What is the gradeability of a vehicle

The city ring is busier than the motorway. Comments on the gradability of the passive state

Transcript

1 The city ring is busier than the motorway. Remarks on the gradability of the passive state Helga Gese and Vera Hohaus (Eberhard Karls University Tübingen) preliminary version, as of: April 5, 2011 Summary The present article is devoted to the passive state from the perspective of gradability. It provides evidence for the thesis that grading is systematically available in the passive state, as is to be expected in a copula-adjective construction, but that it is linked to clear conditions: Grading is possible as long as the property denoted by the passive participle of a state passive provides a degree. This degree can be provided in three different ways, all of which are based on one basic premise: the semantic underdetermination of the state passive. The analysis combines an event-semantic consideration of the passive state with a degree semantic analysis of the comparison and is thus able to combine two independently well-founded models and make them fruitful for a uniform analysis of compared passive states. 1. Introduction 2. Semantics and pragmatics of the passive state 2.1 The underdetermination of the attributed property 2.2 The genericity of the basic verb event 2.3 The adjective affix 3. Semantics of the comparison 4. Gradability of the passive state 4.1 Assignment of Q with a lexically given result state 4.2 Assignment of Q with an ad hoc property 4.3 Allocation of Q with an ad hoc property of the event type: Frequency 5. Concluding remarks 6. Bibliography 1. Introduction The present article is based on two basic assumptions, which justify and question its inclusion in an anthology on non-inflectional parts of speech put: He regards comparison as a case of inflection and the passive state as an adjective construction. As such, the passive state should be * This article is part of a cooperation between the DFG projects A1 Combinatorial Variation of Meaning at the Semantics / Pragmatics Interface (Claudia Maienborn) and C1 Parametric Variation in the Interpretative Component of Grammar (Sigrid Beck) in the Tübingen Collaborative Research Center 833 Constitution of Meaning originated. We would like to thank Nadine Bade, Sigrid Beck, Sebastian Bücking, Frauke Buscher, Lucas Champollion, Andreas Konietzko, Claudia Maienborn, Simone Regazzoni, Britta Stolterfoht and Sonja Tiemann for stimulating discussions and helpful comments. 1

2 have the inflection feature of the comparison so it should be inflectable. The article examines to what extent this is actually the case. He considers state passive clauses like under (1) and (3) and tries to clarify under which conditions these, like in (2) and (4), can be compared and how the semantics of these compared state passives can be derived. (1) The window is open. (2) The more open the window, the more pollen comes in. (3) The road is busy. (4) The road is more traveled than the motorway. As is usual in the more recent discussion about the grammatical classification of the passive state, we understand passive state constructions like under (1) and (3) not as a (verbal) passive, but as an adjectival construction in which the copula sein with an adjective participle II connects (see, inter alia, Rapp 1997, Maienborn 2007). The term state passive has meanwhile been overtaken by the copula adjective analysis, but is mostly retained as a pure label for reasons of recognition, as is the case here. In addition to psycholinguistic evidence (Stolterfoht et al. 2010), a number of adjective-typical properties, such as e.g. coordination with primary adjectives in (5), adjectival word formation patterns as in (6), un-affigierung in (7) and gradeability in (8) (see e.g. Maienborn 2007 and Gese et al for a more detailed discussion). (5) The vehicle is clean and polished. () (6) The Stuttgart star architect is award-winning and hostile. (Der Spiegel 16/1993,) (7) My hair is uncombed and my shirt collar is not ironed. (COSMAS M89 /) (8) The state parliaments are most threatened. (COSMAS V98 / MAY 21411) 2

3 If we value comparison as an inherent inflectional category of adjectives, as indicated 1 above (cf. Stump 1981: 31), it seems problematic for an adjective analysis that graded state passives as in (8) only occur to a limited extent. Contrary to the analysis widespread today, some authors therefore take the (assumed 2) ungrammaticality or marginality of graded state passives as in (9) or (10) as a reason to stick to an analysis as gender verbi (Pittner and Berman 2004: 73; Zifonun et al. 1997 : 1822f.). (9) * My dishes are more washed than yours. (10) ??? The FDP is more voted out than the CDU. (11) * This door is more open than the other. (Pittner and Berman 2004: 73) Under this assumption, the grading in the passive state would have to be seen as an exception that should only appear when the participle is lexicalized as an adjective. Such an analysis is ruled out in the case of spontaneous formations as in (12) and does not appear plausible even in the case of productive formations such as in (13) and (14). (12) ... Author who was much more famous and, above all, much more read than Goethe ... (Die Zeit,; Maienborn 2007: 93) (13) Nowhere in the world are children and young people more protected and guarded than in the USA . () (14) The more open the aperture, the higher the incidence of light. () Another approach is taken by Kratzer (2000: 14), who evaluates differences in the gradeability as an indication of a homonymy of the passive state. There non-gradability serves as (further) evidence for the existence of a perfect-like and thus less adjectival, second reading of the passive state (resultant state reading) 3. However, 1 We therefore share the assessment that the comparison is a type of inflection . As arguments for this, Eisenberg (2006:) cites, among other things, the great regularity of the formation of forms and the negligible tendencies towards lexicalization. He also points out: The non-comparability of individual adjective classes such as color and form adjectives, which is repeatedly mentioned, does not catch on. Comparatives we redder, square and even married or impossible are not formally excluded. They are also well-formed morphologically and violate semantic ... restrictions at best. See also section 3. 2 The rating of (11) as ungrammatic, which originated from Pittner and Berman, cannot be maintained in view of fully acceptable and frequent evidence such as (14). 3 Kratzer's assumption of two readings is based primarily on the combinatorial behavior of the passive state. In their analysis, the combinability with still as well as with a for -pp speaks for the availability of a target state reading (indicating a reversible result state). State passives that do not allow this modification, on the other hand, can only have one (the basically infinite post-state of a 3rd

4, the non-gradability of this resultant state reading is only postulated and not supported by data. The following questions arise, to which the present essay would like to contribute: Is grading only possible in marked exceptions (in which a lexicalization of the participle as an adjective can be assumed)? If not: What are the conditions for grading in the passive state and how do they relate to the semantics and pragmatics of the passive state? To answer these questions, the most important assumptions on the semantics and pragmatics of the passive state are introduced in Section 2. This is followed by a brief introduction to degree semantics in Section 3. Section 4 forms the core of this essay and presents the proposed analysis of comparison in the passive state. Concluding remarks can be found in the section on semantics and pragmatics of the passive state. After the earlier discussion about the grammatical status of the passive state (see the overview in Maienborn 2007), the more recent literature on the passive state focuses on semantic aspects. Based on the basic assumption that the passive state has adjectival semantics and must therefore be interpreted as an attribution of a property to the subject referent, there are now questions about the status of the ascribed property (Maienborn 2009) and the accessibility of the basic verbal event on which this property is based (Gehrke 2011, Gese 2011 ) in the focus of the investigations. For the analysis of graded state passives, two results of the more recent discussion appear to us to be decisive: (i) the underdetermination of the ascribed property and (ii) the genericity of the basic verb event. These two points will now be presented as the basis for the following investigations. The underdetermination of the attributed property If the state passive, as described above, is grammatically classified as a copula-adjective construction, this means that it must also be analyzed semantically as such. Just like the copula adjective sentence with the primary adjective in (15), a state passive sentence like (16) ascribes a property to its subject referent (cf. Maienborn defining a completed event) resultant state reading. For a criticism of this approach see e.g. Maienborn (2007, 2009). 4th

5 2003, 2005 and Engelberg 2005 on the semantics of the copula). 4 (15) The car is blue. s [blue (das-car) (s)] There is a state in which the car is blue. 5 (16) The car is repaired. s [repaired (the-car) (s)] There is a state in which the car is repaired. In contrast to a copula adjective construction with a primary adjective, the property assigned by a state passive clause is complex. It has to be derived from the base verb. This development of a salient property from the underlying basic verb is, according to the thesis of Maienborn (2009), a pragmatic process that is subject to only one semantic restriction: the property must result from the basic verb event, but is otherwise not specified. The assumed semantic underdetermination means, on the one hand, that the property does not have to be given lexically by the basic verb. In other words: There are also state passives with non-resultant basic verbs such as e.g. flatter in (17) and caress in (18). Sometimes these formations require contextual support, e.g. by a so-called job-isdone reading as in (18) (cf. Maienborn 2009, 2007; Kratzer 2000; Rapp 1997) or by a subject who makes a resultant reading appear plausible (18) (cf. Gese et al. 2009). (17) He was flattered when the personality cult around him finally took on grotesque forms. (Der Spiegel 40/1994,) (18) ??? The cat has been petted. (18) Anna has fulfilled her neighbor's duties: the mailbox has been emptied, the flowers have been watered and the cat has been petted. (Maienborn 2009: 36) (18) My soul is caressed. (Gese et al. 2009: 136) On the other hand, semantic underdetermination of the ascribed property means that 4 For reasons of readability, we assume a simplified state analysis in this article and do not go into details on the ontological type of the state argument (but see Maienborn 2003, 2005) . One consequence of this simplification is that a primary adjective like blue does not describe a property, but a relation between an individual and a state. In the following, however, we will continue to speak of the underdetermination of a property. 5 We assume as semantic types for objects, for degrees, for states, for events and for truth values. 5

6 even in the event that the basic verb has a lexically given result state, this can be used to assign the property, but it does not have to. It can also be developed in a purely pragmatic, ad hoc manner. Here it applies that the ascribed property must meet pragmatic optimality requirements. In other words, the context has to justify the effort of ad hoc property attribution, because the most economical / best specification ... is the one that gets by with the fewest contextually unlicensed additional assumptions. If, as in the case of resultative verbs, a corresponding property has already been introduced lexically, then this is a preferred candidate. (Maienborn 2011: 8) However, if the context does not license such an assignment through a lexically given result state, an ad hoc attribute attribution is possible (see Barsalou 1983, 1991, 1992, 2005 on ad hoc categories). Such a contextually triggered ad hoc property can be found, for example, in (19): Here the subject referent is not assigned the lexically given result state of open, namely open, since the sentence is also interpreted as true when the game is closed. The ascribed ad hoc property is here rather not in its original state. The same applies to (20) to (22), see the possible ad hoc interpretation given under the sentence. (19) The game is open, but not played with and absolutely as good as new. (Maienborn 2011: 9) s [not-in-original-state (das-game) (s)] (20) I was worried about how the Japanese would find the Oktoberfest, but it turned out that he was trained in Sweden. (Maienborn 2009: 42) s [drinking festival (der-japaner) (s)] (21) Mappus is switched off! () s [failed-at-the-atomic debate (Mappus) (s)] (22) Now one likes to be totally irradiated. The days when only bulky pocket calculators could be operated with solar energy are long gone. Today solar cells are the energy suppliers for the mobile lifestyle from the solar-powered laptop bag for charging the notebook on the go to the brassiere with solar module. (WELT ONLINE,) That the conclusion on the respectively attributable property refers to the world knowledge available in the respective context becomes particularly clear in the example (21). The occupancy sets the defeat of Stefan Mappus in the Baden-Württemberg state elections and the short-term moratorium on the operation of nuclear power plants as the 6th

7 known ahead. In the case of (22) the occasional character is so strong that it is difficult to fix the ad hoc property to a certain predicate. The important point here is that the underdetermined semantics of the state passive make such occasional formations possible: If the state passive were determined to be attributed to a lexically coded property, formations such as (22) would not be possible Recently, another characteristic of the passive state has attracted the interest of research (Gehrke 2011, Gese 2011): the genericity of the event on which this property is based. The sentence in (23) illustrates both features: On the one hand, the lexically given result state, invisible, is ascribed here, but the contextually triggered property inherently-present, on the other hand, modification always shows that the underlying event is not a spatiotemporal event Is a particular event. (23) The sculptor, who was born in Esslingen in 1944, produces figures from the stone that he believes have always been hidden there. () s [inherent-in-stone-present (figures) (s)] The acceptability of always-modification speaks here for the fact that the actual realization of the basic verb event in the passive state, even if pragmatically strongly preferred, is semantically not required: A The hide event did not even take place here. In addition to the modification by always, there are other indications that speak against an instantiation of the basic verb event in the passive state (cf. Gehrke 2011 and Gese 2011 for a more detailed discussion). In the case of the passive state, the attribute ascription, but not the event on which it is based, can be localized temporally. It is e.g. not possible to specify the time of the opening event by a sentence like (24). (24) * The window opened three hours ago. The implicit event can therefore not be a particular event localized in space and time. Such a particular event is also not available for the passive state in terms of discourse structure, as (25) shows. In (25) and (26) the connection of the weak bridging pronoun 7 requires

8 that is the anaphoric assignment through the assumed implicit event in the passive state (cf. Asher & Lascarides 1998, Härtl 2003 and Härtl 2008 on weak bridging pronouns). Bridging to the implicit event is only possible here if this is not interpreted as a spatiotemporal particular event as in (25), but as a type of event as in (26). 6 (25) Peter's upper arms are tattooed. ??? that hurt a lot. (26) Peter's upper arms are tattooed. This is a bad habit among delinquent young people! The property denoted by the participle of a state passive does not depend on a specific spatiotemporal event, but on a type of event. But what does this mean? Intuitively speaking it means that a particular event of a certain kind in a given context usually brings about a certain state. The relationship between the type of event and the state is not a classic causal relationship between cause and result.Gese (2011) opts for a non-classical causal dependency relation introduced by Kim (1990) into the philosophy of mind, the supervenience relation. In addition to the possibility of combining a type of event with a state that depends on it, another property speaks in favor of supervenience 7: While in a classic-causal dependency relation the result is fully determined by its cause, this does not apply to supervenience: in a supervenience -Relation, the supervening property is only partially determined by its basis. Lexical supervenience, hereinafter referred to as LSV for short, leaves room for the specific ad hoc character of the passive state Basic verbs speak for the semantic underdetermination of the passive state. (ii) The (limited) acceptability of always-modification, the impossibility of the temporal localization of the base verb event and the impossibility of the anaphoric revival of the 6 expressions such as is a bad habit or is a bad habit are commonly assessed as expressions referring to species (see Krifka et al 1995). These enable bridging to the implicit event type. 7 supervenience also enables e.g. non-classical temporal references between the two arguments (e.g. simultaneity or tripod causes) and it does not form causal chains. 8th

9 basic verb event through the bridging pronoun dasenken against the assumption of an implicit particular event in the semantics of the passive state. Successful bridging to a generically interpreted event, on the other hand, suggests that the basic verb event involved in the state passive is to be interpreted generically, i.e. it is a type of event. Accordingly, Gese (2011) argues for an analysis of the passive state, which combines Maienborn's underdetermination analysis with the assumption of event type reference in the passive state. The analysis proposed there assumes the adjectivation of the verbal participle by the zero affix in (27), whereby LSV stands for the non-classical causal supervenience relation described above and is an operator for the formation of species. 8 (27) Ø affix: λp . λx . λs . [q (x) (s) & LSV (P) (s)] The adjective affix in (27) says that the passive state of its subject referent x in state s is a semantically underdetermined property, here captured by the free variable Q, ascribes. Q requires a contextually salient property to be assigned, whereby the only semantic restriction on Q is that it is dependent on the event type P derived from the base verb. The structure of a state passive clause like (1) repeated as (28) is given in (29), its semantics in (30). In the case of (28), the property in (31) would be a plausible occupancy candidate for Q. (28) would then be assigned the interpretation in (32). (28) The window is open. (29) (30) s [Q (x) (s) & LSV ((λe.open (e))) (s)] 8 P is Links (1995: 376) notation for an up operator, the predicates in Species referencing expressions converted: For instance if TIGER is a one-place predicate denoting the set of tigers TIGER is a singular term that denotes the kind Tiger. For a similar operator see e.g. Chierchia

10 There is a state s, which depends on the type of open event, in which Q applies to the window. (31) λx . λs . open (x) (s) (32) s [open (das-window) (s) & LSV ((λe.open (e))) (s)] Under (32) the sentence in (28) is exactly then true if there is a state s, dependent on an open event type, in which the window is open. As an interim conclusion it should be noted: The state passive expresses an attribution of properties to its subject referent. The special thing about the attributed property is that it is semantically underdetermined and depends on a type of event. Both peculiarities, the semantic underdetermination and the event type reference, will be important in the derivation of the comparative semantics of the passive state. Before we can get to this, however, the comparative semantics on which this is based should be briefly introduced below. 3. Semantics of the comparison At the beginning of this section there is a brief introduction to the widely accepted semantics of comparison constructions, enriched with states. The introduction is based on the discussion in von Stechow (1984a, b) and its modernized presentation in Beck (2011). 9 Intuitively, comparison is a linguistic means of expressing degrees. A comparative as in (33) seems to establish a relation between two degrees, namely the degree to which Sonja is tall and the degree to which Anna is tall. (33) Sonja is taller than Anna. The degree to which Sonja is tall exceeds the degree to which Anna is tall. This intuition is reflected in the assumption of a semantic type for degrees. Degrees are abstract entities and elements of scales (cf. von Stechow 1984a: 47). The domain D is the union of all disjoint sets of degrees, including e.g. Grades of size, grades of weight, grades of temperature, etc. Each set has its own relative order. Elements of the set of degrees of weight would be about seven kilograms and two grams. Degrees are derived from degree adjectives such as e.g. large in 9 An up-to-date overview of the semantics of comparison constructions, which does not require a background in formal semantics, is offered by Schwarzschild (2008). 10

11 (34) introduced into the semantics and are of the type >>. (34)! [[Large]] = λd . λx . λs . SIZE (x) (s) d = λd . λx . λs . x is d-size in s In (34), SIZE refers to a measure function. Measure functions assign a degree to an individual. For example, SIZE (x) (s) is the maximum degree to which the individual x in state s is large. The comparative essentially expresses a relationship between two degrees, see the lexical entry for the comparative morpheme in (35). It is one of the elements in German that quantifies through degrees. The maximality operator used in (35) is as defined in (36). He selects the maximum degree contained therein from a set of degrees. (35)! [[-Er]] "= λd . Λd . Max (d)> d (36)! [[Max]] = λd . Ιd [D (d) & d [D (d) dd]] In the following, the comparison with a degree in (37) is discussed as an example. 10 It is based on the structure in (38). (37) Anna is taller than 1.72 meters 10 The comparison standard can, however, also be made available through the context, see (37). (37) (Sonja is 1.72 meters tall.) Anna is taller. Also in the positive (eg Britta is tall.) Is the Comparison not explicit. The degree argument is tied here by a covered positive operator, which is in complementary distribution to the comparative morphology (cf. von Stechow (2007)). The unmarked form of the adjective is used to make a vague or context-dependent statement about the degree , to which the property that the adjective describes applies to an individual (cf. Beck et al. 2009: 12)

12 (38) The als-phrase in (38) is a sister of the comparative morpheme. Their position on the surface is achieved through extra position. The constituent consisting of the comparative morpheme and the als-phrase forms a degree phrase which takes the position of the specifier of the adjective phrase (cf. Heim 2001). Like the inflection categories tense and aspect neglected in the present study, comparison is realized as a functional head with its own maximum projection (cf. also Stump 1981: 36-38). The as-constituent is the first argument of the comparative morphem at the time of the compositional interpretation; see the logical form in (39). The as-phrase denotes (under the simplest possible assumption) a degree, namely 1.72 meters. The main clause denotation is a set of degrees. This interpretation is generated by raising the quantifier of the degree phrase, which is a generalized quantifier of degrees (cf. Heim 2001). Under this analysis, the sentence in (37) has the truth conditions in (40), so it is true if and only if the maximum degree of height that Anna reaches exceeds 1.72 meters. 12th

13 (39) (40) max (λd. S [height (anna) (s) d])> 1.72 m The maximum degree to which Anna is tall is greater than 1.72 meters. The analysis of the comparative presented above can easily be transferred to the differential in (41). It is precisely this transferability that is one of the strengths of the degree semantic analysis of the comparative (cf. von Stechow 1984b:; Beck et al. 2009: 10) and will still be important for the derivation of comparative state passives with basic verbs such as warm or enlarge. It is assumed that the comparative morpheme in (42) provides another place for a degree argument. This can then be substantiated with the indication of the difference, here ten centimeters. The example in (41) receives the interpretation in (43). (41) (Frauke is 1.70 meters tall.) Britta is ten centimeters taller. (42) [[-er diff. ]] = λd . λd . λd . max (d) d + d (43) max (λd. s [britta is d-large in s]) d context + 10 cm The maximum degree to which Britta is large is equal to the contextually provided degree plus ten centimeters. The most important features of this analysis of the semantics of comparative constructions are summarized: (i) Comparative constructions in German use degrees for the semantics. (ii) Degree predicates introduce degree arguments into the syntax. (iii) The comparative is one of the operators that quantify over degrees. 4. Gradeability of the passive state As already mentioned in the introduction, the gradeability of the passive state is one of the arguments for the adjective analysis of the construction. Nevertheless, graded 13

14 state passive forms only to a limited extent. Grading is not only possible in marked or lexicalized exceptions, according to the hypothesis of this article, but is systematically available. However, availability is also linked to a clearly comprehensible condition: The underdetermined property Q introduced in the process of adjectivation must be assigned a gradable property, whereby this must be the preferred assignment candidate in the context. It will be shown that there are three possibilities for this: (i) (ii) Q is assigned the lexically given result state of the base verb. Q is evidenced by the formation of an ad hoc property. (iii) Q is assigned a systematically available ad hoc property of the type of event, the frequency of its realization. In all three cases, so our argumentation, grading is possible. Let us first come to the first case, on the basis of which the basic analysis for the other cases is developed at the same time. 4.1 Assignment of Q with a lexically given result state In the examples in (44a) to (48a), a gradable result state is made available from the event denoted by the basic verb. The free variable Q is usually assigned this degree predicate as long as the context does not trigger any other assignment. Grading is then normally possible, as (44b) to (48b) show. (44) a. Air and water are polluted. b. The air and water in Texas are more polluted than any other state. (Der Spiegel 30/2000,) (45) a. Girls are threatened. b. AIDS threatens to become an epidemic, especially among young women. Girls are more easily infected and more at risk than boys. That said Suman Mehta, AIDS coordinator of the United Nations Population Fund, the largest family planning organization. (dpa,) (46) a. The colleagues are sensitized. b. The colleagues are more sensitized than before thanks to drug detection courses. (Norddeutsche Rundschau,) (47) 14

15 a. The earth is loosened. b. The looser the soil, the better it binds nutrients and prevents their drainage. (Zeit Online,) (48) a. The hair is protected from heat damage. b. [from a hair dryer comparison:] Professional [...] hairdryers have a very high-throughput blower. Drying is faster and the hair is more protected from heat damage. () In the case of (48), for example, the degree predicate is available in (49). The measure function PROTECTION maps an object x to the degree to which it is protected in a state s. The sentence is then based on the already known structure of the adjective phrase in (50) and the participle is interpreted as in (51). (49) λd . λx . λs . PROTECTION (x) (s) d = λd . λx . λs . x is protected to degree d (50) (51) λd . λx . λs . [SCHUTZ (x) (s) d & LSV ((λe.schützen (e))) (s)] The adjective affix maps the semantic type of Q to the adjective and thus also provides the degree argument for further composition ; in this case see (52). This expansion is due to the enrichment of our ontology by degrees, which had no relevance in the previous literature on the passive state. (52) Ø affix: λp . λd . λx . λs . [Q (d) (x) (s) & LSV (P) (s)] Increasing the quantifier corrects the conflict between the semantic types in (50). The logical form is sketched in (53). If we assume the assignment of Q in (49), the sentence is interpreted as in (54). 15th

16 (53) (54) max (λd. S [SCHUTZ (hair) (s) d & LSV ((λe.schützen (e))) (s)])> d context The maximum degree d such that there is a state dependent on a Sagittarius event type in which the hairs are protected to degree d is greater than the degree provided contextually. It should be pointed out at this point that the free variable Q is also assigned the assignment in (49) in the case of (48b) and that the adjective formed has a degree argument, see (51). This is bound in the unmarked form, the positive, by a covered positive operator (see section 3, footnote 10 and the detailed discussion in von Stechow (2007)). Our example in (48b) would be interpreted to the effect that the degree to which the hair is protected is above the degree from which the hair is considered to be protected in the context. Interestingly, the result state made available by the basic verb of a passive state clause, which is used for the property assignment, can also be a differential, see the examples in (55) to (57). (55) The warmer the blood, the more difficult it could be to allow the O 2 to diffuse to the point of consumption. () (56) The boys' critical eye is sharper than in any previous epoch. (W. Berges et al. (Ed.): Zur Geschichte und Problematik der Demokratie) (57) One side is also significantly larger than the other and it hurts when you press it. () What is compared here is not the degree that can be ascribed to an individual, but the degree by which the individual differs from the degree of comparison. In the case of (57) the free variable Q is assigned as in (58). The interpretation of the sentence and a paraphrase of the truth conditions are given in (59). 16

17 (58) λd . λx . λs . SIZE (x) (s) d originally + d = λd . λx . λs . x is in the current state s by d larger than in the original state (59) max (λd. s [SIZE (page 1) (s) d originally + d & LSV ((λe.enlarge (e))) (s) ])> max (λd. s [SIZE (page 2) (s) d originally + d & LSV ((λe.enlarge (e))) (s)]) The maximum degree by which one side is enlarged , is greater than the maximum degree by which the other side is enlarged. The cases of compared passive statuses described in this section could also be captured by a passive status analysis that, e.g. Kratzer (2000) does not start from the semantic underdetermination of the property assigned by a passive state. It is assumed there that state passives that have a lexically given result state (target state) should in principle be gradable: On the current analysis, target state participants have state arguments, just like other adjectives. It is therefore not surpising that they behave like underived adjectives with respect to gradability and degree modifiers ... (Kratzer 2000: 10). However, the analysis of the data presented in the following section is then not obvious. 4.2 Assignment of Q with an ad hoc property In contrast to the cases discussed in the previous section, in which the base verb provides a gradable result state suitable for the assignment of Q, the base verbs in (60) to (65) provide a result state that describes an absolute property. Such a property cannot be graded (see e.g. Bierwisch 1984, 1987), since it does not introduce any degrees into the semantics, as is exemplarily shown in the assignment of Q in (64b) and (65b). This ungrammaticality of the comparative can also be observed in copular constructions with primary adjectives, see (66). (60) * My husband is more civil servant than me. (61) * The task is more completed than any other. (62) * The cardiac arrest occurred more than feared. (63) * The hotplate is switched off more than the oven. (64) a. * The light is more off than before. b. λx . λs . from (x) (s) 17

18 (65) a. * The prey is more shot than it was an hour ago. b. λx . λs . dead (x) (s) (66) a. The girl was mute from birth (* - er). b. λx . λs .mute (x) (s) As described in Section 2, Q is preferably given with the result state lexically given by the base verb for economic reasons. However, an alternative assignment of the free variable through a contextually salient ad hoc property should in principle always be possible. There should therefore be compared passive state passives with basic verbs that have a lexically given non-gradable result state. This is actually the case, as (67) and (68) show. In this case the derived ad hoc property introduces a degree. This is also the case with the following examples (69) (72) of compared ad hoc state passive formations (without lexically given result state). (67) a. The CDU couldn't be more voted out than in Hesse. () b. λd . λx . λs . FAILURE (x, s) = d = λd . λx . λs . x has a d-major failure (68) a. This time my brain is even more switched off than before. (board.nostale.de,) b. λd . λx . λs . ATTENTION (x, s) = d = λd . λx . λs . x is d-attentive (69) The more Swedish-trained someone is, the more fun they will have at the Oktoberfest. (70) Roche is more contaminated than a Castor. () (71) The state is more tamed than society is mobilized. (Der Spiegel 22/1989,) (72) What he wanted to express besides his well-known "humor" opportunism is not accessible to me. But maybe others are smarter and more tried and tested. (Friday) 18

19 Overall, it makes no difference for the gradability of the passive state whether the underdetermined property Q is assigned the result state of the basic verb or a contextually triggered ad hoc property. The only decisive factor is whether Q introduces grades or not. 4.3 Assignment of Q with an ad hoc property of the event type: Frequency The examples given in the last section are partly of a strongly occasional character. The formation of an ad hoc property strongly depends on the respective context, which provides a salient gradable property for the assignment of Q. The grading behavior is then, however, as has been shown, completely regular. We would like to discuss another case in this section. This is also about the attribution of a gradable ad hoc quality to the subject referent. The difference to the cases discussed in Section 4.2 is that the ad hoc property we are concerned with here represents a property that is systematically available for event types, namely the quantity or frequency of the event type realizations. As shown in Section 2, the passive state can in principle also be used with non-resultative basic verbs, e.g. the activity verbs in (73). Such formations usually require licensing through a context that provides a salient characteristic for the assignment of Q. A typical context that licenses the formation of state passives with basic activity verbs is the job-is-done context described in Section 2.1: (73) [from an Internet forum about a game with virtual pets, in which they are looked after as well as possible must:] My horse is groomed, rabbits are petted but not trained and my hamsters are played! () In the literature on state passive, Kratzer (2000) assumed that job-is-done readings should be analyzed as less adjectival. In this way, they do not represent the assumption that the subject referent is attributed to a property, but rather have a perfect-like meaning (resultant state) in which the event is only localized temporally. Due to the perfective semantics, Kratzer (2000: 14) must assume that resultant state passives are not gradable: Resultant state participants are expected to be less adjective-like than target state participants under the current analysis. This seems to be so, 19

20 given that resultant state participants are never gradable, for example. In Kratzer's analysis, all verbs that do not have a lexically given result state that can still be modified form only the resultant-state passive. The non-gradability postulated here in the so-called resultant state reading can, however, easily be refuted by looking at the compared state passives with activity base verbs in (74) to (77). (74) The scale of the training is correct, especially in order to better exercise the horse accordingly. The more let loose and the more ridden the horse, the more lasting joy and success are with horse and rider. () (75) Hardly any other taste topic is currently more discussed than the combination of wine and chocolate. () (76) Nothing is more discussed than the tripled dragon. (Karl August Böttiger: Small writings archaeological and antiquarian) (77) ... Author who was much more famous and above all much more read than Goethe ... (Die Zeit,, Maienborn 2007: 93) Also for the well-known example of the caressed Cat can easily construct a corresponding job-is-done context that licenses the grading of petted: (78) I have now really more than fulfilled my neighbor obligations: The refrigerator is more full than ever and the cat is petted more than she would like . The fundamental gradability of the passive state participles in (74) to (78) confirms that the passive state always has adjectival semantics, even with non-resultative basic verbs. A homonymy solution, as advocated by Kratzer (2000), can thus be ruled out, and the differentiation of readings in the passive state must be treated as a purely pragmatic matter. But what specification does Q experience in the examples in (74) to (78)? Let us consider the following examples: (79) The city ring is more busy than the motorway. Tens of thousands of cars drive over it every day. (80) No topic is currently more debated than Fukushima. There are several special programs on this every day. (81) Saint Exupéry is better read than J.D. Salinger. It already has over 80 million readers, while Salinger only has 65 million. 20th

21 In (79) to (81) it becomes clear that the comparison here focuses on the frequency of the event type realizations. The property used to specify the free variable Q relates accordingly to the frequency of driving on the ring road compared to the motorway. In (80) it concerns the frequency of discussing Fukushima and in (81) that of reading Saint Exupéry compared to J.D. Saligner. The assignment of Q is given as an example for (79) in (82). (82) λd . λx . λs . x is read d often. However, such an assignment is also subject to the above-mentioned optimality requirement for Q: It must therefore be salient in the context. In the case of state passives with basic activity verbs, however, there are usually hardly any competitors for the assignment of Q. The frequency of the realization of the event type seems to be one of the few informative properties here, which explains the prominence of the frequency interpretation of the comparative in these cases. In principle, however, the reading is available to all compared state passives if it delivers a property that is salient in the context, see the examples in (83) and (84) with resultative base verbs. (83) Hyeyoon Park is even more award-winning than Vilde Fang. It has already received six major awards. (84) [about the soccer player Ebbe Sand, top scorer of the Bundesliga 2001]: Hardly any formula is more proven than the following: sand + ball = goal. [Sand has proven it 22 times this season.] () If the frequency interpretation is successful, the subject referent x is assigned a complex property, namely the frequency of the event type realizations, where x is the subject of these events. The intuition on which the assignment of Q given in (82) is based is that FREQUENCY does not operate on particular, i.e. spatiotemporally localized, events. FREQUENCY operates over an event type. The function assigns a degree to this type of event, namely the frequency of its realization (cf. Bücking 2010: 3-4 for a corresponding interpretation of occasionally). Since the passive state, as explained in Section 2.2, refers to an implicit event type, this is of course available for the formation of the ad hoc property. This explains the principle possibility 21

22 a frequency reading of the comparative in the passive state. Conversely, the existence of this reading can be interpreted as a further argument for the event type reference in the passive state. 5. Concluding remarks The present article presents an investigation of graded state passives in German, the main concern of which is to work out the semantic and pragmatic conditions that are linked to the gradability of the state passive. The main finding of the article contradicts the title of the anthology: We have shown that the passive state is a construction with a regularly inflected (namely gradable) adjective participle in a predicative position. But hopefully this realization also deserves its place in the discussion about (non-) flexibility. Our results can be summarized as follows: In the passive state, grading is not limited to marked or lexicalized exceptions, but is systematically available. The availability of the gradation in the passive state is based on two basic semantic properties of the passive state: (i) the semantic underdetermination of the property Q denoted by the participle of a passive state clause; and (ii) the state passive event type reference. The acceptability of compared state passives is linked to a clearly comprehensible condition: The preferred assignment for Q in the context must have a degree argument. Through the adjectivation, the degree argument is transferred into the semantics of the participle and can be called up in the further composition. Only this availability of degrees in the argument structure of the adjective enables the comparison. A total of three possibilities for the assignment for Q were examined: (i) A lexically provided, gradable result state can be selected for the assignment for Q. (ii) A contextually salient, gradable ad hoc characteristic can be formed. (iii) It can be a systematically available ad hoc property of the type of event derived from the basic verb: the frequency of its realization. 22nd

23 It was important to us to establish connections between current event-semantic-oriented investigations of the passive state and degree semantic work, connections that open up perspectives for further research projects. 23

24 6. Bibliography Asher, Nicholas and Alex Lascarides (1998). Bridging. Journal of Semantics 15.1, S Barsalou, Lawrence W. (1983). Ad hoc categories. Memory and Cognition 11, S Barsalou, Lawrence W. (1991). Deriving Categories to Achieve Goals. In: The Psychology of Learning and Motivation: Advances in Research and Theory. Edited by Gordon H. Bower. San Diego: Academic Press, S Barsalou, Lawrence W. (1992). Frames, Concepts, and Conceptual Fields. In: Frames, Fields, and Contrasts: New Essays in Semantic and Lexical Organization. Edited by Eva Feder Kittay and Adrienne Lehrer. Hillsdale: Lawrence Erlbaum, S Barsalou, Lawrence W. (2005). Situated conceptualization. In: Handbook of Categorization in Cognitive Science. Edited by Henri Cohen and Claire Lefebvre. St. Louis: Elsevier, S Beck, Sigrid (2011). Comparison Constructions. Appears in: Semantics: An International Handbook of Natural Language Meaning, Vol. 2. Ed. By Claudia Maienborn, Klaus von Heusinger and Paul Portner. Berlin: De Gruyter. Beck, Sigrid; Daniel Fleischer, Remus Gergel, Stefan Hofstetter, Svetlana Krasikova, Christiane Savelsberg and others (2009). Crosslinguistic Variation in Comparison Constructions. Linguistic Variation Yearbook 9, S Bierwisch, Manfred (1984). Dimensional adjectives: semantic structure and conceptual interpretation. Journal for Phonetics, Linguistics and Communication Research 37, S,, Bierwisch, Manfred (1987). Graduation semantics. In: Grammatical and Conceptual Aspects of Dimensional Adjectives. Edited by Manfred Bierwisch and Ewald Lang. Berlin: Springer, S Bücking, Sebastian (2010). Modification by frequency adjectives in German. Handout, workshop accessibility of implicit events, Eberhard-Karls University of Tübingen. July Chierchia, Gennaro (1998). Reference to Kinds across Languages. Natural Language Semantics 6.4, S Eisenberg, Peter (). Floor plan of the German grammar, Vol. 2, Das Wort. Stuttgart: Verlag J.B. Slaughterer. Engelberg, Stefan (2005). Kimian States and the Grammar of Predicative Adjectives. Theoretical Linguistics 31, S Gehrke, Berit (2011). Tripods Passives and Event Kinds. Appears in: Proceedings of Meaning and Meaning 15. Gese, Helga; Claudia Maienborn and Britta Stolterfoht (2011). Adjectival Conversion of Unaccusatives in German. Journal of Germanic Linguistics 23.2, S Gese, Helga; Britta Stolterfoht and Claudia Maienborn (2009). Context Effects in the Formation of Adjectival Resultatives, in: The Fruits of Empirical Linguistics, Vol. 2, Product. Edited by Susanne Winkler and Sam Featherston. Berlin: De Gruyter, S Gese, Helga (2011). Events in Adjectival Passives. Appears in: Proceedings of Sinn and 24