(Un)predictability has only recently been recognized as an important dimension of animal behavior. Currently, we neither know if (un)predictability encompasses one or multiple traits nor how (un)predictability is dependent on individual conditions. Knowledge about condition dependence, in particular, could inform us about whether predictability or unpredictability is costly in a specific context. Here, we study the condition dependence of (un)predictability in the escape behavior of the steppe grasshopper Chorthippus dorsatus. Predator–prey interactions represent a behavioral context in which we expect unpredictability to be particularly beneficial. By exposing grasshoppers to an immune challenge, we explore if individuals in poor condition become more or less predictable. We quantified three aspects of escape behavior (flight initiation distance, jump distance, and jump angle) in a standardized setup and analyzed the data using a multivariate double-hierarchical generalized linear model. The immune challenge did not affect (un)predictability in flight initiation distance and jump angle, but decreased unpredictability in jump distances, suggesting that unpredictability can be costly. Variance decomposition shows that 3–7% of the total phenotypic variance was explained by individual differences in (un)predictability. Covariation between traits was found both among averages and among unpredictabilities for one of the three trait pairs. The latter might suggest an (un)predictability syndrome, but the lack of (un)predictability correlation in the third trait suggests modularity. Our results indicated condition dependence of (un)predictability in grasshopper escape behavior in one of the traits, and illustrate the value of mean and residual variance decomposition for analyzing animal behavior.

In order to test the condition-dependence of (un)predictability, we exposed steppe grasshoppers Chorthippus dorsatus to an immune challenge (n = 66) and recorded the effect of the treatment on three escape behavior traits, in comparison to untreated individuals (n = 66). Individuals were sampled as nymphs in June/July 2021 from Jena, Germany. Subjects were transferred to the laboratory, separated by sex, and maintained in groups of up to about 60 individuals in large mesh cages of dimensions 47.5 x 47.5 x 93 cm³. After their final molt, individuals were transferred in pairs of the same sex and the same final molt date to smaller mesh cages of dimensions 22 x 16 x 16 cm³. Individuals were maintained with ad libitum access to freshly cut grass provided in small vials with water and a water tube for moisture. Dead individuals and individuals that lost legs were replaced, if possible, with other individuals that matched the cage mate. When a replacement was impossible, the remaining cage mate was not used in the experiment. Around 15 days after the final molt, a behavioral assay was performed as described below. To achieve standardized conditions across the behavioral essay, individuals were placed in refrigerators at 8ᵒC on the evening prior to phenotyping. During overnight cooling, individuals were kept individually in cylindrical vials of dimensions 8 cm height and 5 cm diameter, with 2-3 leaves of grass. For each pair of same sex and final molt date, one individual was randomly selected for an immune challenge, while the other was used as control.

To cause an impairment in the condition of individuals, we performed an immune challenge. Treated individuals were anesthetized by exposing them for 3 minutes to temperatures of -20ᵒC. They were then injected with a solution containing 1.6x108 cells of heat-killed Escherichia coli per microliter diluted in Ringer’s grasshopper solution. The cell concentration was determined after the efficacy of 5 different solutions (with 1.6x108, 1.6x107, 1.6x106, 1.6x105, and 1.6x104 cells/μL) was tested in 20 individuals each (10 males and 10 females). The chosen concentration caused the death of 25% of the individuals after 48h and thus seemed sufficiently strong without being immediately fatal. Males received 1 µL of the solution, and females, 2 µL (females weigh around twice as much as males). In order to maximize the differences between control and treated groups (and because we were not interested in the effect of heat-killed E. coli as such), control individuals did not receive any injection nor were they anesthetized. After injections, all individuals (treatment and control) were color-marked dorsally with yellow paint (for automated analysis, see below). In order to keep track of individual identity, one of the individuals was marked with a stripe from the wings to the pronotum, and the other with a stripe on the wings and a dot on the pronotum. The type of marking was randomly assigned and thus not associated with the treatment groups. After being marked, individuals were left for 1 hour under a heat lamp in mesh cages of dimensions 22 × 16 × 16 cm³, with fresh grass available, as preparation for the behavioral assay.

Each pair of treated and control individuals was phenotyped simultaneously, with the observer being blind to the treatment. The two individuals were placed into a 5.3 x 3.3 m² arena, where they could acclimatize for 5 minutes. The arena floor was coated with foam rubber to provide a non-slippery surface for the grasshoppers to jump. The room temperature was maintained constant at 27ᵒC. Individuals were approached with a moveable device, which was placed on the ground about 50 cm behind the grasshopper and moved smoothly and at a constant speed toward it (approximately 130 cm/s2), mimicking an approaching predator. Each individual had ten escape jumps recorded on video, with 2 minutes time gap between chases. To streamline phenotyping, we alternately simulated attacks on the two individuals in the arena. On average, 14 individuals were phenotyped per phenotyping date. The software Ethovision XT (Noldus, Netherlands), with the Social Interaction Module, was used to track the movements of the chaser and the grasshopper. Flight initiation distance (FID), jump distance, and jump angle measurements were extracted from the tracks.

Recommended to save the .txt file in .csv format to align column data

The dataset consists of a total of 1163 observations.
TrialID refers to each observation.
IndID refers to the identification number of each individual.
FID is the flight initiation distance in centimeters, defined as the minimum distance between the chaser and the grasshopper before the grasshopper jumped.
JumpDistance is the jump's length in centimeters, defined as the distance between the point where the jump started and where the grasshopper landed
JumpAngle is the angle of the jump direction in degrees in relation to the initial grasshopper body position. Straight jumps scored 0ᵒ, while negative values marked jumps to the left, and positive values jumps to the right.
JumpOrder refers to the sequential number of jumps within the 10 trials of an individual.
JumpYN indicates if the grasshopper jumped or not during that trial (Y = jumped, N = didn't jump). 
Sex is coded as M for males and F for females.
Treatment refers whether an individual was from the control group (C) or if it received the immune challenge injection described previously (T)
PhenotypingDate is the date when the experiment occurred (DD_MM_YYYY)
DayTime is coded as 24h, with the minutes converted to decimals (for example: 1:30 pm = 13.5)
PairID is the identification number of a pair of grasshoppers of same age and sex, kept in the same cage and phenotyped together, one assigned to the treatment group and the other to the control group.