This chapter has introduced some basic questions and concepts related to the phenomenon of argument linking and has discussed the role of affectedness as a central semantic component of the syntax/semantics interface, at which the mapping from verbal semantics to syntactic structure takes place, in more detail. A review of the traditional approach to argument linking has shown that thematic roles and thematic hierarchies suffer from a number of drawbacks which alternative accounts involving predicate decompositions and Proto-roles (e.g. Dowty, 1991) aim to avoid. A promising perspective on affectedness as a graded phenomenon is provided by Beavers (2010, 2011); importantly, his scalar approach to affectedness does not only offer a solid basis for defining degrees of affectedness, but also opens up direct links to theories of argument linking making use of Proto-roles as well as to work on semantic determinants of transitivity. It further allows to refine the nature of the relation between affectedness and aspectual notions like telicity and suggests that affectedness is in fact a linguistic primitive which is not reducible to lexical aspect. The following chapter builds upon the concepts and approaches discussed above and presents a first experimental investigation of the role of affectedness in the mapping from predicate meaning to syntactic structure.
Chapter 2
Probing affectedness: a detour
via argument linking in deverbal
nominalisations
The current chapter presents a first experimental exploration of possible effects of affected- ness with processes of argument linking. Its main aims are to introduce, justify and validate the basic experimental paradigm applied in all subsequent experiments, which focusses on argument linking processes in the nominal, rather than the verbal domain, and makes use of deverbal nominalisations as predicates, as already pointed out in the Introduction. Section 2.1 summarises relevant general issues related to parallels and differences in the argument linking processes in the verbal and nominal domain and section 2.2 focusses on crucial argu- ment linking phenomena arising with German -ung -nominalisations of syntactically transitive verbs and postnominal genitives. Section 2.3 presents the semantic account of these pat- terns developed by Ehrich and Rapp (2000), in which the lexical semantic notion of change of state/affectedness plays a prominent role; section 2.4 reinterprets Ehrich and Rapp’s account in terms of semantic prominence of the direct object argument as defined by the Affectedness Hierarchy of Beavers (2010, 2011). After a discussion of the syntactic status of genitive ar- guments of German -ung -nominalisations in section 2.5, section 2.6 outlines the rationale for taking a detour via deverbal nominalisations for the investigation of verbal argument link- ing processes. Section 2.7 presents and discusses an acceptability judgement experiment intended to validate the experimental paradigm and empirically test the predictions derived from the reinterpretation of Ehrich and Rapp’s semantics-based theory based upon Beavers’ Affectedness Hierarchy.
2.1 Argument linking: from verbs to nouns
Among the major building blocks of language are the major parts of speech, i.e. verbs, nouns, adjectives and prepositions (V, N, A and P) – a definition of the properties which distin- guish these major lexical categories from one another on the one hand, and those features which some of them share in common on the other hand is central to a theory of grammar. Chomsky (1970) uses the two features [±N] (‘nominal’) and [±V] (‘verbal’) to classify these categories as [+V,-N] for verbs, [-V,+N] for nouns, [+V,+N] for adjectives and [-V,-N] for prepositions1. Wunderlich (1996) points out that this feature decomposition lacks a con-
ceptual basis and does not characterise some of the most relevant relations between the categories correctly. His alternative classification system makes use of the two features [±dep] for ‘referentially dependent’ and [±art] for ‘articulated’; nouns and verbs share the feature specification [-dep], since items from both classes can be anchored in extra-linguistic contexts via referential arguments, making them referentially independent. Adjectives as well as prepositions are referentially dependent ([+dep]) and need support by the linguistic con- text. In Wunderlich’s system, nouns and verbs are distinguished by the specification for the [art] feature: verbs and prepositions are defined as [+art] because they are specified for (multiple) obligatory arguments, while nouns and adjectives lack such obligatory arguments and are thus specified as [-art].
On this view, the main difference between verbs and nouns lies in the distinction between rich argument structure underlying verbs and impoverished argument structure associated with nouns. Stiebels (2006, pp. 170–171) illustrates the difference in argument structure possibilities between the two word classes in the predicate-argument structure notation of Lexical Decomposition Grammar (e.g. Wunderlich, 1997): the decompositions in (66) ex- emplify different verbs with increasingly complex argument structure, where the argument
s represents the event/situational variable, i.e. the verb’s referential argument. The argu-
ment structure is derived from the lexical decompositions by means of λ-abstraction2of the
variables present in the decompositional structures, placing the lowest-ranked (i.e. most embedded) argument at the leftmost position and the highest argument (the referential ar- gument) at the rightmost. Note that the order of the λ-abstractors matters for argument realisation, as discussed in the following section. As the examples in (66) show, verbs may show any number of structural arguments from none (66a) to three (66d), in some languages even up to five (Stiebels, 2006, p. 170).
(66) a. rain: λs RAIN(s)
1‘+’ and ‘-’ are to be understood as indicating whether the respective feature holds or does not hold for a given
word class, respectively.
2The ‘Lambda calculus’ is a formal and universal system for the representation of functions introduced by Church
(1936) (see Alama, 2015). It is widely used in a number of fields and can be applied to represent the meaning of expressions of artificial and natural languages in an abstract and compact way. As a first example, take the simple mathematical expression x − 2, where x is a variable; in λ-notation, this can be represented as λx [x − 2], where the
λ binds the variable x. To assign a value to x , we can writeλx [x − 2] (5), which indicates that the value ‘5’ is to be
substituted for any occurrences of x in the expression x − 2; this yields 5 − 2 = 3. Abbott (2003, p. 79) gives an example for its use in linguistic semantics, noting that the “function denoted by λx [Mary loves x] maps individuals whom Mary loves to the value True, and others to the value False”.
b. cry: λx λs CRY (x) (s)
c. read: λy λx λs READ(x, y) (s)
d. give: λz λy λx λs [ACT (x) & BECOME( P OSS(y, z))] (s)
On the other hand, a prototypical noun – i.e. a non-relational noun denoting some (more or less) concrete entity – such as cat denotes a one-place predicate as illustrated in (67a). Some nouns such as kinship terms (67b) or body part expressions (67c) are inherently rela- tional and thus denote two-place predicates (see Stiebels, 2006, p. 171).
(67) a. cat: λu CAT (u)
b. mother: λv λu MOT HER(u, v)
c. nose: λv λu [NOSE(u) & P ART _OF (u, v)]
Stiebels uses the metaphor of ‘being in rags’ in referring to this impoverished argument struc- ture associated with prototypical nouns – however, she stresses and illustrates that one can find nominal parallels to operations on verbal arguments structure (e.g. voice/diathesis op- erations such as causative or applicative) and that in some instances nouns can actually ‘be in riches’ in exhibiting more complex argument structure, just as verbs may. Investigating nouns with more complex argument structure involves moving away from prototypical nouns; as Wunderlich (1996, p. 18) remarks, “[n]ouns exhibit the greatest sortal variation in that they may refer to any spatial or temporal or even more abstract entity” as exemplified by basic and underived nouns such as war, event, game or thunderstorm.
The full complexity of verbal argument/event structure, however, only comes to bear in nouns which are derived from verbs, as in deverbal nominalisations like destruction, assassination or examination, which are morphologically derived from the respective base verbs destroy, assassinate and examine (see Alexiadou, 2010, for an introduction to issues related to nom- inalisations). Grimshaw (1990) argues that such deverbal nominalisations may inherit the full argument and event structure of the corresponding base verbs, as in (68).
(68) a. The instructor’s examination of the papers took a long time. b. The enemy’s destruction of the city was awful to watch.
Grimshaw further shows that deverbal nominalisations can be systematically ambiguous be- tween what she calls ‘complex event nominals’ (CENs) on the one hand and ‘simple event nominals’ (SENs) and ‘result nominals’ (RNs) on the other hand – while the former are asso- ciated with true verbal event and argument structure, the latter actually lack these, though they may still be interpreted as events or actions. When derived nominals are interpreted as SENs/RNs, these are internally structured and behave like basic, underived nouns without verbal event/argument structure, as in (69).
The presence of the base verb’s aspectual structure can be witnessed by the compatibility of CENs with the adjectival equivalents of event-modifying adverbs such as frequently or constant : in (70b), frequent/constant can modify the aspectual structure inherited from the base verb examine and the temporal PPs for an hour and in an hour are both compatible with the construction in (70b), due to the possible interpretation of examine patients as either telic or atelic (see Grimshaw, 2011, p. 1300).
(70) a. They frequently/constantly examine patients in/for an afternoon.
b. The frequent/constant examination of patients in/for an afternoon leads to better diagnoses.
On Grimshaw’s view, the presence of aspectual structure in CENs can also go hand in hand with the inheritance of the verb’s argument structure and thus with the obligatory realisation of arguments: since the base verb examine takes an obligatory direct object (cf. (71a)), this must also be present with the CEN examination, explaining the unacceptability of (71b) (from Grimshaw, 2011, p. 1301).
(71) a. * They frequently/constantly examine in/for an afternoon.
b. * The frequent/constant examination in/for an afternoon leads to more accurate diagnoses.
Thus, deverbal event nominalisations are a kind of hybrid category, incorporating purely nom- inal features (e.g. they can function as heads of NP arguments of verbs and are modified by adjectives) as well as verbal properties by inheriting parts of the lexical verbal event and argument structure, with different types of deverbal nominal constructions exhibiting varying degrees of ‘verbiness’ (see Ehrich, 2002, for discussion). When it comes to linking the argu- ments associated with the verbal lexical structure within an NP headed by a deverbal event nominalisation, grammar is facing a challenge though: since the nominal linking resources are based upon the ‘rags’ associated with underived nouns, accommodating the verbal ‘riches’ is not always a trivial enterprise. Since basic nouns are at most associated with two arguments (see (67) above), the nominal linking inventory is often under-equipped to deal with complex verbal argument structure, as discussed by Stiebels (2006, p. 172):
“Since the referential argument does not play a role for argument linking in non-predicative uses of nouns, structural linking is generally rather limited, often restricted to one case/agreement linker, which, in addition, is barred from being doubled in many languages. These restrictions on structural linking also have an impact on the realization of inherited arguments in nominalizations. Often, only one argument may receive a structural nominal linker.”
In this situation, the grammar must sometimes provide ways to compensate for this lack of structural linkers in nouns and sometimes the different arguments of a verb compete for one and the same argument position. In the following sections, I will outline how such a competition in German deverbal event nominalisations can be used to investigate aspects of the role of affectedness on argument linking with nominalisations.