Gaze patterns reflect the retrieval and selection of memories in a context-dependent object location retrieval task

Abstract

Selective retrieval of context-relevant memories is critical for animal survival. A behavioral index that captures its dynamic nature in real time is necessary to investigate this retrieval process. Here, we found a bias in eye gaze towards the locations previously associated with individual objects during retrieval. Participants learned two locations associated with each visual object and recalled one of them indicated by a contextual cue in the following days. Before the contextual cue presentation, participants often gazed at both locations associated with the given object on the background screen (look-at-both), and the frequency of look-at-both gaze pattern increased as learning progressed. Following the cue presentation, their gaze shifted toward the context-appropriate location. Interestingly, participants showed a higher accuracy of memory retrieval in trials where they gazed at both object-associated locations, implying functional advantage of the look-at-both gaze patterns. Our findings indicate that naturalistic eye movements reflect the dynamic process of memory retrieval and selection, highlighting the potential of eye gaze as an indicator for studying these cognitive processes.

Figure 1

Behavioral scheme for learning of the object locations and retrieval of the learned locations. (a) Locations associated with each object are indicated on the background screen. The six dark grey squares were displayed as a background to provide spatial reference points. The black dots positioned at the four corners of the screen indicate the starting point locations. (b) The timeline of the behavioral task scheme. The selective location retrieval task was conducted before the object-location learning task. (c) The task design of the object-location learning task. (d) The task design of the selective location retrieval task (SLRT). Notably, delay periods 1 and 2 were utilized to examine the relationship between the gaze patterns and memory retrieval.

Figure 2

Learning of object locations over days. (a) Example joystick responses of a participant in the selective location retrieval task (SLRT) over the retrieval days. The cyan crosses (‘x’) indicate the participant’s response of the four trials with object D1 for each day. The location of the target object (D1 location and red fractal object) is displayed on the screen. (b) Boxplot diagram of the error distance of the joystick responses in the SLRT. The error distance represents the Euclidian distance between the joystick response location and the corresponding answer location for each trial. Each individual dot on the plot indicates the average error distance of all trials for a participant (n = 20). n.s. not significant.

Figure 3

Gaze time reflects the retrieval of location memories. (a) Illustrative spatial windows for categorizing gaze locations. The green and orange circles represent the target (for this example: A1) and competitor windows (A2), respectively, with a radius of a 4.5° visual angle. (b) Boxplot diagram showing gaze time across trials by gaze location type during delay period 1 in the SLRT. Each individual dot on the plot indicates the total gaze time of all trials for each gaze location type of each participant (n = 20). (c) Boxplot diagram showing the gaze time across trials by gaze location type during delay period 2 in the SLRT. Each individual dot on the plot indicates the total gaze time of all trials for each gaze location type of each participant (n = 20).

Figure 4

Change of gaze patterns as learning progressed and their correlation with memory accuracy. (a) Two example gaze patterns during retrieval phases (delay periods 1 and 2) in the SLRT. The blue and red traces represent the eye positions during delay period 1 and delay period 2, respectively. (b) Boxplot showing the percentage of look-at-one trials during delay period 1. (c) Boxplot showing the percentage of the look-at-both trials during delay period 1. (d) Boxplot depicting the error distance categorized by the different gaze patterns in the trials. Each individual dot represents the mean error distance of each participant in the corresponding gazing pattern trials (n = 20). Blue and red colors indicate the look-at-competitor (comp) and look-at-both trials (both), respectively. Yellow color represents the look-at-target trials (tgt). *p < 0.05, ***p < 0.001.