-ble Adjectives and Middle Constructions : a Problem for Inheritance

-BLE ADJECTIVES AND MIDDLE CONSTRUCTIONS: A PROBLEM FOR INHERITANCE Llulsa Grkia Universitat de Girona The goal of this paper is to study the behavior of -ble adjectives with respect to the inheritance of verbal arguments. Even though -ble adjectives, like middle constructions, have a non-eventive character, they syntactically inherit the intemal arguments of the verbal root, with the most direct one becoming the subject. If it is assumed that syntactic inheritance is possible only when the construct is eventive (cf. Grimshaw (1990) and Picallo (1991) for nominalizations, aud Levin and Rappaport (1988) for -er nominals. among others), then -ble adjectives and middles are a problem for inheritance. We will suggest that the predicative value of adjectives and verbs (both [+VI, in contrast with the [-VI feature of nouns) can explain this aparent contradiction. 1. The Rule of -ble Affixation The rule that forms adjectives adding the suffix -ble to a verbal root has been studied extensively. Linguists genelally agre on the assumption that the suffix -ble requires a transitive verbal root. It explains the contrast between (la-c) and (ld-f): (1) a. adora+ble 'adorable' b. imagina+ble 'imaginable' c. destructi+ble 'destructible' d. *rondina+ble 'grumbleable' e. *neva+ble 'snowable' f. *sembla+ble 'seemable' Catalatt Workirlg Papers in Linguistics (CWPLJ 1992: 163-182. Universitat Authoma de Barcelona The effect of the suffixation rule on the verbal argument structure is double: (a) the object of the verb becomes the subject of the adjective,'and (b) the subject of the verb disappears. See the following examples: (2) a. recomanar: NP, - NP2 'recommend' b. recomanable: NP2 - (*by NP,) 'recommendable' (3) a. El professor va recomanar un llibre. The teacher recommended a book.' b. un Ili bre recomanable (*pel professor) 'a book recommendable (by the teacher)' In Williams' (1981) teminology, the intemal argument has been externalized, and the external argument is lost. Other linguists, as de Miguel (1986), do not define the effect of the rule in terms of the internalexternal argument distinction, nor in tems of syntactic function, but rather in terms of 8roles. She affirms that the subject of -ble adjectives in Spanish is always the Theme of the verb, independently of whether it is the subject or the object. So, the rule of -ble affixation requires that the 8grid of the verb have a Theme; the transitivity of the verb is not important, nor whether the Theme is projected or not as an interna1 argument. Her argument is based, fundamentally, on exarnples such as (4), where the Theme appears in subject position: By 'subject of the adjective' we mean the externalized argument, the NP fuctioning as the antecedent for the agreement features of the adjective. (4) a. El ferro 6s oxidable. 'Iron is oxidizable.' a'. x oxida el ferro 'n oxidize the iron' a". el ferro s'oxida 'iron oxidizes' b. La música 6s agradable. 'Music is pleasant.' b ' . la música agrada 'music pleases' (4a) has two interpretations: one, (&I), is causative, and the Theme is the direct object of the verb, and the other one, (4aU), is inchoative, and the Theme is the subje~t.~ In (4b) the Theme can only appear as the subject, but these Theme subjects come from a D-structure object position, following the traditional analysis of Burzio (1986) and Belletti and Rizzi (1988). So, word formation rules will "see" only that Themes are internal arguments because these mles are sensitive to lexical properties of words, but not to syntactic structures. We assume, for the time being, that the suffix -ble selects verbs with an internal argument. From the examples we have discussed, the two hypotheses seem to be identical, but we will see that the subject of a -ble adjective can be an internal argument bearing a 0-role other than Theme. Given that Themes are always projected in an interna1 argument position (cf. GrAcia (1989)), we can conclude that the rule of -ble affixation is sensitive to the type of argument, internal or extemal, but not to the 8role. 2 According to Roeper (1987303). adjectives like breukable, bounceable,fracturable, and singable, formed from ergative verbs, only have a transitive interpretation, i.e., they imply an implicit Agent. 2. The External Argument It is generally assumed that the external argument of the verb cannot be inherited by the adjective: (5) a. *un llibre recomanable pel professor 'a book recommendable by the teacher' b. *un alumne criticable pel mestre de física 'a student criticizable by the physics teacher' c. *una ciutat anihilable per l'enemic 'a city annihilatable by the enemy' De Miguel presents some examples where the presence of the Agent is acceptable: (6) a. un juguete fAcilmente manejable por 10s niños 'a toy easily handleable by children' b. (??)Pedra es superable por Juan. 'Pedro is surpassable by Juan.' In our opinion, (6b) is not very acceptable, but it becomes better if the by phrase is less specific than por Juan: (i) %ro es superable por cualquiera que haya obtenido el Graduado Escolar. 'Pedro is surpassable by anyone with a primary school certificate.' Indeed, (64 is acceptable, but if we change the Agent complement by a more specific one, or if we eliminate the adverb fkilmente 'easily', the sequence gets worse: (8) a. '??un juguete fácilmente manejable por mi hijo 'a toy easily handleable by my son' (8) b. ?'?un juguete manejable por 10s niños 'a toy handleable by children' c. '??un juguete manejable por mi hijo 'a toy handleable by my son' Observe that if we modify the examples in (5) accordingly, they get better: (9) a. ?un llibre fkilment recomanable per qualsevol que l'hagi llegit 'a book easily recommendable by anyone who's read it' b. ?un alumne fkilment criticable per qualsevol que hagi estat professor seu 'a student easily criticizable by anyone who's been his teacher' c. ?una ciutat kilment anihilable per qualsevol exkrcit minimarnent organitzat 'a city easily annihilatable by any minimally well organized army' Why do these adjectives behave in this way with respect to the Agent? We will analyze the meaning of these adjectives. We will leave aside the adjectives deriving from "intransitive verbs" (actually, inaccusatives) that have an active meaning, as those in (10): (10) a. agradable = que agrada 'pleasant' 'that pleases' b. variable = que varia 'variable' 'that varies' c. perdurable = que perdura 'perdurable' 'that perdures' The other adjectives have a passive or inchoative value, depending on whether the original verbal form is transitive (agentive) or intransitive (ergative). The verbal forn is generally pronominal in the latter case. In (1 1) we have some of the definitions for these adjectives in Fabra (1932): (11) I. quepotserv 'that can be V' a. justificable b. comunicable c. comparable d. aprofitable 11. queespotv that clitic-can V 'that can be V' a. reclamable b. qualificable c. reparable d. maniobrable e. oxidable f. mobilitzable 'justifiable' 'communicable' 'comparable' 'usable' 'claimable' 'qualifiable' 'repairable' 'handleable' 'oxidizable' 'mobilizable' From the data observed it appears that the adjectives defined with the form que pot ser V, those in (11 I), come from transitive verbs that do not have ergative counterparts. The adjectives defined with the paraphrase que es pot V, those in (1 1 11), do not form a homogeneous class: while (121, b) are transitivdergative, (12c, d, e) are purely transitive, (12) a. oxidable 'oxidizable' b. mobilitzable 'mobilizable' c. reclamable 'claimable' d. qualificable 'qualifiable' (12) e. reparable 'repairable' The paraphrase que es pot V can have two values: one is inchoative ('it can oxidize by itself) and the other one is passive, with an impersonal meaning ('one can oxidize it', 'it can be oxidized by one'). The inchoative value corresponds to ergative fonns. The second value resemble the paraphrase que pot ser V; the difference is that the pronominal form implies an impersonal meaning, but the periphrastic forn does not necessarily do it. However, the fact that there is no clear difference between the transitive verbs in (1 1 I) and those in (1 1 11) allows us to say that the impersonal or generic value is common in the two cases. That is, when the verbal form has an Agent, this Agent has a generic value in the adjectival forn. Problably, if the Agent (explicit or not) has a generic value, it is because these adjectives are attributed to a subject when it is generally true that the subject has this quality. That is, we cannot say that the behavior of a person is criticizable only because one person has a reason to criticize it; it will be criticizable if everyone can criticize it. This value is well reflected in Fabra (1932) when he describes some -ble adjectives that, in some way, implicitly carry a subjective valorative load that one wants to generalize or to objectivize: (13) I. a. lamentable digne de lamentaci6 'lamentable' 'worthy of lamentation' b. censurable digne de censura 'censurable' 'worthy of censure' c. admirable digne d1admiraci6 'admirable' 'worthy of admiration' d. adorable digne d'adoraci6 'adorable' 'worthy of adoration' e. respectable digne de respecte 'respectable' 'worthy of respect' (13) 11. a. amable digne de ser amat 'lovable' 'worthy of being loved' b. envejable digne de ser envejat 'enviable' 'worthy of being envied' c. enyorable digne de ser enyorat 'missable' 'worthy of being missed' d. publicable digne de ser publicat 'publishable' 'worthy of being published' All theses facts make constructions with a -ble adjective resemble middle constructions, which have a generic value as well; they describe a quality of the subject and are often modified by b& 'well' or fiicilment 'easily': (14) a. Els burbcrates se subornen ficilment. 'Bureaucrats bribe easily.' b. Aquests llibres es venen bé. These books se11 well.' c. Els teus articles no es llegeixen de bon grat. Lit.: 'Your papers don't read willingly.' Keyser and Roeper (1984384) affirm that middle constructions are often called "generic sentences" and that they are "( ...) state propositions that are held to be generally true. They do not describe particular events in time (...). Middle verbs do not refer to events." If we accept this similarity between -ble adjectives and middle constructions, we have a way of explaining the difficulty in having an Agent in adjectival constructions. Middles are very odd with an explicit Agent as well, but they get better when the Agent is generic or not very specific: (15) a. Els burkrates se subornen fAcilment ... 'Bureaucrates bribe easily ...' a'. *...pel senyor Pere Pi. '...by Mr. Pere Pi.' a". ??...per qualsevol que els ofereixi diners. '...by anyone who offers them money.' b. Aquests llibres es venen bé... These books sell well ...' b'. *...pel llibreter que viu a la cantonada '...by the bookseller who lives in the corner.' b" . ??...pels llibreters de vell. '...by used-book dealers.' c. Els teus articles no es llegeixen de bon grat ... Lit.: 'Your papers don't read willingly ...' c'. *...pel meu gema. '...by my brother.' c". ?...per ningú. '...by no one.' It has been argued that in middles the clitic se absorbs the Agent 9-role, so it cannot reappear as a PP.~ For -ble adjectives, a similar solution has been proposed (cf. Roeper (1987)): the suffix absorbs the Agent 8role. Notice that there is another class of adjectives, formed on a verbal root by the adjunction of the suffix -dar, that have the same value that -ble adjectives have: (16) a. lletra llegidora (*pel nen) 'handwriting readable (by the child)' b. minyona casadora (*pel capell&) 'housemaid marriable (by the priest)' Keyser and Roeper (1984406) propose an abstract clitic for English. parallel to romauce selsi. 171 (24) a. La introducci6 del nou producte al mercat (per part de I'empresa) va ser lenta. The introduction of the new product to the market (by the company) was slow.' b. El reparúment lla divisi6 de l'herkncia entre els familiars (per part del jutge) va ser difícil. The distribution I division of the inheritance among the relatives (by the judge) was difficult.' c. La transformaci6 del club en una S.A. (per part de l'equip directiu) va ser molt lapida. The t~ansformation of the club into a joint-stock company (by the management) was very quick.' If we accept that the origin of these derived nouns is syntactic and that in D-structure the verbal root is the head of a VP, we can explain why the noun maintains the argumental properties of the verb. 4. An Explanation We have seen that the ability of derived nouns to take a PP corresponding to an argument of the verbal base is linked to the eventive value of the noun. If the noun is non-eventive, the PP is not inherited. If this is so, why can-ble adjectives, which are non-eventive, inherit the PP? Now, recall another problem we leaved unsolved above: if -ble adjectives are non-eventive, why can they syntactically realize the intemal argument of the verb as their subject? (i) jugar a 1 *de futbol play to 1 of football 'to play football' (ii) jugador *a 1 de futbol player to 1 of football 'football player' Our proposal is to relate this behavior of -ble adjectives to the fact that the resulting category of the affixation process is precisely an adjective, whereas in the other cases it is a noun. If we accept the traditional analysis of lexical categories in features, adjectives are [+VI and nouns are [-VI. Adjectives, unlike nouns, function as predicates and so they need a subject (this is not necessarily the case for nouns). The need for having a subject can explain the externalization of the internal argument that becomes the subject (recall that they have a passive meaning). We propose that this predicative value of adjectives makes the -ble adjective capable of inheriting all the interna1 arguments (those that would forn the maximal projection VP with the verb). Then, since predicates need a subject, the adjective must externalize one of the internal arguments, generally the Theme. Notice that sometimes, when the verb has no direct interna1 argument (a Theme NP) and it has only one PP complement, it is this complement that is externalized, without the preposition: (25) a. optable 'optable' b. fiable 'reliable' c. opinable 'commentable' (25) d. navegable 'navigable' e. transitable 'travelable' f. prescindible 'dispensable' g. risible 'laughable' optar per 'opt for' fiar-se de 'rel y on' opinar sobre 'comment on' navegar per 'navigate' transitar per 'travel dong' prescindir de 'dispense with' riure de 'laugh at' These adjectives extemalize the most internal argument, independently of the 8role it receives.9 Finally, notice that middles, non eventive verbal constructions, also have a subject that is an extemalized intemal argument; they have lost the extemal argument and they allow the presence of other intemal arguments of the verb: (26) a. Fets wm aquest s'atribueixen fAcilment a ETA. facis like this attribute easily to ETA 'Actions like this can be easily attributed to ETA.' b. Aquest criteri s'aplica fkilment a qualsevol cas. This criterion applies easily to any case.' c. Aquesta conclusid es dedueix kilment de la teva hipbtesi. this conclusion deduces easily from your hypothesis This conclusion easily follows from your hypothesis.' The parallelism between middles and -ble adjectives is more evident if we adopt Pesetsky's (1990) analysis of middles. In his opinion, English middle forms are bimorphemic, they are formed by a verbal root followed by a zero suffix called MIDDLE, which does not change the category of the verb and absorbs the 0-role of the external argument and case: (27) a. Bureaucratsi bribe-MIDDLE ti easily b. [[bribeIv MIDDLEIV c. [[subomaIv ble]* Leaving the category change aside, the job of the suffixes in (27b) and (27c) is analogous. Pesetsky proposes this nul1 suffix for English based on Keyser and Roeper's (1984) observations (see fn. 3). One could entertain the hypothesis that in the Romance languages the Notice that these examples are a counterexample to de Miguel's (1986) hypothesis, according to which only the Theme may be externalized. She analyzes the complement of some of these verbs as a Theme (e.& navegar 'to navigate'). clitic selsi, which corresponds to the English nul1 affix, be analyzed as a morphological affix and not as a syntactic one, as Keyser and Roeper do. We can conclude that the behavior of -ble adjectives is not a real problem for the hypthesis that non-eventive constructions only allow internal inheritance. Independent principies of Grammar explain the apparent factual contradiction. We would like to thank O. Fullana, C. Picallo, G. Rigau, and A. Suñer for their comments. This paper benefited from the supprt of C.1.C.Y .T. Grant PB89-0324. References Abney, S. (1987) The English Noun Phrase in its Sentential Aspect, The Massachussets Institute of Technology Ph.D. dissertation. Anderson, M. (1979) 'Noun Phrase Structure', University of Connecticut Ph.D. dissertation. Belletti, A. and L. Rizzi (1988) 'Psych-Verbs and Theta Theory', Natural Lunguage and Linguistic Theory, 6, 291-352. Booij, G. (1988) The Relation between Inheritance and Argument Linking: Deverbal Nouns in Dutch' in M. Everaert et al. (4s) Morphology and Modularity, Foris, Dordrecht. Burzio, L. (1986) Italian Syntax, Reidel, Dordrecht. Chomsky, N. (1970) 'Remarks on Nominalization' in R. Jacobs and P. Rosenbaum (eds) Readings in English Tramformational Grammar, Ginn, Waltham, Mass. de Miguel, E. (1986) 'Sulla regola di formazione di aggettivi in -ble in spagnolo', Rivista di Grammatica Generativa, 1 1, 127165. Fabra, P. (1932) Diccionari General de la Llengua Catalana (1982 edition), Edhasa, Barcelona. Fabra, P. (1956) Gramcitica Cataluna, Teide, Barcelona Fiengo, R. (1979) Surface Structure: The Interface of Autonomous Components, Harvard University Press, Cambridge, Mass. Grkia, L. (1989) La Teoria Temdica, Universitat Autbnoma de Barcelona, Bellaterra. Grhcia, L. (forthcoming) 'Los nombres agentivos en -dor y la noci6n de herencia en morfologia', Euskal Heniko Unibertsitateaíuniv. del País Vasco, VitoridGasteiz. Grimshaw, J. (1990) Argument Structure, The MIT Press, Cambridge, Mass. Hoekstra, T. and F. van der Putten (1988) 'Inheritance Phenomena' in M. Everaert et al. (eds) Morphology and Modularity, Foris, Dordrecht. Keyser, S. and T. Roeper (1984) 'On the Middle and Ergative Constructions in English', Linguistic lnquiry, 15,381-416. Levin, B. and M. Rappaport (1988) 'Non-event -er Nominals: a Probe into Argument Stmcture' (to appear in Linguistics) Pesetsky, D. (1990) 'Experiencer Predicates and Universal Alignment Principies', Ms. The Massachusetts Institute of Technology. Picallo, M.C. (1991) 'Nominals and Nominalizations in Catalan', Probus, 3,279-316. Randall, J. (1984) Thematic Stmcture and Inheritance', Quaderni di Semantica, V, 92-1 10. Roeper, T. (1987) 'Implicit Arguments and the Head-Complement Relation', Linguistic lnquiry, 18, 267-3 10. Williams, E. (1981) 'Argument Structure and Morphology', The Linguistic Review, 1, 81Departament de Filologia Faculht de Lletres Universitat de Girona Placa Sant Dodnec, 9 E-17071 Girona