Douglas Saddy, Heiner Drenhaus & Stefan Frisch
University of Potsdam
In the linguistic literature, most accounts agree that polarity items must occur in a licensing domain in which the licensing conditions are accessible, where accessibility is determined by the hierarchical constituency rather than linear accessibility. As a general consequence, spurious potential licensers that are linearly near the polarity item play no role in the construction, see (1–3).
(1) NO applicant who a member praised was EVER admitted to the club.
(2) *An applicant who NO member praised was EVER admitted to the club.
(3) An applicant who NO member praised was OFTEN admitted to the club.
From the processing point of view this raises interesting questions regarding the contributions of syntactic and semantic constituency and their influence on potential intrusion effects during processing. Polarity items are not predicted by the presence of a potential licensing element, they can be easily replaced with a non-polarity sensitive adverbial (3). Similarly, the word category of an unlicensed polarity item is well formed; the failure in (2) is one of licensing not that an adverbial cannot occur at that point in the string. Thus at the point of detection of a polarity item the syntactic parse is always well formed but the semantic properties of the construction must be checked.
In a series of ERP and speeded judgment studies, we investigated the processing of both negative (NPCs) and positive (PPCs) polarity constructions in German focusing on the effect of the structural and linear position of negation. In the construction of materials we limited ourselves to simple licensing by overt negation in the case of NPCs involving 'jemals' (English 'ever') and the absence of negation for PPCs involving 'durchaus' (English 'certainly').
| (4) |
Kein / Ein Mann, der einen / keinen Bart hatte, war jemals / durchaus froh. No / a man, that a / no beard had, was ever / certainly happy. |
The ERP data revealed contrastive behaviour that was sensitive to both the type of polarity item and the presence of intrusion. Violations of NPCs produced an N400 component whereas violations of PPCs produced an N400–P600 pattern. In addition, intrusion effects were found in both ERP and judgments when negation occurred in the relative clause. In the case of NPCs, which in our study were ungrammatical when negation was trapped within the relative clause, the N400 was weakened. In the case PPCs, which in our study were grammatical when negation is trapped within the relative clause, a weaker P600 was elicited.
An account of the checking relation induced by the polarity items must explain the following:
(a) The basic grammaticality effects for both NPCs and PPCs
(b) The contrast in the grammaticality effects for both NPCs versus PPCs
(c) That intrusion occurs in both grammatical and ungrammatical strings
(d) That intrusion affects different processes in both types of polarity items
We interpret these findings as support for a mechanism that involves the coupling of the computation of hierarchical constituency with an independent working memory or reactivation process.