Markus Bader

Subject-Object Ambiguities in German Embedded Clauses: An Across-the-Board Comparison

This paper examines the processing of embedded clauses in German which are ambiguous between a subject-before-object and an object-before-subject order. In an experiment using a speeded grammaticality judgment task, four types of locally ambiguous clauses were compared: (i) sentences involving movement of a definite noun phrase (NP), (ii) sentences involving pronoun movement, (iii) relative clauses, and (iv) embedded questions. We found that readers were consistently garden-pathed in the object-before-subject condition, regardless of sentence type. Furthermore, there were considerable differences with respect to garden-path strength. The garden-path effect was strongest for sentences involving scrambling. In addition, sentences involving pronoun movement induced more processing difficulty than embedded questions and relative clauses. We argue that our findings can be best explained within a serial processing model that acknowledges both syntactic and nonsyntactic influences on reanalysis and that can account for graded effects of garden-path strength.

Prosodic Influences on Reading Syntactically Ambiguous Sentences

In this chapter, I will propose a new account for the question of why only some local syntactic ambiguities lead to strong garden-path effects whereas others do not. This account is based on the observation that readers do not only compute syntactic structures during reading but also prosodic structures. These prosodic structures are a product of the process of phonological coding which normally accompanies silent reading. Due to the lack of a one-to-one mapping between syntactic and prosodic structures, recovery from a syntactic misanalysis may be accompanied by the need to replace the original prosodic structure or not. According to the prosodic constraint on reanalysis, the need to revise the original prosodic structure makes recovery from a syntactic misanalysis difficult. Empirical evidence for this claim comes from a series of three self-paced reading experiments which investigated the German variant of the English her-ambiguity. These experiments manipulated the prosodic structure of sentences by inserting focus operators (focus particles, sentence adverbials) into locally ambiguous sentences. Thereby it became possible to independently determine the contributions of syntactic and prosodic factors to reanalysis. The results show that for the ambiguitiy examined, garden-path effects can be predicted on prosodic grounds but not on syntactic grounds. This finding supports the claim that for certain kinds of syntactic ambiguities reanalysis is prosodically constrained.

Morphological underspecification meets oblique case: syntactic and processing effects in German

In German, oblique Cases (dative and genitive) require morphological licensing while structural Cases (nominative and accusative) do not. This difference can be captured by assuming that in German, NPs bearing oblique Case have an extra structural layer Kase phrase (KP) which is missing in NPs bearing structural Case. Focusing on dative NPs, we will show that the postulation of such a phrase-structural difference between oblique and structural case allows for a unified explanation of a wide array of facts both from the domain of grammar and from the domain of language comprehension. First, with regard to grammar, several asymmetries between dative NPs and nominative/accusative NPs follow if the former but not the latter are included within a KP-shell, including asymmetries with respect to function changing operations, clausal licensing, binding and topic drop, among others. Corroborating evidence for our analysis of dative Case in German will be provided by a comparison with data from English and Dutch. Second, when combined with certain independent assumptions about the human sentence parsing mechanism, the postulation of a KP for datives helps explain several recent experimental findings with respect to on-line sentence understanding, including the facts that dative case is dispreferred in situations of local syntactic ambiguity and that dative case may erroneously override structural case during sentence comprehension but not vice versa.

Event-Related Brain Potentials and Case Information in Syntactic Ambiguities

In an ERP study, German sentences were investigated that contain a case-ambiguous NP that may be assigned accusative or dative case. Sentences were disambiguated by the verb in final position of the sentence. As our data show, sentences ending in a verb that assigns dative case to the ambiguous NP elicit a clear garden-path effect. The garden-path effect was indicated by a broad centro-posterior negative shift that occurred between 300 and 900 msec after the dative-assigning verb was presented. No enhanced P600 following the misanalysis was observed. Noun phrases whose case ambiguity was resolved in favor of accusative case and unambiguously dative-marked NPs did not trigger significant ERP differences. We will discuss the implications of our results for parsing and its neuropsychological correlates. The results of this study support a parser design according to which the so-called structural case (nominative or accusative) is assigned without any delay in the absence of morpho-lexical counterevidence. It is argued that the enhancement of a negative ERP component with a ficlassicalfl N400 topography refiects the difficulty of reanalysis due to reaccessing morpho-lexical information that lies outside the domain of the parsing module. Consequently, ERP responses to garden-path effects are not confined to a late positivity but vary depending on the level of processing involved in reanalysis. The fact that garden-path effects may also elicit an N400 can be linked to the nonhomogeneous linguistic properties of the constructions from which they arise.

Ungrammaticality detection and garden path strength: Evidence for serial parsing

It is still a controversial issue whether the human sentence processor computes syntactic representations serially or in parallel. We address this question by comparing the processing of different types of ambiguous garden path sentences with the processing of corresponding ungrammatical sentences. The ungrammatical sentences lead to the same type of syntactic mismatch that arises in the garden path sentences at the point of disambiguation, but they cannot be reanalysed. Using the speeded-grammaticality judgement task it is shown that factors which regulate garden path strength in ambiguous sentences also have an influence on the processing of corresponding ungrammatical sentences in that they determine how reliably the ungrammaticality is detected. We argue that this processing correlation provides evidence for serial parsing models because only serial models lead us to expect that the parser attempts reanalysis in ambiguous and ungrammatical sentences alike.

Mode of Disambiguation and Garden-Path Strength: An Investigation of Subject-Object Ambiguities in German.

The results of three experiments are reported which investigated the processing of locally ambiguous object-subject sentences in German. These sentences are known to elicit garden-path effects because the parser initially prefers the assignment of a subject-object structure (e.g., Schriefers, Friederici, Kühn, 1995). The aim of the experiments was to test whether the type of grammatical information that signals the garden-path (the mode of disambiguation)has an impact on how difficult it is to arrive at the correct structural assignment. We exploited the fact that subject-object ambiguities in German can be disambiguated in two different ways: by agreement or by case. If disambiguation concerning the relative order of subject and object is provided by the number features of the finite verb (agreement disambiguation) a robust garden-path effect results. In contrast, if the disambiguating information is provided by a second NP morphologically marked for nominative, the resulting garden-path effect is weak. This finding poses difficulty for models of reanalysis which relate garden-path strength to revision cost because the revision operations necessary to transform the subject-object structure initially computed into an object-subject structure are the same for both modes of disambiguation.Our results show that different modes of disambiguation can be more or less effective in signaling how to come out of the garden-path, a conclusionin accordance with the diagnosis model of reanalysis as first proposed in Fodor and Inoue (1994).

Case and Reanalysis

In this paper we discuss an asymmetry in the Case system of German and its implications for human sentence processing: the asymmetry between nominative/accusative and dative case. Starting from the assumption that dative case has a distinct grammatical representation--dative DPs are embedded into an extra structural layer KP--the results of two experiments will be presented, which show that dative assignment during reanalysis is accompanied by additional processing operations that are not needed when accusative or nominative are assigned. In particular, we show that dative assignment during reanalysis triggers reaccess to the mental lexicon, giving rise to greater processing difficulty. We conclude with a discussion of empirical and theoretical consequences of our findings.

On Reanalyis: Evidence from German

This paper will explore the question of how the phenomenon of variable garden-path strength might be modeled within a serial model of the human parsing mechanism. The basic property of a serial parser (cf. Frazier, 1979; Frazier & Fodor, 1979) is that it always computes a single, fully specified syntactic representation. This representation, which is called “current partial phrase marker” (CPPM), is continuously updated as each word is read. Due to the existence of local syntactic ambiguities, a serial parser will from time to time compute syntactic structures that are contradicted by following material. In such a situation, the original CPPM has to be revised in order to make it compatible with the current input. This paper will explore the question of what determines whether recovery from a syntactic misanalysis is easy or difficult: why do some misanalyses lead the parser down the garden-path while others are easy to recover from?

Entscheidungsprozesse bei der Verarbeitung lokal ambiger Sätze

Kognitiv adaquate Modelle des menschlichen Sprachverstehens mussen spezifizieren, wie Satzen syntaktische Strukturen zugewiesen werden und in welcher Relation die Berechnung syntaktischer Strukturen (Parsing) zu anderen Komponenten sprachlichen Wissens (lexikalisches, semantisches Wissen) steht. Zur Beantwortung dieser Fragen untersucht man gewohnlich, wie das Sprachverstehenssystem lokale syntaktische Ambiguitaten auflost, d.h. Satze verarbeitet, die temporar mit mehr als einer syntaktischen Struktur kompatibel sind. Experimentelle Untersuchungen lokal ambiger Satze haben zu zwei zentralen Einsichten gefuhrt (Mitchell, 1994): (i) Im allgemeinen wird eine der moglichen Strukturzuweisungen praferiert. (ii) Wird ein lokal ambiger Satz zugunsten der nicht-praferierten Struktur disambiguiert, kommt es zu Verarbeitungsschwierigkeiten, sog. Garden-Path-Effekten, die hinsichtlich ihres Schweregrades erheblich variieren. Aus diesen Befunden ergeben sich zwei zentrale Fragestellungen: Wie verhalt sich das Sprachverstehenssystem, speziell das fur die syntaktische Analyse zustandige Teilsystem (im folgenden als Parser bezeichnet), wenn eine syntaktische Ambiguitat entsteht, und wie verhalt es sich, wenn sich die praferierte Strukturzuweisung als falsch herausstellt? Letztere Fragestellung steht im Mittelpunkt unseres Aufsatzes. Anhand experimenteller Daten wollen wir fur eine Modifikation serieller Modelle syntaktischer Verarbeitung mit Blick auf die Verhaltensweise des Parsers am Punkt der Disambiguierung argumentieren.