Abstract
Does being able to predict forthcoming scenes facilitate target detection? We report a study that explores this issue. Targets were shown in 40 photographs of real-world scenes, with participants learning the set of photographs across eight presentations. The scenes were taken from positions along a route. The repeating sequences were structured to be consistent with that route (route consistent condition) or randomised (route inconsistent condition). In a previous study (VSS, 2015), we reported that route consistency did not influence target detection when targets appeared randomly across scenes and locations. In a new experiment, targets were always presented in particular scenes and, for some scenes, always presented in a particular location (though target identity was randomly determined). As in our previous study, the participants' task was to find and fixate targets. Accuracy, reaction times (RTs) and eye movements were measured. Basic behavioural data showed responses to absent trials to be slower but more accurate than present trials, while RTs and error rates reduced over block. Examining target location (fixed vs. varied) showed that errors reduced over blocks for targets in fixed positions only. There was no effect of presentation sequence (consistent vs. inconsistent) on RTs or accuracy. However, eye movement data showed that scene order reduced fixation durations on present trials for the route consistent condition only. Additionally, the time to first fixate targets was shorter in the fixed vs. varied location condition. This pattern was reversed in the time to verify targets. We conclude that the results show an influence of being able to predict forthcoming scenes on target detection. The ability to predict forthcoming real-world scenes and target location both influence search strategy. We add this finding to previously reported scene context effects (e.g. Brockmole & Henderson, 2006).
Meeting abstract presented at VSS 2016