The term modification is commonly used in a rather loose and intuitive way, unlike other linguistic notions such as argument, head, function, complement, because it is difficult to technically define it in positive terms (cf. Morzycki t.a., McNally t.a.). McNally’s definition proposal for modifier is “an expression which combines with another expression to produce a result with the same semantic value”. Modifiers can be of different categories and they can apply to all kinds of categories. Modification does not have a clear correlate with common syntactic operations assumed in the generative literature and not all modifiers can be treated as adjuncts. Under Frege’s conjecture that all composition is saturating, i.e. Functional Application (FA), there is no clear way to integrate modifiers. There are three different types of proposals in the literature to deal with this problem. One is to add compositional rules to FA, such as Predicate Modification (Heim & Kratzer 1998) or Restrict (Chung & Ladusaw 2004). The second and third type of approach keep the rule system simple (only FA) and either complicate the phrase structure in which modifiers are merged (e.g. Rubin 1994) or complicate the system of types; for example, Kamp & Partee (1995) propose more complex types for adjectives, whereas Siegel (1976) argues for type doublets for adjectives and thus for flexible types. In this course, we start out with a general overview of the problems that modification poses for semantic theory. We will then focus on the semantics of adjectives (adnominal modifiers) and adverbs (adverbial modifiers) and on the related notions of DEGREE and MANNER, addressing commonalities between and applications of them to other domains.
Adjectives (As) can be classified with respect to different criteria. One distinction is between intersective, subsective, and non-intersective As (cf. Kamp & Partee 1995). For intersective As a simple analysis in terms of set intersection suffices (e.g. if x is a carnivorous animal, x is carnivorous and x is an animal). Subsective adjectival modification yields a subset provided by the modified noun (e.g. a skillful violinist is skillful as a violinist, but not necessarily skillful per se). A class of non-intersective As are the privative ones, for which set intersection yields the empty set (e.g. a fake gun is not a gun). We will address various
approaches that aim at reducing non-intersective analyses of different classes of As to intersective (or subsective) ones, e.g. by taking into account events (Lasersohn 1998), kinds (McNally & Boleda 2004, Gehrke & McNally t.a.), or coercion (Partee 2010).
Another group of As that at first sight appear to be subsective, are gradable As, such as tall (e.g. a tall 10-year-old need not be a tall individual). However, the challenges for an intersective analysis in this case are rather due to vagueness (cf. Kennedy 2007, Sassoon 2013, for recent discussion); they disappear if we take into account a standard of comparison with respect to the comparison class involved (e.g. 10-year-olds) (Klein 1980). One prominent approach to gradable As holds that their standard is a degree (an interval) on a scale associated with these As, which thus denote relations between individuals and degrees (Seuren 1973, Cresswell 1976, von Stechow 1984, Heim 1985, Bierwisch 1989, Kennedy 1999, i.a.). Within the gradable As it is common to differentiate between relative (e.g. tall, short) and (partial and total) absolute ones (e.g. full, empty) (Kennedy & McNally 2005, Rotstein & Winter 2004). Kennedy & McNally, for instance, argue that relative As give rise to vagueness effects because they are associated with an open scale so that their standard is some midpoint on the scale. In contrast, absolute As are argued not to give rise to vagueness, as they are associated with closed scales, and the standard is always at the end of the scale. However, context has been shown to play a role in determining the standard also for absolute As (e.g. McNally 2011, Sassoon & Toledo (2011). Sassoon & Toledo argue that it is not properties of the underlying scale that play a role in determining the standard for gradable As, rather it is the nature of the comparison class involved (following van Rooij 2011): Relative As involve comparison between (extensions of) individuals (e.g. someone is tall in comparison to some other entity), whereas absolute As typically involve comparison within (and thus between intensions of) individuals (e.g. a glass is full in comparison to different stages that that particular glass could be in).
A general question is whether degrees should be part of our ontology, and if so whether they are primitives (as in some of the proposals above) or decomposable into or derived from other notions. For example, Moltmann (2004 and subsequent work) argues that degrees can be constructed out of tropes, which are particular instantiations of properties in space and time. Anderson & Morzycki (t.a.) in turn take degrees to be state kinds. Klein (1980), on the other hand, advocates a theory of gradability without degrees, taking degrees to be equivalence classes of entities (see Doetjes et al. 2011, Burnett 2012 for recent proposals). Another question is whether we need degrees only in the adjectival domain. For example, Matushansky (2001) and Morzycki (2009) argue that degrees also play a role in the nominal domain (e.g. in expressions such as a big idiot), and Castroviejo & Schwager (2008) propose a uniform treatment of degrees and kinds as correlates of properties in their analysis of concealed exclamations. Constantinescu (2011), on the other hand, investigates various cases for which it has been argued that degrees play a role in the nominal domain and concludes that most of the effects attributed to degrees can be accounted for without resorting to degrees (but instead to e.g. kinds or prototypicality). Finally, in the verbal domain, event structure is sometimes derived from the proposed scale structure underlying the event (Hay et al. 1999, Kennedy & Levin 2008, Caudal & Nicolas 2005, Winter 2006, Piñón 2008).
Typical modifiers in the verbal domain, in turn, are adverbials that express or specify the manner of the event under description, which is why manner modifiers are standardly analyzed as predicates of events, at least in (Neo-)Davidsonian frameworks (e.g. Parsons 1990). In this course, we focus on adverbs. Based on the relative position that adverbs appear in in a sentence, we can establish various subclasses (Jackendoff 1972). For example, Ernst (2002) divides adverbs with respect to the clausal reading they have, into those that are subject-oriented: agent-oriented (e.g. cleverly), mental-attitude (e.g. reluctantly); speaker-oriented: speech-act (e.g. frankly), epistemic (e.g. probably), evaluative (e.g. unfortunately); and exocomparative (e.g. similarly). He argues that adverbs apply to different ontological categories (speech act, fact, proposition, event), and only those that can apply to events (under the clausal reading) can also have a manner reading. He derives the latter by a general manner rule that turns predicates of events (i.e. events more generally) into predicates of ‘Specified Events’ (e.g. the event of reading a book). Other authors propose various accounts to derive the clausal reading from the manner reading (e.g. McConnell-Ginet 1982, Eckardt 1998, Piñón 2010, Kubota t.a.). We will compare these different options in the course.
Ontological issues arise also for manners. For example, it is a matter of debate whether the unavailability of manner modification for (many) stative predicates calls for an ontological distinction between events (that have a Davidsonian event argument) and (some) states (which do not; e.g. Katz 1995, 2003, 2008, Maienborn 2003, 2005). Alternatively, it has been argued that the difference is rather conceptual in nature, and both events and states come with an event argument (e.g. Rothstein 2005, Mittwoch 2005, Geuder 2006). Another ontological issue is the nature of manner itself. Whereas the most prominent analysis treats manners as denoting (adjectival) properties of events, some argue that manner should be part of our ontology, on a par with ordinary individuals, events, or times (Dik 1975, Schäfer 2003, 2008, Piñón 2007, Alexeyenko 2012). Yet others (Landman & Morzycki 2003, Landman 2006, Anderson & Morzycki to appear, Gehrke 2011 and subsequent work) treat manner as event kind modification. Anderson & Morzycki discuss parallels between degree and manner in the adjectival and adverbial domain to argue in favor of the idea that degrees relate to state kinds whereas manners relate to event kinds. In the course, we will explore these parallels and finish with a general discussion of commonalities between degree and manner.