Adaptive numerical competency in a food-hoarding songbird

Abstract

Most animals can distinguish between small quantities (less than four) innately. Many animals can also distinguish between larger quantities after extensive training. However, the adaptive significance of numerical discriminations in wild animals is almost completely unknown. We conducted a series of experiments to test whether a food-hoarding songbird, the New Zealand robin Petroica australis, uses numerical judgements when retrieving and pilfering cached food. Different numbers of mealworms were presented sequentially to wild birds in a pair of artificial cache sites, which were then obscured from view. Robins frequently chose the site containing more prey, and the accuracy of their number discriminations declined linearly with the total number of prey concealed, rising above-chance expectations in trials containing up to 12 prey items. A series of complementary experiments showed that these results could not be explained by time, volume, orientation, order or sensory confounds. Lastly, a violation of expectancy experiment, in which birds were allowed to retrieve a fraction of the prey they were originally offered, showed that birds searched for longer when they expected to retrieve more prey. Overall results indicate that New Zealand robins use a sophisticated numerical sense to retrieve and pilfer stored food, thus providing a critical link in understanding the evolution of numerical competency.

Figure 1

A New Zealand robin choosing an artificial cache site during a field trial.

Figure 2

Results from two field experiments where birds were allowed to choose between two cache sites containing different numbers of prey. The percentage of birds that chose the cache site containing the greater number of prey is shown on the y-axis. The total number of prey used in each treatment is shown on the x-axis. Treatments are labelled by the number of prey hidden in each of the two compartments. Treatments located above the dotted line (11 or more ‘correct’ decisions) had a greater number of correct decisions than expected by chance based on the binomial distribution (i.e. p<0.05). (a) Results from experiment 1, where only mealworms were presented to birds (uncontrolled). (b) Results from experiment 2, where small rocks were placed in wells containing fewer mealworms to control for time and volume confounds.

Figure 3

Results from a ‘violation of expectancy’ experiment, in which birds were shown a particular number of prey items that were then hidden in an artificial cache site containing a trapdoor. In some trials, birds were allowed to retrieve all the prey items they were shown. In other trials, several prey items were hidden behind the trapdoor, thus allowing birds to retrieve only a fraction of what they were shown. The amount of time spent searching for hidden prey is shown on the x-axis (±s.e.; ^***p<0.001; ^**p<0.01; ^*p<0.05; n.s., not significant). Birds spent more time searching when they were shown more prey than they were allowed to retrieve. However, this effect declined with the total number of prey items they were shown. The bottom pair of treatments controlled for the possibility that birds detected prey hidden behind the trapdoor; 1(6h) refers to a treatment where one prey item was shown to the subject, while six prey identical items were hidden behind the trapdoor in the absence of robins. The pair of treatments listed second from the bottom controlled for the possibility that birds made decisions based on differences in volume; 2(s) refers to a treatment where two small prey items, each measuring half the volume of the prey items used in all other treatments, were used.