Experiment 3: Neutral list 1 and negative list 2

Consistent with Wessel & Merckelbach (2006), using pure lists of one valence leads to similar behavioral results for both emotional categories. Thus, within experiments using only one emotion, emotional material had no influence on the behavioral pattern of directed forgetting. However, as recall for unpleasant material has been shown to be superior when compared with neutral material (e.g. Hamann, 2001), directed forgetting might be modulated when emotional and neutral material are contrasted within one experiment. In experiment 3 a neutral list was followed by an unpleasant list. Assigning neutral words as to-be-forgotten and the following unpleasant items as to-be-remembered might enhance the directed forgetting effect due to memory enhancing effects of emotional material. Additionally, to the investigation of memory enhancing effects of emotional material, the experiments allows to investigate effects of context change:

During the massed presentation of words from one valence category in each list emotional contexts and/or a mood might be created and mental context changes might occur in response to changes of list valence.

3.3.1 Method Subjects

Thirty subjects participated in this experiment (14 male, 16 female) with a mean age of 22.79. Two data sets were excluded due to extreme alpha waves and another set due to extreme artifacts leaving 27 data sets (13 male, 14 female) for analysis. All subjects were right-handed (assessed according to Annett, 1979; Oldfield, 1971) and had normal or corrected to normal vision. Informed consent was provided and all received payment (15 €) or course credit for participation.

Material

Again, 80 words were chosen from the word pool. Half of the words were unpleasant and half neutral. The unpleasant words had a mean valence score of 2.5 (SE: 0.17) on a scale from 1 (very unpleasant) to 9 (very pleasant) and a mean score of 6.17 (SE: 0.16) on an arousal scale ranging from 1 (very un-arousing) to 9 (very arousing). The neutral words were rated on average as 5.3 (0.17) on the valence dimension and had a mean arousal score of 2.66 (SE: 0.16). The unpleasant and neutral words differed significantly in regard to their valence [F(1, 78)=131.95, p<.001] and arousal scores [F(1, 78)=256.52, p<.001]

while they were matched for concreteness, frequency, and word length.

All unpleasant nouns were arranged into two lists consisting of 20 words each with no differences concerning concreteness, frequency, and word length. The same was done for the neutral nouns resulting in two unpleasant and two neutral lists.

Design

The design was identical to the ones in experiments 1 and 2 with the exception that here list 1 in both conditions was comprised of neutral words and list 2 of unpleasant ones.

EEG recording, data analysis, and statistical analysis

The EEG recording, data analysis, and statistical analysis were identical to those used in experiment 1 and 2. In regard to the data analysis, visual inspection of the data revealed that between 550 and 750 ms list 2 of the F condition yielded more positive-going ERPs than list 2 of the R condition in a posterior regions. This sensor group was determined for statistical analysis (figure 15).

3.3.2 Results Recall data

The mean proportions of recalled words for the different conditions are displayed in table 6. A two-way ANOVA with the factors condition (F condition, R condition) and list (list1, list2) was calculated and showed better recall of list 2 compared to list 1 [F(1, 26)=14.80, p<.001]. Directed forgetting occurred as interaction of condition and list [F(1, 26)=7.49, p<.05, fig. 14 right] reflecting trends for both costs (p=.061) and benefits (p=.068).

Additionally, the mean difference of list 1 and list 2 was negative in both conditions reflecting reduced PI with the stronger reduction in the F condition (-.161, p<.001) compared to the R condition (-.029 p=.39). The scatterplot (fig. 14 left) reveals that most subjects recalled more list 1 items in the R condition than in the F condition (costs, more black stars above the linear line) and that also most subjects recalled more list 2 items in the F condition than in the R condition (benefits, more grey dots beneath the line).

table 6: Details for the recall data. Including mean proportion of recalled words (standard error in brackets) and significant results for costs and benefits (p<.001***).

mean difference

F condition R condition

F condition – R condition

list 1 (ntr) .211 (.041) .278 (.036) -.067 (p=.061, costs)

list 2 (neg) .372 (.026) .307 (.029) .065 (p=.068, benefits)

mean difference

-.161***

list 1 – list 2 -.029 (p=.39)

figure 14: Left: Scatterplot of the individual recall data. The line describes a hypothetically symmetrical recall of list 1 and 2 in the different conditions. Right: Bar plot of the interaction condition x list.

EEG data

550 - 750 ms: A repeated measure ANOVA with the factors condition (F condition, R condition) and list (list 1, list 2) was calculated. A main effect of list [F(1, 26)=5.50, p<.05] occurred reflecting more going activity during list 2. Enhanced positive-going activity was found for list 2 in the F condition compared to list 1 (p<.001) and the two lists of the R condition (p<.05) reflected by the interaction between condition and list [F(1, 26)=6.89, p<.05, fig. 15].

figure 15: Effect of the increased positivity during list 2 of the F condition compared to list 2 of the R condition illustrated by centro-parietal electrodes and the difference plot (F list 2 – R list 2).

3.3.3 Discussion

An effect of directed forgetting of a neutral list when followed by an unpleasant list was found reflected by the condition x list interaction. A tendency for costs and benefits occurred which, again, can be explained by the two-factor account of Sahakyan & Delaney (2005). Comparing the recall of the two lists within the conditions yielded a significant reduction of PI in the F condition and a trend for a reduction in the R condition. In relation to experiments 1 and 2 where only one valence was used, here the change of list valence affected recall: While the mean differences between list 1 and list 2 were within similar ranges for the F and R condition in the previous two experiments, in this experiment the difference became bigger in the F condition (-.161, p=.00007) and smaller in the R condition (-.029, p=.391). Thus, assuming negative mean differences to reflect reduced PI, PI was more reduced in both conditions of experiment 3 compared to experiments 1 and 2.

This fits with the assumption that the context change in the F condition, which was initiated in all three experiments by the forget instruction, was additionally supported by

the change of valence in experiment 3 resulting in double context change while in experiments 1 and 2 single context change occurred. Similarly to the F condition, a reduction of PI occurred in the R condition compared to the experiments in which only one valence was used. Again, it seems reasonable to assume that context change was induced by the change of list valence leading to the PI reduction. As already found in the previous experiments, here again a late positive component (550-750 ms after stimulus onset) appeared during encoding of items following the forget instruction which is consistent with findings of late positive potentials in response to attentional demands (Johnson, 1988; Paller et al., 1987). The scalp distribution of this increase in positive activity resembled the topography found in experiment 1. As in both experiments neutral words were subject to directed forgetting, this suggests that the valence of the to-be-forgotten words might be crucial for the strategy during subsequent learning and thereby determining for the scalp distribution (Fabiani & Donchin, 1995; Fabiani et al., 1990;

Fernandez, Weyerts, Tendolkar et al., 1998).