Feature-specific encoding flexibility in visual working memory

1. Feature-specific encoding flexibility in visual working memory Kondo, A. & Saiki, J. (2012). PLoS ONE, 7(12), e50962.

2. Introduction • The binding problem • Multiple features of a single object could be bound together as a single unit in VWM • VWM stores features in an integrated way • It remains unclear how object features are bound together in VWM.

3. Present study (1a) How flexible object features could be bound in VWM? (1b) Are all visual features bound in the same way, or a specific feature plays a special role in binding? (2) Whether specific feature binding can be easily retrieved than other feature bindings?

4. About flexibility of binding • Changes in feature binding led to significant interference. • Binding occurs automatically when attention is directed to an object during encoding. • Feature binding are retrieved through their locations.

5. About flexibility of binding -- predictions • Whether all features of attended objects are always bound? • All object features are always bound:the task-irrelevant features will lead to interference. • Only task-relevant features are bound selectively and flexibly:the task-irrelevant features would not interfere with memory for relevant features

6. About generalization of features • Object file theory: • It is hypothesized that specific feature could play a special role in VWM binding. • The special role of location information in object file cannot be filtered out. • Present study evaluate the flexibility by asking participants to detect whether specific features changed.

7. About generalization of features -- predictions • Partial object file theory: • We can create object representations that only combine the task-relevant features. • There will be no difference in performance between conditions. • Full object file theory: • We always create representations that combine all features. • Strong interference will be observed in all conditions. • Feature-specific theory: • Encoding flexibility is feature-specific. • Strong interference will only be observed in the color-shape condition.

8. About retrieval difficulty • Whether specific feature binding can be remembered easier than other feature bindings. • Single-probe advantage:reduction of interference and/or facilitation of memory retrieval. • Is single-probe advantage feature-specific?

9. About retrieval difficulty -- predictions • Partial and full object file theories: • A comparable advantage across condition • Feature-specific theory: • Larger single-probe advantage in the color-shape condition, due to greater interference from location information.

10. Experiment 1A & 1B • 1A: simple change detection task • 1B: relevant-feature switch detection task • Probe and switch types conditions are all intermixed. • 7 colors, 7 shapes • Same/different response. • 32 trials for practice; 96 trials x 2 blocks in experiment

11. Result • 1A • 1B

12. Discussion • Support the prediction of object file theory. • But the single-probe advantage was not found in the color-location and shape-location tasks. • Suggests that this effect does not simply reflect a cost of number of objects to be compared with the memory representation. • Experiment 2 were designed to confirm the result from experiment 1 were not due to ceiling effect.

13. Experiment 2 • 24 practice trials • 120 trials x 2 blocks • Binding:required memory for color-location or shape-location. • Two-feature:required merely the list of colors or shapes.

14. Result • Relative difficulty across change type conditions was consistent across experiment. • Significant improvement in two-feature memory condition implied the result from Exp1 was not due to ceiling effect.

15. General discussion • Encoding flexibility is feature-specific, which is consistent with the object file theory. • Interference and a single-probe advantage only occurred in the color-shape condition. • The difference result between present study and previous studies may be result from different design.

16. General discussion • An important difference is that earlier studies mentioned above randomized presented feature designated as irrelevant from study to test, the present study did not. • VWM capacity is closely related to an individual’s ability to exclude irrelevant items from current tasks. • Another empirical question is whether the task-irrelevant, but selection-relevant, feature value can be filtered out. (using location cues and feature cues)

17. General discussion • A number of studies are currently limited to selection based on specific feature values. • An interesting future study would be to examine whether dimension-based selection can be carried out with memory representation in VWM.

18. Thank you