Role of time in binding features in visual working memory

Previous research on feature binding in visual working memory has supported a privileged role for location in binding an object’s non-spatial features. However, humans are able to correctly recall feature conjunctions of objects that occupy the same location at different times. In a series of behavioral experiments, we investigated binding errors under these conditions, and specifically tested whether ordinal position can take the role of location in mediating feature binding. We performed two dual-report experiments in which participants had to memorize three colored shapes presented sequentially at the screen center. When participants were cued with the ordinal position of one item and had to report its shape and color, report errors for the two features were largely uncorrelated. In contrast, when participants were cued e.g. with an item’s shape and reported an incorrect ordinal position, they had a high chance of making a corresponding error in the color report. This pattern of error correlations closely matched the predictions of a model in which color and shape are bound to each other only indirectly via an item’s ordinal position. In a third experiment, we directly compared the roles of location and sequential position in feature binding. Participants viewed a sequence of colored disks displayed at different locations, and were cued either by a disk’s location or its ordinal position to report its remaining properties. The pattern of errors supported a mixed strategy with individual variation, suggesting that binding via either time or space could be used for this task.

discrete response errors in Experiment 1, we analyzed the effect of cue feature similarity on response choice.
The response options given in each trial included the target feature, the two non-target features, and a foil feature (except for the ordinal position report, where no foil response was possible).Each of the non-target response options was associated with a certain cue feature in the sample array.We grouped the non-target response options according to how similar their associated cue feature was to the given cue in each trial (which is identical to the cue feature of the target item), and determined the probably that a non-target response option was selected separately for each group.
For the ordinal cue condition, we grouped the non-target response options according to each item's ordinal distance to the target item (in discrete steps).For the shape cue and color cue conditions, we discretized the feature distance (absolute angular difference of feature values in the circular shape and color spaces) into four bins.Bin centers were chosen such that the expected number of items falling into each bin would be equal (due to the minimum distance between the features of different items within a trial, the distributions of feature distances was not uniform).We determined the selection probability for each group as the ratio between the number of times a non-target response option from each group was selected and the number of times a non-target response option from each group was offered.
Errors in the response selection may also occur due to imprecision in memory for the report feature, which may result in selection of a (swap or foil) response option that is similar to the target's report feature.We grouped foil and swap response options based on their ordinal or feature distance to the target feature in the same way as described for cue similarity, and determined the selection probability for each group.
We assessed the influence of cue and report feature similarity on response selection using one-way repeated measures ANOVAs (in the case of ordinal position with only two possible distance values, this is equivalent to a paired samples t-test).For cue feature similarity, foil responses are not included in the analysis, but are shown in the figures for comparison (proportions of swap and foil responses in each task were compared in the main results).

Results
Results for Experiment 1a are shown in Figure A1.In the ordinal cue condition, there was a significant effect of cue similarity (i.e., ordinal proximity) on the selection of non-target response options both for the color report (F (1) = 5.75, p = 0.040) and the shape report (F (1) = 11.7,p = 0.008).Response selection in the color report also showed a significant effect of report feature similarity (F (3) = 3.05, p = 0.046), whereas we found no such effect in the shape report (F (3) = 0.73, p = 0.54).In the shape cue condition, we did not find an effect of cue similarity in the color report (F (3) = 2.11, p = 0.12), but there was a significant effect in the ordinal position report (F (3) = 3.97, p = 0.018).The effect of report feature similarity in the shape cue condition was not significant for the color report (F (3) = 1.07, p = 0.38), but it was significant for the ordinal report (F (1) = 16.6, p = 0.003).
The corresponding results for Experiment 1b are shown in Figure A2.In the ordinal cue condition, response selection was affected by cue similarity for both the shape report (F (1) = 10.0, p = 0.011) and the color report (F (1) = 6.09, p = 0.036).There was no significant effect of report feature similarity in either report (shape: F (3) = 1.11, p = 0.36; color: F (3) = 2.24, p = 0.11).In the shape cue condition, there was no effect of cue similarity in the color report (F (3) = 0.46, p = 0.71), but we did find a significant effect in the ordinal report (F (3) = 3.46, p = 0.030).Similarity in the report feature likewise did not have an effect on color report (F (3) = 1.87, p = 0.16), but there was a highly significant effect on the ordinal report (F (1) = 36.4,p < 0.001).
In summary, effects of feature similarity were predominantly observed for ordinal position (where feature similarity takes the form of temporal proximity in the sample

Figure A2
Effects of feature similarity on response selection for each task condition and response in Experiment 1b, shown in the same format as in Figure A1.sequence).It should be noted that these effects may be driven by the general recency effect described in the main results, which would likely produce more swap errors between the first two items in the sequence than between the first and the third item.Effects of feature similarity in color and shape were relatively limited, indicating that the large minimum distance between feature values within a trial largely prevented confusion of items features.
The finding that cue similarity had a greater effect in the second (ordinal) report than the first report for both the shape cue condition (Experiment 1a) and the color cue condition (Experiment 1b) is consistent with the assumption of indirect binding via the ordinal position.
Figure A1Effects of feature similarity on response selection for each task condition and response in Experiment 1a.The probability of selecting a certain non-target response option is shown as a function of the distance between the cue feature associated with that option and the actual cue (a), and as a function of the similarity between the report feature and the report feature of the target item (b).For the effects of cue feature similarity, the probability of selecting the foil response option (for shape and color report) is shown separately for comparison.Data points show mean value across participants, and error bars indicates ±1 SE.The gray shaded area indicates the minimum feature distance between all items in a trial.