Completed on 11 May 2015 by David Schoppik and David Tingley. Sourced from http://biorxiv.org/content/early/2015/04/10/017863.
Login to endorse this review.
Dreosti et. al. study social behavior in young zebrafish, focusing on its development, sensory origin, and sensitivity to drugs. Using a novel apparatus, they evaluate an individual fish’s preference to swim near one of two connected chambers that contains either a “social” stimulus (conspecifics) or remains empty. They quantify kinematic parameters, including time spent in each chamber, orientation relative to conspecifics, and the temporal dynamics of swimming in the presence of a conspecific. Taken together, their results validate an important novel assay of a behavior of broad interest in an appropriate model organism. Further, there are a number of interesting observations that set the stage for future experiments. The manuscript is appropriately circumspect in its claims, and nicely situates their findings within the broader literature.
To our mind, the findings are generally sound. However, we have identified a number of areas where the authors may have inadvertently left resolving power on the table, or their choice of analysis simply isn’t clear. Below, we suggest a number of different ways to look at the data:
1. Currently, the analysis of preference uses a value of 0.5 on the Social Preference Index as a cutoff for social/antisocial bias. We aren’t clear why the authors adopted this value, rather than one that specifically referred to the expected values from the distribution of SPIs during the acclimation period. Alternatives could be those fish with SPI that fall 2 standard deviations from the average during the acclimation period, or those that fall outside a 95% confidence interval.
2. For the data in Figure 1, we infer that the distributions come from paired data (i.e. each fish had both an acclimation and a test period). If so, the appropriate test would be a paired t-test, not simply a two-tailed t-test as described in the Methods.
3. We aren’t clear why the authors use “the same [t-test] statistic [sic] throughout” (Methods) when some of the data is clearly non-normal. The analyses are multiple comparisons of different populations. If the authors wish to use a single statistical analysis, than a non-parametric ANOVA using age as a factor with post-hoc tests would be preferable.
4. In light of the possibility that fish can be both “social” and “anti-social,” it would make more sense to perform statistical analyses on the absolute value of the SPI, rather than the raw SPI. Consider the possibility that during the test phase fish were evenly split 50/50 with perfect SPI values of either 1 or -1. There, the mean would be precisely zero, which would not be different from the SPI during the acclimation period. Obviously, this would be an inappropriate conclusion. Such a split may explain the null finding in Figure 2d. Naturally, a non-parametric test would be appropriate here.
5. When investigating the coordination between fry, the authors use the bout-triggered average. An ideal measure would report the probability of seeing a bout in one fry, as a function of lag from another — while taking into account the frequency of bouts in both fry. Such an analysis is common in signal processing, and is called coherence (Brillinger 1975, Halliday & Rosenberg 1999, or http://en.wikipedia.org/wiki/C... ). Using coherence would allow the authors to report the relationship between the two as a true fraction of variance explained, rather than their complex trace that is arbitrarily normalized to the % of average bout. We note that the author’s “Bout Triggered Average” is roughly comparable to the numerator of the coherence equation.
6. With the bout-triggered average, a metric in units of motion magnitude, directionality is left in question. With such a tight correlation in magnitude of motion between individuals, we are left to wonder if these animals are swimming in the same or opposing directions.
In the finest ethological tradition, the authors explore a number of relevant aspects of the “social” stimulus. They find that both visual information (dark vs. light) and conspecific size/kinematics (1 wk vs. 3 wks) are crucial for the preferences. Similarly, the coupling between conspecifics is intriguing, although incomplete. We recognize that the space of stimuli is vast and many variations on these themes are possible. However, the following experiments would strengthen the manuscript:
1. Multiple tests of the same fish. We note that the authors appear to agree in the Discussion. Particularly in light of the “anti-social” behavior, and the authors stated interest in investigating the neural underpinnings of the phenomena it is crucial to identify whether such observations reflect the statistics of a population of fish or something truly fish-specific. One manipulation might be to simply switch the three conspecifics to the other side, to see if the fish switch sides to follow the group.
2. We would like to know about the time-course of the synchronization of bouts among conspecifics. This could be investigated by slowly dimming the lights during the time three week old fish are viewing a conspecific, and analyzing the time course over which the fish bouts decouple. We hypothesize that a short time constant for the synchronization to decay would reveal a strong dependence on continuous visual stimulation, while a long time constant would suggest that much of the coupling reflects the statistics of spontaneous movement.
3. The question of tactile/vibrational cues remain unexplored. Partridge & Pitcher 1980 demonstrate that adult fish are capable of utilizing the lateral line to shoal. A mesh window, rather than a glass window, could be used to explore the role of such non-visual cues in social behavior in dark conditions.
4. Did the authors examine the social dynamics of 3 week old fish, with older/larger fish? Does size play a role in the reported “anti-social” behavior?
The authors make a number of claims that would benefit from a bit more detail:
1. The claim is that a 45 degree angle relative to the conspecific chamber is best for viewing; it isn’t clear why this would be preferable to a 90 degree angle?
2. Is Figure 3c discussed in the text?
3. We are curious about the temporal dynamics of the time in the chamber. How often do fish switch chambers? Do the three week old fish leave the conspecifics? Do “anti-social” animals switch chambers more often? Is the bout frequency comparable in all parts of the apparatus?
4. There appears to be a grammatical mistake in the last sentence of third paragraph of the introduction.
Finally, we call the authors attention to the recent work of Maurizio Porfiri, whose work on zebrafish social behavior, including in the presence of alcohol, merits consideration in the Discussion. It may also be interesting to address the relationship between this behavior in two dimensions to the three dimensional world these animals have evolved to inhabit in the discussion.