Abstract: Laboratory experiments with starlings, Sturnus vulgaris, were conducted to investigate the interaction between timing and cost-benefit considerations. The design simulated an environment in which food was distributed in patches. Patches contained a random number of food items (N=0-4) separated by a fixed inter-capture interval or fixed interval. All patches ended with sudden depletion. The time elapsed since the last prey capture was the only way to detect the depletion of the patch. Once the patch was depleted a new patch could be reached by travelling between two perches. Three measures of timing were taken: (1) rate of working for food as function of `waiting' time in a patch, (2) the time of the last response in a patch or `giving-in' time, and (3) the time at which travel was initiated or `moving-on' time. The fixed interval that characterized patches was varied between conditions. The mean time of the peak in working rate was consistently centred around the fixed interval, while the other two measures of timing kept a roughly linear relation to the fixed interval, with slope greater than one. In accordance with Scalar Expectancy Theory, variability in the three forms of timing was proportional to the magnitude of the fixed interval. The birds seemed to take account of this increase in variability as shown by the mean value of their giving-up criterion. These results imply that information-processing constraints are important for modelling behavioural optimality.