The mismatch between saliency and positions of fixation clusters can be attributed to the influence of top–down mechanisms, where attention to meaningful details of the objects determines the location of gaze. This result LY294002 fits well with data from human studies where the choice of fixation positions has been shown to be either driven by bottom–up (exogenous) or by top–down (endogenous) factors ( Cerf et al., 2008 and Mackworth and Morandi, 1967).
It has also been shown that the saliency model does not account for fixations that were directed to the eyes of humans ( Birmingham et al., 2009). Thereby, faces appear to play a particular role, being probably the most important visual stimuli in primate social communication ( Bruce and Young, 1998), as they can provide significant cues Z-VAD-FMK solubility dmso to intention and mental state of other individuals ( Anderson, 1998, Andrew, 1963, Bruce and Young, 1998 and Emery, 2000). Similar observations were found in non-human primates: monkeys make longer fixations on faces ( Guo et al., 2006), and respond appropriately to the expressions of other individuals ( Mendelson et al., 1982), and are able to recognize their faces ( Rosenfeld and Van Hoesen, 1979).
Psychological studies have shown that the sequences of saccades and fixations are relevant for perception (Noton and Stark, 1971b). In humans, during free viewing of still images for long time periods (i.e., > 10 s) saccade amplitudes decrease exponentially (Antes, 1974 and Unema et al., 2005). Pannasch et al. (2008) showed that fixation durations increase after the first 2 s of exploration, revealing a global image exploration that spans the first 2 s, followed by a local, feature exploration phase, evident after 4 s of exploration. The maximum exploration time in our study was 5 s, which could suggest that the higher probability of staying inside a cluster is a consequence of the late, local exploration phase. However, examination of the raw data (see for example Figs. 2A and B, and 5A) reveals that some consecutive fixations are separated by short saccades even during the first seconds of exploration. We find that the monkeys fixate
preferably at certain restricted locations on the images (identified as clusters of fixations), and that the eye movements between these clusters Histidine ammonia-lyase are not random. The Markov chain analysis revealed that the monkeys primarily make short saccades within a cluster of fixations. These short saccades are likely to be followed by a larger saccade that directs the gaze to a new position inside a different cluster. This finding is consistent with the hypothesis that large saccades to new areas are followed by local, short saccades to nearby positions for refinement of the percept (Körner et al., 1999 and Ullman, 1995). Further studies showed that applying a Markov model to humans freely viewing advertisements has revealed similar local vs. global exploration modes (Wedel et al.