Rhesus monkeys employ a procedural strategy to reduce working memory load in a self-ordered spatial search task

Abstract

Several nonhuman primate species have been reported to employ a distance-minimizing, traveling salesman-like, strategy during foraging as well as in experimental spatial search tasks involving lesser amounts of locomotion. Spatial sequencing may optimize performance by reducing reference or episodic memory loads, locomotor costs, competition or other demands. A computerized self-ordered spatial search (SOSS) memory task has been adapted from a human neuropsychological testing battery (CANTAB, Cambridge Cognition, Ltd) for use in monkeys. Accurate completion of a trial requires sequential responses to colored boxes in two or more spatial locations without repetition of a previous location. Marmosets have been reported to employ a circling pattern of search, suggesting spontaneous adoption of a strategy to reduce working memory load. In this study the SOSS performance of rhesus monkeys was assessed to determine if the use of a distance-minimizing search path enhances accuracy. A novel strategy score, independent of the trial difficulty and arrangement of boxes, has been devised. Analysis of the performance of 21 monkeys trained on SOSS over 2 years shows that a distance-minimizing search strategy is associated with improved accuracy. This effect is observed within individuals as they improve over many cumulative sessions of training on the task and across individuals at any given level of training. Erroneous trials were associated with a failure to deploy the strategy. It is concluded that the effect of utilizing the strategy on this locomotion-free, laboratory task is to enhance accuracy by reducing demands on spatial working memory resources.

Figure 1

The 16 possible locations in which boxes might be presented for the SOSS task are illustrated in the left panel. A response path dictated by a distance-minimizing search strategy is illustrated for a sample 4-box trial in the center panel. A suboptimal strategy which does not minimize the distance traveled is represented for the same trial in the right panel. Numbers indicate the sequence of responses and the open box highlights the starting position.

Figure 2

The proportional proximity strategy scores associated with each of 24 possible response sequences are presented for a sample of eight 4-box trials (upper panels). The corresponding raw path lengths in screen pixel units for the same eight trials are represented in the center and lower panels. Pairs of identical length search paths in upper and lower panels indicate the same path traveled in opposite directions.

Figure 3

The mean (N = 21; ± SEM) accuracy, strategy for correct trials and error strategy scores are depicted across 24 months of training for 3-box and 4-box trials. Accuracy depends on trial difficulty whereas strategy scores do not. Accuracy and strategy for correct trials tend to improve with training duration whereas error strategy does not. Open symbols depict a significant difference from Months 1–2 and between difficulty conditions within a given month. Shaded symbols depict a significant difference compared with Months 1–2 only.

Figure 4

Individual (N = 21) accuracy scores are plotted against strategy scores for each of the initial 24 months of training. The achievement of higher average accuracy was associated with a lower average proximity score. Individuals differed substantially in terms of both accuracy and strategy utilization.