Volume 43, Issue 2

Based on the fact that verb learning is a difficult issue in language acquisition, the paper non-empirically addresses the question which inference rules which type of language learner might use in order to infer verb classes from the nominal fillers in the argument positions of verbs. Probabilistic rules based on Bayesian networks allow to decide whether a verb belongs to a class or not. The complexity of the rules might give an explanation why learning the semantics of verbs is difficult and is mastered completely at late stages of the language acquisition process.

In present-day German adnominal genitives usually appear postnominal ( der Vortrag der Präsidentin ), only genitives of proper names can be positioned before the head noun ( Annas Vortrag ). Since genitives of proper names also show different morphological and syntactic behaviour from genitives of appellatives and since they resemble in some of their characteristics possessive pronouns, it is controversial whether prenominal genitives in German should be considered an instance of a normal genitive case or if they have to be classified as possessive markers or possessive adjectives. Against this background, this paper focuses, on the basis of corpus data, on the semantics of prenominal genitives. It is shown that most of the prenominal genitives – be they anthroponyms or toponyms – in fact express a possessive or at least an agent-like role.

The gender specification of personal pronouns in German which refer to lexical hybrids like Mädchen (neuter noun; eng.: girl ) can either be determined directly by the grammatical features of these nouns ( das Mädchen […] es ) or by the sex of their referents ( das Mädchen […] sie ). This phenomenon is a typical example of the alternation of grammatical and semantic agreement, which can be observed in many variations among the world’s languages. In order to explain the distribution of these two agreement types, many authors argue that there is a positive correlation of the linear and/or structural distance between the agreeing elements (controller and target) on the one hand and the acceptability of semantic agreement on the other hand. Based on empirical data, I argue for the additional integration of semantic factors which are strongly related to structural aspects. It is shown that the choice of agreement type for German personal pronouns depends on the interpretation of these words as bound or free variables, whereby semantic agreement is less likely with bound variables.

This paper is concerned with the syntactic properties of hypothetical comparison clauses (= HCCs) in German. These are introduced either by als (‘as’) or wie (‘how’); for example, Ben fährt Rad, {als wenn er betrunken wäre / als ob er betrunken wäre / als wäre er betrunken / wie wenn er betrunken wäre} (‘Ben is cycling as if he were drunk’) . I argue that als in als-HCCs contributes a prepositional C-head that involves idiosyncratic selectional restrictions regarding the embedded conditional clause. HCCs with wie , by contrast, are treated as by and large ordinary free relative clauses; what sets them apart from other free relative clauses is that their initial wh -phrase is extracted from an empty VP, which is licensed by the explicit conditional antecedent. The proposed analyses are backed up by evidence from the following sources: (i) the behavior of HCCs with regard to diagnostics for (dis)integration/subordination as opposed to coordination, (ii) basic semantic properties of HCCs, and (iii) structural contrasts between HCCs introduced by als and HCCs introduced by wie . I spell out the proposals in terms of Sternefeld’s (2006) feature-driven lexicalist syntax model.

German proper names divide in two large classes, with and without obligatory definite articles: die Bismarck ‘the (feminine) Bismarck – a ship’ vs. Ø Rheinhausen ‘ Ø Rheinhausen (neuter) – a town’. Since proper names are inherently definite, the article cannot express definiteness. Rather, the article is part of a classifier system which provides information about the referent: names of ships, rivers and mountains always take the definite (onymic) article whereas names of towns, states and continents do not take the article. Furthermore, the inherited three-gender system in combination with the presence or absence of the article is used to create a new system of six proper name classes: a) with article: names of ships always are feminine ( die Bismarck ), names of cars are masculine ( der Corona ), and names of restaurants are neuter ( das Heiliggeist ); b) without article: names of towns are neuter ( Ø Rheinhausen ), names of males are masculine ( Ø Sascha) , and names of females are feminine ( Ø Sandra ). Thus gender also provides information about the referent. This article deals with the diachronic change from empty categories (“junk”) to a new classifier system and argues for a case of degrammaticalisation with respect to gender and for one of exaptation with respect to the article. This development towards a real classifier system is rather new. Its consolidation can be observed in present German.