Writing an Abstract for Your Research Paper

Definition and Purpose of Abstracts

An abstract is a short summary of your (published or unpublished) research paper, usually about a paragraph (c. 6-7 sentences, 150-250 words) long. A well-written abstract serves multiple purposes:

  • an abstract lets readers get the gist or essence of your paper or article quickly, in order to decide whether to read the full paper;
  • an abstract prepares readers to follow the detailed information, analyses, and arguments in your full paper;
  • and, later, an abstract helps readers remember key points from your paper.

It’s also worth remembering that search engines and bibliographic databases use abstracts, as well as the title, to identify key terms for indexing your published paper. So what you include in your abstract and in your title are crucial for helping other researchers find your paper or article.

If you are writing an abstract for a course paper, your professor may give you specific guidelines for what to include and how to organize your abstract. Similarly, academic journals often have specific requirements for abstracts. So in addition to following the advice on this page, you should be sure to look for and follow any guidelines from the course or journal you’re writing for.

The Contents of an Abstract

Abstracts contain most of the following kinds of information in brief form. The body of your paper will, of course, develop and explain these ideas much more fully. As you will see in the samples below, the proportion of your abstract that you devote to each kind of information—and the sequence of that information—will vary, depending on the nature and genre of the paper that you are summarizing in your abstract. And in some cases, some of this information is implied, rather than stated explicitly. The Publication Manual of the American Psychological Association , which is widely used in the social sciences, gives specific guidelines for what to include in the abstract for different kinds of papers—for empirical studies, literature reviews or meta-analyses, theoretical papers, methodological papers, and case studies.

Here are the typical kinds of information found in most abstracts:

  • the context or background information for your research; the general topic under study; the specific topic of your research
  • the central questions or statement of the problem your research addresses
  • what’s already known about this question, what previous research has done or shown
  • the main reason(s) , the exigency, the rationale , the goals for your research—Why is it important to address these questions? Are you, for example, examining a new topic? Why is that topic worth examining? Are you filling a gap in previous research? Applying new methods to take a fresh look at existing ideas or data? Resolving a dispute within the literature in your field? . . .
  • your research and/or analytical methods
  • your main findings , results , or arguments
  • the significance or implications of your findings or arguments.

Your abstract should be intelligible on its own, without a reader’s having to read your entire paper. And in an abstract, you usually do not cite references—most of your abstract will describe what you have studied in your research and what you have found and what you argue in your paper. In the body of your paper, you will cite the specific literature that informs your research.

When to Write Your Abstract

Although you might be tempted to write your abstract first because it will appear as the very first part of your paper, it’s a good idea to wait to write your abstract until after you’ve drafted your full paper, so that you know what you’re summarizing.

What follows are some sample abstracts in published papers or articles, all written by faculty at UW-Madison who come from a variety of disciplines. We have annotated these samples to help you see the work that these authors are doing within their abstracts.

Choosing Verb Tenses within Your Abstract

The social science sample (Sample 1) below uses the present tense to describe general facts and interpretations that have been and are currently true, including the prevailing explanation for the social phenomenon under study. That abstract also uses the present tense to describe the methods, the findings, the arguments, and the implications of the findings from their new research study. The authors use the past tense to describe previous research.

The humanities sample (Sample 2) below uses the past tense to describe completed events in the past (the texts created in the pulp fiction industry in the 1970s and 80s) and uses the present tense to describe what is happening in those texts, to explain the significance or meaning of those texts, and to describe the arguments presented in the article.

The science samples (Samples 3 and 4) below use the past tense to describe what previous research studies have done and the research the authors have conducted, the methods they have followed, and what they have found. In their rationale or justification for their research (what remains to be done), they use the present tense. They also use the present tense to introduce their study (in Sample 3, “Here we report . . .”) and to explain the significance of their study (In Sample 3, This reprogramming . . . “provides a scalable cell source for. . .”).

Sample Abstract 1

From the social sciences.

Reporting new findings about the reasons for increasing economic homogamy among spouses

Gonalons-Pons, Pilar, and Christine R. Schwartz. “Trends in Economic Homogamy: Changes in Assortative Mating or the Division of Labor in Marriage?” Demography , vol. 54, no. 3, 2017, pp. 985-1005.

“The growing economic resemblance of spouses has contributed to rising inequality by increasing the number of couples in which there are two high- or two low-earning partners. [Annotation for the previous sentence: The first sentence introduces the topic under study (the “economic resemblance of spouses”). This sentence also implies the question underlying this research study: what are the various causes—and the interrelationships among them—for this trend?] The dominant explanation for this trend is increased assortative mating. Previous research has primarily relied on cross-sectional data and thus has been unable to disentangle changes in assortative mating from changes in the division of spouses’ paid labor—a potentially key mechanism given the dramatic rise in wives’ labor supply. [Annotation for the previous two sentences: These next two sentences explain what previous research has demonstrated. By pointing out the limitations in the methods that were used in previous studies, they also provide a rationale for new research.] We use data from the Panel Study of Income Dynamics (PSID) to decompose the increase in the correlation between spouses’ earnings and its contribution to inequality between 1970 and 2013 into parts due to (a) changes in assortative mating, and (b) changes in the division of paid labor. [Annotation for the previous sentence: The data, research and analytical methods used in this new study.] Contrary to what has often been assumed, the rise of economic homogamy and its contribution to inequality is largely attributable to changes in the division of paid labor rather than changes in sorting on earnings or earnings potential. Our findings indicate that the rise of economic homogamy cannot be explained by hypotheses centered on meeting and matching opportunities, and they show where in this process inequality is generated and where it is not.” (p. 985) [Annotation for the previous two sentences: The major findings from and implications and significance of this study.]

Sample Abstract 2

From the humanities.

Analyzing underground pulp fiction publications in Tanzania, this article makes an argument about the cultural significance of those publications

Emily Callaci. “Street Textuality: Socialism, Masculinity, and Urban Belonging in Tanzania’s Pulp Fiction Publishing Industry, 1975-1985.” Comparative Studies in Society and History , vol. 59, no. 1, 2017, pp. 183-210.

“From the mid-1970s through the mid-1980s, a network of young urban migrant men created an underground pulp fiction publishing industry in the city of Dar es Salaam. [Annotation for the previous sentence: The first sentence introduces the context for this research and announces the topic under study.] As texts that were produced in the underground economy of a city whose trajectory was increasingly charted outside of formalized planning and investment, these novellas reveal more than their narrative content alone. These texts were active components in the urban social worlds of the young men who produced them. They reveal a mode of urbanism otherwise obscured by narratives of decolonization, in which urban belonging was constituted less by national citizenship than by the construction of social networks, economic connections, and the crafting of reputations. This article argues that pulp fiction novellas of socialist era Dar es Salaam are artifacts of emergent forms of male sociability and mobility. In printing fictional stories about urban life on pilfered paper and ink, and distributing their texts through informal channels, these writers not only described urban communities, reputations, and networks, but also actually created them.” (p. 210) [Annotation for the previous sentences: The remaining sentences in this abstract interweave other essential information for an abstract for this article. The implied research questions: What do these texts mean? What is their historical and cultural significance, produced at this time, in this location, by these authors? The argument and the significance of this analysis in microcosm: these texts “reveal a mode or urbanism otherwise obscured . . .”; and “This article argues that pulp fiction novellas. . . .” This section also implies what previous historical research has obscured. And through the details in its argumentative claims, this section of the abstract implies the kinds of methods the author has used to interpret the novellas and the concepts under study (e.g., male sociability and mobility, urban communities, reputations, network. . . ).]

Sample Abstract/Summary 3

From the sciences.

Reporting a new method for reprogramming adult mouse fibroblasts into induced cardiac progenitor cells

Lalit, Pratik A., Max R. Salick, Daryl O. Nelson, Jayne M. Squirrell, Christina M. Shafer, Neel G. Patel, Imaan Saeed, Eric G. Schmuck, Yogananda S. Markandeya, Rachel Wong, Martin R. Lea, Kevin W. Eliceiri, Timothy A. Hacker, Wendy C. Crone, Michael Kyba, Daniel J. Garry, Ron Stewart, James A. Thomson, Karen M. Downs, Gary E. Lyons, and Timothy J. Kamp. “Lineage Reprogramming of Fibroblasts into Proliferative Induced Cardiac Progenitor Cells by Defined Factors.” Cell Stem Cell , vol. 18, 2016, pp. 354-367.

“Several studies have reported reprogramming of fibroblasts into induced cardiomyocytes; however, reprogramming into proliferative induced cardiac progenitor cells (iCPCs) remains to be accomplished. [Annotation for the previous sentence: The first sentence announces the topic under study, summarizes what’s already known or been accomplished in previous research, and signals the rationale and goals are for the new research and the problem that the new research solves: How can researchers reprogram fibroblasts into iCPCs?] Here we report that a combination of 11 or 5 cardiac factors along with canonical Wnt and JAK/STAT signaling reprogrammed adult mouse cardiac, lung, and tail tip fibroblasts into iCPCs. The iCPCs were cardiac mesoderm-restricted progenitors that could be expanded extensively while maintaining multipo-tency to differentiate into cardiomyocytes, smooth muscle cells, and endothelial cells in vitro. Moreover, iCPCs injected into the cardiac crescent of mouse embryos differentiated into cardiomyocytes. iCPCs transplanted into the post-myocardial infarction mouse heart improved survival and differentiated into cardiomyocytes, smooth muscle cells, and endothelial cells. [Annotation for the previous four sentences: The methods the researchers developed to achieve their goal and a description of the results.] Lineage reprogramming of adult somatic cells into iCPCs provides a scalable cell source for drug discovery, disease modeling, and cardiac regenerative therapy.” (p. 354) [Annotation for the previous sentence: The significance or implications—for drug discovery, disease modeling, and therapy—of this reprogramming of adult somatic cells into iCPCs.]

Sample Abstract 4, a Structured Abstract

Reporting results about the effectiveness of antibiotic therapy in managing acute bacterial sinusitis, from a rigorously controlled study

Note: This journal requires authors to organize their abstract into four specific sections, with strict word limits. Because the headings for this structured abstract are self-explanatory, we have chosen not to add annotations to this sample abstract.

Wald, Ellen R., David Nash, and Jens Eickhoff. “Effectiveness of Amoxicillin/Clavulanate Potassium in the Treatment of Acute Bacterial Sinusitis in Children.” Pediatrics , vol. 124, no. 1, 2009, pp. 9-15.

“OBJECTIVE: The role of antibiotic therapy in managing acute bacterial sinusitis (ABS) in children is controversial. The purpose of this study was to determine the effectiveness of high-dose amoxicillin/potassium clavulanate in the treatment of children diagnosed with ABS.

METHODS : This was a randomized, double-blind, placebo-controlled study. Children 1 to 10 years of age with a clinical presentation compatible with ABS were eligible for participation. Patients were stratified according to age (<6 or ≥6 years) and clinical severity and randomly assigned to receive either amoxicillin (90 mg/kg) with potassium clavulanate (6.4 mg/kg) or placebo. A symptom survey was performed on days 0, 1, 2, 3, 5, 7, 10, 20, and 30. Patients were examined on day 14. Children’s conditions were rated as cured, improved, or failed according to scoring rules.

RESULTS: Two thousand one hundred thirty-five children with respiratory complaints were screened for enrollment; 139 (6.5%) had ABS. Fifty-eight patients were enrolled, and 56 were randomly assigned. The mean age was 6630 months. Fifty (89%) patients presented with persistent symptoms, and 6 (11%) presented with nonpersistent symptoms. In 24 (43%) children, the illness was classified as mild, whereas in the remaining 32 (57%) children it was severe. Of the 28 children who received the antibiotic, 14 (50%) were cured, 4 (14%) were improved, 4(14%) experienced treatment failure, and 6 (21%) withdrew. Of the 28children who received placebo, 4 (14%) were cured, 5 (18%) improved, and 19 (68%) experienced treatment failure. Children receiving the antibiotic were more likely to be cured (50% vs 14%) and less likely to have treatment failure (14% vs 68%) than children receiving the placebo.

CONCLUSIONS : ABS is a common complication of viral upper respiratory infections. Amoxicillin/potassium clavulanate results in significantly more cures and fewer failures than placebo, according to parental report of time to resolution.” (9)

Some Excellent Advice about Writing Abstracts for Basic Science Research Papers, by Professor Adriano Aguzzi from the Institute of Neuropathology at the University of Zurich:

how long is an abstract research paper

Academic and Professional Writing

This is an accordion element with a series of buttons that open and close related content panels.

Analysis Papers

Reading Poetry

A Short Guide to Close Reading for Literary Analysis

Using Literary Quotations

Play Reviews

Writing a Rhetorical Précis to Analyze Nonfiction Texts

Incorporating Interview Data

Grant Proposals

Planning and Writing a Grant Proposal: The Basics

Additional Resources for Grants and Proposal Writing

Job Materials and Application Essays

Writing Personal Statements for Ph.D. Programs

  • Before you begin: useful tips for writing your essay
  • Guided brainstorming exercises
  • Get more help with your essay
  • Frequently Asked Questions

Resume Writing Tips

CV Writing Tips

Cover Letters

Business Letters

Proposals and Dissertations

Resources for Proposal Writers

Resources for Dissertators

Research Papers

Planning and Writing Research Papers

Quoting and Paraphrasing

Writing Annotated Bibliographies

Creating Poster Presentations

Thank-You Notes

Advice for Students Writing Thank-You Notes to Donors

Reading for a Review

Critical Reviews

Writing a Review of Literature

Scientific Reports

Scientific Report Format

Sample Lab Assignment

Writing for the Web

Writing an Effective Blog Post

Writing for Social Media: A Guide for Academics

Have a language expert improve your writing

Run a free plagiarism check in 10 minutes, generate accurate citations for free.

  • Knowledge Base
  • Dissertation
  • How to Write an Abstract | Steps & Examples

How to Write an Abstract | Steps & Examples

Published on February 28, 2019 by Shona McCombes . Revised on July 18, 2023 by Eoghan Ryan.

How to Write an Abstract

An abstract is a short summary of a longer work (such as a thesis ,  dissertation or research paper ). The abstract concisely reports the aims and outcomes of your research, so that readers know exactly what your paper is about.

Although the structure may vary slightly depending on your discipline, your abstract should describe the purpose of your work, the methods you’ve used, and the conclusions you’ve drawn.

One common way to structure your abstract is to use the IMRaD structure. This stands for:

  • Introduction

Abstracts are usually around 100–300 words, but there’s often a strict word limit, so make sure to check the relevant requirements.

In a dissertation or thesis , include the abstract on a separate page, after the title page and acknowledgements but before the table of contents .

Instantly correct all language mistakes in your text

Upload your document to correct all your mistakes in minutes

upload-your-document-ai-proofreader

Table of contents

Abstract example, when to write an abstract, step 1: introduction, step 2: methods, step 3: results, step 4: discussion, tips for writing an abstract, other interesting articles, frequently asked questions about abstracts.

Hover over the different parts of the abstract to see how it is constructed.

This paper examines the role of silent movies as a mode of shared experience in the US during the early twentieth century. At this time, high immigration rates resulted in a significant percentage of non-English-speaking citizens. These immigrants faced numerous economic and social obstacles, including exclusion from public entertainment and modes of discourse (newspapers, theater, radio).

Incorporating evidence from reviews, personal correspondence, and diaries, this study demonstrates that silent films were an affordable and inclusive source of entertainment. It argues for the accessible economic and representational nature of early cinema. These concerns are particularly evident in the low price of admission and in the democratic nature of the actors’ exaggerated gestures, which allowed the plots and action to be easily grasped by a diverse audience despite language barriers.

Keywords: silent movies, immigration, public discourse, entertainment, early cinema, language barriers.

Here's why students love Scribbr's proofreading services

Discover proofreading & editing

You will almost always have to include an abstract when:

  • Completing a thesis or dissertation
  • Submitting a research paper to an academic journal
  • Writing a book or research proposal
  • Applying for research grants

It’s easiest to write your abstract last, right before the proofreading stage, because it’s a summary of the work you’ve already done. Your abstract should:

  • Be a self-contained text, not an excerpt from your paper
  • Be fully understandable on its own
  • Reflect the structure of your larger work

Start by clearly defining the purpose of your research. What practical or theoretical problem does the research respond to, or what research question did you aim to answer?

You can include some brief context on the social or academic relevance of your dissertation topic , but don’t go into detailed background information. If your abstract uses specialized terms that would be unfamiliar to the average academic reader or that have various different meanings, give a concise definition.

After identifying the problem, state the objective of your research. Use verbs like “investigate,” “test,” “analyze,” or “evaluate” to describe exactly what you set out to do.

This part of the abstract can be written in the present or past simple tense  but should never refer to the future, as the research is already complete.

  • This study will investigate the relationship between coffee consumption and productivity.
  • This study investigates the relationship between coffee consumption and productivity.

Next, indicate the research methods that you used to answer your question. This part should be a straightforward description of what you did in one or two sentences. It is usually written in the past simple tense, as it refers to completed actions.

  • Structured interviews will be conducted with 25 participants.
  • Structured interviews were conducted with 25 participants.

Don’t evaluate validity or obstacles here — the goal is not to give an account of the methodology’s strengths and weaknesses, but to give the reader a quick insight into the overall approach and procedures you used.

Next, summarize the main research results . This part of the abstract can be in the present or past simple tense.

  • Our analysis has shown a strong correlation between coffee consumption and productivity.
  • Our analysis shows a strong correlation between coffee consumption and productivity.
  • Our analysis showed a strong correlation between coffee consumption and productivity.

Depending on how long and complex your research is, you may not be able to include all results here. Try to highlight only the most important findings that will allow the reader to understand your conclusions.

Finally, you should discuss the main conclusions of your research : what is your answer to the problem or question? The reader should finish with a clear understanding of the central point that your research has proved or argued. Conclusions are usually written in the present simple tense.

  • We concluded that coffee consumption increases productivity.
  • We conclude that coffee consumption increases productivity.

If there are important limitations to your research (for example, related to your sample size or methods), you should mention them briefly in the abstract. This allows the reader to accurately assess the credibility and generalizability of your research.

If your aim was to solve a practical problem, your discussion might include recommendations for implementation. If relevant, you can briefly make suggestions for further research.

If your paper will be published, you might have to add a list of keywords at the end of the abstract. These keywords should reference the most important elements of the research to help potential readers find your paper during their own literature searches.

Be aware that some publication manuals, such as APA Style , have specific formatting requirements for these keywords.

It can be a real challenge to condense your whole work into just a couple of hundred words, but the abstract will be the first (and sometimes only) part that people read, so it’s important to get it right. These strategies can help you get started.

Read other abstracts

The best way to learn the conventions of writing an abstract in your discipline is to read other people’s. You probably already read lots of journal article abstracts while conducting your literature review —try using them as a framework for structure and style.

You can also find lots of dissertation abstract examples in thesis and dissertation databases .

Reverse outline

Not all abstracts will contain precisely the same elements. For longer works, you can write your abstract through a process of reverse outlining.

For each chapter or section, list keywords and draft one to two sentences that summarize the central point or argument. This will give you a framework of your abstract’s structure. Next, revise the sentences to make connections and show how the argument develops.

Write clearly and concisely

A good abstract is short but impactful, so make sure every word counts. Each sentence should clearly communicate one main point.

To keep your abstract or summary short and clear:

  • Avoid passive sentences: Passive constructions are often unnecessarily long. You can easily make them shorter and clearer by using the active voice.
  • Avoid long sentences: Substitute longer expressions for concise expressions or single words (e.g., “In order to” for “To”).
  • Avoid obscure jargon: The abstract should be understandable to readers who are not familiar with your topic.
  • Avoid repetition and filler words: Replace nouns with pronouns when possible and eliminate unnecessary words.
  • Avoid detailed descriptions: An abstract is not expected to provide detailed definitions, background information, or discussions of other scholars’ work. Instead, include this information in the body of your thesis or paper.

If you’re struggling to edit down to the required length, you can get help from expert editors with Scribbr’s professional proofreading services or use the paraphrasing tool .

Check your formatting

If you are writing a thesis or dissertation or submitting to a journal, there are often specific formatting requirements for the abstract—make sure to check the guidelines and format your work correctly. For APA research papers you can follow the APA abstract format .

Checklist: Abstract

The word count is within the required length, or a maximum of one page.

The abstract appears after the title page and acknowledgements and before the table of contents .

I have clearly stated my research problem and objectives.

I have briefly described my methodology .

I have summarized the most important results .

I have stated my main conclusions .

I have mentioned any important limitations and recommendations.

The abstract can be understood by someone without prior knowledge of the topic.

You've written a great abstract! Use the other checklists to continue improving your thesis or dissertation.

If you want to know more about AI for academic writing, AI tools, or research bias, make sure to check out some of our other articles with explanations and examples or go directly to our tools!

Research bias

  • Anchoring bias
  • Halo effect
  • The Baader–Meinhof phenomenon
  • The placebo effect
  • Nonresponse bias
  • Deep learning
  • Generative AI
  • Machine learning
  • Reinforcement learning
  • Supervised vs. unsupervised learning

 (AI) Tools

  • Grammar Checker
  • Paraphrasing Tool
  • Text Summarizer
  • AI Detector
  • Plagiarism Checker
  • Citation Generator

An abstract is a concise summary of an academic text (such as a journal article or dissertation ). It serves two main purposes:

  • To help potential readers determine the relevance of your paper for their own research.
  • To communicate your key findings to those who don’t have time to read the whole paper.

Abstracts are often indexed along with keywords on academic databases, so they make your work more easily findable. Since the abstract is the first thing any reader sees, it’s important that it clearly and accurately summarizes the contents of your paper.

An abstract for a thesis or dissertation is usually around 200–300 words. There’s often a strict word limit, so make sure to check your university’s requirements.

The abstract is the very last thing you write. You should only write it after your research is complete, so that you can accurately summarize the entirety of your thesis , dissertation or research paper .

Avoid citing sources in your abstract . There are two reasons for this:

  • The abstract should focus on your original research, not on the work of others.
  • The abstract should be self-contained and fully understandable without reference to other sources.

There are some circumstances where you might need to mention other sources in an abstract: for example, if your research responds directly to another study or focuses on the work of a single theorist. In general, though, don’t include citations unless absolutely necessary.

The abstract appears on its own page in the thesis or dissertation , after the title page and acknowledgements but before the table of contents .

Cite this Scribbr article

If you want to cite this source, you can copy and paste the citation or click the “Cite this Scribbr article” button to automatically add the citation to our free Citation Generator.

McCombes, S. (2023, July 18). How to Write an Abstract | Steps & Examples. Scribbr. Retrieved February 29, 2024, from https://www.scribbr.com/dissertation/abstract/

Is this article helpful?

Shona McCombes

Shona McCombes

Other students also liked, how to write a thesis or dissertation introduction, shorten your abstract or summary, how to write a literature review | guide, examples, & templates, what is your plagiarism score.

  • Features for Creative Writers
  • Features for Work
  • Features for Higher Education
  • Features for Teachers
  • Features for Non-Native Speakers
  • Learn Blog Grammar Guide Community Academy FAQ
  • Grammar Guide

How to Write an Abstract (With Examples)

Sarah Oakley headshot

Sarah Oakley

how to write an abstract

Table of Contents

What is an abstract in a paper, how long should an abstract be, 5 steps for writing an abstract, examples of an abstract, how prowritingaid can help you write an abstract.

If you are writing a scientific research paper or a book proposal, you need to know how to write an abstract, which summarizes the contents of the paper or book.

When researchers are looking for peer-reviewed papers to use in their studies, the first place they will check is the abstract to see if it applies to their work. Therefore, your abstract is one of the most important parts of your entire paper.

In this article, we’ll explain what an abstract is, what it should include, and how to write one.

An abstract is a concise summary of the details within a report. Some abstracts give more details than others, but the main things you’ll be talking about are why you conducted the research, what you did, and what the results show.

When a reader is deciding whether to read your paper completely, they will first look at the abstract. You need to be concise in your abstract and give the reader the most important information so they can determine if they want to read the whole paper.

Remember that an abstract is the last thing you’ll want to write for the research paper because it directly references parts of the report. If you haven’t written the report, you won’t know what to include in your abstract.

If you are writing a paper for a journal or an assignment, the publication or academic institution might have specific formatting rules for how long your abstract should be. However, if they don’t, most abstracts are between 150 and 300 words long.

A short word count means your writing has to be precise and without filler words or phrases. Once you’ve written a first draft, you can always use an editing tool, such as ProWritingAid, to identify areas where you can reduce words and increase readability.

If your abstract is over the word limit, and you’ve edited it but still can’t figure out how to reduce it further, your abstract might include some things that aren’t needed. Here’s a list of three elements you can remove from your abstract:

Discussion : You don’t need to go into detail about the findings of your research because your reader will find your discussion within the paper.

Definition of terms : Your readers are interested the field you are writing about, so they are likely to understand the terms you are using. If not, they can always look them up. Your readers do not expect you to give a definition of terms in your abstract.

References and citations : You can mention there have been studies that support or have inspired your research, but you do not need to give details as the reader will find them in your bibliography.

how long is an abstract research paper

Good writing = better grades

ProWritingAid will help you improve the style, strength, and clarity of all your assignments.

If you’ve never written an abstract before, and you’re wondering how to write an abstract, we’ve got some steps for you to follow. It’s best to start with planning your abstract, so we’ve outlined the details you need to include in your plan before you write.

Remember to consider your audience when you’re planning and writing your abstract. They are likely to skim read your abstract, so you want to be sure your abstract delivers all the information they’re expecting to see at key points.

1. What Should an Abstract Include?

Abstracts have a lot of information to cover in a short number of words, so it’s important to know what to include. There are three elements that need to be present in your abstract:

Your context is the background for where your research sits within your field of study. You should briefly mention any previous scientific papers or experiments that have led to your hypothesis and how research develops in those studies.

Your hypothesis is your prediction of what your study will show. As you are writing your abstract after you have conducted your research, you should still include your hypothesis in your abstract because it shows the motivation for your paper.

Throughout your abstract, you also need to include keywords and phrases that will help researchers to find your article in the databases they’re searching. Make sure the keywords are specific to your field of study and the subject you’re reporting on, otherwise your article might not reach the relevant audience.

2. Can You Use First Person in an Abstract?

You might think that first person is too informal for a research paper, but it’s not. Historically, writers of academic reports avoided writing in first person to uphold the formality standards of the time. However, first person is more accepted in research papers in modern times.

If you’re still unsure whether to write in first person for your abstract, refer to any style guide rules imposed by the journal you’re writing for or your teachers if you are writing an assignment.

3. Abstract Structure

Some scientific journals have strict rules on how to structure an abstract, so it’s best to check those first. If you don’t have any style rules to follow, try using the IMRaD structure, which stands for Introduction, Methodology, Results, and Discussion.

how to structure an abstract

Following the IMRaD structure, start with an introduction. The amount of background information you should include depends on your specific research area. Adding a broad overview gives you less room to include other details. Remember to include your hypothesis in this section.

The next part of your abstract should cover your methodology. Try to include the following details if they apply to your study:

What type of research was conducted?

How were the test subjects sampled?

What were the sample sizes?

What was done to each group?

How long was the experiment?

How was data recorded and interpreted?

Following the methodology, include a sentence or two about the results, which is where your reader will determine if your research supports or contradicts their own investigations.

The results are also where most people will want to find out what your outcomes were, even if they are just mildly interested in your research area. You should be specific about all the details but as concise as possible.

The last few sentences are your conclusion. It needs to explain how your findings affect the context and whether your hypothesis was correct. Include the primary take-home message, additional findings of importance, and perspective. Also explain whether there is scope for further research into the subject of your report.

Your conclusion should be honest and give the reader the ultimate message that your research shows. Readers trust the conclusion, so make sure you’re not fabricating the results of your research. Some readers won’t read your entire paper, but this section will tell them if it’s worth them referencing it in their own study.

4. How to Start an Abstract

The first line of your abstract should give your reader the context of your report by providing background information. You can use this sentence to imply the motivation for your research.

You don’t need to use a hook phrase or device in your first sentence to grab the reader’s attention. Your reader will look to establish relevance quickly, so readability and clarity are more important than trying to persuade the reader to read on.

5. How to Format an Abstract

Most abstracts use the same formatting rules, which help the reader identify the abstract so they know where to look for it.

Here’s a list of formatting guidelines for writing an abstract:

Stick to one paragraph

Use block formatting with no indentation at the beginning

Put your abstract straight after the title and acknowledgements pages

Use present or past tense, not future tense

There are two primary types of abstract you could write for your paper—descriptive and informative.

An informative abstract is the most common, and they follow the structure mentioned previously. They are longer than descriptive abstracts because they cover more details.

Descriptive abstracts differ from informative abstracts, as they don’t include as much discussion or detail. The word count for a descriptive abstract is between 50 and 150 words.

Here is an example of an informative abstract:

A growing trend exists for authors to employ a more informal writing style that uses “we” in academic writing to acknowledge one’s stance and engagement. However, few studies have compared the ways in which the first-person pronoun “we” is used in the abstracts and conclusions of empirical papers. To address this lacuna in the literature, this study conducted a systematic corpus analysis of the use of “we” in the abstracts and conclusions of 400 articles collected from eight leading electrical and electronic (EE) engineering journals. The abstracts and conclusions were extracted to form two subcorpora, and an integrated framework was applied to analyze and seek to explain how we-clusters and we-collocations were employed. Results revealed whether authors’ use of first-person pronouns partially depends on a journal policy. The trend of using “we” showed that a yearly increase occurred in the frequency of “we” in EE journal papers, as well as the existence of three “we-use” types in the article conclusions and abstracts: exclusive, inclusive, and ambiguous. Other possible “we-use” alternatives such as “I” and other personal pronouns were used very rarely—if at all—in either section. These findings also suggest that the present tense was used more in article abstracts, but the present perfect tense was the most preferred tense in article conclusions. Both research and pedagogical implications are proffered and critically discussed.

Wang, S., Tseng, W.-T., & Johanson, R. (2021). To We or Not to We: Corpus-Based Research on First-Person Pronoun Use in Abstracts and Conclusions. SAGE Open, 11(2).

Here is an example of a descriptive abstract:

From the 1850s to the present, considerable criminological attention has focused on the development of theoretically-significant systems for classifying crime. This article reviews and attempts to evaluate a number of these efforts, and we conclude that further work on this basic task is needed. The latter part of the article explicates a conceptual foundation for a crime pattern classification system, and offers a preliminary taxonomy of crime.

Farr, K. A., & Gibbons, D. C. (1990). Observations on the Development of Crime Categories. International Journal of Offender Therapy and Comparative Criminology, 34(3), 223–237.

If you want to ensure your abstract is grammatically correct and easy to read, you can use ProWritingAid to edit it. The software integrates with Microsoft Word, Google Docs, and most web browsers, so you can make the most of it wherever you’re writing your paper.

academic document type

Before you edit with ProWritingAid, make sure the suggestions you are seeing are relevant for your document by changing the document type to “Abstract” within the Academic writing style section.

You can use the Readability report to check your abstract for places to improve the clarity of your writing. Some suggestions might show you where to remove words, which is great if you’re over your word count.

We hope the five steps and examples we’ve provided help you write a great abstract for your research paper.

Get started with ProWritingAid

Drop us a line or let's stay in touch via :

  • Resources Home 🏠
  • Try SciSpace Copilot
  • Search research papers
  • Add Copilot Extension
  • Try AI Detector
  • Try Paraphraser
  • Try Citation Generator
  • April Papers
  • June Papers
  • July Papers

SciSpace Resources

Abstract Writing: A Step-by-Step Guide With Tips & Examples

Sumalatha G

Table of Contents

step-by-step-guide-to-abstract-writing

Introduction

Abstracts of research papers have always played an essential role in describing your research concisely and clearly to researchers and editors of journals, enticing them to continue reading. However, with the widespread availability of scientific databases, the need to write a convincing abstract is more crucial now than during the time of paper-bound manuscripts.

Abstracts serve to "sell" your research and can be compared with your "executive outline" of a resume or, rather, a formal summary of the critical aspects of your work. Also, it can be the "gist" of your study. Since most educational research is done online, it's a sign that you have a shorter time for impressing your readers, and have more competition from other abstracts that are available to be read.

The APCI (Academic Publishing and Conferences International) articulates 12 issues or points considered during the final approval process for conferences & journals and emphasises the importance of writing an abstract that checks all these boxes (12 points). Since it's the only opportunity you have to captivate your readers, you must invest time and effort in creating an abstract that accurately reflects the critical points of your research.

With that in mind, let’s head over to understand and discover the core concept and guidelines to create a substantial abstract. Also, learn how to organise the ideas or plots into an effective abstract that will be awe-inspiring to the readers you want to reach.

What is Abstract? Definition and Overview

The word "Abstract' is derived from Latin abstractus meaning "drawn off." This etymological meaning also applies to art movements as well as music, like abstract expressionism. In this context, it refers to the revealing of the artist's intention.

Based on this, you can determine the meaning of an abstract: A condensed research summary. It must be self-contained and independent of the body of the research. However, it should outline the subject, the strategies used to study the problem, and the methods implemented to attain the outcomes. The specific elements of the study differ based on the area of study; however, together, it must be a succinct summary of the entire research paper.

Abstracts are typically written at the end of the paper, even though it serves as a prologue. In general, the abstract must be in a position to:

  • Describe the paper.
  • Identify the problem or the issue at hand.
  • Explain to the reader the research process, the results you came up with, and what conclusion you've reached using these results.
  • Include keywords to guide your strategy and the content.

Furthermore, the abstract you submit should not reflect upon any of  the following elements:

  • Examine, analyse or defend the paper or your opinion.
  • What you want to study, achieve or discover.
  • Be redundant or irrelevant.

After reading an abstract, your audience should understand the reason - what the research was about in the first place, what the study has revealed and how it can be utilised or can be used to benefit others. You can understand the importance of abstract by knowing the fact that the abstract is the most frequently read portion of any research paper. In simpler terms, it should contain all the main points of the research paper.

purpose-of-abstract-writing

What is the Purpose of an Abstract?

Abstracts are typically an essential requirement for research papers; however, it's not an obligation to preserve traditional reasons without any purpose. Abstracts allow readers to scan the text to determine whether it is relevant to their research or studies. The abstract allows other researchers to decide if your research paper can provide them with some additional information. A good abstract paves the interest of the audience to pore through your entire paper to find the content or context they're searching for.

Abstract writing is essential for indexing, as well. The Digital Repository of academic papers makes use of abstracts to index the entire content of academic research papers. Like meta descriptions in the regular Google outcomes, abstracts must include keywords that help researchers locate what they seek.

Types of Abstract

Informative and Descriptive are two kinds of abstracts often used in scientific writing.

A descriptive abstract gives readers an outline of the author's main points in their study. The reader can determine if they want to stick to the research work, based on their interest in the topic. An abstract that is descriptive is similar to the contents table of books, however, the format of an abstract depicts complete sentences encapsulated in one paragraph. It is unfortunate that the abstract can't be used as a substitute for reading a piece of writing because it's just an overview, which omits readers from getting an entire view. Also, it cannot be a way to fill in the gaps the reader may have after reading this kind of abstract since it does not contain crucial information needed to evaluate the article.

To conclude, a descriptive abstract is:

  • A simple summary of the task, just summarises the work, but some researchers think it is much more of an outline
  • Typically, the length is approximately 100 words. It is too short when compared to an informative abstract.
  • A brief explanation but doesn't provide the reader with the complete information they need;
  • An overview that omits conclusions and results

An informative abstract is a comprehensive outline of the research. There are times when people rely on the abstract as an information source. And the reason is why it is crucial to provide entire data of particular research. A well-written, informative abstract could be a good substitute for the remainder of the paper on its own.

A well-written abstract typically follows a particular style. The author begins by providing the identifying information, backed by citations and other identifiers of the papers. Then, the major elements are summarised to make the reader aware of the study. It is followed by the methodology and all-important findings from the study. The conclusion then presents study results and ends the abstract with a comprehensive summary.

In a nutshell, an informative abstract:

  • Has a length that can vary, based on the subject, but is not longer than 300 words.
  • Contains all the content-like methods and intentions
  • Offers evidence and possible recommendations.

Informative Abstracts are more frequent than descriptive abstracts because of their extensive content and linkage to the topic specifically. You should select different types of abstracts to papers based on their length: informative abstracts for extended and more complex abstracts and descriptive ones for simpler and shorter research papers.

What are the Characteristics of a Good Abstract?

  • A good abstract clearly defines the goals and purposes of the study.
  • It should clearly describe the research methodology with a primary focus on data gathering, processing, and subsequent analysis.
  • A good abstract should provide specific research findings.
  • It presents the principal conclusions of the systematic study.
  • It should be concise, clear, and relevant to the field of study.
  • A well-designed abstract should be unifying and coherent.
  • It is easy to grasp and free of technical jargon.
  • It is written impartially and objectively.

the-various-sections-of-abstract-writing

What are the various sections of an ideal Abstract?

By now, you must have gained some concrete idea of the essential elements that your abstract needs to convey . Accordingly, the information is broken down into six key sections of the abstract, which include:

An Introduction or Background

Research methodology, objectives and goals, limitations.

Let's go over them in detail.

The introduction, also known as background, is the most concise part of your abstract. Ideally, it comprises a couple of sentences. Some researchers only write one sentence to introduce their abstract. The idea behind this is to guide readers through the key factors that led to your study.

It's understandable that this information might seem difficult to explain in a couple of sentences. For example, think about the following two questions like the background of your study:

  • What is currently available about the subject with respect to the paper being discussed?
  • What isn't understood about this issue? (This is the subject of your research)

While writing the abstract’s introduction, make sure that it is not lengthy. Because if it crosses the word limit, it may eat up the words meant to be used for providing other key information.

Research methodology is where you describe the theories and techniques you used in your research. It is recommended that you describe what you have done and the method you used to get your thorough investigation results. Certainly, it is the second-longest paragraph in the abstract.

In the research methodology section, it is essential to mention the kind of research you conducted; for instance, qualitative research or quantitative research (this will guide your research methodology too) . If you've conducted quantitative research, your abstract should contain information like the sample size, data collection method, sampling techniques, and duration of the study. Likewise, your abstract should reflect observational data, opinions, questionnaires (especially the non-numerical data) if you work on qualitative research.

The research objectives and goals speak about what you intend to accomplish with your research. The majority of research projects focus on the long-term effects of a project, and the goals focus on the immediate, short-term outcomes of the research. It is possible to summarise both in just multiple sentences.

In stating your objectives and goals, you give readers a picture of the scope of the study, its depth and the direction your research ultimately follows. Your readers can evaluate the results of your research against the goals and stated objectives to determine if you have achieved the goal of your research.

In the end, your readers are more attracted by the results you've obtained through your study. Therefore, you must take the time to explain each relevant result and explain how they impact your research. The results section exists as the longest in your abstract, and nothing should diminish its reach or quality.

One of the most important things you should adhere to is to spell out details and figures on the results of your research.

Instead of making a vague assertion such as, "We noticed that response rates varied greatly between respondents with high incomes and those with low incomes", Try these: "The response rate was higher for high-income respondents than those with lower incomes (59 30 percent vs. 30 percent in both cases; P<0.01)."

You're likely to encounter certain obstacles during your research. It could have been during data collection or even during conducting the sample . Whatever the issue, it's essential to inform your readers about them and their effects on the research.

Research limitations offer an opportunity to suggest further and deep research. If, for instance, you were forced to change for convenient sampling and snowball samples because of difficulties in reaching well-suited research participants, then you should mention this reason when you write your research abstract. In addition, a lack of prior studies on the subject could hinder your research.

Your conclusion should include the same number of sentences to wrap the abstract as the introduction. The majority of researchers offer an idea of the consequences of their research in this case.

Your conclusion should include three essential components:

  • A significant take-home message.
  • Corresponding important findings.
  • The Interpretation.

Even though the conclusion of your abstract needs to be brief, it can have an enormous influence on the way that readers view your research. Therefore, make use of this section to reinforce the central message from your research. Be sure that your statements reflect the actual results and the methods you used to conduct your research.

examples-of-good-abstract-writing

Good Abstract Examples

Abstract example #1.

Children’s consumption behavior in response to food product placements in movies.

The abstract:

"Almost all research into the effects of brand placements on children has focused on the brand's attitudes or behavior intentions. Based on the significant differences between attitudes and behavioral intentions on one hand and actual behavior on the other hand, this study examines the impact of placements by brands on children's eating habits. Children aged 6-14 years old were shown an excerpt from the popular film Alvin and the Chipmunks and were shown places for the item Cheese Balls. Three different versions were developed with no placements, one with moderately frequent placements and the third with the highest frequency of placement. The results revealed that exposure to high-frequency places had a profound effect on snack consumption, however, there was no impact on consumer attitudes towards brands or products. The effects were not dependent on the age of the children. These findings are of major importance to researchers studying consumer behavior as well as nutrition experts as well as policy regulators."

Abstract Example #2

Social comparisons on social media: The impact of Facebook on young women’s body image concerns and mood. The abstract:

"The research conducted in this study investigated the effects of Facebook use on women's moods and body image if the effects are different from an internet-based fashion journal and if the appearance comparison tendencies moderate one or more of these effects. Participants who were female ( N = 112) were randomly allocated to spend 10 minutes exploring their Facebook account or a magazine's website or an appearance neutral control website prior to completing state assessments of body dissatisfaction, mood, and differences in appearance (weight-related and facial hair, face, and skin). Participants also completed a test of the tendency to compare appearances. The participants who used Facebook were reported to be more depressed than those who stayed on the control site. In addition, women who have the tendency to compare appearances reported more facial, hair and skin-related issues following Facebook exposure than when they were exposed to the control site. Due to its popularity it is imperative to conduct more research to understand the effect that Facebook affects the way people view themselves."

Abstract Example #3

The Relationship Between Cell Phone Use and Academic Performance in a Sample of U.S. College Students

"The cellphone is always present on campuses of colleges and is often utilised in situations in which learning takes place. The study examined the connection between the use of cell phones and the actual grades point average (GPA) after adjusting for predictors that are known to be a factor. In the end 536 students in the undergraduate program from 82 self-reported majors of an enormous, public institution were studied. Hierarchical analysis ( R 2 = .449) showed that use of mobile phones is significantly ( p < .001) and negative (b equal to -.164) connected to the actual college GPA, after taking into account factors such as demographics, self-efficacy in self-regulated learning, self-efficacy to improve academic performance, and the actual high school GPA that were all important predictors ( p < .05). Therefore, after adjusting for other known predictors increasing cell phone usage was associated with lower academic performance. While more research is required to determine the mechanisms behind these results, they suggest the need to educate teachers and students to the possible academic risks that are associated with high-frequency mobile phone usage."

quick-tips-on-writing-a-good-abstract

Quick tips on writing a good abstract

There exists a common dilemma among early age researchers whether to write the abstract at first or last? However, it's recommended to compose your abstract when you've completed the research since you'll have all the information to give to your readers. You can, however, write a draft at the beginning of your research and add in any gaps later.

If you find abstract writing a herculean task, here are the few tips to help you with it:

1. Always develop a framework to support your abstract

Before writing, ensure you create a clear outline for your abstract. Divide it into sections and draw the primary and supporting elements in each one. You can include keywords and a few sentences that convey the essence of your message.

2. Review Other Abstracts

Abstracts are among the most frequently used research documents, and thousands of them were written in the past. Therefore, prior to writing yours, take a look at some examples from other abstracts. There are plenty of examples of abstracts for dissertations in the dissertation and thesis databases.

3. Avoid Jargon To the Maximum

When you write your abstract, focus on simplicity over formality. You should  write in simple language, and avoid excessive filler words or ambiguous sentences. Keep in mind that your abstract must be readable to those who aren't acquainted with your subject.

4. Focus on Your Research

It's a given fact that the abstract you write should be about your research and the findings you've made. It is not the right time to mention secondary and primary data sources unless it's absolutely required.

Conclusion: How to Structure an Interesting Abstract?

Abstracts are a short outline of your essay. However, it's among the most important, if not the most important. The process of writing an abstract is not straightforward. A few early-age researchers tend to begin by writing it, thinking they are doing it to "tease" the next step (the document itself). However, it is better to treat it as a spoiler.

The simple, concise style of the abstract lends itself to a well-written and well-investigated study. If your research paper doesn't provide definitive results, or the goal of your research is questioned, so will the abstract. Thus, only write your abstract after witnessing your findings and put your findings in the context of a larger scenario.

The process of writing an abstract can be daunting, but with these guidelines, you will succeed. The most efficient method of writing an excellent abstract is to centre the primary points of your abstract, including the research question and goals methods, as well as key results.

Interested in learning more about dedicated research solutions? Go to the SciSpace product page to find out how our suite of products can help you simplify your research workflows so you can focus on advancing science.

Literature search in Scispace

The best-in-class solution is equipped with features such as literature search and discovery, profile management, research writing and formatting, and so much more.

But before you go,

You might also like.

Consensus GPT vs. SciSpace GPT: Choose the Best GPT for Research

Consensus GPT vs. SciSpace GPT: Choose the Best GPT for Research

Sumalatha G

Literature Review and Theoretical Framework: Understanding the Differences

Nikhil Seethi

Types of Essays in Academic Writing

  • Privacy Policy

Buy Me a Coffee

Research Method

Home » Research Paper Abstract – Writing Guide and Examples

Research Paper Abstract – Writing Guide and Examples

Table of Contents

Research Paper Abstract

Research Paper Abstract

Research Paper Abstract is a brief summary of a research pape r that describes the study’s purpose, methods, findings, and conclusions . It is often the first section of the paper that readers encounter, and its purpose is to provide a concise and accurate overview of the paper’s content. The typical length of an abstract is usually around 150-250 words, and it should be written in a concise and clear manner.

Research Paper Abstract Structure

The structure of a research paper abstract usually includes the following elements:

  • Background or Introduction: Briefly describe the problem or research question that the study addresses.
  • Methods : Explain the methodology used to conduct the study, including the participants, materials, and procedures.
  • Results : Summarize the main findings of the study, including statistical analyses and key outcomes.
  • Conclusions : Discuss the implications of the study’s findings and their significance for the field, as well as any limitations or future directions for research.
  • Keywords : List a few keywords that describe the main topics or themes of the research.

How to Write Research Paper Abstract

Here are the steps to follow when writing a research paper abstract:

  • Start by reading your paper: Before you write an abstract, you should have a complete understanding of your paper. Read through the paper carefully, making sure you understand the purpose, methods, results, and conclusions.
  • Identify the key components : Identify the key components of your paper, such as the research question, methods used, results obtained, and conclusion reached.
  • Write a draft: Write a draft of your abstract, using concise and clear language. Make sure to include all the important information, but keep it short and to the point. A good rule of thumb is to keep your abstract between 150-250 words.
  • Use clear and concise language : Use clear and concise language to explain the purpose of your study, the methods used, the results obtained, and the conclusions drawn.
  • Emphasize your findings: Emphasize your findings in the abstract, highlighting the key results and the significance of your study.
  • Revise and edit: Once you have a draft, revise and edit it to ensure that it is clear, concise, and free from errors.
  • Check the formatting: Finally, check the formatting of your abstract to make sure it meets the requirements of the journal or conference where you plan to submit it.

Research Paper Abstract Examples

Research Paper Abstract Examples could be following:

Title : “The Effectiveness of Cognitive-Behavioral Therapy for Treating Anxiety Disorders: A Meta-Analysis”

Abstract : This meta-analysis examines the effectiveness of cognitive-behavioral therapy (CBT) in treating anxiety disorders. Through the analysis of 20 randomized controlled trials, we found that CBT is a highly effective treatment for anxiety disorders, with large effect sizes across a range of anxiety disorders, including generalized anxiety disorder, panic disorder, and social anxiety disorder. Our findings support the use of CBT as a first-line treatment for anxiety disorders and highlight the importance of further research to identify the mechanisms underlying its effectiveness.

Title : “Exploring the Role of Parental Involvement in Children’s Education: A Qualitative Study”

Abstract : This qualitative study explores the role of parental involvement in children’s education. Through in-depth interviews with 20 parents of children in elementary school, we found that parental involvement takes many forms, including volunteering in the classroom, helping with homework, and communicating with teachers. We also found that parental involvement is influenced by a range of factors, including parent and child characteristics, school culture, and socio-economic status. Our findings suggest that schools and educators should prioritize building strong partnerships with parents to support children’s academic success.

Title : “The Impact of Exercise on Cognitive Function in Older Adults: A Systematic Review and Meta-Analysis”

Abstract : This paper presents a systematic review and meta-analysis of the existing literature on the impact of exercise on cognitive function in older adults. Through the analysis of 25 randomized controlled trials, we found that exercise is associated with significant improvements in cognitive function, particularly in the domains of executive function and attention. Our findings highlight the potential of exercise as a non-pharmacological intervention to support cognitive health in older adults.

When to Write Research Paper Abstract

The abstract of a research paper should typically be written after you have completed the main body of the paper. This is because the abstract is intended to provide a brief summary of the key points and findings of the research, and you can’t do that until you have completed the research and written about it in detail.

Once you have completed your research paper, you can begin writing your abstract. It is important to remember that the abstract should be a concise summary of your research paper, and should be written in a way that is easy to understand for readers who may not have expertise in your specific area of research.

Purpose of Research Paper Abstract

The purpose of a research paper abstract is to provide a concise summary of the key points and findings of a research paper. It is typically a brief paragraph or two that appears at the beginning of the paper, before the introduction, and is intended to give readers a quick overview of the paper’s content.

The abstract should include a brief statement of the research problem, the methods used to investigate the problem, the key results and findings, and the main conclusions and implications of the research. It should be written in a clear and concise manner, avoiding jargon and technical language, and should be understandable to a broad audience.

The abstract serves as a way to quickly and easily communicate the main points of a research paper to potential readers, such as academics, researchers, and students, who may be looking for information on a particular topic. It can also help researchers determine whether a paper is relevant to their own research interests and whether they should read the full paper.

About the author

' src=

Muhammad Hassan

Researcher, Academic Writer, Web developer

You may also like

Research Paper Conclusion

Research Paper Conclusion – Writing Guide and...

Appendices

Appendices – Writing Guide, Types and Examples

Research Paper Citation

How to Cite Research Paper – All Formats and...

Delimitations

Delimitations in Research – Types, Examples and...

Scope of the Research

Scope of the Research – Writing Guide and...

Research Contribution

Research Contribution – Thesis Guide

The Writing Center • University of North Carolina at Chapel Hill

What this handout is about

This handout provides definitions and examples of the two main types of abstracts: descriptive and informative. It also provides guidelines for constructing an abstract and general tips for you to keep in mind when drafting. Finally, it includes a few examples of abstracts broken down into their component parts.

What is an abstract?

An abstract is a self-contained, short, and powerful statement that describes a larger work. Components vary according to discipline. An abstract of a social science or scientific work may contain the scope, purpose, results, and contents of the work. An abstract of a humanities work may contain the thesis, background, and conclusion of the larger work. An abstract is not a review, nor does it evaluate the work being abstracted. While it contains key words found in the larger work, the abstract is an original document rather than an excerpted passage.

Why write an abstract?

You may write an abstract for various reasons. The two most important are selection and indexing. Abstracts allow readers who may be interested in a longer work to quickly decide whether it is worth their time to read it. Also, many online databases use abstracts to index larger works. Therefore, abstracts should contain keywords and phrases that allow for easy searching.

Say you are beginning a research project on how Brazilian newspapers helped Brazil’s ultra-liberal president Luiz Ignácio da Silva wrest power from the traditional, conservative power base. A good first place to start your research is to search Dissertation Abstracts International for all dissertations that deal with the interaction between newspapers and politics. “Newspapers and politics” returned 569 hits. A more selective search of “newspapers and Brazil” returned 22 hits. That is still a fair number of dissertations. Titles can sometimes help winnow the field, but many titles are not very descriptive. For example, one dissertation is titled “Rhetoric and Riot in Rio de Janeiro.” It is unclear from the title what this dissertation has to do with newspapers in Brazil. One option would be to download or order the entire dissertation on the chance that it might speak specifically to the topic. A better option is to read the abstract. In this case, the abstract reveals the main focus of the dissertation:

This dissertation examines the role of newspaper editors in the political turmoil and strife that characterized late First Empire Rio de Janeiro (1827-1831). Newspaper editors and their journals helped change the political culture of late First Empire Rio de Janeiro by involving the people in the discussion of state. This change in political culture is apparent in Emperor Pedro I’s gradual loss of control over the mechanisms of power. As the newspapers became more numerous and powerful, the Emperor lost his legitimacy in the eyes of the people. To explore the role of the newspapers in the political events of the late First Empire, this dissertation analyzes all available newspapers published in Rio de Janeiro from 1827 to 1831. Newspapers and their editors were leading forces in the effort to remove power from the hands of the ruling elite and place it under the control of the people. In the process, newspapers helped change how politics operated in the constitutional monarchy of Brazil.

From this abstract you now know that although the dissertation has nothing to do with modern Brazilian politics, it does cover the role of newspapers in changing traditional mechanisms of power. After reading the abstract, you can make an informed judgment about whether the dissertation would be worthwhile to read.

Besides selection, the other main purpose of the abstract is for indexing. Most article databases in the online catalog of the library enable you to search abstracts. This allows for quick retrieval by users and limits the extraneous items recalled by a “full-text” search. However, for an abstract to be useful in an online retrieval system, it must incorporate the key terms that a potential researcher would use to search. For example, if you search Dissertation Abstracts International using the keywords “France” “revolution” and “politics,” the search engine would search through all the abstracts in the database that included those three words. Without an abstract, the search engine would be forced to search titles, which, as we have seen, may not be fruitful, or else search the full text. It’s likely that a lot more than 60 dissertations have been written with those three words somewhere in the body of the entire work. By incorporating keywords into the abstract, the author emphasizes the central topics of the work and gives prospective readers enough information to make an informed judgment about the applicability of the work.

When do people write abstracts?

  • when submitting articles to journals, especially online journals
  • when applying for research grants
  • when writing a book proposal
  • when completing the Ph.D. dissertation or M.A. thesis
  • when writing a proposal for a conference paper
  • when writing a proposal for a book chapter

Most often, the author of the entire work (or prospective work) writes the abstract. However, there are professional abstracting services that hire writers to draft abstracts of other people’s work. In a work with multiple authors, the first author usually writes the abstract. Undergraduates are sometimes asked to draft abstracts of books/articles for classmates who have not read the larger work.

Types of abstracts

There are two types of abstracts: descriptive and informative. They have different aims, so as a consequence they have different components and styles. There is also a third type called critical, but it is rarely used. If you want to find out more about writing a critique or a review of a work, see the UNC Writing Center handout on writing a literature review . If you are unsure which type of abstract you should write, ask your instructor (if the abstract is for a class) or read other abstracts in your field or in the journal where you are submitting your article.

Descriptive abstracts

A descriptive abstract indicates the type of information found in the work. It makes no judgments about the work, nor does it provide results or conclusions of the research. It does incorporate key words found in the text and may include the purpose, methods, and scope of the research. Essentially, the descriptive abstract describes the work being abstracted. Some people consider it an outline of the work, rather than a summary. Descriptive abstracts are usually very short—100 words or less.

Informative abstracts

The majority of abstracts are informative. While they still do not critique or evaluate a work, they do more than describe it. A good informative abstract acts as a surrogate for the work itself. That is, the writer presents and explains all the main arguments and the important results and evidence in the complete article/paper/book. An informative abstract includes the information that can be found in a descriptive abstract (purpose, methods, scope) but also includes the results and conclusions of the research and the recommendations of the author. The length varies according to discipline, but an informative abstract is rarely more than 10% of the length of the entire work. In the case of a longer work, it may be much less.

Here are examples of a descriptive and an informative abstract of this handout on abstracts . Descriptive abstract:

The two most common abstract types—descriptive and informative—are described and examples of each are provided.

Informative abstract:

Abstracts present the essential elements of a longer work in a short and powerful statement. The purpose of an abstract is to provide prospective readers the opportunity to judge the relevance of the longer work to their projects. Abstracts also include the key terms found in the longer work and the purpose and methods of the research. Authors abstract various longer works, including book proposals, dissertations, and online journal articles. There are two main types of abstracts: descriptive and informative. A descriptive abstract briefly describes the longer work, while an informative abstract presents all the main arguments and important results. This handout provides examples of various types of abstracts and instructions on how to construct one.

Which type should I use?

Your best bet in this case is to ask your instructor or refer to the instructions provided by the publisher. You can also make a guess based on the length allowed; i.e., 100-120 words = descriptive; 250+ words = informative.

How do I write an abstract?

The format of your abstract will depend on the work being abstracted. An abstract of a scientific research paper will contain elements not found in an abstract of a literature article, and vice versa. However, all abstracts share several mandatory components, and there are also some optional parts that you can decide to include or not. When preparing to draft your abstract, keep the following key process elements in mind:

  • Reason for writing: What is the importance of the research? Why would a reader be interested in the larger work?
  • Problem: What problem does this work attempt to solve? What is the scope of the project? What is the main argument/thesis/claim?
  • Methodology: An abstract of a scientific work may include specific models or approaches used in the larger study. Other abstracts may describe the types of evidence used in the research.
  • Results: Again, an abstract of a scientific work may include specific data that indicates the results of the project. Other abstracts may discuss the findings in a more general way.
  • Implications: What changes should be implemented as a result of the findings of the work? How does this work add to the body of knowledge on the topic?

(This list of elements is adapted with permission from Philip Koopman, “How to Write an Abstract.” )

All abstracts include:

  • A full citation of the source, preceding the abstract.
  • The most important information first.
  • The same type and style of language found in the original, including technical language.
  • Key words and phrases that quickly identify the content and focus of the work.
  • Clear, concise, and powerful language.

Abstracts may include:

  • The thesis of the work, usually in the first sentence.
  • Background information that places the work in the larger body of literature.
  • The same chronological structure as the original work.

How not to write an abstract:

  • Do not refer extensively to other works.
  • Do not add information not contained in the original work.
  • Do not define terms.

If you are abstracting your own writing

When abstracting your own work, it may be difficult to condense a piece of writing that you have agonized over for weeks (or months, or even years) into a 250-word statement. There are some tricks that you could use to make it easier, however.

Reverse outlining:

This technique is commonly used when you are having trouble organizing your own writing. The process involves writing down the main idea of each paragraph on a separate piece of paper– see our short video . For the purposes of writing an abstract, try grouping the main ideas of each section of the paper into a single sentence. Practice grouping ideas using webbing or color coding .

For a scientific paper, you may have sections titled Purpose, Methods, Results, and Discussion. Each one of these sections will be longer than one paragraph, but each is grouped around a central idea. Use reverse outlining to discover the central idea in each section and then distill these ideas into one statement.

Cut and paste:

To create a first draft of an abstract of your own work, you can read through the entire paper and cut and paste sentences that capture key passages. This technique is useful for social science research with findings that cannot be encapsulated by neat numbers or concrete results. A well-written humanities draft will have a clear and direct thesis statement and informative topic sentences for paragraphs or sections. Isolate these sentences in a separate document and work on revising them into a unified paragraph.

If you are abstracting someone else’s writing

When abstracting something you have not written, you cannot summarize key ideas just by cutting and pasting. Instead, you must determine what a prospective reader would want to know about the work. There are a few techniques that will help you in this process:

Identify key terms:

Search through the entire document for key terms that identify the purpose, scope, and methods of the work. Pay close attention to the Introduction (or Purpose) and the Conclusion (or Discussion). These sections should contain all the main ideas and key terms in the paper. When writing the abstract, be sure to incorporate the key terms.

Highlight key phrases and sentences:

Instead of cutting and pasting the actual words, try highlighting sentences or phrases that appear to be central to the work. Then, in a separate document, rewrite the sentences and phrases in your own words.

Don’t look back:

After reading the entire work, put it aside and write a paragraph about the work without referring to it. In the first draft, you may not remember all the key terms or the results, but you will remember what the main point of the work was. Remember not to include any information you did not get from the work being abstracted.

Revise, revise, revise

No matter what type of abstract you are writing, or whether you are abstracting your own work or someone else’s, the most important step in writing an abstract is to revise early and often. When revising, delete all extraneous words and incorporate meaningful and powerful words. The idea is to be as clear and complete as possible in the shortest possible amount of space. The Word Count feature of Microsoft Word can help you keep track of how long your abstract is and help you hit your target length.

Example 1: Humanities abstract

Kenneth Tait Andrews, “‘Freedom is a constant struggle’: The dynamics and consequences of the Mississippi Civil Rights Movement, 1960-1984” Ph.D. State University of New York at Stony Brook, 1997 DAI-A 59/02, p. 620, Aug 1998

This dissertation examines the impacts of social movements through a multi-layered study of the Mississippi Civil Rights Movement from its peak in the early 1960s through the early 1980s. By examining this historically important case, I clarify the process by which movements transform social structures and the constraints movements face when they try to do so. The time period studied includes the expansion of voting rights and gains in black political power, the desegregation of public schools and the emergence of white-flight academies, and the rise and fall of federal anti-poverty programs. I use two major research strategies: (1) a quantitative analysis of county-level data and (2) three case studies. Data have been collected from archives, interviews, newspapers, and published reports. This dissertation challenges the argument that movements are inconsequential. Some view federal agencies, courts, political parties, or economic elites as the agents driving institutional change, but typically these groups acted in response to the leverage brought to bear by the civil rights movement. The Mississippi movement attempted to forge independent structures for sustaining challenges to local inequities and injustices. By propelling change in an array of local institutions, movement infrastructures had an enduring legacy in Mississippi.

Now let’s break down this abstract into its component parts to see how the author has distilled his entire dissertation into a ~200 word abstract.

What the dissertation does This dissertation examines the impacts of social movements through a multi-layered study of the Mississippi Civil Rights Movement from its peak in the early 1960s through the early 1980s. By examining this historically important case, I clarify the process by which movements transform social structures and the constraints movements face when they try to do so.

How the dissertation does it The time period studied in this dissertation includes the expansion of voting rights and gains in black political power, the desegregation of public schools and the emergence of white-flight academies, and the rise and fall of federal anti-poverty programs. I use two major research strategies: (1) a quantitative analysis of county-level data and (2) three case studies.

What materials are used Data have been collected from archives, interviews, newspapers, and published reports.

Conclusion This dissertation challenges the argument that movements are inconsequential. Some view federal agencies, courts, political parties, or economic elites as the agents driving institutional change, but typically these groups acted in response to movement demands and the leverage brought to bear by the civil rights movement. The Mississippi movement attempted to forge independent structures for sustaining challenges to local inequities and injustices. By propelling change in an array of local institutions, movement infrastructures had an enduring legacy in Mississippi.

Keywords social movements Civil Rights Movement Mississippi voting rights desegregation

Example 2: Science Abstract

Luis Lehner, “Gravitational radiation from black hole spacetimes” Ph.D. University of Pittsburgh, 1998 DAI-B 59/06, p. 2797, Dec 1998

The problem of detecting gravitational radiation is receiving considerable attention with the construction of new detectors in the United States, Europe, and Japan. The theoretical modeling of the wave forms that would be produced in particular systems will expedite the search for and analysis of detected signals. The characteristic formulation of GR is implemented to obtain an algorithm capable of evolving black holes in 3D asymptotically flat spacetimes. Using compactification techniques, future null infinity is included in the evolved region, which enables the unambiguous calculation of the radiation produced by some compact source. A module to calculate the waveforms is constructed and included in the evolution algorithm. This code is shown to be second-order convergent and to handle highly non-linear spacetimes. In particular, we have shown that the code can handle spacetimes whose radiation is equivalent to a galaxy converting its whole mass into gravitational radiation in one second. We further use the characteristic formulation to treat the region close to the singularity in black hole spacetimes. The code carefully excises a region surrounding the singularity and accurately evolves generic black hole spacetimes with apparently unlimited stability.

This science abstract covers much of the same ground as the humanities one, but it asks slightly different questions.

Why do this study The problem of detecting gravitational radiation is receiving considerable attention with the construction of new detectors in the United States, Europe, and Japan. The theoretical modeling of the wave forms that would be produced in particular systems will expedite the search and analysis of the detected signals.

What the study does The characteristic formulation of GR is implemented to obtain an algorithm capable of evolving black holes in 3D asymptotically flat spacetimes. Using compactification techniques, future null infinity is included in the evolved region, which enables the unambiguous calculation of the radiation produced by some compact source. A module to calculate the waveforms is constructed and included in the evolution algorithm.

Results This code is shown to be second-order convergent and to handle highly non-linear spacetimes. In particular, we have shown that the code can handle spacetimes whose radiation is equivalent to a galaxy converting its whole mass into gravitational radiation in one second. We further use the characteristic formulation to treat the region close to the singularity in black hole spacetimes. The code carefully excises a region surrounding the singularity and accurately evolves generic black hole spacetimes with apparently unlimited stability.

Keywords gravitational radiation (GR) spacetimes black holes

Works consulted

We consulted these works while writing this handout. This is not a comprehensive list of resources on the handout’s topic, and we encourage you to do your own research to find additional publications. Please do not use this list as a model for the format of your own reference list, as it may not match the citation style you are using. For guidance on formatting citations, please see the UNC Libraries citation tutorial . We revise these tips periodically and welcome feedback.

Belcher, Wendy Laura. 2009. Writing Your Journal Article in Twelve Weeks: A Guide to Academic Publishing Success. Thousand Oaks, CA: Sage Press.

Koopman, Philip. 1997. “How to Write an Abstract.” Carnegie Mellon University. October 1997. http://users.ece.cmu.edu/~koopman/essays/abstract.html .

Lancaster, F.W. 2003. Indexing And Abstracting in Theory and Practice , 3rd ed. London: Facet Publishing.

You may reproduce it for non-commercial use if you use the entire handout and attribute the source: The Writing Center, University of North Carolina at Chapel Hill

Make a Gift

Have a language expert improve your writing

Run a free plagiarism check in 10 minutes, automatically generate references for free.

  • Knowledge Base
  • Dissertation
  • How to Write an Abstract | Steps & Examples

How to Write an Abstract | Steps & Examples

Published on 1 March 2019 by Shona McCombes . Revised on 10 October 2022 by Eoghan Ryan.

An abstract is a short summary of a longer work (such as a dissertation or research paper ). The abstract concisely reports the aims and outcomes of your research, so that readers know exactly what your paper is about.

Although the structure may vary slightly depending on your discipline, your abstract should describe the purpose of your work, the methods you’ve used, and the conclusions you’ve drawn.

One common way to structure your abstract is to use the IMRaD structure. This stands for:

  • Introduction

Abstracts are usually around 100–300 words, but there’s often a strict word limit, so make sure to check the relevant requirements.

In a dissertation or thesis , include the abstract on a separate page, after the title page and acknowledgements but before the table of contents .

Instantly correct all language mistakes in your text

Be assured that you'll submit flawless writing. Upload your document to correct all your mistakes.

upload-your-document-ai-proofreader

Table of contents

Abstract example, when to write an abstract, step 1: introduction, step 2: methods, step 3: results, step 4: discussion, tips for writing an abstract, frequently asked questions about abstracts.

Hover over the different parts of the abstract to see how it is constructed.

This paper examines the role of silent movies as a mode of shared experience in the UK during the early twentieth century. At this time, high immigration rates resulted in a significant percentage of non-English-speaking citizens. These immigrants faced numerous economic and social obstacles, including exclusion from public entertainment and modes of discourse (newspapers, theater, radio).

Incorporating evidence from reviews, personal correspondence, and diaries, this study demonstrates that silent films were an affordable and inclusive source of entertainment. It argues for the accessible economic and representational nature of early cinema. These concerns are particularly evident in the low price of admission and in the democratic nature of the actors’ exaggerated gestures, which allowed the plots and action to be easily grasped by a diverse audience despite language barriers.

Keywords: silent movies, immigration, public discourse, entertainment, early cinema, language barriers.

The only proofreading tool specialized in correcting academic writing

The academic proofreading tool has been trained on 1000s of academic texts and by native English editors. Making it the most accurate and reliable proofreading tool for students.

how long is an abstract research paper

Correct my document today

You will almost always have to include an abstract when:

  • Completing a thesis or dissertation
  • Submitting a research paper to an academic journal
  • Writing a book proposal
  • Applying for research grants

It’s easiest to write your abstract last, because it’s a summary of the work you’ve already done. Your abstract should:

  • Be a self-contained text, not an excerpt from your paper
  • Be fully understandable on its own
  • Reflect the structure of your larger work

Start by clearly defining the purpose of your research. What practical or theoretical problem does the research respond to, or what research question did you aim to answer?

You can include some brief context on the social or academic relevance of your topic, but don’t go into detailed background information. If your abstract uses specialised terms that would be unfamiliar to the average academic reader or that have various different meanings, give a concise definition.

After identifying the problem, state the objective of your research. Use verbs like “investigate,” “test,” “analyse,” or “evaluate” to describe exactly what you set out to do.

This part of the abstract can be written in the present or past simple tense  but should never refer to the future, as the research is already complete.

  • This study will investigate the relationship between coffee consumption and productivity.
  • This study investigates the relationship between coffee consumption and productivity.

Next, indicate the research methods that you used to answer your question. This part should be a straightforward description of what you did in one or two sentences. It is usually written in the past simple tense, as it refers to completed actions.

  • Structured interviews will be conducted with 25 participants.
  • Structured interviews were conducted with 25 participants.

Don’t evaluate validity or obstacles here — the goal is not to give an account of the methodology’s strengths and weaknesses, but to give the reader a quick insight into the overall approach and procedures you used.

Prevent plagiarism, run a free check.

Next, summarise the main research results . This part of the abstract can be in the present or past simple tense.

  • Our analysis has shown a strong correlation between coffee consumption and productivity.
  • Our analysis shows a strong correlation between coffee consumption and productivity.
  • Our analysis showed a strong correlation between coffee consumption and productivity.

Depending on how long and complex your research is, you may not be able to include all results here. Try to highlight only the most important findings that will allow the reader to understand your conclusions.

Finally, you should discuss the main conclusions of your research : what is your answer to the problem or question? The reader should finish with a clear understanding of the central point that your research has proved or argued. Conclusions are usually written in the present simple tense.

  • We concluded that coffee consumption increases productivity.
  • We conclude that coffee consumption increases productivity.

If there are important limitations to your research (for example, related to your sample size or methods), you should mention them briefly in the abstract. This allows the reader to accurately assess the credibility and generalisability of your research.

If your aim was to solve a practical problem, your discussion might include recommendations for implementation. If relevant, you can briefly make suggestions for further research.

If your paper will be published, you might have to add a list of keywords at the end of the abstract. These keywords should reference the most important elements of the research to help potential readers find your paper during their own literature searches.

Be aware that some publication manuals, such as APA Style , have specific formatting requirements for these keywords.

It can be a real challenge to condense your whole work into just a couple of hundred words, but the abstract will be the first (and sometimes only) part that people read, so it’s important to get it right. These strategies can help you get started.

Read other abstracts

The best way to learn the conventions of writing an abstract in your discipline is to read other people’s. You probably already read lots of journal article abstracts while conducting your literature review —try using them as a framework for structure and style.

You can also find lots of dissertation abstract examples in thesis and dissertation databases .

Reverse outline

Not all abstracts will contain precisely the same elements. For longer works, you can write your abstract through a process of reverse outlining.

For each chapter or section, list keywords and draft one to two sentences that summarise the central point or argument. This will give you a framework of your abstract’s structure. Next, revise the sentences to make connections and show how the argument develops.

Write clearly and concisely

A good abstract is short but impactful, so make sure every word counts. Each sentence should clearly communicate one main point.

To keep your abstract or summary short and clear:

  • Avoid passive sentences: Passive constructions are often unnecessarily long. You can easily make them shorter and clearer by using the active voice.
  • Avoid long sentences: Substitute longer expressions for concise expressions or single words (e.g., “In order to” for “To”).
  • Avoid obscure jargon: The abstract should be understandable to readers who are not familiar with your topic.
  • Avoid repetition and filler words: Replace nouns with pronouns when possible and eliminate unnecessary words.
  • Avoid detailed descriptions: An abstract is not expected to provide detailed definitions, background information, or discussions of other scholars’ work. Instead, include this information in the body of your thesis or paper.

If you’re struggling to edit down to the required length, you can get help from expert editors with Scribbr’s professional proofreading services .

Check your formatting

If you are writing a thesis or dissertation or submitting to a journal, there are often specific formatting requirements for the abstract—make sure to check the guidelines and format your work correctly. For APA research papers you can follow the APA abstract format .

Checklist: Abstract

The word count is within the required length, or a maximum of one page.

The abstract appears after the title page and acknowledgements and before the table of contents .

I have clearly stated my research problem and objectives.

I have briefly described my methodology .

I have summarized the most important results .

I have stated my main conclusions .

I have mentioned any important limitations and recommendations.

The abstract can be understood by someone without prior knowledge of the topic.

You've written a great abstract! Use the other checklists to continue improving your thesis or dissertation.

An abstract is a concise summary of an academic text (such as a journal article or dissertation ). It serves two main purposes:

  • To help potential readers determine the relevance of your paper for their own research.
  • To communicate your key findings to those who don’t have time to read the whole paper.

Abstracts are often indexed along with keywords on academic databases, so they make your work more easily findable. Since the abstract is the first thing any reader sees, it’s important that it clearly and accurately summarises the contents of your paper.

An abstract for a thesis or dissertation is usually around 150–300 words. There’s often a strict word limit, so make sure to check your university’s requirements.

The abstract is the very last thing you write. You should only write it after your research is complete, so that you can accurately summarize the entirety of your thesis or paper.

Avoid citing sources in your abstract . There are two reasons for this:

  • The abstract should focus on your original research, not on the work of others.
  • The abstract should be self-contained and fully understandable without reference to other sources.

There are some circumstances where you might need to mention other sources in an abstract: for example, if your research responds directly to another study or focuses on the work of a single theorist. In general, though, don’t include citations unless absolutely necessary.

The abstract appears on its own page, after the title page and acknowledgements but before the table of contents .

Cite this Scribbr article

If you want to cite this source, you can copy and paste the citation or click the ‘Cite this Scribbr article’ button to automatically add the citation to our free Reference Generator.

McCombes, S. (2022, October 10). How to Write an Abstract | Steps & Examples. Scribbr. Retrieved 26 February 2024, from https://www.scribbr.co.uk/thesis-dissertation/abstract/

Is this article helpful?

Shona McCombes

Shona McCombes

Other students also liked, how to write a thesis or dissertation introduction, thesis & dissertation acknowledgements | tips & examples, dissertation title page.

  • Affiliate Program

Wordvice

  • UNITED STATES
  • 台灣 (TAIWAN)
  • TÜRKIYE (TURKEY)
  • Academic Editing Services
  • - Research Paper
  • - Journal Manuscript
  • - Dissertation
  • - College & University Assignments
  • Admissions Editing Services
  • - Application Essay
  • - Personal Statement
  • - Recommendation Letter
  • - Cover Letter
  • - CV/Resume
  • Business Editing Services
  • - Business Documents
  • - Report & Brochure
  • - Website & Blog
  • Writer Editing Services
  • - Script & Screenplay
  • Our Editors
  • Client Reviews
  • Editing & Proofreading Prices
  • Wordvice Points
  • Partner Discount
  • Plagiarism Checker
  • APA Citation Generator
  • MLA Citation Generator
  • Chicago Citation Generator
  • Vancouver Citation Generator
  • - APA Style
  • - MLA Style
  • - Chicago Style
  • - Vancouver Style
  • Writing & Editing Guide
  • Academic Resources
  • Admissions Resources

How to Write an Abstract for a Research Paper | Examples

how long is an abstract research paper

What is a research paper abstract?

Research paper abstracts summarize your study quickly and succinctly to journal editors and researchers and prompt them to read further. But with the ubiquity of online publication databases, writing a compelling abstract is even more important today than it was in the days of bound paper manuscripts.

Abstracts exist to “sell”  your work, and they could thus be compared to the “executive summary” of a business resume: an official briefing on what is most important about your research. Or the “gist” of your research. With the majority of academic transactions being conducted online, this means that you have even less time to impress readers–and increased competition in terms of other abstracts out there to read.

The APCI (Academic Publishing and Conferences International) notes that there are  12 questions or “points” considered in the selection process  for journals and conferences and stresses the importance of having an abstract that ticks all of these boxes. Because it is often the ONLY chance you have to convince readers to keep reading, it is important that you spend time and energy crafting an abstract that faithfully represents the central parts of your study and captivates your audience.

With that in mind, follow these suggestions when structuring and writing your abstract, and learn how exactly to put these ideas into a solid abstract that will captivate your target readers.

Before Writing Your Abstract

How long should an abstract be.

All abstracts are written with the same essential objective: to give a summary of your study. But there are two basic styles of abstract: descriptive and informative . Here is a brief delineation of the two:

Of the two types of abstracts, informative abstracts are much more common, and they are widely used for submission to journals and conferences. Informative abstracts apply to lengthier and more technical research and are common in the sciences, engineering, and psychology, while descriptive abstracts are more likely used in humanities and social science papers. The best method of determining which abstract type you need to use is to follow the instructions for journal submissions and to read as many other published articles in those journals as possible.

Research Abstract Guidelines and Requirements

As any article about research writing will tell you, authors must always closely follow the specific guidelines and requirements indicated in the Guide for Authors section of their target journal’s website. The same kind of adherence to conventions should be applied to journal publications, for consideration at a conference, and even when completing a class assignment.

Each publisher has particular demands when it comes to formatting and structure. Here are some common questions addressed in the journal guidelines:

  • Is there a maximum or minimum word/character length?
  • What are the style and formatting requirements?
  • What is the appropriate abstract type?
  • Are there any specific content or organization rules that apply?

There are of course other rules to consider when composing a research paper abstract. But if you follow the stated rules the first time you submit your manuscript, you can avoid your work being thrown in the “circular file” right off the bat.

Identify Your Target Readership

The main purpose of your abstract is to lead researchers to the full text of your research paper. In scientific journals, abstracts let readers decide whether the research discussed is relevant to their own interests or study. Abstracts also help readers understand your main argument quickly. Consider these questions as you write your abstract:

  • Are other academics in your field the main target of your study?
  • Will your study perhaps be useful to members of the general public?
  • Do your study results include the wider implications presented in the abstract?

Outlining and Writing Your Abstract

What to include in an abstract.

Just as your  research paper title  should cover as much ground as possible in a few short words, your abstract must cover  all  parts of your study in order to fully explain your paper and research. Because it must accomplish this task in the space of only a few hundred words, it is important not to include ambiguous references or phrases that will confuse the reader or mislead them about the content and objectives of your research. Follow these  dos  and  don’ts  when it comes to what kind of writing to include:

  • Avoid acronyms or abbreviations since these will need to be explained in order to make sense to the reader, which takes up valuable abstract space. Instead, explain these terms in the Introduction section of the main text.
  • Only use references to people or other works if they are well-known. Otherwise, avoid referencing anything outside of your study in the abstract.
  • Never include tables, figures, sources, or long quotations in your abstract; you will have plenty of time to present and refer to these in the body of your paper.

Use keywords in your abstract to focus your topic

A vital search tool is the research paper keywords section, which lists the most relevant terms directly underneath the abstract. Think of these keywords as the “tubes” that readers will seek and enter—via queries on databases and search engines—to ultimately land at their destination, which is your paper. Your abstract keywords should thus be words that are commonly used in searches but should also be highly relevant to your work and found in the text of your abstract. Include 5 to 10 important words or short phrases central to your research in both the abstract and the keywords section.

For example, if you are writing a paper on the prevalence of obesity among lower classes that crosses international boundaries, you should include terms like “obesity,” “prevalence,” “international,” “lower classes,” and “cross-cultural.” These are terms that should net a wide array of people interested in your topic of study. Look at our nine rules for choosing keywords for your research paper if you need more input on this.

Research Paper Abstract Structure

As mentioned above, the abstract (especially the informative abstract) acts as a surrogate or synopsis of your research paper, doing almost as much work as the thousands of words that follow it in the body of the main text. In the hard sciences and most social sciences, the abstract includes the following sections and organizational schema.

Each section is quite compact—only a single sentence or two, although there is room for expansion if one element or statement is particularly interesting or compelling. As the abstract is almost always one long paragraph, the individual sections should naturally merge into one another to create a holistic effect. Use the following as a checklist to ensure that you have included all of the necessary content in your abstract.

how to structure an abstract list

1) Identify your purpose and motivation

So your research is about rabies in Brazilian squirrels. Why is this important? You should start your abstract by explaining why people should care about this study—why is it significant to your field and perhaps to the wider world? And what is the exact purpose of your study; what are you trying to achieve? Start by answering the following questions:

  • What made you decide to do this study or project?
  • Why is this study important to your field or to the lay reader?
  • Why should someone read your entire article?

In summary, the first section of your abstract should include the importance of the research and its impact on related research fields or on the wider scientific domain.

2) Explain the research problem you are addressing

Stating the research problem that your study addresses is the corollary to why your specific study is important and necessary. For instance, even if the issue of “rabies in Brazilian squirrels” is important, what is the problem—the “missing piece of the puzzle”—that your study helps resolve?

You can combine the problem with the motivation section, but from a perspective of organization and clarity, it is best to separate the two. Here are some precise questions to address:

  • What is your research trying to better understand or what problem is it trying to solve?
  • What is the scope of your study—does it try to explain something general or specific?
  • What is your central claim or argument?

3) Discuss your research approach

Your specific study approach is detailed in the Methods and Materials section .  You have already established the importance of the research, your motivation for studying this issue, and the specific problem your paper addresses. Now you need to discuss  how  you solved or made progress on this problem—how you conducted your research. If your study includes your own work or that of your team, describe that here. If in your paper you reviewed the work of others, explain this here. Did you use analytic models? A simulation? A double-blind study? A case study? You are basically showing the reader the internal engine of your research machine and how it functioned in the study. Be sure to:

  • Detail your research—include methods/type of the study, your variables, and the extent of the work
  • Briefly present evidence to support your claim
  • Highlight your most important sources

4) Briefly summarize your results

Here you will give an overview of the outcome of your study. Avoid using too many vague qualitative terms (e.g, “very,” “small,” or “tremendous”) and try to use at least some quantitative terms (i.e., percentages, figures, numbers). Save your qualitative language for the conclusion statement. Answer questions like these:

  • What did your study yield in concrete terms (e.g., trends, figures, correlation between phenomena)?
  • How did your results compare to your hypothesis? Was the study successful?
  • Where there any highly unexpected outcomes or were they all largely predicted?

5) State your conclusion

In the last section of your abstract, you will give a statement about the implications and  limitations of the study . Be sure to connect this statement closely to your results and not the area of study in general. Are the results of this study going to shake up the scientific world? Will they impact how people see “Brazilian squirrels”? Or are the implications minor? Try not to boast about your study or present its impact as  too  far-reaching, as researchers and journals will tend to be skeptical of bold claims in scientific papers. Answer one of these questions:

  • What are the exact effects of these results on my field? On the wider world?
  • What other kind of study would yield further solutions to problems?
  • What other information is needed to expand knowledge in this area?

After Completing the First Draft of Your Abstract

Revise your abstract.

The abstract, like any piece of academic writing, should be revised before being considered complete. Check it for  grammatical and spelling errors  and make sure it is formatted properly.

Get feedback from a peer

Getting a fresh set of eyes to review your abstract is a great way to find out whether you’ve summarized your research well. Find a reader who understands research papers but is not an expert in this field or is not affiliated with your study. Ask your reader to summarize what your study is about (including all key points of each section). This should tell you if you have communicated your key points clearly.

In addition to research peers, consider consulting with a professor or even a specialist or generalist writing center consultant about your abstract. Use any resource that helps you see your work from another perspective.

Consider getting professional editing and proofreading

While peer feedback is quite important to ensure the effectiveness of your abstract content, it may be a good idea to find an academic editor  to fix mistakes in grammar, spelling, mechanics, style, or formatting. The presence of basic errors in the abstract may not affect your content, but it might dissuade someone from reading your entire study. Wordvice provides English editing services that both correct objective errors and enhance the readability and impact of your work.

Additional Abstract Rules and Guidelines

Write your abstract after completing your paper.

Although the abstract goes at the beginning of your manuscript, it does not merely introduce your research topic (that is the job of the title), but rather summarizes your entire paper. Writing the abstract last will ensure that it is complete and consistent with the findings and statements in your paper.

Keep your content in the correct order

Both questions and answers should be organized in a standard and familiar way to make the content easier for readers to absorb. Ideally, it should mimic the overall format of your essay and the classic “introduction,” “body,” and “conclusion” form, even if the parts are not neatly divided as such.

Write the abstract from scratch

Because the abstract is a self-contained piece of writing viewed separately from the body of the paper, you should write it separately as well. Never copy and paste direct quotes from the paper and avoid paraphrasing sentences in the paper. Using new vocabulary and phrases will keep your abstract interesting and free of redundancies while conserving space.

Don’t include too many details in the abstract

Again, the density of your abstract makes it incompatible with including specific points other than possibly names or locations. You can make references to terms, but do not explain or define them in the abstract. Try to strike a balance between being specific to your study and presenting a relatively broad overview of your work.

Wordvice Resources

If you think your abstract is fine now but you need input on abstract writing or require English editing services (including paper editing ), then head over to the Wordvice academic resources page, where you will find many more articles, for example on writing the Results , Methods , and Discussion sections of your manuscript, on choosing a title for your paper , or on how to finalize your journal submission with a strong cover letter .    

How to Write an Abstract APA Format

Saul Mcleod, PhD

Editor-in-Chief for Simply Psychology

BSc (Hons) Psychology, MRes, PhD, University of Manchester

Saul Mcleod, Ph.D., is a qualified psychology teacher with over 18 years experience of working in further and higher education. He has been published in peer-reviewed journals, including the Journal of Clinical Psychology.

Learn about our Editorial Process

Olivia Guy-Evans, MSc

Associate Editor for Simply Psychology

BSc (Hons) Psychology, MSc Psychology of Education

Olivia Guy-Evans is a writer and associate editor for Simply Psychology. She has previously worked in healthcare and educational sectors.

An APA abstract is a brief, comprehensive summary of the contents of an article, research paper, dissertation, or report.

It is written in accordance with the guidelines of the American Psychological Association (APA), which is a widely used format in social and behavioral sciences. 

An APA abstract summarizes, usually in one paragraph of between 150–250 words, the major aspects of a research paper or dissertation in a prescribed sequence that includes:
  • The rationale: the overall purpose of the study, providing a clear context for the research undertaken.
  • Information regarding the method and participants: including materials/instruments, design, procedure, and data analysis.
  • Main findings or trends: effectively highlighting the key outcomes of the hypotheses.
  • Interpretations and conclusion(s): solidify the implications of the research.
  • Keywords related to the study: assist the paper’s discoverability in academic databases.

The abstract should stand alone, be “self-contained,” and make sense to the reader in isolation from the main article.

The purpose of the abstract is to give the reader a quick overview of the essential information before reading the entire article. The abstract is placed on its own page, directly after the title page and before the main body of the paper.

Although the abstract will appear as the very first part of your paper, it’s good practice to write your abstract after you’ve drafted your full paper, so that you know what you’re summarizing.

Note : This page reflects the latest version of the APA Publication Manual (i.e., APA 7), released in October 2019.

Structure of the Abstract

[NOTE: DO NOT separate the components of the abstract – it should be written as a single paragraph. This section is separated to illustrate the abstract’s structure.]

1) The Rationale

One or two sentences describing the overall purpose of the study and the research problem(s) you investigated. You are basically justifying why this study was conducted.

  • What is the importance of the research?
  • Why would a reader be interested in the larger work?
  • For example, are you filling a gap in previous research or applying new methods to take a fresh look at existing ideas or data?
  • Women who are diagnosed with breast cancer can experience an array of psychosocial difficulties; however, social support, particularly from a spouse, has been shown to have a protective function during this time. This study examined the ways in which a woman’s daily mood, pain, and fatigue, and her spouse’s marital satisfaction predict the woman’s report of partner support in the context of breast cancer.
  • The current nursing shortage, high hospital nurse job dissatisfaction, and reports of uneven quality of hospital care are not uniquely American phenomena.
  • Students with special educational needs and disabilities (SEND) are more likely to exhibit behavioral difficulties than their typically developing peers. The aim of this study was to identify specific risk factors that influence variability in behavior difficulties among individuals with SEND.

2) The Method

Information regarding the participants (number, and population). One or two sentences outlining the method, explaining what was done and how. The method is described in the present tense.

  • Pretest data from a larger intervention study and multilevel modeling were used to examine the effects of women’s daily mood, pain, and fatigue and average levels of mood, pain, and fatigue on women’s report of social support received from her partner, as well as how the effects of mood interacted with partners’ marital satisfaction.
  • This paper presents reports from 43,000 nurses from more than 700 hospitals in the United States, Canada, England, Scotland, and Germany in 1998–1999.
  • The study sample comprised 4,228 students with SEND, aged 5–15, drawn from 305 primary and secondary schools across England. Explanatory variables were measured at the individual and school levels at baseline, along with a teacher-reported measure of behavior difficulties (assessed at baseline and the 18-month follow-up).

3) The Results

One or two sentences indicating the main findings or trends found as a result of your analysis. The results are described in the present or past tense.

  • Results show that on days in which women reported higher levels of negative or positive mood, as well as on days they reported more pain and fatigue, they reported receiving more support. Women who, on average, reported higher levels of positive mood tended to report receiving more support than those who, on average, reported lower positive mood. However, average levels of negative mood were not associated with support. Higher average levels of fatigue but not pain were associated with higher support. Finally, women whose husbands reported higher levels of marital satisfaction reported receiving more partner support, but husbands’ marital satisfaction did not moderate the effect of women’s mood on support.
  • Nurses in countries with distinctly different healthcare systems report similar shortcomings in their work environments and the quality of hospital care. While the competence of and relation between nurses and physicians appear satisfactory, core problems in work design and workforce management threaten the provision of care.
  • Hierarchical linear modeling of data revealed that differences between schools accounted for between 13% (secondary) and 15.4% (primary) of the total variance in the development of students’ behavior difficulties, with the remainder attributable to individual differences. Statistically significant risk markers for these problems across both phases of education were being male, eligibility for free school meals, being identified as a bully, and lower academic achievement. Additional risk markers specific to each phase of education at the individual and school levels are also acknowledged.

4) The Conclusion / Implications

A brief summary of your conclusions and implications of the results, described in the present tense. Explain the results and why the study is important to the reader.

  • For example, what changes should be implemented as a result of the findings of the work?
  • How does this work add to the body of knowledge on the topic?

Implications of these findings are discussed relative to assisting couples during this difficult time in their lives.

  • Resolving these issues, which are amenable to managerial intervention, is essential to preserving patient safety and care of consistently high quality.
  • Behavior difficulties are affected by risks across multiple ecological levels. Addressing any one of these potential influences is therefore likely to contribute to the reduction in the problems displayed.

The above examples of abstracts are from the following papers:

Aiken, L. H., Clarke, S. P., Sloane, D. M., Sochalski, J. A., Busse, R., Clarke, H., … & Shamian, J. (2001). Nurses’ reports on hospital care in five countries . Health affairs, 20(3) , 43-53.

Boeding, S. E., Pukay-Martin, N. D., Baucom, D. H., Porter, L. S., Kirby, J. S., Gremore, T. M., & Keefe, F. J. (2014). Couples and breast cancer: Women’s mood and partners’ marital satisfaction predicting support perception . Journal of Family Psychology, 28(5) , 675.

Oldfield, J., Humphrey, N., & Hebron, J. (2017). Risk factors in the development of behavior difficulties among students with special educational needs and disabilities: A multilevel analysis . British journal of educational psychology, 87(2) , 146-169.

5) Keywords

APA style suggests including a list of keywords at the end of the abstract. This is particularly common in academic articles and helps other researchers find your work in databases.

Keywords in an abstract should be selected to help other researchers find your work when searching an online database. These keywords should effectively represent the main topics of your study. Here are some tips for choosing keywords:

Core Concepts: Identify the most important ideas or concepts in your paper. These often include your main research topic, the methods you’ve used, or the theories you’re discussing.

Specificity: Your keywords should be specific to your research. For example, suppose your paper is about the effects of climate change on bird migration patterns in a specific region. In that case, your keywords might include “climate change,” “bird migration,” and the region’s name.

Consistency with Paper: Make sure your keywords are consistent with the terms you’ve used in your paper. For example, if you use the term “adolescent” rather than “teen” in your paper, choose “adolescent” as your keyword, not “teen.”

Jargon and Acronyms: Avoid using too much-specialized jargon or acronyms in your keywords, as these might not be understood or used by all researchers in your field.

Synonyms: Consider including synonyms of your keywords to capture as many relevant searches as possible. For example, if your paper discusses “post-traumatic stress disorder,” you might include “PTSD” as a keyword.

Remember, keywords are a tool for others to find your work, so think about what terms other researchers might use when searching for papers on your topic.

The Abstract SHOULD NOT contain:

Lengthy background or contextual information: The abstract should focus on your research and findings, not general topic background.

Undefined jargon, abbreviations,  or acronyms: The abstract should be accessible to a wide audience, so avoid highly specialized terms without defining them.

Citations: Abstracts typically do not include citations, as they summarize original research.

Incomplete sentences or bulleted lists: The abstract should be a single, coherent paragraph written in complete sentences.

New information not covered in the paper: The abstract should only summarize the paper’s content.

Subjective comments or value judgments: Stick to objective descriptions of your research.

Excessive details on methods or procedures: Keep descriptions of methods brief and focused on main steps.

Speculative or inconclusive statements: The abstract should state the research’s clear findings, not hypotheses or possible interpretations.

  • Any illustration, figure, table, or references to them . All visual aids, data, or extensive details should be included in the main body of your paper, not in the abstract. 
  • Elliptical or incomplete sentences should be avoided in an abstract . The use of ellipses (…), which could indicate incomplete thoughts or omitted text, is not appropriate in an abstract.

APA Style for Abstracts

An APA abstract must be formatted as follows:

Include the running head aligned to the left at the top of the page (professional papers only) and page number. Note, student papers do not require a running head. On the first line, center the heading “Abstract” and bold (do not underlined or italicize). Do not indent the single abstract paragraph (which begins one line below the section title). Double-space the text. Use Times New Roman font in 12 pt. Set one-inch (or 2.54 cm) margins. If you include a “keywords” section at the end of the abstract, indent the first line and italicize the word “Keywords” while leaving the keywords themselves without any formatting.

Example APA Abstract Page

Download this example as a PDF

APA Style Abstract Example

Further Information

  • APA 7th Edition Abstract and Keywords Guide
  • Example APA Abstract
  • How to Write a Good Abstract for a Scientific Paper or Conference Presentation
  • How to Write a Lab Report
  • Writing an APA paper

How long should an APA abstract be?

An APA abstract should typically be between 150 to 250 words long. However, the exact length may vary depending on specific publication or assignment guidelines. It is crucial that it succinctly summarizes the essential elements of the work, including purpose, methods, findings, and conclusions.

Where does the abstract go in an APA paper?

In an APA formatted paper, the abstract is placed on its own page, directly after the title page and before the main body of the paper. It’s typically the second page of the document. It starts with the word “Abstract” (centered and not in bold) at the top of the page, followed by the text of the abstract itself.

What are the 4 C’s of abstract writing?

The 4 C’s of abstract writing are an approach to help you create a well-structured and informative abstract. They are:

Conciseness: An abstract should briefly summarize the key points of your study. Stick to the word limit (typically between 150-250 words for an APA abstract) and avoid unnecessary details.

Clarity: Your abstract should be easy to understand. Avoid jargon and complex sentences. Clearly explain the purpose, methods, results, and conclusions of your study.

Completeness: Even though it’s brief, the abstract should provide a complete overview of your study, including the purpose, methods, key findings, and your interpretation of the results.

Cohesion: The abstract should flow logically from one point to the next, maintaining a coherent narrative about your study. It’s not just a list of disjointed elements; it’s a brief story of your research from start to finish.

What is the abstract of a psychology paper?

An abstract in a psychology paper serves as a snapshot of the paper, allowing readers to quickly understand the purpose, methodology, results, and implications of the research without reading the entire paper. It is generally between 150-250 words long.

Print Friendly, PDF & Email

  • USC Libraries
  • Research Guides

Organizing Your Social Sciences Research Paper

  • 3. The Abstract
  • Purpose of Guide
  • Design Flaws to Avoid
  • Independent and Dependent Variables
  • Glossary of Research Terms
  • Reading Research Effectively
  • Narrowing a Topic Idea
  • Broadening a Topic Idea
  • Extending the Timeliness of a Topic Idea
  • Academic Writing Style
  • Choosing a Title
  • Making an Outline
  • Paragraph Development
  • Research Process Video Series
  • Executive Summary
  • The C.A.R.S. Model
  • Background Information
  • The Research Problem/Question
  • Theoretical Framework
  • Citation Tracking
  • Content Alert Services
  • Evaluating Sources
  • Primary Sources
  • Secondary Sources
  • Tiertiary Sources
  • Scholarly vs. Popular Publications
  • Qualitative Methods
  • Quantitative Methods
  • Insiderness
  • Using Non-Textual Elements
  • Limitations of the Study
  • Common Grammar Mistakes
  • Writing Concisely
  • Avoiding Plagiarism
  • Footnotes or Endnotes?
  • Further Readings
  • Generative AI and Writing
  • USC Libraries Tutorials and Other Guides
  • Bibliography

An abstract summarizes, usually in one paragraph of 300 words or less, the major aspects of the entire paper in a prescribed sequence that includes: 1) the overall purpose of the study and the research problem(s) you investigated; 2) the basic design of the study; 3) major findings or trends found as a result of your analysis; and, 4) a brief summary of your interpretations and conclusions.

Writing an Abstract. The Writing Center. Clarion University, 2009; Writing an Abstract for Your Research Paper. The Writing Center, University of Wisconsin, Madison.

Importance of a Good Abstract

Sometimes your professor will ask you to include an abstract, or general summary of your work, with your research paper. The abstract allows you to elaborate upon each major aspect of the paper and helps readers decide whether they want to read the rest of the paper. Therefore, enough key information [e.g., summary results, observations, trends, etc.] must be included to make the abstract useful to someone who may want to examine your work.

How do you know when you have enough information in your abstract? A simple rule-of-thumb is to imagine that you are another researcher doing a similar study. Then ask yourself: if your abstract was the only part of the paper you could access, would you be happy with the amount of information presented there? Does it tell the whole story about your study? If the answer is "no" then the abstract likely needs to be revised.

How to Write a Research Abstract. Office of Undergraduate Research. University of Kentucky; Staiger, David L. “What Today’s Students Need to Know about Writing Abstracts.” International Journal of Business Communication January 3 (1966): 29-33; Swales, John M. and Christine B. Feak. Abstracts and the Writing of Abstracts . Ann Arbor, MI: University of Michigan Press, 2009.

Structure and Writing Style

I.  Types of Abstracts

To begin, you need to determine which type of abstract you should include with your paper. There are four general types.

Critical Abstract A critical abstract provides, in addition to describing main findings and information, a judgment or comment about the study’s validity, reliability, or completeness. The researcher evaluates the paper and often compares it with other works on the same subject. Critical abstracts are generally 400-500 words in length due to the additional interpretive commentary. These types of abstracts are used infrequently.

Descriptive Abstract A descriptive abstract indicates the type of information found in the work. It makes no judgments about the work, nor does it provide results or conclusions of the research. It does incorporate key words found in the text and may include the purpose, methods, and scope of the research. Essentially, the descriptive abstract only describes the work being summarized. Some researchers consider it an outline of the work, rather than a summary. Descriptive abstracts are usually very short, 100 words or less. Informative Abstract The majority of abstracts are informative. While they still do not critique or evaluate a work, they do more than describe it. A good informative abstract acts as a surrogate for the work itself. That is, the researcher presents and explains all the main arguments and the important results and evidence in the paper. An informative abstract includes the information that can be found in a descriptive abstract [purpose, methods, scope] but it also includes the results and conclusions of the research and the recommendations of the author. The length varies according to discipline, but an informative abstract is usually no more than 300 words in length.

Highlight Abstract A highlight abstract is specifically written to attract the reader’s attention to the study. No pretense is made of there being either a balanced or complete picture of the paper and, in fact, incomplete and leading remarks may be used to spark the reader’s interest. In that a highlight abstract cannot stand independent of its associated article, it is not a true abstract and, therefore, rarely used in academic writing.

II.  Writing Style

Use the active voice when possible , but note that much of your abstract may require passive sentence constructions. Regardless, write your abstract using concise, but complete, sentences. Get to the point quickly and always use the past tense because you are reporting on a study that has been completed.

Abstracts should be formatted as a single paragraph in a block format and with no paragraph indentations. In most cases, the abstract page immediately follows the title page. Do not number the page. Rules set forth in writing manual vary but, in general, you should center the word "Abstract" at the top of the page with double spacing between the heading and the abstract. The final sentences of an abstract concisely summarize your study’s conclusions, implications, or applications to practice and, if appropriate, can be followed by a statement about the need for additional research revealed from the findings.

Composing Your Abstract

Although it is the first section of your paper, the abstract should be written last since it will summarize the contents of your entire paper. A good strategy to begin composing your abstract is to take whole sentences or key phrases from each section of the paper and put them in a sequence that summarizes the contents. Then revise or add connecting phrases or words to make the narrative flow clearly and smoothly. Note that statistical findings should be reported parenthetically [i.e., written in parentheses].

Before handing in your final paper, check to make sure that the information in the abstract completely agrees with what you have written in the paper. Think of the abstract as a sequential set of complete sentences describing the most crucial information using the fewest necessary words. The abstract SHOULD NOT contain:

  • A catchy introductory phrase, provocative quote, or other device to grab the reader's attention,
  • Lengthy background or contextual information,
  • Redundant phrases, unnecessary adverbs and adjectives, and repetitive information;
  • Acronyms or abbreviations,
  • References to other literature [say something like, "current research shows that..." or "studies have indicated..."],
  • Using ellipticals [i.e., ending with "..."] or incomplete sentences,
  • Jargon or terms that may be confusing to the reader,
  • Citations to other works, and
  • Any sort of image, illustration, figure, or table, or references to them.

Abstract. Writing Center. University of Kansas; Abstract. The Structure, Format, Content, and Style of a Journal-Style Scientific Paper. Department of Biology. Bates College; Abstracts. The Writing Center. University of North Carolina; Borko, Harold and Seymour Chatman. "Criteria for Acceptable Abstracts: A Survey of Abstracters' Instructions." American Documentation 14 (April 1963): 149-160; Abstracts. The Writer’s Handbook. Writing Center. University of Wisconsin, Madison; Hartley, James and Lucy Betts. "Common Weaknesses in Traditional Abstracts in hte Social Sciences." Journal of the American Society for Information Science and Technology 60 (October 2009): 2010-2018; Procter, Margaret. The Abstract. University College Writing Centre. University of Toronto; Riordan, Laura. “Mastering the Art of Abstracts.” The Journal of the American Osteopathic Association 115 (January 2015 ): 41-47; Writing Report Abstracts. The Writing Lab and The OWL. Purdue University; Writing Abstracts. Writing Tutorial Services, Center for Innovative Teaching and Learning. Indiana University; Koltay, Tibor. Abstracts and Abstracting: A Genre and Set of Skills for the Twenty-First Century . Oxford, UK: 2010; Writing an Abstract for Your Research Paper. The Writing Center, University of Wisconsin, Madison.

Writing Tip

Never Cite Just the Abstract!

Citing to just a journal article's abstract does not confirm for the reader that you have conducted a thorough or reliable review of the literature. If the full-text is not available, go to the USC Libraries main page and enter the title of the article [NOT the title of the journal]. If the Libraries have a subscription to the journal, the article should appear with a link to the full-text or to the journal publisher page where you can get the article. If the article does not appear, try searching Google Scholar using the link on the USC Libraries main page. If you still can't find the article after doing this, contact a librarian or you can request it from our free i nterlibrary loan and document delivery service .

  • << Previous: Research Process Video Series
  • Next: Executive Summary >>
  • Last Updated: Feb 29, 2024 9:59 AM
  • URL: https://libguides.usc.edu/writingguide

How To Write an Abstract for Any Subject and Publication (With Examples)

How To Write an Abstract for Any Subject and Publication (With Examples)

Table of contents

how long is an abstract research paper

Christian Rigg

An abstract is a short summary of a longer work, such as a study or research paper. The goal is to provide readers with an overview of the purpose, methodology, results, conclusion, and importance of this text.

As a writing coach and part-time academic editor and translator, I’ve read hundreds of abstracts and helped authors draft and refine dozens more. I’ve found that, when writing an abstract, the greatest difficulty lies in balancing brevity, detail, and accessibility.

Fortunately, there’s a simple formula you can use to write a solid abstract for publication, regardless of the subject. What’s more, you can leverage AI to help you write a clear, concise abstract — without losing your voice or sounding unprofessional.

Below you’ll find step-by-step instructions, best practices, examples, and a helpful checklist. 

Key Takeaways

  • An abstract offers a succinct overview of the aims, results, and importance of your research.
  • Check submission guidelines, write clearly and concisely, and use language to “guide” readers through your abstract. 
  • The IMRaD (Introduction, Methodology, Results, and Discussion) approach is simple and effective. 
  • More and more authors are using AI to do the heavy lifting. With the right prompts, AI can save you time and create a cohesive abstract.

Writing an abstract: First steps and best practices

Keep the following in mind as you write your abstract:

  • If you’re submitting to a publication , check for specific guidelines regarding overall length, format, keywords, and the presence or absence of section headings (e.g. “Purpose”). Follow these guidelines exactly.
  • Write concisely and clearly . If you struggle to write concisely, consider using an AI-writing assistant like Wordtune . Simply select text to receive suggestions on how to write a sentence or paragraph more concisely without losing any value.
  • Make your abstract self-contained . Don’t refer to passages in your article or research. If you must include terms that your audience may not be familiar with, such as highly technical jargon or concepts borrowed from another field, offer a brief definition.
  • Use connecting phrases like “for this reason,” “as a result,” and “this led us” to “guide” the reader through your abstract and help them see the connections between your research goal, methodology, results, and conclusions. ‍
  • Read abstracts on similar studies . This gives you a good benchmark and can help you get started. If you’re submitting your abstract to a particular publication, it also gives you a good idea of the type of language and structure they prefer.

Wordtune offers suggestions to make your text clear and concise.

Get Wordtune for free > Get Wordtune for free >

How to write an abstract: The IMRaD Structure

IMRaD stands for Introduction, Methodology, Results, and Discussion (or Conclusion). 

It’s the most common way to structure a research paper and a very simple way to approach your abstract. In some cases, authors even include these section headings in their abstracts. 

Step One: Introduction

Length : About 25% of your abstract

Purpose : Provide context for your research and describe your research objectives. 

Start by introducing your topic. There are two main parts to this:

  • Your research question stated simply and straightforwardly (what missing knowledge does your study aim to answer?). You can use words like “investigate,” “review,” “test,” “analyze,” “study,” and “evaluate” to make it clear how your work relates to the context.
  • A brief overview of the academic, historical, social, or scientific context. This helps the reader understand the importance and relevance of your work. In many cases, starting with context before your research question makes more sense, so feel free to write in that order. 

Regarding context, consider the following: 

how long is an abstract research paper

For example:

Psychologists and neuroscientists have long studied the role of sleep in the formation of new memories. Previous research into how sleep affects memory has often struggled because it’s difficult to measure the quality, stages, and overall impact of sleep accurately. As a result, there’s ongoing debate in the scientific community , and recent research suggests sleep may not be as important as researchers once thought. In this study, we review the evidence and offer a novel conclusion : the same mechanisms thought to mediate sleep-related memory formation also operate during waking hours, particularly quiet wakefulness.  In this example, several contextual cues are offered: it’s a long-standing topic in the literature; previous research is limited due to a specific issue , and there is active scientific debate . The section closes with the research aims: to review the evidence and offer a new conclusion. 

Step Two: Methodology

Purpose : Clearly describe what you did and highlight novelty. 

In this section, provide a clear description of your research methodology. While it’s important to be concise, make sure you’re not being vague. Mention specific frameworks and tools. 

‍ To explore the impact of social media on political engagement, we conducted a study with 200 participants, divided into two groups. The first was exposed to curated political content on social media, while the control group received a neutral feed. Our mixed-method approach combined quantitative engagement metrics analysis and qualitative interviews to assess changes in political participation.

There’s no need to provide an in-depth justification of your approach, although if it’s a novel one, it’s worth highlighting this and explaining what makes it appropriate. For example, " We chose this approach because it offers a clearer image of the structure of proteins involved in the transfer of electrons during cellular respiration ."

Finally, you can omit methodological limitations; we’ll cover these later. 

Step Three: Results

Length : About 35% of your abstract

Purpose : Provide a clear, specific account of your results. 

This section is arguably the most important (and interesting) part of your abstract.

Explain the results of your analysis in a specific and detailed fashion. This isn’t the time to be vague or bury the lead. For example:

“Our survey indicates a marked shift in sedimentary rock composition. In three locations, we observed significant erosion, and mineralogical analysis revealed a high concentration of quartz. Further analysis suggests two major events in the past 200 years, correlating with disturbances in the region.”
"Our survey of the Redstone Canyon region identified a marked shift in sedimentary rock composition from predominantly sandstone to shale, particularly evident in the lower strata. Quantitative analysis showed a 40% increase in shale content compared to previous surveys. In three distinct locations, we observed significant erosion, with up to two meters of topsoil displacement, primarily due to water runoff. Mineralogical analysis revealed an unexpectedly high concentration of quartz (up to 22%) in these eroded areas. Additionally, our seismic retrogression analysis suggests two major seismic events in the past 200 years, correlating with the observed stratification disturbances."

Incidentally, you don’t need to include all of your findings here, only those that will help the reader to understand the next section: your discussion and conclusion (i.e., what the results mean). This will help you keep the results section concise and relevant. 

Step Four: Discussion/Conclusion

Length : About 15%

Purpose : Present what new knowledge you’ve found and why it matters.

Bearing in mind your research question, give a clear account of your conclusions. What new knowledge has been gained? 

The simplest way to do this is in the present tense: “We conclude that…”

You should also briefly explain why this matters. What are the implications of your findings? Be specific and avoid making claims that aren’t directly supported by your research. 

If there are any important limitations (such as population or control group size), you can mention them now. This helps readers assess the credibility and generalizability of your findings. 

You can use these samples for inspiration.

They are divided into introduction , methodology , results , and conclusion.

The rising urbanization rate poses challenges to mental health, an issue garnering increasing attention in recent years. This study aims to analyze the impact of urban green spaces on the mental health of city dwellers. The focus is on how access to parks and natural environments within urban settings contributes to psychological well-being . For this purpose, we employed a cross-sectional survey methodology, targeting residents in three major cities with varying levels of green space availability. We used a combination of GIS mapping to determine green space distribution and structured questionnaires to assess mental health indicators among 1,000 participants . Our results show a clear correlation between access to green spaces and improved mental health outcomes. Residents with frequent access to parks reported 30% lower stress levels and a 25% reduction in symptoms related to anxiety and depression, compared to those with limited access. Additionally, our analysis revealed that green spaces in dense urban areas had a more significant impact than those in less populated districts . We conclude that urban green spaces play a crucial role in enhancing mental health. This underscores the importance of urban planning policies that prioritize green space development as a public health strategy. These findings have significant implications for city planning and public health policy, advocating for the integration of green spaces in urban development to foster mental well-being .

The phenomenon of antibiotic resistance is a growing concern in medical science. This study investigates the effectiveness of novel synthetic peptides as potential antibiotics against multi-drug resistant bacteria. The research specifically examines the impact of these peptides on the cellular integrity and replication processes of resistant bacterial strains . Our methodology involved in vitro testing of three newly synthesized peptides against a panel of bacteria known for high resistance to conventional antibiotics. The bacterial strains included methicillin-resistant Staphylococcus aureus (MRSA) and vancomycin-resistant Enterococci (VRE). We used a combination of microbiological assays and electron microscopy to evaluate the antibacterial activity and the cellular changes induced by the peptides . The results were promising, showing that two of the three peptides effectively inhibited the growth of MRSA and VRE at low concentrations. Electron microscopy revealed significant disruption of bacterial cell walls and membranes, leading to cell lysis. These peptides also demonstrated low toxicity in preliminary mammalian cell culture tests, suggesting a high therapeutic index . Our study provides promising evidence for the use of synthetic peptides in combating antibiotic-resistant bacteria. These findings open new avenues for developing effective treatments against infections caused by drug-resistant pathogens and highlight the potential of peptide-based therapies in future pharmaceutical applications .

The integration of artificial intelligence (AI) in education is a rapidly evolving area of study. This research explores the effectiveness of AI-driven personalized learning systems in enhancing student performance in high school mathematics. The study focuses on understanding how AI customization impacts learning outcomes compared to traditional teaching methods . We conducted a randomized controlled trial involving 500 high school students from five schools, divided into two groups. The experimental group used an AI-based learning platform that adapted to each student's learning pace and style, while the control group continued with standard classroom instruction. The study measured improvements in mathematical understanding and problem-solving skills over a six-month period . The results indicated a significant improvement in the AI group, with a 40% increase in test scores and a 35% rise in problem-solving abilities compared to the control group. Additionally, students using the AI system reported higher levels of engagement and satisfaction with the learning process . In conclusion, the use of AI-driven personalized learning systems shows considerable promise in enhancing educational outcomes in mathematics. This study suggests that AI personalization can be a valuable tool in modern educational strategies, potentially revolutionizing how subjects are taught and learned in schools .

What is the main objective of an abstract?

The goal of an abstract is to provide readers with a concise overview of the purpose, methodology, results, conclusion, and importance of a longer work, such as a research paper or study. 

How long should an abstract be?

Depending on the publication, an abstract should be anywhere from 150 to 250 words. 

What should an abstract include?

An abstract should include an introduction (context + research question), the methodology, the results, and a conclusion (what you found and why it matters).

IMRaD is a simple formula you can follow to write a great abstract for any topic and publication type. Simply follow the instructions above to write each section: Introduction, Methodology, Results, and Discussion/Conclusion.

Be careful to balance detail with brevity, as abstracts are meant to be a short overview of your study. If you struggle with writing concisely and clearly, consider using a writing aid like Wordtune to handle some of the heavy lifting. 

Want to learn more key writing tips? Check out these articles:

  • How to Write Concisely and Effectively (+Examples)
  • Transition Word Examples and How to Use Them Effectively
  • How to Write a Research Paper (+Free AI Research Paper Writer)

Share This Article:

How To Make Money Writing on X (Twitter) in 2024

How To Make Money Writing on X (Twitter) in 2024

How to Write an Impactful Product Review

How to Write an Impactful Product Review

Mastering the Art of Goodbyes: Crafting a Positive Resignation Letter

Mastering the Art of Goodbyes: Crafting a Positive Resignation Letter

Looking for fresh content, thank you your submission has been received.

 alt=

Academic & Employability Skills

Subscribe to academic & employability skills.

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 400 other subscribers.

Email Address

' src=

Writing an abstract - a six point checklist (with samples)

Posted in: abstract , dissertations

how long is an abstract research paper

The abstract is a vital part of any research paper. It is the shop front for your work, and the first stop for your reader. It should provide a clear and succinct summary of your study, and encourage your readers to read more. An effective abstract, therefore should answer the following questions:

  • Why did you do this study or project?
  • What did you do and how?
  • What did you find?
  • What do your findings mean?

So here's our run down of the key elements of a well-written abstract.

  • Size - A succinct and well written abstract should be between approximately 100- 250 words.
  • Background - An effective abstract usually includes some scene-setting information which might include what is already known about the subject, related to the paper in question (a few short sentences).
  • Purpose  - The abstract should also set out the purpose of your research, in other words, what is not known about the subject and hence what the study intended to examine (or what the paper seeks to present).
  • Methods - The methods section should contain enough information to enable the reader to understand what was done, and how. It should include brief details of the research design, sample size, duration of study, and so on.
  • Results - The results section is the most important part of the abstract. This is because readers who skim an abstract do so to learn about the findings of the study. The results section should therefore contain as much detail about the findings as the journal word count permits.
  • Conclusion - This section should contain the most important take-home message of the study, expressed in a few precisely worded sentences. Usually, the finding highlighted here relates to the primary outcomes of the study. However, other important or unexpected findings should also be mentioned. It is also customary, but not essential, to express an opinion about the theoretical or practical implications of the findings, or the importance of their findings for the field. Thus, the conclusions may contain three elements:
  • The primary take-home message
  • Any additional findings of importance
  • Implications for future studies 

abstract 1

Example Abstract 2: Engineering Development and validation of a three-dimensional finite element model of the pelvic bone.

bone

Abstract from: Dalstra, M., Huiskes, R. and Van Erning, L., 1995. Development and validation of a three-dimensional finite element model of the pelvic bone. Journal of biomechanical engineering, 117(3), pp.272-278.

And finally...  A word on abstract types and styles

Abstract types can differ according to subject discipline. You need to determine therefore which type of abstract you should include with your paper. Here are two of the most common types with examples.

Informative Abstract

The majority of abstracts are informative. While they still do not critique or evaluate a work, they do more than describe it. A good informative abstract acts as a surrogate for the work itself. That is, the researcher presents and explains all the main arguments and the important results and evidence in the paper. An informative abstract includes the information that can be found in a descriptive abstract [purpose, methods, scope] but it also includes the results and conclusions of the research and the recommendations of the author. The length varies according to discipline, but an informative abstract is usually no more than 300 words in length.

Descriptive Abstract A descriptive abstract indicates the type of information found in the work. It makes no judgements about the work, nor does it provide results or conclusions of the research. It does incorporate key words found in the text and may include the purpose, methods, and scope of the research. Essentially, the descriptive abstract only describes the work being summarised. Some researchers consider it an outline of the work, rather than a summary. Descriptive abstracts are usually very short, 100 words or less.

(Adapted from: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3136027/ )

Share this:

  • Click to print (Opens in new window)
  • Click to email a link to a friend (Opens in new window)
  • Click to share on WhatsApp (Opens in new window)
  • Click to share on LinkedIn (Opens in new window)
  • Click to share on Twitter (Opens in new window)
  • Click to share on Facebook (Opens in new window)

Click here to cancel reply.

  • Email * (we won't publish this)

Write a response

' src=

Navigating the dissertation process: my tips for final years

Imagine for a moment... After months of hard work and research on a topic you're passionate about, the time has finally come to click the 'Submit' button on your dissertation. You've just completed your longest project to date as part...

Vanda Sigel and another HSS student working on laptops.

8 ways to beat procrastination

Whether you’re writing an assignment or revising for exams, getting started can be hard. Fortunately, there’s lots you can do to turn procrastination into action.

A post-it note reading 'Procrastination' surrounded by balls of screwed-up paper

My takeaways on how to write a scientific report

If you’re in your dissertation writing stage or your course includes writing a lot of scientific reports, but you don’t quite know where and how to start, the Skills Centre can help you get started. I recently attended their ‘How...

Person in a lab coat looking into a microscope doing an experiment in a laboratory. There's a row of test tubes on the bench. The person is writing on a clipboard.

U.S. flag

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

  • Publications
  • Account settings
  • Advanced Search
  • Journal List
  • Saudi J Anaesth
  • v.13(Suppl 1); 2019 Apr

Writing the title and abstract for a research paper: Being concise, precise, and meticulous is the key

Milind s. tullu.

Department of Pediatrics, Seth G.S. Medical College and KEM Hospital, Parel, Mumbai, Maharashtra, India

This article deals with formulating a suitable title and an appropriate abstract for an original research paper. The “title” and the “abstract” are the “initial impressions” of a research article, and hence they need to be drafted correctly, accurately, carefully, and meticulously. Often both of these are drafted after the full manuscript is ready. Most readers read only the title and the abstract of a research paper and very few will go on to read the full paper. The title and the abstract are the most important parts of a research paper and should be pleasant to read. The “title” should be descriptive, direct, accurate, appropriate, interesting, concise, precise, unique, and should not be misleading. The “abstract” needs to be simple, specific, clear, unbiased, honest, concise, precise, stand-alone, complete, scholarly, (preferably) structured, and should not be misrepresentative. The abstract should be consistent with the main text of the paper, especially after a revision is made to the paper and should include the key message prominently. It is very important to include the most important words and terms (the “keywords”) in the title and the abstract for appropriate indexing purpose and for retrieval from the search engines and scientific databases. Such keywords should be listed after the abstract. One must adhere to the instructions laid down by the target journal with regard to the style and number of words permitted for the title and the abstract.

Introduction

This article deals with drafting a suitable “title” and an appropriate “abstract” for an original research paper. Because the “title” and the “abstract” are the “initial impressions” or the “face” of a research article, they need to be drafted correctly, accurately, carefully, meticulously, and consume time and energy.[ 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 ] Often, these are drafted after the complete manuscript draft is ready.[ 2 , 3 , 4 , 5 , 9 , 10 , 11 ] Most readers will read only the title and the abstract of a published research paper, and very few “interested ones” (especially, if the paper is of use to them) will go on to read the full paper.[ 1 , 2 ] One must remember to adhere to the instructions laid down by the “target journal” (the journal for which the author is writing) regarding the style and number of words permitted for the title and the abstract.[ 2 , 4 , 5 , 7 , 8 , 9 , 12 ] Both the title and the abstract are the most important parts of a research paper – for editors (to decide whether to process the paper for further review), for reviewers (to get an initial impression of the paper), and for the readers (as these may be the only parts of the paper available freely and hence, read widely).[ 4 , 8 , 12 ] It may be worth for the novice author to browse through titles and abstracts of several prominent journals (and their target journal as well) to learn more about the wording and styles of the titles and abstracts, as well as the aims and scope of the particular journal.[ 5 , 7 , 9 , 13 ]

The details of the title are discussed under the subheadings of importance, types, drafting, and checklist.

Importance of the title

When a reader browses through the table of contents of a journal issue (hard copy or on website), the title is the “ first detail” or “face” of the paper that is read.[ 2 , 3 , 4 , 5 , 6 , 13 ] Hence, it needs to be simple, direct, accurate, appropriate, specific, functional, interesting, attractive/appealing, concise/brief, precise/focused, unambiguous, memorable, captivating, informative (enough to encourage the reader to read further), unique, catchy, and it should not be misleading.[ 1 , 2 , 3 , 4 , 5 , 6 , 9 , 12 ] It should have “just enough details” to arouse the interest and curiosity of the reader so that the reader then goes ahead with studying the abstract and then (if still interested) the full paper.[ 1 , 2 , 4 , 13 ] Journal websites, electronic databases, and search engines use the words in the title and abstract (the “keywords”) to retrieve a particular paper during a search; hence, the importance of these words in accessing the paper by the readers has been emphasized.[ 3 , 4 , 5 , 6 , 12 , 14 ] Such important words (or keywords) should be arranged in appropriate order of importance as per the context of the paper and should be placed at the beginning of the title (rather than the later part of the title, as some search engines like Google may just display only the first six to seven words of the title).[ 3 , 5 , 12 ] Whimsical, amusing, or clever titles, though initially appealing, may be missed or misread by the busy reader and very short titles may miss the essential scientific words (the “keywords”) used by the indexing agencies to catch and categorize the paper.[ 1 , 3 , 4 , 9 ] Also, amusing or hilarious titles may be taken less seriously by the readers and may be cited less often.[ 4 , 15 ] An excessively long or complicated title may put off the readers.[ 3 , 9 ] It may be a good idea to draft the title after the main body of the text and the abstract are drafted.[ 2 , 3 , 4 , 5 ]

Types of titles

Titles can be descriptive, declarative, or interrogative. They can also be classified as nominal, compound, or full-sentence titles.

Descriptive or neutral title

This has the essential elements of the research theme, that is, the patients/subjects, design, interventions, comparisons/control, and outcome, but does not reveal the main result or the conclusion.[ 3 , 4 , 12 , 16 ] Such a title allows the reader to interpret the findings of the research paper in an impartial manner and with an open mind.[ 3 ] These titles also give complete information about the contents of the article, have several keywords (thus increasing the visibility of the article in search engines), and have increased chances of being read and (then) being cited as well.[ 4 ] Hence, such descriptive titles giving a glimpse of the paper are generally preferred.[ 4 , 16 ]

Declarative title

This title states the main finding of the study in the title itself; it reduces the curiosity of the reader, may point toward a bias on the part of the author, and hence is best avoided.[ 3 , 4 , 12 , 16 ]

Interrogative title

This is the one which has a query or the research question in the title.[ 3 , 4 , 16 ] Though a query in the title has the ability to sensationalize the topic, and has more downloads (but less citations), it can be distracting to the reader and is again best avoided for a research article (but can, at times, be used for a review article).[ 3 , 6 , 16 , 17 ]

From a sentence construct point of view, titles may be nominal (capturing only the main theme of the study), compound (with subtitles to provide additional relevant information such as context, design, location/country, temporal aspect, sample size, importance, and a provocative or a literary; for example, see the title of this review), or full-sentence titles (which are longer and indicate an added degree of certainty of the results).[ 4 , 6 , 9 , 16 ] Any of these constructs may be used depending on the type of article, the key message, and the author's preference or judgement.[ 4 ]

Drafting a suitable title

A stepwise process can be followed to draft the appropriate title. The author should describe the paper in about three sentences, avoiding the results and ensuring that these sentences contain important scientific words/keywords that describe the main contents and subject of the paper.[ 1 , 4 , 6 , 12 ] Then the author should join the sentences to form a single sentence, shorten the length (by removing redundant words or adjectives or phrases), and finally edit the title (thus drafted) to make it more accurate, concise (about 10–15 words), and precise.[ 1 , 3 , 4 , 5 , 9 ] Some journals require that the study design be included in the title, and this may be placed (using a colon) after the primary title.[ 2 , 3 , 4 , 14 ] The title should try to incorporate the Patients, Interventions, Comparisons and Outcome (PICO).[ 3 ] The place of the study may be included in the title (if absolutely necessary), that is, if the patient characteristics (such as study population, socioeconomic conditions, or cultural practices) are expected to vary as per the country (or the place of the study) and have a bearing on the possible outcomes.[ 3 , 6 ] Lengthy titles can be boring and appear unfocused, whereas very short titles may not be representative of the contents of the article; hence, optimum length is required to ensure that the title explains the main theme and content of the manuscript.[ 4 , 5 , 9 ] Abbreviations (except the standard or commonly interpreted ones such as HIV, AIDS, DNA, RNA, CDC, FDA, ECG, and EEG) or acronyms should be avoided in the title, as a reader not familiar with them may skip such an article and nonstandard abbreviations may create problems in indexing the article.[ 3 , 4 , 5 , 6 , 9 , 12 ] Also, too much of technical jargon or chemical formulas in the title may confuse the readers and the article may be skipped by them.[ 4 , 9 ] Numerical values of various parameters (stating study period or sample size) should also be avoided in the titles (unless deemed extremely essential).[ 4 ] It may be worthwhile to take an opinion from a impartial colleague before finalizing the title.[ 4 , 5 , 6 ] Thus, multiple factors (which are, at times, a bit conflicting or contrasting) need to be considered while formulating a title, and hence this should not be done in a hurry.[ 4 , 6 ] Many journals ask the authors to draft a “short title” or “running head” or “running title” for printing in the header or footer of the printed paper.[ 3 , 12 ] This is an abridged version of the main title of up to 40–50 characters, may have standard abbreviations, and helps the reader to navigate through the paper.[ 3 , 12 , 14 ]

Checklist for a good title

Table 1 gives a checklist/useful tips for drafting a good title for a research paper.[ 1 , 2 , 3 , 4 , 5 , 6 , 12 ] Table 2 presents some of the titles used by the author of this article in his earlier research papers, and the appropriateness of the titles has been commented upon. As an individual exercise, the reader may try to improvise upon the titles (further) after reading the corresponding abstract and full paper.

Checklist/useful tips for drafting a good title for a research paper

Some titles used by author of this article in his earlier publications and remark/comment on their appropriateness

The Abstract

The details of the abstract are discussed under the subheadings of importance, types, drafting, and checklist.

Importance of the abstract

The abstract is a summary or synopsis of the full research paper and also needs to have similar characteristics like the title. It needs to be simple, direct, specific, functional, clear, unbiased, honest, concise, precise, self-sufficient, complete, comprehensive, scholarly, balanced, and should not be misleading.[ 1 , 2 , 3 , 7 , 8 , 9 , 10 , 11 , 13 , 17 ] Writing an abstract is to extract and summarize (AB – absolutely, STR – straightforward, ACT – actual data presentation and interpretation).[ 17 ] The title and abstracts are the only sections of the research paper that are often freely available to the readers on the journal websites, search engines, and in many abstracting agencies/databases, whereas the full paper may attract a payment per view or a fee for downloading the pdf copy.[ 1 , 2 , 3 , 7 , 8 , 10 , 11 , 13 , 14 ] The abstract is an independent and stand-alone (that is, well understood without reading the full paper) section of the manuscript and is used by the editor to decide the fate of the article and to choose appropriate reviewers.[ 2 , 7 , 10 , 12 , 13 ] Even the reviewers are initially supplied only with the title and the abstract before they agree to review the full manuscript.[ 7 , 13 ] This is the second most commonly read part of the manuscript, and therefore it should reflect the contents of the main text of the paper accurately and thus act as a “real trailer” of the full article.[ 2 , 7 , 11 ] The readers will go through the full paper only if they find the abstract interesting and relevant to their practice; else they may skip the paper if the abstract is unimpressive.[ 7 , 8 , 9 , 10 , 13 ] The abstract needs to highlight the selling point of the manuscript and succeed in luring the reader to read the complete paper.[ 3 , 7 ] The title and the abstract should be constructed using keywords (key terms/important words) from all the sections of the main text.[ 12 ] Abstracts are also used for submitting research papers to a conference for consideration for presentation (as oral paper or poster).[ 9 , 13 , 17 ] Grammatical and typographic errors reflect poorly on the quality of the abstract, may indicate carelessness/casual attitude on part of the author, and hence should be avoided at all times.[ 9 ]

Types of abstracts

The abstracts can be structured or unstructured. They can also be classified as descriptive or informative abstracts.

Structured and unstructured abstracts

Structured abstracts are followed by most journals, are more informative, and include specific subheadings/subsections under which the abstract needs to be composed.[ 1 , 7 , 8 , 9 , 10 , 11 , 13 , 17 , 18 ] These subheadings usually include context/background, objectives, design, setting, participants, interventions, main outcome measures, results, and conclusions.[ 1 ] Some journals stick to the standard IMRAD format for the structure of the abstracts, and the subheadings would include Introduction/Background, Methods, Results, And (instead of Discussion) the Conclusion/s.[ 1 , 2 , 7 , 8 , 9 , 10 , 11 , 12 , 13 , 17 , 18 ] Structured abstracts are more elaborate, informative, easy to read, recall, and peer-review, and hence are preferred; however, they consume more space and can have same limitations as an unstructured abstract.[ 7 , 9 , 18 ] The structured abstracts are (possibly) better understood by the reviewers and readers. Anyway, the choice of the type of the abstract and the subheadings of a structured abstract depend on the particular journal style and is not left to the author's wish.[ 7 , 10 , 12 ] Separate subheadings may be necessary for reporting meta-analysis, educational research, quality improvement work, review, or case study.[ 1 ] Clinical trial abstracts need to include the essential items mentioned in the CONSORT (Consolidated Standards Of Reporting Trials) guidelines.[ 7 , 9 , 14 , 19 ] Similar guidelines exist for various other types of studies, including observational studies and for studies of diagnostic accuracy.[ 20 , 21 ] A useful resource for the above guidelines is available at www.equator-network.org (Enhancing the QUAlity and Transparency Of health Research). Unstructured (or non-structured) abstracts are free-flowing, do not have predefined subheadings, and are commonly used for papers that (usually) do not describe original research.[ 1 , 7 , 9 , 10 ]

The four-point structured abstract: This has the following elements which need to be properly balanced with regard to the content/matter under each subheading:[ 9 ]

Background and/or Objectives: This states why the work was undertaken and is usually written in just a couple of sentences.[ 3 , 7 , 8 , 9 , 10 , 12 , 13 ] The hypothesis/study question and the major objectives are also stated under this subheading.[ 3 , 7 , 8 , 9 , 10 , 12 , 13 ]

Methods: This subsection is the longest, states what was done, and gives essential details of the study design, setting, participants, blinding, sample size, sampling method, intervention/s, duration and follow-up, research instruments, main outcome measures, parameters evaluated, and how the outcomes were assessed or analyzed.[ 3 , 7 , 8 , 9 , 10 , 12 , 13 , 14 , 17 ]

Results/Observations/Findings: This subheading states what was found, is longer, is difficult to draft, and needs to mention important details including the number of study participants, results of analysis (of primary and secondary objectives), and include actual data (numbers, mean, median, standard deviation, “P” values, 95% confidence intervals, effect sizes, relative risks, odds ratio, etc.).[ 3 , 7 , 8 , 9 , 10 , 12 , 13 , 14 , 17 ]

Conclusions: The take-home message (the “so what” of the paper) and other significant/important findings should be stated here, considering the interpretation of the research question/hypothesis and results put together (without overinterpreting the findings) and may also include the author's views on the implications of the study.[ 3 , 7 , 8 , 9 , 10 , 12 , 13 , 14 , 17 ]

The eight-point structured abstract: This has the following eight subheadings – Objectives, Study Design, Study Setting, Participants/Patients, Methods/Intervention, Outcome Measures, Results, and Conclusions.[ 3 , 9 , 18 ] The instructions to authors given by the particular journal state whether they use the four- or eight-point abstract or variants thereof.[ 3 , 14 ]

Descriptive and Informative abstracts

Descriptive abstracts are short (75–150 words), only portray what the paper contains without providing any more details; the reader has to read the full paper to know about its contents and are rarely used for original research papers.[ 7 , 10 ] These are used for case reports, reviews, opinions, and so on.[ 7 , 10 ] Informative abstracts (which may be structured or unstructured as described above) give a complete detailed summary of the article contents and truly reflect the actual research done.[ 7 , 10 ]

Drafting a suitable abstract

It is important to religiously stick to the instructions to authors (format, word limit, font size/style, and subheadings) provided by the journal for which the abstract and the paper are being written.[ 7 , 8 , 9 , 10 , 13 ] Most journals allow 200–300 words for formulating the abstract and it is wise to restrict oneself to this word limit.[ 1 , 2 , 3 , 7 , 8 , 9 , 10 , 11 , 12 , 13 , 22 ] Though some authors prefer to draft the abstract initially, followed by the main text of the paper, it is recommended to draft the abstract in the end to maintain accuracy and conformity with the main text of the paper (thus maintaining an easy linkage/alignment with title, on one hand, and the introduction section of the main text, on the other hand).[ 2 , 7 , 9 , 10 , 11 ] The authors should check the subheadings (of the structured abstract) permitted by the target journal, use phrases rather than sentences to draft the content of the abstract, and avoid passive voice.[ 1 , 7 , 9 , 12 ] Next, the authors need to get rid of redundant words and edit the abstract (extensively) to the correct word count permitted (every word in the abstract “counts”!).[ 7 , 8 , 9 , 10 , 13 ] It is important to ensure that the key message, focus, and novelty of the paper are not compromised; the rationale of the study and the basis of the conclusions are clear; and that the abstract is consistent with the main text of the paper.[ 1 , 2 , 3 , 7 , 9 , 11 , 12 , 13 , 14 , 17 , 22 ] This is especially important while submitting a revision of the paper (modified after addressing the reviewer's comments), as the changes made in the main (revised) text of the paper need to be reflected in the (revised) abstract as well.[ 2 , 10 , 12 , 14 , 22 ] Abbreviations should be avoided in an abstract, unless they are conventionally accepted or standard; references, tables, or figures should not be cited in the abstract.[ 7 , 9 , 10 , 11 , 13 ] It may be worthwhile not to rush with the abstract and to get an opinion by an impartial colleague on the content of the abstract; and if possible, the full paper (an “informal” peer-review).[ 1 , 7 , 8 , 9 , 11 , 17 ] Appropriate “Keywords” (three to ten words or phrases) should follow the abstract and should be preferably chosen from the Medical Subject Headings (MeSH) list of the U.S. National Library of Medicine ( https://meshb.nlm.nih.gov/search ) and are used for indexing purposes.[ 2 , 3 , 11 , 12 ] These keywords need to be different from the words in the main title (the title words are automatically used for indexing the article) and can be variants of the terms/phrases used in the title, or words from the abstract and the main text.[ 3 , 12 ] The ICMJE (International Committee of Medical Journal Editors; http://www.icmje.org/ ) also recommends publishing the clinical trial registration number at the end of the abstract.[ 7 , 14 ]

Checklist for a good abstract

Table 3 gives a checklist/useful tips for formulating a good abstract for a research paper.[ 1 , 2 , 3 , 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 17 , 22 ]

Checklist/useful tips for formulating a good abstract for a research paper

Concluding Remarks

This review article has given a detailed account of the importance and types of titles and abstracts. It has also attempted to give useful hints for drafting an appropriate title and a complete abstract for a research paper. It is hoped that this review will help the authors in their career in medical writing.

Financial support and sponsorship

Conflicts of interest.

There are no conflicts of interest.

Acknowledgement

The author thanks Dr. Hemant Deshmukh - Dean, Seth G.S. Medical College & KEM Hospital, for granting permission to publish this manuscript.

Book cover

How to Practice Academic Medicine and Publish from Developing Countries? pp 179–184 Cite as

How to Write an Abstract?

  • Samiran Nundy 4 ,
  • Atul Kakar 5 &
  • Zulfiqar A. Bhutta 6  
  • Open Access
  • First Online: 24 October 2021

53k Accesses

5 Altmetric

An abstract is a crisp, short, powerful, and self-contained summary of a research manuscript used to help the reader swiftly determine the paper’s purpose. Although the abstract is the first paragraph of the manuscript it should be written last when all the other sections have been addressed.

Research is formalized curiosity. It is poking and prying with a purpose. — Zora Neale Hurston, American Author, Anthropologist and Filmmaker (1891–1960)

Download chapter PDF

1 What is an Abstract?

An abstract is usually a standalone document that informs the reader about the details of the manuscript to follow. It is like a trailer to a movie, if the trailer is good, it stimulates the audience to watch the movie. The abstract should be written from scratch and not ‘cut –and-pasted’ [ 1 ].

2 What is the History of the Abstract?

An abstract, in the form of a single paragraph, was first published in the Canadian Medical Association Journal in 1960 with the idea that the readers may not have enough time to go through the whole paper, and the first abstract with a defined structure was published in 1991 [ 2 ]. The idea sold and now most original articles and reviews are required to have a structured abstract. The abstract attracts the reader to read the full manuscript [ 3 ].

3 What are the Qualities of a Good Abstract?

The quality of information in an abstract can be summarized by four ‘C’s. It should be:

C: Condensed

C: Critical

4 What are the Types of Abstract?

Before writing the abstract, you need to check with the journal website about which type of abstract it requires, with its length and style in the ‘Instructions to Authors’ section.

The abstract types can be divided into:

Descriptive: Usually written for psychology, social science, and humanities papers. It is about 50–100 words long. No conclusions can be drawn from this abstract as it describes the major points in the paper.

Informative: The majority of abstracts for science-related manuscripts are informative and are surrogates for the research done. They are single paragraphs that provide the reader an overview of the research paper and are about 100–150 words in length. Conclusions can be drawn from the abstracts and in the recommendations written in the last line.

Critical: This type of abstract is lengthy and about 400–500 words. In this, the authors’ own research is discussed for reliability, judgement, and validation. A comparison is also made with similar studies done earlier.

Highlighting: This is rarely used in scientific writing. The style of the abstract is to attract more readers. It is not a balanced or complete overview of the article with which it is published.

Structured: A structured abstract contains information under subheadings like background, aims, material and methods, results, conclusion, and recommendations (Fig. 15.1 ). Most leading journals now carry these.

figure 1

Example of a structured abstract (with permission editor CMRP)

5 What is the Purpose of an Abstract?

An abstract is written to educate the reader about the study that follows and provide an overview of the science behind it. If written well it also attracts more readers to the article. It also helps the article getting indexed. The fate of a paper both before and after publication often depends upon its abstract. Most readers decide if a paper is worth reading on the basis of the abstract. Additionally, the selection of papers in systematic reviews is often dependent upon the abstract.

6 What are the Steps of Writing an Abstract?

An abstract should be written last after all the other sections of an article have been addressed. A poor abstract may turn off the reader and they may cause indexing errors as well. The abstract should state the purpose of the study, the methodology used, and summarize the results and important conclusions. It is usually written in the IMRAD format and is called a structured abstract [ 4 , 5 ].

I: The introduction in the opening line should state the problem you are addressing.

M: Methodology—what method was chosen to finish the experiment?

R: Results—state the important findings of your study.

D: Discussion—discuss why your study is important.

Mention the following information:

Important results with the statistical information ( p values, confidence intervals, standard/mean deviation).

Arrange all information in a chronological order.

Do not repeat any information.

The last line should state the recommendations from your study.

The abstract should be written in the past tense.

7 What are the Things to Be Avoided While Writing an Abstract?

Cut and paste information from the main text

Hold back important information

Use abbreviations

Tables or Figures

Generalized statements

Arguments about the study

figure a

8 What are Key Words?

These are important words that are repeated throughout the manuscript and which help in the indexing of a paper. Depending upon the journal 3–10 key words may be required which are indexed with the help of MESH (Medical Subject Heading).

9 How is an Abstract Written for a Conference Different from a Journal Paper?

The basic concept for writing abstracts is the same. However, in a conference abstract occasionally a table or figure is allowed. A word limit is important in both of them. Many of the abstracts which are presented in conferences are never published in fact one study found that only 27% of the abstracts presented in conferences were published in the next five years [ 6 ].

Table 15.1 gives a template for writing an abstract.

10 What are the Important Recommendations of the International Committees of Medical Journal of Editors?

The recommendations are [ 7 ]:

An abstract is required for original articles, metanalysis, and systematic reviews.

A structured abstract is preferred.

The abstract should mention the purpose of the scientific study, how the procedure was carried out, the analysis used, and principal conclusion.

Clinical trials should be reported according to the CONSORT guidelines.

The trials should also mention the funding and the trial number.

The abstract should be accurate as many readers have access only to the abstract.

11 Conclusions

An Abstract should be written last after all the other sections of the manuscript have been completed and with due care and attention to the details.

It should be structured and written in the IMRAD format.

For many readers, the abstract attracts them to go through the complete content of the article.

The abstract is usually followed by key words that help to index the paper.

Andrade C. How to write a good abstract for a scientific paper or conference presentation? Indian J Psychiatry. 2011;53:172–5.

Article   Google Scholar  

Squires BP. Structured abstracts of original research and review articles. CMAJ. 1990;143:619–22.

CAS   PubMed   PubMed Central   Google Scholar  

Pierson DJ. How to write an abstract that will be accepted for presentation at a national meeting. Respir Care. 2004 Oct;49:1206–12.

PubMed   Google Scholar  

Tenenbein M. The abstract and the academic clinician. Pediatr Emerg Care. 1995;11:40–2.

Article   CAS   Google Scholar  

Bahadoran Z, Mirmiran P, Kashfi K, Ghasemi A. The principles of biomedical scientific writing: abstract and keywords. Int J Endocrinol Metab. 2020;18:e100159.

PubMed   PubMed Central   Google Scholar  

Grover S, Dalton N. Abstract to publication rate: do all the papers presented in conferences see the light of being a full publication? Indian J Psychiatry. 2020;62:73–9.

Preparing a manuscript for submission to a medical journal. Available on http://www.icmje.org/recommendations/browse/manuscript-preparation/preparing-for-submission.html . Accessed 10 May 2020.

Download references

Author information

Authors and affiliations.

Department of Surgical Gastroenterology and Liver Transplantation, Sir Ganga Ram Hospital, New Delhi, India

Samiran Nundy

Department of Internal Medicine, Sir Ganga Ram Hospital, New Delhi, India

Institute for Global Health and Development, The Aga Khan University, South Central Asia, East Africa and United Kingdom, Karachi, Pakistan

Zulfiqar A. Bhutta

You can also search for this author in PubMed   Google Scholar

Rights and permissions

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License ( http://creativecommons.org/licenses/by/4.0/ ), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

Reprints and permissions

Copyright information

© 2022 The Author(s)

About this chapter

Cite this chapter.

Nundy, S., Kakar, A., Bhutta, Z.A. (2022). How to Write an Abstract?. In: How to Practice Academic Medicine and Publish from Developing Countries?. Springer, Singapore. https://doi.org/10.1007/978-981-16-5248-6_15

Download citation

DOI : https://doi.org/10.1007/978-981-16-5248-6_15

Published : 24 October 2021

Publisher Name : Springer, Singapore

Print ISBN : 978-981-16-5247-9

Online ISBN : 978-981-16-5248-6

eBook Packages : Medicine Medicine (R0)

Share this chapter

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

  • Publish with us

Policies and ethics

  • Find a journal
  • Track your research

Logo for M Libraries Publishing

Want to create or adapt books like this? Learn more about how Pressbooks supports open publishing practices.

13.1 Formatting a Research Paper

Learning objectives.

  • Identify the major components of a research paper written using American Psychological Association (APA) style.
  • Apply general APA style and formatting conventions in a research paper.

In this chapter, you will learn how to use APA style , the documentation and formatting style followed by the American Psychological Association, as well as MLA style , from the Modern Language Association. There are a few major formatting styles used in academic texts, including AMA, Chicago, and Turabian:

  • AMA (American Medical Association) for medicine, health, and biological sciences
  • APA (American Psychological Association) for education, psychology, and the social sciences
  • Chicago—a common style used in everyday publications like magazines, newspapers, and books
  • MLA (Modern Language Association) for English, literature, arts, and humanities
  • Turabian—another common style designed for its universal application across all subjects and disciplines

While all the formatting and citation styles have their own use and applications, in this chapter we focus our attention on the two styles you are most likely to use in your academic studies: APA and MLA.

If you find that the rules of proper source documentation are difficult to keep straight, you are not alone. Writing a good research paper is, in and of itself, a major intellectual challenge. Having to follow detailed citation and formatting guidelines as well may seem like just one more task to add to an already-too-long list of requirements.

Following these guidelines, however, serves several important purposes. First, it signals to your readers that your paper should be taken seriously as a student’s contribution to a given academic or professional field; it is the literary equivalent of wearing a tailored suit to a job interview. Second, it shows that you respect other people’s work enough to give them proper credit for it. Finally, it helps your reader find additional materials if he or she wishes to learn more about your topic.

Furthermore, producing a letter-perfect APA-style paper need not be burdensome. Yes, it requires careful attention to detail. However, you can simplify the process if you keep these broad guidelines in mind:

  • Work ahead whenever you can. Chapter 11 “Writing from Research: What Will I Learn?” includes tips for keeping track of your sources early in the research process, which will save time later on.
  • Get it right the first time. Apply APA guidelines as you write, so you will not have much to correct during the editing stage. Again, putting in a little extra time early on can save time later.
  • Use the resources available to you. In addition to the guidelines provided in this chapter, you may wish to consult the APA website at http://www.apa.org or the Purdue University Online Writing lab at http://owl.english.purdue.edu , which regularly updates its online style guidelines.

General Formatting Guidelines

This chapter provides detailed guidelines for using the citation and formatting conventions developed by the American Psychological Association, or APA. Writers in disciplines as diverse as astrophysics, biology, psychology, and education follow APA style. The major components of a paper written in APA style are listed in the following box.

These are the major components of an APA-style paper:

Body, which includes the following:

  • Headings and, if necessary, subheadings to organize the content
  • In-text citations of research sources
  • References page

All these components must be saved in one document, not as separate documents.

The title page of your paper includes the following information:

  • Title of the paper
  • Author’s name
  • Name of the institution with which the author is affiliated
  • Header at the top of the page with the paper title (in capital letters) and the page number (If the title is lengthy, you may use a shortened form of it in the header.)

List the first three elements in the order given in the previous list, centered about one third of the way down from the top of the page. Use the headers and footers tool of your word-processing program to add the header, with the title text at the left and the page number in the upper-right corner. Your title page should look like the following example.

Beyond the Hype: Evaluating Low-Carb Diets cover page

The next page of your paper provides an abstract , or brief summary of your findings. An abstract does not need to be provided in every paper, but an abstract should be used in papers that include a hypothesis. A good abstract is concise—about one hundred fifty to two hundred fifty words—and is written in an objective, impersonal style. Your writing voice will not be as apparent here as in the body of your paper. When writing the abstract, take a just-the-facts approach, and summarize your research question and your findings in a few sentences.

In Chapter 12 “Writing a Research Paper” , you read a paper written by a student named Jorge, who researched the effectiveness of low-carbohydrate diets. Read Jorge’s abstract. Note how it sums up the major ideas in his paper without going into excessive detail.

Beyond the Hype: Abstract

Write an abstract summarizing your paper. Briefly introduce the topic, state your findings, and sum up what conclusions you can draw from your research. Use the word count feature of your word-processing program to make sure your abstract does not exceed one hundred fifty words.

Depending on your field of study, you may sometimes write research papers that present extensive primary research, such as your own experiment or survey. In your abstract, summarize your research question and your findings, and briefly indicate how your study relates to prior research in the field.

Margins, Pagination, and Headings

APA style requirements also address specific formatting concerns, such as margins, pagination, and heading styles, within the body of the paper. Review the following APA guidelines.

Use these general guidelines to format the paper:

  • Set the top, bottom, and side margins of your paper at 1 inch.
  • Use double-spaced text throughout your paper.
  • Use a standard font, such as Times New Roman or Arial, in a legible size (10- to 12-point).
  • Use continuous pagination throughout the paper, including the title page and the references section. Page numbers appear flush right within your header.
  • Section headings and subsection headings within the body of your paper use different types of formatting depending on the level of information you are presenting. Additional details from Jorge’s paper are provided.

Cover Page

Begin formatting the final draft of your paper according to APA guidelines. You may work with an existing document or set up a new document if you choose. Include the following:

  • Your title page
  • The abstract you created in Note 13.8 “Exercise 1”
  • Correct headers and page numbers for your title page and abstract

APA style uses section headings to organize information, making it easy for the reader to follow the writer’s train of thought and to know immediately what major topics are covered. Depending on the length and complexity of the paper, its major sections may also be divided into subsections, sub-subsections, and so on. These smaller sections, in turn, use different heading styles to indicate different levels of information. In essence, you are using headings to create a hierarchy of information.

The following heading styles used in APA formatting are listed in order of greatest to least importance:

  • Section headings use centered, boldface type. Headings use title case, with important words in the heading capitalized.
  • Subsection headings use left-aligned, boldface type. Headings use title case.
  • The third level uses left-aligned, indented, boldface type. Headings use a capital letter only for the first word, and they end in a period.
  • The fourth level follows the same style used for the previous level, but the headings are boldfaced and italicized.
  • The fifth level follows the same style used for the previous level, but the headings are italicized and not boldfaced.

Visually, the hierarchy of information is organized as indicated in Table 13.1 “Section Headings” .

Table 13.1 Section Headings

A college research paper may not use all the heading levels shown in Table 13.1 “Section Headings” , but you are likely to encounter them in academic journal articles that use APA style. For a brief paper, you may find that level 1 headings suffice. Longer or more complex papers may need level 2 headings or other lower-level headings to organize information clearly. Use your outline to craft your major section headings and determine whether any subtopics are substantial enough to require additional levels of headings.

Working with the document you developed in Note 13.11 “Exercise 2” , begin setting up the heading structure of the final draft of your research paper according to APA guidelines. Include your title and at least two to three major section headings, and follow the formatting guidelines provided above. If your major sections should be broken into subsections, add those headings as well. Use your outline to help you.

Because Jorge used only level 1 headings, his Exercise 3 would look like the following:

Citation Guidelines

In-text citations.

Throughout the body of your paper, include a citation whenever you quote or paraphrase material from your research sources. As you learned in Chapter 11 “Writing from Research: What Will I Learn?” , the purpose of citations is twofold: to give credit to others for their ideas and to allow your reader to follow up and learn more about the topic if desired. Your in-text citations provide basic information about your source; each source you cite will have a longer entry in the references section that provides more detailed information.

In-text citations must provide the name of the author or authors and the year the source was published. (When a given source does not list an individual author, you may provide the source title or the name of the organization that published the material instead.) When directly quoting a source, it is also required that you include the page number where the quote appears in your citation.

This information may be included within the sentence or in a parenthetical reference at the end of the sentence, as in these examples.

Epstein (2010) points out that “junk food cannot be considered addictive in the same way that we think of psychoactive drugs as addictive” (p. 137).

Here, the writer names the source author when introducing the quote and provides the publication date in parentheses after the author’s name. The page number appears in parentheses after the closing quotation marks and before the period that ends the sentence.

Addiction researchers caution that “junk food cannot be considered addictive in the same way that we think of psychoactive drugs as addictive” (Epstein, 2010, p. 137).

Here, the writer provides a parenthetical citation at the end of the sentence that includes the author’s name, the year of publication, and the page number separated by commas. Again, the parenthetical citation is placed after the closing quotation marks and before the period at the end of the sentence.

As noted in the book Junk Food, Junk Science (Epstein, 2010, p. 137), “junk food cannot be considered addictive in the same way that we think of psychoactive drugs as addictive.”

Here, the writer chose to mention the source title in the sentence (an optional piece of information to include) and followed the title with a parenthetical citation. Note that the parenthetical citation is placed before the comma that signals the end of the introductory phrase.

David Epstein’s book Junk Food, Junk Science (2010) pointed out that “junk food cannot be considered addictive in the same way that we think of psychoactive drugs as addictive” (p. 137).

Another variation is to introduce the author and the source title in your sentence and include the publication date and page number in parentheses within the sentence or at the end of the sentence. As long as you have included the essential information, you can choose the option that works best for that particular sentence and source.

Citing a book with a single author is usually a straightforward task. Of course, your research may require that you cite many other types of sources, such as books or articles with more than one author or sources with no individual author listed. You may also need to cite sources available in both print and online and nonprint sources, such as websites and personal interviews. Chapter 13 “APA and MLA Documentation and Formatting” , Section 13.2 “Citing and Referencing Techniques” and Section 13.3 “Creating a References Section” provide extensive guidelines for citing a variety of source types.

Writing at Work

APA is just one of several different styles with its own guidelines for documentation, formatting, and language usage. Depending on your field of interest, you may be exposed to additional styles, such as the following:

  • MLA style. Determined by the Modern Languages Association and used for papers in literature, languages, and other disciplines in the humanities.
  • Chicago style. Outlined in the Chicago Manual of Style and sometimes used for papers in the humanities and the sciences; many professional organizations use this style for publications as well.
  • Associated Press (AP) style. Used by professional journalists.

References List

The brief citations included in the body of your paper correspond to the more detailed citations provided at the end of the paper in the references section. In-text citations provide basic information—the author’s name, the publication date, and the page number if necessary—while the references section provides more extensive bibliographical information. Again, this information allows your reader to follow up on the sources you cited and do additional reading about the topic if desired.

The specific format of entries in the list of references varies slightly for different source types, but the entries generally include the following information:

  • The name(s) of the author(s) or institution that wrote the source
  • The year of publication and, where applicable, the exact date of publication
  • The full title of the source
  • For books, the city of publication
  • For articles or essays, the name of the periodical or book in which the article or essay appears
  • For magazine and journal articles, the volume number, issue number, and pages where the article appears
  • For sources on the web, the URL where the source is located

The references page is double spaced and lists entries in alphabetical order by the author’s last name. If an entry continues for more than one line, the second line and each subsequent line are indented five spaces. Review the following example. ( Chapter 13 “APA and MLA Documentation and Formatting” , Section 13.3 “Creating a References Section” provides extensive guidelines for formatting reference entries for different types of sources.)

References Section

In APA style, book and article titles are formatted in sentence case, not title case. Sentence case means that only the first word is capitalized, along with any proper nouns.

Key Takeaways

  • Following proper citation and formatting guidelines helps writers ensure that their work will be taken seriously, give proper credit to other authors for their work, and provide valuable information to readers.
  • Working ahead and taking care to cite sources correctly the first time are ways writers can save time during the editing stage of writing a research paper.
  • APA papers usually include an abstract that concisely summarizes the paper.
  • APA papers use a specific headings structure to provide a clear hierarchy of information.
  • In APA papers, in-text citations usually include the name(s) of the author(s) and the year of publication.
  • In-text citations correspond to entries in the references section, which provide detailed bibliographical information about a source.

Writing for Success Copyright © 2015 by University of Minnesota is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License , except where otherwise noted.

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • View all journals
  • My Account Login
  • Explore content
  • About the journal
  • Publish with us
  • Sign up for alerts
  • Open access
  • Published: 22 February 2024

Blood–brain barrier disruption and sustained systemic inflammation in individuals with long COVID-associated cognitive impairment

  • Chris Greene   ORCID: orcid.org/0000-0003-4192-9433 1 ,
  • Ruairi Connolly 2 ,
  • Declan Brennan 2 ,
  • Aoife Laffan 2 ,
  • Eoin O’Keeffe 1 ,
  • Lilia Zaporojan 2 ,
  • Jeffrey O’Callaghan   ORCID: orcid.org/0000-0001-8818-4331 1 ,
  • Bennett Thomson 1 ,
  • Emma Connolly 3 ,
  • Ruth Argue 4 ,
  • Ignacio Martin-Loeches 5 ,
  • Aideen Long 6 ,
  • Cliona Ni Cheallaigh 6 , 7 ,
  • Niall Conlon 7 , 8 ,
  • Colin P. Doherty   ORCID: orcid.org/0000-0002-8869-8567 2 , 9 , 10 &
  • Matthew Campbell   ORCID: orcid.org/0000-0003-3325-240X 1 , 10  

Nature Neuroscience ( 2024 ) Cite this article

40k Accesses

4475 Altmetric

Metrics details

  • Diseases of the nervous system
  • Neuro–vascular interactions
  • Neuroimmunology

Vascular disruption has been implicated in coronavirus disease 2019 (COVID-19) pathogenesis and may predispose to the neurological sequelae associated with long COVID, yet it is unclear how blood–brain barrier (BBB) function is affected in these conditions. Here we show that BBB disruption is evident during acute infection and in patients with long COVID with cognitive impairment, commonly referred to as brain fog. Using dynamic contrast-enhanced magnetic resonance imaging, we show BBB disruption in patients with long COVID-associated brain fog. Transcriptomic analysis of peripheral blood mononuclear cells revealed dysregulation of the coagulation system and a dampened adaptive immune response in individuals with brain fog. Accordingly, peripheral blood mononuclear cells showed increased adhesion to human brain endothelial cells in vitro, while exposure of brain endothelial cells to serum from patients with long COVID induced expression of inflammatory markers. Together, our data suggest that sustained systemic inflammation and persistent localized BBB dysfunction is a key feature of long COVID-associated brain fog.

Similar content being viewed by others

how long is an abstract research paper

Blood–brain barrier link to human cognitive impairment and Alzheimer’s disease

Giuseppe Barisano, Axel Montagne, … Berislav V. Zlokovic

how long is an abstract research paper

Cerebral hypoperfusion in post-COVID-19 cognitively impaired subjects revealed by arterial spin labeling MRI

Miloš Ajčević, Katerina Iscra, … Paolo Manganotti

how long is an abstract research paper

Severe Neuro-COVID is associated with peripheral immune signatures, autoimmunity and neurodegeneration: a prospective cross-sectional study

Manina M. Etter, Tomás A. Martins, … Gregor Hutter

Coronavirus disease 2019 (COVID-19) is a clinical syndrome caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). COVID-19 primarily affects the respiratory tract and can progress to respiratory compromise, severe acute respiratory distress syndrome (ARDS) and death 1 , 2 . ARDS due to COVID-19 has been associated with encephalopathy, agitation, confusion and corticospinal tract dysfunction. Such symptoms, however, including anosmia (although not to the extent seen in the first wave of COVID-19), may be expected in anyone recovering from a severe viral illness because of cytokine release, critical illness encephalopathy or medication 3 . Neurological sequelae of COVID-19, colloquially known as ‘brain fog’, are increasingly being reported and include headache, fatigue, malaise and altered levels of consciousness. For example, clinical observations of neurological complications in 236,379 patients in the 6 months after a COVID-19 diagnosis found that 33.62% of patients had demonstrated clinically important neurological or psychiatric dysfunction 4 . Neurological problems have been reported in other respiratory viral infections including influenza, coronavirus and metapneumovirus, with febrile or afebrile seizures, status epilepticus, encephalopathies and encephalitis being the most frequently reported 5 . However, there is still little understanding of the pathogenesis and long-term outcome of neurological problems after SARS-CoV-2 infection. SARS-CoV-2 gains cellular entry via its receptors angiotensin-converting enzyme 2 and transmembrane protease serine 2, but it may enter via other receptors, including neuropilin and vimentin, all of which are enriched in cells of the neurovascular unit 6 , 7 , 8 , 9 , 10 . There are, however, conflicting reports regarding the neuroinvasiveness of SARS-CoV-2 and indeed the cellular expression of the receptors 11 , 12 , 13 , 14 , 15 , suggesting that other mechanisms are responsible for the neurological problems reported. A recent study suggested persistence of viral RNA in multiple anatomic sites, including the brain, for up to 230 days after symptom onset, although these data were from postmortem donor tissues, which represent the sickest of individuals 16 .

Several lines of research suggested that breakdown to the integrity of the blood–brain barrier (BBB) and subsequent brain penetration of serum components and cytokines is responsible for the neurological manifestations after SARS-CoV-2 infection 17 , 18 . The BBB is formed by endothelial cells lining cerebral blood vessels and supported by surrounding cells including astrocytes, pericytes, microglia, neurons and the acellular basement membrane 19 . The barrier is characterized by an enrichment of interendothelial tight junction proteins, several luminal and abluminal transporters, and luminal efflux transporters, which together maintain separation of the blood and brain and tightly regulate molecular trafficking between the blood and brain and vice versa 20 .

There is clear evidence of microvascular injury in the brains of deceased patients with COVID-19, including fibrinogen leakage and thinning of the endothelial cell basal laminae in the olfactory bulb 14 , 21 . A more comprehensive evaluation of the same cohort using spatial transcriptomics revealed more detailed vascular and immunological features of microvessels in the brain, including serum protein extravasation, platelet accumulation and coagulation system activation 22 . Numerous studies also examined BBB-related changes and responses to SARS-CoV-2 infection or spike (S) protein treatment in postmortem tissue and animal models 14 , 15 , 21 , 23 , 24 , 25 , 26 , 27 , 28 . However, the cerebrovascular pathology in patients and the underlying mechanisms of pathology are still unclear, especially in individuals with long COVID.

The lack of a specific neurological signature of the disease is interesting because other zoonotic betacoronaviruses often produce robust and predictable neurological injury 29 . In humans, data from SARS and Middle East respiratory syndrome also showed that neurological injury in humans is rare, strongly suggesting that, normally, the BBB provides robust neuroprotection from viral CNS invasion in most patients 30 . The clinical manifestation of SARS-CoV2-induced BBB alterations in patients has not yet been reported.

In this study, we hypothesized that the neurological response to COVID may be due to BBB breakdown and subsequent extravasation of serum components. We show that BBB disruption is evident in patients with acute COVID with brain fog and a cohort of patients with persistent long COVID-associated brain fog. We suggest that measurement of BBB integrity may be a clinically useful biomarker of the neurological sequelae associated with COVID-19 in some patients. Added to this, targeted regulation of BBB integrity may also represent a new method of clinically managing patients with long COVID.

Acute COVID-induced brain fog is associated with BBB dysfunction

We collected serum and plasma samples from 76 inpatients with acute COVID-19 recruited as part of the St James’s Hospital, Tallaght University Hospital, Trinity College Dublin Allied Researchers Bioresource collection during the initial wave of COVID-19 in March and April 2020 (Extended Data Fig. 1a ) 31 . Twenty-five unaffected control samples were collected before the COVID-19 pandemic. The mean age of the control and COVID samples was 44 and 44.7, respectively. The most frequent presenting symptoms included dyspnea (47), loss of smell and taste (46), cough (45), fatigue (40) and fever (36). Serum and plasma samples were screened with multiplex Luminex and ProcartaPlex panels for inflammatory, coagulation and BBB dysfunction markers. In total, we profiled 50 analytes in serum and plasma. The severity of COVID-19 was determined according to the World Health Organization (WHO) Severity Guidelines with 25 unaffected, 43 mild, 10 moderate and 23 severe. Of the 50 markers investigated, 4, 11 and 25 serum and plasma analytes were significantly different from controls in mild, moderate and severe groups, respectively after false discovery rate (FDR) correction and included several well-defined pro-inflammatory cytokines, including interferon-γ (IFNγ), interleukin-6 (IL-6), interleukin-1β (IL-1β), interleukin-1RA (IL-1RA), interleukin-8 (IL-8) and 10 kDa interferon gamma-induced protein (IP-10); growth factors, including granulocyte colony-stimulating factor (G-CSF) and granulocyte-macrophage colony-stimulating factor (GM-CSF); and markers of thrombosis and endothelial cell activation including plasminogen activator inhibitor-1 (PAI-1), protein C, protein S, Von Willebrand factor (vWF), factor IX, intercellular adhesion molecule 1 (ICAM-1) and vascular cell adhesion protein 1 (VCAM-1) (Fig. 1a–e and Extended Data Fig. 1b–d,h ). Most markers were increased in moderate and severe cases except for coagulation markers, which were increased in all COVID groups (Fig. 1a and Extended Data Fig. 1g ). Next, we determined if segregating according to brain fog status could reveal changes in the inflammatory profile of patients. Patients with brain fog had a higher mean age (53.7 versus 42.7) and were more likely to be hospitalized and require oxygen therapy; therefore, age, sex, comorbidities and severity of infection were included in the statistical model to identify differences between groups (Supplementary Table 1 ). Stratification of patients according to the presence or absence of brain fog revealed a general increase in most markers in the brain fog cohort (Fig. 1f ), with significantly increased serum levels of protein S100β (Fig. 1g ), a marker indirectly associated with BBB dysfunction. There were also increased levels of basic fibroblast growth factor (bFGF), interleukin-13 (IL-13) and monocyte chemoattractant protein-1 (MCP-1) in patients with brain fog (Fig. 1h–j ). Correlation analysis revealed a significant correlation between WHO Severity of COVID-19 and age, duration of hospitalization and sum of comorbidities (Extended Data Fig. 2a–c ). Therefore, partial correlations were performed with age, sex and all comorbidities as covariates, which revealed positive associations between serum concentrations of tumor necrosis factor (TNF), interleukin-6 (IL-6), IL-1β and IP-10 with COVID severity and an inverse association with plasma protein S (Extended Data Fig. 2d–h ). We also found a significant association between serum S100β and age (Extended Data Fig. 2i ). Of the 76 patients, 36 had a second blood sample drawn because of deterioration of clinical symptoms; thus, the serum concentrations of all analytes were assessed between time 1 and 2 (T1 and T2) to monitor disease progression. There was a significant decrease in serum concentrations of coagulation factors including PAI-1 and the cell adhesion molecules VCAM-1 and ICAM-1, while there was an increase in IL-13 and IL-8 between T1 and T2 (Extended Data Fig. 3a,b ).

figure 1

a , Analyte abundance plots showing serum concentrations of blood biomarkers in unaffected patients and patients with mild, moderate and severe SARS-CoV-2. Each cytokine was normalized to the respective mean cytokine level in unaffected individuals. BDNF, brain-derived neurotrophic factor; CCL5, C-C motif chemokine 5; MIP-1α, macrophage inflammatory protein-1 alpha; PAI1, plasminogen activator inhibitor-1; PDGF-BB, platelet-derived growth factor-BB; VEGF, vascular endothelial growth factor. b – e , Levels of IL-6 ( P  = 0.009 moderate versus control, P  < 0.0001 severe versus control) ( b ), IL-8 ( P  = 0.027 moderate versus control, P  = 0.003 severe versus control) ( c ), IFNγ ( P  = 0.02 moderate versus control, P  = 0.015 severe versus control) ( d ) and IL-1RA ( P  = 0.0007 moderate versus control, P  < 0.0001 severe versus control) ( e ) according to COVID severity. f , Analyte abundance plots showing serum concentrations of blood biomarkers in cases with brain fog versus cases without. Each cytokine was normalized to the respective mean cytokine level in individuals without brain fog. g – j , Levels of serum S100β ( P  = 0.0002) ( g ), bFGF ( P  = 0.027) ( h ), IL-13 ( P  = 0.005) ( i ) and MCP-1 ( P  = 0.028) ( j ) according to brain fog status. Data were analyzed using analysis of covariance (ANCOVA) adjusting for age, sex, COVID severity and comorbidities. The violin plots show the median (solid line) and interquartile (dashed lines) values; each data point represents one patient.

Source data

Bbb dysfunction is associated with long covid-induced cognitive impairment.

Given the significantly increased serum concentrations of S100β, our data indicated that active and acute SARS-CoV-2 infection is associated with potential BBB dysfunction in individuals with neurological impairment. However, to directly visualize BBB function, we recruited ten recovered participants, 11 with long COVID and 11 with long COVID with brain fog who were diagnosed with COVID-19 during the first outbreak of disease in Ireland in April 2020 (Fig. 2a and Supplementary Table 2 ). All participants were recruited from St James Hospital Dublin and were PCR-confirmed cases of COVID-19. None of the patients in this cohort had received a vaccine and all had an initially mild course of disease that did not require hospitalization or antiviral treatment. We used a quick smell identification test (Q-SIT)-based method to determine objective anosmia status in participants and determined a strong correlation of reported anosmia status and Q-SIT score, providing an excellent readout of the utility of objective anosmia measurement in prolonged anosmia after COVID-19. Participants were grouped according to the presence or absence of self-reported cognitive issues termed ‘brain fog’ (brain fog (−) or brain fog (+)). Participants were considered as recovered when they reported no recurring symptoms following recovery from active SARS-CoV-2 infection. We hypothesized that COVID-19-associated cognitive impairment may be a strong predictor of BBB disruption in patients with COVID-19. There were no differences in age between each group (Fig. 2b ). Imaging took place at a median time of 46, 175 and 211 days after PCR-confirmed SARS-CoV-2 infection for the recovered, long COVID and brain fog cohorts, respectively (Fig. 2c and Supplementary Table 2 ). Sixteen (50%) participants reported anosmia, which was confirmed using Q-SIT testing (average score 1 out of 3; 159 ± 88 days duration) at the time of scanning. Six participants with brain fog showed mild-to-moderate cognitive impairment on the Montreal Cognitive Assessment (MOCA) test (score 18–25) along with deficits in recall, executive functioning and word finding (Supplementary Table 2 ).

figure 2

a , Patient cohort for dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI). b , Age distribution across cohorts ( n  = 10 recovered, n  = 11 without brain fog (−), n  = 11 with brain fog (+)). c , Time from COVID + PCR test to scan across cohorts ( n  = 10 recovered, n  = 11 without brain fog (−), n  = 11 with brain fog (+)). Data were analyzed using a two-sided Kruskal–Wallis test with Dunn’s correction for multiple comparisons ( P  = 0.0157 without brain fog (−) versus recovered; P  = 0.0004 with brain fog (+) versus recovered). d , Averaged BBB permeability maps in cases with and without brain fog. e , Quantification of the percentage of brain volume with leaky blood vessels in the cohort with brain fog compared to recovered cases ( P  = 0.0057) and cases without brain fog ( P  = 0.0004). Data were analyzed using a one-way analysis of variance with Tukey’s correction. f , Frequency distribution of the percentage of BBB-disrupted voxels in cases with and without brain fog. g , Representative BBB permeability maps at the level of the TLs, FLs and OLs showing enhanced BBB permeability in cases with brain fog. h – k , Quantification of regional BBB permeability in the right TL ( P  = 0.0095) ( h ), left TL ( P  = 0.0202) ( i ), right frontal cortex ( P  = 0.0123) ( j ) and left frontal cortex ( P  = 0.0047) ( k ). Data were analyzed using a two-sided Mann–Whitney U -test. The box plots display the minimum and maximum values (whiskers), median (solid line) and interquartile range (IQR) (upper and lower box). The violin plots show the median (solid line) and IQR (dashed lines); each data point represents one patient. Schematics in a were created with BioRender.com .

While standard diagnostic MRI scans showed no clinically relevant pathological findings in any participant, DCE-MRI imaging revealed significantly increased whole-brain leakage in patients with long COVID with brain fog (Fig. 2d–f ), with an increased percentage of brain volume with leaky blood vessels in the cohort with brain fog compared to the cohort without brain fog. Stratifying the cohort into recovered, long COVID without brain fog and long COVID with brain fog revealed significantly increased BBB permeability in the cohort with brain fog compared to recovered patients and patients with long COVID without brain fog. Region of interest analysis identified significantly increased leakage in the right and left temporal lobes (TLs) and right and left frontal cortex (Fig. 2g–k ). Stratifying the groups according to recovered, long COVID or brain fog revealed significantly increased BBB permeability in the group with brain fog only in the right and left TL and right and left frontal cortex (Extended Data Fig. 4 ). There was no difference in age between those with or without brain fog, with age not being associated with BBB dysfunction (Extended Data Fig. 4b,c ). There was no association between BBB permeability and anosmia status, duration of anosmia, Q-SIT or MOCA scores (Extended Data Fig. 5a–c and Supplementary Table 3 ); however, regional BBB permeability in the right and left TLs was correlated with the duration of anosmia (Extended Data Fig. 5a,b ).

Long COVID-associated brain fog induces structural changes in the brain

To explore if there were structural brain changes accompanying increased BBB permeability in our cohorts, we conducted volume and thickness measurements on recovered, long COVID and 60 age-matched healthy controls from the publicly available IXI dataset (Supplementary Table 2 ) and examined global brain volume (GBV), cerebrospinal fluid (CSF) volume and the right and left volumes of the cerebral and cerebellar white matter (WM) and gray matter (GM), and the brainstem, hippocampus and amygdala. Comparing all individuals with previous COVID infection to unaffected controls revealed volumetric deficits predominantly in the FLs and TLs and increases in the lateral ventricles and occipital lobes (OLs) (Fig. 3a ), while group-wise comparisons of macrostructures revealed decreased GBV in patients with brain fog along with significantly reduced cerebral WM volume in both hemispheres in the recovered and brain fog cohorts along with reduced cerebellar WM volume in the recovered, long COVID and brain fog cohorts (Fig. 3b–d and Supplementary Table 3 ). There was a significantly increased CSF volume in the cohort with brain fog only (Fig. 3e and Supplementary Table 3 ). Cortical thinning was also evident predominantly in the TLs and frontal lobes (FLs) when analyzing all patients with previous SARS-CoV-2 infection compared to unaffected controls (Fig. 3f ). When comparing groups, there was reduced thickness in the frontal pole in the recovered, long COVID and brain fog cohorts; the superior frontal gyrus in the cohorts with long COVID and brain fog; the middle temporal gyrus in the cohort with brain fog only; and the superior temporal gyrus in the cohort with brain fog only (Fig. 3g–j ). Spearman partial correlations revealed significant negative associations between the number of BBB-disrupted voxels with GBV, right and left WM volume, and right and left cerebral volume and was positively associated with CSF volume (Fig. 4a–g ). Regionally, we observed a negative correlation between BBB disruption in the right frontal cortex with the volume of the right frontal cortex and right frontal pole (Extended Data Fig. 5c ).

figure 3

a , Voxel-based morphometry map indicating brain regions with reduced volume in patients with previous SARS-CoV-2 infection. b – e , Group-wise comparison of total brain volume ( P  = 0.008 brain fog (+) versus control) ( b ), CSF volume ( P  = 0.021 recovered versus control; P  = 0.006 brain fog (+) versus control) ( c ), right WM volume ( P  < 0.0001 recovered versus control; P  = 0.00061 brain fog (+) versus control) ( d ) and left WM volume ( P  = 0.00014 recovered versus control; P  = 0.00098 brain fog (+) versus control) ( e ) in unaffected individuals, recovered individuals and individuals with long COVID and brain fog. Data were analyzed using an ANCOVA, adjusting for age, sex and total intracranial volume (TIV), with Bonferroni correction. f , Surface-based morphometry map indicating brain regions with reduced cortical thickness in patients with previous SARS-CoV-2 infection. g – j , Group-wise comparison of frontal pole thickness ( P  = 0.003 recovered versus control; P  = 0.002 brain fog (−) versus control; P  = 0.001 brain fog (+) versus control) ( g ), superior frontal gyrus thickness ( P  = 0.003 brain fog (−) versus control; P  = 0.008 brain fog (+) versus control) ( h ), middle temporal gyrus ( P  = 0.027 brain fog (+) versus control) ( i ) and superior temporal gyrus ( P  = 0.00012 brain fog (+) versus control) ( j ) in the unaffected, recovered, long COVID and brain fog cohorts. Data were analyzed using an ANCOVA adjusting for age and sex with Bonferroni correction. Maps were generated with Computational Anatomy Toolbox (CAT12) running in the Statistical Parametric Mapping (SPM12) software on MATLAB 2021a. The violin plots show the median (solid line) and IQR (dashed lines). Cohorts were compared with an unpaired t -test, with a family-wise error of less than 0.05, adjusted for age, sex and TIV. Volumetric and thickness region of interest measurements were obtained from volBrain .

figure 4

a – f , Spearman partial correlation between the percentage of BBB-disrupted voxels and GBV ( a ), WM right volume ( b ), WM left volume ( c ), right cerebrum volume ( d ), left cerebrum volume ( e ) and CSF volume ( f ). The dotted lines represent the 95% confidence intervals (CIs). g , Plot of Spearman correlations between BBB permeability and brain volume measurements. Each data point represents one patient. Dot size corresponds to the Spearman correlation coefficient, while color represents the P value. Spearman partial correlation analysis for all panels was adjusted for age, sex and TIV.

Immunovascular dysregulation in long COVID blood samples

Next, we analyzed blood-based biomarkers of neuroinflammation and BBB dysfunction in the recovered and long COVID cohort using multiplex Luminex panels as done for the acute cohort (Extended Data Fig. 6a ). We examined 50 markers of BBB integrity and inflammation. Several markers were increased across all groups compared to controls including IL-1RA, IL-1β, bFGF and IL-13, while IL-9 was the only cytokine decreased in all groups (Fig. 5a–c and Extended Data Fig. 6 ). Glial fibrillary acidic protein (GFAP) was increased in the cohort with brain fog compared to recovered individuals while only transforming growth factor-β (TGFβ) was selectively increased in the cohort with brain fog compared to the cohort with long COVID without brain fog (Fig. 5c,d ). There were also changes in plasma levels of coagulation markers with significantly increased levels of proteins C and S in the recovered cohort and the cohort with brain fog (Extended Data Fig. 6g ). Next, we performed correlation analysis adjusting for age and sex to identify any associations between neuroinflammatory and BBB dysfunction markers with BBB permeability assessed using DCE-MRI. Levels of TGFβ and D-dimer were significantly associated with the percentage of voxels with abnormal leakage but only TGFβ was significant after correcting for multiple comparisons (Fig. 5e ). The levels of TGFβ were significantly associated with GBV, CSF volume, brainstem volume and amygdala volume (Fig. 5f–i ).

figure 5

a – d , Serum and plasma concentrations of IL-8 ( P  = 0.014 brain fog (−) versus control; P  = 0.009 brain fog (+) versus control) ( a ), bFGF ( P  = 0.002 brain fog (−) versus control; P  < 0.0001 brain fog (+) versus control) ( b ), GFAP ( P  = 0.0016 brain fog (+) versus recovered) ( c ) and TGFβ ( P  = 0.0045 brain fog (+) versus control; P  = 0.0115 brain fog (+) versus recovered; P  = 0.0115 brain fog (+) versus brain fog (−)) ( d ) between each cohort. a , b , Data were analyzed using an ANCOVA adjusted for age and sex with Bonferroni correction. c , d , Data were analyzed using a two-tailed Kruskal–Wallis test with Dunn’s correction. e , Correlation plot between analyte levels and BBB permeability. f – i , Spearman correlation between levels of TGFβ and percentage of BBB dysfunction ( f ), percentage of CSF volume ( g ), brainstem volume ( h ) and amygdala volume ( i ). The dashed lines represent the 95% CIs. The violin plots show the median (solid line) and IQR (dashed lines). Each data point represents one patient. a – d , Kruskal–Wallis test. f – i , Spearman partial correlation analysis controlling for age, sex and TIV. Multiple comparisons were Benjamini–Hochberg-corrected, with P  < 0.026 considered discoveries.

White blood cells from patients with COVID-19 activate brain endothelial cells

Given the prevalence of circulating markers indicative of BBB dysfunction and immune cell activation, we examined gene expression changes in peripheral blood mononuclear cells (PBMCs) isolated from unaffected ( n  = 7), recovered ( n  = 5) and patients with long COVID with ( n  = 6) or without ( n  = 5) brain fog using RNA sequencing (RNA-seq). Compared to unaffected individuals, there were 950 differentially expressed genes (DEGs) in recovered individuals, 481 in individuals with long COVID and 126 in individuals with brain fog in our cohorts (Extended Data Fig. 7a–c ). Next, we performed gene ontology (GO) analysis. In the recovered cohort and the cohort without brain fog, upregulated terms included those related to the coagulation system, such as blood coagulation (for example, F13A1 , PROS1 ), platelet activation (for example, F2R , PF4 , PF4V1 ), platelet degranulation (for example, SELP , VCL , CLU ) as well as megakaryocyte development, immunoglobulin production and complement activation (Extended Data Fig. 7d,e and Supplementary Tables 4 and 5 – 12 ). In the cohort with brain fog, there were changes in genes associated with vitamin A metabolism (for example, DGAT2 , DHRS9 ) and regulation of leukocyte homeostasis (for example, CXCL10 , IL6R ) (Extended Data Fig. 7f and Supplementary Tables 4 and 5 – 12 ).

When comparing the cohort with brain fog to the recovered and long COVID cohorts, there were 1,156 and 1,078 DEGs, respectively (Fig. 6a–d ). Principal component analysis (PCA) plots showed a clear separation of the cohort with brain fog from the recovered and long COVID cohorts (Fig. 6a–d ). Compared to the recovered cases, there was a strong enrichment in upregulated terms for pathways related to T cell differentiation and activation (for example, PRDM1, TNF), TGFβ signaling (for example, SMAD3 , SNAI1 , SMURF1 ) and regulation of angiogenesis (for example, HES1 , DLL1 , HIF1A ), while there was downregulation in genes involved in platelet activation, signaling and aggregation (for example, PF4V1 , PF4 , TREML1 ) and hemostasis (for example, F13A1 , GP1BA , GP1BB ) (Fig. 6e and Supplementary Tables 4 and 5 – 12 ). We also compared the transcriptome profile of individuals with and without brain fog in our cohort with long COVID. Upregulated genes were enriched in pathways related to T cell differentiation and activation (for example, RUNX3 , IFNG , TNFSF9 ), negative regulation of the immune response (for example, WASL , ID2 , TNFAIP3 ) and circadian regulation of gene expression (for example, RORA , PER1 , NRIP1 ) (Fig. 6f,j–l ). Pathways related to immunoglobulin, production, defense responses and B cell activation were among those downregulated, including immunoglobulin production ( IGKV1–12 , IGKV1–17 ), adaptive immune response (for example, CX3CR1 , FCGR1BP ) and B cell activation ( HDAC9 , CD180 , MNDA ) (Extended Data Fig. 7 and Supplementary Tables 4 and 5 – 12 ). In agreement with previous studies, several factors involved in the coagulation pathway were downregulated specifically in the cohort with brain fog, including PF4 , PF4V1 and SELP (Fig. 6g–i ) 32 .

figure 6

a , PCA plot of brain fog versus recovered PBMC samples. b , Volcano plot depicting DEGs (red circles) with a log 2 fold change > 0.58 or < −0.58 (vertical dashed lines) and P  < 0.05 (horizontal dashed line). All DEGs with log 2 fold change < 0.58 or > −0.58 and P < 0.05 are also displayed (blue circles). Data were analyzed using a Wald test with multiple comparisons controlled with an FDR. c , PCA plot of brain fog versus recovered PBMC samples. d , Volcano plot of DEGs (red circles) with a log 2 fold change > 0.58 or < −0.58 (vertical dashed lines) and P  < 0.05 (horizontal dashed line). DEGs with a log 2 fold change < 0.58 or or > −0.58 and P < 0.05 are also displayed (blue circles). Data were analyzed using a Wald test with multiple comparisons controlled with an FDR. e , f , Top five upregulated and downregulated terms from brain fog versus recovered ( e ) and brain fog versus long COVID ( f ) cohorts. g – i , Normalized counts of PF4V1 ( g ), PF4 ( h ) and SELP ( i ) in brain fog versus recovered cohorts ( n  = 5 recovered, n  = 5 with brain fog). j – l , Normalized counts of PER1 ( j ), NR1D2 ( k ) and RORA ( l ) in the cohort with brain fog versus the cohort with long COVID ( n  = 6 without brain fog (−), n  = 5 with brain fog (+)). Data were analyzed using a Wald test with multiple comparisons controlled with an FDR. The box plots display the minimum and maximum values (whiskers), the median (solid line) and the IQR (upper and lower box) with significance set at P  < 0.05. Statistical significance was assessed using DESeq2 with a Wald test and Benjamini–Hochberg correction.

We next examined immunovascular interactions in PBMCs isolated from patients with COVID-19 and found increased adhesion of PBMCs to human brain endothelial cells in the cohort with long COVID compared to unaffected individuals, which was heightened in the presence of TNF and only modestly affected by blocking antibodies against ICAM-1 and VCAM-1 (Extended Data Fig. 8a,b ). Furthermore, exposure of human brain endothelial cells to 10% serum from recovered individuals and individuals with long COVID resulted in the upregulation of ICAM1 , VCAM1 and TNF transcripts compared to sera from unaffected individuals (Extended Data Fig. 8c,d ). Previous studies indicated a role for S protein persistence in coagulation dysregulation and brain injury, so we explored how S protein affects endothelial cell activity. Exposure of human brain endothelial cells to S1 protein led to a dose-dependent increase in TNF , TGFβ , ICAM1 and VCAM1 mRNA (Extended Data Fig. 8e ) after 72-h treatment with 0–400 nM S1 spike protein.

Our results suggest that long COVID-derived brain fog is associated with BBB disruption and sustained systemic inflammation. BBB dysfunction was unique to the cohort with brain fog, with disruption evident up to 1 year after active infection in multiple neuroanatomical regions, including the TLs and frontal cortex. BBB dysfunction was not apparent in patients with anosmia without accompanying brain fog implying this might not be a major driver of this symptom. Instead, accumulation of infiltrating T cells expressing IFNγ depleted sensory neurons in patients with anosmia 33 . Inflammation in the olfactory epithelium may result in cerebrovascular damage; higher-resolution MRI may help to tease apart these changes. We observed a significant correlation between BBB disruption in the TLs with the duration of anosmia. BBB disruption in the TL may be linked to anosmia because it contains important regions that form part of the primary olfactory cortex, including the piriform cortex, amygdala and entorhinal cortex, with direct connections to and from the olfactory bulb 33 , 34 , 35 . Our transcriptome data also point to aberrant T cell activation and regulation of IFN production. Previous work revealed an upregulation of activated T cells up to 8 months after infection 36 . T cell activation and waning of the innate and humoral immune response in patients recovering from COVID-19 has also been observed 37 . Dysregulation of the coagulation system has been strongly suggested as a key driver of long COVID and this was the most significantly dysregulated pathway in our cohort with brain fog 32 , 38 .

Structurally, there was reduced brain and WM volume in individuals with brain fog and recovered patients, suggesting that these changes do not primarily drive the fatigue and cognitive impairment associated with brain fog. Similar findings have been reported by others, including longitudinal changes in brain volume and cortical thinning after mild SARS-CoV-2 infection 39 . Neuroimaging has been used to detect other cerebrovascular changes in the brain after SARS-CoV-2 infection, including cerebral microbleeds, hypometabolism and cerebral hypoperfusion 40 , 41 , 42 , 43 , 44 .

BBB dysfunction was associated with neurological impairment during the active phase of SARS-CoV-2 infection, with increased serum levels of the astrocytic protein S100β together with increased levels of IL-6, bFGF and IL-13 suggesting that a heightened systemic inflammatory response may drive BBB dysfunction. Serum levels of S100β are elevated in several neurological disorders including epilepsy, traumatic brain injury and schizophrenia 45 , 46 , 47 . BBB dysfunction also increases with aging, which is an important risk factor for COVID severity 48 , 49 . S100β was associated with age in our acute cohort, which may explain the differences observed in those with brain fog; however, controlling for age still revealed a strong association with acute brain fog. Importantly, participant age was not associated with BBB permeability in our cohort with long COVID, suggesting that BBB disruption is more probably due to the neurological symptoms of long COVID. BBB dysfunction correlated with changes in brain volume and cortical thickness, most notably reduced GBV and increased CSF volume. Similar associations have been reported in bipolar disorder and systemic lupus erythematosus, where individuals with severe BBB disruption had more extensive brain volume loss or greater psychiatric morbidity 50 , 51 . This implies that changes in BBB function are closely related to changes in brain structure and ultimately function. However, longitudinal studies are needed to determine if BBB disruption during acute infection predisposes to the development of long COVID-associated brain fog.

Patients with long COVID had elevated levels of IL-8, GFAP and TGFβ, with TGFβ specifically increased in the cohort with brain fog. GFAP is a robust marker of cerebrovascular damage and is elevated after repetitive head trauma, reflecting BBB disruption, as seen in contact sport athletes and in individuals with self-reported neurological symptoms in long COVID 26 , 52 , 53 . Interestingly, TGFβ was strongly associated with BBB disruption and structural brain changes. TGFβ has been implicated in the pathogenesis of chronic fatigue syndrome, a condition with clinical similarities to long COVID 54 , 55 , 56 .

Insights from animal models and postmortem tissue examined the impact of acute infection on BBB integrity. Brain sections from patients who died from COVID-19 showed fibrinogen extravasation and coagulation system dysregulation 21 , 22 , while mouse models revealed changes in blood vessel morphology with the appearance of string vessels, that is, pathological ‘ghost’ vessels without endothelial cells 28 . Biomarker studies in patients convalescing from COVID-19 also consistently highlighted the involvement of inflammation and coagulation system dysregulation 32 , 38 .

Persistence of viral components, such as S protein, has been hypothesized to be responsible for long COVID-associated neurological symptoms 57 , 58 , 59 . S protein persistence may be involved in neurological sequelae as direct brain injection was associated with coagulation dysregulation and neurodegeneration. This suggests that S protein may have a long half-life in the body. In support of this, immune cells were identified with S protein up to 15 months after infection 60 . Furthermore, we showed that exposure of brain endothelial cells to S protein resulted in an activated endothelial cell phenotype with upregulation of inflammatory cytokines and cell adhesion molecules and probably has a role in long COVID-associated brain fog. Reinforcing these findings, previous studies showed that S protein promoted tight junction degradation, endothelial cell activation and increased adherence of immune cells 23 , 61 . The long-lasting influence of S protein on cerebrovascular function is unknown and should be investigated in future studies, especially considering the longevity of brain endothelial cells.

Long COVID is a substantial burden in many patients after recovery from COVID-19. Patients describe fatigue, memory loss and dyspnea as some of the key symptoms of long COVID, while another subset of patients describe ‘brain fog’ like the one commonly reported in postconcussive syndrome and chronic fatigue syndrome 62 , 63 . Our data suggest that BBB disruption occurs during acute infection and long COVID, where it is strongly associated with cognitive impairment. Our work provides objective evidence for a link between BBB disruption and cognitive impairment within a cohort of patients with long COVID. Further longitudinal studies are required to examine changes in BBB permeability over time and in other postviral illnesses; however, targeted regulation of BBB integrity could now potentially be considered for the treatment of patients with brain fog associated with long COVID.

Our study has some limitations. First, we did not have access to CSF samples from our cohort to confirm molecular BBB breakdown in those with brain fog. However, other studies found increased CSF permeability and BBB disruption in a subset of patients infected with SARS-CoV-2 with elevated Q-albumin ratios that correlated with markers of inflammation 64 . Patients with severe neurological complications from SARS-CoV-2 infection have an increased CSF Q-albumin ratio indicating BBB disruption, coupled with elevated CSF proteins, such as IL-8, which are associated with BBB disruption 65 . Blood–CSF barrier breakdown was the most frequent pathological finding in a multicenter study of 127 patients with COVID-19 with neurological impairment 66 .

Second, we did not examine longitudinal changes in BBB function in our cohort with long COVID; it will be important to determine how long it takes for individuals to recover and if there is resolution of BBB function and prolonged inflammation. As many as half of those infected with SARS-CoV-2 reported no, or incomplete, recovery between 6 and 18 months after infection with 11% reporting deterioration in symptoms 67 . Understanding the long-term outcome of long COVID will be critical to develop treatment options for this large group of individuals.

Our study is also limited by a small sample size. Future studies with larger patient cohorts should perform unbiased proteome profiling on blood and CSF samples. In agreement with our study, another study also found elevated levels of markers of neurological injury and BBB disruption, such as GFAP, in individuals with long COVID with self-reported neurological symptoms 53 . Ultimately, expanding the use of clinical tools focused on understanding the role of the BBB in postviral illnesses may lead to better treatment and management strategies for patients in the future.

Study participants

Participants included patients who had recovered from COVID-19, male or female aged 18 and above with and without neurological symptoms. Participants with long COVID, with symptom persistence over 12 weeks from infection, were also recruited. Candidates were excluded if they had a history of a neurological disorder that may better explain the results of the study such as epilepsy, brain trauma, neuropsychiatric disorder or mild cognitive impairment. Suitable candidates proceeded to assessment with DCE-MRI imaging, Q-SIT olfactory testing and a review of pulmonary imaging and hematological parameters at the time of the COVID-19 diagnosis. The Joint Research Ethics Committee of St James’s and Tallaght Hospital’s approved the study and written informed consent was obtained from all participants. Research was performed according to the principles of the Declaration of Helsinki of 2013. The legal basis for the study was consent according to General Data Protection Regulation principles.

Olfactory testing

The olfactory function of participants was assessed using the Q-SIT. The Q-SIT is a standardized and validated three-item odor identification screen 68 . A score of 2 or more is a normal test and a cutoff score of 1 or less is an abnormal test for anosmia. Q-SIT has displayed high positive and negative predictive value in detecting olfactory dysfunction in patients with COVID-19. In addition, the Q-SIT is a tear-off card test that is disposable, so there is no concern about contamination and transmission of disease from patients with COVID-19 (ref. 69 ).

BBB permeability maps were created using the slope of contrast agent concentration in each voxel over time, calculated using a linear fit model as described previously 70 , 71 , 72 . Thresholds of high permeability were defined by the 95th percentile of all slopes in a previously examined control group. Imaging was performed with a 3T Philips Achieva scanner. Sequences included a T1-weighted anatomical scan (3D gradient echo; time to echo (TE)/repetition time (TR) = 3 ms/6.7 ms; acquisition matrix 268 × 266; voxel size: 0.83 mm × 0.83 mm × 9 mm), T2-weighted imaging (TE/TR = 80 ms/3,000 ms; voxel size: 0.45 mm × 0.45 mm × 4 mm), fluid-attenuated inversion recovery (TE/TR = 125 ms/11,000 ms; voxel size: 0.45 mm × 0.45 mm × 4 mm). For the calculation of precontrast longitudinal relaxation time (T10), the variable flip angle method was used (3D T1W-FFE; TE/TR = 2.78 ms/5.67 ms; acquisition matrix: 240 × 184; voxel size: 0.68 mm × 0.68 mm × 5 mm; flip angles: 10°, 15°, 20°, 25° and 30°). The DCE sequence was then acquired (axial, 3D T1w-FFE; TE/TR = 2.78 ms/5.6 ms; acquisition matrix: 240 × 184; voxel size: 0.68 mm × 0.68 mm × 5 mm; flip angle: 20°; Δ t  = 22.2 s; temporal repetitions: 61; total scan length: 22.6 min). An intravenous bolus injection of the contrast agent gadobenate dimeglumine (Bracco Diagnostics Inc.) was administered using an automatic injector after the first three DCE repetitions. To control for interindividual variabilities due to heart rate, blood flow or rate of contrast injection, each voxel’s leakage rate was normalized to that of the superior sagittal sinus. Each of the 61 temporal repetitions was manually inspected for movement artifacts and was manually excluded from the DCE analysis protocol. If movement artifacts were detected on more than 10% of the entire scan, then the individual was excluded from the analysis. No patients were excluded from the DCE-MRI analysis in this study. One patient from the cohort with long COVID underwent T1 imaging only. Quantification of BBB dysfunction was calculated as described previously 70 , 72 . Briefly, image preprocessing involved image segmentation, registration and normalization to Montreal Neurological Institute (MNI) space using SPM12. To calculate the slow accumulation of contrast agent, a linear fit was applied to min 6–22 of the concentration curve of each voxel with normalization to the venous input function, which is the leakage rate of the superior sagittal sinus. The percentage of the suprathreshold voxels was used as a measure reflecting global BBB leakage.

Volumetric and thickness measurements

T1-weighted anatomical images were uploaded to the volBrain online brain volumetry software ( https://www.volbrain.net/ ) 73 and analyzed with vol2brain 1.0, which is an online pipeline that registers images to the MNI space and reports the volumes of expert-labeled anatomical structures as a percentage of the TIV. We analyzed the volume of the right and left cerebral and cerebellar GM/WM, FLs, TLs, OLs, parietal lobes and CSF along with the thickness of the FLs, TLs, OLs and parietal lobes. All volume data was normalized to the TIV, which is the sum of GM, WM and CSF. Volumes were expressed as a percentage of the TIV. Sixty age-matched and sex-matched healthy control scans were randomly selected from the IXI dataset ( https://brain-development.org/ixi-dataset/ ), which represents 10% of the entire dataset. All scans were performed on the same Philips 3T system at Hammersmith Hospital. Volumetric maps for comparisons between COVID + and COVID − groups were generated in xjView after automatic brain segmentation in the CAT12 toolbox with default parameters and subsequent smoothing with an 8-mm kernel. Thickness maps for comparisons between COVID + and COVID − groups were generated in the CAT12 toolbox run in SPM12 in MATLAB R2021a after brain segmentation, as above, and smoothing with a 15-mm kernel. A two-sample t -test was used for statistical analysis with age, sex and TIV as covariates.

Sample collection

Blood samples were collected into serum separator tubes and EDTA-coated tubes for serum and PBMC isolation, respectively. Serum was separated by centrifugation at 800 g for 10 min at room temperature. PBMCs were separated by layering the blood samples diluted twofold in PBS (cat. no. 14190, Thermo Fisher Scientific) over a Lymphoprep density gradient medium (cat. no. 07851, STEMCELL Technologies) followed by centrifugation at 400 g for 25 min at room temperature at 0 break and 0 acceleration. Plasma was collected and stored at −80 °C and the PBMC layer was collected into a new 50-ml Falcon tube, resuspended to 50 ml with PBS and centrifuged at 800 g for 5 min at room temperature. PBMCs were resuspended in 50 ml PBS and centrifuged at 400 g for 10 min at room temperature. PBMCs were resuspended to 2 × 10 6 cells per ml in Roswell Park Memorial Institute (RPMI) 1640 medium with l -glutamine (cat. no. LZBE12-702F, Lonza) supplemented with 50% FCS (cat. no. F7524, Merck) and 10% dimethylsulfoxide (cat. no. D5879, Merck) and frozen at −80 °C overnight before being moved to liquid nitrogen.

Multiplex assays

A 10-plex Luminex assay (cat. no. LXSAHM-10, R&D Systems) was used for cytokine profiling. Serum samples were diluted twofold in sample dilution buffer. Then, 50 µl of sample or standard was pipetted in duplicate into each wall of an assay 96-well plate. Then, 50 µl of diluted Microparticle Cocktail was added to each well, the plate was covered and incubated for 2 h at room temperature on a shaker at 300 r.p.m. Wells were washed three times with wash buffer before addition of 50 µl of diluted Biotin-Antibody Cocktail. The plate was covered and incubated for 1 h at room temperature on a shaker at 300 r.p.m. Wells were washed as above before the addition of 50 µl of diluted streptavidin-phycoerythrin to each well. The plate was covered and incubated for 30 min at room temperature on a shaker at 300 r.p.m. Wells were washed as above before microparticles were resuspended in 100 µl of wash buffer. The plate was incubated for 2 min at room temperature on a shaker at 300 r.p.m. and was read on a MAGPIX plate reader with the xPONENT software (Luminex). A Bio-Plex Pro Human Cytokine 27-plex Assay (cat. no. M500KCAF0Y, Bio-Rad Laboratories) was used for cytokine profiling. Samples and standards were diluted fourfold in sample dilution and plates were processed according to the manufacturer’s instructions. A ProcartaPlex Human Coagulation Panel 3 4-Plex (cat. no. EPX040-10825-901, Thermo Fisher Scientific) was used for coagulation factor profiling. Plasma samples were collected in citrate tubes and spun at 2,000 g for 10 min. Plasma was diluted 1:500 and assayed according to the manufacturer’s instructions. Separate enzyme-linked immunosorbent assays (ELISAs) were used for additional inflammatory, coagulation and BBB markers and included human tissue factor (1:10 dilution, cat. no. HUFI00258, AssayGenie), human D-dimer (1:20,000 dilution, cat. no. EHDDIMER, Thermo Fisher Scientific), human plasminogen activator inhibitor-1 (1:200 dilution, cat. no. DY9387-05, Bio-Techne), human UCH-L1 (1:2 dilution, cat. no. DY6007-05, Bio-Techne), human VCAM-1 (1:2,000 dilution, cat. no. DY809, Bio-Techne), human ICAM-1 (1:500 dilution, cat. no. DY720-05, Bio-Techne), human ECM-1 (1:1,000 dilution, cat. no. DY3937-05, Bio-Techne), human VEGF (1:5 dilution, cat. no. DY293B, Bio-Techne) and human BDNF (1:5 dilution, cat. no. DY248, Bio-Techne). For analytes at the lower limit, the lower limit of detection was used.

Plasma samples were spotted (2 µl) onto a 0.2-µm nitrocellulose membrane (cat. no. 10401391, Whatman) and allowed to dry for 30 min. Membranes were blocked in 5% BSA (cat. no. A7906, Merck) in PBS with 0.1% Tween 20 (PBST) for 1 h at room temperature. Membranes were incubated overnight in primary antibody in blocking buffer. Membranes were washed three times for 5 min each in PBST, followed by incubation in secondary horseradish peroxidase (HRP)-conjugated antibodies. Membranes were washed three times for 5 min each in PBST and incubated with strong ECL substrate (cat. no. K-12045-D50, Advansta) for 2 min before being developed on a C-Digit (LI-COR Biosciences). Protein bands were quantified in ImageJ (National Institutes of Health). Primary antibodies used were mouse anti-GFAP (1:500 dilution, cat. no. G3893, Merck), rabbit anti-TGFβ (1:500 dilution, cat. no. ab92486, Abcam) and mouse anti-Phospho-Tau (1:500 dilution, cat. no. 10599853, Thermo Fisher Scientific). Secondary antibodies used were anti-mouse HRP (1:5,000 dilution, cat. no. A4416, Merck) and anti-rabbit HRP (1:5,000 dilution, cat. no. A6154, Merck).

Quantitative PCR with reverse transcription

RNA was isolated from PBMCs and the human brain endothelial cell line hCMEC/d3 (cat. no. SCC066, Merck Millipore) with the Omega RNA Isolation Kit (cat. no. R6834-02, VWR) according to the manufacturer’s instructions. Complementary DNA was reverse-transcribed from 500 ng RNA with the High Capacity cDNA Reverse Transcription Kit (cat. no. 4368814, Applied Biosystems). Transcript levels were quantified on a StepOne Plus instrument (Applied Biosystems) with FastStart Universal SYBR Green Master (Rox) master mix (cat. no. 04913914001, Roche). Quantitative PCR with reverse transcription was performed with the following conditions: 95 °C × 2 min (95 °C × 5 s, 60 °C × 30 s) ×40, 95 °C × 15 s, 60 °C × 1 min, 95 °C × 15 s, 60 °C × 15 s. Relative gene expression levels were quantified using the comparative CT method (ΔΔ Ct ). The expression levels of target genes were normalized to β-actin.

Adhesion assay

hCMEC/d3 cells were cultured in EGM2-MV growth medium (cat. no. CC-3202, Lonza) and were stimulated with 10 ng ml −1 recombinant human TNF (cat. no. 300-01A, PeproTech) for 4 h in the presence or absence of 1 µg ml −1 anti-ICAM-1, anti-VCAM-1 or anti-PECAM-1 antibodies before incubation with 1 × 10 5 MitoTracker Orange-labeled PBMCs (cat. no. M7510, Thermo Fisher Scientific) for 1 h at 37 °C. Cells were washed three times in PBS to remove unbound PBMCs and fixed in 4% formaldehyde (cat. no. F1635, Merck) for 10 min at room temperature. The number of adhered PBMCs was counted with the ImageJ cell counter plugin. Images were imported and converted to 8-bit and thresholded. Noise was removed with the despeckle function and images were converted to binary. The cell counter plugin was then used to count adhered PBMCs. Counts were averaged from five images per treatment.

Serum and S protein treatment

hCMEC/d3 cells were seeded in 12-well plates at 2 × 10 5 cells per well and grown to confluence. Medium was replaced with medium containing 10% serum from individuals with COVID and unaffected controls and incubated for up to 72 h. This was followed by RNA isolation. hCMEC/d3 cells were cultured in 12-well plates as described above and stimulated with 4, 40 and 400 nM Recombinant SARS-CoV-2 Spike S1 Subunit Biotin Protein (cat. no. BT10569, R&D Systems) for up to 72 h. RNA was isolated as described above.

PBMCs were thawed in a bead bath at 37 °C, topped up with RPMI medium to 10 ml and centrifuged at 300 g for 5 min at room temperature. PBMCs were allowed to recover for 1 h at 37 °C in RPMI medium supplemented with 10% FCS before RNA isolation. RNA was isolated from 1 × 10 6 PBMCs with the E.Z.N.A. Total RNA Kit I (Omega Bio-tek) according to the manufacturer’s recommendations. RNA samples were analyzed on a TapeStation 2200 (Agilent Technologies) and samples with an RNA integrity number value greater than 7 and ribosomal RNA ratio greater than 1 were used for library preparation and RNA-seq; 10 ng RNA was used for the SMART-Seq v4 Ultra Low Input RNA Kit for sequencing with the Nextera XT DNA Library Preparation Kit. RNA-seq with 100 bp paired-end reads and more than 20 million reads per sample was performed on a NovaSeq 6000 (Illumina). Raw FASTQ files were trimmed with Cutadapt (part of Trim Galore) and aligned to the Gencode GRCh38 Release 43 reference using STAR. The resulting BAM files were sorted with STAR and indexed with Samtools. Gene quantification was performed using the RSEM’s rsem-calculate-expression tool. Count matrices produced by RSEM were processed with tximport and differential expression analysis was performed using DESeq2 v.1.38.3 in R. Alignment metrics were generated using the Genome Analysis Toolkit (GATK)’s CollectAlignmentSummaryMetrics. The distribution of the bases within the transcripts was determined using GATK’s CollectRnaSeqMetrics. Quality control reports were generated with FastQC (part of Trim Galore) and amalgamated into a single report with MultiQC. GO was performed with the clusterProfiler package. Data were filtered for P  < 0.05. The enrichGO function of clusterProfiler was then used with Benjamini–Hochberg adjustment and a cutoff of P  < 0.05. GO enrichment was performed by selecting the biological processes subontology.

Ethical approval

Informed consent was obtained from each participant. All ethical approvals were in place before the initiation of studies on humans. All experiments conformed to the principles set out in the World Medical Association Declaration of Helsinki and the Department of Health and Human Services Belmont Report. The St James’ Hospital ethics committee approved these studies.

Statistical analysis

No statistical methods were used to predetermine sample sizes. SPSS v.28 (IBM Corporation) and Prism v.9 (GraphPad Software) were used for the statistical analysis. Prism v.9 was used to generate the graphs. Categorical variables were compared between groups with chi-squared tests. Cytokine data were log-transformed and analyzed with a general linear model, with age, sex and all comorbidities as covariates. For structural MRI analysis, a multivariate general linear model was used, with age, sex and TIV as covariates. Correlations were assessed with Pearson or Spearman rho correlation tests using partial correlations to control for age, sex and TIV. For repeated blood samples, matched samples were compared using a Wilcoxon signed-rank test. To control for multiple comparisons in multiplex assays, brain region MRI analysis and correlation analysis, FDR was applied using the Benjamini–Hochberg correction. A P  < 0.05 was considered statistically significant. All quantitative PCR, ELISA and adhesion assays were performed in duplicate.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

Data supporting the findings of this study are available from the corresponding authors upon reasonable request. The RNA-seq data have been deposited at the NCBI’s Gene Expression Omnibus and are accessible under accession no. GSE251849 . Publicly available MRI datasets can be accessed at https://brain-development.org/ixi-dataset/ . This study did not generate new or unique reagents. Source data are provided with this paper.

Code availability

No custom software was developed during this study. Open-source packages and libraries and corresponding versions used during the computational analysis are described in the Methods and Reporting Summary . MATLAB scripts and notebooks are available upon reasonable request from the corresponding authors.

Wu, Z. & McGoogan, J. M. Characteristics of and important lessons from the coronavirus disease 2019 (COVID-19) outbreak in China: summary of a report of 72 314 cases from the Chinese Center for Disease Control and Prevention. JAMA 323 , 1239–1242 (2020).

Article   CAS   PubMed   Google Scholar  

Zhu, N. et al. A novel coronavirus from patients with pneumonia in China, 2019. N. Engl. J. Med. 382 , 727–733 (2020).

Article   CAS   PubMed   PubMed Central   Google Scholar  

Helms, J. et al. Neurologic features in severe SARS-CoV-2 infection. N. Engl. J. Med. 382 , 2268–2270 (2020).

Article   PubMed   Google Scholar  

Taquet, M., Geddes, J. R., Husain, M., Luciano, S. & Harrison, P. J. 6-month neurological and psychiatric outcomes in 236 379 survivors of COVID-19: a retrospective cohort study using electronic health records. Lancet Psychiatry 8 , 416–427 (2021).

Article   PubMed   PubMed Central   Google Scholar  

Bohmwald, K., Gálvez, N. M. S., Ríos, M. & Kalergis, A. M. Neurologic alterations due to respiratory virus infections. Front. Cell. Neurosci. 12 , 386 (2018).

Amraei, R. et al. Extracellular vimentin is an attachment factor that facilitates SARS-CoV-2 entry into human endothelial cells. Proc. Natl Acad. Sci. USA 119 , e2113874119 (2022).

Cantuti-Castelvetri, L. et al. Neuropilin-1 facilitates SARS-CoV-2 cell entry and infectivity. Science 370 , 856–860 (2020).

Article   ADS   CAS   PubMed   PubMed Central   Google Scholar  

Hoffmann, M. et al. SARS-CoV-2 cell entry depends on ACE2 and TMPRSS2 and is blocked by a clinically proven protease inhibitor. Cell 181 , 271–280 (2020).

Ni, W. et al. Role of angiotensin-converting enzyme 2 (ACE2) in COVID-19. Crit. Care 24 , 422 (2020).

Vanlandewijck, M. et al. A molecular atlas of cell types and zonation in the brain vasculature. Nature 554 , 475–480 (2018).

Article   ADS   CAS   PubMed   Google Scholar  

Iadecola, C., Anrather, J. & Kamel, H. Effects of COVID-19 on the nervous system. Cell 183 , 16–27 (2020).

Lersy, F. et al. Cerebrospinal fluid features in patients with coronavirus disease 2019 and neurological manifestations: correlation with brain magnetic resonance imaging findings in 58 patients. J. Infect. Dis. 223 , 600–609 (2021).

Schweitzer, F. et al. Cerebrospinal fluid analysis post-COVID-19 is not suggestive of persistent central nervous system infection. Ann. Neurol. 91 , 150–157 (2022).

Thakur, K. T. et al. COVID-19 neuropathology at Columbia University Irving Medical Center/New York Presbyterian Hospital. Brain 144 , 2696–2708 (2021).

Yang, A. C. et al. Dysregulation of brain and choroid plexus cell types in severe COVID-19. Nature 595 , 565–571 (2021).

Stein, S. R. et al. SARS-CoV-2 infection and persistence in the human body and brain at autopsy. Nature 612 , 758–763 (2022).

Monje, M. & Iwasaki, A. The neurobiology of long COVID. Neuron 110 , 3484–3496 (2022).

Natale, N. R., Lukens, J. R. & Petri, W. A. Jr The nervous system during COVID-19: caught in the crossfire. Immunol. Rev. 311 , 90–111 (2022).

Abbott, N. J., Patabendige, A. A., Dolman, D. E. M., Yusof, S. R. & Begley, D. J. Structure and function of the blood–brain barrier. Neurobiol. Dis. 37 , 13–25 (2010).

Greene, C., Hanley, N. & Campbell, M. Claudin-5: gatekeeper of neurological function. Fluids Barriers CNS 16 , 3 (2019).

Lee, M.-H. et al. Microvascular injury in the brains of patients with Covid-19. N. Engl. J. Med. 384 , 481–483 (2021).

Lee, M. H. et al. Neurovascular injury with complement activation and inflammation in COVID-19. Brain 145 , 2555–2568 (2022).

DeOre, B. J., Tran, K. A., Andrews, A. M., Ramirez, S. H. & Galie, P. A. SARS-CoV-2 spike protein disrupts blood–brain barrier integrity via RhoA activation. J. Neuroimmune Pharmacol. 16 , 722–728 (2021).

Krasemann, S. et al. The blood–brain barrier is dysregulated in COVID-19 and serves as a CNS entry route for SARS-CoV-2. Stem Cell Reports 17 , 307–320 (2022).

Pellegrini, L. et al. SARS-CoV-2 infects the brain choroid plexus and disrupts the blood–CSF barrier in human brain organoids. Cell Stem Cell 27 , 951–961.e5 (2020).

Savarraj, J. et al. Brain injury, endothelial injury and inflammatory markers are elevated and express sex-specific alterations after COVID-19. J. Neuroinflammation 18 , 277 (2021).

Schwabenland, M. et al. Deep spatial profiling of human COVID-19 brains reveals neuroinflammation with distinct microanatomical microglia-T-cell interactions. Immunity 54 , 1594–1610 (2021).

Wenzel, J. et al. The SARS-CoV-2 main protease M pro causes microvascular brain pathology by cleaving NEMO in brain endothelial cells. Nat. Neurosci. 24 , 1522–1533 (2021).

Montalvan, V., Lee, J., Bueso, T., De Toledo, J. & Rivas, K. Neurological manifestations of COVID-19 and other coronavirus infections: a systematic review. Clin. Neurol. Neurosurg. 194 , 105921 (2020).

Ng Kee Kwong, K. C., Mehta, P. R., Shukla, G. & Mehta, A. R. COVID-19, SARS and MERS: a neurological perspective. J. Clin. Neurosci. 77 , 13–16 (2020).

O’Doherty, L. et al. Study protocol for the St James’s Hospital, Tallaght University Hospital, Trinity College Dublin Allied Researchers’ (STTAR) Bioresource for COVID-19. HRB Open Res. 5 , 20 (2022).

Ryan, F. J. et al. Long-term perturbation of the peripheral immune system months after SARS-CoV-2 infection. BMC Med. 20 , 26 (2022).

Finlay, J. B. et al. Persistent post–COVID-19 smell loss is associated with immune cell infiltration and altered gene expression in olfactory epithelium. Sci. Transl. Med. 14 , eadd0484 (2022).

Zhou, G., Lane, G., Cooper, S. L., Kahnt, T. & Zelano, C. Characterizing functional pathways of the human olfactory system. eLife 8 , e47177 (2019).

Saive, A.-L., Royet, J.-P. & Plailly, J. A review on the neural bases of episodic odor memory: from laboratory-based to autobiographical approaches. Front. Behav. Neurosci. 8 , 240 (2014).

Phetsouphanh, C. et al. Immunological dysfunction persists for 8 months following initial mild-to-moderate SARS-CoV-2 infection. Nat. Immunol. 23 , 210–216 (2022).

Zheng, H.-Y. et al. Longitudinal transcriptome analyses show robust T cell immunity during recovery from COVID-19. Signal Transduct. Target. Ther. 5 , 294 (2020).

Taquet, M. et al. Acute blood biomarker profiles predict cognitive deficits 6 and 12 months after COVID-19 hospitalization. Nat. Med. 29 , 2498–2508 (2023).

Douaud, G. et al. SARS-CoV-2 is associated with changes in brain structure in UK Biobank. Nature 604 , 697–707 (2022).

Donegani, M. I. et al. Brain metabolic correlates of persistent olfactory dysfunction after SARS-Cov2 infection. Biomedicines 9 , 287 (2021).

Fitsiori, A., Pugin, D., Thieffry, C., Lalive, P. & Vargas, M. I. COVID-19 is associated with an unusual pattern of brain microbleeds in critically ill patients. J. Neuroimaging 30 , 593–597 (2020).

Guedj, E. et al. 18 F-FDG brain PET hypometabolism in patients with long COVID. Eur. J. Nucl. Med. Mol. Imaging 48 , 2823–2833 (2021).

Qin, Y. et al. Long-term microstructure and cerebral blood flow changes in patients recovered from COVID-19 without neurological manifestations. J. Clin. Invest. 131 , e147329 (2021).

Tian, T. et al. Long-term follow-up of dynamic brain changes in patients recovered from COVID-19 without neurological manifestations. JCI Insight 7 , e155827 (2022).

Aleksovska, K. et al. Systematic review and meta-analysis of circulating S100B blood levels in schizophrenia. PLoS ONE 9 , e106342 (2014).

Article   ADS   PubMed   PubMed Central   Google Scholar  

Thelin, E. P., Nelson, D. W. & Bellander, B.-M. A review of the clinical utility of serum S100B protein levels in the assessment of traumatic brain injury. Acta Neurochir. 159 , 209–225 (2017).

Greene, C. et al. Microvascular stabilization via blood–brain barrier regulation prevents seizure activity. Nat. Commun. 13 , 2003 (2022).

Montagne, A. et al. Blood–brain barrier breakdown in the aging human hippocampus. Neuron 85 , 296–302 (2015).

Senatorov, V. V. et al. Blood–brain barrier dysfunction in aging induces hyperactivation of TGFβ signaling and chronic yet reversible neural dysfunction. Sci. Transl. Med. 11 , eaaw8283 (2019).

Kamintsky, L. et al. Blood–brain barrier imaging as a potential biomarker for bipolar disorder progression. Neuroimage Clin. 26 , 102049 (2020).

Kamintsky, L. et al. Blood–brain barrier leakage in systemic lupus erythematosus is associated with gray matter loss and cognitive impairment. Ann. Rheum. Dis. 79 , 1580–1587 (2020).

Abdelhak, A. et al. Blood GFAP as an emerging biomarker in brain and spinal cord disorders. Nat. Rev. Neurol. 18 , 158–172 (2022).

Peluso, M. J. et al. Plasma markers of neurologic injury and inflammation in people with self-reported neurologic postacute sequelae of SARS-CoV-2 infection. Neurol. Neuroimmunol. Neuroinflamm. 9 , e200003 (2022).

McCarthy, M. J. Circadian rhythm disruption in myalgic encephalomyelitis/chronic fatigue syndrome: implications for the post-acute sequelae of COVID-19. Brain Behav. Immun. Health 20 , 100412 (2022).

Lee, W. K. et al. Exogenous transforming growth factor-β in brain-induced symptoms of central fatigue and suppressed dopamine production in mice. Int. J. Mol. Sci. 22 , 2580 (2021).

Montoya, J. G. et al. Cytokine signature associated with disease severity in chronic fatigue syndrome patients. Proc. Natl Acad. Sci. USA 114 , E7150–E7158 (2017).

Chen, B., Julg, B., Mohandas, S. & Bradfute, S. B. Viral persistence, reactivation, and mechanisms of long COVID. eLife 12 , e86015 (2023).

Proal, A. D. et al. SARS-CoV-2 reservoir in post-acute sequelae of COVID-19 (PASC). Nat. Immunol. 24 , 1616–1627 (2023).

Rong, Z. et al. SARS-CoV-2 spike protein accumulation in the skull-meninges-brain axis: potential implications for long-term neurological complications in post-COVID-19. Preprint at bioRxiv https://doi.org/10.1101/2023.04.04.535604 (2023).

Patterson, B. K. et al. Persistence of SARS CoV-2 S1 protein in CD16 + monocytes in post-acute sequelae of COVID-19 (PASC) up to 15 months post-infection. Front. Immunol. 12 , 746021 (2022).

Buzhdygan, T. P. et al. The SARS-CoV-2 spike protein alters barrier function in 2D static and 3D microfluidic in-vitro models of the human blood–brain barrier. Neurobiol. Dis. 146 , 105131 (2020).

Rass, V. et al. Neurological outcomes one year after COVID-19 diagnosis: a prospective longitudinal cohort study. Eur. J. Neurol. 29 , 1685–1696 (2022).

Whitaker, M. et al. Persistent COVID-19 symptoms in a community study of 606,434 people in England. Nat. Commun. 13 , 1957 (2022).

Normandin, E. et al. Intrathecal inflammatory responses in the absence of SARS-CoV-2 nucleic acid in the CSF of COVID-19 hospitalized patients. J. Neurol. Sci. 430 , 120023 (2021).

Etter, M. M. et al. Severe Neuro-COVID is associated with peripheral immune signatures, autoimmunity and neurodegeneration: a prospective cross-sectional study. Nat. Commun. 13 , 6777 (2022).

Jarius, S. et al. Cerebrospinal fluid findings in COVID-19: a multicenter study of 150 lumbar punctures in 127 patients. J. Neuroinflammation 19 , 19 (2022).

Hastie, C. E. et al. Outcomes among confirmed cases and a matched comparison group in the Long-COVID in Scotland study. Nat. Commun. 13 , 5663 (2022).

Jackman, A. H. & Doty, R. L. Utility of a three-item smell identification test in detecting olfactory dysfunction. Laryngoscope 115 , 2209–2212 (2005).

Lechien, J. R. et al. Olfactory and gustatory dysfunctions as a clinical presentation of mild-to-moderate forms of the coronavirus disease (COVID-19): a multicenter European study. Eur. Arch. Otorhinolaryngol. 277 , 2251–2261 (2020).

O’Keeffe, E. et al. Dynamic blood–brain barrier regulation in mild traumatic brain injury. J. Neurotrauma 37 , 347–356 (2020).

Weissberg, I. et al. Imaging blood–brain barrier dysfunction in football players. JAMA Neurol. 71 , 1453–1455 (2014).

Veksler, R. et al. Slow blood-to-brain transport underlies enduring barrier dysfunction in American football players. Brain 143 , 1826–1842 (2020).

Manjón, J. V. & Coupé, P. volBrain: an online MRI brain volumetry system. Front. Neuroinform. 10 , 30 (2016).

Download references

Acknowledgements

This work was supported in part by a Science Foundation Ireland (SFI) COVID-19 rapid response grant (no. 20/COV/0312) to M.C. and C.P.D. The laboratory is also supported by grants from the SFI (Eye-D-21/SPP/3732) to M.C., the Irish Research Council and by a research grant from SFI under grant no. 21/RC/10294_P2 and cofunded under the European Regional Development fund by FutureNeuro industry partners to M.C. and C.P.D. The Campbell laboratory is also supported by a European Research Council grant, ‘Retina‐Rhythm’ (no. 864522). The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. We thank N. Bourke and M. McElheron of Trinity College Dublin for assistance with the Luminex assays.

Author information

Authors and affiliations.

Smurfit Institute of Genetics, Trinity College Dublin, Dublin, Ireland

Chris Greene, Eoin O’Keeffe, Jeffrey O’Callaghan, Bennett Thomson & Matthew Campbell

Department of Neurology, Health Care Centre, St James’s Hospital, Dublin, Ireland

Ruairi Connolly, Declan Brennan, Aoife Laffan, Lilia Zaporojan & Colin P. Doherty

The Irish Longitudinal Study on Ageing, School of Medicine, Trinity College Dublin, Dublin, Ireland

Emma Connolly

Clinical Research Facility, St James’s Hospital, Dublin, Ireland

Department of Intensive Care Medicine, Multidisciplinary Intensive Care Research Organization, Trinity Centre for Health Sciences, St James’s University Hospital, Dublin, Ireland

Ignacio Martin-Loeches

Trinity Translational Medicine Institute, Trinity College Dublin, St James’s Hospital, Dublin, Ireland

Aideen Long & Cliona Ni Cheallaigh

Department of Immunology, St James’s Hospital, Dublin, Ireland

Cliona Ni Cheallaigh & Niall Conlon

St James’s Hospital, Tallaght University Hospital, Trinity College Dublin Allied Researchers (STTAR) Bioresource, Trinity College Dublin, Dublin, Ireland

Niall Conlon

Academic Unit of Neurology, Biomedical Sciences Institute, Trinity College Dublin, Dublin, Ireland

Colin P. Doherty

FutureNeuro, Science Foundation Ireland Research Centre for Chronic and Rare Neurological Diseases, Royal College of Surgeons in Ireland, University of Medicine and Health Sciences, Dublin, Ireland

Colin P. Doherty & Matthew Campbell

You can also search for this author in PubMed   Google Scholar

Contributions

C.G. designed the research, performed the experiments, collected and analyzed the data, and wrote the manuscript. R.C. carried out patient recruitment and assessment. E.O.K. prepared the samples. D.B., A. Laffan and L.Z. carried out patient recruitment and assessment. J.O.C. and B.T. carried out the RNA-seq analysis. E.C. carried out data maintenance and the statistical analysis. N.C., R.A., I.M.-L., A. Long and C.N.C. collected the samples and took part in patient recruitment. C.P.D. and M.C. conceived the project, designed the experiments and edited the manuscript.

Corresponding authors

Correspondence to Colin P. Doherty or Matthew Campbell .

Ethics declarations

Competing interests.

The authors declare no competing interests.

Peer review

Peer review information.

Nature Neuroscience thanks Stig Cramer and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended data fig. 1 disease severity is associated with changes in serum markers of inflammation..

a ) Study design for acute COVID cohort. b – d ) Volcano plot of differentially expressed analytes in mild, moderate, and severe COVID patient’s vs healthy controls. e ) Volcano plot of differentially expressed analytes between patients with vs without brain fog. f ) Volcano plot of differentially expressed proteins between patients receiving vs not receiving COVID-directed medication. g ) Levels of expression of Factor IX (p < 0.0001 Mild vs Control, p = 0.001 Moderate vs Control, p < 0.0001 Severe vs Control), Protein C (p < 0.0001 Mild vs Control, p < 0.0001 Moderate vs Control, p < 0.0001 Severe vs Control), Protein S (p < 0.0001 Mild vs Control, p = 0.012 Moderate vs Control), Tissue Factor, VCAM-1 (p = 0.0003 Severe vs Control), vWF, PAI-1 (p < 0.0001 Severe vs Control), D-dimer (p = 0.002 Severe vs Control). Data was analysed by ANCOVA adjusting for age and sex with Bonferroni correction for multiple comparisons. h ) Levels of expression of IL-9, IL-13, IL-17, IL-1β (p = 0.013 Moderate vs Control, p < 0.0001 Severe vs Control), 10 kDa interferon gamma-induced protein (p < 0.0001 Moderate vs Control, p < 0.0001 Severe vs Control), G-CSF (p = 0.007 Moderate vs Control, p = 0.006 Severe vs Control), GM-CSF (p = 0.011 Severe vs Control), ICAM-1 (p = 0.0001 Severe vs Control). Data was analysed by ANCOVA adjusting for age and sex with Bonferroni correction for multiple comparisons. Violin plots show median (solid line) and interquartile values (dashed lines). Each datapoint represents one patient. All plots were generated in GraphPad PRISM. Multiple comparisons were controlled for by Benjamini-Hochberg correction. Schematic in a created with BioRender.com .

Extended Data Fig. 2 COVID severity is associated with blood levels of inflammatory cytokines.

a ) Pearson correlation between patient age and COVID severity. b ) Pearson correlation between patient age and duration of hospitalisation. c ) Pearson correlation between comorbidity score and COVID severity. d – h ) Pearson correlation between levels of IL-6, TNF, IL-1B, IP-10 and Protein S with COVID severity. i ) Pearson correlation between age and levels of S100β. Dashed lines represent 95 % confidence intervals. Data in d-i were analysed by partial correlations with age, sex, and comorbidities as covariates. Multiple comparisons were controlled for by Benjamini-Hochberg correction.

Extended Data Fig. 3 Longitudinal changes in serum analytes.

a ) Volcano plot of differentially expressed analytes between sample 1 (T1) and sample 2 (T2) following deterioration of clinical symptoms in hospitalised patients. b ) Levels of sVCAM1, PAI-1, sICAM-1, IL-4, IL-13 and IL-8 between T1 and T2 (n = 27 paired samples). Error bars are mean ± s.d. For IL-8, ICAM-1, VCAM-1, PAI-1, VEGF, BDNF (n = 34 paired samples). Matched samples were compared with two-sided Wilcoxon signed-rank test. Multiple comparisons were controlled for by Benjamini-Hochberg correction.

Extended Data Fig. 4 BBB disruption persists up to one year post SARS-CoV-2 infection in individuals with brain fog.

a ) Study design for Long COVID cohort. b ) Signal to noise ratio in T1 weighted scans (n = 21 Brain Fog (-), n = 11 Brain Fog (+)). c ) Summary statistics for the effects of age and brain fog on BBB permeability (effect of brain fog status on % BBB disruption p = 0.0006). d ) Representative DCE-MRI scans from 3 recovered, Long COVID and brain fog participants. e – h ) Quantification of regional BBB permeability in the right temporal lobe (p = 0.037 Brain Fog (+) vs Recovered), left temporal lobe (p = 0.036 Brain Fog (+) vs Recovered), right frontal cortex (p = 0.039 Brain Fog (+) vs Brain Fog (-)) and left frontal cortex (p = 0.033 Brain Fog (+) vs Brain Fog (-). Data analysed by two-sided Kruskal-Wallis test with Bonferroni correction. i ) Whole brain 95th percentile values in participants with or without brain fog (p = 0.001). Data was analysed by two-sided Mann-Whitney test. j ) 95th percentile Ktrans values in participants with or without brain fog (p = 0.0072). Data was analysed by two-sided Mann-Whitney test. Box plots display min and max values (whiskers), median values (solid line) and interquartile range (upper/lower box). Violin plots show median (solid line) and interquartile values (dashed lines). Schematic in a created with BioRender.com .

Extended Data Fig. 5 BBB disruption in the temporal lobes correlates with duration of anosmia.

a , b ) Spearman correlations between regional BBB disruption in the temporal lobes and the duration of anosmia. Dashed lines represent 95 % confidence intervals. c ) Correlation heatmap between regional BBB disruption, anosmia status and cognitive performance as assessed by the Montreal Cognitive Assessment (MOCA). Data was analysed by Spearman partial correlations adjusted for age and sex.

Extended Data Fig. 6 Systemic inflammatory changes in recovered and Long COVID patients.

a ) Study design for Long COVID cohort. b ) Demographics of all cohorts. Data was analysed by One-Way ANOVA with Bonferroni correction for multiple comparisons. c – e ) Volcano plots of differentially expressed analytes in recovered, Long COVID and brain fog cohorts compared to healthy controls. Multiple comparisons were controlled for by Benjamini-Hochberg correction. f ) Level of expression of Factor IX, Protein C (p = 0.0001 Recovered vs Control, p = 0.001 Brain Fog (+) vs Control), Protein S (p = 0.0008 Recovered vs Control, p = 0.013 Brain Fog (+) vs Control), vWF, D-dimer, PAI-1 (p = 0.001 Recovered vs Control, p = 0.017 Brain Fog (-) vs Control), VCAM-1 and Tissue Factor according to disease status. Data was analysed by ANCOVA adjusting for age and sex with Bonferroni correction for multiple comparisons. g ) Levels of expression of IL-9 (p < 0.0001 Recovered vs Control, p < 0.0001 Brain Fog (-) vs Control, p < 0.0001 Brain Fog (+) vs Control), IL-13 (p < 0.0001 Recovered vs Control, p < 0.0001 Brain Fog (-) vs Control, p < 0.0001 Brain Fog (+) vs Control), IL-6, IL-1β (p = 0.046 Recovered vs Control, p < 0.0001 Brain Fog (-) vs Control, p = 0.007 Brain Fog (+) vs Control), IL-1RA (p = 0.002 Recovered vs Control, p < 0.0001 Brain Fog (-) vs Control, p = 0.0005 Brain Fog (+) vs Control), IP-10, G-CSF and ICAM-1 (p = 0.020 Recovered vs Control, p = 0.003 Brain Fog (-) vs Control) according to disease status. Data was analysed by ANCOVA adjusting for age and sex with Bonferroni correction for multiple comparisons. Violin plots show median (solid line) and interquartile values (dashed lines). Each datapoint represents one patient. Schematic in a created with BioRender.com .

Extended Data Fig. 7 Transcriptional characterisation of PBMCs across all patient groups.

a – c ) Volcano plots of differentially expressed analytes in recovered, Long COVID and brain fog patient’s vs healthy controls. Data was analysed by Wald test with multiple comparisons controlled with Benjamini-Hochberg correction. d – f ) Bubble plots of enriched Gene Ontology biological processes in each cohort. Bubble plots show top 20 enriched terms with absLog2FC cut-off < 0.59 and adjusted p-value < 0.05. Pathway enrichment p values were calculated in clusterProfiler with the enrichGO function with Benjamini-Hochberg adjustment and cutoff p-value of 0.05.

Extended Data Fig. 8 Adhesion of Long COVID PBMCs to human brain endothelial cells.

a , b ) Peripheral blood mononuclear cells (PBMCs) from Long COVID participants show greater adherence to the human brain endothelial cell line, hCMEC/d3, in the presence or absence of 10 ng/ml TNFa (n = 5 Control, n = 7 Long COVID). Error bars are mean ± s.d. Data in a was analysed by two-way ANOVA with Tukey correction. Each datapoint represents one patient. c ) PBMCs from long COVID patients in the presence of IgG, VCAM-1 or ICAM-1 blocking antibodies (n = 6 long COVID patients). Data was analysed by repeated measures one-way ANOVA with Tukey correction for multiple comparisons. Each datapoint represents one patient. d ) Schematic of experiments to assess the effect of long COVID patient serum or spike protein on human brain endothelial cells. e ) Exposure of hCMEC/d3 cells to 10 % serum from healthy or long COVID participants and quantification of gene expression changes by q-RT-PCR. Long COVID serum significantly increased TNF ( P = 0.0006) and VCAM1 ( P < 0.0001) vs control serum. f ) Exposure of hCMEC/d3 cells to vehicle or 4, 40 or 400 nM S1 spike protein and quantification of gene expression changes by q-RT-PCR. 400 nM S1 spike protein significantly increased TNF ( P = 0.024), TGFB1 ( P = 0.0274), VCAM1 ( P < 0.0001), MCP1 ( P = 0.0269) and SNAI1 ( P = 0.0389) vs vehicle. For e, n = 4 healthy serum samples and n = 8 long COVID serum samples were used. Scale bar – 50 µm. Schematics in a and d created with BioRender.com .

Supplementary information

Supplementary information.

Supplementary Tables 1–4.

Reporting Summary

Supplementary tables 5–13.

GO biological analysis of long COVID PBMCs.

Source Data Fig. 1

Statistical source data.

Source Data Fig. 2

Source data fig. 3, source data fig. 4, source data fig. 5, source data fig. 6, source data extended data fig. 1, source data extended data fig. 3, source data extended data fig. 4, source data extended data fig. 5, source data extended data fig. 6, source data extended data fig. 8, rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Greene, C., Connolly, R., Brennan, D. et al. Blood–brain barrier disruption and sustained systemic inflammation in individuals with long COVID-associated cognitive impairment. Nat Neurosci (2024). https://doi.org/10.1038/s41593-024-01576-9

Download citation

Received : 16 November 2022

Accepted : 09 January 2024

Published : 22 February 2024

DOI : https://doi.org/10.1038/s41593-024-01576-9

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

Quick links

  • Explore articles by subject
  • Guide to authors
  • Editorial policies

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

how long is an abstract research paper

HTML conversions sometimes display errors due to content that did not convert correctly from the source. This paper uses the following packages that are not yet supported by the HTML conversion tool. Feedback on these issues are not necessary; they are known and are being worked on.

  • failed: kantlipsum
  • failed: textpos

Authors: achieve the best HTML results from your LaTeX submissions by following these best practices .

redacted \correspondingauthor Ashley Edwards ( [email protected] ), Jack Parker-Holder ( [email protected] ). \banner figures/hook.pdf A whole new world : Genie is capable of converting a variety of different prompts into interactive, playable environments that can be easily created, stepped into, and explored. This is made possible via a latent action interface, learned fully unsupervised from Internet videos. On the right we see a few generated steps for taking two latent actions. See more examples on our website .

Genie: Generative Interactive Environments

We introduce Genie, the first generative interactive environment trained in an unsupervised manner from unlabelled Internet videos. The model can be prompted to generate an endless variety of action-controllable virtual worlds described through text, synthetic images, photographs, and even sketches. At 11B parameters, Genie can be considered a foundation world model . It is comprised of a spatiotemporal video tokenizer, an autoregressive dynamics model, and a simple and scalable latent action model. Genie enables users to act in the generated environments on a frame-by-frame basis despite training without any ground-truth action labels or other domain-specific requirements typically found in the world model literature. Further the resulting learned latent action space facilitates training agents to imitate behaviors from unseen videos, opening the path for training generalist agents of the future.

1 1. Introduction

The last few years have seen an emergence of generative AI , with models capable of generating novel and creative content. Driven by breakthroughs in architectures such as transformers (Vaswani et al., 2017 ) , advances in hardware, and a recent focus on scaling models and datasets, we can now generate coherent, conversational language (Radford et al., 2018 , 2019 ; Brown et al., 2020 ) , as well as crisp and aesthetically pleasing images from a text prompt (Ramesh et al., 2021 , 2022 ; Saharia et al., 2022 ; Rombach et al., 2022 ) . Early signs indicate video generation will be yet another frontier, with recent results suggesting that such models may also benefit from scale (Hong et al., 2023 ; Ho et al., 2022a ; Esser et al., 2023 ; Blattmann et al., 2023a ) . Still, there remains a gulf between the level of interactions and engagement of video generative models and language tools such as ChatGPT, let alone more immersive experiences.

What if, given a large corpus of videos from the Internet, we could not only train models capable of generating novel images or videos, but entire interactive experiences? We propose generative interactive environments , a new paradigm for generative AI whereby interactive environments can be generated from a single text or image prompt. Our approach, Genie, is trained from a large dataset of over 200,000 hours of publicly available Internet gaming videos and, despite training without action or text annotations , is controllable on a frame-by-frame basis via a learned latent action space (see Table   1 for a comparison to other approaches). At 11B parameters, Genie exhibits properties typically seen in foundation models—it can take an unseen image as a prompt making it possible to create and play entirely imagined virtual worlds (e.g Figure   1 ).

Refer to caption

Genie builds on ideas from state-of-the-art video generation models (Villegas et al., 2023 ; Gupta et al., 2023 ) , with a core design choice being spatiotemporal (ST) transformers (Xu et al., 2020 ) which are used in all of our model components. Genie utilizes a novel video tokenizer, and extracts latent actions via a causal action model. Both the video tokens and latent actions are passed to a dynamics model, which autoregressively predicts the next frame using MaskGIT (Chang et al., 2022 ) . We provide a rigorous scaling analysis of our architecture with respect to both batch and model size, which we vary from 40M to 2.7B parameters. The results show that our architecture scales gracefully with additional computational resources, leading to a final 11B parameter model. We train Genie on a filtered set of 30,000 hours of Internet gameplay videos from hundreds of 2D platformer games, producing a foundation world model for this setting.

To demonstrate the generality of our approach, we also train a separate model on action-free robot videos from the RT1 dataset (Brohan et al., 2023 ) , learning a generative environment with consistent latent actions. Finally, we show that latent actions learned from Internet videos can be used for inferring policies from unseen action-free videos of simulated reinforcement learning (RL) environments, indicating that Genie may hold the key to unlocking unlimited data for training the next generation of generalist agents (Open Ended Learning Team et al., 2021 ; Bauer et al., 2023 ; Reed et al., 2022 ; Clune, 2019 ) .

Refer to caption

2 2. Methodology

Genie is a generative interactive environment trained from video-only data. In this section we begin with preliminaries before explaining the main components of our model.

Several components in the Genie architecture are based on the Vision Transformer (ViT) (Vaswani et al., 2017 ; Dosovitskiy et al., 2021 ) . Notably, the quadratic memory cost of transformers poses challenges for videos, which can contain up to O ⁢ ( 10 4 ) 𝑂 superscript 10 4 O(10^{4}) italic_O ( 10 start_POSTSUPERSCRIPT 4 end_POSTSUPERSCRIPT ) tokens. We thus adopt a memory efficient ST-transformer architecture (inspired by Xu et al. ( 2020 ) , see Figure   3 ) across all model components, balancing model capacity with computational constraints.

Refer to caption

Unlike a traditional transformer where every token attends to all others, an ST-transformer contains L 𝐿 L italic_L spatiotemporal blocks with interleaved spatial and temporal attention layers, followed by a feed-forward layer (FFW) as standard attention blocks. The self-attention in the spatial layer attends over the 1 × H × W 1 𝐻 𝑊 1\times H\times W 1 × italic_H × italic_W tokens within each time step, and in the temporal layer attends over T × 1 × 1 𝑇 1 1 T\times 1\times 1 italic_T × 1 × 1 tokens across the T 𝑇 T italic_T time steps. Similar to sequence transformers, the temporal layer assumes a causal structure with a causal mask. Crucially, the dominating factor of computation complexity (i.e. the spatial attention layer) in our architecture scales linearly with the number of frames rather than quadratically, making it much more efficient for video generation with consistent dynamics over extended interactions. Further, note that in the ST block, we include only one FFW after both spatial and temporal components, omitting the post-spatial FFW to allow for scaling up other components of the model, which we observe to improve results significantly.

2.1 Model Components

As shown in Figure   2 , our model contains three key components: 1) a latent action model that infers the latent action 𝒂 𝒂 \bm{a} bold_italic_a between each pair of frames and 2) a video tokenizer that converts raw video frames into discrete tokens 𝒛 𝒛 \bm{z} bold_italic_z and 3) a dynamics model that, given a latent action and past frame tokens, predicts the next frame of the video. The model is trained in two phases following a standard autoregressive video generation pipeline: we train the video tokenizer first, which is used for the dynamics model. We then co-train the latent action model (directly from pixels) and the dynamics model (on video tokens).

Latent Action Model (LAM) To achieve controllable video generation, we condition each future frame prediction on the action taken at the previous frame. However, such action labels are rarely available in videos from the Internet and action annotation can be costly to obtain. Instead, we learn latent actions in a fully unsupervised manner (see Figure   4 ).

Refer to caption

𝑡 1 \hat{x}_{t+1} over^ start_ARG italic_x end_ARG start_POSTSUBSCRIPT italic_t + 1 end_POSTSUBSCRIPT .

To train the model, we leverage a VQ-VAE-based objective (van den Oord et al., 2017 ) , which enables us to limit the number of predicted actions to a small discrete set of codes. We limit the vocabulary size | A | 𝐴 |A| | italic_A | of the VQ codebook, i.e. the maximum number of possible latent actions, to a small value to permit human playability and further enforce controllability (we use | A | = 8 𝐴 8 |A|=8 | italic_A | = 8 in our experiments). As the decoder only has access to the history and latent action, a ~ t subscript ~ 𝑎 𝑡 \tilde{a}_{t} over~ start_ARG italic_a end_ARG start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT should encode the most meaningful changes between the past and the future for the decoder to successfully reconstruct the future frame. Note that this decoder exists only to give the LAM training signal. In fact, apart from the VQ codebook, the entire LAM is discarded at inference time and replaced with actions from the user.

We utilize our ST-transformer architecture for the latent action model. The causal mask in the temporal layer allows us to take the entire video 𝒙 1 : T subscript 𝒙 : 1 𝑇 \bm{x}_{1:T} bold_italic_x start_POSTSUBSCRIPT 1 : italic_T end_POSTSUBSCRIPT as input and generate all latent actions between each frame 𝒂 ~ 1 : T − 1 subscript ~ 𝒂 : 1 𝑇 1 \tilde{\bm{a}}_{1:T-1} over~ start_ARG bold_italic_a end_ARG start_POSTSUBSCRIPT 1 : italic_T - 1 end_POSTSUBSCRIPT .

Video Tokenizer Following prior work (Villegas et al., 2023 ; Gupta et al., 2023 ; Yan et al., 2023 ) , we compress videos into discrete tokens to reduce dimensionality and enable higher quality video generation (see Figure   5 ). We again make use of VQ-VAE, which takes in T 𝑇 T italic_T frames of video 𝒙 1 : T = ( x 1 , x 2 , ⋯ , x T ) ∈ ℝ T × H × W × C subscript 𝒙 : 1 𝑇 subscript 𝑥 1 subscript 𝑥 2 ⋯ subscript 𝑥 𝑇 superscript ℝ 𝑇 𝐻 𝑊 𝐶 \bm{x}_{1:T}=(x_{1},x_{2},\cdots,x_{T})\in\mathbb{R}^{T\times H\times W\times C} bold_italic_x start_POSTSUBSCRIPT 1 : italic_T end_POSTSUBSCRIPT = ( italic_x start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , italic_x start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT , ⋯ , italic_x start_POSTSUBSCRIPT italic_T end_POSTSUBSCRIPT ) ∈ blackboard_R start_POSTSUPERSCRIPT italic_T × italic_H × italic_W × italic_C end_POSTSUPERSCRIPT as input, generating discrete representations for each frame 𝒛 1 : T = ( z 1 , z 2 , ⋯ , z T ) ∈ 𝕀 T × D subscript 𝒛 : 1 𝑇 subscript 𝑧 1 subscript 𝑧 2 ⋯ subscript 𝑧 𝑇 superscript 𝕀 𝑇 𝐷 \bm{z}_{1:T}=(z_{1},z_{2},\cdots,z_{T})\in\mathbb{I}^{T\times D} bold_italic_z start_POSTSUBSCRIPT 1 : italic_T end_POSTSUBSCRIPT = ( italic_z start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , italic_z start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT , ⋯ , italic_z start_POSTSUBSCRIPT italic_T end_POSTSUBSCRIPT ) ∈ blackboard_I start_POSTSUPERSCRIPT italic_T × italic_D end_POSTSUPERSCRIPT , where D 𝐷 D italic_D is the size of the discrete latent space. The tokenizer is trained using a standard VQ-VQAE objective over the entire video sequence.

Refer to caption

Unlike prior works that focus on spatial-only compression in the tokenization phase (Hong et al., 2022 ; Wu et al., 2022 ; Gupta et al., 2023 ) , we utilize the ST-transformer in both the encoder and decoder to incorporate temporal dynamics in the encodings, which improves the video generation quality. By the causal nature of the ST-transformer, each discrete encoding z t subscript 𝑧 𝑡 z_{t} italic_z start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT contains information from all previously seen frames of the video 𝒙 1 : t subscript 𝒙 : 1 𝑡 \bm{x}_{1:t} bold_italic_x start_POSTSUBSCRIPT 1 : italic_t end_POSTSUBSCRIPT . Phenaki (Villegas et al., 2023 ) also uses a temporal-aware tokenizer, C-ViViT, but this architecture is compute intensive, as the cost grows quadratically with the number of frames—in comparison, our ST-transformer based tokenizer (ST-ViViT) is much more compute efficient with the dominating factor in its cost increasing linearly with the number of frames.

Refer to caption

Dynamics Model The dynamics model is a decoder-only MaskGIT (Chang et al., 2022 ) transformer ( Figure   6 ). At each time step t ∈ [ 1 , T ] 𝑡 1 𝑇 t\in[1,T] italic_t ∈ [ 1 , italic_T ] , it takes in the tokenized video 𝒛 1 : t − 1 subscript 𝒛 : 1 𝑡 1 \bm{z}_{1:t-1} bold_italic_z start_POSTSUBSCRIPT 1 : italic_t - 1 end_POSTSUBSCRIPT and stopgrad latent actions 𝒂 ~ 1 : t − 1 subscript ~ 𝒂 : 1 𝑡 1 \tilde{\bm{a}}_{1:t-1} over~ start_ARG bold_italic_a end_ARG start_POSTSUBSCRIPT 1 : italic_t - 1 end_POSTSUBSCRIPT and predicts the next frame tokens z ^ t subscript ^ 𝑧 𝑡 \hat{z}_{t} over^ start_ARG italic_z end_ARG start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT . We again utilize an ST-transformer, whose causal structure enables us to use tokens from all ( T − 1 ) 𝑇 1 (T-1) ( italic_T - 1 ) frames 𝒛 1 : T − 1 subscript 𝒛 : 1 𝑇 1 \bm{z}_{1:T-1} bold_italic_z start_POSTSUBSCRIPT 1 : italic_T - 1 end_POSTSUBSCRIPT and latent actions 𝒂 ~ 1 : T − 1 subscript ~ 𝒂 : 1 𝑇 1 \tilde{\bm{a}}_{1:T-1} over~ start_ARG bold_italic_a end_ARG start_POSTSUBSCRIPT 1 : italic_T - 1 end_POSTSUBSCRIPT as input, and generate predictions for all next frames 𝒛 ^ 2 : T subscript ^ 𝒛 : 2 𝑇 \hat{\bm{z}}_{2:T} over^ start_ARG bold_italic_z end_ARG start_POSTSUBSCRIPT 2 : italic_T end_POSTSUBSCRIPT . The model is trained with a cross-entropy loss between the predicted tokens 𝒛 ^ 2 : T subscript ^ 𝒛 : 2 𝑇 \hat{\bm{z}}_{2:T} over^ start_ARG bold_italic_z end_ARG start_POSTSUBSCRIPT 2 : italic_T end_POSTSUBSCRIPT and ground-truth tokens 𝒛 2 : T subscript 𝒛 : 2 𝑇 \bm{z}_{2:T} bold_italic_z start_POSTSUBSCRIPT 2 : italic_T end_POSTSUBSCRIPT . At train time we randomly mask the input tokens 𝒛 2 : T − 1 subscript 𝒛 : 2 𝑇 1 \bm{z}_{2:T-1} bold_italic_z start_POSTSUBSCRIPT 2 : italic_T - 1 end_POSTSUBSCRIPT according to a Bernoulli distribution masking rate sampled uniformly between 0.5 0.5 0.5 0.5 and 1 1 1 1 . Note that a common practice for training world-models, including transformer-based models, is to concatenate the action at time t 𝑡 t italic_t to the corresponding frame (Micheli et al., 2023 ; Robine et al., 2023 ) . However, we found that treating the latent actions as  additive embeddings for both the latent action and dynamics models helped to improve the controllability of the generations.

2.2 Inference: Action-Controllable Video Generation

Refer to caption

We now describe how to use Genie for action-controllable video generation at inference time (see Figure   7 ). A player first prompts the model with an image x 1 subscript 𝑥 1 x_{1} italic_x start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT that serves as the initial frame 1 1 1 The model can be conditioned on a varying number of prompt frames. Here we start from one image as an example. . The image is tokenized using the video encoder, yielding z 1 subscript 𝑧 1 z_{1} italic_z start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT . The player then specifies a discrete latent action a 1 subscript 𝑎 1 a_{1} italic_a start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT to take by choosing any integer value within [ 0 , | A | ) 0 𝐴 [0,|A|) [ 0 , | italic_A | ) . 2 2 2 When first interacting with the model, it is unclear how each latent action will impact the next frame generation. However, we found that the meaning of each action remained consistent across different inputs. Hence, interpreting the mapping of latent actions is akin to learning the buttons on a new controller. The dynamics model takes the frame tokens z 1 subscript 𝑧 1 z_{1} italic_z start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT and corresponding latent action a ~ 1 subscript ~ 𝑎 1 \tilde{a}_{1} over~ start_ARG italic_a end_ARG start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , which is obtained by indexing into the VQ codebook with the discrete input a 1 subscript 𝑎 1 a_{1} italic_a start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , to predict the next frame tokens z 2 subscript 𝑧 2 z_{2} italic_z start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT . This process is repeated to generate the rest of the sequence 𝒛 ^ 2 : T subscript ^ 𝒛 : 2 𝑇 \hat{\bm{z}}_{2:T} over^ start_ARG bold_italic_z end_ARG start_POSTSUBSCRIPT 2 : italic_T end_POSTSUBSCRIPT in an autoregressive manner as actions continue to be passed to the model, while tokens are decoded into video frames 𝒙 ^ 2 : T subscript ^ 𝒙 : 2 𝑇 \hat{\bm{x}}_{2:T} over^ start_ARG bold_italic_x end_ARG start_POSTSUBSCRIPT 2 : italic_T end_POSTSUBSCRIPT with the tokenizer’s decoder. Note that we can regenerate ground truth videos from the dataset by passing the model the starting frame and inferred actions from the video, or generate completely new videos (or trajectories) by changing the actions.

3 3. Experimental Results

Refer to caption

Datasets We train Genie on a large-scale dataset collected from publicly available Internet videos of 2D Platformer games (referred to from here on as “Platformers”). We construct the Platformers dataset by filtering publicly available videos for keywords relating to platformers, yielding 55M 16s video clips at 10FPS, with 160x90 resolution. The final dataset contains 6.8M 16s video clips (30k hours), within an order of magnitude of other popular Internet video datasets (Wang et al., 2023 ; Bain et al., 2021 ) . More details can be found in Section   B.1 . Unless otherwise specified, results are with a 11B-parameter model trained on this dataset.

To verify the generality of our method, we also consider the robotics datasets used to train RT1 Brohan et al. ( 2023 ) , combining their dataset of ∼ 130 ⁢ k similar-to absent 130 𝑘 {\sim}130k ∼ 130 italic_k robot demonstrations with a separate dataset of simulation data and the 209k episodes of real robot data from prior work (Kalashnikov et al., 2018 ) . Note that we do not use actions from any of these datasets, and simply treat them as videos. For simplicity, from here on we refer to this dataset as “Robotics”.

Metrics We examine the video generation performance of Genie via two factors, namely video fidelity , i.e. the quality of video generation, and controllability , i.e. how much impact the latent actions have in video generation. For video fidelity we use the Frechet Video Distance (FVD), a video-level metric, which has been shown to have a high level of alignment to human evaluation on video quality (Unterthiner et al., 2019 ) . For controllability, we devise a metric based on peak signal-to-noise ratio (PSNR) which we call Δ t ⁢ PSNR subscript Δ 𝑡 PSNR \Delta_{t}\text{PSNR} roman_Δ start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT PSNR , that measures how much the video generations differ when conditioned on latent actions inferred from ground-truth ( x ^ t subscript ^ 𝑥 𝑡 \hat{x}_{t} over^ start_ARG italic_x end_ARG start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) vs. sampled from a random distribution ( x ^ t ′ superscript subscript ^ 𝑥 𝑡 ′ \hat{x}_{t}^{\prime} over^ start_ARG italic_x end_ARG start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT ):

where x t subscript 𝑥 𝑡 x_{t} italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT denotes the ground-truth frame at time t 𝑡 t italic_t , x ^ t subscript ^ 𝑥 𝑡 \hat{x}_{t} over^ start_ARG italic_x end_ARG start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT denotes the frame from latent actions 𝒂 ~ 1 : t subscript ~ 𝒂 : 1 𝑡 \tilde{\bm{a}}_{1:t} over~ start_ARG bold_italic_a end_ARG start_POSTSUBSCRIPT 1 : italic_t end_POSTSUBSCRIPT inferred from ground-truth frames, and x ^ t ′ superscript subscript ^ 𝑥 𝑡 ′ \hat{x}_{t}^{\prime} over^ start_ARG italic_x end_ARG start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT the same frame generated from a sequence of latent actions randomly sampled from a categorical distribution. As such, the greater Δ t ⁢ PSNR subscript Δ 𝑡 PSNR \Delta_{t}\text{PSNR} roman_Δ start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT PSNR  is, the more the video generated from random latent actions differs from ground-truth, which indicates a higher level of controllability from the latent actions. For all experiments we report Δ t ⁢ PSNR subscript Δ 𝑡 PSNR \Delta_{t}\text{PSNR} roman_Δ start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT PSNR  with t = 4 𝑡 4 t=4 italic_t = 4 .

Training Details Our video tokenizer uses 200M parameters, a patch size of 4 and a codebook with embedding size 32 and 1024 unique codes, which we found to be the most effective given the trade-off between reconstruction quality of the tokenizer and downstream performance of video prediction. The latent action model has 300M parameters, a patch size of 16, and a codebook with embedding size 32 and 8 unique codes (latent actions). For all modelling components we use a sequence length of 16 frames with an FPS of 10. Further, we employ bfloat16 and QK norm for training our dynamics model, which has been shown to stabilize training at large scale (Henry et al., 2020 ; Dehghani et al., 2023 ) . At inference time, we perform 25 MaskGIT steps for the sampling of each frame with a temperature of 2 using random sampling. See Appendix   C for more details.

3.1 Scaling Results

In this section, we investigate the scaling behavior of our model. To this end, we conduct studies that explore the impact of both model size and batch size. See Appendix   D for more details on architecture and compute usage.

Refer to caption

Scaling Model Size Given a fixed video tokenizer and action model architecture, we train a series of dynamics models ranging from 40M to 2.7B parameters. Figure   8 shows our architecture scales gracefully with model parameters, with each increase in size corresponding to a consistent decrease in the final training loss. This is a strong indication that our approach benefits from scaling, which we exploit with our main Genie model.

Scaling Batch Size We also investigate the effect of scaling the batch size, considering a 2.3B model with batch sizes of 128, 256, and 448, equating to 1.9M, 3.8M and 6.6M tokens. As shown in Figure  8 , increasing the batch size leads to a similarly favorable gain in terms of model performance.

Genie Model It is clear that increasing both model size and batch size helps improve model performance. As a result, for our final model, we train a 10.1B dynamics model with a batch size of 512, for a total of 125k steps, using 256 TPUv5p. When combined with the tokenizer and action model this brings the total to 10.7B parameters, trained on 942B tokens, which we refer to as the Genie model. For our website, we train a larger decoder mapping tokens to 360p videos, adding additional parameters.

3.2 Qualitative Results

We now present qualitative results from the Genie model. We showcase a 11B parameter model trained on the Platformers dataset and a smaller model trained on the Robotics dataset. Our model generates high-quality, controllable videos across diverse domains. Notably, we qualitatively evaluate our Platformers-trained model using only out-of-distribution (OOD) image prompts , including those generated from text-to-image models, hand-drawn sketches, and even realistic photos. The ability to generalize to such significantly OOD inputs underscores the robustness of our approach and the value of training on large-scale data, which would not have been feasible with real actions as input.

Refer to caption

Platformers-trained model Figure   9 showcases examples of our model’s generations prompted from OOD images, including (top row) images generated from Imagen2 (Ho et al., 2022a ; van den Oord et al., ) , (second row) hand-drawn sketches and (bottom row) real-world photos. Genie is able to bring these imagined worlds to life, as we see game-like behaviour when interacting with each example. We showcase more generations by our model in Appendix   A , additionally highlighting the consistency of the latent actions.

Refer to caption

Another emergent capability of our model is its ability to understand 3D scenes and emulate parallax, which is commonly seen in platformer games. In Figure   11 we show an image generated by Imagen2, where taking a latent action moves the foreground at a different rate to the background (as indicated by the length of different colored arrows).

Refer to caption

Robotics-trained model We trained a 2.5B-parameter model on the Robotics dataset using the same hyperparameters found to be best on Platformers, achieving an FVD of 82.7 on the test split. As shown in Figure   12 , this model successfully learns distinct and consistent actions from video data, requiring neither text nor action labels (as in e.g. Yang et al. ( 2023 ) ). Notably, our model learns not only the controls of the robotic arm but also the interactions and deformations of various objects ( Figure   10 ). We believe this shows our approach presents a path to using larger video datasets from the Internet to create a foundational world model for robotics, with low-level controllable simulation that could be used for a variety of applications.

3.3 Training Agents

We believe Genie could one day be used as a foundation world model for training generalist agents. In Figure  13 we show that the model can already be used for generating diverse trajectories in unseen RL environments given starting frames. We further investigate if latent actions learnt from Internet videos can be used for imitating behaviors from unseen videos. We use a frozen LAM to label a sequence of expert videos from a target environment with discrete latent actions and then train a policy that predicts the likelihood of the expert taking a latent action given an observation. We then use a small dataset with expert ground-truth actions for mapping latent to real actions (see Appendix   E for more details).

Refer to caption

We evaluate in both hard and easy settings of a procedurally generated 2D-platformer environment, CoinRun (Cobbe et al., 2020 ) , and compare against an oracle behavioral cloning (BC) model that has access to expert actions as an upper bound, and a random agent as a lower bound (Figure  14 ). The LAM-based policy achieves the same score as the oracle given as few as 200 expert samples to adapt, despite almost certainly never seeing CoinRun before. This provides evidence that the learnt latent actions are consistent and meaningful for transfer, as the mapping from latent to real contains no information about the current observation.

Refer to caption

3.4 Ablation Studies

Design choices for latent action model In designing our latent action model, we carefully considered the type of input to use. While we ultimately chose to use the original images (pixels), we evaluated this choice against the alternative of using tokenized images (replacing x with z in Figure   4 ). We refer to this alternative approach as the “token-input" model (see Table   2 ).

While this model achieved a slightly lower FVD score on the Platformers dataset, it did not maintain this advantage on the Robotics dataset. More importantly, in both environments, the token-input model exhibited worse controllability (as measured by Δ t ⁢ PSNR subscript Δ 𝑡 PSNR \Delta_{t}\text{PSNR} roman_Δ start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT PSNR ). This suggests that some information about video dynamics and movement might have been lost during tokenization, and as a result it is beneficial for the latent action model to take in raw videos as input.

Tokenizer architecture ablations We compare the performance of three choices of tokenizers, including 1) (spatial-only) ViT, 2) (spatial-temporal) ST-ViViT and 3) (spatial-temporal) C-ViViT ( Table   3 ). For comparison we use similar number of parameters for all tokenizers, with patch size 10, batch size 128 and sequence length 16. We then train the same dynamics and latent action model on these three different tokenizers, and report their FVD as well as Δ t ⁢ PSNR subscript Δ 𝑡 PSNR \Delta_{t}\text{PSNR} roman_Δ start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT PSNR .

Our proposed ST-ViViT architecture provides both improved video generation (FVD) and Δ t ⁢ PSNR subscript Δ 𝑡 PSNR \Delta_{t}\text{PSNR} roman_Δ start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT PSNR , for a reasonable trade-off in memory, as compared to to C-ViViT and the spatial-only ViT. This demonstrates its ability to generate videos of high fidelity and controllability, respectively. While C-ViViT employs a full space-time attention mechanism, resulting in significantly higher memory consumption compared to the other two architectures at the same parameter count, this does not translate to improved performance. In fact, C-ViViT exhibits a tendency towards overfitting, necessitating strong regularization during training, which might explain its considerably lower performance.

4 4. Related Work

World models Generative interactive environments can be considered a class of World Models (Ha and Schmidhuber, 2018 ; Oh et al., 2015 ) , which enable next-frame prediction that is conditioned on action inputs (Nunes et al., 2020 ; Hafner et al., 2020 , 2021 ; Micheli et al., 2023 ; Robine et al., 2023 ; Kim et al., 2020 , 2021 ; Bamford and Lucas, 2020 ; Chiappa et al., 2017 ; Pan et al., 2022 ; Eslami et al., 2018 ) . Such models can be useful for training agents, as they can be used for learning policies without direct environment experience at agent training time. However, learning the models themselves typically requires action-conditioned data obtained directly from the environment. In contrast, our approach seeks to learn a world model in an unsupervised fashion from videos alone. Recently, there has been renewed emphasis on scaling world models. GAIA-1 (Hu et al., 2023 ) and UniSim (Yang et al., 2023 ) learn world models for autonomous driving and robotic manipulation respectively. These approaches require both text and action labels, while we focus on training from video-only data from publicly available Internet videos.

Video models Our work is related to video models , which typically condition on initial frames (or text) and predict the remaining frames in a video (Kalchbrenner et al., 2017 ; Clark et al., 2019 ; Finn et al., 2016 ; Luc et al., 2020 ; Lotter et al., 2017 ; Yan et al., 2021 ; Blattmann et al., 2023b ; Walker et al., 2021 ; Le Moing et al., 2021 ; Höppe et al., 2022 ; Singer et al., 2023 ; Ho et al., 2022a , b ; Brooks et al., 2024 ; Yu et al., 2023 ) . Our approach most resembles recent transformer based models such as Phenaki (Villegas et al., 2023 ) , TECO (Yan et al., 2023 ) and MaskViT (Gupta et al., 2023 ) , as we use MaskGIT (Chang et al., 2022 ) and an ST-Transformer (Xu et al., 2020 ) over tokenized images. While video models are becoming increasingly controllable (e.g. (Huang et al., 2022 ) ), we seek a more agentic goal and explicitly learn a latent action space from data, allowing users or agents to “play” the model using latent action-conditioned predictions.

Playable Video Generation Genie generalizes beyond Playable Video Generation (PVG) (Menapace et al., 2021 ) , where latent actions are used for controlling world models learnt directly from videos (Menapace et al., 2021 , 2022 ) . In contrast to Genie, PVG considers domain-specific static examples, rather than generating entirely new environments via prompting. Thus, scaling beyond this setting required non-trivial architectural changes, dropping inductive biases in exchange for a general method.

Environment generation Our work is also related to Procedural Content Generation   (PCG, e.g. Risi and Togelius, 2020a , b ) where machine learning has proven highly effective for generating game levels (Summerville et al., 2018 ) , recently via language models that directly write game code (Sudhakaran et al., 2023 ; Todd et al., 2023 ) . Language models themselves can also be considered to be interactive environments (Wong et al., 2023 ) , albeit lacking a visual component. By contrast in our setting the levels can be learnt and generated directly from pixels, which enables us to utilize the diversity of Internet video data.

Training agents with latent actions Prior works have used latent actions for imitation from observation (Edwards et al., 2019 ) , planning (Rybkin* et al., 2019 ) and pre-training RL agents (Ye et al., 2022 ; Schmidt and Jiang, 2024 ) . These approaches have similar objectives to our latent action model, though have not been applied at scale. VPT  (Baker et al., 2022 ) is a recent approach that uses an inverse dynamics model learnt from human-provided action labeled data, to label Internet-scale videos with actions that can then be used for training a policy. We showed, in contrast, that we can use latent actions learnt from Internet videos to infer policies for arbitrary environments, avoiding the need for ground-truth actions that are costly and may not generalize.

5 5. Conclusion and Future Work

We proposed Genie, a new form of generative AI that enables anyone, even children, to dream up, create, and step into generated worlds as we can with human-designed simulated environments. Genie can be prompted to generate a diverse set of interactive and controllable environments despite training from video-only data.

There are clear improvements that can be made to the model. Genie inherits some of the weaknesses of other autoregressive transformer models, and can hallucinate unrealistic futures. And while we have made progress with spatiotemporal representations, we are still limited to 16 frames of memory which makes it challenging to get consistent environments over long horizons. Finally, Genie currently operates around 1FPS and requires future advances to achieve an efficient frame rate for interaction.

Still, we believe Genie opens up vast potential for future research. Given its generality, the model could be trained from an even larger proportion of Internet videos to simulate diverse, realistic, and imagined environments. Furthermore, we only briefly touched upon the capabilities of using Genie for training agents, but given that the lack of rich and diverse environments is one of the key limitations in RL, we could unlock new paths to creating more generally capable agents.

Broader Impact

Societal Impact Genie could enable a large amount of people to generate their own game-like experiences. This could be positive for those who wish to express their creativity in a new way, for example children who could design and step into their own imagined worlds. We also recognize that with significant advances, it will be critical to explore the possibilities of using this technology to amplify existing human game generation and creativity—and empowering relevant industries to utilize Genie to enable their next generation of playable world development.

Training Data and Weights : We have chosen not to release the trained model checkpoints, the model’s training dataset, or examples from that data to accompany this paper or the website. We would like to have the opportunity to further engage with the research (and video game) community and to ensure that any future such releases are respectful, safe and responsible.

Reproducibility : We understand that it may be challenging for researchers with fewer computational to reproduce our main results. In order to mitigate this issue, we describe a smaller scale, fully reproducible example in Appendix   F that can run on a single mid-range TPU (or GPU). Given that many design choices translate between the two settings, we believe this will make it possible for the broader community to investigate future architectural improvements as well as additional research directions resulting from our work.

Acknowledgements

We thank Mateusz Malinowski, Philip Ball and Louis Kirsch for reviewing a draft of our paper; Cassidy Hardin, David Bridson, Eric Lau, Lars Lowe Sjoesund, Lucas Smaira and Bernardo Avila Pires for help with our Platformers dataset; Ruben Villegas for valuable discussions on our video model training and evaluation; and Adrian Bolton, Rushil Mistry, Hannah Openshaw, Zoubin Ghahramani, Raia Hadsell, Koray Kavukcuoglu, Daan Wierstra, Doina Precup and Ed Hirst for strategic advice and guidance. We make use of the DeepMind Jax ecosystem (Babuschkin et al., 2010 ) and specifically thank Andy Brock for building the internal framework we used for our model training and Arthur Brussee who provided an initial interface that enabled us to “play” our models. Finally, thank you to Seneca and Caspian Clune for their creative sketches, potentially making them the youngest ever game designers.

Author Contributions

We list authors alphabetically by last name. Please direct all correspondence to Ashley Edwards ( [email protected] ) and Jack Parker-Holder ( [email protected] ).

Core Contributors

Jake Bruce : project leadership, video tokenizer research, action model research, dynamics model research, scaling, model demo, infrastructure

Michael Dennis : dynamics model research, scaling, metrics, model demo, infrastructure

Ashley Edwards : genie concept, project leadership, action model research, agent training, model demo

Edward Hughes : dynamics model research, infrastructure

Matthew Lai : dataset curation, infrastructure

Aditi Mavalankar : action model research, metrics, agent training

Jack Parker-Holder : genie concept, project leadership, dynamics model research, scaling, dataset curation

Yuge (Jimmy) Shi : video tokenizer research, dynamics model research, dataset curation, metrics

Richie Steigerwald : dataset curation, metrics

Partial Contributors and Advisors

Chris Apps : project management

Yusuf Aytar : technical advice

Sarah Bechtle : technical advice

Feryal Behbahani : strategic advice

Stephanie Chan : technical advice

Jeff Clune : technical advice, strategic advice

Lucy Gonzalez : project management

Nicolas Heess : strategic advice

Simon Osindero : technical advice

Sherjil Ozair : technical advice

Scott Reed : technical advice

Jingwei Zhang : technical advice

Konrad Zolna : scaling, technical advice

Nando de Freitas : strategic advice

Tim Rocktäschel : genie concept, project leadership

Satinder Singh : strategic advice

  • Babuschkin et al. (2010) I. Babuschkin, K. Baumli, A. Bell, S. Bhupatiraju, J. Bruce, P. Buchlovsky, D. Budden, T. Cai, A. Clark, I. Danihelka, et al. The deepmind jax ecosystem, 2020. URL http://github. com/deepmind , 2010.
  • Bain et al. (2021) M. Bain, A. Nagrani, G. Varol, and A. Zisserman. Frozen in time: A joint video and image encoder for end-to-end retrieval. In 2021 IEEE/CVF International Conference on Computer Vision (ICCV) , pages 1708–1718, Los Alamitos, CA, USA, oct 2021. IEEE Computer Society. 10.1109/ICCV48922.2021.00175 .
  • Baker et al. (2022) B. Baker, I. Akkaya, P. Zhokov, J. Huizinga, J. Tang, A. Ecoffet, B. Houghton, R. Sampedro, and J. Clune. Video pretraining (vpt): Learning to act by watching unlabeled online videos. Advances in Neural Information Processing Systems , 35:24639–24654, 2022.
  • Bamford and Lucas (2020) C. Bamford and S. M. Lucas. Neural game engine: Accurate learning ofgeneralizable forward models from pixels. In Conference on Games , 2020.
  • Bauer et al. (2023) J. Bauer, K. Baumli, F. Behbahani, A. Bhoopchand, N. Bradley-Schmieg, M. Chang, N. Clay, A. Collister, V. Dasagi, L. Gonzalez, K. Gregor, E. Hughes, S. Kashem, M. Loks-Thompson, H. Openshaw, J. Parker-Holder, S. Pathak, N. Perez-Nieves, N. Rakicevic, T. Rocktäschel, Y. Schroecker, S. Singh, J. Sygnowski, K. Tuyls, S. York, A. Zacherl, and L. M. Zhang. Human-timescale adaptation in an open-ended task space. In A. Krause, E. Brunskill, K. Cho, B. Engelhardt, S. Sabato, and J. Scarlett, editors, Proceedings of the 40th International Conference on Machine Learning , volume 202 of Proceedings of Machine Learning Research , pages 1887–1935. PMLR, 23–29 Jul 2023.
  • Blattmann et al. (2023a) A. Blattmann, T. Dockhorn, S. Kulal, D. Mendelevitch, M. Kilian, D. Lorenz, Y. Levi, Z. English, V. Voleti, A. Letts, V. Jampani, and R. Rombach. Stable video diffusion: Scaling latent video diffusion models to large datasets, 2023a.
  • Blattmann et al. (2023b) A. Blattmann, R. Rombach, H. Ling, T. Dockhorn, S. W. Kim, S. Fidler, and K. Kreis. Align your latents: High-resolution video synthesis with latent diffusion models. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages 22563–22575, 2023b.
  • Brohan et al. (2023) A. Brohan, N. Brown, J. Carbajal, Y. Chebotar, J. Dabis, C. Finn, K. Gopalakrishnan, K. Hausman, A. Herzog, J. Hsu, J. Ibarz, B. Ichter, A. Irpan, T. Jackson, S. Jesmonth, N. J. Joshi, R. Julian, D. Kalashnikov, Y. Kuang, I. Leal, K.-H. Lee, S. Levine, Y. Lu, U. Malla, D. Manjunath, I. Mordatch, O. Nachum, C. Parada, J. Peralta, E. Perez, K. Pertsch, J. Quiambao, K. Rao, M. Ryoo, G. Salazar, P. Sanketi, K. Sayed, J. Singh, S. Sontakke, A. Stone, C. Tan, H. Tran, V. Vanhoucke, S. Vega, Q. Vuong, F. Xia, T. Xiao, P. Xu, S. Xu, T. Yu, and B. Zitkovich. Rt-1: Robotics transformer for real-world control at scale. In Robotics: Science and Systems , 2023.
  • Brooks et al. (2024) T. Brooks, B. Peebles, C. Homes, W. DePue, Y. Guo, L. Jing, D. Schnurr, J. Taylor, T. Luhman, E. Luhman, C. W. Y. Ng, R. Wang, and A. Ramesh. Video generation models as world simulators. 2024. URL https://openai.com/research/video-generation-models-as-world-simulators .
  • Brown et al. (2020) T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askell, et al. Language models are few-shot learners. Advances in neural information processing systems , 33:1877–1901, 2020.
  • Chang et al. (2022) H. Chang, H. Zhang, L. Jiang, C. Liu, and W. T. Freeman. Maskgit: Masked generative image transformer. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages 11315–11325, June 2022.
  • Chiappa et al. (2017) S. Chiappa, S. Racaniere, D. Wierstra, and S. Mohamed. Recurrent environment simulators. In International Conference on Learning Representations , 2017.
  • Clark et al. (2019) A. Clark, J. Donahue, and K. Simonyan. Efficient video generation on complex datasets. CoRR , abs/1907.06571, 2019. URL http://arxiv.org/abs/1907.06571 .
  • Clune (2019) J. Clune. Ai-gas: Ai-generating algorithms, an alternate paradigm for producing general artificial intelligence. arXiv preprint arXiv:1905.10985 , 2019.
  • Cobbe et al. (2020) K. Cobbe, C. Hesse, J. Hilton, and J. Schulman. Leveraging procedural generation to benchmark reinforcement learning. In Proceedings of the 37th International Conference on Machine Learning , pages 2048–2056, 2020.
  • Dehghani et al. (2023) M. Dehghani, J. Djolonga, B. Mustafa, P. Padlewski, J. Heek, J. Gilmer, A. P. Steiner, M. Caron, R. Geirhos, I. Alabdulmohsin, R. Jenatton, L. Beyer, M. Tschannen, A. Arnab, X. Wang, C. Riquelme Ruiz, M. Minderer, J. Puigcerver, U. Evci, M. Kumar, S. V. Steenkiste, G. F. Elsayed, A. Mahendran, F. Yu, A. Oliver, F. Huot, J. Bastings, M. Collier, A. A. Gritsenko, V. Birodkar, C. N. Vasconcelos, Y. Tay, T. Mensink, A. Kolesnikov, F. Pavetic, D. Tran, T. Kipf, M. Lucic, X. Zhai, D. Keysers, J. J. Harmsen, and N. Houlsby. Scaling vision transformers to 22 billion parameters. In A. Krause, E. Brunskill, K. Cho, B. Engelhardt, S. Sabato, and J. Scarlett, editors, Proceedings of the 40th International Conference on Machine Learning , volume 202 of Proceedings of Machine Learning Research , pages 7480–7512. PMLR, 23–29 Jul 2023.
  • Dosovitskiy et al. (2021) A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale. In International Conference on Learning Representations , 2021. URL https://openreview.net/forum?id=YicbFdNTTy .
  • Edwards et al. (2019) A. Edwards, H. Sahni, Y. Schroecker, and C. Isbell. Imitating latent policies from observation. In International conference on machine learning , pages 1755–1763. PMLR, 2019.
  • Eslami et al. (2018) S. M. A. Eslami, D. J. Rezende, F. Besse, F. Viola, A. S. Morcos, M. Garnelo, A. Ruderman, A. A. Rusu, I. Danihelka, K. Gregor, D. P. Reichert, L. Buesing, T. Weber, O. Vinyals, D. Rosenbaum, N. Rabinowitz, H. King, C. Hillier, M. Botvinick, D. Wierstra, K. Kavukcuoglu, and D. Hassabis. Neural scene representation and rendering. Science , 360(6394):1204–1210, 2018. 10.1126/science.aar6170 .
  • Esser et al. (2023) P. Esser, J. Chiu, P. Atighehchian, J. Granskog, and A. Germanidis. Structure and content-guided video synthesis with diffusion models. In 2023 IEEE/CVF International Conference on Computer Vision (ICCV) , 2023.
  • Finn et al. (2016) C. Finn, I. Goodfellow, and S. Levine. Unsupervised learning for physical interaction through video prediction. In Proceedings of the 30th International Conference on Neural Information Processing Systems , NIPS’16, page 64–72, Red Hook, NY, USA, 2016. Curran Associates Inc. ISBN 9781510838819.
  • Gupta et al. (2023) A. Gupta, S. Tian, Y. Zhang, J. Wu, R. Martín-Martín, and L. Fei-Fei. Maskvit: Masked visual pre-training for video prediction. In The Eleventh International Conference on Learning Representations , 2023.
  • Ha and Schmidhuber (2018) D. Ha and J. Schmidhuber. Recurrent world models facilitate policy evolution. In Proceedings of the 32Nd International Conference on Neural Information Processing Systems , NeurIPS’18, pages 2455–2467, 2018.
  • Hafner et al. (2020) D. Hafner, T. Lillicrap, J. Ba, and M. Norouzi. Dream to control: Learning behaviors by latent imagination. In International Conference on Learning Representations , 2020.
  • Hafner et al. (2021) D. Hafner, T. P. Lillicrap, M. Norouzi, and J. Ba. Mastering atari with discrete world models. In International Conference on Learning Representations , 2021.
  • He et al. (2016) K. He, X. Zhang, S. Ren, and J. Sun. Deep residual learning for image recognition. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , pages 770–778, 2016. 10.1109/CVPR.2016.90 .
  • Henry et al. (2020) A. Henry, P. R. Dachapally, S. S. Pawar, and Y. Chen. Query-key normalization for transformers. In Findings of the Association for Computational Linguistics: EMNLP 2020 , pages 4246–4253, Online, Nov. 2020. Association for Computational Linguistics. 10.18653/v1/2020.findings-emnlp.379 .
  • Ho et al. (2022a) J. Ho, W. Chan, C. Saharia, J. Whang, R. Gao, A. Gritsenko, D. P. Kingma, B. Poole, M. Norouzi, D. J. Fleet, and T. Salimans. Imagen video: High definition video generation with diffusion models, 2022a.
  • Ho et al. (2022b) J. Ho, T. Salimans, A. Gritsenko, W. Chan, M. Norouzi, and D. J. Fleet. Video diffusion models. In S. Koyejo, S. Mohamed, A. Agarwal, D. Belgrave, K. Cho, and A. Oh, editors, Advances in Neural Information Processing Systems , volume 35, pages 8633–8646. Curran Associates, Inc., 2022b.
  • Hong et al. (2022) W. Hong, M. Ding, W. Zheng, X. Liu, and J. Tang. Cogvideo: Large-scale pretraining for text-to-video generation via transformers. arXiv preprint arXiv:2205.15868 , 2022.
  • Hong et al. (2023) W. Hong, M. Ding, W. Zheng, X. Liu, and J. Tang. Cogvideo: Large-scale pretraining for text-to-video generation via transformers. In The Eleventh International Conference on Learning Representations , 2023. URL https://openreview.net/forum?id=rB6TpjAuSRy .
  • Höppe et al. (2022) T. Höppe, A. Mehrjou, S. Bauer, D. Nielsen, and A. Dittadi. Diffusion models for video prediction and infilling. Transactions on Machine Learning Research , 2022. ISSN 2835-8856.
  • Hu et al. (2023) A. Hu, L. Russell, H. Yeo, Z. Murez, G. Fedoseev, A. Kendall, J. Shotton, and G. Corrado. Gaia-1: A generative world model for autonomous driving, 2023.
  • Huang et al. (2022) J. Huang, Y. Jin, K. M. Yi, and L. Sigal. Layered controllable video generation. In Computer Vision – ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23–27, 2022, Proceedings, Part XVI , page 546–564, Berlin, Heidelberg, 2022. Springer-Verlag. ISBN 978-3-031-19786-4.
  • Jouppi et al. (2020) N. P. Jouppi, D. H. Yoon, G. Kurian, S. Li, N. Patil, J. Laudon, C. Young, and D. Patterson. A domain-specific supercomputer for training deep neural networks. Communications of the ACM , 63(7):67–78, 2020.
  • Kalashnikov et al. (2018) D. Kalashnikov, A. Irpan, P. Pastor, J. Ibarz, A. Herzog, E. Jang, D. Quillen, E. Holly, M. Kalakrishnan, V. Vanhoucke, et al. Qt-opt: Scalable deep reinforcement learning for vision-based robotic manipulation. arXiv preprint arXiv:1806.10293 , 2018.
  • Kalchbrenner et al. (2017) N. Kalchbrenner, A. van den Oord, K. Simonyan, I. Danihelka, O. Vinyals, A. Graves, and K. Kavukcuoglu. Video pixel networks. In D. Precup and Y. W. Teh, editors, Proceedings of the 34th International Conference on Machine Learning , volume 70 of Proceedings of Machine Learning Research , pages 1771–1779. PMLR, 06–11 Aug 2017. URL https://proceedings.mlr.press/v70/kalchbrenner17a.html .
  • Kapturowski et al. (2018) S. Kapturowski, G. Ostrovski, J. Quan, R. Munos, and W. Dabney. Recurrent experience replay in distributed reinforcement learning. In International conference on learning representations , 2018.
  • Kim et al. (2020) S. W. Kim, Y. Zhou, J. Philion, A. Torralba, and S. Fidler. Learning to simulate dynamic environments with gamegan. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , June 2020.
  • Kim et al. (2021) S. W. Kim, J. Philion, A. Torralba, and S. Fidler. Drivegan: Towards a controllable high-quality neural simulation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages 5820–5829, June 2021.
  • Le Moing et al. (2021) G. Le Moing, J. Ponce, and C. Schmid. Ccvs: Context-aware controllable video synthesis. In M. Ranzato, A. Beygelzimer, Y. Dauphin, P. Liang, and J. W. Vaughan, editors, Advances in Neural Information Processing Systems , volume 34, pages 14042–14055. Curran Associates, Inc., 2021.
  • Lotter et al. (2017) W. Lotter, G. Kreiman, and D. Cox. Deep predictive coding networks for video prediction and unsupervised learning. In International Conference on Learning Representations , 2017. URL https://openreview.net/forum?id=B1ewdt9xe .
  • Luc et al. (2020) P. Luc, A. Clark, S. Dieleman, D. de Las Casas, Y. Doron, A. Cassirer, and K. Simonyan. Transformation-based adversarial video prediction on large-scale data. CoRR , abs/2003.04035, 2020.
  • Menapace et al. (2021) W. Menapace, S. Lathuilière, S. Tulyakov, A. Siarohin, and E. Ricci. Playable video generation. In IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2021, virtual, June 19-25, 2021 , pages 10061–10070. Computer Vision Foundation / IEEE, 2021.
  • Menapace et al. (2022) W. Menapace, S. Lathuilière, A. Siarohin, C. Theobalt, S. Tulyakov, V. Golyanik, and E. Ricci. Playable environments: Video manipulation in space and time. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022.
  • Micheli et al. (2023) V. Micheli, E. Alonso, and F. Fleuret. Transformers are sample-efficient world models. In The Eleventh International Conference on Learning Representations , 2023.
  • Nunes et al. (2020) M. S. Nunes, A. Dehban, P. Moreno, and J. Santos-Victor. Action-conditioned benchmarking of robotic video prediction models: a comparative study. In 2020 IEEE International Conference on Robotics and Automation (ICRA) , pages 8316–8322, 2020. 10.1109/ICRA40945.2020.9196839 .
  • Oh et al. (2015) J. Oh, X. Guo, H. Lee, R. Lewis, and S. Singh. Action-conditional video prediction using deep networks in atari games. In Proceedings of the 28th International Conference on Neural Information Processing Systems - Volume 2 , NIPS’15, page 2863–2871, Cambridge, MA, USA, 2015. MIT Press.
  • Open Ended Learning Team et al. (2021) Open Ended Learning Team, A. Stooke, A. Mahajan, C. Barros, C. Deck, J. Bauer, J. Sygnowski, M. Trebacz, M. Jaderberg, M. Mathieu, N. McAleese, N. Bradley-Schmieg, N. Wong, N. Porcel, R. Raileanu, S. Hughes-Fitt, V. Dalibard, and W. M. Czarnecki. Open-ended learning leads to generally capable agents. CoRR , abs/2107.12808, 2021.
  • Oquab et al. (2023) M. Oquab, T. Darcet, T. Moutakanni, H. Vo, M. Szafraniec, V. Khalidov, P. Fernandez, D. Haziza, F. Massa, A. El-Nouby, et al. Dinov2: Learning robust visual features without supervision. arXiv preprint arXiv:2304.07193 , 2023.
  • Pan et al. (2022) M. Pan, X. Zhu, Y. Wang, and X. Yang. Iso-dream: Isolating and leveraging noncontrollable visual dynamics in world models. In S. Koyejo, S. Mohamed, A. Agarwal, D. Belgrave, K. Cho, and A. Oh, editors, Advances in Neural Information Processing Systems , volume 35, pages 23178–23191. Curran Associates, Inc., 2022.
  • Radford et al. (2018) A. Radford, K. Narasimhan, T. Salimans, and I. Sutskever. Improving language understanding by generative pre-training. 2018.
  • Radford et al. (2019) A. Radford, J. Wu, R. Child, D. Luan, D. Amodei, I. Sutskever, et al. Language models are unsupervised multitask learners. OpenAI blog , 1(8):9, 2019.
  • Rajbhandari et al. (2020) S. Rajbhandari, J. Rasley, O. Ruwase, and Y. He. Zero: Memory optimizations toward training trillion parameter models. In SC20: International Conference for High Performance Computing, Networking, Storage and Analysis , pages 1–16. IEEE, 2020.
  • Ramesh et al. (2021) A. Ramesh, M. Pavlov, G. Goh, S. Gray, C. Voss, A. Radford, M. Chen, and I. Sutskever. Zero-shot text-to-image generation. In M. Meila and T. Zhang, editors, Proceedings of the 38th International Conference on Machine Learning , volume 139 of Proceedings of Machine Learning Research , pages 8821–8831. PMLR, 18–24 Jul 2021.
  • Ramesh et al. (2022) A. Ramesh, P. Dhariwal, A. Nichol, C. Chu, and M. Chen. Hierarchical text-conditional image generation with clip latents, 2022.
  • Reed et al. (2022) S. Reed, K. Zolna, E. Parisotto, S. G. Colmenarejo, A. Novikov, G. Barth-maron, M. Giménez, Y. Sulsky, J. Kay, J. T. Springenberg, T. Eccles, J. Bruce, A. Razavi, A. Edwards, N. Heess, Y. Chen, R. Hadsell, O. Vinyals, M. Bordbar, and N. de Freitas. A generalist agent. Transactions on Machine Learning Research , 2022. ISSN 2835-8856. Featured Certification, Outstanding Certification.
  • Risi and Togelius (2020a) S. Risi and J. Togelius. Increasing generality in machine learning through procedural content generation. Nature Machine Intelligence , 2, 08 2020a. 10.1038/s42256-020-0208-z .
  • Risi and Togelius (2020b) S. Risi and J. Togelius. Procedural content generation: From automatically generating game levels to increasing generality in machine learning. Nature , 2020b.
  • Robine et al. (2023) J. Robine, M. Höftmann, T. Uelwer, and S. Harmeling. Transformer-based world models are happy with 100k interactions. In The Eleventh International Conference on Learning Representations , 2023.
  • Rombach et al. (2022) R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages 10684–10695, June 2022.
  • Rybkin* et al. (2019) O. Rybkin*, K. Pertsch*, K. G. Derpanis, K. Daniilidis, and A. Jaegle. Learning what you can do before doing anything. In International Conference on Learning Representations , 2019.
  • Saharia et al. (2022) C. Saharia, W. Chan, S. Saxena, L. Li, J. Whang, E. Denton, S. K. S. Ghasemipour, R. Gontijo-Lopes, B. K. Ayan, T. Salimans, J. Ho, D. J. Fleet, and M. Norouzi. Photorealistic text-to-image diffusion models with deep language understanding. In A. H. Oh, A. Agarwal, D. Belgrave, and K. Cho, editors, Advances in Neural Information Processing Systems , 2022.
  • Schmidt and Jiang (2024) D. Schmidt and M. Jiang. Learning to act without actions. In The Twelfth International Conference on Learning Representations , 2024.
  • Shoeybi et al. (2019) M. Shoeybi, M. Patwary, R. Puri, P. LeGresley, J. Casper, and B. Catanzaro. Megatron-lm: Training multi-billion parameter language models using model parallelism. CoRR , abs/1909.08053, 2019. URL http://arxiv.org/abs/1909.08053 .
  • Singer et al. (2023) U. Singer, A. Polyak, T. Hayes, X. Yin, J. An, S. Zhang, Q. Hu, H. Yang, O. Ashual, O. Gafni, D. Parikh, S. Gupta, and Y. Taigman. Make-a-video: Text-to-video generation without text-video data. In The Eleventh International Conference on Learning Representations , 2023.
  • Sudhakaran et al. (2023) S. Sudhakaran, M. González-Duque, C. Glanois, M. Freiberger, E. Najarro, and S. Risi. Prompt-guided level generation. In Proceedings of the Companion Conference on Genetic and Evolutionary Computation , pages 179–182, 2023.
  • Summerville et al. (2018) A. Summerville, S. Snodgrass, M. Guzdial, C. Holmgård, A. K. Hoover, A. Isaksen, A. Nealen, and J. Togelius. Procedural content generation via machine learning (PCGML). IEEE Trans. Games , 10(3):257–270, 2018.
  • Todd et al. (2023) G. Todd, S. Earle, M. U. Nasir, M. C. Green, and J. Togelius. Level generation through large language models. In Proceedings of the 18th International Conference on the Foundations of Digital Games , pages 1–8, 2023.
  • Torabi et al. (2018) F. Torabi, G. Warnell, and P. Stone. Behavioral cloning from observation. arXiv preprint arXiv:1805.01954 , 2018.
  • Unterthiner et al. (2019) T. Unterthiner, S. van Steenkiste, K. Kurach, R. Marinier, M. Michalski, and S. Gelly. FVD: A new metric for video generation, 2019.
  • (72) A. van den Oord, A. Razavi, B. Uria, Çağlar Ünlü, C. Nash, C. Wolff, C. Durkan, D. Ding, D. Górny, E. Gladchenko, F. Riedel, H. Qi, J. Kelly, J. Bauer, J. Donahue, J. Zhang, M. Malinowski, M. Bińkowski, P. Luc, R. Riachi, R. Strudel, T. P. I. Sander Dieleman, Y. Ganin, and Z. Eaton-Rosen. Imagen 2. URL https://deepmind.google/technologies/imagen-2/ .
  • van den Oord et al. (2017) A. van den Oord, O. Vinyals, and K. Kavukcuoglu. Neural discrete representation learning. In Proceedings of the 31st International Conference on Neural Information Processing Systems , NIPS’17, page 6309–6318, Red Hook, NY, USA, 2017. Curran Associates Inc. ISBN 9781510860964.
  • Vaswani et al. (2017) A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin. Attention is all you need. In Advances in Neural Information Processing Systems , pages 5998–6008, 2017.
  • Villegas et al. (2023) R. Villegas, M. Babaeizadeh, P.-J. Kindermans, H. Moraldo, H. Zhang, M. T. Saffar, S. Castro, J. Kunze, and D. Erhan. Phenaki: Variable length video generation from open domain textual descriptions. In International Conference on Learning Representations , 2023.
  • Walker et al. (2021) J. C. Walker, A. Razavi, and A. van den Oord. Predicting video with VQVAE, 2021.
  • Wang et al. (2023) Y. Wang, Y. He, Y. Li, K. Li, J. Yu, X. Ma, X. Chen, Y. Wang, P. Luo, Z. Liu, Y. Wang, L. Wang, and Y. Qiao. Internvid: A large-scale video-text dataset for multimodal understanding and generation, 2023.
  • Wong et al. (2023) L. Wong, G. Grand, A. K. Lew, N. D. Goodman, V. K. Mansinghka, J. Andreas, and J. B. Tenenbaum. From word models to world models: Translating from natural language to the probabilistic language of thought, 2023.
  • Wu et al. (2022) C. Wu, J. Liang, L. Ji, F. Yang, Y. Fang, D. Jiang, and N. Duan. Nüwa: Visual synthesis pre-training for neural visual world creation. In European conference on computer vision , pages 720–736. Springer, 2022.
  • Xu et al. (2020) M. Xu, W. Dai, C. Liu, X. Gao, W. Lin, G.-J. Qi, and H. Xiong. Spatial-temporal transformer networks for traffic flow forecasting. arXiv preprint arXiv:2001.02908 , 2020.
  • Yan et al. (2021) W. Yan, Y. Zhang, P. Abbeel, and A. Srinivas. Videogpt: Video generation using vq-vae and transformers, 2021.
  • Yan et al. (2023) W. Yan, D. Hafner, S. James, and P. Abbeel. Temporally consistent transformers for video generation. In A. Krause, E. Brunskill, K. Cho, B. Engelhardt, S. Sabato, and J. Scarlett, editors, Proceedings of the 40th International Conference on Machine Learning , volume 202 of Proceedings of Machine Learning Research , pages 39062–39098. PMLR, 23–29 Jul 2023.
  • Yang et al. (2023) M. Yang, Y. Du, K. Ghasemipour, J. Tompson, D. Schuurmans, and P. Abbeel. Learning interactive real-world simulators. arXiv preprint arXiv:2310.06114 , 2023.
  • Ye et al. (2022) W. Ye, Y. Zhang, P. Abbeel, and Y. Gao. Become a proficient player with limited data through watching pure videos. In The Eleventh International Conference on Learning Representations , 2022.
  • Yu et al. (2023) L. Yu, Y. Cheng, K. Sohn, J. Lezama, H. Zhang, H. Chang, A. G. Hauptmann, M. Yang, Y. Hao, I. Essa, and L. Jiang. Magvit: Masked generative video transformer. In 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages 10459–10469, Los Alamitos, CA, USA, jun 2023. IEEE Computer Society. 10.1109/CVPR52729.2023.01008 .

Appendix A Additional Example Trajectories

Refer to caption

Appendix B Dataset

B.1 platformers dataset, initial dataset.

We generated a dataset by filtering publicly available Internet videos, using the following criteria:

The title contains keywords relating to 2D platformer games.

The title or description must contain an action word, such as “speedrun” or “playthrough”.

The title must not contain negating words such as “movie” or “unboxing”.

We then split each video into 16s clips at 10 FPS, which corresponds to 160 frames per clip. Our resulting dataset contains 55M videos, which totals around 244k hours. When selecting keywords, we manually spot checked results to check that they typically produced 2D platformer gameplay videos which are not outnumbered by other sorts of videos which happen to share similar keywords.

Filter Pipeline

We noticed that many of the videos in the dataset were of poor quality, impacting our model performance. We propose a scalable approach to systematically filter the data, using a learned classifier as in Baker et al. ( 2022 ) . First, we define high quality videos as those that display clear gameplay and do not contain distractor items such as menu screen or streamer faces. We then filter this data as follows:

Our team hand labelled 10k videos, with roughly ten hours of total human effort. The labels ranged from 5 (best) to 1 (worst) quality.

We trained a 11M parameter ResNet18 (He et al., 2016 ) with binary classification where we deleted all entries rated 2-4 and classified 5 as good and 1 as bad.

We then apply a decision rule based on model prediction and confidence to determine whether to keep the video.

Consistent to findings in prior work Baker et al. ( 2022 ); Oquab et al. ( 2023 ) , having high quality data outweighs the quantity of data – even though the curated datasaet is only just over 10% the size of the original dataset, the model trained on the curated dataset outperforms in terms of FVD, see Table   4 . Our final dataset is 6.8M videos for a total of over 30k hours.

Appendix C Training details

C.1 latent action model training.

We found a benefit from increasing the number of codes (i.e. number of actions), at the cost of reduced playability for human and AI agents.

Note that the model inputs are normalized between 0 0 and 1 1 1 1 and the final outputs of the decoder are placed through a sigmoid.

C.2 Video Tokenizer Training

Here we describe our video tokenizer training. We found it more effective to scale our decoder than the encoder, and a marginal gain from increasing batch size (see Table  6 ).

We train our video tokenizer for 300k steps using the AdamW optimizer, with cosine decay, using the hyperparameters in Table  8 .

C.3 Dynamics Model Training

Appendix d scaling experiments details.

In this section we provide more details on the architecture as well as compute budget for the scaling experiments.

Scaling model size

For all models we use a batch size of 256. We train all models for 200k steps, thus use a total of 750B training tokens for each run. All runs make use of batch parallelism and stage-3 ZeRO sharding (Rajbhandari et al., 2020 ) , while our larger models also make use of tensor parallelism (Shoeybi et al., 2019 ) . For this experiment we make use of TPUv2 and TPUv3 (Jouppi et al., 2020 ) . See Table  10 for more details.

Scaling batch size

All models use the same architecture with 2.3B parameters, as shown in Table  11 , and train for 200k steps. The only difference between the three runs is hardware—the 128, 256 and 448 batch size models train on 64 TPUv3, 128 TPUv3 and 64 TPUv5p respectively.

Genie Model

The parameter count, model architecture as well as compute usage of the dynamics model for the final Genie model is listed in Table   12 . We train a 10.1B dynamics model with a batch size of 512, for a total of 125k steps using 256 TPUv5.

Appendix E Behavioral Cloning Details

In this section we provide more details about our behavioral cloning experiments. We train within the Procgen CoinRun environment  (Cobbe et al., 2020 ) and evaluate in a held out test set. We assume we have a dataset of expert sequences in this environment from an agent trained with R2D2  (Kapturowski et al., 2018 ) . We then train an agent to imitate from this data. Notably, the oracle agent has access to the corresponding ground-truth expert actions. We now discuss how we can utilize a pre-trained LAM to infer the actions taken.

E.1 Genie LAM

𝑡 1 a_{t}\leftarrow LAM(x_{t},x_{t+1}) italic_a start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ← italic_L italic_A italic_M ( italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT , italic_x start_POSTSUBSCRIPT italic_t + 1 end_POSTSUBSCRIPT ) . We then train a policy π ⁢ ( a t | x t ) 𝜋 conditional subscript 𝑎 𝑡 subscript 𝑥 𝑡 \pi(a_{t}|x_{t}) italic_π ( italic_a start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT | italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) to predict the likelihood of the expert taking latent action a t subscript 𝑎 𝑡 a_{t} italic_a start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT given observation x t subscript 𝑥 𝑡 x_{t} italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT . Note that this procedure is similar to prior works that learn from videos (Torabi et al., 2018 ; Baker et al., 2022 ) . However, these approaches use ground-truth actions for labeling videos whereas we utilize latent actions learnt completely offline.

𝑡 1 \langle x_{t},u_{t},x_{t+1}\rangle ⟨ italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT , italic_u start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT , italic_x start_POSTSUBSCRIPT italic_t + 1 end_POSTSUBSCRIPT ⟩ (we denote u t subscript 𝑢 𝑡 u_{t} italic_u start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT for ground-truth actions to avoid confusion with predicted latent actions), we use the LAM to obtain a latent action a t subscript 𝑎 𝑡 a_{t} italic_a start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT and fill a dictionary D 𝐷 D italic_D consisting of mapped latents to a list of corresponding real actions. In summary, given an observation x t subscript 𝑥 𝑡 x_{t} italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT from the environment, we can obtain the most likely latent action as a t ∼ π ⁢ ( s t ) similar-to subscript 𝑎 𝑡 𝜋 subscript 𝑠 𝑡 a_{t}\sim\pi(s_{t}) italic_a start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ∼ italic_π ( italic_s start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) , and then take the corresponding real action as u t ∼ D ⁢ [ a t ] similar-to subscript 𝑢 𝑡 𝐷 delimited-[] subscript 𝑎 𝑡 u_{t}\sim D[a_{t}] italic_u start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ∼ italic_D [ italic_a start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ] .

Note that other works have used data extracted from the agent’s policy to obtain a mapping from latent to real actions (Edwards et al., 2019 ; Ye et al., 2022 ) , but we found using expert data enabled us to better evaluate the quality of the learnt policy. As shown in the main text, the agent was capable of adapting with as few as 200 200 200 200 expert labels.

E.2 Architecture

We train a transformer as the policy for both the oracle and latent BC agents. We utilize our proposed ST-ViViT architecture for encoding the frames 𝒙 1 : t = ( x 1 , ⋯ ⁢ x t ) subscript 𝒙 : 1 𝑡 subscript 𝑥 1 ⋯ subscript 𝑥 𝑡 \bm{x}_{1:t}=(x_{1},\cdots x_{t}) bold_italic_x start_POSTSUBSCRIPT 1 : italic_t end_POSTSUBSCRIPT = ( italic_x start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , ⋯ italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) . All previous actions are placed through a one-hot and then combined with the corresponding frame encoding as an additive embedding. We use a sequence length of 4 4 4 4 during both training and inference and a batch size of 16 16 16 16 .

Both the oracle and Genie LAM are trained with a cross-entropy loss where targets are either real or latent actions, respectively. During inference, we obtain the final prediction by sampling from the predicted logits. Note we found the oracle agent performed better when we randomly sampled actions 10 % percent 10 10\% 10 % of the time.

Appendix F Reproducible Case Study

In this section we describe a self-contained, fully reproducible case study that can be trained with a single mid range TPU/GPU in under a week.

F.1 Data Collection

First we need to collect the data to train our model. We use the CoinRun environment from the Procgen benchmark (Cobbe et al., 2020 ) since it has thousands of visually diverse levels with fairly simple platformer-like dynamics. Using the “hard” mode, we collect data using a random policy with no action repeats. We sample level seeds between zero and 10,000 and collect 1,000 timesteps for each level, for a total of 10M transitions.

F.2 Video Tokenizer Training

Our video tokenizer for CoinRun follows the same setup as described in Section  2.1 , trained with the optimizer configuration as in Section  C.2 . The primary difference in this example is we use smaller model sizes (see Table  15 ), and then use a batch size of 48 sequences, of length 16, for a total of 768 images per batch. This is sufficient to fit in a single TPU with 16G memory. The model is trained for three days using a single TPU which is sufficient to complete 300k steps.

F.3 Dynamics + Latent Action Model Training

Once we have trained the video tokenizer we can then jointly train the latent action and dynamics models. Once again we seek to fit our model training inside 16G memory, so we use a batch size of 36 sequences consisting of 16 frames each, for a total of 576 images. We train both the latent action model and dynamics model in parallel, using the setup described above (see: Section  C.1 for the latent action model and Section  C.3 for the dynamics model).

We train both the latent action and dynamics models in parallel for 200k steps, using the optimizer hyperparameters in Table   9 . We find this model generates consistent playable latent actions, resembling the original environment.

IMAGES

  1. How to Write an Abstract for a Research Paper

    how long is an abstract research paper

  2. what is an abstract in a thesis paper

    how long is an abstract research paper

  3. 002 Essay Abstract Example ~ Thatsnotus

    how long is an abstract research paper

  4. A Complete Guide on How to Write an Abstract for a Research Paper

    how long is an abstract research paper

  5. Calaméo

    how long is an abstract research paper

  6. How To Write Abstract For Research Paper

    how long is an abstract research paper

VIDEO

  1. Congrats

  2. How To Write an Abstract. #research #researchtips

  3. research -Abstract

  4. How to write a good abstract for your research paper

  5. HOW TO WRITE ABSTRACT IN A RESEARCH PAPER

  6. Choosing the Perfect Abstract 📑✨ #Research #WritingTips #USMLEPrep #foryou #reels #medicalstudent

COMMENTS

  1. Writing an Abstract for Your Research Paper

    Definition and Purpose of Abstracts An abstract is a short summary of your (published or unpublished) research paper, usually about a paragraph (c. 6-7 sentences, 150-250 words) long. A well-written abstract serves multiple purposes: an abstract lets readers get the gist or essence of your paper or article quickly, in order to decide whether to….

  2. APA Abstract (2020)

    Follow these five steps to format your abstract in APA Style: Insert a running head (for a professional paper—not needed for a student paper) and page number. Set page margins to 1 inch (2.54 cm). Write "Abstract" (bold and centered) at the top of the page. Place the contents of your abstract on the next line.

  3. How to Write an Abstract

    An abstract is a short summary of a longer work (such as a thesis, dissertation or research paper). The abstract concisely reports the aims and outcomes of your research, so that readers know exactly what your paper is about. ... You can shorten your abstract by not using excessively long sentences. Avoiding the following five things is an easy ...

  4. PDF Abstract and Keywords Guide, APA Style 7th Edition

    It provides an overview of the paper and helps readers decide whether to read the full text. Limit your abstract to 250 words. 1. Abstract Content . The abstract addresses the following (usually 1-2 sentences per topic): • key aspects of the literature review • problem under investigation or research question(s) • clearly stated ...

  5. How to Write an Abstract (With Examples)

    How Long Should an Abstract Be? If you are writing a paper for a journal or an assignment, the publication or academic institution might have specific formatting rules for how long your abstract should be. However, if they don't, most abstracts are between 150 and 300 words long.

  6. Abstract Writing: A Step-by-Step Guide With Tips & Examples

    You can, however, write a draft at the beginning of your research and add in any gaps later. If you find abstract writing a herculean task, here are the few tips to help you with it: 1. Always develop a framework to support your abstract. Before writing, ensure you create a clear outline for your abstract.

  7. Research Paper Abstract

    Research Paper Abstract is a brief summary of a research paper that describes the study's purpose, methods, findings, and conclusions. It is often the first section of the paper that readers encounter, and its purpose is to provide a concise and accurate overview of the paper's content. The typical length of an abstract is usually around ...

  8. Abstracts

    Authors abstract various longer works, including book proposals, dissertations, and online journal articles. There are two main types of abstracts: descriptive and informative. A descriptive abstract briefly describes the longer work, while an informative abstract presents all the main arguments and important results.

  9. How to Write an Abstract

    How to Write an Abstract | Steps & Examples. Published on 1 March 2019 by Shona McCombes.Revised on 10 October 2022 by Eoghan Ryan. An abstract is a short summary of a longer work (such as a dissertation or research paper).The abstract concisely reports the aims and outcomes of your research, so that readers know exactly what your paper is about.

  10. How to Write an Abstract for a Research Paper

    Include 5 to 10 important words or short phrases central to your research in both the abstract and the keywords section. For example, if you are writing a paper on the prevalence of obesity among lower classes that crosses international boundaries, you should include terms like "obesity," "prevalence," "international," "lower ...

  11. How to Write an Abstract in APA Format with Examples

    An APA abstract is a brief, comprehensive summary of the contents of an article, research paper, dissertation, or report. ... How long should an APA abstract be? An APA abstract should typically be between 150 to 250 words long. However, the exact length may vary depending on specific publication or assignment guidelines. ...

  12. How to Write an Abstract for Your Paper

    Set page margins at 1 inch (2.54 cm). Write the word "Abstract" at the top of the page, centered and in a bold font. Don't indent the first line. Keep your abstract under 250 words. Include a running header and page numbers on all pages, including the abstract.

  13. The Writing Center

    An abstract is a 150- to 250-word paragraph that provides readers with a quick overview of your essay or report and its organization. It should express your thesis (or central idea) and your key points; it should also suggest any implications or applications of the research you discuss in the paper. According to Carole Slade, an abstract is ...

  14. 3. The Abstract

    An abstract summarizes, usually in one paragraph of 300 words or less, the major aspects of the entire paper in a prescribed sequence that includes: 1) the overall purpose of the study and the research problem(s) you investigated; 2) the basic design of the study; 3) major findings or trends found as a result of your analysis; and, 4) a brief summary of your interpretations and conclusions.

  15. 15 Abstract Examples: A Comprehensive Guide

    When it comes to writing an abstract for a research paper, striking a balance between consciousness and informative detail is essential. Our examples of abstracts will help you grasp this balance better. ... While there is no absolute rule for how long an abstract should be, a general rule of thumb is around 100-150 words. However, some ...

  16. How To Write an Abstract for Any Subject and Publication ...

    An abstract is a short summary of a longer work, such as a study or research paper. The goal is to provide readers with an overview of the purpose, methodology, results, conclusion, and importance of this text. ... How long should an abstract be? Depending on the publication, an abstract should be anywhere from 150 to 250 words. ...

  17. Writing an abstract

    Methods - The methods section should contain enough information to enable the reader to understand what was done, and how. It should include brief details of the research design, sample size, duration of study, and so on. Results - The results section is the most important part of the abstract. This is because readers who skim an abstract do so ...

  18. How To Write an Abstract in 7 Steps (With an Example)

    An abstract is a concise summary of a longer work, such as a dissertation or research paper, and allows readers to decide whether to read the full paper. Abstracts should be written after the full paper is written, and are usually about 150-250 words and one to two paragraphs long. An abstract should include a statement of the problem you are ...

  19. Learn How to Write an Abstract for a Research Paper

    According to the rules of academic writing, an abstract for a research paper should be around six sentences long, which equals about 150 words or less. If you are majoring in English Literature, Healthcare, or Psychology, your abstract may go up to 250 words. Still, writing less for an abstract section of a research paper is recommended because ...

  20. Writing the title and abstract for a research paper: Being concise

    The title and abstracts are the only sections of the research paper that are often freely available to the readers on the journal websites, search engines, and in many abstracting agencies/databases, whereas the full paper may attract a payment per view or a fee for downloading the pdf copy.[1,2,3,7,8,10,11,13,14] The abstract is an independent ...

  21. How to Write an Abstract?

    Abstract. An abstract is a crisp, short, powerful, and self-contained summary of a research manuscript used to help the reader swiftly determine the paper's purpose. Although the abstract is the first paragraph of the manuscript it should be written last when all the other sections have been addressed. Research is formalized curiosity.

  22. 13.1 Formatting a Research Paper

    Set the top, bottom, and side margins of your paper at 1 inch. Use double-spaced text throughout your paper. Use a standard font, such as Times New Roman or Arial, in a legible size (10- to 12-point). Use continuous pagination throughout the paper, including the title page and the references section.

  23. How to write long abstracts?

    An abstract summarizes, usually in one paragraph of 300 words or less, the major aspects of the entire paper in a prescribed sequence that includes: 1) the overall purpose of the study and the ...

  24. CD19 CAR T-Cell Therapy in Autoimmune Disease

    Supported by Deutsche Forschungsgemeinschaft through the Leibniz Award (to Dr. Schett), the research group FOR2886, Collaborative Research Center (CRC) 1181, CRC 1483, CRC-Transregio (TRR) 305 ...

  25. Blood-brain barrier disruption and sustained systemic ...

    Long COVID is a major public health issue since 2020 and exhibits frequent neurological symptoms. ... Abstract. Vascular disruption has been implicated in coronavirus disease 2019 (COVID-19 ...

  26. [2402.15391] Genie: Generative Interactive Environments

    We introduce Genie, the first generative interactive environment trained in an unsupervised manner from unlabelled Internet videos. The model can be prompted to generate an endless variety of action-controllable virtual worlds described through text, synthetic images, photographs, and even sketches. At 11B parameters, Genie can be considered a foundation world model. It is comprised of a ...

  27. Genie: Generative Interactive Environments

    The last few years have seen an emergence of generative AI, with models capable of generating novel and creative content.Driven by breakthroughs in architectures such as transformers (Vaswani et al., 2017), advances in hardware, and a recent focus on scaling models and datasets, we can now generate coherent, conversational language (Radford et al., 2018, 2019; Brown et al., 2020), as well as ...

  28. Why Read Adam Smith Today? by Peter J. Boettke :: SSRN

    Abstract. The scholarly world celebrated Adam Smith's 300th birthday throughout 2023. This article attempts to lay out the case on why we should still be reading Adam Smith today. The argument isn't because we want to honor the founder of a scientific discipline.

  29. CRB1-associated retinal degeneration is dependent on bacterial

    The Crumbs homolog 1 (CRB1) gene is associated with retinal degeneration, most commonly Leber congenital amaurosis (LCA) and retinitis pigmentosa (RP).Here, we demonstrate that murine retinas bearing the Rd8 mutation of Crb1 are characterized by the presence of intralesional bacteria. While normal CRB1 expression was enriched in the apical junctional complexes of retinal pigment epithelium and ...

  30. The Cost of Justice at the Dawn of AI

    Abstract. Justice isn't free, but it might soon get much less expensive. Policies concerning issues such as arbitration, class actions, and plea bargaining depend on how much legal services cost, but the legal literature has generally ignored past and future cost trends and their implications.