There is a very common assumption among researchers preparing proposals for the ERC calls that the text must have graphics and photos and figures liberally embedded within it for it to be easy to read and evaluate. However, my experiences of reading these proposals and often stumbling over and struggling to make sense of the visuals led me to think that lots of researchers felt under pressure to include graphics of some kind or other in commonly held belief that they are easier to read and evaluate but didn’t really know how make them very well or understand what a graphic is capable of. Recent experience of project reviews for the latest open calls for proposals have done nothing to change my view that there are too many confusing and distracting graphics being developed and that to get a good one in the text that adds some value and says something in a neater and more efficient way than it could be said in words is actually pretty difficult and rare.
I shared a platform recently at an ERC proposal development training event with a colleague from a company dedicated to teaching researchers how to develop visuals that make a difference in their scientific writing in general and in proposals in particular – so, specialised training does exist in this area that those struggling with effective visual communication could make use of and it is quite clearly a recognised problem that can be solved. The ideas of this expert in visual communication and graphics reflected my own experience quite well. He showed that over-complicated figures that try to convey too many ideas and cannot be read in simple and logical ways are a barrier to understanding rather than a way of improving the clarity and simplifying the presentation of the ideas. It can be as simple as there being no obvious starting point to a swirling mass of arrows and data, but anything that makes a hard-pressed evaluator think and wonder and puzzle is a negative and to be avoided at all costs.
A very interesting research paper has been published recently[1] which helps to throw some light on the potential benefits of figures and graphics and which types used at what frequency might be most effective when creating high impact proposals. A research team from the University of Washington set out to find out whether papers with figures in them reach larger audiences, are read more widely and as a result become more influential.
After training a computer algorithm to distinguish between different types of figures with a high level of accuracy the team assessed the PubMed Central archive of biomedical and life sciences research articles which contains more than 650,000 papers containing more than 10 million figures (once the component parts of multichart figures had been broken out).
The types of figures the team defined were diagrams, equations, photographs, plots (e.g., bar charts and scatter graphs) and tables. The most common type of figures turned out to be data plots which comprised 35% of the total, followed by photos (22%), diagrams (20%), equations (17%) and tables (5%). The team used Eigenfactor scores to measure a paper’s influence – this is a powerful scoring system that gives weights to the citations rather than simply noting the number so that a citation in a highly cited paper counts for more.
The research showed that figures do in fact appear to make a difference. Papers with more diagrams than average tended to be more influential and high impact ideas tend to be conveyed visually. An average paper in PubMed Central has one diagram for every three pages and scores 1.67 citations. Papers with more diagrams per page got better results; on average a paper scored two more citations per extra diagram per page. Plots were also effective as papers got one more citation for every extra plot per page. In brief, the authors conclude that “The key result is that higher proportions of diagrams are linked to higher impact, while high proportions of photos are linked to lower impact”
The team say there are two possible explanations for this effect: “That visual information improves the clarity of the paper, leading to more citations, and higher impact, or that high impact papers naturally tend to include new, complex ideas that require visual explanation”
It is interesting to note that certain types of figure had a negative effect. Including photographs and equations seemed to reduce the chance of a paper being cited by others. I have long used the example in the training courses on how to write for the ERC of a very turgid proposal that sticks in my mind which opened up with complex equations on the very first line of the opening paragraph to illustrate how not to start a persuasive argument. It seems this research bears that out more generally : use equations carefully and probably right down in the technical sections of the methodology descriptions to avoid deterring readers. A previous study by the same research team found that in a dataset of over 600 biology papers each additional equation per page reduced the number of citations by 22%.
The authors have not shown the causal relations between increased use of diagrams and citations, in fact, exploring this link is part of the future research agenda they have set themselves: “Next steps will be refining this question and interpreting these preliminary results to understand how figures influence impact.” Papers with lots of diagrams for example might be exploring issues on the edge of knowledge and be capturing lots of new information in them which would make them very likely to be highly cited. And the negative effect of equations is possibly very field specific – the biologists which are the audience of the PubMed Central database papers are likely to find complex maths more off putting than physicists, for example.
The research team recognise the probable limitations of their focus to date on life science and biomedicine and plan to widen the research to include the physical sciences. They also want to draw better conclusions about how the different types of diagram convey information and begin to make the process of creating scientific diagrams more of a science in itself. They have coined a term for this new science of visual information, the study of “how ideas are communicated” – ‘viziometrics’ (to convey the shared goals with bibliometrics and scinetometrics). They have created a searchable database of the figures used in their research work at www.viziometrics.org which allows scientists to search the scientific literature in a new way using search terms to call up all the stored visual data dealing with that theme.
I think the key idea to take from this fascinating paper in relation to writing ERC proposals is to be found in their conclusions when they try to interpret the fact that high impact papers have a higher proportion of diagrams relative to other figure types. They write, “A possible interpretation is that clarity is critical for impact: illustrating an original idea may be more influential than quantitative experimental results.” ERC projects are by definition dealing with original ideas, things that are being uncovered, evaluated and controlled for the first time. It is very likely, then that these are the types of pioneering projects which will benefit from excellent figures at higher densities than average for proposals.
Currently my experience of reading some hundreds of these texts is that most don’t contain any figures and that if they do then they are complex flow diagrams of the total project concept with background, objectives and process all represented in a compound of difficult-to-read visual information. I take from this paper that a number of excellent graphics capturing each of the key ideas or categories of ideas in the proposal would make the texts clearer and more competitive.
However, the research paper itself doesn’t assess the readability or otherwise of the graphics and the correlation seems to be between number, type and impact without reference to any external evaluation of the quality of the figures: but I am assuming that high numbers of good ones will do a better job in ERC proposals than the same number of opaque ones. I am also assuming here, of course, that we can use greater readability and wider readership which leads to greater impact with the openness and simplicity that winning proposals have i.e., behaviours in making high impact papers will also help make winning proposals and that the best and most communicative visual elements will make the greatest impact of all.
Clearly, the figures will need to be high quality and add something beyond the job that words could in the same space. Researchers will have to learn how to do their visual work better as a bad diagram is far more difficult and ambiguous than simple, reasonably well-structured text. So, the tendency that I see in many proposals to put in diagrams even if they don’t mean very much, are too large and difficult to read might actually be tapping into the right idea even if it is not managed or exploited very well in many instances. There is clearly some competitive advantage to using graphics intensively and expertly to capture complex ideas at every stage of the project from setting out the objectives and explaining expected results as well as the methods in place of large complex and compound graphics that try to do that all in one go – reduce the complexity and increase the number would seem to be a good strategy in light of these first viziometrics findings.
There are many ways to improve the attractiveness and readability of graphics in proposals. As I mentioned at the start of this post, there are companies that specialise in training scientists to create better graphics and I am sure this would make a beneficial difference for many. There are also lots of resources online – three to explore are mindthegraph, Canva and Piktochart but there are many, many others out there. After reading this new viziometrics research I have come to think that figures play a more critical role in successful communication and, therefore, in proposal writing than before and it appears that higher numbers of high quality graphics have an influential part to play in tipping the balance in your favour. It is probably one of those things that makes a slight difference – but all advantages need to be exploited and in the end the difference between funding and not is a percentage point – they have to make the cut somewhere so all fragments of marks need to be carefully marshalled and maximised.
[1] Viziometrics: Analyzing Visual Information in the Scientific Literature. arXiv:1605.04951v2 [cs.SI] 27 May 2016. http://arxiv.org/pdf/1605.04951v2.pdf.