Aspectual Properties of Spanish Absolute Small Clauses

Aspectual Properties of Spanish Absolute Small Clauses* Rafael Marín Universitat Autbnoma de Barcelona. Departament de Filologia Espanyola. 08193 Bellaterra (Barcelona) Catalonia (Spain) rafaG2liceu.uab.e~ Received: January 10th 1997 Accepted: July 10th 1997 Abstract The major aim of this paper consists in determining the aspectual constraints affecting absolute small clauses (ASCs) headed by a past participle. A detailed analysis of these constructions seems to indicate that there are enough arguments to postulate the existence of two types of ASCs: those having an NP direct interna1 argument (AASCs), and those in which this NP is phonetically unrealized (PASCs). The different temporal interpretation that these two types of ASCs can receive is one of the facts that most clearly support this distinction. From the analysis of the aspectual properties of ASCs, the following generalization arises: AASCs can only be formed from eventive predicates, whereas PASCs are not affected by these aspectual restrictions. Finally, an implementation in HPSG is proposed on the aspectual information necessary to capture our results. Key words: absolute small clauses, aspectual analysis, HPSG. Resum. Les propietats aspectuals de les oracions redui'des absolutes L'objectiu principal d'aquest article consisteix a intentar determinar les restriccions aspectuals que afecten les oraciones redu'ides absolutes (ORAs) el nucli de les quals és un participi. Una anhlisi detallada d'aquest tipus de construccions sembla indicar que hi ha arguments suficients per postular l'existbncia de dos tipus d'ORAs: aquelles que contenen un SN argument intern directe i aquelles en les quals aquest SN es troba elidit. La diferent interpretació temporal que poden rebre aquest dos tipus d'ORAs és un dels factors que més clarament incideixen en aquesta distinció. * I would like to thank Lourdes Aguilar and Jala Pfaff for their help in the final English version of this paper. Special thanks are due to Sergio Balari and José Maria Brucart for detailed discussions on more than one version of the ideas presented here. I also benefited from the cornments of two anonymous reviewers. This research has been partially supported by a grant from Generalitat de Catalunya (AP-LPl95-2704). 184 CatWPL 512, 1996 Rafael Marín L'anhlisi de les propietats aspectuals de les ORAs també sembla incidir en la plausibilitat d'aquesta distinció. Així, les ORAs amb SN realitzat fonbticament només podran ésser constitu'ides per predicats eventius, mentre que les ORAs amb SN elidit no es veuen afectades per aquestes restriccions aspectuals. Finalment, exposem una proposta d'implementació en el model HPSG de la informació aspectual necesskia per cop~ar aquests fenbmens. Paraules clau: oracions redu'ides absolutes, analisi aspectual, HPSG. Table of Contents 1. Introduction 4. Aspectual Properties of ASCs 2. A Formal Description of Absolute 5. An Analysis in HPSG Small Clauses 6. Conclusions 3. A Compositional Analysis of Aktionsart References 1. Introduction This paper is devoted to the analysis of the aspectual properties of Spanish absolute small clauses (henceforth, ASCs), which are headed by a past participle in HPSG. It is structured as follows. Section 2 provides an account of the main characteristics of ASCs, focusing on the temporal interpretation that these constructions adopt. Likewise, we will propose, following Dini (1994), the existence of two types of ASCs: aspectual ASCs and predicative ASCs. In section 3, a general treatment of Aktionsart is given. We will present a strictly compositional analysis of Aktionsart starting from the aspectual information that lexical entries contain. We will discuss, moreover, what the most adequate level for the aspectual calculus is. Section 4 deals with the aspectual properties of ASCs by means of the general treatment of Aktionsart proposed in section 3. We will see that the distinction between two types of ASCs offered in section 2 is especially useful in this matter. Finally, in section 5, our analysis is implemented in HPSG. We will see that a strictly lexicalist approach allows us to perform a precise compositional analysis of Aktionsart, as well as an accurate treatment of ASCs. 2. A Formal Description of Absolute Small Clauses López (1994) defines ASCs in the following terms: c<Absolute Small Clauses are adjuncts composed of a predicate -usually, but not necessarily, a past participle- and a noun phrase, which is optional in many cases>>. Hernanz (1991) describes ASCs on the basis of these four properties: ccAbsolute Constructions are not lexically selected by the predicate of the clause they modify; they adopt an adverbial interpretation which does not correlate with any overt Aspectual Properties of Spanish Absolute Small Clauses CatWPL 512, 1996 185 complementizer; they have a lexical NP subject, coreferent or not with an NP in the main clause, and they are headed by an heterogeneous class of predicates, ranging from non-finite verbal forms to adjectives, adverbs, and PPsP. 2.1. Some Syntactic Constraints 2.1.1. Only Transitive and Unaccusative Verbs. According to López (1 994), the NP subject of an ASC has to be the deep object of the verb; that is, the superficial subject of an unaccusative verb, as in (I), or the object of a transitive verb, as in (2): (1) a. Desaparecidas las joyas, llamaron a la policia. disappeared the jewels, called-3pl to the police 'The jewels having been disappeared, they called the police.' [apud López (1994)l b. Muerto el perro, se acabó la rabia. dead the dog, SE finished the rabies 'Once the dog is dead, the rabies is over.' [apud Hernanz (1 99 1 )I (2) a. Arreglado el coche, pudimos seguir el viaje. repaired the car, couldlpl continue the trip 'Once the car was repaired, we could continue the trip.' b. (Una vez) perpretado el robo, 10s ladrones huyeron. (once) perpetrated the theft, the thieves run-away 'Once the theft was perpetrated, the thieves run away.' Deep subjects cannot appear in these constructions, so intransitive verbs cannot be projected as predicative heads of ASCs: (3) a. "Trabajado el torero, Madonna pagó el dinero. worked the bullfighter, Madonna paid the money b. "Corrido Antonio, Melanie se pus0 nerviosa. run Antonio, Melanie SE got nervous 2.1.2. Order. The order past participle + NP is obligatory in ASCs, as shown in the following examples: (4) a. Rotos 10s vínculos familiares, 10s adolescentes se deprimen. broken the bonds familiar, the adolescents SE depress [apud López (1994)l b. *Los vínculos familiares rotos, 10s adolescentes se deprimen. the bonds familiar broken, the adolescents SE depress 186 CatWPL 512, 1996 Rafael Marín As pointed out by Hernanz (1991) and de Migue1(1992), this syntactic constraint may be explained as a consequence of Case assignment mechanisms. 2.1.3. Free Adjuncts. ASCs are not selected by the predicate of the main clause; they are free adjuncts. Therefore, it is not obligatory for them to precede the main clause, as evidenced in (5): (5) a. Una vez construida su casa, Juan se trasladó. once built his house, Juan SE moved b. Juan, una vez construida su casa, se trasladó. Juan, once built his house, SE moved 2.1.4. Agreement and Case Assignment. The NP of an ASC must agree in numberl with the past participle: (6) a. Desautorizado el presidente, . . . unauthorized-masc-sg the president-masc-sg b. *Desautorizados el presidente, . . . unauthorized-masc-pl the president-masc-sg c. *Desautorizado 10s presidentes, . . . unauthorized-masc-sg the presidents-masc-pl This proves that NPs in ASCs receive nominative Case, as shown in (20), where we found pronominal NPs: (7) a. Desautorizado yo, la reunión se acabó. unauthorized I, the meeting SE finished b. *Desautorizado mi, la reunión se acabó. unauthorized me, the meeting SE finished Nevertheless, the inclusion of personal pronouns in this kind of constructions is much more restricted than for ASCs headed by a gerund or an infinitive. 2.1.5. Agent Complement. If the predicate is transitive, an agent complement can be included in an ASC: (8) a. Destruido el litoral por el petróleo, . . . destroyed the seaboard by the oi1 b. Detenido el ministro por la policia, . . . arrested the minister by the police 1. They must also agree in gender, but this constraint is not relevant now. Aspectual Properties of Spanish Absolute Small Clauses CatWPL 512, 1996 187 However, we can find numerous exarnples where an ASC makes it difficult for an agent complement to appear: (9) a. ?(Una vez) escrita la novela por ese profesor, la editorial (once) written the novel by that professor, the publishing-house decidió publicarla. decided publish-INF-it b. ?(Una vez) descubierta la vacuna por un investigador, (once) discovered the vaccine by a researcher, se comercializará. SE commercialize-FUT-3-sg As can be observed in (10), these constructions are completely felicitous in the absence of an agent complement: (10) a. (Una vez) escrita la novela, la editoria decidió (once) written the novel, the publishing-house decided publicarla. publish-INF-it b. (Una vez) descubierta la vacuna, se comercializará. (once) discovered the vaccine, SE commercialize-FUT-3-sg From these examples, we can conclude that ASCs, regardless of their passive character, show a strong resistance to agent complement inclusion. 2.1.6. Specific NPs. NPs direct interna1 arguments must be specific so as to be included in ASCs: (1 1) a. Construida la casa, . . . built the house b. "Construida casa, . . . built house c. *Construidas casas, ... built houses 2.1.7. Negation. As pointed out by several authors (Hernanz (1991), de Miguel (1992), López (1994), among others), negation is incompatible with ASCs: (12) a. *No disuelto el azúcar, ... not dissolved the sugar b. *No disuelto, el azúcar . . . not dissolved, the sugar 188 CatWPL 512, 1996 Rafael Marín (13) a. *No arreglado el coche, ... not repaired the car b. *No arreglado, el coche . . not repaired, the car Nevertheless, contrary to the opinion of these authors, it is possible to find an adverb of negation in an ASC if this element coappears with adverbs such aún (yet) or todavía (still). (14) a. No disuelto (aúrdtodavia] el azúcar, . . . not dissolved (yeilstill) the sugar b. No disuelto (aún/todavia) , el azúcar, . . not dissolved ( yeilstill), the sugar (15) a. No arreglado {aúnltodavia} el coche, . . . not repaired {yet/still} the car b. No arreglado {aúrdtodavia), el coche, . . . not repaired ( yet/still), the car in our opinion, the explanation of this phenomenon is related to the presuppositional information this kind of adverbs contain (see section 4.4.). 2.2. Two Types of ASCs There is disagreement among researchers as to the characterization of the role of the NP subject in ASCs. As mentioned above, López (1994) considers the NP subject in ASCs as being optional, whereas Hernanz (1991) affirms that ASCs must have a lexical NP subject. Concerning ASCs headed by a past participle, we state, contrary to Hernanz (1991), that sequences with empty NP subjects are also ASCs. Following Dini (1994 ), we assume that two types of ASCs exist: those having a lexical NP subject -generally not coreferent with an NP in the main clause-, and those having an empty NP subject -generally coreferent with an NP in the main clause-. 2.2.1. Sequences with Empty NP Subjects. First, we can observe a great sirnilarity between the following sentences: (16) a. La piscina, una vez vaciada, parecia más grande. the swirnming-pool, once emptied seemed more big b. La piscina, una vez vaciada toda el agua, parecia más the swimrning-pool, once emptied all the water, seemed more grande. big Asoectual Prooerties of Svanish Absolute Small Clauses CatWPL 512, 1996 189 (17) a. La novela, (una vez) publicada, fue un Cxito. the novel, (once) published, was a success b. La novela, (una vez) publicada la primera edición, fue un éxito. the novel, (once) published the first edition, was a success It is hard to believe that (16b) and (17b) -both having a lexical NP- are ASCs, and that (16a) and (17a) -both having a phonetically unrealized NP- are not. Second, ASCs with an empty NP subject are similar, but not equal, to some related sentences such as relative clauses. Let us compare the following examples: (18) a. La casa, construida en el acantilado, aguantaba 10s embates the house, built on the cliff, withstand-IMP the lashings del mar. of-the sea b. La casa construida en el acantilado aguantaba 10s embates the house built on the cliff withstand-IMP the lashings del mar. of-the sea (19) a. La casa, que estaba construida en el acantilado, aguantaba the house, which be-IMP built on the cliff, withstood 10s embates del mar. the lashings of-the sea b. La casa que estaba construida en el acantilado aguantaba the house which be-IMP built on the cliff withstood 10s embates del mar. the lashings of-the sea In spite of the similarities between (18) and (19), there is an important difference: ASCs are free adjuncts while relative clauses are not. In (21), it is indicated that relative clauses, contrary to ASCs and other free adjuncts, cannot appear in absolute initial position: (20) a. Construida en un acantilado, la casa se divisaba desde la bahia. built on a cliff, the house SE spy-IMP from the bay b. Talado el árbol, la casa se divisaba desde la bahia. felled the tree, the house SE spy-IMP from the bay (21) a. *Que estaba construida en un acantilado, la casa se divisaba which be-IMP built on a cliff, the house SE spy-IMP desde la bahia. from the bay b. *Que se taló el árbol, la casa se divisaba desde la bahia. which SE felled the tree, the house SE spy-IMP from the bay 190 CatWF'L 512, 1996 Rafael Marin Third, as pointed out by Hemanz (1991), ASCs with a lexical NP subject have a tonal inflection that contributes to set them off from the main sentence. However, in our opinion, not only ASCs with a lexical NP subject, but also ASCs with an empty NP subject have this property. Fourth, both Hernanz (1991) and de Miguel(1992) argue against considering the constructions with an empty NP subject as ASCs. They note that these constructions share the NP subject with another NP in the main sentence. However, the NP subject of ASCs headed by a gerund has the same behavior, and we are quite sure about treating them as ASCs: (22) a. Jugando proi al póker, Harveyi es el mejor. playing to-the poker, Harvey is the best b. Harveyi, jugando proi al póker, es el mejor. Harvey, playing to-the poker, is the best Besides, the main sentence and the ASC with lexical NP subject can share, on some occasions, the NP: (23) a. Consultados 10s juecesi, proi prefirieron no revelar el secreto. consulted the judges, preferred-3pl not disclose-INF the secret b. Interrogados 10s detenidosi, proi acabaron confesando su crimen. interrogated the inmates, ended-up-3pl confessing their crime Finally, we can find examples where ASCs with an empty NP subject and the main sentence do not share the NP. In these cases, the referent appears in a preceding sentence of the discourse. (24) a. Aprobado finalmente con una Única abstención, el Consejo approved finally with a single abstention, the Council pas6 a discutir otro asunto. went-on to discuss-INF another topic b. Llegados al salón, S. M. pronunció unas palabras. arrived at-the room, H. M. spoke some words Examples such as (24) show that ASCs with an empty NP subject can appear completely separated from the sentence they modify. 2.2.2. Aspectual ASCs and Predicative ASCs. As far as ASCs headed by a past participle are concemed, we think, following Dini (1994), that there are enough arguments to conclude that two types of ASCs exist. Dini (1994) distinguishes between Aspectual ASCs (henceforth, AASCs), that is, ASCs with a lexical NP subject, and Predicative ASCs (henceforth, PASCs), in other words, ASCs with an empty NP subject. The main argument adduced by Dini (1994) is the different temporal interpretation these two types of constructions receive. Dini (1994) provides the following examples for Italian: Aspectual Properties of Spanish Absolute Small Clauses CatWPL 512, 1996 191 (25) a. Insultatolo, i1 linguista si suicidb. 'Mistreated him, the linguist committed suicide.' b. Insultato da tutti i professori, i1 linguista si suicidb. 'Mistreated by all the professors, the linguist committed suicide.' (26) a. Interrogato i1 presidente dal Pubblico Ministero, i telespettatori scoppiarono a piangere. 'Questioned the president by the State Prosecutor, the television viewers burst into tears.' b. Interrogato dal Pubblico Ministero, Silvio scoppib a piangere. 'Questioned by the State Prosecutor, Silvio burst into tears.' AASCs such as (25a) and (26a) denote a time interval previous to the sentence they modify, whereas PASCs such as (25b) and (26b) can be previous or simultaneous to the main sentence. We find a similar behavior in Spanish: (27) a. Leído el discurso por el Señor Rosa, el público se rió. read the speech by the Mr. Pink, the public SE laughed b. Leído por el Señor Rosa, el discurso hizo reír al público. read by the Mr. Pink, the speech made laugh-INF to-the public Only (28a) is an adequate paraphrase for (27a). This fact shows that the tense denoted by AASCs is previous to the tense denoted by the main sentence: (28) a. Después de que el Señor Rosa leyera el discurso, after of that the Mr. Pink read-IMP-SUBJ the speech, el públic0 se rió. the public SE laughed b. Mientras el Señor Rosa leia el discurso, el públic0 se rió. while theMr. Pink read-IMP the speech, the public SE laughed As mentioned above, PASCs have two possible temporal interpretations. In this sense, we can observe that both (28a) and (28b) are valid paraphrases from (27b). Dini (1994) proposes the following diagrams to represent these two different time intervals: (29) [terminative] insult suicide (30) [durative] insult suicide 198 CatWPL 512, 1996 Rafael Marín Smith (1991) differentiates between accomplishments (to win a match) and semelfactives (to cough, to ~neeze).~ 3.3.2. Our Proposal. As far as the aspectual labeling of lexical entries is concerned, our proposa1 is based on the following points: (iii) Eventized verbs do not exist; that is to say, there are no verbs labeled in the lexicon as events regardless of the properties of their arguments. (iv) Uneventizable verbs are verbs labeled in the lexicon as states or processes regardless of the kind of NP subcategorized as direct object. (v) Eventizable verbs are verbs which can form VPs labeled as events depending on the properties of their arguments. If the NP is [-SQA], then these verbs will form processive VPs; if the NP is [+SQA], then these verbs will form eventive VPs. The global procedure of aspectual labeling of lexical entries is presented in the following figure: verbs uneventizables eventizables [lexical level] states processes events [VP level] As a consequence of our proposal, we can establish an important link between syntax and semantics: eventizable verbs must be transitive or unaccusative. This syntactic restriction does not apply in the case of uneventizable verbs.1° 4. Aspectual Properties of ASCs As shown in the previous section, the aspectual markedness of verbs is not enough to determine the aspectual class to which a predicate belongs. On the contrary, it is necessary to calculate the Aktionsart compositionally by means of the information given by the verb and the NP direct interna1 argument. 9. Smith (1991) considers, moreover, that semelfactives are punctual and atelic. Dini and Bertinetto (1995) reach a similar conclusion. In fact, there are not strong reasons to vinculate the punctuality of a predicate to the class of events. 10. The behavior of some uneventizable verbs is especially interesting. So, we can find transitive verbs like to caress, to push or to use that always form a processive predicate. Aspectual Properties of Spanish Absolute Small Clauses CatWPL 512, 1996 199 As regards the aspectual constraints determining the good formation of AASCs, in Marín (1994) we proposed the following generalization: (vi) Only VPs aspectually marked as events can appear in AASCs. It is hereby inferred that if a VP pertains to one of the classes state or process, it cannot appear in an AASC. It is worthwhile to compare the grammaticality of the sentences in (45) with the ungrammaticality of (46): (45) a. Construida la casa, . . . built the house [accomplishment] b. Descubierta la vacuna, . . . [achievement] discovered the vaccine (46) a. "Acariciada su mujer, . . . [process] caressed his wife b. *Amada su mujer, . . . loved his wife [state] We have also shown, concerning the contribution of the NP direct object in the calculus of the Aktionsart, that VPs where the NP appears without specification, either singular or plural, cannot form AASCs. This is held for the exarnples in (47). (47) a. "Construida casa, . . built house b. "Construidas casas, . . ; built houses However, we cannot conclude that the nominal specification on its own constitutes a good criterion in the analysis of AASCs. Although non-specific NPs cannot appear in this type of predications, some AASCs are ungrammatical when including specific NPs, as shown in (46). In the previous section, the differences between uneventizable and eventizable verbs were stated. Uneventizable verbs are those verbs that will always form noneventive VPs, with independence of their NP. This implies states (to know, to fear) and some of the processes (to caress, to use). Eventizable verbs are those verbs (to build, to discover) that can form eventive and non-eventive predicates, depending on the value, plus or minus, of the feature SQA of the NP direct internal argument. For the sake of clarity, let us compare the examples in (48) and (49): (48) a. *Amada su mujer, . . . loved his wife b. "Acariciada su mujer, . . . caressed his wife [uneventizable (state)] [uneventizable (process)] 200 CatWPL 512, 1996 Rafael Marín (49) a. *Construidas casas, . . . built houses b. Construidas las casas, . . . built the houses [eventizable (process)] [eventizable (event)] In order to incorporate these distinctions in our analysis, the generalization in (vi) could be reformulated in the following tems: (vii) Only VPs formed by an eventizable verb and an NP [+SQA] can appear in AASCs. It is inferred from (vii) that neither a VP formed by an uneventizable verb and an NP [-SQA] nor a VP with an uneventizable verb can appear in an AASC. With this approach, following either (vi) or (vii), one need not include the specificity of the NP in the syntactic characterization of AASCs. In addition, it is explained why some verbs cannot appear in AASCs even if they select specific NPs. These arguments make our proposal more adequate when compared to others.ll 4.2. PASCs As far as PASCs are concerned, the generalizations proposed above are not adequate: PASCs do not share the aspectual constraints described for AASCs. As we can see in the examples below, not only eventive predicates but also stative and processive ones, can form PASCs: (50) a. El decano, interrogado por la policia, se sintió abatido. the dean, questioned by the police, SE felt dejected b. La vacuna, (una vez) descubierta, resultó ser muy adecuada. the vaccine, (once) discovered, turned-out be-INF very adequate c. Madonna, amada por el torero, era feliz. Madonna, loved by the bullfighter, be-IMP happy d. Antonio, acariciado por Melanie, se excitó. Antonio, caressed by Melanie, SE excited Differences concerning aspectual constraints in AASCs and PASCs confirm the distinction between these two types of ASCs proposed in section 2. As we have mentioned formerly, Dini (1994) observes that Italian PASCs can receive a durative or a terminative interpretation. In Spanish, we can observe a relationship between aspectual constraints and temporal interpretation: PASCs formed by an eventive predicate are ambiguous between a durative and a terminative interpretation, while PASCs formed by a non-eventive predicate can only receive 11. The aspectual analysis proposed by de Miguel(1992) does not take into account these syntactic specifications and, as a consequence, they have to be included as a characteristic. Moreover, de Miguel(1992) does not consider that not only the verb must be aspectually marked, but also the complete predicate. Aspectual Properties of Spanish Absolute Small Clauses CatWPL 512, 1996 201 a durative interpretation. Thls is shown in (51), where PASCs formed by a non-even- tive predicate cannot cooccur with una vez (once), an adverb which forces a tenninative meaning: I (5 1) a. "Madonna, una vez amada por el torero, era feliz. Madonna, once loved by the bullfighter, be-IMP happy b. "Antonio, una vez acariciado por Melanie, se excitó. Antonio, once caressed by Melanie, SE excited On the contrary, PASCs formed by eventive predicates can coappear with this adverb, as can be seen in (52): (52) a. El decano, una vez interrogado por la policia, se sintió abatido. the dean, once questioned by the police, SE felt dejected b. El discurso, una vez leido por el presidente, hizo reír al pdblico. the speech, once read by the president, made laugh to-the public Another test which allows us to argue that PASCs formed by eventive predicates can receive a double interpretation is observed in the following examples: (53) a. El decano, una vez que fue interrogado por la policia, se sintió the dean, once that was questioned by the police, SE felt abatido. dejected b. El decano, siendo interrogado por la policia, se sintió abatido. the dean, being questioned by the police, SE felt dejected It should be noted that both (53a) and (53b) are valid paraphrases for (50a); however, (54b) is the only adequate paraphrase for (50c): (54) a. "Madonna, una vez que fue amada por el torero, era feliz. Madonna, once that was loved by the bullfighter, be-IMP happy b. Madonna, siendo amada por el torero, era feliz. Madonna, being loved by the bullfighter, be-IMP happy From the data analyzed throughout this section, we can extract the following generalizations: (viii) PASCs formed by an eventive predicate can denote a time interval previous or simultaneous to the main sentence. (ix) PASCs formed by a non-eventive predicate can only denote a time interval simultaneous to the sentence they modify. 202 CatWPL 512, 1996 Rafael Marin In (53, the global results with respect to the temporal interpretation of both AASCs and PASCs are shown: (55) terminative durative AASCs + - PASCs (eventive VP) + + PASCS (non-eventive VP) - + As we can observe in (53, a strong interaction between temporal meaning and aspectual labeling is found for Spanish. 4.3. The Generic Interpretation of ASCs Concerning the relevance of the NP direct object in the interpretation of the ASCs, it must be said that the ASCs with a non-definite NP can have a generic interpretation. This is demonstrated in the following examples: (56) a. Acabado un ciclo, comienza otro. finished one cycle, begins another b. Una vez cometido un robo, te conviertes en un ladrón. once committed a theft, you turn into a thief However, this generic interpretation is only possible when the verb of the main sentence is in the present or the imperfect indicative tenses.12 This is the reason why the examples of (57) are preferred to those of (58). (57) a. Acabado un imperio, empieza otro. finished one empire, begins another b. Acabado un imperio, empezaba otro. finished one empire, begin-IMP another (58) a. ?Acabado un imperio, empezó otro. finished one empire, began another b. ??Acabado un imperio, ha empezado otro. finished one empire, has begun another The doubts about the grammaticality of the constructions in (58) make us think that the ASCs having a non-definite NP cannot complement a main sentence which is in the past tense. 12. The future can favor a generic interpretation too, but its behavior is not truly consistent, as it is the case with the present and the imperfect. Aspectual Properties of Spanish Absolute Srnall Clauses CatWPL 512, 1996 203 This behavior is shown in (59). If in the same examples the nature of the NP changes from non-definite to definite they become acceptable, as evidenced in (60): (59) a. ??Una vez acabada una película, Quentin se ha sentido muy once finished a film, Quentin SE has felt very satisfecho. satisfied b. ?Detenido un ladrón de joyas, el comisario se sintió muy arrested a thief of jewels, the police-chief SE felt very satisfecho. satisfied (60) a. Una vez acabada la película, Quentin se ha sentido muy satisfecho. once finished the film, Quentin SE has felt very satisfied b. Detenido el ladrón de joyas, el comisario se sintió muy satisfecho. arrested the thief of jewels, the police-chief SE felt very satisfied The main reason for the grammaticality of (60) in contrast with the doubtful acceptability of (59) lies, in our opinion, on the generic interpretation that the ASCs with non-definite NP tend to receive. 4.4. Negation As mentioned in section 2, Hernanz (1991) and de Miguel (1992) contend that negative polarity elements cannot appear in ASCs. (61) a. *No construida la casa, . . . not built the house b. *No abierta la puerta, . . . not opened the door Nevertheless, the constructions in (61) become acceptable if an adverb such as aún (yet) or todavía (still) is included: (62) a. No construida todavía la casa, . . not built still the house b. No abierta aún la puerta, . . . not opened yet the door In our opinion, the well-formedness of (62) is related to the presuppositional nature that these adverbs give to the ASC.'^ 13. See Marín (1996) for a more detailed discussion. 204 CatWPL 512, 1996 Rafael Marin However, despite these considerations, the ungramrnaticality of the examples in (61) is covered by the generalizations proposed formerly. Negation turns a delimited predicate into a non-delimited one. Thus, no construir una casa indicates a non-eventive situation and, as a consequence, it cannot appear in an ASC. Once more, the aspectual analysis of basic predcation proposed in our study shows a high degree of generalization. If we state that only the eventive predicates can appear in ASCs, we have a solution for the problem posed by negation and no other specification needs to be added. 5. An Analysis in HPSG 5.1. The Aspectual Information in HPSG In section 3, we have seen that, although the VP is the appropriate level of the aspectual analysis, it is necessary to specify the aspectual information contained in lexical entries. This is especially relevant if we are interested in an implementation of our results in a linguistic model such as HPSG. According to the analysis proposed, stative and processive predicates may be headed by uneventizable verbs. On the other hand, eventizable verbs can form processive or eventive predicates depending on the type of NP direct interna1 argument they select. These three kinds of predicates can be obtained by means of markedness with respect to ADD TO and delirnitedness: (63) ADD TO DELIMITEDNESS states - - processes + - events + + As far as the treatment of Aktionsart in HPSG is concerned, our proposa1 can be synthesized as follows: aspectual information is contained in the AKTIONSART (AKTIO) attribute of the lexical entries which is part of the CONTENT attribute. The value of AKTIO is an object of the type situation which is defined by the DELIMITEDNESS (DELIM) and ADD TO Boolean features, as shown in (64) : lent ton j ] AKTIO ADD TO boolean DELIM boolean In the case of uneventizable verbs, the treatment of aspectual information and the subsequent markedness of the predicate is not as difficult as in the case of eventizable verbs. Aspectual Properties of Spanish Absolute Small Clauses CatWPL 512, 1996 205 Stative predicates are headed by verbs containing both features [ADD TO -1 and [DELIM -1, and adopt the same aspectual information as these verbs: content 1 lstote 1 1 AKTIO ADD TO - L [DELIM - jj Some processive predicates are formed by uneventizable verbs containing both features [ADD TO +] and [DELIM -1 and adopt the same aspectual information as these verbs: rient Drocen ] AKTIO ADD TO + DELIM - The analysis of eventizable verbs is slightly more complex. The information contained in lexical entries is not sufficient to obtain either the aspectual valence of eventive predicates or a great number of processive predicates. We have to take into account these two possibilities: 1) if an eventizable verb selects an NP [SQA -1 as direct interna1 argument, then this verb will be the head of a processive VP; 2) if an eventizable verb selects an NP [SQA +] as direct interna1 argument, then this verb will be the head of an eventive VP. A possible solution would consist of considering the DELIM attribute as having a disjunctive value: content 1 loccurrence 1 1 AKTIO ADD TO + I I DELIM + v- 1 1 If we adopt this representation of lexical entries, it would be possible to compositionally obtain the aspectual valence of processive and eventive VPs in this way: -PHON () - SYNSEM I LOC I CONTENT I AKTIO [~~~&O~P [2~ PHON () HEAD-DTR I I ADD TO + SYNSEM I LOC I CONTENT I AKTIO [DELIM + - DTRS PHON () - I) - I I - 206 CatWPL 512, 1996 Rafael Man'n In the case of eventive predicates, for example, we would obtain the following representation: PHON (construir la casa) SYNSEM I LOC I CONTENT l AKTIO [bEg&Od PHON ( construir ) HEAD-DTR COMP-DTRS ([ PHON ( la casa ) SYNSEM I LOC I CONTENT I AKTIO [SQA +] However, this analysis is not adequate in HPSG. There are relevant technical reasons that exclude it. Firstly, lexical entries cannot hold an attribute with a disjunction as value. Secondly, as postulated in the Content Principle, the CONTENT of the semantic head is token-identical to the CONTENT of the mother. Obviously, (65) and (66) violate this principle because the information contained in the CONTENT of the mother is different from the information of the CONTENT of the head-daughter.14 We can solve these two problems by means of a lexical rule as in (67): occurrence 1 1 Ioccurrence 1 1 AKTIO ADD TO + The use of this rule respects the requirements of the Content Principle: the DELIM attribute adopts a fixed value -positive or negativein the lexical level. This implies that the semantic information does not change between the lexical level and the phrasal level. As we can observe in (67), the DELIM attribute takes the same value as SQA. Thus, for example, if the SQA attribute has a negative value, then the DELIM value will be negative too, and the subsequent VP will be processive. Nevertheless, given that in HPSG the information contained in the COMPS list is accessible to the head, the lexical rule in (70) is not necessary. We can offer a more accurate analysis as in (71): 14. We would be able to solve this problem proposing a modification of the Content Principle. Related to this, we believe that increasingly there are arguments to treat semantic information in a configurational way, like syntactic information. Aspectual Properties of Spanish Absolute Small Clauses CatWPL 512, 1996 207 The lexical entry of eventizable verbs that form processive predicates is illustrated in (71). Eventizable verbs that can head an eventive VP have the following lexical entry: I i' ,j 'OMPS NP CASE acc Note that the CASE value of the NP selected by the verb is accusative. It accounts for the fact that only NP direct interna1 arguments are involved in the aspectual calculus of predicates. As far as a general treatment of Aktionsart is concerned, results obtained in this section allow us to propose the following type-hierarchy: situation [2:&::::11 1 r occurrence ADD TO + , . rrocess ] Ievent ] ADD TO + ADD TO + DELIMDELIM + Observe that by means of this type-hierarchy the relationship between processes and events is captured (Verkuyl(1989)). 5.2. The Temporal Information in HPSG In order to deal with the adverbial interpretation of the ASCS and, more specifically, with the temporal relation between these constructions and the main sentence, Dini (1994) includes in his work the analysis of tense developed by Reichenbach (1947)15 as well as later applications such as Hornstein (1977, 1990) and van Eynde (1991). 15. Acero (1990) provides an examination of Reichenbach (1990)'s proposal especially usefull for the analysis of Spanish.