Prioritizing new objects for eye fixation in real-world scenes: Effects of object–scene consistency

Abstract

Recent research suggests that new objects appearing in real-world scenes are prioritized for eye fixations and by inference, for attentional processing. We examined whether semantic consistency modulates the degree to which new objects appearing in a scene are prioritized for viewing. New objects were added to photographs of real-world scenes during a fixation (new object with transient onset) or during a saccade (new object without transient onset). The added object was either consistent or inconsistent with the scene's meaning. Object consistency did not affect the efficacy with which transient onsets captured attention, suggesting that transient motion signals capture attention in a bottom-up manner. Without a transient motion signal, the semantic consistency of the new object affected its prioritization with new inconsistent objects fixated sooner than new consistent objects, suggesting that attention prioritization without capture is a top-down memory-based phenomenon at least partially controlled by object identity and meaning.

Acknowledgements

JRB is supported by the British Academy and the Edinburgh Fund. JMH is supported by the US Army Research Office (W911NF-04-1-0078). We thank Devon Witherell for his help with data collection and two anonymous reviewers for their helpful comments on a previous version of this manuscript.

Notes

¹Although the extent to which perception is suppressed during a saccade may be a matter of some debate, we have demonstrated that under some circumstances objects that suddenly appear in (or disappear from) real-world scenes during a saccade are fixated no more often than expected by chance (Brockmole & Henderson, 2005a, 2005b). As such, in this context the transient motion signal was functionally eliminated by saccadic suppression.

²We note that in the baseline condition, compared to consistent critical objects, inconsistent critical objects enjoyed reliably longer individual fixation durations (463 ms vs. 326 ms), first pass gaze durations (853 ms vs. 441 ms.), and total fixation time (1862 ms vs. 996 ms), replicating previous research.

³In a previous report using the same consistent scenes used here, Brockmole and Henderson (2005a) found that in a condition where no onsets where used, 10% of fixations were directed to the critical objects. It would therefore be expected that the first object selected for fixation by the observer would be the critical object on approximately 10% of the trials, as observed here.

⁴By dividing trials into 500 ms viewing bins, our ability to detect an effect of semantic consistency in the new object condition on the order of hundreds of milliseconds may have been limited. To investigate this possibility, we calculated the elapsed time from the appearance of the new object to the start of the first fixation on that object when those first fixations began within the first 500 ms (viewing bin 1). In these situations, consistent objects were first fixated 286 ms after their appearance and inconsistent objects were first fixated 275 ms after their appearance, t(22) < 1. There is no evidence of a semantic consistency effect in the new object condition even in the very first viewing bin.