Subjects performed a virtual patch-foraging task: a discrete-trial adaptation of a class of tasks from the ecology literature (Cuthill et al., 1990 ; Hayden et al., 2011 (link); Charnov, 1976 (link); Stephens and Krebs, 1986 ). On each trial, subjects were presented with a tree and had to decide whether to harvest it for apples or go to a new tree. Subjects indicated their choice by one of two key presses when prompted by a response cue. If they decided to harvest the tree they incurred a short harvest time delay, during which the tree shook and the harvested apples were displayed (as an integer number of apple icons plus a fractional apple icon for the remainder), followed by a response cue. As the subject continued to harvest apples at the same tree, the apples returned were exponentially depleted.
If the subject chose to go to a new, replenished tree, they incurred a travel time delay, during which the old tree faded and moved off the screen while a new tree moved on to the screen, followed by a response cue. Trees were never revisited; each new tree had never been harvested and its starting quality was correlated with subsequent outcomes (and thus signaled the quality of the overall tree) in experiment 2 and uncorrelated in experiment 1. The total time in the game was fixed and each choice’s reaction time was counted toward the ensuing harvest or travel delay. (Subjects who responded too slowly were penalized by a timeout lasting the length of a single harvest trial.) Thus, subjects visited a different number of trees depending on their harvest decisions, but apart from timeouts (which occurred on a negligible 1.7% of trials) they were able to influence the reward rate only through their harvest or leave choices, not their reaction times. This design ensured that the optimal choice policy was invariant to the speed of responding.
Subjects experienced four foraging environments in a counterbalanced block design. The decision-relevant parameters that define an environment are the harvest time, the travel time, the rate at which apples were depleted and the tree quality distribution. By varying travel time and depletion rate across blocks, we produced environments that differed in terms of richness, with some having a higher achievable average reward rate than others.
The environment changed every 14 minutes and this was signaled by a change in background color and a short message. Subjects were not instructed about the type of environment they were entering or what aspects of the environment had changed. They were also not told the form or rate of the depletion or the exact duration of a foraging environment but were informed that they would have a fixed and equal time in all four environments and that the experiment would last approximately one hour, that trees could never be revisited, that new trees had never been harvested and were a priori identical, and that harvesting a tree would tend to return fewer apples over time. They were told that they would be paid a half-cent for every apple collected and that they should try to collect as many apples as possible.
If the subject chose to go to a new, replenished tree, they incurred a travel time delay, during which the old tree faded and moved off the screen while a new tree moved on to the screen, followed by a response cue. Trees were never revisited; each new tree had never been harvested and its starting quality was correlated with subsequent outcomes (and thus signaled the quality of the overall tree) in experiment 2 and uncorrelated in experiment 1. The total time in the game was fixed and each choice’s reaction time was counted toward the ensuing harvest or travel delay. (Subjects who responded too slowly were penalized by a timeout lasting the length of a single harvest trial.) Thus, subjects visited a different number of trees depending on their harvest decisions, but apart from timeouts (which occurred on a negligible 1.7% of trials) they were able to influence the reward rate only through their harvest or leave choices, not their reaction times. This design ensured that the optimal choice policy was invariant to the speed of responding.
Subjects experienced four foraging environments in a counterbalanced block design. The decision-relevant parameters that define an environment are the harvest time, the travel time, the rate at which apples were depleted and the tree quality distribution. By varying travel time and depletion rate across blocks, we produced environments that differed in terms of richness, with some having a higher achievable average reward rate than others.
The environment changed every 14 minutes and this was signaled by a change in background color and a short message. Subjects were not instructed about the type of environment they were entering or what aspects of the environment had changed. They were also not told the form or rate of the depletion or the exact duration of a foraging environment but were informed that they would have a fixed and equal time in all four environments and that the experiment would last approximately one hour, that trees could never be revisited, that new trees had never been harvested and were a priori identical, and that harvesting a tree would tend to return fewer apples over time. They were told that they would be paid a half-cent for every apple collected and that they should try to collect as many apples as possible.
