Planning sequences of arm - hand movements to maximise expected gain

We examined how subjects plan movement sequences in a 'foraging' task involving three disks that appeared simultaneously at a small number of possible locations on a touch-screen monitor. Subjects received monetary rewards for the disks they touched within 1.6 s. Disk colour signaled the value. The subjects could choose to attempt three disks in any order and could choose to attempt fewer than three disks. The dependent variable was the estimated probability of hitting each disk ( pA, pB, pC) as a function of sequence. It might be (0.9, 0.6, 0.5) for sequence ABC but (0.9, 0, 0.9) for sequence AC (no attempt to hit B). The sequence ABC offers the higher probability of hitting all three disks but, if A and C are made valuable enough, the sequence AC has higher expected gain. We tested whether subjects planned movement sequences that maximised expected gain. The experiment consisted of three sessions. In the first session, ten naive subjects were trained until their performance was stable. In the second session, we measured how accurately each subject could execute each of the 12 possible sequences of length 2 or 3 on three disks of equal value. In the last session, the values of the disks were altered by amounts that varied between blocks of trials. Subjects were told the values of the colour-coded disks before each block and were free to choose any sequence. On the basis of second-session performance we could predict the strategy each subject should adopt on each trial in order to maximise expected value for each assignment of values to disks. Subjects did change strategies in the expected direction but typically favoured strategies that maximised probability of hitting three disks over the maximum expected gain strategy.