5.3.3.2. Results and discussion

The trimming procedure for correct RTs was the same as in Experiments 1 and 2 (see results of Experiment 1). The RTs of one participant, with more than 30% errors, were also discarded. The RT data were subjected to a 2 (ISI: 90 ms, 0 ms) x 2 (congruency levels of Stroop) x 3 (emotional valence: negative, neutral and positive) x 3 (block of trials) ANOVA. RT data obtained from the without-face condition were excluded from this analysis.

Three main effects proved to be significant. The mean RTs were greater in the incongruent condition (631 ms) than in the congruent condition (612 ms), F (1, 57) = 22.42, p < .001. Response latencies decreased significantly across blocks, F (2, 112) = 17.14; p < .0001 (640, 612 msec, and 610 msec for blocks 1, 2, and 3, respectively). There was also a significant effect of emotional face, F (2, 112) = 6.19; p < .01 (negative: 643 ms, neutral: 612 ms and positive: 610 ms). The expected interaction between emotional face type and ISI level was significant, F (2, 112) = 4.79, p < .01 (see Table 2). There were no other significant effects.

Bias scores. We calculated facilitation and interference effects as in Experiment 1 (bias scores = mean RT in each condition – mean RT measured in the without-face condition). Scores were submitted to a 2 (ISI: 90 ms, 0 ms) x 2 (congruency levels of Stroop) x 3 (emotional valence: negative, neutral and positive) x 3 (block) ANOVA. The main effects of emotional face and block were significant, F (2, 112) = 6.19, p < .01, and F (2, 112) = 3.53, p < .05, respectively. There was a significant interaction between emotional face and ISI, F (2, 112) = 4.79, p < .01 (see Figure 11), and a significant interaction between ISI and block, F (2, 112) = 3.42, p < .05. Post hoc comparisons in the interaction between ISI and emotional valence of face revealed a significant difference only between positive and negative face in ISI level of 90 msec (p < .05).

There were no other significant effects. There was no interaction of block and face type in the ISI level of the 90-ms condition, and the rapid response latency with the negative face stimulus appeared in Block 1 and continued in Blocks 2 and 3.

As shown in Figure 11, participants did not show any particular response tendency to negative stimuli in the 0-ms ISI level condition, as compared to positive or neutral stimuli. Ongoing target processing might play a role by interrupting the processing of irrelevant stimuli, regardless of their emotional valence. This agrees with the original emotional Stroop effect, in which normal participants show no specific effects of negative irrelevant information. On the contrary, in the 90-ms ISI level condition, participants demonstrated faster colour-naming latency with negative face than with positive and neutral face, and thus seemed to allocate more attention to task-relevant processing after priming with the negative face. This strongly suggests that when target and emotional irrelevant processing occur simultaneously, controlled task demands related to processing per se can interrupt all kinds of task-irrelevant stimulus processing, regardless of its content or valence, resulting in minimal interference. However, when target processing is postponed with a time interval, normal participants are more likely to use an avoidant processing mode with negative emotional stimuli, and thus can more effectively protect target processing from interference from negative information.

Contrary to our findings in Experiment 1, there was no interaction between blocks and emotional face types in the 90-ms ISI condition. Response latencies in all three blocks were nearly similar, regardless of the valence of the emotional face. This could imply a more stable automatic attentional pattern to negative information in normal participants rather than a task-related strategy. In Experiment 1, although there was a clear interaction between blocks and emotional faces, response latencies with the negative face were always shorter than with the neutral face. However, as results from Experiment 1 suggest that processing emotional stimuli may be sensitive to participant strategies, more research will be needed on this question.