ERP-Evidence on the Influence of Verbal Working Memory Capacity on Individual Differences in Processing Negative Polarity Items
Juliane Domke
Humboldt-Universität zu Berlin
Using Event-related potentials (ERPs) we investigated the influence of working memory capacity
(WMC) on individual differences in the processing of the German negative polarity item (NPI) jemals.
We compared the NPI appearance in an appropriate licensed with an ungrammatical unlicensed context as in (1):
(1) Kein Schüler hat jemals den Lehrer geärgert.
*Der Schüler hat jemals den Lehrer geärgert. literally translated as “No/*The pupil has ever the teacher annoyed.”
Structures including an NPI are complex, since their meaning contains additional components which restrict the environments of their occurrences (cf. Giannakidou, 2011). Psycholinguistically NPI processing involves syntactic and semantic processes (i.e. Saddy, 2004, Drenhaus et al., 2005), emphasizing on the structural complexity. ERP evidence on the processing of the unlicensed NPI jemals has repeatedly reported a biphasic N400-P600 processing pattern when the licensor was not appropriate (e.g.. Drenhaus et al., 2005).
As to the role of verbal WMC in language comprehension there is the assumption that (syntactic) complexity is one major factor influencing the way in which a structure is processed (i.e. Just &
Carpenter, 1992). Low WMC readers are less accurate and their processing is slowed down when complexity is high.
According to this, we were interested whether to see an influence of verbal WMC when the
German NPI jemals is not licensed. That is, the comparison of an ill-formed complex sentence structure (unlicensed NPI) with the analogous well-formed structure was correlated with the individual verbal WMC. According to the assumptions made by i.e. the verbal WMC model of Just & Carpenter
(1992) it would be expected for low WMC readers to show a decreased and delayed processing pattern compared to high WMC readers.
Using the violation paradigm (see example (1)) WMC and EEG of 26 native German speakers were recorded. Participants were grouped according to their WMC test results (Lewandowsky et al.,
2010) in HIGH WMC and LOW WMC. Statistical analyses on the ERP data were performed using the linear-mixed-effects-model method (i.e. Baayen et al., 2008). According to visual inspection (see figure 1 below) six time windows post-stimulus jemals (in ms) were analysed (early/late “Negativity”
= 300-450, 500-650, early/late “Positivity” = 600-900, 800-1000, early/late “Late-Negativity” 1200-
1400, 1400-1600).
Figure 1: ERP waves of the two WMC groups
Figure 1 above illustrates the averages of the ERP waves of the two WMC groups. As to visual impression both groups show an early negative going wave between 300 and approx. 600 ms followed by a late positive going wave (600-1000 ms). Both effects seem to be delayed in the LOW WMC group.
Also, the early negativity appears less clear or smaller in the LOW WMC group. There further appears a large late negativity in the LOW WMC group. Statistical analysis reveals the biphasic NPI processing pattern, confirming the more negative going wave in the unlicensed NPI condition in the “Negativity” time-windows and the more positive going ERP wave in the “Positivity” time windows in both WMC groups. Significant interactions with WMC and time windows indicate clear ERP wave differences between the two groups. Analysis reveals significant differences for HIGH WMC in the early and for
LOW WMC in the late time windows, respectively.
Both time windows calculated for “Negativity” yield strong interactions of ungrammaticality by group (WMC) supporting the visual impression that the effects are of different strength for each WMC group.
L OW WMC additionally reveals a large “Late
Negativity” which is statistically validated for both time windows. This “Late Negativity” does not yield significance in HIGH WMC in either time window.
Results suggest that the processing of the unlicensed German NPI jemals is affected by WMC in that LOW WMC processing is delayed and effects are smaller, confirming prior expectations.
Interestingly LOW WMC also shows a large late negativity which leaves open questions: There are at least two possible but vague explanations: On the one hand, this late negativity might be indicating a higher processing load because prediction is not met, that is: the upcoming NPI was not expected due to prior context. Processing problems due to prediction mismatch is reflected by a late negativity which is enlarged when WMC is low (see i.e. Otten & van Berkum, 2009). This, however, is suspicious, since a subject NP including a negative quantifier in initial sentence position would not per se expect an NPI to come up as the sentence proceeds. On the other hand, this late negativity could be a reflection of higher processing load of controlled “post-interpretative” processes (Caplan & Waters,
1999). This would assume that the failure of integrating an unlicensed NPI into the sentence context further triggers controlled post-interpretative processes, which are harder to deal with when WMC is low. However, following the Caplan & Water’s model of verbal WMC (Caplan & Waters, 1999) would not expect any differences in prior ERP correlates between the two WMC groups, since they should reflect “interpretative processes” not being affected by WMC differences.
In sum, our study reveals verbal WMC influence on individual processing differences when the
German NPI jemals is not appropriately licensed. When WMC is low typical ERP correlates (that appear in processing of an unlicensed NPI) are delayed and appear to be less strong. They are followed by a large late negative component which supposedly indicates that additional late controlled processes to take place when the NPI is not licensed and these processes are influenced by verbal
WMC.
References:
Baayen, R.H., D.J. Davidson & D.M. Bates, (2008) Mixed-effects modeling with crossed random effects for subjects and items. Journal of Memory and Language 59, 390-412.
Caplan, D.& G.S. Waters, (1999). Verbal memory and sentence comprehension. Behavioral and Brain
Sciences, 22, 77-94.
Drenhaus, H., D. Saddy & S. Frisch, (2005) Processing negative polarity items: When negation comes through the backdoor. In: Kepser, S. & M. Reis, (eds.), Linguistic Evidence -- Empirical,
Theoretical, and Computational Perspectives. Berlin: Mouton de Gruyter, 145-165.
Giannakidou, A., (2011). Positive polarity items and negative polarity items: variation, licensing, and compositionality. In: C. Maienborn, K. von Heusinger & P. Portner (eds.) Semantics: An
International Handbook of Natural Language Meaning . Berlin: Mouton de Gruyter.
Just, M.A., & P.A. Carpenter, (1992). A capacity theory of comprehension: Individual differences in working memory. Psychological Review , 98,122-149.
Lewandowsky S., K. Oberauer, L.-X. Yang & U.K.H. Ecker, (2010) A working memory test battery for MatLab. Behavior Research Methods, 42 , 571-585.
Otten M. & J. van Berkum, (2009), Does working memory capacity affect the ability to predict upcoming words in discourse? Brain Research , 1291, 92-101.
Saddy D., Drenhaus, H. & S. Frisch, (2004), Processing polarity items: Contrastive licensing costs.
Brain and Language , 90 , 495-502.