James Hartley, Keele University
Keywords:abstracts; scientific communication; written communication; readability; writing
-------------------------------------------------------------------------------------------------------------------
ABSTRACT
BACKGROUND. Structured abstracts (which use
sub-headings like this one) have now become widespread in medical research
journals.
AIM. The aim of this paper is to consider
whether or not such structured abstracts can be used effectively in social
science journals.
METHOD. The paper reviews a selection of studies
as carried out by the author and his colleagues to see if structured abstracts
written for social science journals are more informative, easier to read,
and easier to search than are their traditional equivalents.
RESULTS. The results suggest that structured
abstracts do, indeed, have these virtues, and that they are thus appropriate
for social science journals.
CONCLUSION. The author therefore recommends that the editors of social science journals consider adopting structured abstracts.
-----------------------------------------------------------------------------------------------------
Readers of this article will notice that the abstract that precedes it is set in a different format from the one normally employed in Learned Publishing. The abstract to this present article has a structured format. Such structured abstracts (which contain sub-headings - such as background, aims, methods, results and conclusions) have now replaced traditional abstracts in most medical research journals, and they have been subjected to considerable research in this context. This research has shown that structured abstracts:
· are of a higher quality 1
· facilitate peer review 3,4 and
· (still) sometimes omit important information 1,5.
-----------------------------------------------------------------------------------------------------
Table 1 Changes in the frequency, length and
informativeness of abstracts
published in a selection of science journals over
time.
(Data adapted with permission from Huckin 6.)
---------------------------------------------------------------------------------------------------
Year
No. of Number of
Average number Average number of
Journals abstracts
of sentences
results statements
----------------------------------------------------------------------------------------------------
1944
5
11
4.4
2.6
1959
7
36*
5.6
4.7
1974
11
110
6.0
4.7
1989
11
120
6.5
5.0
*Includes one abstract with 39 sentences and 38 statements
of results.
-----------------------------------------------------------------------------------------------------
Recent studies in linguistics 8, 9 have shown that - in general - authors have much the same goals in writing abstracts in different subject disciplines. Dos Santos 9, for example, writes:
In the last few years or so, my colleagues and I have been examining how far it might be possible to include such structured abstracts in social
science journals as well as in medical ones. Many papers in social science journals report experimental data, and - with a moderate adjustment of the sub-headings - qualitative papers, review articles and discursive papers can also be accommodated (as shown in the abstract to this article).
In this article I present an overview of a selection of our studies to date. Basically we have repeated the kinds of studies done in the medical literature, and we have extended the range of questions asked. We have, for example, considered the effectiveness of different typographic designs for structured abstracts, we have looked at their readability, and we have also examined whether or not it is easier to search and to find information in structured abstracts.
Writing structured abstracts
In order to carry out our research we first had to create a pool of traditional and structured abstracts - as none existed in this context when we began. To do this we used thirty traditional abstracts, fourteen of which had been prepared for an earlier study 10 . These abstracts were based on original abstracts taken from consecutive issues of the British Journal of Educational Psychology published in 1992 and 1993. To form the matching structured abstracts we re-wrote each of these traditional abstracts in a structured format, using the sub-headings Aim, Method and Results (and sometimes, Comment).
In doing this we found that the main changes to the traditional abstracts came in the Method sections. Here more detail was required. Many of the traditional abstracts did not give details of, for example, sample sizes, numbers of comparison groups, the timing of delayed tests, etc.. The Results sections, too, were often weak in this respect.
We found that most of our original abstracts could easily be recast in the structured format, but that we nearly always had to read the accompanying articles to obtain the necessary data to write the Method and Results sections. Sometimes this procedure revealed results that were, in our view, at odds with what was stated (or more likely left out) of the original traditional abstracts. However, we only had to re-write one abstract completely to make it fit our structured format.
Thus, for our initial experiments, we had thirty traditional and thirty structured abstracts. (In our later studies 11 we were able to use abstracts from four British psychology journals that now publish them.) None of these thirty abstracts was fabricated, but most were simplified/shortened from the originals to make them easier for non-specialists to understand. This was achieved for both forms of abstract by:
· simplifying the technical vocabulary;
· reducing the number of conditions reported in experiments; and
· simplifying the results.
Presentation
The structured abstracts published in medical journals come in a variety of typographic formats. Not only do the positioning and the typographic settings of the abstracts vary generally, but there is also a wide variety of typographic detailing used to denote the sub-headings. Thus the sub-headings may be printed in bold, bold capitals, italic, italic capitals, bold italic capitals, etc. Sub-headings may start new lines, be indented, or just run-on.
Our approach was to start by asking readers for their preferences for subsets of possibilities drawn from a pool of 28 different typographic settings, and then to 'home-down', as it were, on the most preferred versions. We eventually found that our readers (academics, research workers, postgraduates and undergraduate students) preferred structured abstracts (i) to be set apart by printers' rules from the body of the text, (ii) to have the sub-headings printed in bold capital letters, (iii) to have each sub-section start on a new line, and (iv), if possible, for each sub-section to be separated by a line-space. Furthermore, our readers preferred their abstracts to be centred across the top of a single- or a double-column format, and not to be positioned at the top of the left-hand column in a double-column format. The abstract to this present article illustrates these preferences. (Hartley and Sydes 12 provide further details of this study.)
Readability
We were impressed when we were writing our own structured abstracts by how far using the sub-headings as a guide seemed to make the resulting abstracts easier to write - and possibly easier to read. We tested out this last possibility by assessing the readability of traditional and structured abstracts in various conditions (see 11,13). To cut a long story short, we found no significant differences between the readability of traditional and structured abstracts (as measured by computer-based readability formulae) when we compared traditional and structured abstracts published before and after the introduction of structured abstracts in the British Medical Journal, the British Journal of Psychiatry and the British Journal of Educational Psychology. However, we found that when authors were asked by editors or conference organisers to re-write traditional abstracts in a structured form, then the resulting abstracts did in fact score significantly better on measures of readability.
Furthermore, we may note here that although some of our traditional and structured abstracts were equally difficult to read (as measured by readability formulae) our readers rated the structured versions as significantly more readable 13 . This, of course, stemmed from their clearer spatial organisation 10, 14.
Search
Because structured abstracts convey their internal structure more clearly than do traditional ones, we would expect readers to be able to search structured abstracts more easily than traditional ones. We conducted two studies to test this proposition (see 15). First of all we compared the abilities of readers to search traditional and structured abstracts presented in an electronic database, and then we carried out a similar study with a printed database.
For the electronic database, undergraduate psychology students were presented with an abstract on screen, together with a question about it. When they had indicated their response - 'yes' or 'no' by pressing an appropriate button - they were then presented with a second question. After responding to the second question a new abstract appeared, and the process continued until eight abstracts had been completed. The times taken to do these tasks were recorded electronically.
For the printed database students were timed searching for particular abstracts that reported studies using a particular sample (e.g. university students) and a particular methodology (e.g. questionnaires). Here the total time to find five abstracts - in a set of 29 - and the error rates were recorded by hand.
In both studies we found that the participants searched the structured abstracts more quickly than the traditional ones, and that they made fewer errors.
Conclusions
Our studies show that it is possible to produce structured abstracts for journal articles in the social sciences. As noted above, we had little difficulty in re-writing traditional abstracts in the structured format - although we often had to read the original article in order to obtain the information required under each sub-heading.
On the basis of our experience we recommend the use of six sub-headings - Background, Aim, Method, Results, Conclusion and (optionally) Comment. We think that these six sub-headings can be used with review and discursive articles as well as with scientific reports. We also recommend that these sub-headings are conveyed in bold capital letters (as shown in the abstract to this article). Furthermore, if the abstract can be set apart from the text by the use of printers' rules, and if the text for each sub-section can be set apart by the use of a line-space, then so much the better.
In line with the earlier medical research we found that our structured abstracts were typically longer than the traditional ones. In our recent study of thirty abstracts actually published in four British psychology journals 11, the structured abstracts were on average 33% longer than the traditional ones.
With this extra text, of course, comes extra information. Analyses of the gains from the traditional to the structured formats in this study showed that most increases in information lay under the sub-heading ‘Participants’ . Here 40% of thirty structured abstracts had higher scores than did the traditional ones. The next highest gains were for ‘Background’ (37%), ‘Aims’ (33%) and ‘Conclusions’ (33%). Details of ‘Method’ did improve (15%), but the ‘Results’ were little different from each other (7% gain). These findings show, like those from medical journals, that structured abstracts are more informative than traditional ones.
However, the fact that the typographic settings of the structured abstracts are more spacious, together with the fact that structured abstracts are typically longer, has implications for journal editors and printers. If journal space is at a premium, and the journal circulation is large, then it might be expensive to introduce structured abstracts into printed journals. (This problem does not arise for electronic ones.) However, most of the printed social science journals that we have examined (including Learned Publishing) allow themselves the luxury of starting new articles on a fresh right-hand page, and thus there is considerable room for manoeuvre. The issue, therefore, is one of cost-effectiveness.
It is my belief, of course, that structured abstracts are an improvement on traditional ones - and hence are cost-effective. I thus recommend that editors of social science journals seriously consider publishing them.
Acknowledgments
Much of the research outlined in this paper was supported
by a grant from the British Library, to whom I am indebted. The views expressed
in this paper are the views of the author and not necessarily those of
the British Library.
References
Taddio, A., Pain, T., Fassos, F. F. et al. Quality
of nonstructured and structured abstracts of original research articles
in the British medical Journal, the Canadian Medical Association
Journal and the Journal of the American Medical Association. Canadian
Medical Association Journal 1994: 150 (10) 1611-1615.
E-mail j.hartley@keele.ac.uk