Spatial clustering during memory search

Abstract

In recalling a list of previously experienced items, participants are known to organize their responses on the basis of the items' semantic and temporal similarities. Here, we examine how spatial information influences the organization of responses in free recall. In Experiment 1, participants studied and subsequently recalled lists of landmarks. In Experiment 2, participants played a game in which they delivered objects to landmarks in a virtual environment and later recalled the delivered objects. Participants in both experiments were simply asked to recall as many items as they could remember in any order. By analyzing the conditional probabilities of recall transitions, we demonstrate strong spatial and temporal organization of studied items in both experiments.

Figure 1.

In this theoretical example, the participant has just recalled the location Central Park, and there are only three possible locations that have not been recalled to which the participant can transition. Each possible transition has the corresponding pairwise similarity value shown. If the participant next recalls the farthest possible location (Galapagos), the transition will receive a percentile rank of 0; if Gettysburg is next recalled, the obtained value will be .5; and if Brooklyn Bridge is recalled, the percentile rank will be 1.

Figure 2.

Conditional response probability (CRP) as a function of spatial distance for Experiment 1. The x value for each point represents the mean of that bin’s spatial distance values (i.e., 1-spatial similarity value), while the y value represents the mean CRP across that entire bin of pairwise values. The probability of recalling a given location immediately subsequent to the previously recalled location decreased as the distance between those two locations increased (and thus as their spatial similarity decreased). The x values represent all possible spatial distances, binned into quintiles. Error bars are ± standard error of the mean. A: The spatial-CRP for location-only immediate free recall periods. B: The spatial-CRP for mixed-list immediate free recall periods. We observe higher conditional probabilities because there are fewer locations on mixed lists and thus fewer possible location-to-location transitions. C: The spatial-CRP for the final free recall periods.

Figure 3.

Conditional response probability (CRP) as a function of temporal lag for Experiment 1. The CRP as a function of lag (or lag-CRP) shows the probability of recalling an item from serial position i + lag immediately following an item from serial position i. A: The lag-CRP for location-only immediate free recall periods. B: The lag-CRP for mixed-list immediate free recall periods. C: The lag-CRP for the final free recall periods. Only within-list transitions were included in this analysis.

Figure 4.

A: An overhead map of the layout of the virtual town. Shaded blue areas represent locations of nonstore buildings. Shaded red areas represent locations of stores. Shaded green areas represent grass, and the small dark blue, brown, and yellow boxes represent mailboxes, benches, and street lights, respectively. B: A view down one of the streets in the town. C: An example store.

Figure 5.

Conditional response probability (CRP) as a function of spatial distance for Experiment 2. The CRP as a function of spatial distance (or spatial-CRP) shows the probability of recalling an item presented at a given store location immediately following recall of an item presented at a different store location. The x values represent all possible spatial distances between stores, binned into thirds. Error bars are ± standard error of the mean. A: The spatial-CRP for the store recall periods. B: The spatial-CRP for the immediate free recall periods. C: The spatial-CRP for the final free recall periods.

Figure 6.

Conditional response probability (CRP) as a function of temporal lag for Experiment 2. The CRP as a function of lag (or lag-CRP) shows the probability of recalling an item from serial position i + lag immediately following an item from serial position i. A: The lag-CRP for the immediate free recall periods. B: The lag-CRP for the final free recall periods. Only within-list transitions were included in the analysis.