The temporal dynamics of visual short-term memory retrieval

Poster Presentation 43.318: Monday, May 20, 2024, 8:30 am – 12:30 pm, Banyan Breezeway
Session: Visual Memory: Encoding, retrieval

Tianye Ma1 (), Weiwei Zhang1; 1University of California, Riverside

Recent research on short-term memory (STM) has revealed a combination of continuous and categorical reports in color recall, indicating a complex interplay between these memory formats. Nonetheless, the temporal dynamics of recruiting these STM representations during retrieval have remained unclear. The current study aims to elucidate how memory retrieval of continuous and categorical information unfolds over time by examining participants' computer mouse cursor movement trajectories during the recall phase of a continuous estimation task. First, with empirically derived color categories from an independent color naming task, the data from the recall task exhibited a between-trial mixture of categorical and continuous recall of colors, replicating some previous findings. Second, we partitioned the entire time course of the recall trajectories into different bins and fit a categorical-continuous mixture model to each bin. The estimated model parameters showed a gradual increase in the categorical memory over time, accompanied by an incremental shift towards categorical centers in the recall trajectories. Third, a formal model comparison indicated that a model with no categorical memory responses outperformed other models in the early stages of recall, whereas the categorical mixture model peaked towards the end of the trajectory. This pattern was replicated in another continuous estimation task with a different response method, ruling out an alternative account based on bias in the decisional stage. Overall, our data suggests that the retrieval of continuous color short-term memory precedes categorical color memory retrieval. This observation aligns with the notion that during memory encoding and retention, continuous representations undergo a gradual transition towards discrete stable attractor states, correcting accumulated errors across time.