Abstract
Lower level decisions involving the perceptual and perceptuo-motor systems (e.g., choosing an aim point for a hand movement) appear near-optimal. High level decisions (e.g., choosing between financial options), on the other hand, are commonly described as sub-optimal. In most studies, however, methods differ between investigations of perceptual and cognitive decision making. To facilitate comparison, we studied both in a common context: deciding how much time to spend on a task. Our experiments involved measuring time-accuracy relationships and a decision stage. In the latter, observers were given a limited amount of time to complete an unspecified number of trials. They faced a choice of completing many trials with low accuracy, or completing fewer trials but with a higher accuracy. Observers were told the values of correct and incorrect responses, were paid a performance related bonus, but received no feedback. To assess performance, we mapped observers' choices onto efficiency functions derived from the independently measured time-accuracy relationships. When the underlying task was perceptual (motion discrimination) observers made highly efficient decisions. Moreover, performance was excellent whether the task was easy or hard. Apparently, people have good knowledge of their task performance and use it to make highly efficient decisions. For asymmetric reward structures people changed their behaviour in a qualitatively appropriate manner (e.g., slowed down), but did not behave in a quantitatively appropriate manner (slowed down by the right amount). With baseline perceptual performance established, we explored cognitive performance (mental arithmetic & mental rotation) under a symmetric reward structure. Performance was virtually indistinguishable from perceptual performance. Finally, we obtained excellent perceptual- and cognitive performance even when task difficulty changed unpredictably from trial to trial. The results may be interpreted as optimisation of actions, rather than specific sub-systems, in an uncertain environment.
Andreas Jarvstad was supported by the ESRC, UK.