Abstract
There is evidence that attention and eye movements during scene viewing are preferentially allocated to semantically inconsistent objects compared to their consistent controls. However, there has been a dispute over how early during scene viewing such inconsistencies affect eye movement control. While the classical “octopus in a farmyard” study by Loftus and Mackworth (1978) and more recent studies by Underwood and colleagues (e.g., Underwood, Humphreys, & Cross, 2007) argue for extrafoveal detection of object-scene inconsistencies, other work has failed to find such early effects of semantic inconsistency on eye movement control (e.g., Henderson, Williams, & Hollingworth, 1999; Gareze & Findlay, 2007). The study presented here extends previous work by using highly controlled 3D-rendered images of real-world scenes instead of line drawings or photographs, which are either less realistic or more difficult to control for bottom-up saliency. In addition, we directly compared the effects of both semantic and syntactic object-scene inconsistencies on eye movement control during scene viewing. We therefore introduced syntactic object-scene inconsistencies (i.e., floating objects) in addition to semantic inconsistencies to investigate the degree to which they attract attention during scene viewing. In Experiment 1 participants viewed scenes in preparation for a subsequent memory task, while in Experiment 2 participants were instructed to search for target objects. In neither experiment were we able to find evidence for extrafoveal detection of either type of inconsistency. However, upon fixation both semantically and syntactically inconsistent objects led to increased object processing as seen in elevated gaze durations and number of fixations. Interestingly, the semantic inconsistency effect was diminished for floating objects, which suggests an interaction of semantic and syntactic scene processing. This study is the first to provide evidence for the influence of syntactic in addition to semantic object-scene inconsistencies on eye movement behavior during real-world scene viewing.