Artificial selection for reversal learning reveals limited repeatability and no heritability of cognitive flexibility in great tits (Parus major)

Abstract

Cognitive flexibility controls how animals respond to changing environmental conditions. Individuals within species vary considerably in cognitive flexibility but the micro-evolutionary potential in animal populations remains enigmatic. One prerequisite for cognitive flexibility to be able to evolve is consistent and heritable among-individual variation. Here we determine the repeatability and heritability of cognitive flexibility among great tits (Parus major) by performing an artificial selection experiment on reversal learning performance using a spatial learning paradigm over three generations. We found low, yet significant, repeatability (R = 0.15) of reversal learning performance. Our artificial selection experiment showed no evidence for narrow-sense heritability of associative or reversal learning, while we confirmed the heritability of exploratory behaviour. We observed a phenotypic, but no genetic, correlation between associative and reversal learning, showing the importance of prior information on reversal learning. We found no correlation between cognitive and personality traits. Our findings emphasize that cognitive flexibility is a multi-faceted trait that is affected by memory and prior experience, making it challenging to retrieve reliable values of temporal consistency and assess the contribution of additive genetic variation. Future studies need to identify what cognitive components underlie variation in reversal learning and study their between-individual and additive genetic components.

Keywords: animal personality; artificial selection; cognition; quantitative genetics; reversal learning.

Figure 1.

Response to selection per generation. (a) Boxplot with horizontal lines representing median and interquartile range of TTC (ln-transformed); circle and triangles indicate mean TTC, corrected for feeder type. Eye plots represent density distributions for TTC. Dots indicate numbers of individuals within each bin. (b) Realized heritability (h²) of reversal learning performance in an artificial selection experiment. Cumulative response to selection (R) is plotted as function of the cumulative selection differential (S) for each line selected on number of trials to reach learning criterion for two generations. The realized heritability (h²) was calculated as the slope of the linear regression between R and S (black solid line). Error bars represent s.e. Dashed line indicates expected function if slope (h²) were 0.15. Orange, downward pointing arrows: fast line; blue, upward pointing arrows: slow line.

Figure 2.

Relationship between the number of TTC (ln-transformed) in the reversal learning phase and the number of TTC (ln-transformed) in the associative learning phase. Higher values = slower learning. This is shown for (a) feeder type 1 (light grey, continuous line and dots) and type 2 (dark grey, dashed line and triangles) and for (b) the interaction between associative learning and the ‘correct visit lag number’ (grouping levels based on standard deviation; +1 s.d. is dark grey solid line and −1 s.d. is light grey dashed line). Darker dots reflect a higher ‘correct visit lag number’. For both (a) and (b), plotted lines are marginal effects of the interactions term. Lines and shaded regions represent the model prediction (±95% confidence interval). The points of the scatterplot represent the actual data points.