The model likely generalized well because it was grounded in a core set of gaze features that captured consistent orienting behaviors across the two studies, that is, fixation detection, the angular displacement between the current and previous fixation centroid, the angular displacement between the current and previous saccade centroid, the angular displacement between the current and previous saccade landing positions, vertical component of saccade acceleration skew, and saccade acceleration skew (
Figure 11). When it comes to the probability of fixation, there was a greater probability of fixation as people oriented to a target of encoding, suggesting that people fixate as they anticipate orienting toward a target of encoding. This finding is consistent with the hypothesis that the oculomotor system maintains fixation so that people can detect and precisely orient themselves toward encoding targets in their periphery (
Chen et al., 2021;
Schuetz et al., 2019). Additionally, the angular displacement between saccade centroids and saccade landing points was smaller as people oriented to the memory target, which suggests that gaze was focused in the region of the encoding target at the end of search and was more exploratory during early search periods. Furthermore, saccades were more variable in size during early search, which suggests that gaze patterns are more diverse when people search without the intent to orient to a target of encoding. Together, these findings suggest that the oculomotor system executes long, orienting saccades during early search and then executes small, fine-tuning saccades as it anticipates precisely orienting to, and encoding, a target (
Chen et al., 2021;
Schuetz et al., 2019). The angular displacement between fixation centroids was also smaller when people oriented to an encoding target. This finding provides converging evidence that the oculomotor system focused on the general region of the encoding target for visual analysis. Finally, a negative skew in the vertical and total saccade acceleration of gaze samples was observed, meaning that saccade decelerations were slower than accelerations as people oriented to the target. Prior work finds that saccades with faster accelerations have greater standard deviations in their saccadic end points (
Abrams et al., 1989), thereby suggesting that faster saccades might be subject to undershoot or overshoot a target (
Plamondon & Alimi, 1997). Therefore, saccade decelerations might have been slower to avoid overshooting or undershooting task-relevant target information (
Plamondon & Alimi, 1997). This result provides converging evidence that, as people orient toward a target of encoding, they limit exploratory sampling behaviors. Here, they execute fewer, smaller, and slower eye movements to avoid overshooting the target of encoding.