Paradigm and Hypotheses

This document plots the results of a 1-hour version of LTM AB. Subjects completed 20 blocks of 3 tasks in the following order. First, the subject was shown 12 objects for 4 seconds each and asked to remember those images for later report. The subjects then performed 8 trials of an Attentional Blink task - this time with the target images embedded in noise and in the midst of completely scrambled noise distractors. T1 was an object and T2 was a gabor patch turned 45 degrees either clockwise or counterclockwise. At the end of each trial, subjects reported which of two objects in a 2AFC was presented on that trial, and were asked to report which direction the gabor was turned. Following the AB task, subjects performed 4 trials of a 2AFC memory task where they reported which of two images they remembered seeing in the first part of the experiment.

Half of the AB trials presented a T1 image that were currently in LTM, whereas the other half presented a new image. The gabor was equally likely to be turned clockwise and counterclockwise. The LTM test task at the end of each block only included images that were seen in the first part of the experiment but not in the AB task, and the foils were completely new images.

This experiment is testing whether having an LTM representation of an object allows you to forgo encoding the image into WM, which should be evident as a decreased AB magnitude on trials in which an old image was T1.

Note: all error bars are within-subject SEM. T2 performance is plotted both overall and, as is customary in AB research, for trials in which T1 was correctly identified. Subjects who had a d’ of less than .5 in the LTM task were excluded from all T1 and T2 analyses.

Subject (N = 107) performance on Long-Term Memory 2AFC, before exclusion

d’

RT

Attentional Blink Task T1 performance (N = 85) after exclusion

T1 d’

T1 d’ based upon lag (2 or 8) and whether T1 was seen before (old or new). Error bars are within-subject SEM.

T1 RT

T1 RT based upon lag (2 or 8) and whether T1 was seen before (old or new)

Attentional Blink Task T2 performance

Overall T2 d’

Overall T2 d’ based upon lag (2 or 8) and whether T1 was seen before (old or new)

T2 d’ given correct T1 response

T2 d’ based upon lag (2 or 8) and whether T1 was seen before (old or new), only for trials in which T1 was accurately identified

T2 RT given correct T1 response

T2 RT based upon lag (2 or 8) and whether T1 was seen before (old or new)

Old/New d’ effect for each subject

Difference between old and new d’ for each subject at lag 2 (left) and lag 8 (right). We hypothesized that T2 performance should be better following old vs. new T1s at the short lag, where the presence of the LTM representation may allow subjects to sometimes forgo encoding and therefore let them encode T2. This effect should not be very pronounced at the long lag, given that the bottleneck in encoding should be alleviated. Therefore, we would expect subjects to show a positve d’ old-new difference at the short lag and not the long lag.

Old/New RT effect for each subject

Difference between old and new RT for each subject at lag 2 (left) and lag 8 (right). We hypothesized that T2 performance should be better following old vs. new T1s at the short lag, where the presence of the LTM representation may allow subjects to sometimes forgo encoding and therefore let them encode T2. This effect should not be very pronounced at the long lag, given that the bottleneck in encoding should be alleviated. Therefore, we would expect subjects to show a negative RT old-new difference at the short lag and not the long lag.

Statistics

T1 d’ ANOVA

Effect DFn DFd SSn SSd F p p<.05 ges partial_eta_squared
(Intercept) 1 84 1751.907 47.838 3076.227 0.000 * 0.954 0.973
trialLag 1 84 0.456 10.457 3.659 0.059 0.005 0.042
presOld 1 84 2.311 13.498 14.384 0.000 * 0.027 0.146
trialLag:presOld 1 84 0.456 11.916 3.213 0.077 0.005 0.037


T1 RT ANOVA

Effect DFn DFd SSn SSd F p p<.05 ges
(Intercept) 1 84 310798449.97 33314090.0 783.665 0.000 * 0.898
trialLag 1 84 12714.11 597042.0 1.789 0.185 0.000
presOld 1 84 670526.70 637808.6 88.309 0.000 * 0.019
trialLag:presOld 1 84 864.24 669736.7 0.108 0.743 0.000




T2 accuracy ANOVA (for trials in which T1 was correctly identified)

Effect DFn DFd SSn SSd F p p<.05 ges partial_eta_squared
(Intercept) 1 84 970.206 156.983 519.148 0.000 * 0.821 0.861
trialLag 1 84 14.583 28.298 43.289 0.000 * 0.065 0.340
presOld 1 84 3.437 15.128 19.086 0.000 * 0.016 0.185
trialLag:presOld 1 84 0.334 10.991 2.556 0.114 0.002 0.030


T2 RT ANOVA

Effect DFn DFd SSn SSd F p p<.05 ges
(Intercept) 1 84 2.701670e+08 28643161.4 792.302 0.000 * 0.895
trialLag 1 84 6.904846e+03 1589991.5 0.365 0.547 0.000
presOld 1 84 4.369725e+04 526702.6 6.969 0.010 * 0.001
trialLag:presOld 1 84 5.092387e+03 940086.7 0.455 0.502 0.000