3.2.2. Brain-behavior relationships
In the next step, we correlated N250 amplitude differences between the
first and second part of the experiment with the performance in the easy
versus difficult face memory tests. As common in psychometrics, we used
accuracy as a memory measure in the difficult task and the inverted RTs
as a performance indicator in the easy tasks.
For the N250 amplitude differences to target faces the correlations were
small for both the easy versus the difficult memory tasks, r =
-0.035 and r = -0.070, respectively. For the N250 in response to
non-target faces, we calculated amplitude differences between
experimental halves separately for high- and low-distinctive faces. The
correlations with recognition performance of high-distinctive faces with
easy versus difficult task performance were r = -0.201 andr = -0.181, respectively. The corresponding correlations for
low-distinctive faces were r = -0.355 and r = -0.163,
respectively.
According to the permutation test described in Section 2.4, we can
reject the null hypothesis with a probability of 5% that all six
considered pairs of variables are independent. Hence, for at least one
pair of variables (i.e., the largest N250 low-distinct non-targets X2
vs. memory speed X5) there is a statistically significant linear
relationship. The same permutation test was used to test the H1 that the
second largest correlation coefficient (for the modulus) is greater than
zero. The corresponding null hypothesis could not be rejected even at a
level of 0.1. Thus, there is only one significant correlation, which is
between the change in N250 amplitude to low-distinctive non-target faces
and the psychometric face recognition speed (r = -0.355). The scatter
plot of corresponding data of this distribution is shown in Figure 3. Of
note, there are at least four data points where performance in the face
recognition test was rather slow. Although, upon closer inspection these
participants did not show any unusual behavior, we excluded these four
participants to assess the stability of the association. As a
consequence, this particular correlation dropped to r = -.25 and
the permutation test of all 6 correlations of interest failed
significance.