Feature versus object in interpreting working memory capacity

Abstract

The debate about whether the capacity of working memory (WM) varies with the complexity of memory items continues. This study employed novel experimental materials to investigate the role of complexity in WM capacity. Across seven experiments, we explored the relationship between complexity and WM capacity. The results indicated that the complexity of memory items significantly affects WM capacity. However, given the non-linear relationship between complexity and WM capacity, we propose that WM may not allocate resources directly to each individual item. Instead, it might integrate these items to some extent before storage.

Fig. 1. Experimental materials.

A Sample memory items in simple and complex conditions. The dotted lines were included for better illustration. They were not presented to participants on the experimental monitor. B Samples for experiment materials. There were four items in each sample.

Fig. 2. The response time in three conditions.

The dependent variable was the visual search rate, estimated by taking the slope of the line relating the target-presentation reaction time to the quantity of items in the search array in the correct response trials. the visual search rate in the complex condition was significantly higher than in the simple condition, and the visual search rate in the colored square condition was significantly higher than in the simple condition. Error bars indicate 95% CI, and the same is true below.

Fig. 3. The averaged K values in simple and complex conditions.

In simple condition, the K values increased with the number of items. In complex condition, the difference between set size was not significant.

Fig. 4. The averaged K values in the simple, moderate, and complex conditions.

In each set size condition, the K values in the simple condition were significantly higher than those in the moderate condition, which were significantly higher than those in the complex condition.

Fig. 5. The averaged K values at presentation times of 100 ms, 500 ms, and 1000 ms.

There were significant main effects in complexity and presentation time. But the interaction between complexity and presentation time was not significant.

Fig. 6. The averaged K values in the simple, moderate, and complex conditions.

In the 2 items condition, there was no significant difference between the simple and moderate conditions, with K values significantly higher in the moderate condition compared to the complex condition. In the 4 items and 6 items conditions, the K values were significantly higher in the simple condition compared to the moderate condition, and in the moderate condition compared to the complex condition.

Fig. 7. Mean amplitude of SPN in the simple, moderate, and complex conditions.

A In the 2 items condition, the amplitude of moderate condition was significantly higher than that of simple condition, and the amplitude of complex condition was significantly higher than that of moderate condition. B In the 4 items condition, there were no significant differences. C In the 6 items condition, there were no significant differences.

Fig. 8. Experimental materials.

The target was included in the search array in half of the trials and excluded in the other half. Participants were asked to indicate whether the target was present or absent as quickly as possible.

Fig. 9. Procedure for Experiment 2.

In this example, four memory items were presented to a participant.

Fig. 10. Procedure for Experiment 7.

In this example, four complex memory items were presented to a participant. The items on which the arrow points are different between the two arrays.