Thank you for your comments on our paper. We appreciate that you and the journal club have taken the time to read and critique our work. We hope that all of your questions are addressed sufficiently below, and we welcome further discussion!

1. In Fig. 4, was a DIP test for unimodality performed? Is that test applicable in this situation?

No tests for unimodality were performed, and there are no doubt more sophisticated ways to choose the threshold between what seemed to be the ‘near-zero’ distribution and the ‘finite speed’ distribution. Although the distributions do appear to be a mixture, you correctly point out that in order to say definitively whether a mixture actually exists, one should perform a statistical test. We visually chose a threshold merely for simplicity and for generality.

2. It never became entirely clear why values below the cutoff were not included in many evaluations. Why wasn’t it possible to set all the data below the noise cutoff to zero and include them in the analysis?

The values below cutoff were included in all of our calculations. If you refer to the section “Segmentation of Velocity”, we labeled these as ‘stops’ and set the velocity to zero. We don’t show them in the probability distribution plots simply for clarity, since there were quite a lot of points that fall into this single bin (refer to Fig. 6).

3. Why weren’t more arbitrary zones included, e.g. a rim zone of only .5cm? Or three different zones instead of two? The graph in Fig. 9 suggests the zones were specifically chosen to match the length of stay in each zone.

In this paper, we wanted to stress that there is no need to delineate arbitrary spatial zones. You can draw any sort of zone that you want, but that doesn’t make it behaviorally relevant. We chose the location of the zones based solely on the behavior of the fly, as illustrated by Fig. 3 and discussed in the text. In effect, the fly “chose” the zones. Although Fig. 9 shows approximately equal lengths of stay, the zones were not chosen specifically to cause this result. This is a consequence of the fly’s behavior.

4. Why were the speed-measurements of this animal not compared to published speed data for walking flies?

As stated in the introduction, “Because only a single fly is studied in this paper, it should be noted that the results are not meant as generalizations of fly behavior. The emphasis in this paper, rather, is on the methods used for analysis of the trajectory.” Our aim for this paper was to show the utility of examining these joint probability distributions in obtaining a quantitative characterization of the fly’s exploratory behavior. To our knowledge, these joint distributions have not been used in this context. Admittedly, it is entirely possible that our fly behaved completely uncharacteristically, walking much faster or slower than normal, but that does not negate the metrics used to present these data or their method of calculation. In addition, we have only seen average speeds quoted in the literature. The shape of the distribution has not been previously mentioned, and so the average speeds can be quite misleading from a behavioral standpoint. That being said, from the data we have seen in the literature for arenas of similar sizes, our results seems to be on par with what to expect.

5. What was the color of the arena, the material and the lighting? Was special care taken that the lighting exceeded the flies’ flicker fusion frequency of ~200Hz? What was the height of the chamber the fly was enclosed in?

We used off-white Perspex arena 15cm diameter by 0.7cm with a circular 40W neon lamp placed around the video camera, and a silk paper cone covering the arena to evenly spread light and shelter the fly from distal visual cues.

6. What was the sex and age of the fly and when was it last fed? When in its circadian activity cycle was it measured?

The fly was a Canton-S, 3 days old male, last fed immediately before the onset of the experiment, and videotaped at the morning part of the daylight part of the circadian cycle.

7. Some members of the journal club even expressed their dissatisfaction with the strategy of just measuring a single fly over two hours, analyzing some of the data and then publishing a paper. They argued that it is difficult to assess the usefulness of the authors’ evaluations due to the lack of comparisons both within the paper (by including more flies) and to published data. Concerns were raised as to what the smallest publishable unit in PLoS One will be.

A reasonable argument, but again, we want to stress that this paper was meant solely to comment on the utility of the techniques—we were not trying to obtain a refined model of a fly’s behavior. We see large benefits in treating the trajectory as a stochastic process and examining joint distributions as summaries of this behavior. These techniques are of widespread use in describing particle trajectories in statistical physics, and it was of our opinion that many researchers in the Drosophila research community (and other animal locomotor studies) may find these techniques useful.

In addition, not only do these distributions yield information about the fly, but they also provide a wealth of information about the environment as measured by the fly. Therefore, they can also be used as diagnostics to find systematic errors in a particular assay (e.g. a tilting of the camera, misalignment of lighting, etc.). So whereas multiple fly statistics may certainly refine what these distributions look like, it will not change the fact that one should look at them, nor does it change how one should calculate them, which were the main points we emphasized in this paper. To illustrate how to use/calculate the metrics, we felt that analysis of two hours from a single fly was sufficient.

Because PLoS ONE emphasizes fantastic interactive discussions like this one, and quickly disseminates the information, we felt it was the best venue for publishing this work.

Sincerely,

The Authors