Supplementary Material
SM1:
Videos of grackles performing on the lid removal (SM1a), stick pulling (SM1b) and discrimination learning (SM1c) tasks.
SM2:
Analyses considering performance in the lid removal and stick pulling tasks as a binary variable (comparison of individuals successful within the 10 first trials vs. individuals needing more trials to succeed).
Correlations between time variables and success within each problem solving task
For both the lid removal and stick pulling tasks, individuals that were successful within the first 10 trials spent more time interacting with the task (for the lid removal task: Wilcoxon test: p < 0.001; average time spent interacting with the task per trial in successful individuals = 12.577 ± 2.745 s; in unsuccessful individuals = 0.594 ± 0.202 s; for the stick pulling task: Wilcoxon test: p < 0.001; average time spent interacting with the task per trial in successful individuals = 51.321 ± 6.755 s; in unsuccessful individuals = 7.964 ± 1.800 s).
These trends in individual differences could be associated with shyness or body condition. We first tested whether shyness and body condition had an effect on success in the tasks and whether the relationships between success within the 10 first trials and time spent interacting with the lid removal and stick pulling tasks were due to these confounding effects. To that aim, we used GLMs fitting success at a given problem solving task with a binomial distribution. We found an effect of shyness and body condition for the lid removal task only: individuals succeeding in solving the lid removal task within 10 trials were less shy (estimate = -0.002 ± 0.001; z = -3.031; p = 0.002) and had a lower body condition (estimate = -0.340 ± 0.115; z = -2.945; p = 0.003). In spite of this, we found that the significant difference between successful and unsuccessful individuals in time spent interacting with the lid removal task remained significant even after taking into account shyness and body condition effects (effect of time spent interacting with the lid removal task on success at lid removal, in a generalised linear model including body condition and shyness as fixed effect, and considering a binomial distribution: estimate = 0.047 ± 0.021; p = 0.026). None of the other variables affected lid removal success. None of the explanatory variables explained stick pulling success.
Correlations between success at the two problem solving tasks and number of trials before success at colour discrimination learning
Individuals that succeeded within 10 trials in the lid removal task were also more likely to succeed within 10 trials in the stick pulling task (χ² = 11.177; Df = 1; p < 0.001). This relationship was still significant after considering the effects of shyness and body condition (estimate = 4.517 ± 1.523; z = 2.965; p = 0.003). In contrast, the number of trials before success at discrimination colour learning did not differ between successful and unsuccessful individuals at the lid removal task (Wilcoxon test: p = 0.120) nor at the stick pulling task (Wilcoxon test: p = 0.281).
SM3:
Canonical correlation analysis of the relationship between behaviour in the discrimination learning and problem solving tasks.
dimension / canonical correlation / Wilks λ / df / p1 / 0.42 / 0.697 / 8,96 / 0.022
2 / 0.39 / 0.847 / 3,49 / 0.042
CA1 / CA2
Canonical variables (problem solving)
Trials to success obstacle removal / 0.257 / 0.956
Trials to success stick pulling / 0.909 / 0.332
Time interacting with the task obstacle removal / -0.177 / -0.510
Time interacting with the task stick pulling / -0.805 / 0.030
Canonical variables (discrimination learning)
Trials to success at discrimination learning / 0.351 / -0.936
Average time before colour choice / 0.755 / 0.655
1