Effective Abstracts in Science Class

T. R. Girill
Technical Literacy Project
trgirill@acm.org

Core Characteristics and Skills

Abstracts are one of those rare literary forms important in real life as well as just in school (more on that at the end). An abstract summarizes a paper (or, in real life, a published article) in 250 to 500 words, but in a special way. Abstracts circulate widely independently of the articles that they summarize (in vast searchable online databases sold commercially or offered by professional societies). Hence astute abstract writers have learned the importance of constructing a compact surrogate for their much larger paper. A good abstract is brief and entirely self-contained, yet someone who has never seen the paper that it summarizes should be able to reliably judge the content and relevance of that paper (to their own interests) by reading the abstract alone.

Abstract writing thus demands sophisticated text design skills:

So teaching these skills to science students presents special challenges as well as unusual practical opportunities.

Field-Specific Features

Most high-school students learn about abstracts (first) in language arts or history class. So what does a good authentic abstract look like in language arts or history? That is an empirical question, which Helen Tibbo answered in her revealing comparative study "Abstracting across the disciplines" (1992). Tibbo performed a sentence-by-sentence analysis of 30 abstracts randomly selected from a "high-impact" (influential, widely cited) American history journal and compared them with 30 abstracts randomly picked from a high-impact journal in chemistry (p. 43).

The history abstracts had this content profile:
Percentage of sentences
in each content category
Categories
History
Background 5
Purpose/scope 9
Hypotheses 1
Methods 5
Results 2
Conclusions 14
None of above 64

Most real-life abstracts in science and engineering, however, are very different from this. Technical abstracts are highly structured and divide their words very consistently by task. Tibbo's analysis of chemistry abstracts, for example, found this pattern:
Percentage of sentences
in each content category
Categories
Chemistry
Background 9
Purpose/scope 24
Hypotheses 0
Methods 28
Results 22
Conclusions 16
None of above 0

In other words, science abstracts strongly reflect the IMRD (introduction [= background], methods, results, discussion [= conclusions]) structure of most science reseach articles, while (even well-written) abstracts in other fields do not.

Few students draft science abstracts consistent with this "chemistry" pattern without special coaching. A likely explanation is that they learned how to craft abstracts from history or language arts teachers mostly familiar with the upper (unstructured) pattern rather than the more structured lower pattern typical in science publications. The drafting sophistication required together with the highly structured division of the text means that your students will probably learn how to write good science abstracts only from their science teacher (if they learn it at all).

Structured Abstracts

Because the structure of science abstracts is so consistent and useful to readers, a few technical journals--mostly in medicine where confusion or ambiguity can have drastic negative clinical consequences--now make the IMRD chunks of their abstracts explicit with imbedded headings. Your students can see other students using this approach (misleadingly call "structured" abstracts instead of "labeled" abstracts) at the California State Science Fair (CSSF) website (go to http://www.usc.edu/CSSF/History for a choice of recent years).

Heading details vary by publication (or year) but the usual pattern involves overtly subdividing the abstract text with these labels:

Objective/Goal(s)
Materials and Methods
Results
Conclusion(s)/Discussion
Even these slight prompts are seldom deployed, however. Most professional journals still leave to the scientist (or engineer) the task of rationing their scarce abstract space among the IMRD topics to adequately represent the work done and signaling the abstract's organization without the help of any displayed headings. Few students do this well even if their lab or field work is strong and interesting.

A Content Checklist for Abstracts

Giving students examples of well-crafted and of ineffective abstracts is the usual practice to encourage good drafting. Unfortunately, this assumes that the students can learn from examples, always a cognitively demanding task and especially so when the content distinctions are subtle. Even poor abstracts are often full of "science content," but that content is inappropriate to thoughtfully represent the corresponding paper or report.

Another way to scaffold abstract drafting is with an itemized list of what science abstracts include and exclude:
When constructing your abstract...
Include
Exclude
Your topic or problem (purpose), but usually not your hypothesis Background information (except to briefly frame the problem)
Scope of work Historical details
Treatment (experimental, theoretical, methodological, practical) Literature review, cross references, footnotes
Novel methods or algorithms
(but always age, sex, genus, species of biological subjects)
Procedural details, apparatus diagrams
Key numerical or statistical results Equations, formulas, data tables, graphs
Significance; interesting conclusions General principles or trends, common knowledge
"Keywords"--searchable terms, distinctions, comparisons that identify your work to others Definitions of technical terms

This chart offers a chance to not only tour the important features for your students, but to explain why they fall on one side of the chart or the other. For example, school science projects often stress testing some explicit hypothesis, but (as Tibbo's study confirms) virtually no published abstracts include this. Instead, they make clear the purpose of the reported work, the specific research or design problem addressed. Readers naturally filter abstracts looking for those whose reported problems link to their own.

Since students spend much time mastering techniques ("methods") new to them but standard in a field, the focus of their work is often strongly methodological. Projects with a theoretical or therapeutic focus should point this out. Biologists always filter the work of others by organism studied (hence the mention of subject genus, species, sex, and age). And every reader wants to know how a project turned out--the key quantitative results and their significance from the author's perspective. Abstracts are searched in large databases by keyword, so smart abstracters take care that their vocabulary contains those terms and distinctions that others are most likely to use during such a search.

The exclusion column surprises some students because most items listed there seem obviously relevant and important. But an abstract needs to devote its scarce space to the work at hand, not to general science principles or background information readily available from reference sources. Almost all of the other exclusions reflect the text-only character of abstract databases: diagrams, equations, graphs, or tables are not plain text (in journals they are often encoded PDF or TeX) and they will not survive sharing in plain text formats. Astute writers learn ways to mention or summarize these otherwise valuable features using text alone (in retrieval-friendly ways).

A Science-Abstract Template

More Space Scaffolding

Explicit as the foregoing checklist is, it still leaves many students confused about how to budget their scarce space in a science abstract and how to implement the checklist's content advice. Those students need a further scaffold of a different kind when they draft an abstract: a top-down guide to space allocation.

I find that this template or "action matrix" meets that need well (click on the chart for a printable PDF version):

Action Matrix for Building a Good Science Abstract
(1)
Allot roughly this percentage of total WORDS:
(2)
To describe this TOPIC:
(3)
Then review the draft and REVISE it to be sure that:
--These subtopics are included (if relevant).
--These questions are addressed.
24% Purpose: what we sought
  • Goal(s) clear and unambiguous?
  • Research/experimental design explicit?
24% Methods: what we did
  • What was actually measured, and how?
  • Any new or improved methods?
  • Steps or sequence clear?
  • Which controls used? Comparisons?
  • Risks, attrition, adverse events disclosed?
  • Subject selection, randomization, limits?
  • Blinding of subjects? Investigators?
24% Results: what we found
  • Key numerical/statistical trends revealed?
  • Outcome assessment well defined?
  • Sample size adequate?
  • Calculations or analysis appropriate? Tests stated?
  • Explains if no results (yet)?
24% Conclusions (discussion): so what
  • What is omitted that could mislead a reader?
  • Terminology inaccurate, inappropriate, needlessly hard?
  • Text too condensed or confusing (for ESL readers)?
  • Disorganized or wrongly organized claims?
4-5% Background: place first but draft last so you don't waste too many words here
  • Focused on problem-framing context?
  • Acronyms explained, jargon minimized?

The matrix assigns to each of the four key topics that a good science abstract includes (IMRD, col. 2) roughly one-quarter of the available total words (col. 1). Although a little "background" comes first, I advise students to draft it last, using a sliver of space saved from each of the other topics. Software makes word counts and percentage calculations easy for anyone now, so students can give themselves simple quantitative feedback as they iterate toward this division of space.

Of course, student projects are often not well balanced among the four IMRD topics--results (or interpretive conclusions) may be light while (newly learned) methods are disproportionately important. So rebalancing the division of space from the even four-way split suggested here can be quite appropriate: the template still provides a disciplined framework from which to start.

More Content Scaffolding

A somewhat different problem is that some students are intimidated by an abstract's space constraints and include too few details about any of their specific work. An abstract is no place for vague generalities, however true they are.

The template's third (right) column addresses this problem. Here for each topic are a few focused questions to prompt students to include relevant details. (Not all questions pertain to all projects, of course, but if they pertain their answers are very interesting to technical readers eagerly searching for work related to their own.) The methods row contains the most questions because (1) students often carelessly overlook their own key methodological details, and (2) methods refinements are often more innovative than results or conclusions in student-level research.

Using this template along with the include/exclude itemized chart more reliably yields good abstracts than using the latter alone. Template cols. 1, 2, and 3 together more thoroughly externalize the drafting actions that students need to try.

The Authentic Payoff

One really satisfying aspect of teaching students how to design effective science abstracts is that they are avidly read by real scientists, not just by teachers grading them in science class. And reading such abstracts often has significant practical consequences.

Henry C. Barry and his colleagues reported a particularly striking example of science abstract influence in their 2001 article in the Journal of the American Board of Family Practice. Barry et al. asked almost 300 family-practice physicians how they would treat corneal abrasions (e.g., patch or not) and how they would treat fibromyalgia. Then these physicians read only abstracts, not whole research articles, summarizing briefly research done by others relevant to those conditions. After reading the abstracts alone, without consulting the nuanced subject-control discussions and applicability warnings in the original articles, 76% of the physicians were willing to change their treatment of corneal abrasions and 73% would change their use of drugs with fibromyalgia patients. Because of time pressures in clinical practice, these doctors routinely scanned abstract databases and adjusted their therapeutic behavior based only on 300-word summaries of medical research done elsewhere by complete strangers.

So unlike many things that your students learn (or should learn) to write in school, science and engineering abstracts have a high impact on other practitioners in real life. Introducing students to checklists and templates that promote abstract effectiveness may therefore increase their influence on others during their professional lives much more than many other superficially cogent lessons or techniques. Abstracts are small but mighty.

References

Barry, Henry C., et al. (2001).
Family physicians' use of medical abstracts to guide decision making: style or substance? Journal of the American Board of Family Practice, 14(6), 437-442. http://www.jabfm.org/content/14/6/437.full.pdf
Tibbo, Helen. (1992).
Abstracting among the disciplines. Library and Information Science Research, 14(1), 31-56.