Learning of paw preference in mice is strain dependent, gradual and based on short-term memory of previous reaches

We studied the dynamics of paw preference learning in unbiased symmetrical test chambers where mice, Mus musculus, could freely choose which paw to use. When compared to nonlearner model mice, three strains exhibited different degrees of learning within and between two training sessions 1 week apart. While paw preference was probabilistic, positive autocorrelation between paw choices made by individual mice in a training session showed that bias in paw preference changed gradually with successive paw reaches and was concordant between sessions. Within a session, the degree of positive autocorrelation between consecutive paw choices differed between strains and decreased faster with increasing lag between reaches in stronger than in weak learners, suggesting that the degree of learning is the genetic variable in mice. We propose that strong biases in individual mice result from weak biases that appear by chance early in training and increase by a positive feedback mechanism because of learning during training. This explains how individual mice of each strain were strongly biased when there was almost no bias at the population level. The decrease in autocorrelation with lag shows that constitutive behaviours play a minor role and, hence, paw preference can adapt to environmental changes. We conclude that paw preference learning in mice is genetically dependent on past experience, changes gradually within training sessions, and is based on short-term memory; it also has long-term effects, provided there is time for memory consolidation.