67.3 August 2020

A Scheme for Understanding and Writing Summaries

By David K. Farkas


Purpose: Writing a summary is not necessarily a straightforward task. Presented here is a scheme of twelve descriptors that reveals much of the complexity of summaries. The scheme can also serve as a design heuristic.

Method: The scheme draws upon the psychology of text processing and applied linguistics, and derives theoretically from the intertextual relationship between the summary and the summarized text and the contrast between summary as microcosm (miniature of the summarized text) and the many ways in which summaries are not microcosmic.

Results: The scheme consists of these twelve descriptors. The Purpose of the summary. The possibility of a Specification constraint that complicates the writing of the summary. Reduction, the length of a summary as a ratio of the summarized text. The Phrasing—informative or descriptive—of the summary. Proportion and Exclusion, how fully each part of the text is represented in the summary. Structure, how the ideas in the summary are arranged in relation to the text. Placement, where in relation to the text (before, after, within, alongside, or independent of) a summary (or multiple summaries) appears. Addition, the possibility of adding content not in the text to the summary. Authorship, whether the author of the text (or an associate) or someone not associated with the text wrote the summary. The Stance, Style, and Format of a summary.

Conclusion: The scheme advances our understanding of summaries and, especially when used as a heuristic, can help technical communicators in their work. It may also be the basis of further research.

Keywords: summarization, summary writing, abstracts, design, heuristics

Practitioner’s Takeaway:

  • There are many kinds of summaries other than those that appear at the beginning of a document. Also, while it’s easy to assume that summaries are written as simple miniatures of the document, this is often not the case. In fact, there is considerable variation and much recent evolution in the writing of summaries.
  • Presented here is a scheme of twelve descriptors that—with their accompanying discussions—provides a comprehensive and up-to-date look at summaries.
  • The descriptors also function as a design heuristic that enables better summary writing and innovative design.


Summary writing may at first seem a straightforward task requiring no more than basic textbook knowledge and an acquaintance with familiar models. In fact, summaries are complex and vary along many dimensions. They are written to fulfill multiple purposes and to meet diverse constraints. Also, the documents we summarize belong to very different genres and, even within a genre, differ greatly.

Because summary writing is a complex task, technical communicators should be able to draw upon a comprehensive understanding of summaries and a wide range of design strategies when they plan and write a summary. I therefore set forth a scheme consisting of twelve rhetorically focused descriptors. The scheme is, first and foremost, a conceptual description and synthesis of much of what we know about summaries. It can also serve as a design heuristic that can help technical communicators and all professionals who write summaries identify a wide range of useful design strategies, both familiar and unfamiliar. Each descriptor prompts design thinking about a particular aspect of writing a summary. The scheme may also contribute to further research and support teaching.

This scheme applies across genres to summaries of all kinds, from research article abstracts, to the longer executive summaries that precede reports, proposals, business plans, and other organizational and business documents, to the brief summaries that appear before articles in news magazines, both in print and on the Web. These and other before-the-body summaries are separate components that are visually bounded from the body of the text through such means as box borders, a distinctive font, and extra spacing. But the scheme applies as well to the reviews that, especially in instructional documents, are visually bounded components that follow the body of the text.

Furthermore, the scheme applies to the unbounded summaries we embed within the body of a document: Unbounded summaries are regularly embedded within the introduction of a document, where they preview the main ideas, and they are often embedded within the conclusion of a document, where they review the main ideas. Indeed, unbounded summaries can be embedded anywhere in the body of the text, and they can either summarize some or all of the preceding content (signaled by such phrases as “Summing up what has been said so far”) or preview some or all of the upcoming content.

Here is a capsule presentation of the scheme:

  • Purpose: What is the purpose or purposes of the summary?
  • Specification constraint: Must the summary be written according to a gatekeeper’s specification, or is it unspecified?
  • Reduction: What is the ratio, in word count or by some other measure, of the summary to the summarized text? How was the reduction accomplished?
  • Phrasing: Is the summary written with informative or descriptive phrasing or a mixture of both?
  • Proportion and Exclusion: Is each part of the summarized text reduced by the same amount? Is any important content excluded from the summary? What are the reasons for improportion or exclusion?
  • Structure: How is the summary organized in relation to the text? How closely does the sequence of ideas in the summary map the sequence in the text?
  • Placement: Is there a single summary that appears before the body of the text (the default), a review that appears after the body of the text, or multiple summaries distributed within the body of the text or in the margins? Is the summary a free-standing module that excludes references to the text?
  • Addition: Is the content of a before-the-body summary strictly limited to what appears in the text? Is there any new content? Does a summarizing conclusion add new thoughts about the broader implications of the topic or point to future developments?
  • Authorship: Was the summary written by the author of the text or an associate? Or, was it written by a commentator or someone else with no direct connection to the text?
  • Stance: Is there a difference in stance between the summary and the text? How did this difference come about? What are the consequences?
  • Style: How do the style of the summary and the stylistic differences between summary and text affect the reader?
  • Format: How is the summary formatted and how do the format elements impact the reader’s experience?


The research on summarization and summaries is vast and varied. Much of the research comes from the fields of psycholinguistics and educational psychology, often deriving from the model devised by Kintsch and van Dijk (1978) of the creation of macropropositions through the macrorules of deletion, selection, generalization, and construction. This model has been modified and adapted by subsequent researchers (Brown & Day, 1983; Sherrard, 1989; Li & Hoey, 2014), and this broad line of research encompasses the processes and strategies entailed in writing summaries (e.g., comprehension, condensation, and production of a new text), how these linguistic skills develop, and how best to teach these skills in school settings (Kintsch & van Dijk, 1978; Sherrard, 1989; Brown & Day, 1983; Westby et al., 2010).

From the perspective of text processing, psychologists study how summaries, functioning as signals and advance organizers, improve comprehension and retention (Mayer et al., 1984; Kardash & Noel, 2000; Lorch & Lorch, 1996; Lemarié et al., 2008). Applied linguists, often with an interest in training early career researchers and L2 (second language) speakers of English, study research article abstracts at a very fine-grained level, often using genre move analysis (Swales, 1990; Swales & Feak, 2009; Ayers, 2008; Breeze, 2016; Samraj, 2005; Stotesbury, 2003; Supatranont, 2012). Other researchers with an interest in the enterprise of science (often medical science) also focus on research article abstracts (Berwanger et al., 2009; Dupuy et al., 2003). Information scientists study how successfully abstracts return literature citations when users conduct keyword searches of databases (Lin, 2009). Computer scientists in machine learning and related fields are trying to improve the ability of computers to summarize text (Lloret & Palomar, 2012).

In the field of technical and professional communication, most work on summaries and abstracts takes the form of textbooks and books on writing up research for publication (e.g., Mogull, 2018; Montgomery, 2003). These treatments are necessarily brief and are based on familiar models. The specifications of summaries in published guidelines are usually—though not always (JAMA Network, 2019)—brief and narrowly tailored to the specific publication. The American National Standards Institute/National Information Standards Organization (ANSI/NISO) (2015) standards document Guidelines for Abstracts only codifies familiar models and covers only a limited number of text features. I draw upon much of this prior work, in particular, that of psychologists who study text processing and that of applied linguists.

This investigation, however, is distinct from all prior work I have seen because it addresses the challenges faced by writers and information designers, looks closely at many summary features and design options, and encompasses a very wide range of summaries. Also, this scheme makes clear that many design issues apply to a broad range of summaries. Most research overlooks the commonality underlying diverse summaries (e.g., bounded vs. unbounded summaries), and so we fail to apply what is known about writing one kind of summary to the writing of other kinds.


A significant part of my recent academic work has been motivated by the conviction that those who wish to communicate substantial textual content must reckon with readers’ increasing resistance to long texts and that facilitating selective reading within a document and improving the effectiveness of summaries will therefore become increasingly important (Zhou & Farkas, 2009; van der Meij et al., 2013; Farkas & Raleigh, 2013). The starting point of this project was the observation that while a summary can be written as a microcosm (miniature) of the full text, summaries very often differ from the full text along numerous dimensions. Indeed, present-day summaries appear to be increasingly less microcosmic.

The reasoning underlying the project is that the idea of a summary as microcosm directs us to the ways in which many of the summaries we write are not microcosms—and why. That is, there are reasons why an individual writer or a particular organization or discipline chooses a proportionality, structure, or rhetorical stance that is different from that of the summarized text. Therefore, a systematic look at how summaries are very often not microcosms should lead to a better understanding of summaries and, hence, better summary writing and innovative design.

My method was a two-year process of analysis and classification, based on an informal but wide-ranging survey of summaries—every kind of summary I could recall from over 40 years working in this field; a fresh examination of how summaries are currently written in most major genres of professional writing; and a systematic review of the academic and professional literature on summaries and summarization.

I made a provisional list of the text features and other design considerations that differentiate the broad universe of summaries. This became my scheme of descriptors. Through an iterative process, I sought to make the scheme truly comprehensive and to make the individual descriptors distinct, meaningful, and useful. I did not construct a study sample of summaries according to highly specific criteria so as to draw a limited number of highly focused conclusions using descriptive statistics—a research method known as corpus linguistics. Nor was my focus on how specific genres of documents are summarized, but rather on summarization in its most general form. Most of the descriptors are based on widely recognized aspects of summarization. However, the options embodied in the Placement descriptor have largely been overlooked, and Addition has only recently drawn the attention of researchers after the wide adoption of non-specialist synopses.

An invitation to talk about summaries in a graduate seminar led me to more carefully consider and articulate the theory that was guiding my thinking, and the more refined theory aided in my ongoing analysis. I recognized that the summary as microcosm idea is an instance of the broader concept of intertextuality—the idea that texts may be related to one another in important ways (Orr, 2010). While each summary is certainly a unique text crafted to succeed in its own rhetorical circumstances, it is equally true that summaries are closely tied to their summarized text (Heard, 2016; Holtz, 2009; Hyland, 2005; Samraj, 2005). Indeed, in the world of professional documents, there is hardly a closer relationship than that between a summary and the summarized text. This makes clear why several of the descriptors (e.g., Reduction, Proportion and Exclusion, and Addition) are meaningless except in relation to a summarized text and why other descriptors (e.g., Purpose, Style, and Format), while characteristic of all texts, are discussed within this scheme primarily in regard to the close and derivative relationship between summary and text. The following two considerations both support and are foregrounded by the idea of intertextuality:

  • Readers notice significant differences as they transition from summary to text (or from text to summary). The difference can be dramatic, for example, when a reader transitions from a findings-first abstract to a standard research article.
  • Authors write their abstracts and other summaries with much attention to a completed (or mostly completed) text they have labored over.

This statement of method and theoretical rationale still leaves open an important question: Why these twelve descriptors and not others? The best response is that complex domains such as summaries can be conceptually divided in multiple ways, and no one way is the “correct” one. Certainly, the theory that guided my analysis is nowhere near strong enough to yield a definitive set of descriptors. Rather, the appropriate test for the twelve descriptors presented here, or for a competing set of descriptors, is how fully, meaningfully, and usefully they reveal the complexity of the domain of summaries, while being crisply differentiated and sufficiently limited in number to be manageable in application. The proof is in the pudding. Consider, for example, that this scheme might have included the descriptor “Improving retention,” but this potential descriptor was folded into the broader Purpose descriptor. The descriptor with the least practical application is Authorship, but it is included because it adds significantly to the conceptual breadth of the scheme.

The twelve descriptors belong to two categories. Nine are text variables, actual features of texts that can be manipulated in the writing process. But Purpose, the Specification constraint, and Authorship are circumstances that surround and precede the creation of summaries. Purpose is a central consideration in planning a summary. Likewise, if the summary must adhere to a gatekeeper’s specification, this too is an important planning consideration. Authorship—whether you are summarizing your own or another person’s text—impacts the writing process.


Months after teaching the graduate seminar, I gave a presentation to working technical communicators about writing effective summaries. The process of planning the presentation along with an informal follow-up discussion with several audience members led me to see that the scheme could serve as a design heuristic.

The scheme is a design heuristic of a specific kind. Many design heuristics, including collections of design patterns, are explicitly prescriptive and, often, research based—for example, the HHS Web Standards and Usability Guidelines (2019). Note, however, that the highly prescriptive Journalist’s 5 Ws (Who, What, Where, When, Why) is supported by venerable tradition and rhetorical theory but not by empirical research. Another category of heuristics, that of invention heuristics, is non-prescriptive but instead supports deep engagement with the subject matter and creative thinking. Invention heuristics include the particle, wave, field perspective (Young et al., 1970), and cubing (Axelrod & Cooper, 2012, pp. 322–323).

This scheme, in its function as a heuristic, is fundamentally non-prescriptive. It calls attention to design considerations which the writer can then address using his or her own background and experience. I do, however, offer some judgments about the good and bad design of summaries.

As a design heuristic, the scheme can assist students, early career researchers, and workplace writers when they write specified or unspecified summaries. In addition, the heuristic can assist expert writers, publications managers, content strategists, and others who are tasked with specifying a summary, re-thinking a specification, designing a summary for a special situation, or attempting to innovate broadly. For use as a heuristic, the scheme can be modified to suit the background and needs of those who will use it. The conceptual parts of each of the twelve discussions can be trimmed, the Authorship descriptor can be deleted, and the heuristic can be narrowed in scope so that it applies only to abstracts, executive summaries, and other before-the-body summaries. Following my discussion of the twelve descriptors appear brief scenarios that demonstrate how the scheme can be used as a heuristic.


Summaries serve various purposes, and the chosen purpose or purposes will govern many design decisions. One purpose is filtering: Often, readers read a summary, decide whether or not the topic is relevant to their needs, and, if so, decide whether to read the full text or whether the summary itself is sufficiently informative. In other instances (for example, abstracts of journal articles that are only available to online journal subscribers), a summary must suffice for some readers because the text is unavailable. When readers read both the summary and the text, the previewing function of the summary can improve the reader’s comprehension and retention of the text (Lorch & Lorch, 1996; Mayer et al., 1984).

When summaries map the structure of the text, readers can use the summary to find the section containing specific information they are interested in. Some summaries, as shown below, are specifically designed to support selective reading within a text. Readers can comfortably bypass sections of the text that interest them less on the basis of what they have learned about the content of those sections from reading the summary. Reviews very often appear in textbooks and training materials as an after-the-body component. Their purpose is to improve retention.


Many organizations issue or reference guidelines specifying how the before-the-body summaries of documents prepared for that organization are to be written. Writers of such summaries may find the specifications burdensome. Of course, other kinds of documents written in organizational settings are often written to specifications, but special problems may arise in summary writing because summaries are so closely tied to an existing text. Problems are especially likely to arise if the document being summarized diverges in some way from the class of documents the specification of the summary was devised for. For example, if the specification states that summaries should be “objective” in regard to their stance, but the summarized text is explicitly polemical, the summary writer is caught in this divergence. Similarly, when a specification mandates a structure for the summary that is different from the structure of the text, the summary’s structure may be an awkward fit with that of the summarized text.

Summary writers should consider how much leeway they have in interpreting the specification or, in certain situations, should ask for an exemption from the burdensome part of the specification. But often they will need to deal creatively with certain specification constraints. For example, if a guidelines document specifies one-paragraph abstracts, and the text divides sharply into multiple sections, the summary writer may look for a way to reveal the structure of the text within the summary. Further implications of the specification constraint are noted below.


The length of a summary in relation to the text largely determines how much information the summary conveys. Therefore, reduction, meaning the degree of reduction, strongly influences the uses of a summary. For example, a reader with only a moderate interest in a topic may well be able to satisfy their information needs by reading a 250-word summary, but this is less likely if there is only a 100-word summary. On the other hand, when summaries are long, a reader confronted with numerous summaries may choose to read fewer of them.

In many situations, summary writers make their own decisions about how long a summary should be, but still more often the summary’s length is specified by guidelines. But whether the limit has been imposed by a guideline or represents the author’s best judgment, limiting the length of a summary is a necessary and often challenging task. In creating a summary, the writer selects the most important ideas in the text. The threshold for “important” is a function of the length and the information density of both the text and the summary. For example: The threshold for inclusion will be much higher (ideas must be more important) for a 200-word vs. a 500-word summary—especially so if the summarized text is a dense 50-page engineering document versus a more casually written 20-page project update. Importance, furthermore, should be generally understood to mean the most important (superordinate) ideas in the summarized text in preference to what the summarizer judges to be important to the reader. However, as we will see below, rhetorical considerations may result in summaries that depart—sometimes quite dramatically—from expressing the summarized text’s superordinate ideas.

In conjunction with selection, the ideas in the summarized text are condensed through various stylistic means. These include generalized phrasing in place of a set of specific items, tighter syntax (including nominalization), and making larger assumptions about the relevant background knowledge the audience brings to the text (Holtz, 2009; Kintsch & van Dijk, 1978; van Dijk, 2003). My summary of Lincoln’s “Gettysburg Address,” though very brief, should still be meaningful to an educated United States audience because these readers presumably know that Abraham Lincoln, President of the United States during the American Civil War, is speaking about the great many Union soldiers who died in the crucial battle that took place in Gettysburg, Pennsylvania:

Lincoln’s message was simple: Only by dedicating ourselves to winning the war will we keep faith with those who gave their lives to preserve liberty and equality.

In regard to condensation, the single most consequential decision is whether to employ standard informative phrasing, descriptive (“indicative”) phrasing, or a mixture of the two. Descriptive phrasing, the basis of the familiar descriptive abstract, much reduces a summary’s length because descriptive phrasing only indicates what can be found in the text (ANSI/NISO, 2015; Paradis & Zimmerman, 2002). In this next example, we do not learn the actual relationships that are referred to: “This study identifies relationships between the physical and genetic characteristics of bones in mice.” When used consistently, descriptive phrasing limits the purposes served by the summary: The summary does well at performing the filtering function, but it cannot substitute for reading the text. Because executive summaries are especially likely to be read in place of the text, executive summaries are not only relatively lengthy in relation to the text (mild reduction), but they primarily use informative phrasing.

Reduction in all its forms comes at a price. Less of the content of the summarized text is communicated, there is more potential for misunderstanding, and inadequacies in the writing style may arise. Specifically, two big problems that may occur are decreased readability and alterations of meaning arising from the elimination of hedging words. For example, “can lead to” becomes “leads to.” The implications of reduction will be discussed further as we proceed.


The distinction between informative and descriptive phrasing was introduced as a stylistic technique for reduction, which it is. But I classify this distinction as a separate descriptor within this scheme because it is so consequential: A summary that tells you what you will learn by reading the text is very different from a summary that provides immediately useful information.

When a specification does not call for either informative or descriptive (indicative) phrasing, authors should take note of this extra flexibility. A good strategy, I suggest, is to favor informative phrasing in the interest of communicating more information to the audience, but to strategically switch to descriptive phrasing when greater reduction is necessary to summarize especially complex parts of the text (especially if they are of lesser importance) and in other instances when ideas in the text resist informative phrasing.

Reviews typically appear as components of textbooks, training manuals, online tutorials, and other documents with an explicit goal of retention. They may be written with informative phrasing or may exploit the reader’s familiarity with the text and use descriptive phrasing. The distinction between informative and descriptive phrasing reappears in different contexts below.


Proportion is the comparative level of detail of the ideas in the summary in relation to the level of detail of those ideas as they appear in the text. The naïve assumption is that a summary is a microcosm that, fractal-like, retains the text’s original proportions by reducing all parts of the text to the same degree. However, although accurate proportion serves readers well in many situations and is often specified (or simply assumed), both summary writers and gatekeepers often have reasons to depart from accurate proportion. The most extreme improportion is the outright exclusion of an idea that a judicious reader or a capable computer algorithm would rate as superordinate.

One reason for improportion is simply that some ideas cannot be summarized as concisely as others and will therefore require a disproportionately lengthier explanation in the summary, unless descriptive phrasing is used.

In other situations, a gatekeeper or summary writer may decide that the goals of readers of a summary are sufficiently different from the goals of those same individuals as readers of the text to warrant different proportions. This difference in goals often arises in regard to the Methods section of research articles, which may be disproportionately briefer or entirely omitted in the abstract (Samraj, 2005; Ayers, 2008). Such improportion presumably arises because methods are often routine or standardized and because the reader of the abstract initially assumes that the research is methodologically sound and recognizes that, in any case, identifying problems with the methods will require a careful examination of the article.

The introductions of many documents include a preview of the upcoming sections, but these summaries routinely exclude the conclusion section if the conclusion section does nothing more than summarize the main ideas.

Extreme improportion also occurs in some “promotional abstracts” that appear in websites advertising the presentations in professional or trade conferences. The writers of this kind of abstract are usually consultants, industry experts, or representatives of a vendor but may be academics as well. A great many abstracts of all kinds, mirroring the text, follow a problem-solution structure (Jordan, 1988). These include the abstracts of research articles, because the canonical IMRAD pattern of research articles (Introduction, Materials and Methods, Results, and Discussion) or something similar itself embodies a problem-solution structure. Below, we see the problem-solution structure in a promotional abstract for a presentation to be given at a Gartner conference. While the abstract explains in relative detail a significant problem faced by members of the audience, the abstract, switching to a higher level of reduction and to descriptive phrasing, provides scanty information about the solution:

The Evolution of Sales Development Reps
to Support B2B Demand Generation

B2B companies are increasingly relying on sales
development reps (SDRs) to support their demand
generation and prospecting efforts, but the success
of these teams varies widely from company to
company. In this session, you’ll learn how to enable
SDRs for success and why quality matters more
than quantity. (Berkowitz, 2018)

A summary of Berkowitz’s presentation posted by Gartner after the presentation indicates that Berkowitz’s presentation, unsurprisingly, focused on solutions (Bryan, 2018).

There are other rhetorical strategies that can lead to exclusion in a summary. A writer addressing a lay audience might omit from a summary an important but dauntingly complex idea in order to draw readers into the text. This practice, I suggest, is ethical if the summary employs a heading or other descriptive phrase which signals that the summary is highly selective: “Things you will learn in this article.” Finally, taking on the perspective of genre moves, we note that both aggressive reduction and improportion and exclusion often result in abstracts that contain only a subset (sometimes just purpose and results) of a research article’s genre moves (Swales & Feak, 2009).


Structure is the arrangement of the ideas making up the summary. In this scheme the emphasis is on the difference between the structure of the summary and the structure of the text. Note that structure is distinct from proportion. For example, you can change the proportions of a summary without changing the structure, much as the structure of a sandwich remains the same (bread, avocado, bread) regardless of whether the bread slices are thick or thin or how much avocado is inside. Conversely, you can re-arrange the parts without changing the size of the parts.


One broad and far-reaching design choice regarding structure is mapping: organizing the summary so that its structure (whether understood in terms of major topics, structural divisions, genre moves, etc.) corresponds to that of the text. In the case of before-the-body summaries, gatekeepers often specify or at least imply mapping. In addition, reviews and other kinds of summaries are very often mapped. Both readers and writers of summaries, unless they have been conditioned by reading unmapped summaries, will assume that summaries map texts; and, very often, the logic of the summary will require at least parts of the summary to map the text.

Various factors result in loose mapping. When texts are informally structured or, as in many humanities texts, when they employ an elaborate but inexplicit structure in which the author returns to ideas developed previously, only loose mapping is possible.

Reduction can also require looser mapping, even when texts are rigorously and explicitly structured. Envision a text organized with these topics and subtopics: (Introduction) (A¹ A² A³) () (B¹ B² B³) (C² C³) (D¹ D² D³) (Conclusion). Subtopic has been moved from its natural and expected location because subtopic cannot be made clear unless the groundwork has been laid by ideas contained in . The author of the article will likely explain this departure from the expected sequence of topics, and readers will have no difficulty with it.

In the summary, however, has been condensed along with  and  into the broader idea B, and has been condensed along with and  into the broader idea C. Nothing in this condensed B requires any information from C as a prerequisite, and the summary writer is now free to follow the natural and expected sequence of ideas: (Introduction) (A) (B) (C) (D) (Conclusion). Readers will likely not notice that the sequence of ideas in the text differs slightly from what they read in the summary. We can say, then, that complexity in the organization of the text has been “smoothed out” in the much briefer summary.

Unmapped Summaries

Many summaries do not map at all. Aggressive reduction can cause writers to abandon mapping. The brief summary of Lincoln’s “Gettysburg Address” that appeared above reverses the sequence of the main ideas in Lincoln’s speech.

Summaries can be planned around a structure that is fundamentally different from that of the text. The ANSI/NISO standards document for abstracts (2015) allows for a “results-oriented” arrangement in which the most important results and conclusions are placed first. The JAMA Network’s “Instructions for Authors” (2019) specifies numerous structures for JAMA network research articles and various structures for the abstracts that do not map the structures of the articles. As the JAMA Network’s “Instructions for Authors” suggests, the myriad possibilities for unmapped summaries are an avenue for innovation in the design of summaries.

Although rhetorical considerations may well justify unmapped summaries, it is probable that the recall of the complete text (both summary and text together) decreases when the mental model of the content structure that the reader creates from reading a summary is not carried through in the text (Meyer & Poon, 2001; Ritchey et al., 2008).


As we have seen, bounded summaries can appear before or after the body of the text, and unbounded summaries can appear within an introduction, within a conclusion, or anywhere else within the body of a text. Here, we look at three other placement options: multiple within-text summaries, marginal summaries, and free-standing summaries.

Multiple Within-Text Summaries

Multiple summaries may be placed before chapters, sections, or other divisions within a text. A website  in which a summary appears on each page (except for the introductory page) is shown in Figure 1 (BBC News, 2002).

This and similar multiple-summary designs serve the purpose of supporting selective reading. That is, for each of the events described on the website, the reader is invited to read just the summary, the text, or both. To support selective reading effectively, the reduction in length must be mild (longish summaries). This is to ensure that a reader who has chosen to read only the summaries on several of the webpages has accumulated enough information to be able to productively read the full text of a later section (Farkas & Raleigh, 2013). QuikScan is a similar, though more elaborate, multiple summary design whose benefits have been established through empirical research (www.quikscan.org/research, 2019; van der Meij & van der Meij, 2011).

Marginal Summaries

Multiple summaries may be placed alongside the text in wide page margins. This practice is rare today but was more prevalent in past centuries. These days, wide page margins usually contain relevant factoids and definitions, most famously in Richard Saul Wurman’s Information Anxiety (1989), or else contain definitions and tips (as in certain computer manuals). Figure 2 illustrates the informatively phrased marginal summaries that authors can write for their documents, most likely in conjunction with a conventional abstract.

Much like multiple within-text designs, these marginal-summary designs support selective reading, as long as the reduction is mild. Marginal summaries also help readers find specific ideas in the text by scanning the marginal summaries. If descriptive phrasing is used, the marginal summaries will not support selective reading, but they can be greatly reduced in length.

Free-Standing Summaries

Finally, a great many summaries are written so that they can be read independently, and so the free-standing summary becomes an additional category, an overlapping category, within the Placement descriptor. Research journals often specify in their author guidelines that abstracts must be free standing. Thus, the guidelines—in addition to specifying informative phrasing—prohibit in-text citations to items in the article’s reference list. If an item in the reference list must be mentioned in the abstract, an abbreviated reference, not a citation, is provided.

As a negative example, envision a social scientist who has completed the draft of a book but decides to modify the first chapter so that it is more likely to be included in print or electronic course packs in university courses. He greatly reduces contextual content that served as background for the subsequent chapters and deletes all allusions to the subsequent chapters and to the book as a whole. He has turned his introductory chapter into a free-standing summary and a microcosm of the rest of the book. But it is less effective as an introduction.


Addition refers to adding information to a summary that does not appear within the text (or, in the case of summaries embedded in the body of the text, does not appear in any other part of the text). At first glance, adding any new content to a summary seems antithetical to the notion of a summary, but certain additions are allowed and even specified.

One form of addition is adding information to before-the-body summaries. The executive summaries of reports primarily intended for specialists are often written for generalist audiences such as managers with broad responsibilities and elected officials, such as city council members. Therefore, guidelines for executive summaries may cautiously indicate that summaries may provide additional content in the interest of making the summary easier to understand. The Oxfam guideline on writing executive summaries states that “the summary is the bridge between the often-technical arguments and findings of the main report, and the target audience, whether media, decision makers or aid professionals” (Green, 2015).

Breeze (2016) describes non-specialist synopses, which are gaining popularity, primarily as a supplement to standard research article abstracts. They are meant to communicate with a broader audience. Non-specialist synopses include context, explanations, and examples and may conclude with an explanation of the relevance of the research—all of which may not be present in the article itself. The Journal of Bacteriology (American Society for Microbiology, 2019) requires abstracts to include a separate “Importance section” that provides “a nontechnical explanation of the significance of the study to the field.” Such additions are valid and useful provided that they do not in any way misrepresent the text.

The minutes of major civic meetings held in a public setting will likely summarize, without additions, all the important ideas that were expressed. But when small civic bodies meet without outside attendees, it is often necessary for the writer of the minutes to supply information that was known to the meeting participants and was therefore left unstated during their discussion. Below is a (modified) excerpt from the minutes of a meeting of the commissioners of a small city. Their discussion of the successful grant application was far more fragmentary than the “filled out” rendition we see in the minutes:

The city has been awarded a $10,000 grant from the King County Conservation District. Trees will be planted in Blue Heron Park to replace invasive trees and the trees that were removed because they were interfering with power lines. This will serve as a demonstration project, with interpretive signage explaining what trees should be planted in right-of-way areas. The trees will be planted in March or April.

Some conclusions are strictly summaries. But many conclusions blend summarization with additional information as they make reference to the broader implications of the topic or point to future developments.


Most summaries are written by the author of the summarized text or by an associate. However, one can write a summary of another author’s text. Therefore, we distinguish between an authorial and a non-authorial summary. My brief, non-authorial summary of Lincoln’s “Gettysburg Address” was written more than 150 years after Lincoln’s death rather than by Lincoln or a staff member.

Non-authorial summaries are generally embedded in the document of the writer who wrote the summary and, generally, are blends of summary and interpretive commentary. For example, my summary of the “Gettysburg Address” reads as though it were part of an interpretive discussion of the speech or the historical context of the speech. Book and movie reviews embed summaries into the reviewer’s text and most often fully blend the summary with interpretive commentary. On the other hand, this non-authorial, unbounded summary does not blend summary content with interpretive content:

Before discussing Joseph Browne’s ideas about infectious diseases, I will summarize his Practical Treatise on the Plague, written 1720.

Note that like the Purpose descriptor and the Specification constraint descriptor, authorship is not a design choice. Rather, it is a matter of circumstance whether one is summarizing one’s own ideas or summarizing another author’s text. However, the decision whether or not to blend a non-authorial summary is a design choice.


Stance is a broad concept that encompasses a writer’s presentation of self and the writer’s attitude toward the content of the text and the audience (Hyland, 2005). A writer, for example, might take an impersonal or a personal stance and might present the content as either a collegial contribution to or as a radical departure from the research, thinking, or values of a discourse community. In discussing a text written by another author, a neutral stance, a mildly or strongly laudatory stance, a disparaging stance, or some other stance is possible. Now we consider stance in regard to a summary and the summarized text.

The stance of an authorial summary, as we would expect, is almost always similar to that of the summarized text. The expectation is that the same authorial voice is addressing us at two different levels of detail. Even so, differences are certainly possible. First, using the terminology of Hyland (2005), the pressure for reduction can lead authors to delete hedging words (thereby accentuating the stance) or delete boosters (thereby moderating the stance). Beyond this, because summaries inherently “advertise” the summarized text (Ayers, 2008; Heard, 2016), authors—with the goal of being read—may consciously accentuate the stance of their summaries, for example, by writing abstracts which make stronger claims for the value of the research or which are more polemical. Stotesbury (2003) notes a tendency for research article abstracts, especially in the humanities, to be more explicitly evaluative than the texts themselves. Authors should be thoughtful and cautious about any significant difference between the stance of their summary and their text, for these differences, although they may be subtle, are highly consequential. When we summarize another author’s work, there is an ethical imperative to be fair-minded and scrupulous, especially because the reader is apt not to examine the summarized text or have it immediately available for examination.


Style refers to a wide range of linguistic techniques, especially those at the word and sentence level, and includes both techniques that directly affect ease of reading and those that help establish stance. Style has previously been discussed as a means of reduction and in regard to the distinction between informative vs. descriptive phrasing. In this discussion, we are concerned with stylistic differences between summary and text, and my focus is before-the-body summaries.

A great many documents intended for the workplace do not have a conspicuous writing style, and workplace authors often have no reason to employ a special writing style when writing the summary. If there is a difference, it is apt to stem largely from the pressure of reduction (Holtz, 2009).

In some instances, however, authors intend a different style. Non-specialist synopses may employ techniques that add to word count but result in a more readable and “livelier” style than does the text (Breeze, 2016). Promotional abstracts employ a lively, informal style and often include a range of devices (pronouns referring to the reader, rhetorical questions, and references to shared knowledge) designed to promote engagement (Hyland, 2005).

When the text has a conspicuous style—for example, the more complex syntax and Classical rhetorical elements we often find in 19th century texts—there is no need for the summary to mimic that style. Note, however, that my brief summary of Abraham Lincoln’s “Gettysburg Address” includes a syntactical construction (“Only by . . . ”), which, at the cost of a few extra words, retains just a bit of Lincoln’s intensity and lofty eloquence.


As noted, one key role of format is to visually bound summaries from the text. But there are significant format options within a bounded summary. The structured abstract—an abstract with headings—is an effective and highly prevalent format option (Dupuy et al., 2003; Hartley, 2004). Structured abstracts may be mapped or (as in this journal) unmapped. Summaries can also include bulleted lists, and summaries—especially non-specialist synopses and other kinds of supplementary summaries—may consist entirely of bulleted lists (as does this journal’s “Practitioner’s Takeaway”). Another format option is to follow the longstanding convention—arguably dysfunctional—of limiting research article abstracts to a single paragraph when paragraph breaks might otherwise be expected.

Although unbounded summaries, by definition, have no special formatting to distinguish them from the surrounding text, they may have their own internal formatting, such as division into bulleted lists or boldfacing of key terms. Furthermore, it is possible that writers might take the unusual step of bounding summary content that appears within the body of a text. The summarizing portion of a document’s conclusion might be boxed. Also, looking back to the (imagined) book chapter explaining Joseph Browne’s beliefs regarding infectious diseases, we can imagine the author distinctively formatting the embedded summary of Browne’s Practical Treatise on the Plague.

In some cases, format decisions have stylistic implications. Swales and Feak (2009, p. 12) note that structured abstracts and, presumably, summaries with bulleted lists contain more sentences that introduce new topics. The notion of format can be extended to digital behaviors such as hyperlinking and pop-up windows that make possible entirely new kinds of summaries.


We can plausibly demonstrate how this scheme, used as a design heuristic, can benefit both less sophisticated and highly sophisticated writers, and how it can benefit publications managers, content strategists, editors, designers, and writing instructors:

  • An early-career computer consultant is submitting a report to his client and is writing an abstract that must conform to the client’s publication guidelines. The report is built around five technical problems the consultant has uncovered. Because the abstract is limited to 250 words, he has contextualized the report in the first 50 words of the abstract and now has described each of the five problems in a sentence or two (about 30 words). But he has 50 words left over. What to do with the 50 words? There is nothing helpful in the client’s guidelines. The consultant’s instinct is to describe each of the problems at the same level of detail, and so he is considering making the abstract 50 words shorter than the guidelines allow for. However, recalling the Proportion descriptor, he is willing to depart from accurate proportion, and he uses his remaining 50 words to write lengthier explanations of two of the problems, the two that will benefit significantly from further elaboration.
  • A researcher who is following author guidelines that call for the mapping of summaries recalls that “smoothing out” the organizational complexities of the summarized text is often necessary. The researcher, therefore, decides that the guidelines allow for some leeway in this regard.
  • Another researcher, reviewing the entire scheme, stops at the Stance descriptor to perform a quick check to ensure that she has not inadvertently accentuated the stance.
  • A publications manager, writing the summary section of the organization’s style guide, includes the guideline that authors should raise the threshold for “importance” so as to include fewer ideas rather than pack in a larger number of sketchily explained ideas.
  • Entrepreneurs planning a new company that will publish digital books are looking for innovations that will give them a competitive advantage. After perusing this scheme for ideas, they are considering a design that includes multiple within-text summaries consisting of numbered summary items. When the reader selects a summary item, the first word of the corresponding portion of the text is highlighted so that readers can easily locate that part of the text. The entrepreneurs are debating whether to employ informative phrasing, descriptive phrasing, or a combination for the summaries.
  • A writing instructor decides to use the heuristic as the basis of both a general discussion of summaries and classroom activities. As an activity, the instructor asks students to write a 500-word summary that attempts to be a microcosm of the summarized text (or else provides such a microcosmic summary). Then, she asks the students to successively reduce word count while taking note of how the summary changes—other than in length—at each level of reduction. Alternatively, she can provide a summary and text and ask students to experiment with different placements, change a summary’s mapping, or make other changes prompted by an examination of the descriptors.


Summaries take many forms, serve numerous purposes, and often present significant challenges to writers, editors, and publications managers. The scheme of twelve descriptors presented here advances our understanding of summaries and summarization and, serving as a design heuristic, can directly assist in the better design and writing of summaries. As a basis for further research, it could, for example, be used to formulate research studies on the effectiveness of various forms of summarization in specific rhetorical settings. Such studies might determine how readers respond to summaries written at different levels of reduction, or with different structures, or in different formats. Or, from the standpoint of the design and writing process, studies might examine how a change in one of the descriptors prompts writers to make other changes.


American National Standards Institute/National Information Standards Organization (ANSI/NISO). (2015). Guidelines for abstracts [ANSI/NISO Z39.14-1997 (R2015)]. http://groups.niso.org/apps/group_public/download.php/14601/Z39-14-1997_r2015.p%E2%80%A6

American Society for Microbiology. (2019). Journal of Bacteriology/For Authors/Organization and Format. https://jb.asm.org/content/organization-and-format

Axelrod, R. B., & Cooper, C. R. (2012). Concise guide to writing (6th ed.). Bedford/St. Martin’s.

Ayers, G. (2008). The evolutionary nature of genre: An investigation of the short texts accompanying research articles in the scientific journal Nature. English for Specific Purposes, 27, 22-41.

BBC News. (2002). India-Pakistan: Troubled relations [Online multimedia]. http://news.bbc.co.uk/hi/english/static/in_depth/south_asia/2002/india_pakistan/timeline/default.stm

Berkowitz, T. (2018). The evolution of sales development reps to support B2B demand generation (conference abstract). Gartner Sales & Marketing Conference. Las Vegas, NV, October 9–11. https://www.gartner.com/en/conferences/na/sales-us/speakers/featured-speakers/todd-berkowitz

Berwanger, O., Ribeiro, R. A., Finkelsztajn, A., Watanabe, M., Suzumura, E. A., Duncan, B. B., Devereaux P. J., & Cook, D. (2009). The quality of reporting of trial abstracts is suboptimal: Survey of major general medical journals. Journal of Clinical Epidemiology, 62(4), 387–392.

Breeze, R. (2016). Tracing the development of an emergent part-genre: The author summary. English for Specific Purposes, 42, 50–65. https://doi.org/10.1016/j.esp.2015.11.003

Brown, A., & Day, J. D. (1983). Macrorules for summarizing texts: The development of expertise. Journal of Verbal Learning and Verbal Behaviour, 22, 1–4.

Bryan, J. (2018). Help your B2B sales development reps to evolve. https://www.gartner.com/smarterwithgartner/help-your-b2b-sales-development-reps-to-evolve

Buitendag, A. A. K., van der Walt, J., Malebane, T., & de Jager, L. (2012). Addressing knowledge support services as part of a Living Lab environment. Issues in Informing Science and Information Technology, 9, 221–242. https://doi.org/10.28945/1618

Dupuy, A., Khosrotehrani, K., Lebbe´, C., Rybojad, M., & Morel, M. D. (2003). Quality of abstracts in 3 clinical dermatology journals. Arch Dermatology, 139(5), 589–593. JAMA Dermatology. https://doi.org/10.1001/archderm.139.5.589

Farkas, D. K., & Raleigh, C. A. (2013). Designing documents for selective reading. Information Design Journal, 20(1). 2–15. https://doi.org/10.1075/idj.20.1.01far.

Green, D. (2015). Writing an executive summary. (Oxfam Research Guidelines). https://oxfamilibrary.openrepository.com/bitstream/handle/10546/579649/ml-writing-executive-summary-051115-en.pdf.pdf?sequence=6&isAllowed=y

HHS Web Standards and Usability Guidelines. (2019). https://webstandards.hhs.gov

Hartley, J. (2004). Current findings from research on structured abstracts. Journal of the Medical Library Association, 92(3), 368–371.

Heard, S. B. (2016). The scientist’s guide to writing: How to write more easily and effectively throughout your scientific career. Princeton University Press.

Holtz, M. (2009). Nominalization in scientific discourse: A corpus-based study of abstracts and research articles. In M. Mahlberg, V. González-Díaz, & C. Smith (Eds.), Proceedings of the Corpus Linguistics Conference CL2009 (article 341). http://ucrel.lancs.ac.uk/publications/cl2009/

Hyland, K. (2005). Stance and engagement: A model of interaction in academic discourse. Discourse Studies, 7, 173–192.

JAMA Network. (2019). Instructions for authors. https://jamanetwork.com/journals/jama/pages/instructions-for-authors

Jordan, M. P. (1988). How can problem-solution structures help writers plan and write technical documents? In L. Beene, & P. White (Eds.), Solving problems in technical writing (pp. 3–26). Oxford University Press.

Kardash, C. A. M., & Noel, L. K. (2000). How organizational signals, need for cognition, and verbal ability affect text recall and recognition. Contemporary Educational Psychology, 25, 317–331. https://doi.org/10.1006/ceps.1999.1011

Kintsch, W., & van Dijk, T. A. (1978). Toward a model of text comprehension and production. Psychological Review, 85, 363–394.

Lemarié, J., Lorch, R. F., Eyrolle, H., & Virbel, J. (2008). SARA: A text-based and reader-based theory of signaling. Educational Psychologist, 43, 27–48. https://doi.org/10.1080/00461520701756321

Li, Y. k., & Hoey, M. (2014). Strategies of writing summaries for hard news texts: A text analysis approach, Discourse Studies, 16(1), 89–105. https://doi.org/10.1177/1461445613496356

Lin, J. (2009). Is searching full text more effective than searching abstracts? BMC Bioinformatics, 10(46). doi:10.1186/1471-2105-10-46.

Lloret, E., & Palomar, M. (2012). Text summarisation in progress: A literature review. Artificial Intelligence Review, 37, 1–41. https://doi.org/10.1007/s10462-011-9216-z

Lorch, R. F., & Lorch, E. P. (1996). Effects of organizational signals on free recall of expository text. Journal of Educational Psychology, 88(1), 38–48.

McEnery, T., & Hardie, A. (2012). Corpus linguistics: Method, theory and practice. Cambridge University Press.

Mayer, R. E., Dyck, L. K., & Cook, J. L. (1984). Techniques that help readers build mental models from scientific text: Definitions pretraining and signaling. Journal of Educational Psychology, 76(6), 1089–1105.

Meyer, B. J. F., & Poon, L. W. (2001). Effects of structure strategy training and signaling on recall of text. Journal of Educational Psychology, 93, 141–159.

Mogull, S. A. (2018). Scientific and medical communication: A guide for effective practice (ATTW series in technical and professional communication). Routledge.

Montgomery, S. L. (2003). The Chicago guide to communicating science. University of Chicago Press.

Orr, M. (2010). Intertextuality. The encyclopedia of literary and cultural theory. https://doi.org/10.1002/9781444337839.wbelctv2i002

Paradis, J., & Zimmerman, M. (2002). The MIT guide to science and engineering communication (2nd ed.). MIT Press.

QuikScan Project website. (2019). Retrieved from https://www.quikscan.org

Ritchey, K., Schuster, J., & Allen, J. (2008). How the relationship between text and headings influences readers’ memory. Contemporary Educational Psychology, 33(4), 859–874. https://doi.org/10.1016/j.cedpsych.2007.11.001

Samraj, B. (2005). An exploration of a genre set: Research article abstracts and introductions in two disciplines. English for Specific Purposes, 24(2), 141–156.

Sherrard, C. (1989). Teaching students to summarize: Applying text linguistics. System, 17(1), 1–11.

Stotesbury, H. (2003). Evaluation in research article abstracts in the narrative and hard sciences. Journal of English for Academic Purposes, 2(4), 327–341.

Supatranont, P. (2012). Developing a writing template of research article abstracts: A corpus-based method. Procedia – Social and Behavioral Sciences, 8th International Language for Specific Purposes (LSP) Seminar, 66, 144–156. https://doi.org/10.1016/j.sbspro.2012.11.256

Swales, J. (1990). Genre analysis. Cambridge University Press.

Swales, J., & Feak, C. B. (2009). Abstracts and the writing of abstracts. Volume 1 of the revised and expanded edition of English in today’s research world. Michigan Series in English for Academic & Professional Purposes. University of Michigan Press.

van der Meij, H., & van der Meij, J. (2011). Improving text recall with multiple summaries. British Journal of Educational Psychology, 82(4), 257–69. https://doi.org/10.1111/j.2044-8279.2011.02024.x

van der Meij, H., van der Meij, J., & Farkas, D. K. (2013). QuikScan formatting as a means to improve text recall. Journal of Documentation, 69(1), 81-97. https://doi.org/10.1108/00220411311295333

van Dijk, T. A. (2003). The discourse-knowledge interface. In R. Wodak, & G. Weiss (Eds.), Critical discourse analysis: Theory and interdisciplinarity (pp. 85–109). Palgrave.

Westby, C., Culatta, B., Lawrence, B., & Hall-Kenyon, K. (2010). Summarizing expository texts. Topics in Language Disorders, 30(4), 275–287.

Wurman, R. S. (1989). Information anxiety. Doubleday.

Young, R. E., Becker, A. L., & Pike, K. L. (1970). Rhetoric: Discovery and change. Harcourt, Brace & World.

Zhou, Q., & Farkas, D. K. (2010). QuikScan: Formatting documents for better comprehension and navigation, Technical Communication, 57(2), 197–209.


David K. Farkas is Professor Emeritus in the Department of Human Centered Design & Engineering at the University of Washington. He has published on the editing process, the structure of expository texts, procedures, software user assistance, slideware, and website navigation. He is an STC Fellow and a member of the Puget Sound STC chapter. His email is farkas@uw.edu, his professional website is https://faculty.washington.edu/farkas, and he posts his publications on ResearchGate. He has been publishing in this journal since 1981.