There are effects of magnetism on behavior in fruit flies
Reply to “Comment on Magnetic Field Effects in the Drosophila Butterfly” by Gegear et al
Our conclusion3 that the magnetic field effects reported by Gegear et al.4 were most likely false positives was based on the incorrect choice of statistical tests by these authors. We have discussed these matters extensively in the Supplementary Information of our paper3 and in ref. 8. Here we provide only a short summary.
Fly conditioning is an example of a behavioral paradigm that is sensitive to the environment in which it is carried out. It is arguably the most complex of these types of Drosophila phenotype and requires considerable skill and experience to obtain reliable results. Although the experiments of Bassetto et al.1 might have been optimally shielded against interfering outside magnetic effects, it is evident in their Methods section that the critical ‘positive conditional control’ utilizing olfactory conditioning was not carried out under the same conditions as the failed magnetic conditioning studies. Instead, these ‘controls’ were carried out under temperature- and humidity-controlled conditions in Oxford, UK. It is impossible to determine whether a shielded location in Oldenburg had the appropriate environment for robust sugar-reinforced conditioning without controlling it under the same location and conditions. The lack of an appropriate ‘positive conditional control’ in Oldenburg is a substantial criticism and suggests that there may be other important variables that differ between the studies in Bassetto et al.1 and those in Gegear et al. 2008 (ref. 2).
Our conclusion that the approach of Bassetto et al.1 is overly conservative is based on a much more straightforward, non-parametric approach (the Wilcoxon rank sum test, also known as the Mann–Whitney U-test). The data for the naive and trained groups of flies in the synthetic dataset is vastly different by this analysis.
Reppert2 regrets that we did not consider the monarch butterfly as a model organism for studying the mechanism of light-dependent magnetoreception. Given the reports that monarchs should be able to orient in the Earth’s magnetic field (reviewed in ref. 15), these genetically tractable insects could be a potential alternative to Drosophila. In two studies, monarchs did not show such an ability, with one of the studies showing a clearly directed southwest orientation and the other showing random orientation. Furthermore, monarchs first flown in the normal magnetic field did not react to a horizontal 120° turn of the field even when they were kept flying in the rotated field for up to 2 h (ref. 17).
Kyriacou1 remarks on the accuracy of the video tracking of fly movements in our negative geotaxis experiments3. There were flies that were out of bounds that led to the number of frames being recorded. The analysis did not include frames with flies hidden by the supports at the base of the tubes. There were flies that came to the top of the tubes but did not log them. The flies were being tracked while they were climbing. Moreover, it is not necessary to track a fly in every frame to determine its climbing rate. By contrast, Fedele et al.7 simply reported the proportion of flies that climbed to an arbitrarily chosen height within an arbitrarily chosen time period with no further data or photographic documentation.
Bassetto et al.1 criticize the use of parametric statistical testing in the Gegear et al. 2008 paper2. There are times when analysis of the data is done using statistics. Indeed, Krashes and Waddell7,8 advise using parametric statistical testing of performance index values derived from appetitive and aversive olfactory conditioning assays and recommend a sample size of 8–10 replicates per condition per genotype. Instead, Bassetto et al.1 have selected an extremely conservative approach to reanalysis of the data in Gegear et al. 2008 (ref. 2). The statistical power of the original analysis can be found in this choice.
The statistical analysis in Gegear et al. 2008 was re-visited by Bassetto et al. The reanalysis doesn’t support the idea that most of the results were false positives and not statistically significant.
When using an ordinal logistic fit model to assess the synthetic dataset, which is equivalent to the type of generalized linear model used by Bassetto et al.1 (based on the group averages in Gegear et al. 2008, Fig. 1b2; discussed in the text and Supplementary Fig. The statistical results are dependent on the number of flies in the model. With ‘batch’ included as an independent variable, the effect of training is minimal (P = 0.33), whereas omission of ‘batch’ altogether leads to a highly significant effect of training (P < 0.0001). The former option is likely to be chosen by the authors.
Mimicking the approach of Bassetto et al.1 to generate a single synthetic dataset, we generated an additional 20 synthetic datasets. The conservative approach of Bassetto et al.1 led to 5 of 20 datasets showing a significant effect of training, whereas 15 did not. When using three other approaches (t-tests, ordinal logistic binominal models without ‘batch’ or non-parametric rank tests), all 20 synthetic replicates demonstrated highly significant differences between the groups (P < 0.0001). Thus, Bassetto et al.1 seem to have selected a statistical approach with extremely poor sensitivity for detecting differences when reanalysing the data in Gegear et al. 2008 (ref. 2). There is no way that the results represent a false positive. Moreover, if false positives occurred in previous studies, they would be expected to occur in a variety of treatments and not in a way that consistently provides evidence for magnetosensitivity.
