The effects of working memory load on negative priming in an N-back task

Ewald Neumann Brain-Inspired Cognitive Systems (BICS) July, 2010

Example of a high memory load trial sequence involving a Congruent and an Incongruent attention display. Correct answer to memory item is 4.

500 ms

1500 ms

850 ms

2 to 4 attention displays are

presented for 500 ms, eachfollowed by a 1250 ms blank

responseinterval.

3000 msLow

LoadHighLoad

Exp. 1 N = 10Interference

de Fockert, Rees, Frith, & Lavie (2001). The Role of Working Memory in Visual Selective Attention. Science, 291.

500 ms

1500 ms

850 ms

2 to 4 attention displays are

presented for 500 ms, eachfollowed by a 1250 ms blank

responseinterval.

3000 ms

LowLoad

HighLoad

Exp. 2 N = 6Interference

de Fockert, Rees, Frith, & Lavie (2001). The Role of Working Memory in Visual Selective Attention. Science, 291.

• RTs in the selective attention task revealed thatinterference from incongruent (versus congruent) distractor faces was significantly greater with high working memory load than with low working memory load.

Summary of results and conclusions

de Fockert, Rees, Frith, & Lavie (2001). The Role of Working Memory in Visual Selective Attention. Science, vol. 291.

• High memory load leads to greater intrusion of irrelevant distractors, because of difficulty in prioritizing targets from distractors.

• Low memory load enables more successful filtering, or blocking out of irrelevant distractors, thus minimizingintrusion from distractors.

Why is it important to eliminate Congruent stimuli in Stroop-like conflict tasks?

Because it is known that having congruent stimuliin Stroop-like conflict tasks induces subjects to maintain less attentional selectivity solely to the target.

That is why participants consistently show greateramounts of interference in Incongruenttrials as a function of increasing the proportion ofCongruent trials in such conflict tasks (e.g., Lindsay & Jacoby, 1994).

Why is it important to eliminate Congruent stimuli in Stroop-like conflict tasks (cont’d)?

Since one of the stated objectives in using the present paradigm was to investigate selective attention, rather than more diffuse or divided attention, it is important to try to induce maximal attentional selectivity.

Including Congruent stimuli in the task could be counter-productive in that regard.

In addition, assessing “interference” by contrasting Congruent vs. Incongruent conditions is another potential problem in their study. By doing so, it is not possible to disentangle the interference cost from incongruent stimuli from the potential benefit in processing due to congruent stimuli.

To illustrate, consider their results from Experiment 1 again:

500 ms

1500 ms

850 ms

2 to 4 attention displays are

presented for 500 ms, eachfollowed by a 1250 ms blank

responseinterval.

3000 msLow

LoadHighLoad

Exp. 1 N = 10Interference

de Fockert, Rees, Frith, & Lavie (2001). The Role of Working Memory in Visual Selective Attention. Science, 291.To assess “interference” more purely, the Incongruent conditionshould be contrasted with a neutralControl condition, not a Congruent condition.

Why is it important to have an “ignored repetition” (negative priming) condition in Stroop-like conflict tasks?

In some instances negative priming effects can provide a more sensitive behavioral index, than concurrent interference effects, regarding the depth to which distractors have been processed.

Why is it important to have an “ignored repetition” (negative priming) condition in Stroop-like conflict tasks?

Name the color of the “odd man out” letter taskProblems with inferences from null results X X X X X vs G R E E N (no Stroop interference)

X X X X B L U E (significant negative priming) The Stroop color-naming effect has often been taken as evidence for the automaticity of word processing. However, Besner et al. (1997) reported that coloring a single letter instead of the whole word eliminated the Stroop effect. From this finding, they concluded that word processing could not be purely automatic, since it can be prevented. Mari-Beffa et al. (2000) asked whether the elimination of the Stroop effect is sufficient evidence for concluding that the word is not processed. Combining Besner et al.'s manipulation with a negative-priming procedure, she found intact negative priming from the prime color word in the absence of a Stroop effect. This result clearly indicates that the meaning of the prime word was processed. These findings highlight the importance of using converging methods to evaluate lack of processing.

Main goal of the present study:Test implications of the working memory load theoryin a new experiment designed to eliminate some of the potential flaws I just outlined.

Method - regarding memory load: The N-back task is used extensively as a way to manipulate working memory (WM) load. Typically, in the N-back task participants are presented with a stream of stimuli, and the task is to decide for each stimulus if it matches the one N items before.

The WM processing load can be varied systematically by manipulating the value of N. WM load is deemed to be low in a 1-back task and high in a 2-back task.

Method - regarding selective attention: Because this study is designed to investigate how WM and visual selective attention interact, this N-back task also involved selective attention.

Sample 1-back task: Respond “same” if the immediately preceding picture is of the same object, otherwiserespond “different.”

Red Picture Targets (ignore green letters)

Schematic illustration of stimulus configuration

Sample 2-back task: Respond “same” only if a picture is of the same objectthat was 2 pictures back, otherwiserespond “different.”

Red Picture Targets (ignore green letters)

ResponseIgnored Repetition Control

Sample experimental trials

Same(bird then bird)

Different(ashtray then bird)

Different(bird then finger)

Probe Displays

Prime Displays

Insertion here of an unrelated picture

creates a 2-back trial.

1-back sample stream of stimuli

2-back sample stream of stimuli

ResponseIgnored Repetition Control

Sample experimental trials

Same(bird then bird)

Different(ashtray then bird)

Different(bird then finger)

Probe Displays

Prime Displays

Insertion here of an unrelated picture

creates a 2-back trial.

Negative Priming (RT) Negative Priming % Errors

1-backLow Load

2-backHighLoad

1-backLow

Load

2-backHighLoad

Results Mean RT and % Error for Control vs. Ignored Repetition (IR) conditions as a function of Low vs. High Working Memory Load

Results summary and implications:

RTs and Errors much greater in 2-back than 1-back task suggests WM load manipulation was successful.

RTs significantly longer in Ignored Repetition (IR) condition than Control condition, and Errors occurred significantly more often in the IR than Control condition, which indicates that ignoring a picture’s name can impede response to the picture on the following trial. This negative priming effect was produced in both the 1- and 2-back versions of the task.

For RTs there was no interaction between the N- back and priming conditions; however, the error rate analysis for this interaction was significant. In this case, it appears that greater WM demands in the 2-back task did yield an exaggerated difference in the error rates between the IR and Control conditions, compared to the 1-back task. This provides some support for the idea that participants may be better able to block or filter out distractors under low WM load.

Summary and Conclusions regarding how WM and selective attention interact:

At least according to the error rates, the present findings could be interpreted to suggest that the working memory system is modulating attention and affecting processing within the brain’s recognition system.

High WM load might reduce the efficiency of early selection and increase the effects of irrelevant stimuli.

In contrast, Low WM load might enable relatively increased efficiency of early selection so that participants are better able to block or filter out the irrelevant distractors under low memory load conditions, as proposed by de Fockert et al. (2001).

End