Selection for evasive mimicry imposed by an arthropod predator
Data files
Dec 05, 2023 version files 25.54 KB
-
mantis_evasive_test_data.xlsx
-
mantis_evasive_training_data.xlsx
-
README.md
Abstract
It has long been hypothesized that a species that is relatively easy to catch by predators may face selection to resemble a species that is harder to catch. Several experiments using avian predators have since supported this “evasive mimicry” hypothesis. However, the sudden movement of artificial evasive prey in each of the above experiments may have startled the predators, generating an avoidance response unrelated to difficulty of capture. Additionally, in the above experiments, the catchability of prey was all or nothing, while in nature predators may occasionally catch evasive prey or fail to catch slower species, which might inhibit learning. Here, using mantids as predators, we conducted an experimental test of the evasive mimicry hypothesis that circumvents these limitations, using live painted calyptrate flies with modified evasive capabilities as prey. We found that mantids readily learned to avoid pursuing the more evasive prey types. Warning signals based on evasiveness and their associated mimicry may be widespread phenomena in nature. These findings not only further support its plausibility but demonstrate that even arthropod predators can select for it.
README: Selection for evasive mimicry imposed by an arthropod predator
https://doi.org/10.5061/dryad.s1rn8pkff
This paper describes an experiment conducted to test whether mantids can learn to use a signal (black or white) to discriminate between flies that are hard to catch (wings not clipped) from flies that are easy to catch (wings clipped).
Description of the data and file structure
We have two data sets and associated code. The first data set represents the training data over seven presentations of pairs of flies (one black and one white) to mantids (80 in total, split evenly over four training treatments). The second data set represents the test data in which mantids were presented with a black and a white fly with unclipped wings.
Training data set
The training data set provided in MS Excel format comprises 80 rows (20 mantids for each of 4 treatments) and 9 columns, comprising the following variables:
Individual : a unique ID for the individual mantid (1-80)
Treatment : 1-4 (1 = BN/WC, 2 = BC/WN, 3 = BC/WC, 4 = BN/WN)
D1-D7 : type of fly (if any) attacked on given training day D1-D7 (0 = No fly attacked; 1 = White fly attacked; 2 = Black fly attacked)
Test data set
The test data set provided in MS Excel format comprises 80 rows (20 mantids for each of 4 treatments) and 3 columns, comprising the following variables:
Individual : a unique ID for the individual mantid (1-80)
Training teatment : 1-4 (1 = BN/WC, 2 = BC/WN, 3 = BC/WC, 4 = BN/WN)
WBN120 : whether white (= 1), black (= 2) or no fly (= 0) was attacked
Sharing/Access information
The experiments were run by KLH and the data were analyzed by TNS. Further details of the experiment are available as Supplementary Methods associated with the main paper.
Code/Software
All visualization and analysis of the above data was conducted in R.
R Training code:
Data preparation and visualization
We first read the above data file, then convert it from wide to long format and define some variables based on transformations of the existing variables. We then sum the number of outcomes (no fly attacked, white fly attacked, black fly attacked) over all mantids within each treatment over each training day and plot these distributions out in a stacked bar chart. For ease of interpretation, we have used white bars to represent the number of white flies attacked, black bars to represent the total number of black flies attacked and grey bars to represent the total number of times no fly was attacked.
Statistics
Our analysis of the training data was limited to testing whether more white or black flies were consumed by mantids over the seven days of presentations within any given treatment, using a simple Wilcoxon Signed rank test (with the mantid as the pairing variable). To do so, we first subsetted the data by treatment, then calculated the frequency of black and white (and no) flies caught by each mantid over the course of their training. Sign tests use less information but here we show that they result in identical qualitative conclusions.
R Test code:
Data preparation and visualization
The data are already in long format, so we simply sum the number of times a white fly, black fly or no fly was attacked for each of the 4 treatments and plot the results as a stacked barchart.
Statistics
For ease of interpretation, we have focused on instances in which one of the two flies (black or white) were attacked (90% of cases). We then fitted two binary logistic models (intercept only and a model in which treatment affected the log odds of being attacked) and compared them using a log likelihood ratio test (LRT). Following evidence for an overall effect of training treatment on the log odds of attacking a black (as opposed to white) fly, we compared among treatments with a post-hoc Tukey HSD implemented using the glht function of the multcomp package. Finally, to test the null hypothesis that attack on the two types of fly were equally likely we compare the total number of white and black flies attacked by each mantid using a separate sign test for each of the four treatments.
Methods
Training phase
The training phase of our experiment involved repeatedly presenting two similar-sized flies, one with thorax painted black and one with thorax painted white, to individual mantids (Tenodera sinensis). The presentations (seven in total) were made every second day over a period of two weeks. The two presented flies were clipped and painted as part of a 2 (Black B or White W) x 2 (Clipped C or Not Clipped N) treatment design i.e. only the white fly was clipped (BN/WC; treatment 1), only the black fly was clipped (BC/WN; treatment 2), both flies were clipped (BC/WC; treatment 3), neither fly was clipped (BN/WN; treatment 4). Each mantid only experienced a single training treatment and there were 20 mantids per treatment.
Presentations of flies to mantids took place in sequential order based on treatment, so that flies were first given to a mantid in treatment 1, then a mantid in treatment 2, then 3 and 4. After setup, the flies and mantids were checked every 20 minutes over a three-hour duration. Once a mantid had caught one fly, the remaining fly was removed. If no flies were caught after three hours both were removed. Since it takes mantids of the size (L3-L4) used in this experiment more than 20 minutes to consume a fly, the mantids were never able to consume both flies between consecutive checks. Note that failed capture attempts (or the precise timing of successful captures) were not recorded in the training phase, only the fly type that was successfully caught.
Test phase
The test phase was conducted once for each mantid and involved presenting unclipped (BN/WN) yet tethered flies and observing which colour of fly, if any, was first attacked. The mantids were tested following the same sequential order of treatment as above and monitored in the same manner as in the training phase.