The thesis examines how spatial expectations affect endogenous attention using visual psychophysics and information theoretic models. We consider three main questions (1) how probabilistic cues affect response latencies and discrimination accuracy and whether these effects exhibit the behavioural hallmarks of decision under risk. (2) Whether there are limitations to how spatial attention is distributed across multiple locations. We consider specifically the role of working memory capacity in setting these limitations when attending multiple locations. (3) Finally we investigate whether the reliability of endogenous, attentional cues can be learned.We carried out a number of experiments using central endogenous cues indicating one or more locations. Each location contained a random dot kinematogram (RDK), but only one RDK contained coherent motion. Experiment 2 was a simple reaction time task requiring a simple, speeded response to the appearance of supra-threshold expanding coherent motion. Experiment 6 employed a fine motion discrimination task. In all other experiments, a coarse discrimination task was used.
Experiments 1 and 2 used probabilistic cues, ranging from a non-informative 25% to a highly informative 86% reliability, in a motion discrimination accuracy task and a motion detection reaction time task, respectively. We found that cue reliability modulated the size of the validity effect for both accuracy and reaction times. However, a two-process model was consistent with a probability matching strategy in the motion discrimination task but an under-matching strategy in the speeded task.Experiments 3-5 compared motion discrimination performance following probabilistic, one location cues and multiple location cues, which provided the same amount of spatial information. With four RDKs performance was very similar for single and multiple location cues. We concluded that attention could be flexibly distributed over multiple locations with no costs or limitations. However, when six RDKs were used, motion discrimination accuracy was lower following cues indicating multiple locations than information matched, one location probabilistic cues, suggesting that limitations in distributing spatial attention do occur with more locations. When participants were required, in the same block of trails, to either recall the cued locations or discriminate motion at the cued locations, their recall of the cue locations was worse than blocks when they were only required to recall the cued locations. We conclude that attention to the cued locations interferes with spatial working memory of those same locations.
Experiment 6 used a probabilistic cue, whose reliability was not explicitly communicated to the participants and which changed several times across an experimental session. We find that only some of the participants showed behavioural evidence that they had learned changes in cue reliability. In those participants, who showed evidence of learning, learning took place over sequences greater than four trials.