astronomers vs. thirty seconds

This year at the Astronomical Society of Australia annual meeting in Adelaide, significant talk oversubscription meant that there were over 120 posters presented by both students and supervisors alike. The goal of a scientific poster is to summarise your work (possibly the result of many years research) in a brief but clear, scientific but visually appealing way - no easy task! Last year in Hobart, the optional choice of giving a poster sparkler was met with pretty positive reception and so this year everyone was given a poster sparkler slot. This slot amounts to 30 seconds in which to show a one-slide summary of your poster (in some cases, the poster itself) and promote your work, hopefully resulting in other researchers coming to talk to you in the breaks about it. It's not about trying to explain your work in detail, of course - it's about trying to get people interested! Sparklers are a neat idea, with the strict time limit really making you consider how best to say what you want to say and also testing how quickly you can say it.

But how well do astronomers do at keeping to the time limit?

We decided to time every sparkler, with the noble goal of investigating an astronomer's concept of time. Is it to the second, or maybe an order of magnitude? There's no easy way to quantitatively judge the quality of a sparkler talk, so we went for the stopwatch option and assembled a database consisting of all 119 sparklers that were given (not all chose to deliver one), easily divisible into students (57) or non-students (62) for an additional parameter separation. Note that non-students are further in their careers than students and hence more senior, research-wise. If you are interested in errorbars, the timer was able to resolve to the nearest second so you can assume about ±0.5 seconds of error on the times. And for the statistically inclined, the median student time was 27±6 seconds and the median non-student time was 27±8 seconds; this is also evident in the histograms below.


large scale structure

This histogram is binned in steps of 5 seconds, and so shows the overall trends of the distribution. Students are shown in red, non-students in blue, and the entire distribution in grey. You can see from this that students hover most dominantly around the 30 second mark, with some students finishing before and others accidentally going a bit over. Non-students, on the other hand, either finish much quicker than 30 seconds or go significantly over. This may be an indication that students prepare more rigidly for exactly 30 seconds, whereas non-students might say whatever they think needs to be said, whether it be over or under 30 seconds.

small scale structure

Interestingly, if we bin instead in steps of 2 seconds, a bimodal distribution appears. Very few sparklers fall around 25 seconds; instead, they either cluster at 20 or 30 seconds with the tail of the distribution reaching as high as 45 seconds. As seen in the above histogram, the first peak is dominated by non-students while the second peak is dominated by students. Special mention goes to those few sparklers (3) that were 15 seconds or less, it may be worth noting that these were also the rhyming sparklers.

Is the student's concept of time better than the non-student's? One might be tempted to conclude this since the students clumped around 30 seconds with very little spread compared to the non-students. But perhaps this is also an indication that students are more likely to follow a time restriction, while a significant number of non-students proved that you can say what you need to say in only 20 seconds. Maybe next year the sparkler talks should be shortened to 20 seconds in order to assess whether this is true or not, and a repeated distribution analysis be carried out. Based on the extremely positive reception of those few who chose to do an interpretive rhyme or song about their research, we make the strong recommendation that all poster sparklers be performed similarly at next year's ASA in Sydney.

30/09/2011 update
This pioneering study was recently continued at another conference! Here are the results, thanks to he who likes high-mass stars:

He offers, "Here's the histogram of the length of sparkler talks from the Rome conference, as requested. Notice that once again, there is a bimodal distribution of speakers. One group tended to go overtime by about 5-10 seconds, the other tended to blow out their time by 30-40 seconds." So it may be that this bimodal distribution we saw in the Adelaide conference was statistically significant and not unique to Adelaide!

Unless otherwise stated, the content of this page is licensed under Creative Commons Attribution-ShareAlike 3.0 License