We argue for a split semantics of German predicative participle constructions, depending on whether or not the formation of the participle involves prefixation with the dedicated morpheme ge-. Against the background of the analysis of participles of German be-prefixed verbs proposed in Pross (2019), and using the licensing of superlative constructions and ung-nominalizations as tests, we show that ge-prefixed participles denote a result relation between a property of an event and an individual. In contrast, be-prefixed participles, like adjectives, denote properties of individuals. We cast the distinction between event properties and individual properties in a compositional semantics of ge-and be-prefixed participles and show how the resulting semantic distinction allows to predict the distinction between target and resultant state participles drawn in Kratzer (2000) without using the questionable immer noch ‘still’ test.
This paper is about German participles. More specifically, its focus is the analysis of predicative uses of German participles, the formation of which requires prefixation of the base verb with a dedicated morpheme ge- as in (1-b). In a predicative participle construction, the participle is the main predicate of a sentence and follows the copula sein (BE).
German participles that require prefixation of a base verb with ge- as in (1-b) contrast with participles of already prefixed constructions as in (2-b), which do not show prefixation with ge-.
In the present paper we argue that the split formation of German participles corresponds to a split semantics of their predicative use. In this respect, our analysis differs importantly from previous work on German participles (like Rapp 1997; Kratzer 2000; Maienborn 2009; Gehrke 2015), where the semantics of German participles is defined independently of whether or not the participle requires prefixation with ge- and consequently, the formation of participles, and in particular the distribution of ge-, is assumed to be solely a matter of morphophonology. But, as Rathert (2009) shows, the formation of participles, and in particular the distribution of the prefix ge-, cannot be explained in terms of morphophonology alone (e.g. as proposed in Neef 1996) but is to a large extent semantically determined. Rathert considers only those semantic factors that concern the distribution of ge-, but does not connect the distribution of ge- to the interpretation of the participle constructions in which ge- figures. Against this background, the main goal of the present paper is to spell out in detail a semantics for the prefix ge- that explains its function in the formation of participles and sets ge-prefixed participles semantically apart from participles of prefix-verbs as in (2-b). The semantic distinction we aim to correlate with the split formation of participles in German concerns the type of property that underlies the state described in predicative constructions with ge-prefixed participles as in (1-b) and participles of prefix-verbs as in (2-b).
Following Pross (2019), we assume that the state described with participles of prefix-verbs as in (2-b) derives from an “individual property”. Individual properties are prototypically exemplified by predicative constructions with adjectives as in (3).1
In the present paper, we contrast states derived from individual properties with states derived from “event properties”. We argue that participles that require prefixation with ge- describe a state based on an event property, a type of state that is prototypically exemplified by present perfect constructions as in (4).
The semantic distinction between individual properties and event properties reproduces semantic distinctions that have previously been proposed in the literature, like the distinction between target and resultant states (Kratzer 2000), or Kimian and Davidsonian States (Maienborn 2009). But previous work assumes that the states described by predicative participles are derived independently of their morphological makeup, and in particular independently of whether or not ge- figures in the formation of a participle. In contrast, the analysis of participles we develop in the present paper computes the split semantics of participles and the restrictions on their use from their morphosemantic constituents.
It is important to note right in advance that in the present paper we are concerned with adjectival participles, but not with non-adjectival participles of the type discussed in Wasow (1977). Wasow (1977) is also traditionally credited for having brought up the question whether participles are derived in the lexicon or in the syntax, and participles have played a crucial role in advancing specific conceptions of the division of labor between syntax and the lexicon (see in particular Bruening 2014). We took great care to make our analysis independent of specific assumptions about word formation. In particular, we want to leave open whether the morphosemantic constituents of German participles we define in the present paper are syntactic heads that are selected and combined according to syntactic principles or lexical units that are selected and combined according to lexical principles. We assume that the principles of meaning composition are the same in both cases, and thus our semantic analysis can be mapped to syntactic and lexical approaches to word formation alike.
In section 2 we introduce the background assumptions we make about the event structure of German verbs and the predication of properties with participles of prefix verbs like bemalen in (2-a). In section 3 we argue that German predicative participles fall into two subclasses, one of which patterns with adjectival predicatives. We explain the difference between the two classes by showing that German predicative participles can be used to predicate two different types of properties, which we label “individual” and “event” properties, respectively. We propose the licensing of ung-nominalizations and superlative constructions as tests to distinguish between the two. With a systematic assessment of those verb classes that form their participles with the prefix ge-, we argue for a correlation of the predication of event properties with the presence of the prefix ge-. In section 4, we compare our proposal with previous approaches to predicative participles with respect to the ontology of states. We show that the distinction between target state participles and resultant state participles proposed in Kratzer (2000) can be grounded in the distinction between individual and event properties. Our approach can thus be understood as a restatement of Kratzer’s analysis which, however, avoids the problems that arise when compatibility of a participle with immer noch ‘still’ is used as a test to distinguish target and resultant states. Section 5 concludes.
In this section we provide a short overview of the basic regularities of the formation of German participles; readers who are familiar with the morphology of German participles can safely skip this section.
There are two participle suffixes in German, a suffix -t and a suffix -en. Weak verbs do not involve stem-vowel changes in any tense (comparable to regular verbs in English) and systematically form their participle with -t (5). Strong verbs often involve stem-vowel changes (comparable to irregular verbs in English) and form their participle with -en (6). Mixed verbs as in (7) form their participles with -t, like weak verbs, but involve a stem-vowel change, like strong verbs.
Verbs like those in (5)–(7) require an additional morphological operation, prefixation with a dedicated morpheme ge- to form a participle, whereas verbs that already have a prefix, as in (8) and (9) do not.
(5)/(6) show that ge-prefixation is independent of the weak/strong distinction and that the participle suffixes -t and -en are independent of the ge-prefix. But the requirement for ge-prefixation to form a participle is independent of morphological complexity, or so it seems, insofar as morphologically complex verbs like those in (10)/(11) also require ge-prefixation of the verb stem.
The difference between the verbs in (8) and (9) on the one hand and those in both (5)/(6) and (10)/(11) on the other correlates with the separability of preverbal morphemes. Preverbal morphemes like be- in (8), ver- in (9) and the prefix ge- in (5)–(7) are inseparable from the verb. We call such inseparable preverbal morphemes prefixes, and distinguish “prefix verbs” from “particle verbs” as in (10)/(11), where the morphemes an- and ein- are separated from the remainder of the verb in the verb second configuration of main clauses (e.g. when the complementizer is causal denn ‘(causal) since’), see(12-a) vs. (12-c). Note that there is no separation of either verbal prefixes or particles in subordinate clauses (e.g. when the complementizer is weil ‘since’), see (12-b) vs. (12-d).
Verbal prefixes like be- are in mutually exclusive distribution with the prefix ge-, see (13). In general, no more than one prefix can occur on a German verb.
For a more detailed discussion of the peculiarities of the formation of German participles, we refer the reader to Rathert (2009).
The main goal of this paper is to derive a fine-grained typology of German predicative participles based on a systematic investigation of the event structure and morphological make-up of the input to participle formation. In this section we introduce the background assumptions underlying the tests which we use to tease apart participles that predicate individual properties and participles that predicate event properties. Following Roßdeutscher & Kamp (2010), we use the licensing of ung-nominalizations as a probe into the event structure of German verbs, and following Pross (2019), we use the licensing of superlative constructions as a probe into the predication of properties with participles. Before we introduce these tests in more detail, it is necessary to point out that using constraints on word formation as diagnostic tools rests on the assumption that there are systematic patterns of word formation. In this section, and in the analysis of ge-prefixed participles we develop in the next section, we focus on systematic patterns of word formation and discuss prototypical examples to illustrate these patterns. But there are of course always exceptions to the rule which are to a greater or lesser extent difficult to fit under a certain constraint of word formation. We discuss such exceptions to the rule in section 2.4, after we have established the set of generalizations against which such exceptions appear after all.
Rappaport Hovav & Levin (2010) distinguish two broad semantic classes of verbs. The first class of “manner verbs” encodes the manner in which some action is carried out and describes internally caused events. The second class of “result verbs” encodes the coming about of some particular result state and describes externally caused events. Result verbs have a bi-eventive lexical-semantic structure in that they describe a relation between two events, e1 and e2, where – following Dowty (1979) – e1 is a CAUSE event that causes a change of state event e2 (a BECOME event). Bi-eventivity is thus to be understood in contrast to mono-eventivity, which is associated with manner verbs that describe an activity, i.e. a single event e (a DO event). Roßdeutscher & Kamp (2010) correlate bi-eventivity of a verb with the licensing of an ung-nominalization: only transitive result verbs but not manner verbs license ung-nominalizations. E.g., the verb säubern ‘to clean’ in (14-a) is a result verb. It predicates a change from being dirty to being clean of the obligatory direct object. Accordingly, säubern has an ung-nominalization, see (14-b).
Intransitive German verbs generally “do not have ung-nominals; this is the case irrespective of whether the verb is unergative [or] unaccusative” (Roßdeutscher & Kamp 2010: 176), cp. (15).
An important qualification of the claim that only transitive result verbs license ung-nominalizations is that the transitivity requirement is understood in a strong sense that excludes verbs like essen ‘to eat’ in (16) that participate in the unspecified object alternation (16-a)/(16-b).
The example (16-b) may appear like a perfect instance of a transitive description of a change of state in the sense that Peter’s eating causes the apple to undergo a change of state from being not eaten to being eaten. But this resemblance is only superficial. Verbs like essen that participate in the unspecified object alternation, typically so-called “incremental theme verbs”, have been argued to be manner verbs (Rappaport Hovav 2008), with the transitive construction being constructed out of the manner verb (see e.g. Kratzer 2004; Kennedy 2012).
The restriction of the formation of ung-nominalizations to “core transitive” verbs (in the terminology of Kratzer 2005, see also Levin 1999 for discussion) is in particular telling with respect to the event structure of German prefix-verbs. (17) and (18) are one but many examples where the unprefixed base verb has no ung-nominalization but the prefixed verb has an ung-nominalization (see e.g. the corpus study in Roßdeutscher 2010 and further discussion in Pross 2019).
Roßdeutscher & Kamp (2010) argue that the contrasts in (17)/(18) are difficult to explain under the assumption that the prefixed verb bemalen is derived from the verb malen. Instead, they propose that the same “root” √mal can be inserted into the templatic representation of a manner verb (which may be either a syntactic or a lexical-conceptual structure) or into the templatic representation of the event structure of a result verb, deriving either a core-transitive result verb that licenses an ung-nominalization or a non-core transitive manner verb that does not license an ung-nominalization.
Given that we are interested in the predication of properties in predicative constructions, a natural starting point are predicative constructions where the main predicate is a lexical (i.e. underived) adjective like kühl ‘cool’. A widely accepted view is that gradable adjectives predicate a property of an individual by determining a value on a scale with respect to some contextual standard (see e.g. Kennedy & Levin 2008) as in (19).
|(19)||Das Bier ist kühl (relativ zum relevanten Standard für Kühle)|
|‘The beer is cool (relative to the relevant standard for coolness)’|
In the following, we refer to properties that are predicated with lexical adjectives in copula constructions with BE (relative to a standard of comparison) as individual properties. A unique property of German adjectival constructions that denote individual properties is the licensing of superlative constructions that are derived by suffixation of the dedicated morpheme -st- as in (20).
Individual properties can also be predicated with adjectives like hungrig ‘hungry’, where the adjective is derived from the noun Hunger ‘hunger’. Francez & Koontz-Garboden (2017) propose that nouns like Hunger denote what they call a “quality”. They argue that because qualities cannot be predicated of individuals directly (21-a), but portions of qualities are possessed by individuals (21-b), the derivation of an adjective from a quality-denoting noun requires possessive morphology (21-c).
(22) is a simplified denotation for hungrig according to the possessive analysis of Francez & Koontz-Garboden (2017).
|(22)||hungrig ⇝ λx.∃p.[POSS(x)(p) ∧ p ⊆ hunger]|
According to (22), hungrig predicates of an individual possession of a portion p of the quality denoted by Hunger.
Unlike Roßdeutscher & Kamp (2010), who follow Wunderlich (1987) and analyze the prefix be- as an instance of preposition incorporation, Pross (2019) argues that be- is an adjectival possessive morpheme (like -ig) which allows to predicate a quality of an individual, similar to the adjectival case (21-c). The prototypical case of quality possession predication with be-prefixed constructions are be-prefixed participles as in (24) which have no corresponding verb, see (24) and the parallel examples (25) (cp. Dewell 2015; Günther 1974). Notably, since the class of be-prefixed participles illustrated with (24)-(26) does not have a corresponding verb, the participle constructions cannot be explained under the traditional assumption that “[o]nly verbs […] can appear in German adjectival passives.” (Gehrke 2015: 908–909, generalizations 1 and 2).
The qualities predicated with be-prefixed participles in (24) and (25) are quite tangible and closely related to the meaning of the underlying noun Gabe ‘gift’ in (24) or Ruhm ‘fame’ in (26). But in general, the qualities predicated with be-prefixed participles are often abstract and shifted remnants of diachronic meaning change. For example, while the synchronic meaning of the noun Mal ‘mark, spot’ is restricted to a mole or memorial, diachronically Mal referred to a sign or picture (Grimm & Grimm 1854: Bd. 12, Sp. 1494) and (Kluge 2002: 592), and it seems to be this no longer existent meaning of Mal that is still inherent in the quality that is predicated in (23-b), and also underlying the meaning of the corresponding verb malen (Grimm & Grimm 1854: Bd. 12, Sp. 1501), (Kluge 2002: 593). Note that it is for this reason that we gloss the quality associated with the root √mal in (23) as ‘mark’.
While determining the meaning of a particular be-prefixed construction may turn out to be difficult, the actual value of an analysis of the prefix be- as in Pross (2019) shows up in comparison with adjectives. Like adjectives, be-prefixed participles license superlative constructions (27).
Superlative constructions with adjectives do not make reference to an underlying event, or allow for a temporal interpretation: (27-a) cannot be understood to mean “the most often painted wall”. Thus, Pross (2019) argues, the licensing of superlatives by be-prefixed participles shows that like adjectives these participles do not make reference to an event. We discuss the formation of superlatives as a test for the denotation of an individual property in the absence of an underlying event in more detail in section 3.1. Moreover, like adjectives (and verbs derived from them), be-prefixed constructions license ung-nominalizations (28).
One further piece of evidence Pross cites in favor of an analysis of be-prefixed participles according to which they are formed independently of the corresponding verbal construction is that such an analysis allows a straightforward explanation of the otherwise mysterious asymmetry of copredication with the ung-nominalization (28-b) in (29) and (30).
According to the tests for event and state denotation of German ung-nominalizations proposed in Ehrich & Rapp (2000), the data in (29) shows that the event denotation of Bemalung makes available a state denotation which can serve as the antecedent of the anaphoric pronoun sie ‘it’ in (29). But the state denotation of Bemalung in (30) does not make available an event denotation which can serve as the antecedent of the anaphoric pronoun in (30). Pross argues that copredication asymmetries with be-prefixed ung-nominalizations as in (29)/(30) show that the state denotation of Bemalung is not derived from the lexical entry of the verb bemalen but from the lexical entry of the participle bemalt, which does not make reference to an event. The formation of ung-nominalizations from be-prefixed participles that lack a corresponding verb is attested e.g. by examples as in (31).
The possibility to form ung-nominalizations from participles, and in the absence of a corresponding verb, leads Pross to a redefinition of the constraint on the formation of ung-nominalizations put forward by Roßdeutscher & Kamp (2010). According to Pross (2019), a construction has an ung-nominalization iff it predicates an individual property, and it is with respect to this constraint that we will consider ung-nominalization as a test for the denotation of an individual property in the remainder of the present paper. That is, if a construction does not license an ung-nominalization, then it does not predicate an individual property.
In the analysis of Pross (2019), the quality underlying the participle construction in (23-b) is derived from the noun Mal. Pross associates the derivation of a quality from a noun with the function of the participle suffix -t and proposes that one way to formally approximate the derivation of a quality from an individual-denoting noun is by intensionalization of the denotation of that noun.2 For ease of exposition, we recast the main point of Pross’ DRT-based analysis in a typed λ-calculus. Assuming that the type of individuals is e, the type of events and states is ev, the type of truth values is t and the type of propositions is s, the denotation of the participle suffix -t is a function that derives a propositional function (an intension) of type ⟨e, ⟨s, t ⟩⟩ from a noun denotation of type ⟨e, t⟩ as in (32).
|(32)||participle -t ⇝ λQ.λx.^[Q(x)]|
The application of participle -t to a noun denotation constitutes an intermediate stage in the construction of the participle. In particular, regardless of whether the semantics of -t in (32) is linked to a lexical or syntactic operation, it is difficult to assign this operation to a certain category, since there is no word that would correspond to the combination of (32) and a noun denotation. The category-determining head of the participle construction is the possessive prefix be- in (33). It predicates an intension of type ⟨e, ⟨s, t⟩⟩ of an individual of type e via extensionalization of that intension. While Pross correlates this semantic function of be- with that of an adjectival head in the syntax, nothing speaks against associating the semantics of be- in (33) with a lexical adjectivization operation.
|(33)||possessive be- ⇝ λZ.λx.[POSS(Z)(x)]|
The denotation in (34-b) is the property predicated by the adjectival participle bemalt that results from the composition of the denotation of the noun Mal represented as in (34-a) with (32) and subsequent application of (33).
|(34)||a.||mark’ ⇝ λy.[mark(y)]|
Following Maienborn (2005), Pross assumes a denotation of the copula BE as in (35) to derive the state denotation of the predicative participle (23-b) as in (36). Following Asher (1993), Maienborn employs ‘≈’ to relate the abstract concept associated with the state variable s to the expression that characterizes the state.
|(35)||Maienborn (2005: 114)|
|copula BE ⇝ λP.λx.λs.[s ≈ P(x)]|
|(36)||(23-b) ⇝ λx.λ s.[s ≈ POSS(x)(λy.^[mark(y)])]|
Concluding, the purpose of this section was to introduce systematic constraints on the formation of ung-nominalizations and superlatives. Simple adjectives and be-prefixed participles both ascribe individual properties in predicative constructions, and in this respect are semantically similar. In particular, according to the analysis of Pross (2019), neither adjectives nor be-prefixed participles make reference to an event, or are derived from a verb. The properties of be-prefixed participles are thus to be expected to differ importantly from the properties of participles which are derived from verbs. In section 3, we discuss the relevant differences with respect to the formation of ung-nominalizations and superlatives for participles the derivation of which requires prefixation of a verb with the prefix ge-.
One challenge to an analysis that aims to make falsifiable claims about productive morphosemantics are lexical idioms. Like phrasal idioms, lexical idioms are constructed according to the pattern of the compositional construction, but have a fixed and shifted meaning that cannot be decomposed into the meanings of constituents of the construction. One major source of lexical idioms is diachronic change, which often conceals the original semantic composition of morphological complex verbs to an extent where a synchronic semantic decomposition of construction is no longer possible albeit the morphology of the original composition is still present. For example, we followed Pross (2019) in choosing be- as a prefix to be compared with ge-, because be- is a “remarkably simple and regular prefix” (Dewell 2015: 53). But there are also lexical idioms where be- cannot be understood as a prefix, and where Pross’ analysis is thus not applicable. E.g, as a reviewer notes for the verb beginnen ‘to begin’, there is no lexical root √ginn to which be- could attach although diachronically, beginnen has been analyzed as being derived from a root that does not contain the prefix be-, see e.g. Kluge (2002: 102). Similar considerations about lexical idiomacy apply to other constructions mentioned by reviewers and editors involving be- like e.g. the verb bedauern ‘to regret’. The special status of lexical idioms like beginnen or bedauern is also often indicated by the ungrammaticality of such be-prefixed idioms in predicative constructions with BE, see (37-a) and (37-b).
According to our analysis, the ungrammaticality of (37-a) and (37-b) in predicative constructions with BE indicates a meaning shift of the be-prefixed constructions towards a perfective interpretation (where perfective interpretations are discussed in more detail in section 3.2) rather than an individual property. This meaning shift is made plain by the ungrammaticality of the construction with a construction BE but the grammaticality of the present perfect construction in (37-c), where the auxiliary HAVE is selected. Looked at in this light, the ungrammaticality of the copula constructions (37-a) and (37-b) can be understood to indicate that the corresponding verbs are not of the same construction type as those be-prefixed verbs that are grammatical in copula constructions with BE and thus do not fall under the generalization about be-prefixed participles we have put forward.
In the previous section we introduced an analysis of BE-prefixed participles according to which such participles pattern with adjectives in copula constructions. Participles prefixed with BE- and adjectives both predicate individual properties, are not derived from a verb, and do not make reference to an event. This parallel analysis accounts for the patterning of adjectives and BE-prefixed participles with respect to the formation of ung-nominalizations and superlative constructions, and for the large number of BE-prefixed participles which do not have a corresponding verb.
In the current section, we show that adjectives and BE-prefixed participles that predicate individual properties differ importantly from ge-prefixed participles derived from verbs with respect to the formation of ung-nominalizations and superlatives. Our account thus differs from the extensive literature on German predicative participles – Rapp (1997); Kratzer (2000); Maienborn (2007); Welke (2007); Gehrke (2015), to name just a few prominent works – in that we do not follow the traditional assumption that German be-prefixed participles are derived from verbs. In particular, we claimed that the form, use and meaning of be-participles is not determined by properties of a (possibly) underlying verb.
We contrast the pattern exemplified by adjectives and be-prefixed participles with the absence of ung-nominalizations and superlatives with ge-prefixed participles, and argue that this contrast cannot be explained under the traditional assumption of a uniform event-based semantics of German participles. We argue that the predication of event-based properties systematically correlates with the presence of the prefix ge-, and that the acceptability of ge-prefixed predicatives depends on whether or not an individual property can be reconstructed from the lexical semantics of the base verb, as has been previously observed in the literature. We cast the observations about the role of ge- with respect to the predication of event properties into a compositional semantics of ge- and its interaction with the semantics of the participle suffix -t by systematically going over the verb classes that form their participles with the prefix ge-. Eventually, we thus arrive at a fine-grained typology of German participles, which in section 4 we employ to predict the distinction between target state participles and resultant state participles proposed in Kratzer (2000).
We begin our discussion with verbs derived from adjectives like kühl ‘cool’ which regularly form their participles with the prefix ge-, and are generally grammatical in predicative constructions with BE (they are “good inputs” to predicative formation according to Gehrke 2015).
The relevant observation with which we motivate our distinction between the meaning of ge-prefixed participles and the meaning of constructions that ascribe individual properties is that unlike individual property predicates (such as be-prefixed participles or adjectives) ge-prefixed event property predicatives like (38-c) cannot be used in superlative constructions,3 see (39-a) and do not license ung-nominalizations, see (39-b).
It is important to note at this point that we use superlative constructions without further modification and in prenominal position as a test to distinguish individual properties from event properties but not comparative constructions. The reason for this is that comparative constructions can also be interpreted relative to a temporal standard of comparison. In the grammatical example (40-a) (ex. (30-d) of Maienborn 2007), the ge-prefixed participle gelesen ‘read’ is compared with respect to the frequency of how often the works of an author have been read, i.e. an event property. In contrast, the unmodified superlative of the same participle in prenominal position is clearly ungrammatical, which we take to indicate that the participle gelesen denotes an event property but not an individual property.
The important semantic difference between adjectives as in (38-a) (kühl ‘cool’) and the participles of the verbs related to them as in (38-c) (gekühlt ‘cooled’) pertains to the way in which the property of being cool is linguistically predicated of the beer with (38-a) and (38-c), respectively. The truth of (38-a) depends just on whether or not the beer is cool (relative to a standard of comparison for coolness). In contrast, the truth of (38-c) depends on whether or not the beer is now cooler than it was. That is, (38-a) can be true if the temperature of the beer never changed and was always the same, and consequently there was no event of cooling the beer. But (38-c) is false if the temperature of the beer has never undergone a change of degree. Following the analysis of Kennedy & Levin (2008), we assume a denotation for the verb kühlen as in (41), according to which the event described by the verb kühlen measures out the degree of change Δ in terms of the difference between the degree of coolness at t and the degree of coolness at t0 < t, where t0 is the initial and t the final time point of the event e described by the verb.4
|(41)||kühlen (verb) ⇝ λy.λe.[coolΔ(y)(e) ≥ stnd(coolΔ)]|
To further establish the pattern indicative for the semantic difference between an adjective and the ge-prefixed participles of the verb derived from that adjective, consider the adjective trocken ‘dry’ in (42) and the corresponding participle in (43).
As for (38-a) and (38-c), the truth-conditions for (42-a) and (43-a) differ in that the truth of (43-a) requires that the laundry was damp and is dry now, but (42-a) can be true if the laundry has always been dry. The truth of adjectival predicatives like (38) and (42-a) depend on individual properties (and accordingly, (42-a) has an ung-nominalization as in (42-c)). In contrast, the truth of participles like (38-c) and (43-a) depends on properties of the relevant event, a type of property which we call “event properties” in the following. The truth-conditions of the predication of an event property are independent of the truth-conditions of individual properties. Consequently, the difference between individual properties and event properties is a categorical difference rather than a matter of complexity or underspecification as is assumed e.g. in the analysis of (Maienborn 2009: 45), who proposes that “adjectival passives only differ from adjectival base predicates in that they express an internally more complex and semantically underspecified property.”
We illustrated the truth-conditional contrast between individual property predicatives and event property predicatives with the example of adjectives and ge-prefixed participles of the corresponding deadjectival verb. But as we argued in section 2, be-prefixed participles pattern with adjectives with respect to the predication of an individual property and thus the same contrast also arises between be-prefixed participles and ge-prefixed participles. To make this contrast formally explicit, we propose to distinguish event properties and individual properties in an analysis of ge-prefixed participle predicatives like (38-c) that parallels the analysis of be-prefixed participle predicatives like that in Pross (2019) – but with one important difference. Whereas the underlying qualities in be-predicatives are derived from nouns, the underlying event properties in ge-predicatives are derived from verbs. It is this difference, we argue in the following, that separates ge-prefixed participles from be-prefixed participles and adjectives (neither of which makes reference to an event).
For the denotation of the participle suffix -t, we have to take into account that the lexical entry in (41) involves two variables, one for the internal argument and one for the event described, i.e. the type of (41) is ⟨e, ⟨ev, t⟩⟩. We thus modify the derivation of a propositional function with -t in (32) as in (44), where Z is of type ⟨e, ⟨ev, t⟩⟩.
|(44)||-t (+event argument) ⇝ λZ.λy.λe.^[Z(e)(y)]|
Recall that Pross (2019) analyzes the prefix be- as an adjectival head (or a lexical adjectivizer) that predicates a quality of an individual via the concept of possession. In a parallel manner, we propose to analyze the prefix ge- as an adjectival head (or a lexical adjectivizer) the function of which is to predicate an event property. More precisely, ge- allows to predicate an event property of an individual via a “result” relation, which we represent as a relation RES(U)(x) between an event property U and an individual x. We use the “result”-relation RES between an individual x and an event property U to reflect the unique truth-conditions of the ascription of event properties. Whether or not the internal argument x has an event property U is to be judged on the basis of whether or not the property U comes about as a result, consequence, or effect of the event described by the underlying verb in which the internal argument x participated. The consequent denotation of the prefix ge- we propose is given in (45), where U (the denotation of (44) applied to a one-place predicate of events like (41)) is of type ⟨e, ⟨ev, ⟨s, t⟩⟩⟩.
|(45)||ge- ⇝ λU.λx.[RES(U)(x)]|
On these premises, we can now compute the denotation of the predicative participle (38-c) as follows. First, we derive an event property by functional application of the denotation of the participle suffix -t in (44) to the denotation of the verb kühlen (41). Several instances of λ-conversion yield the denotation in (46-c).
|(46)||a.||λZ.[λx.λe.^[Z(e)(x)]](λy.λe.[coolΔ(y)(e) ≥ stnd(coolΔ)]|
|// by functional application: [(44)]((41))|
|b.||λx.λe.^[λyλe. coolΔ(y)(e) ≥ stnd(coolΔ)](e)(x)|
|// by λ-conversion: (Z/λy.λe. coolΔ(y)(e) ≥ stnd(coolΔ)|
|c.||λx.λe.^[coolΔ(x)(e) ≥ stnd(coolΔ)]|
|// by two instances of λ-conversion: (e/e), (y/x)|
Next, we apply the denotation of ge- (45) to the event property denotation in (46-c) and after several instances of λ-conversion arrive at the individual property denotation in (47-c).
|(47)||a.||λU.λx.[RES(U)(x)](λy.λe.^[coolΔ (y)(e) ≥ stnd(coolΔ))]|
|// by functional application: [(45)]((46-c))|
|b.||λx.[RES(λy.λe.^[coolΔ (x)(e) ≥ stnd(coolΔ)(y)])(x)|
|// by λ-conversion: (U/λyλe.[^coolΔ(y)(e) ≥ stnd (coolΔ)])|
|c.||λx.[RES(λe.^[coolΔ(x)(e))(x) ≥ stnd(coolΔ)(x)]]|
|// by λ-conversion: (y/x)|
Finally, we stativize the derived individual property predicate by applying the denotation of BE (35) to (47-c), which yields the state denotation of (1-b) in (48-c).
|(48)||a.||λP.λy.λs.[s≈ P(y)](λx.[RES(λe.^[coolΔ(x)(e) ≥ stnd(coolΔ)(x)])]|
|// by functional application: [(35)]((47-c))|
|b.||λy.λs.[s≈ λx.[RES(λe.^[coolΔ(x)(e) ≥ stnd(coolΔ)(x)])](y)|
|// by λ-conversion: (P/λx.[RES(λe.^[coolΔ(x)(e))(x) ≥ stnd(coolΔ)(x)]]|
|c.||λy.λs.[s≈[RES(λe.^[coolΔ(y)(e)) ≥ stnd(coolΔ)(y)])]|
|// by λ-conversion: (x/y)|
The truth-conditions of sentences in which a ge-participle of a deadjectival verb serves as complement to a copula, as in (38-c) or (43-a), thus have to be evaluated in terms of the difference over time between the degrees to which the underlying adjective applies to a given individual (given in sentences like (38-c)/(43-a) as the grammatical subject). Consequently, when the event property predicated by the participle of a deadjectival verb is defined in terms of (degrees of differences of) individual properties, and if the function of the copula BE is to predicate individual properties, this explains why participles of deadjectival verbs can occur as copula complement.5
Summarizing, we argued that participles prefixed with be- predicate individual properties and pattern morphologically and semantically with lexical adjectives whereas ge-prefixed participles of deadjectival verbs predicate event properties and do not pattern morphologically or semantically with lexical adjectives.
So far, we considered be-prefixed participles and ge-prefixed participles of otherwise unprefixed deadjectival verbs like kühlen and trocknen. Both these kinds of participles are grammatical, acceptable and productive when used in copula constructions with BE. In this respect, ge-prefixed participles of unprefixed deadjectival verbs differ importantly from the other classes of ge-prefixed constructions with morphologically simple verbs which are “bad inputs” to predicative formation according to Gehrke (2015). In the remainder of this section of the present paper, we systematically go over these “bad” classes of ge-prefixed participles of otherwise unprefixed verbs.
We begin our investigation of ge-prefixed participles of non-deadjectival verbs with the example of incremental theme verbs (Dowty 1991; Krifka 1992) as in (49-a). Broadly speaking, these verbs are distinguished by participation in the unspecified object alternation (Levin 1993) as in (49-a)/(49-b).
|(49)||a.||Peter kochte die Kartoffel.||‘Peter cooked the potato.’|
|b.||Peter kochte.||‘Peter cooked.’|
Following Rappaport Hovav (2008); Kennedy (2012), and like Gehrke (2015), we analyze incremental theme verbs as non-core transitive verbs in the sense of Levin (1999); Kratzer (2005), i.e. as manner verbs in which a direct object is only optional but not subcategorized by the verb. The incremental theme verb kochen thus denotes a property of events (the manner of the action described by the verb) as in (50).
|(50)||kochen ⇝ λe.[cook(e)]|
No individual property is made available by manner verbs. Accordingly, manner verbs do not license a superlative construction (51) nor an ung-nominalization (52).
We proposed that predicative participles are only grammatical when an individual property is predicated. Thus, it is expected that participles of incremental theme verbs are questionable out of the blue, an expectation which is borne out in line with the literature (Rapp 1997; Kratzer 2000; Maienborn 2007; Gehrke 2015), see (53).6
Note that like incremental theme verbs, otherwise unprefixed core transitive and ditransitive verbs that are not derived from adjectives form their participles with the prefix ge-. Accordingly, we expect that in general ge-prefixed predicative participles of such verbs are questionable out of the blue, see (54).
Traditionally, the direct object in transitive constructions with non-core transitive verbs is analyzed as an incremental theme that measures out the progress of the event described by the verb. That is, the relation between the theme and the event described by the verb is a thematic relation of incrementality. The exact nature of this incrementality relation has been subject to debate in the literature, see e.g. Krifka (1992); Kratzer (2004); Kennedy (2012) for some opposing proposals. As the goal of the present paper is rather independent of the exact specification of the incrementality relation, in the following we represent the relation between the incremental theme and the activity event in transitive uses of non-core-transitive verbs as an “aspect”-relation ASP between an individual x and an event e, leaving open the exact formalization of the incremental semantics of the ASP-relation. We define the morphologically empty ASP-operation involved in incremental theme verbs as in (55).
|(55)||ASP (incremental) ⇝ λx.λe.[ASP(e)(x)]|
Following Kennedy (2012), the application of the ASP-operator to the manner denotation in (50) extends the argument structure of the verb with a direct object via the composition principle of event identification, yielding the denotation in (56).
|(56)||kochen (+theme) ⇝ λx.λe.[ASP(e)(x) ∧ cook(e)]|
When the denotation in (56) is applied to the denotation of participle morphology -t (44) the event property denotation in (57) is derived.
|(57)||λy.λe.^[ASP(e)(y) ∧ cook(e)]|
It should be clear at this stage of the derivation what the main difference between deadjectival and non-core transitive participles is. According to our analysis, the property that would be predicated with the denotation in (57) is not an individual property, nor can an individual property be reconstructed from the base verb. Instead, the ge-prefixed participle gekocht derived in (57) denotes an event property. But what property is predicated when we combine the event property denotation (57) with the denotation of potato and then apply the denotation of the copula BE (35), so as to derive the denotation of (53) in (58)?
|(58)||(53) ⇝ λs.[s≈ λe.^[ASP(e)(potatoe′) ∧ cook(e)]]|
In the next section, we assess the “aspectual” property that would be predicated with (58) in more detail by considering the interaction between predicative participles and perfective aspect.
It is a well-established observation in the literature (Rapp 1996; Kratzer 2000; Maienborn 2007; Gehrke 2015) that the acceptability of (53) is improved when the context in which such constructions are interpreted enforces a shift from the event property ‘be cooked’ denoted by (57), in which the theme figures as a measure of the cooking event, to an interpretation that solely focuses on the result of the event described, i.e. that the potato ‘has been cooked’, see (59).
The truth-conditions of a property like that of ‘having been cooked’ in (59) do not make reference to individual or event properties but in fact resemble the truth-conditions standardly associated with the present perfect tense. Whether or not something has been cooked depends only on whether or not there was a completed cooking event, but not on individual properties of what has been cooked. Formally speaking, the relevant property predicated in (59) comes about through determining the ASP-relation between the underlying manner event and the theme to be a perfective or resultative relation (and this is the contribution of bereits in (59)). One way to make resultative aspect plain in German is to use aspectual particles, and consequently we expect, as also observed by Gehrke (2015) that predicative participles of particle verbs are grammatical when they predicate a resultative property. Consider for example the fully grammatical particle construction (60), where the perfective interpretation towards which (53) is in the example (59) is determined with the particle ab ‘off’.
It is because of the close parallel to perfect constructions (which determine completion of the event described by the verb at some time preceding the evaluation time) that, following established conventions, we propose to call resultative properties the properties predicated with (i) shifted event property predicatives such as (59), (ii) particle constructions as in (60). Accordingly, we call the state that consists in something having a resultative property a resultative state.
The crux of our proposal for the semantics of ge-, however, is that we do not require ge- to operate on a state argument provided by the verb. In this respect, our analysis of ge- differs importantly from previous analyses of ge- like that of Gehrke (2015) in (61).
|(61)||ge-en/t ⇝ λPλyλsλe∃x[P(e)(s)(x)(y)]|
|(Gehrke 2015: (42-b)), P(e)(s)(x)(y) is the input to participle formation.|
Since Gehrke’s ge- in (61) can only be applied to inputs that provide a state argument, it can by definition not occur on incremental theme verbs (which Gehrke, like we do, analyzes as a manner verb). Consequently, examples like (59), where the enforcement of a perfective interpretation renders the predicative participle acceptable cannot be explained semantically. Instead, like Maienborn & Herdtfelder (2015), Gehrke has to assume that examples like (59) are licensed pragmatically, which raises the question why and how ge- can be “pragmatically” combined with a manner verb as in (59).
We noted that the event property predicated with the prefix ge- leaves in principle open the possibility that the event property receives a perfective interpretation. As it happens, ge-prefixed predicative participles and the corresponding perfect participles do not only share the same form but, as Gehrke (2015) argues for the case of unaccusative verbs, also cannot be semantically distinguished. Consider (62).
Because unaccusative verbs select BE as a perfect auxiliary, whether (62) is a predicative or present perfect construction is difficult to determine. As Gehrke (2015: fn. 12) argues, there is no reliable test that would tease apart the predicative and present perfect interpretation of (62). Ge-prefixed predicative participles with unergative verbs are ungrammatical (63-a), because unergative verbs are manner verbs. But it may not be by accident that the same surface form that is ungrammatical in predicative constructions is grammatical in perfect constructions with the auxiliary HAVE (see (63-b)).
The close parallel between predicative participles and perfective aspect has been previously recognized in the literature, see in particular Welke (2007) for an over-view and extensive discussion. Since a detailed discussion of ge-prefixed participles in perfect constructions is beyond the scope of the present paper, we leave a further exploration of our analysis of ge-prefixed participles as a starting point for the possibly uniform semantic analysis of predicative and perfect participle forms to future research.
A summary of the correlation of the form, use and meaning of German participles which we argued for in this section is given in Table 1.
We examined two kinds of predicative constructions, using the licensing of ung-nominalizations and superlatives as tests to tease the two classes apart. On the one hand, predicative constructions with adjectives and be-prefixed participles ascribe individual properties. On the other, predicative constructions with ge-prefixed participles ascribe event properties. We correlated the (un)grammaticality of predicative constructions with the predication of individual and event properties. Adjectives and be-prefixed participles predicate individual properties and are fully grammatical whereas ge-prefixed constructions with alternating transitives do not predicate individual properties at any stage of their derivation and thus in predicative constructions are bad out of the blue. Participles of deadjectival verbs are between the two ends of acceptable and unacceptable predicatives. They are formed with ge- and thus ascribe an event property, but the copula construction is generally acceptable because the predicated event property is by itself derived from an individual property.
A challenge to the analysis proposed in this paper are systematic exceptions to the typology of predicative participles developed so far. For example, a reviewer asks about whether and how our analysis can account of psychological verbs in (64), as some (but not all) psychological verbs license predicative constructions, ge-prefixed superlatives and ung-nominalizations.
If, as the reviewer suggests, (64-a) (and similar constructions with psychological verbs like fürchten ‘to fear’ or lieben) are grammatical in predicative constructions, we believe that systematic patterns as exemplified by (64) show that our typology of ge-prefixed participles is not yet fine-grained enough. According to our tests (ung-nominalization and superlatives), constructions as in (64) suggest that there are ge-prefixed participles that denote individual properties. Consequently, our analysis of the prefix ge- has to be further refined in that ge- can not only be used to predicate an event property (as discussed in section 3) but can also be used to predicate possession of a quality.
In turn, such a possessive analysis of ge-prefixed psychological predicatives would be very much in the spirit of the analysis of psychological qualities like wisdom in Francez & Koontz-Garboden (2017). Given the limitations and scope of the present paper, we leave a further exploration of this parallel to future research. According to our analysis, quality possession predication with ge- is also involved in participles that are puzzling under the assumption that participles are always derived from verbs as in (65).
According to our analysis, the licensing of ung-nominalizations and superlatives indicates that participles as in (65) predicate individual properties rather than event properties. Since individual property predicatives do not make reference to an event, the lack of a corresponding verb is accounted for. Finally, our proposal also suggests a quality possession analysis of predicative participles as in (66), when the possessive prefix can be morphologically silent.910
In summary, we suggest that our proposal for a morphosemantically grounded typology of predicative participles be understood as a starting point for future research into deviant classes of participles, and consequent refinement of the basic distinctions we have established, rather than a final definition.1314
The goal of this section is to evaluate our analysis with respect to the proposal of Kratzer (2000), who distinguishes participles that describe target states from participles that describe resultant states. We argue that our proposal distinguishes target and resultant state participles without invoking the questionable immer noch test on which Kratzer’s analysis rests.
According to Kratzer (2000), two types of predicative participles are distinguished by their compatibility with the modifier immer noch (‘still’) as in (67) and (68).
Adopting terminology from Parsons (1990), Kratzer proposes that participles that allow for modification with immer noch denote “target states” (67) and that those that don’t denote “resultant states” (68). The intuition behind the distinction between these two kinds of states is that target states but not resultant states are alterable, hence target states but not resultant states allow modification with immer noch.
Assuming that overt participle morphology is “meaningless” (Kratzer 2000: 391), Kratzer argues that “only those verbs can have target state passives that characterize states as part of their meaning” (Kratzer 2000: 393). Accordingly, the target state participle aufgepumpt ‘pumped up’ in (67) would be derived from a lexical representation that makes available a target state argument – the result state of the event described by the verb aufpumpen ‘pump up’ in (69-a) – and the participle bewiesen ‘proved’ in (68) would be derived from a lexical representation as in (69-b), which does not make available a state argument.
|(69)||a.||Kratzer (2000: 390)|
|aufpump- ⇝ λs.λe.[pump(e) ∧ event(e) ∧ inflated(the–boat)(s) ∧ cause(e)(s)]|
|b.||Kratzer (2000: 395)|
|beweis- ⇝ λe.[prove(the–theorem)(e)]|
Kratzer’s analysis is challenged by deadjectival verbs as in (70).
According to Kratzer’s immer noch test, (70) describes a resultant state, and thus the lexical stem from which the participle geleert ‘emptied’ is derived should not make available a target state argument. But deadjectival verbs lexicalize a target state, and thus should be compatible with immer noch. To account for this mismatch, Kratzer proposes that deadjectival verbs like leeren should be reanalyzed as periphrastic causatives in which an adjective incorporated into a light verb machen (‘make’) as in (71).
Because (71) is ungrammatical, Kratzer suggests that target state participles of deadjectival verbs are ruled out for independent reasons: verbs derived through incorporation into a light verb never license target state participles. Besides bringing up new problems (i.e. whether a periphrastic causative analysis of deadjectival verbs is viable, see e.g. the competing well-established proposals by Hale & Keyser 1993; Kennedy & Levin 2008), the proposed analysis falls short of deadjectival verbs in general. For the overwhelming majority of deadjectival verbs, the participle is compatible with immer noch, although the corresponding periphrastic causative is ungrammatical, see (72) and (73).
While Kratzer attempts to explain that participles of deadjectival verbs are incompatible with immer noch although deadjectival verbs lexicalize a target state with an (unsuccessful) redefinition of their lexical semantics, Irmer & Mueller-Reichau (2018) propose to explain the incompatibility of participles of deadjectival verbs and immer noch with a refined view of the conditions under which a participle is (in)compatible with immer noch. Irmer and Mueller-Reichau explain (70) by proposing that an emptied mailbox is in a state in which it does not afford the same change of state that is implied by the participle geleert and expected by immer noch, since that same change of state would require that “one and the same stuff participant is operated on” (Irmer & Mueller-Reichau 2018: 593) both by the participle and the future change of state. That is, (70) requires that the same letters are taken from the mailbox that have been removed before, and this requirement is not compatible with our world knowledge that mailboxes are never filled twice with the same letters.
The heavy reliance of Irmer & Mueller-Reichau (2018) on world knowledge about particular entities and processes requires that compatibility of immer noch is decided on a case by case basis, which takes away much of the appeal of Kratzer’s original proposal to use immer noch as a genuinely linguistic test to tease apart two types of participles. In particular, it is difficult to evaluate the predictions of the approach of Irmer & Mueller-Reichau (2018) for verbs derived from absolute adjectives other than leer. For example, the participle getrocknet in (74) is incompatible with immer noch albeit the same participant is operated on.
Moreover, the immer noch test also makes wrong predictions for result verbs like verbrennen (‘to burn’) and sterben (‘to die’), and these wrong predictions are not accounted for in the reanalysis of Irmer & Mueller-Reichau (2018). According to the established tests (Beavers 2010; Rappaport Hovav & Levin 1998) for the lexical entailment of a result state (a target state in Kratzer’s terminology) these verbs differ in their entailment of a change of state from manner verbs like wischen (‘to wipe’), compare (75-a) vs. (75-b)/(75-c).
|(75)||a.||Er hat den Tisch gewischt, (aber er ist nicht sauber/aber nichts an ihm ist anders).|
|‘He has wiped the table (but it is not clean/but nothing is different about it).’|
|b.||#Das Papier ist verbrannt, aber (es ist nicht verbrannt/aber nichts an ihm ist anders).|
|‘The paper is burnt, but (it isn’t burnt/nothing is different about it).’|
|c.||#Peter ist gestorben, aber (er ist nicht tot/aber nichts an ihm ist anders).|
|‘Peter died, but (he isn’t dead/nothing is different about him.)’|
Given that verbrennen and sterben lexicalize a target state, Kratzer’s analysis would thus predict that verbrennen and sterben are compatible with immer noch, a prediction which – as (68) and (76-a)/(76-b) show – is not borne out.
|(76)||a.||Das Papier ist (*immer noch) verbrannt.|
|‘The paper is (*still) burnt.’|
|b.||Peter ist (*immer noch) gestorben.|
|‘Peter is (*still) died.’|
Kratzer explains the unexpected incompatibility of result verb participles with immer noch as in (76) by proposing that incompatibility with immer noch can only be used as a test for “true” resultant state participles. Only the states described by true resultant state participles are irreversible, but those described by “fake” resultant states (i.e. target state participles) as in (76) are in principle reversible, since (76-b) is fine in “contexts in which people are assumed to come back to life” (Kratzer 2000: 387). But this raises the question for whether and how “false” resultant state participles like (76-a) or (76-b) can be distinguished from “true” resultant state passives like (68) and (70), without presupposing what the immer noch test is intended to probe for. Actually, similar contextual considerations as for (76-b) should be applicable to every result verb participle, and then (70) is just as fine as (76-b) in contexts in which mailboxes may fill up after being emptied. Once the immer noch test is subject to contextual considerations, seemingly “true” resultant state participles like (68) appear to be similarly questionable. As (77) shows, beweisen ‘to prove’ may entail a result state (makes available a target state argument in Kratzer’s terminology), contrary to the analysis (69-b) proposed by Kratzer.
|(77)||#Die Hypothese ist bewiesen, aber (sie ist nicht bewiesen/aber nichts an ihr ist anders).|
|‘The hypothesis is proved, but (it isn’t proved/nothing is different about it.)’|
(68) is just as fine as (76-b) in a context in which a proof may be disproved (e.g. in the context of the Riemann hypothesis).
In summary, it appears that the distinction between target states and resultant states is difficult to implement using the immer noch test as a guiding principle. Instead, the distinction between individual and event properties we have established underlies the distinction between resultant states and target states as we understand it to be characterized in Parsons (1990). Parsons illustrates his conception of a target state with (78).
|(78)||The ball is on the roof.|
Whether or not a ball is on a roof can be determined by inspection of the ball’s individual properties like its position in space relative to the roof. In contrast, a resultant state is the “state of my having thrown the ball onto the roof” (Parsons 1990: 235). As for deadjectival verbs and their underlying adjectives, the difference in truth-conditions manifests itself in that fact that (78) can be true even if the ball has never been thrown, whereas a description of the state of having thrown the ball onto the roof is false if the ball has never been thrown. From this point of view, the relevant difference that Kratzer aims to test for with the immer noch test, is not the (ir)reversibility of states but the irreversibility of time. Once a ball has been thrown, this event cannot be undone in time, but since a ball’s being on the roof is an individual property independent of the temporal order of causation presupposed by events, the individual property of the ball can change over time.
The distinction between individual property states and event property states also reproduces the distinction between “Davidsonian” and “Kimian” states that Maienborn (2005) argues to be central to predicatives with BE in general. According to Maienborn, Kimian states “are abstract objects for the exemplification of a property P at a holder x at time t” (Maienborn 2005: 113) and do not make reference to an event whereas Davidsonian states are defined in terms of the causal effects of an event. This parallels our analysis of individual and event property states insofar as individual property states are defined independently of an event whereas event property states are defined with respect to an event. Concerning the addition of “resultative” states to the ontology, event properties can either be stativized as resultant state when combined with the copula BE or perfectivized as resultative states when combined with the auxiliary BE. The improvement of the acceptability of event-property predicatives in perfective contexts is then to be explained as the coercion of an aspectual event property (like culmination or termination) into a resultative property.
The overall ontology of states according to our proposal is summarized in figure (79), where the dotted arrow indicates coercion and k-state and d-state are short for Kimian and Davidsonian states, respectively.
The main goal of the present paper was to familiarize the reader with some of the systematic patterns of the formation and interpretation of German participles. We restricted our attention to ge-prefixed participles in predicative constructions. But we believe the shift of perspective adopted in this paper with respect to participles, from a primary focus on verbs and events to adjectives and properties, may also inform the analysis of other uses of ge-prefixed participles. In particular, although ge-prefixed participles have the same surface form in all their different usages like predicative participle, perfect and passive, the participle is assumed to have a quite different meaning in each of these constructions. In this respect, the present paper constitutes an attempt to identify event properties as the lowest common denominator of ge-prefixed participles in their different possible uses, independent of whether or not the underlying verb makes available a result or target state argument. One direction of future research that arises from the discussion in the present paper is thus whether, and if yes how, a rather abstract semantic analysis of ge-prefixed participles as ours can serve as a uniform input to a further specification of the meaning of the participles in the context of its use.
We already mentioned in section 3.2.2 the close relationship our analysis suggests between ge-prefixed predicative participles and perfect participles. But our analysis also suggests a similarly close cross-connection between ge-prefixed predicative participles and passive constructions. We have dubbed participle constructions with the copula BE “predicative participles”. We chose this term to avoid the more established term “adjectival passive”. The reason for this is that as it stands, nothing in the analysis of predicative participles we proposed involves the passivization of a verbal phrase. Our proposal can thus be understood as an implementation of the idea that “in adjectival passives, the verb’s external argument is truly missing. It’s not that it has been eliminated or suppressed. It was never there to begin with.” (Kratzer 2000: 391). According to our proposal, the states described by predicative participles do not follow from the implications of verbal passivization but are derived from genuinely adjectival constructions.
An interesting observation in this respect, due to Maienborn & Herdtfelder (2017), is that von ‘from/by’-phrases usually associated with verbal passives are licit in adjectival predicative constructions like (80). In the examples in (80), the matrix predicate is an adjective that does not make an event available, and consequently the von-phrase cannot be understood to pick up a passivized external argument of a verb.
|(80)||a.||Peter ist müde von der Reise.|
|‘Paul is tired from the trip.’|
|b.||Der Platz ist weiß von den Hagelkörnern.|
|‘The square is white from the hailstones.’|
In light of non-eventive, thus non-passive, examples involving von-phrases like (80), deadjectival ge-prefixed predicative participles with von-phrases as in (81) do thus not provide conclusive evidence for a passive analysis of predicative participles, but may as well be analyzed along the lines suggested by Maienborn & Herdtfelder (2017) for the adjectival cases in (80).
|(81)||a.||Die Wangen sind gerötet von der Anstrengung.|
|‘The cheeks are reddened by the effort.’|
|b.||Der Körper ist geschwächt von der Krankheit.|
|‘The body is weakened by the illness.’|
One further point we haven’t addressed yet is the open question why no more than one prefix can occur on a German verb. Given that we associated prefixes with the predication of a property, a possible explanation may be that more than one prefix is disallowed for general semantic reasons, i.e. that predicates can only predicate one property of an argument at a time. Thus, when a prefix is applied to a verb, further prefixes are blocked because there is no property bearer available that the predication operation associated with the prefix could target. Such an explanation would be very much in the spirit of the analysis of Rathert (2009), albeit locating the relevant explanatory feature at the more fundamental level of predication in general rather than making assumptions about the semantics of specific prefixes.
Concluding, we developed an analysis of German ge-prefixed participles in copula constructions according to which there is a systematic correlation between the presence or absence of the prefix ge- and the type of state described by the predicative. Our analysis of ge- was developed by comparing the properties of ge-predicatives with predicatives prefixed with be-. But we believe our analysis can in principle be extended to account for the other prefixes as well, based on the observation of Rathert (2000) that prefixes often have an aspectual function. From this point of view, prefixes like ver- may be analyzed as RES-operators that do not operate on events (like ge- does), but rather on those scales or paths in terms of which the result of the event described is conceptualized. Naturally, the same considerations hold for the wide range of particles (which we suggested to be overt realizations of lexical aspect ASP) German verbs can combine with. A further exploration of the morphosemantics of particles and prefixes other than those discussed in the paper is left to future research. But we like to stress that work like Dewell (2015) or Roßdeutscher (2016) shows that a systematic account of the semantics of German prefixes and particles is not as hopeless as is sometimes suggested, e.g. in Kratzer (2000; 2004; 2005). But quick generalizations are also not possible in the other direction. We believe that the complex patterns of German word formation and interpretation can only be revealed step by step through detailed case studies, and thus the discussion of the prefix ge- in the present paper is no more than just one piece of the big puzzle that arises with the investigation of the patterns underlying the form, use and meaning of morphologically complex words.
1We use the term “individual property” to refer to a property of an individual, but note that this does bear any connection to the term “individual argument” of a verb.
2Francez & Koontz-Garboden (2017) analyze qualities as atomic entities of the model-theory. To account for qualities derived from nouns, Pross’ proposal makes use of the analysis of individual concepts in Higher-Order Intensional Logic (Montague 1973), where intensionalization is employed to turn the individual denotation of a noun like man into the propositional function of having characteristic properties of a man (a function from individuals to sets of possible worlds) that returns for each world the extension of the predicate man.
3Note that we are concerned with the productive use of ge- as a prefix in participle constructions only, i.e. where ge- is not present in other forms of the verb such as its infinitive. Accordingly, examples like (i) (from Rapp 1997) do not contradict our claim about the lack of superlative ge-prefixed participles: the infinitive of the verb underlying the participle in (i) is gefährden ‘to endanger’ and thus ge- in (i) is not a prefix of the type under discussion.
4Following Kratzer (1996), we assume that the external argument of verbs is introduced in a separate Voice projection above the actual verbal phrase.
5Formally speaking, in our analysis the acceptability of deadjectival predicatives is due to the fact that the domain of the intensionalization operation in (44) is restricted to event predicates and thus does not affect predicates which although they are in the scope of the intensionalization operator are not event predicates. The relevant non-eventive predicate underlying the predicative in (38-c) is the predicate that denotes the degree of difference in coolness from which the measure function coolΔ is derived according to Kennedy & Levin (2008).
6It should be noted for the sake of completeness that the acceptability of predicative participles of verbs that are not associated with the predication of individual properties can also be improved in a purely pragmatic manner when the event property described by the participle is coerced into a comparable individual property, for example with the help of a contrastive construction as in (i) (see Rapp 1996).
7https://www.facebook.com/ardbuffet/posts/die-gestreifteste-himbeertorte-aller-zeiten/1636035973086702/ Last accessed: May 14, 2019.
8https://de.wikipedia.org/wiki/Bunter_Loewe Last accessed: May 15, 2019.
9https://antolin.westermann.de/all/bookdetail.jsp?book_id=83900 Last accessed: May 15, 2019.
10Title of a painting by Oskar Schlemmer.
We use the following glossing conventions for German examples: NMLZ = nominalizer, PRFX = prefix, PRTC = particle, PTCP = participle, SPL = superlative, CMP = comparative, ADJ = adjectivizer. If there is a suitable translation of a German prefix or particle into English, we use this translation in the morpheme gloss.
We would like to thank the handling editors Berit Gehrke and Olga Borik as well as three anonymous reviewers for having carefully read earlier versions of the paper. The paper profited very much from their instructive and detailed comments. We would particularly like to thank Hans Kamp for reading and commenting on numerous earlier versions of the paper. All remaining errors are our own.
The research reported in this paper was supported by DFG grants CRC-732/B4 (Roßdeutscher and Pross) and PR-1860/1-1 (Pross).
The authors have no competing interests to declare.
Asher, Nicholas. 1993. Reference to abstract objects in discourse. Dordrecht: Kluwer. DOI: https://doi.org/10.1007/978-94-011-1715-9
Beavers, John. 2010. The structure of lexical meaning: Why semantics really matters. Language 86(4). 821–864. DOI: https://doi.org/10.1353/lan.2010.0040
Bruening, Benjamin. 2014. Word formation is syntactic: Adjectival passives in English. Natural Language and Linguistic Theory 32(2). 363–422. DOI: https://doi.org/10.1007/s11049-014-9227-y
Dewell, Robert B. 2015. The semantics of German verb prefixes. Amsterdam: John Benjamins. DOI: https://doi.org/10.1075/hcp.49
Dowty, David R. 1979. Word meaning und Montague Grammar. New York: Springer. DOI: https://doi.org/10.1007/978-94-009-9473-7
Dowty, David R. 1991. Thematic proto-roles and argument selection. Language 67(3). 547–619. DOI: https://doi.org/10.1353/lan.1991.0021
Ehrich, Veronika & Irene Rapp. 2000. Sortale Bedeutung und Argumentstruktur: ung-Nominalisierungen im Deutschen. Zeitschrift für Sprachwissenschaft 19(2). 245–300. DOI: https://doi.org/10.1515/zfsw.2000.19.2.245
Francez, Itamar & Andrew Koontz-Garboden. 2017. Semantics and morphosyntactic variation. Oxford: Oxford University Press. DOI: https://doi.org/10.1093/acprof:oso/9780198744580.001.0001
Gehrke, Berit. 2015. Adjectival participles, event kind modification and pseudoincorporation. Natural Language and Linguistic Theory 33(3). 897–938. DOI: https://doi.org/10.1007/s11049-015-9295-7
Günther, Hartmut. 1974. Das System der Verben mit be- in der deutschen Sprache der Gegenwart. Tübingen: Max Niemeyer. DOI: https://doi.org/10.1515/9783111354613
Hale, Kenneth & Samuel Jay Keyser. 1993. On argument structure and the lexical expression of syntactic relations. In Kenneth Hale & Samuel Jay Keyser (eds.), The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, 53–109. Cambridge, MA: MIT Press.
Irmer, Matthias & Olav Mueller-Reichau. 2018. Script knowledge and the felicity of phase particles in German adjectival passives. Journal of Semantics 35(4). 585–637. DOI: https://doi.org/10.1093/jos/ffy008
Kennedy, Chris & Beth Levin. 2008. Measure of change: The adjectival core of degree achievements. In Louise McNally & Chris Kennedy (eds.), Adjectives and adverbs: Syntax, semantics and discourse, 156–182. Oxford: Oxford University Press.
Kennedy, Christopher. 2012. The composition of incremental change. In Violeta Demonte & Louise McNally (eds.), Telicity, change, and state: A crosscategorial view of event structure, 103–121. Oxford: Oxford University Press. DOI: https://doi.org/10.1093/acprof:oso/9780199693498.003.0004
Kratzer, Angelika. 1996. Severing the external argument from its verb. In Johan Rooryck & Laurie Ann Zaring (eds.), Phrase structure and the lexicon, 109–137. Dordrecht: Kluwer Academic Publishers. DOI: https://doi.org/10.1007/978-94-015-8617-7_5
Kratzer, Angelika. 2000. Building statives. In Proceedings of the Twenty-Sixth Annual Meeting of the Berkeley Linguistics Society, 385–399. DOI: https://doi.org/10.3765/bls.v26i1.1131
Maienborn, Claudia. 2005. On the limits of the Davidsonian approach: The case of copula sentences. Theoretical Linguistics 31(3). 275–316. DOI: https://doi.org/10.1515/thli.2005.31.3.275
Maienborn, Claudia. 2007. Das Zustandspassiv: Grammatische Einordnung – Bildungsbeschränkungen – Interpretationsspielraum. Zeitschrift für Germanistische Linguistik 35. 83–114. DOI: https://doi.org/10.1515/ZGL.2007.005
Maienborn, Claudia & Johanna Herdtfelder. 2015. A compositional account of the event/stative ambiguity of German causal von-modifiers. In Proceedings of SALT 25, 163–183. DOI: https://doi.org/10.3765/salt.v25i0.3061
Maienborn, Claudia & Johanna Herdtfelder. 2017. Eventive versus stative causation: the case of German causal von-modifiers. Linguistics and Philosophy 40(3). 279–320. DOI: https://doi.org/10.1007/s10988-016-9201-8
Montague, Richard. 1973. The proper treatment of quantification in ordinary English. Approaches to Natural Language 49. 221–242. DOI: https://doi.org/10.1007/978-94-010-2506-5_10
Pross, Tillmann. 2019. What about lexical semantics if syntax is the only generative component of the grammar? A case study on word meaning in German. Natural Language and Linguistic Theory 37. 215–261. DOI: https://doi.org/10.1007/s11049-018-9410-7
Rappaport Hovav, Malka. 2008. Lexicalized meaning and the internal temporal structure of events. In Susan Rothstein (ed.), Theoretical and crosslinguistic approaches to the semantics of aspect, 13–42. Amsterdam: John Benjamins. DOI: https://doi.org/10.1075/la.110.03hov
Rappaport Hovav, Malka & Beth Levin. 2010. Reflections on manner/result complementarity. In Edit Doron, Malka Rappaport Hovav & Ivy Sichel (eds.), Syntax, lexical semantics, and event structure, 21–38. Oxford: Oxford University Press. DOI: https://doi.org/10.1093/acprof:oso/9780199544325.003.0002
Rapp, Irene. 1996. Zustand? Passiv? Überlegungen zum sogenannten “Zustandspassiv”. Zeitschrift für Sprachwissenschaft 15(2). 231–265. DOI: https://doi.org/10.1515/zfsw.19188.8.131.52
Rathert, Monika. 2000. Morpho-phonology of the past participle in German: where is the place of ge-? In Lew N. Zybatow (ed.), Europa der Sprachen: Sprachkompetenz – Mehrsprachigkeit – Translation, 109–114. Frankfurt a. M.: Peter Lang.
Roßdeutscher, Antje. 2016. Scale based particles and prefixes in German deadjectival verbs. Handout presented at the Workshop on Aspect and Argument Structure of Adjectives and Participles (WAASAP III), University of Lille III.
Roßdeutscher, Antje & Hans Kamp. 2010. Syntactic and semantic constraints on the formation and interpretation of ung-Nouns. In Artemis Alexiadou & Monika Rathert (eds.), Nominalisations across languages and frameworks, 169–214. Berlin: de Gruyter Mouton. DOI: https://doi.org/10.1515/9783110226546.169
Welke, Klaus. 2007. Das Zustandspassiv: Pragmatische Beschränkungen und Regelkonflikte. Zeitschrift für germanistische Linguistik 35. 115–145. DOI: https://doi.org/10.1515/ZGL.2007.006
Wunderlich, Dieter. 1987. An investigation of lexical composition: The case of German be-verbs. Linguistics 25. 283–331. DOI: https://doi.org/10.1515/ling.19184.108.40.2063