International Journal of Comparative Psychology

Domestic dogs (Canis familiaris) were administered progressive elimination tasks in which they had to visit and deplete either 3 or 4 baited sites. They were brought back to the starting point after each visit.  When administered a 4-choice task with small angular deviation between adjacent targets, the dogs chose first an inner target (e.g., right inner target) and the opposite outer target (e.g., the left target) as a second correct choice.  So they relied on divergence; that is they chose the farthest target as the next choice.  Varying angular deviation did not modulate divergence.  Decreasing the number of targets (3-choice task, Experiment 2) did relax divergence, though target selection was not totally random.  The dogs still chose as a first choice an inner target (i.e., the middle target) when selecting the most divergent patterns of elimination. Finally, in Experiment 3, the dogs were administered a 3-choice task with large angular deviation but in which all targets had been hidden.  The dogs chose first an outer target (i.e., right or left)) and the other outer target as the second correct choice.  That is they relied on divergence.  The results suggest that divergence is the outcome of a flexibility/cognitive load tradeoff when facing novelty and uncertainty.