Early experience affects foraging behavior of wild fruit bats more than their original behavioral predispositions

To assess bats’ personality, we ran young bats through the multiple-foraging box paradigm, which was developed in our lab and has been shown to measure behavioral variability along multiple axes (Harten et al., 2021). A total of 40 bats participated in the foraging box experiments, which were repeated a maximum of five times per individual over a period of almost 5 months (Figure 1A, see Methods). The foraging box experimental setup was designed to assess the innate personality traits in bats, which we could then use as parameters when trying to explain their foraging outdoors. Repeated measurements were used to confirm that the traits met the criteria for personality, such as consistency over time. The experiment comprised baseline trials for each bat on two consecutive nights, upon the bats reaching an average age of 4.1±1.4 months (trials 1–2; Mean ± STD). Subsequently, we randomly divided the bats into two colonies, each exposed to different environmental conditions. Both colonies were provided with the same basic diet sufficient for all bats in the colony. Importantly, bats in both colonies were raised in large social groups of similar size. The enriched colony underwent frequent changes in their surroundings for a period of ~10 weeks, through the introduction of various enrichments that forced the bats to practice trial and error to obtain (Methods). While the control impoverished colony was exposed to a stable environment with minimal changes. A post-enrichment boxes trial was performed immediately after this period (Trial 3, on average 72 days after the previous). Finally, two post-release box trials were performed after the bats were released into our open colony and could freely forage in the wild and explore their natural environment (Trials 4–5, average 45 and 58 days after their release into the wild). Post-enrichment and post-release trials were conducted to assess whether early life experiences had an effect on the development of bat personality and in order to identify persistent personality traits that can be used as a proxy of the individual innate predispositions. We used these trials to assess proxies of the three above-noted behavioral traits: boldness, exploration, and activity. To assess the behavioral traits during the foraging box experiments, we scored the bats’ actions. Specifically, we recorded how many times they approached and took food from a foraging box, the number of different boxes they explored (i.e. landed on), and the total number of actions they performed during the night (Figure 1A and see Methods).

Assessing personality in the laboratory.

(A) Schematic illustration of the experiments. (B) A 3D plot of the three behavioral traits, estimated during Trial 1. (C1-3) Behavioral traits were consistent over time. The behavioral traits of the first and third (post-enrichment) trial are presented. Positive correlations were found for all three behavioral traits over a period of more than 10 weeks. For the enriched (blue lines), the impoverished (orange lines), and both environments together (not shown), respectively: (C1) Boldness; (C2) exploration; and (C3) activity. (D) Boldness of the first and fifth (post-release 2) trial. Dashed line represents the Y=X line.

When plotting the three behavioral traits (boldness, exploration, and activity) in 3D, they seem to form a triangle (Figure 1B) – reminiscent of a Pareto front (Shoval et al., 2012) – suggesting a trade-off between the different traits. The three vertices of the triangle (often referred to as the archetypes when analyzing a Pareto front) represent individuals that exhibit: (a) high exploration, high activity, and medium-low boldness; (b) high boldness, low activity, and low exploration; and (c) low boldness, low activity, and low exploration (very few bats). These three archetypes suggest a trade-off between boldness on the one hand, and activity and exploration on the other hand.

For all three behavioral traits, we found a significant positive correlation between baseline (Trial 1) and the post-enrichment trial (Trial 3) examined on average 72 days apart. A significant positive correlation was also found between bats’ boldness during the baseline (Trial 1) and the second post-release trial (Trial 5), examined on average 144 days apart. These findings suggest that these behavioral traits, and especially boldness, remain consistent at least throughout the bat’s early life (Figure 1C1-3, Pearson correlation test: r=0.66 p<9.7e-0.5, r=0.66 p<9.1e-0.5, and r=0.51 p<0.004; for Trials 1–3 for boldness, exploration, and activity, respectively, and r=0.6 p<0.021, for Trials 1–5 boldness, Table 1). These results did not change when examining the enriched and impoverished colonies separately. To compute an integrative measurement of personality, we ran a PCA analysis on the data from trials 1 and 5. We found a significant Pearson correlation (p=0.03, r=0.58, Figure 1—figure supplement 1) between the individuals’ scores on PC1 in trials 1 and 5 suggesting that this measure captures a persistent representation of behavior that takes both boldness and exploration into account (the activity weight was very low).

Pearson’s correlation test.

Boldness 1–3Exploration 1–3Activity 1–3Boldness 1–5Pearson r value0.660.660.510.6Pearson p-value<9.7e-059.1e-0.50.0040.021

Notably, all three personality traits exhibited significant correlations in the two baseline trials suggesting that indeed they capture behavioral tendencies (Figure 2).

Personality measurements over time.

(A) The behavioral traits of the first and second baseline trials, performed on two consecutive nights, are presented. Positive correlations were found for all three behavioral. Boldness (Pearson r=0.59, p=5.8e-05) (A1); exploration (Pearson r=0.70, p=5.3e-07) (A2); and activity (Pearson r=0.52, p=0.0005) (A3). Dashed line represents the Y=X line. (B) Box plots depicting the bats’ behavior across the five trials according to three traits: Boldness (B1), exploration (B2), and activity levels (B3). (C) Paired box plots show individual bat values in the first trial (baseline 1) and the fifth trial (post-release 144 days apart on average) for each trait: Boldness (C1), exploration (C2), and activity levels (C3). Lines connect repeated measurements from the same individual, illustrating within-individual behavioral changes over time. The number of bats participated in each trial is given in Methods, Table 8. In the baseline trials, no significant differences were found between the two environmental conditions (enriched / impoverished) for any of the three behavioral traits we measured (p=0.62, p=0.84, p=0.72 for boldness, exploration, and activity levels, respectively, n=40, Supplementary file 1). Figure 2—figure supplement 1 also shows the indication. The experimental phase and the time in days from the day of the first trial are depicted on the X-axis. Box plots show the 25% and 75% percentiles. Medians and whiskers based on 1.5 IQR are shown.

The temporal dynamics of boldness and exploration revealed similar patterns: both increased between the first two consecutive baseline trials (Figure 2). Following the enrichment condition (i.e. in Trial 3), they decreased back to the level of the first baseline and then increased again in the two post-release trials (4-5). Boldness was significantly higher in the last trial compared to the first one (p<1.9e-07, see discussion). A mixed-effect Generalized Linear Model (GLMM) was used, with the behavioral traits set as the response variable, trial number (1 or 5) as a fixed factor, and bat ID as a random effect (see Table 2). While exploration showed a fairly similar pattern. Activity, in contrast, showed a different pattern. It decreased continuously throughout the trials and was significantly lower in the last trial compared to in the first one (p<0.0001, GLMM as above). An individual-level comparison between the first and last trial is presented in Figure 2C.

Mixed generalized linear model (GLM) results for the comparison between trials 1 and 5 response ~trial + (1 | bat ID).

ResponseAICBICLogLikelihoodDevianceBoldness–36.379–28.49822.189–44.379Fixed effects coefficients (95% CIs)NameEstimateSEtStatDFp-valueLowerUpperIntercept0.4080.03113.16515.03e-180.3460.471Trial number0.0560.0096.02511.902e-070.0370.075ResponseAICBICLogLikelihoodDevianceExploration12.07419.956–2.0374.074Fixed effects coefficients (95% CIs)NameEstimateSEtStatDFp-valueLowerUpperIntercept0.5450.04911.02514.248e-150.4460.644Trial number0.0180.0161.157510.252–0.013–0.013ResponseAICBICLogLikelihoodDevianceActivity–580.34–572.46294.17–588.34Fixed effects coefficients (95% CIs)NameEstimateSEtStatDFp-valueLowerUpperIntercept0.0020.000113.737519.050e-190.0020.003Trial number–0.00027.135e-05–4.065510.0001–0.0004–0.0001

Despite the above-reported changes in behavior over time, the environment that the animals experienced as juveniles (enriched or impoverished) did not have a significant effect on their personality as measured in the lab (GLMM – with the behavioral traits set as the response parameter, the trial number and the interaction between trial number and the environmental condition as fixed factors, and bat ID as a random effect). The interaction between trial and environmental condition was not significant (Table 3). When examining only Trials 4–5, i.e., following their release into nature, the enriched bats were observed to be significantly bolder than the non-enriched bats (p=0.003, GLMM as above). However, this seems like a result of post-selection, resulting from which bats remained in our colony and could be tested (see Discussion).

Mixed generalized linear model (GLM) results for all tested trials (1-5) response ~1 + Trial × EnvironmentalCondition + (1 | bat ID).

ResponseAICBICLogLikelihoodDevianceBoldness–100.5–82.80756.25–112.5Fixed effects coefficients (95% CIs)NameEstimateSEtStatDFp-valueLowerUpperIntercept0.4480.044410.0781373.266e-10.360.536Environmental condition Impoverished0.0070.0610.1171370.906–0.1150.129Trial number0.0380.0142.7441370.0060.0100.066Environmental condition Impoverished: Trial–0.0110.019–0.581370.561–0.0490.026ResponseAICBICLogLikelihoodDeviance Exploration–0.4612817.2316.230–12.461Fixed effects coefficients (95% CIs)NameEstimateSEtStatDFp-valueLowerUpperIntercept0.6250.0669.4671371.13e-160.4940.755Environmental condition Impoverished–0.0100.092–0.1121370.910–0.1920.171Trial number–0.0030.019–0.1821370.855–0.0410.034Environmental condition Impoverished:Trial0.0030.0060.0250.2471370.804–0.0430.054ResponseAICBICLogLikelihoodDevianceActivity–1586–1568.3799.01–1598Fixed effects coefficients (95% CIs)NameEstimateSEtStatDFp-valueLowerUpperIntercept0.0020.000211.0171371.319e-20.0020.003Environmental condition Impoverished0.00020.00030.9081370.365–30e-50.0009Trial number–0.00027.292e-05–3.273970.001–30e-3–9.e-05Environmental condition Impoverished:Trial–9.2e-059.640e-05–0.9591370.338–20e-59.8e-05