Here come the science…
First, a warning: The maths may well be crucially flawed (I consulted with statisticians but you know what those guys are like), and my professors did not understand the results well enough to mark it with any degree of confidence. The anonymised data is available at the end of the post, as I would sincerely like statisticians with more resources and brain power than me to take a proper look at them. The results apply to the UK only.
I didn’t want to fall into the trap most graphic design dissertations fall into of taking a sample of 5 graphic designers as my subjects and expecting them to be representative of a whole population – a deeply flawed and far too commonly repeated approach in both design and pyschology studies. So I asked the internet instead. More than 300 of you obliging little blighters replied. I love you guys!
I displayed two things on a screen and asked people to tell me which one represented more people.
Those things were a mixture of graphics, written ratios, and written percentages. Like was compared with like, and like was compared with both of the other options. The questions were randomised so participants never saw the questions in the same order. There were a lot of questions and it got a little dull. I only analysed the results of people who got to the end of the survey, which gave me 223 participants, and for the questions alone over 7,000 datapoints. With demographic information it comes to more like 11,000 datapoints.
You can download the data in Excel format.
I have edited them to take out any identifying data, to merge the demographics columns, and to add in which answers were correct and which were incorrect.
The final dissertation
Download the whole thing, including some analysis of charity communications and the current state of psychological models of risk-perception: The perception of risk in charity health communications. This includes the survey questions (in the appendix at the end) so you can see how the survey was run. It also includes the particular pitfalls of this kind of methodology (and no doubt misses a few out).
Above all, I don’t care if I was right or not – but I would love it if you guys with the power and mathematical skills can look at this and tell me if there’s anything useful there.