Experiment 4: Memory task

Participants

Sixteen participants (8 male, age range: 18-27 years, mean: 22.12) were recruited from the University of Konstanz by advertising. All subjects were native speakers of Standard German.

Materials and Procedure

The materials, the random lists, and the rotation of the runs were the same in this experiment as they were in Experiment 3. This decision was determined by the objective to study the influence of the experimental task on the error-sensitivity. To be able to assess this factor, we had to keep the rest of the factors intact. The experimental procedure of this study differed from the previous one in the following points: (i) the stimuli were presented auditorily in blocks of 3-8 words; (ii) after each block a word was presented on a computer screen; (iii) the subjects were instructed to press a corresponding key if they heard this visually presented probe in the last presented auditory block (Yes/No).

Data Analysis

The data were processed in the same way as in the lexical decision task experiment. The filtering and averaging procedures resulted in the rejection of approximately 6% of the data.

Figure 9. German Participants: Memory Task

Grand averages (A) and topographies (B) of the difference waveforms are shown for all experimental conditions in the memory task. The LAN effect is highlighted at the F5 electrode. The violation conditions demonstrate the following pattern: the RD (green solid line) condition is the most negative curve; the UD (red dashed line) and the NC (purple dashed line) conditions are more positive than the RD condition. The W (black solid line) condition is the most positive curve. The topographies show a similar scalp distribution for all conditions with a slight left hemispheric shift for the NC and UD conditions.

3.1.2.2.2 Results

Figure 9 displays the grand averages and the topographies for the difference waveforms of the type Violation condition–W in the time window of 400-600 msec. The LAN effect is highlighted at the F5 electrode site. The RD condition (a green solid line) has the most negative value, while the UD (a red

dashed line) and NC (a purple dashed line) conditions pattern together and have more positive values than the RD condition.

To make sure that the change of the experimental task resulted in differential brain activity, we first ran an omnibus test within the time window of 400-600 msec with Task as a between factor, which revealed a four-way interaction of Task, Anteriority, Laterality, and Stem type (F (5.55, 166.56) = 10.84, p < 0.001). Further one-way ANOVAs with the factor Stem type were run within each of nine regions (Anteriority x Laterality). The one-way ANOVA reached significance at the left anterior electrode sites: F (2.08,41.7) = 2.84, p <

0.05, which is where the LAN component had been observed (Friederici, 2002, 2011). Analyses on mean amplitudes were run at the left anterior electrode sites (F5, AF3, AF7) within the time window of 400 – 600 msec. The mean amplitude values for the LAN component are illustrated in Figure 8. The contrast analysis revealed a significant difference of the RD condition from the rest of the conditions: t (60) = -2.74, p < 0.01.

Figure 10. German Participants: the LAN Effect

The LAN effect at the F5 electrode is shown for the memory task. Note that the (R)elated (D)erived condition has the greatest negative value.

3.1.2.2.3 Discussion

We predicted that the change of the experimental task would trigger a different processing strategy as compared to the lexical decision task experiment.

We further argued that the RD items would elicit a formal error-detection mechanism, indexed by a morphosyntactic ERP component, if our hypothesis were valid. The results of the study validated our hypothesis yielding a spatial-temporal activation pattern of the brain responses that was different from that of Experiment 3. The results of the memory task experiment revealed a LAN effect in the time window of 400-600 msec evoked only by the RD items. The

irreparable violation conditions failed to trigger any error-detection mechanisms and had no significant difference from each other and from the W condition. No other ERP components were observed in this experiment.

In the predictions for the memory task experiment we argued that the presence of the error-detection mechanisms or the lack thereof for the irreparable nonwords might be explained by different processing strategies. Thus, if the meaning of the items was relevant to their maintenance in the short term memory, a semantic ERP component would have been observed. If, however, the new structure, introduced by the irreparable nonwords, was accommodated into the short term memory as a novel word, the processing of these items should not have differed significantly from that of the existing words. The fact that irreparable items failed to induce any violation effects provided evidence in favor of the latter processing strategy. The shift of the focus from the lexical semantics to the whole structure processing led to the reevaluation of processing priorities.

Considering the effort required for the simultaneous maintenance of three to eight words in the short term memory, a new processing hierarchy had to be established in order to increase the task fluency. The lexical retrieval and semantic judgment were no longer a prerequisite for the successful task performance; therefore, they were relinquished in favor of the formal structure analysis. Without the possibility of finding a stem representation in the mental lexicon, the irreparable items could not be classified as structurally deviant.

Therefore, both the novelty of the item and the novelty of the structure contributed to the accommodation of the irreparable nonwords into the short term memory.

The maintenance of the RD items in the short term memory presented a greater challenge than the rest of the conditions did. Considering the two-step derivation, required for the deverbal noun formation, the parser had to trace back all the steps in search for the missing manipulation. This particular manipulation also had to be memorized in order to produce the correct response during the memory task. Therefore, the omission of the conversion of the adjectival stem into the verbal, as reflected in the umlaut of the stem vowel, triggered a morphosyntactic ERP component LAN (A. D. Friederici, E. Pfeifer, & A. Hahne, 1993; Gunter et al., 1997; Rösler et al., 1993). This finding delivered a solid body of evidence supporting our hypothesis. The morphosyntactic combinatorial possibilities must be listed in the mental lexicon entry, as reflected in the amplitude of the LAN effect. In the case of the invalidity of our hypothesis, the processing of the RD items should have been similar to that induced by the rest

of conditions, i.e. without an equivalent structure in the mental lexicon the new item would have been accommodated as a novel word.