Gaze-centered gating, reactivation, and reevaluation of economic value in orbitofrontal cortex

Economic choice often involves sequential consideration of options until a decision is made. However, neuroeconomics typically overlooks the dynamic nature of this deliberation process. Existing research suggests a causal link between eye movements and choice deliberation, with longer fixations on chosen options. While sequential fixations may seem inherent to visual processing, sequential evaluation might be a default computational strategy, advantageous for focused evaluation compared to simultaneous assessment of all options. This is supported by computational benefits, anatomical considerations, and evolutionary perspectives. Previous studies have shown that neurons in the OFC and vmPFC modulate their firing rates according to the economic value of choice targets or gaze transitions. However, it remains unclear how many options can be simultaneously encoded, the extent of parallel value computation and comparison, and the influence of gaze and attention. To address these questions, a paradigm decoupling visual encoding and evaluation is necessary, achieved here by interleaving offer presentations with empty screen delays. The study hypothesized that the 'look-at-nothing' effect, a tendency to gaze at empty locations of previously presented stimuli, is associated with the retrieval of choice options for reevaluation.

A substantial body of research indicates the significant role of eye fixations in decision-making. Studies demonstrate that individuals tend to examine options sequentially before committing to a choice, and that the duration of fixation correlates with the final choice. These observations suggest that eye movements play a causal role in the decision-making process. Neural activity in OFC and vmPFC correlates with the economic value of visual choice targets. Some studies suggest that distinct neural populations independently compute the value of each offer. Others show that value-coding regions represent offer values sequentially, not in parallel. The existing literature is unclear regarding the capacity for simultaneous encoding of multiple offers, the extent of parallel value computation, and the interplay between gaze patterns and these processes. Prior research largely focused on tasks where offers remain continuously visible, making it challenging to separate the visual encoding of offers from their evaluation. This study aimed to resolve this by introducing delay epochs where offers are invisible, enabling a clearer assessment of the sequential nature of offer evaluation.

Two male macaque monkeys participated in a visually cued, two-alternative reward gambling task. Two vertical bar stimuli (offers) were presented sequentially at opposite screen sides, each followed by an empty screen delay (600ms). Subjects freely moved their gaze throughout. After a re-fixation period, the offers reappeared side-by-side, prompting a saccade to select an offer. Offer reward magnitude (small, medium, large) was color-coded, and the probability of reward was indicated by bar height. Expected value (EV) = magnitude * probability, and risk was defined as variance. Subjective value (SV) was computed via logistic regression of choices. Oculomotor data were collected using an infrared eye-monitoring system (1kHz). Neural activity was recorded from 248 neurons in OFC areas 11 and 13 using multi-contact V-probes. Behavioral data analysis included logistic regressions modelling choice probability as a function of EV differences, looking times, and SV. Neural data analysis used linear regression models, comparing spike counts against offer EVs and SV. Trials were categorized by gaze conditions (LookL, LookR) and analyzed for gaze-dependent encoding and reactivation. Finally, choice-related activity was analyzed using residuals of linear models, incorporating SVs and choice, to assess reevaluation during the delay epochs.

Both macaques successfully performed the task, showing value-based choices. Eye position closely tracked the location of visible offers. A 'look-at-nothing' bias was observed during delay epochs, with gaze directed towards empty locations where previous offers appeared. Longer fixations on either visible offers or empty locations predicted choice. A logistic regression model showed that the fraction of time spent looking at each side predicted choice, even after controlling for EV, magnitude, and risk. OFC neural activity encoded the value of the gazed offer. Crucially, gazing at an empty location reactivated the value encoding of the previous offer at that location, even when a different offer was presented and encoded in between. This reactivation correlated with choice, suggesting active reevaluation. Overlapping cell populations participated in value encoding and reactivation. Analysis time-locked to midline-crossing gaze shifts confirmed that reactivation occurred following gaze shifts to the previously presented offer's location. Analysis of regression residuals showed that residual activity fluctuations during reactivation, which could not be explained by the offers' subjective value alone, predicted choice. This indicates that neural activity in OFC reflects a reevaluation process, not just a passive memory trace.

The study's findings demonstrate a previously unrecognized link between gaze, neural reactivation, and reevaluation during economic decision-making. The results support the hypothesis of sequential value encoding, where the OFC processes one offer at a time, leveraging gaze to facilitate evaluation and reevaluation, even for invisible offers held in working memory. The observation that gaze shifts to an empty location caused reduced encoding of the previous offer's value implies that OFC might not retain comprehensive memory of all considered options, focusing instead on the most recently sampled item. This aligns with the theory that sequential, focused evaluation is more efficient than parallel processing of multiple options. While a motor preparation component cannot be entirely excluded, the results suggest that gaze-dependent reactivation contributes substantially to decision-making by enabling memory recall and subsequent reevaluation.

This research provides compelling evidence for an active role of gaze in economic choice, highlighting its contribution to memory reactivation and reevaluation within the OFC. The findings support a sequential model of value encoding, where gaze directs neural resources to a single offer at a time. Further studies could investigate the causal relationship between gaze and reevaluation, examining other brain areas involved and exploring the interaction between gaze and other attentional mechanisms.

The study was conducted with two macaque monkeys, limiting generalizability to humans. The specific experimental design might not fully capture the complexities of real-world decision-making. While the analysis controlled for various factors, other influences on choice or gaze behavior might exist. Further research with larger sample sizes and across different species would strengthen the conclusions.