Supplemental Material for

“The Role of Color in the Implicit Memory Performance of Healthy Older Adults and Individuals With Alzheimer’s Disease.”

Toby J. Lloyd-Jones

Neuropsychology, Vol. 19, No. 1, 44-53

View article

Note on Experiment 1

Although there was equivalent priming for untransformed and transformed objects for both AD and HC groups in the response time data, this was not robust in the error data where there was no priming for transformed objects for the HC group. This is likely because in this kind of paradigm error data are generally less sensitive and more variable than response time data. However, there is a more important concern here: Perhaps there is equivalent priming for RTs to untransformed and transformed objects because the error rates differ for these two conditions (i.e., due to speed-accuracy trade off). To test this hypothesis, analyses of covariance with RT as the dependent variable and error rate as the covariate were conducted. This enabled examination of the least squares means, i.e., the dependent variable means adjusted for effects of the covariate. If equivalent priming in RTs for untransformed and transformed objects remains when the effects of error rate have been partialled out statistically, then we can be sure the result is not due to trade off between speed and accuracy. For both RTs and errors ‘priming scores’ were calculated for both untransformed and transformed objects, i.e., new minus same and new minus changed conditions (averaged over correct and incorrectly colored conditions for simplicity, and because this variable did not qualify the main effect of interest). Two analyses of covariance were then conducted; (1) with group as the between subjects factor (AD vs. HC), untransformed RT priming score as the dependent variable, and untransformed error priming score as the covariable; and (2) with group as the between subjects factor (AD vs. HC), transformed RT priming score as the dependent variable, and transformed error priming score as the covariable. The results were as follows. For the untransformed condition, there was a main effect of group, F(1,20)=6.88, p<.05, MSe=124266. For the transformed condition, there was a main effect of group, F(1,20)=8.69, p<.01, MSe=122373. These results are consistent with greater priming for the AD relative to the HC group, but the interest lies in examining the least squares means and comparing them to the original means calculated from Table 2 in the article. The priming score data (in ms) presented below does this.

Table

Untransformed and Transformed Priming Scores Comparing the Original Means with Least Squares Means Derived from an Analysis of Covariance

AD HC

Least Squares Means

Untransformed condition 549 195

(new-same condition)

Transformed condition 642 209

(new-changed condition)

Original Means (calculated from Table 2)

Untransformed condition 560 227

(new-same condition)

Transformed condition 626 176

(new-changed condition)

[AD=Alzheimer’s disease, HC=Healthy Older Adult Controls]

It can be seen quite clearly that, for the HC group, when the effects of error rate are statistically partialled out from the RT measure the difference between untransformed and transformed priming (195 ms vs 209 ms) is smaller than in the original RT data which does not take error rate into account (227 ms vs. 176 ms). Therefore, differences in error rate are not responsible for observing equivalent RT priming for untransformed and transformed conditions.