Optimal Reflexivity in Dutch

Binding principles applied to Dutch reflexives are supplemented with optimality considerations and a general principle of linguistic economy to account for differences in reflexivization strategies of Dutch and English. Dutch SE-reflexives are optimally encoded coreference in contrast to the English ordinary bound pronouns they are translated to. Burzio's stated general dispreference for SE-reflexives is replaced by a preference based on economy considerations for SE-reflexivization. 1 I N T R O D U C T I O N : OPTIMALITY IN SEMANTICS The view that natural languages are adaptive, biologically based systems for the efficient and economical communication of information sheds new light on the issues of the semantics of natural languages and opens up an interesting path towards novel explanatory insights into the structural processes of human communication. The central thesis of linguistic economy considers information to be a negotiable commodity, where the producer is maximizing profits by restricting resources, such as time, articulatory effort, memory, and attention, and the consumer is seeking to maximize his understanding by updating his information state extracting all information from what is said, while minimizing his cognitive effort and economizing on processing cost. Producer and consumer often may have conflicting interests within an overarching common concern to cooperate in sharing information. The producer serves his interest, for instance, in expressing his thought in a way that he expects to fit in with what he thinks the consumer already believes, and in minimizing the time needed to express it, i.e. preferring fewer and shorter words. The consumer will serve his interests if he maximizes binding of definite expressions like anaphora or definite descriptions, accommodating presuppositions where necessary, and getting as much out of the locutionary act as his information state permits him, while maintaining cohesion within the information he assumes to be true. The agenda for semantic theory, so reconceived, consists now in sorting out what constraints govern the actions of both producer and consumer, which different currencies may play a role, how conflicts of interests may arise, as well as what strategies may be exploited to cooperatively resolve the conflicts. Of course, the syntactic phrase structure 264 Optimal Reflexivity in Dutch rules and configurationally defined notions such as c-command and locality conditions do their share, and compositional semantic rules still play their own role in grammar. These are hard, inviolable constraints, whose exceptions result in (degrees of) ungrammaticality or semantic uninterpretability. Binding principles provide the point of departure for the account of Dutch reflexives offered in this paper, but optimality considerations are needed in addition as principle of linguistic economy to account for differences in reflexivization strategies of Dutch and English. Against this generalized pragmatics background, where optimizing use of the language in different circumstances is at stake and processing considerations enter into consideration, this paper outlines an OT account of the semantics of the two Dutch reflexive pronouns: the short form zich, glossed as SE, and the morphologically complex zichzelf, glossed as SELF. The referential economy hierarchy, first developed in various functionalist or pragmatic accounts, and adopted subsequently within Optimality Theory in Burzio (1998), assumes that reflexives have no descriptive semantic content, pronouns have some, and other referring expressions, including proper names, have full referential content. This paper will argue that Dutch SE-reflexivization is an economical way to (restructure information in comparison to SELF-reflexivization or to ordinary coreferential pronouns. Linguistic economy can be abstractly defined as a stable equilibrium in the equation relating the parameters minimizing the processing load for the consumer of the information to the parameters maximizing the ease of expression for the producer. This mini-max strategy may not always reach a unique best result, as conflicts of interest may arise between speaker and hearer, or informational power struggles may end in stalemate. Reflexives that only take impersonal or third-person singular or plural antecedents in Dutch support Burzio's view that their lack of agreement features makes them match best with local impersonal antecedents, or the weakened feature structure of local third-person antecedents. They are contrasted to the firstand second-person NPs that additionally have richer, partly contextually determined features and bind case-, genderand number-sensitive compound reflexives, consisting of the corresponding pronoun and the bound morpheme -zelf. I will argue, contrary to Burzio, however, that SE-reflexives in Dutch provide a maximally economical 1 The notion of a Nash equilibrium, defined in Dekker & van Rooy (2000), may well be a good step towards formalization of the mini-max strategy informally characterized here. It requires the spelling out of the preference order of actions for speaker and hearer in matching representations to possible meanings to determine which is the optimal alternatively. This would take us considerably beyond our current concerns, which is to describe the role of SE-reflexivization as optimizing strategy in comparing Dutch to English. Alice G. B. ter Meulen 265 strategy for local binding, optimally encoding coreference in comparison to the English counterparts with ordinary bound pronouns. Burzio's stated general dispreference for SE-reflexives is consequently dismissed and replaced by a preference based on economy considerations for SE-reflexivization. 2 THE ECONOMY OF REFLEXIVIZATION Linguistic economy leads us to expect that whenever lexically extensional substitution salva veritate is observed in a construction, there must be another context where the two expressions do make a different contribution to the content of the sentence. This meaningful difference could well be located in the discourse relations of the sentence, i.e. in the way its content is merged with the context created by preceding discourse, or in the way either the producer may continue or the consumer react to it. Contextindependent true synonymy, as strongly intensional substitution in any context not only salva veritate, but also preserving content and context change potential, is inefficient and uneconomical. Then why does Dutch, unlike English, need two reflexive pronouns? (1) Peter scheert zich. (2) Peter scheert zichzelf. [Peter shaves SE] [Peter shaves SELF] Peter shaves. Peter shaves himself. In (1) and (2) prima facie, no truth conditional difference can be detected between the two sentences in isolation, so zich and zichzelf could be considered synonymous expressions in Dutch in this context, both lacking case, number, and gender agreement features. But there is a hidden, perhaps to some extent culturally determined assumption at play here, for this observation is correct only when Peter is the agent of the predicate. In (1), both Dutch and English, the identity of the shaver could, but need not be Peter. In (2), however, Peter must be the shaver, and zichzelf is a full-fledged argument of the predicate, assigning it thematic role PATIENS. If someone other than Peter shaves Peter, (1) is still true, but (2) is not, but (2) entails (1). Focus is sometimes considered to give additional content to zichzelf so (2) expresses that of all people who could possibly shave Peter, Peter himself was the only one who did so. The -zelf morpheme in Dutch combines with the short form zich to yield an internal argument of the predicate, interpreted as a generalized quantifier, stating the identity of the internal and external argument; for (2) resulting in the focussed content Vx 3y[(shave (x, y) & Patient (y) & Agent (x) & y = Peter) —> x = Peter] 266 Optimal Reflexivity in Dutch In English the morpheme —self combines with the ordinary, case sensitive pronoun to form a reflexive. In Dutch quantificational adverbs, in Spec of VP, may modify such predications, as in (3), where the internal argument zelf remains in situ within the VP, while zich is cliticized in I-projection. (3) Peter scheert zich [VP nooit zelf]. Peter shaves SE [yp never SELF]. Peter never shaves himself. This indicates the first important difference between Dutch and English reflexivization strategies. In (3) Dutch uses the short form reflexive zich, economizing on agreement features and overt expression of the agent argument, to form a reflexive predicate, encoding that zelf must be identified with the external argument Peter in Spec of IP and still carrying the PATIENT role, as it is raised from internal argument to the external position. English must resort in (4a or b) to a less economical way of expressing the equivalent content of (3), with an overt existentially quantified shaver argument, using an ordinary, case sensitive, bound pronoun in a light verb exceptional case marking (ECM) construction with a bare infinitival complement. (4) a. Peter always has someone (else) shave him. b. Peter never has himself shave him. Resolving bound pronouns is relatively more costly for a consumer of information, as it requires calls to available referents stored in memory, exploits lexical information, as well as perceptual sources in the situation of use and common sense. Of course, him in (4a) could always be interpreted deictically or otherwise bound by an antecedent in prior discourse, an ambiguity the Dutch (3) lacks. Alternatively, (4a) could be paraphrased, less naturally, by its dual (4b). In any case, situations in which (3) and (4a, b) are true, (5) must be false. (5) Peter never shaves. In Dutch there is also a less economical, marked way of expressing (3), which may overtly refer to the shaver role in an optional PP, as in passives, using a light verb with a non-finite complement in (3'). The lack of 2 Modifying PPs are not arguments of the verbal predicate, but adjuncts to it, hence are not taken into consideration in this paper. 3 Light verbs are verbs that have little or no descriptive content, but that may restructure the arguments of a predicate or reassign thematic roles. Causative verbs like have or let are examples of light verbs, that serve to make the agent causing the action overt. In the classical example oiJane gave Sally a kiss, give serves to turn a transitive predicate structure into a di-transitive one and facilitates discourse pronominal reference to the nominalized predicate as internal argument, a kiss, in subsequent sentence, e.g. It landed on her cheek. Alice G. B. ter Meulen 267 agreement features that require checking by the verbal predicate allow both zkh and zichzelf to raise to subject of a non-finite complement of a light verb, as in (3b), to corefer with the subject of the main IP. (3') Peter laat zich(zelf) (altijd) ([PP door een ander]) scheren. Peter lets SE/SELF (always) ([PP by someone else]) shave. Peter always let someone else shave him. When the shaving agent is overtly expressed, (3') can be continued with a simple pronoun coreferring with it, as in (6a). (6) a. Hij betaalt hem tien dollar. He pays him ten dollars, b. Hij betaalt de barbier tien dollar. He pays the barber ten dollars. When the shaver argument is implicit as in (3) or in (3') without PP, a definite description de barbier/the barber must be used to express the coreference with the shaver and establish reference to him, as in (6b). Universally, pronouns require overtly asserted antecedents, but definite descriptions allow accommodated antecedents to be interpreted as coreferential with indefinites inferred from implicit arguments in prior discourse. 3 FURTHER PROPERTIES OF SE REFLEXIVES SE reflexives, like clitics, do not constitute acceptable short answers to whquestions (7), cannot be fronted in wh-questions (8), nor do they carry high pitch accents marking focus or nuclear scope (9), nor admit of topicalization (10), nor coordinate with syntactic arguments of verbs (11). (7) Wie scheert Peter? *Zich/Zichzelf. [Who shaves Peter? * SE/SELF] Who does Peter shave? Himself. (8) Jim laat Peter zich scheren. *Wie laat Jan Peter t scheren? [Jim let Peter SE shave. * Who let Jim Peter t shave?] Jim let Peter shave. Who does Jim let Peter shave? 4 This is an important difference in creating discourse cohesion with two classes of definites, pronouns, and definite descriptions. Only rather special scenarios may support an inferred antecedent as in Jane is pregnant. It is a boy. A definite description allows its presupposed referent to be identified with a referent inferred from prior discourse. No such bridging is available for ordinary pronouns, although high pitch pronouns with appositive relative clauses (e.g. He who Jane is pregnant with, (H*LL%)) could very well in this respect have the same semantic properties as definite descriptions. Cf. ter Meulen (1999). 268 Optimal Reflexivity in Dutch (9) Als Peter *ZICH/ZICHZELF scheert, dan mag hij naar het feest. [If Peter *SE/SELF shaves, then he may go to the party.] If Peter shaves himself, he may go to the party. (10) *Zich/Zichzelf heeft Peter t goed geschoren. [*SE/SELF has Peter t shaved well] Himself Peter has shaven t well. (11) Peter scheerde *zich/zichzelf en Jim. [Peter shaved *SE/SELF and Jim.] Peter shaved himself and Jim. In all of the contexts (7)-(i 1) the SELF reflexive is perfectly acceptable, supporting the claim above that it is a full-fledged argument of a predicate, in spite of its lack of agreement features. The SE-reflexivization is best regarded as an operator reducing binary simple reflexive relations to unary reflexive marked properties of individuals, according to: operator O(t>t): Ax OX(R (x, x)) = Ax SE-R (x) (syntactic reflexive reduction) But this SE-reflexivization reduction is not applicable to just any binary relation one may have to oneself. The meaning of the predicate must be constrained in specific ways that first need further description. If the verbal predicate describes an involuntary, internal bodily action (12), or an emotional state denoted by intransitive verbs (13a), only SE reflexives are acceptable, and SELF reflexives are not. This class of predicates is included in what Reinhart & Reuland (1993) call 'inherent reflexive predicates'. (12) Peter verslikte zich/*zichzelf/*Mary [Peter ver-swallow-PAST SE/* SELF/* Mary] Peter choked (IV) (13) a. Peter schaamt zich/*zichzelf/*Mary [Peter shames SE/* SELF/* Mary] Peter is ashamed (of himself) b. Peter haat *zich/zichzelf/Mary [Peter hate *SE/SELF/Mary] Peter hates himself/Mary c. Peter verbaasde zich/zichzelf/Mary [Peter surprised SE/SELF/Mary] Peter surprised himself/Mary. As a final point in this description of the differences between SEand SELF-reflexives, it should be noted that the famous Russellian selfreferential barber paradox arises only with SELF-reflexives, assuming in addition that the definite description referring to the barber picks a unique Alice G. B. ter Meulen 269 individual, while ignoring an inverse scope interpretation of the barber as dependent definite NP. An SE-reflexive in the same context creates an outright contradiction in (14), for the statement could only be true just in case the unique barber shaves everyone, including the barber himself, who neither shaves himself nor is shaven by anyone other than himself. This appeals to the fact that zich scheren does not assign agent to external arguments, as discussed above, hence zich met scheren is true only of those whom no one shaves, i.e. men with a beard. Accordingly, (14) constitutes a contradiction, in English expressed without internal argument. (14) De kapper scheert iedereen die zich niet scheert. [The barber shaves everyone who SE not shaves.] The barber shaves everyone who does not shave. In (15) the SELF-reflexive creates an internal argument in the restrictive relative clause of the quantifier, that binds it and is itself the internal argument of the main transitive predicate. (15) De kapper scheert iedereen die zichzelf niet scheert. [The barber shaves everyone who SELF not shaves.] The barber shaves everyone who does not shave himself. The paradox arises in attempting to answer the question whether the barber shaves himself, assuming that (15) is true and there is only one barber. For if he does shave himself, then he is excluded from the domain of non-selfshavers, but they are just the ones that get shaven by that unique barber: contradiction! And if the barber does not shave himself, then he must be included in that domain and hence get shaven by himself after all: contradiction again! Although (is) is perfectly grammatical and even interpretable, it can neither be verified nor falsified for the barber, who must be included in any case in the domain of any model for (15). Dutch speakers often prefer a passive construction over (15), as in (16). (16) Iedereen die zichzelf niet scheert wordt geschoren door de barbier. [Everyone who SELF not shaves is shaven by the barber] Everyone who does not shave himself is shaven by the barber. The interpretation of (16) easily allows for a dependent definite, referring to several different barbers, dependent upon the choice on non-self-shaver. Under that interpretation, (16) does not create a paradox, for one barber may shave another one, getting everyone shaven in the domain and yet having no barber shave himself. An insightful explanation of the semantics of self-referential constructions with the two Dutch reflexives, to my 5 Here I am disregarding the tacit exclusion of women of the domain of this universal quantifier. 270 Optimal Reflexivity in Dutch knowledge not yet addressed in the logical literature on self-reference and truth functional paradoxes, must be deferred to another occasion. 4 REFERENTIAL ECONOMY: TOWARDS AN OT ANALYSIS The referential hierarchy alluded to above considers reflexives easiest to process as they are most economical, only requiring a call to local memory for the referent of the subject. Pronouns may be interpreted deictically, requiring a call to the situation of use, or anaphorically, calling to memory any available, possibly non-local antecedents, stored by the interpretation of preceding discourse. Proper names refer independently in any context, so they introduce a new referent at their first use and corefer with it at each subsequent use, regardless of the level of information structure. If coarguments of a verbal predicate are not marked for coreference by reflexives, they should be interpreted as disjoint in reference (Binding condition B). It is always possible that ignorance or intention to mislead on the part of the speaker makes him fail to encode the coreference of two co-arguments, even though in fact they do corefer. In other words, the interpreter should always allow for the possibility that his source, the speaker, may not have perfect information, and hence cannot be blamed for failing to encode coreference, even though two co-arguments do corefer in the intended model. Consumers of information should always be on the alert for a possible discrepancy between what the source of their information wants them to believe is true and what in fact is true, even if their sources do not intend to be lying. Ignorance is more often the parent of misunderstanding than the intention to mislead. For present purposes, however, we are interested in analysing what information a speaker conveys by his assertions—how he shares his beliefs and thoughts by encoding them in expressions with that information content. Given these assumptions, consider the data in (i7)-(io). (17) De barbier; knipte Peteij. [The barber cut (3PS past) Peter.] The barber cut Peter. (18) Peterj werd geknipt (door de barbier;). [Peter was cut (by the barber).] Peter was cut (by the barber). 6 Each NP is assigned its own index, their coreference is indicated by an identity. Here i, j , and k are used as distinct indices and the table in Figure I presents their coreference conditions. Alice G. B. ter Meulen 271 (19) Peterj liet zich^ knippen (door de barbier;). [Peter let SE cut by the barber.] Peter let himself be cut (by the barber). In (17) the two definite co-arguments of the predicate are interpreted as disjoint in reference (i ̂ j). The same holds for the passive form of (17) in (18), but (18) may economize on the agent argument by suppressing the PP. The marked construction with the light verb let in (19) is on the one hand less economical than (17) and (18) in using two predicates, instead of one. On the other hand, the light verb let in (19) assigns control of the subject over the situation described in its non-finite complement. The SE reflexive in (19) is raised to subject of the non-finite clause, and the agent of that complement is again in an optional PP, just as the passive in the English translation. Furthermore, (19) with overt PP contains stronger information, as it entails (17) and (18), but not vice versa. (20) Peterj liet (alleen) zichzelf^ knippen. [Peter let (only) SELF cut.] * Peter let (only) himself cut. In (20) the SELF reflexive, optionally in the scope of the focus adverb, creates an ambiguity that (19) lacks, for SELF could, but need not be an agentive subject of the non-finite complement knippen if its internal argument is implicit. If it is agentive, of course, the exceptional case marking construction would be ungrammatical with a PP specifying another agent of the predicate. Economizing on the agentive PP as in a passive in (18), (20) shares an interpretation with (19) where SELF as internal argument would only add the contrastive focus over SE. But (20) adds the ECM interpretation with an implicit existential internal argument, unacceptable in its English translation. (21) Peterj liet hem^ knippen. [Peter let him cut.] Peter let him cut. A non-reflexive pronoun cannot be interpreted as coreferential with the subject, as in (21), when they are co-arguments of the predicate (Binding condition B). The interpretation of (21) requires disjoint reference, with either a deictic interpretation of the pronoun or one coreferring with a referent available in memory, introduced by a referential NP in prior discourse. In either case, (21) is ambiguous between assigning the accusative case pronoun either the agent role of the predicate cutting in an ECM 7 Without overt PP (18) and (19) would still entail that some cut Peter, existentially generalizing over the agent subject of the non-finite complement272 Optimal Reflexivity in Dutch construction, or the patient role, just as in (20), where the implicit argument in both cases is existential. The focus adverb alleen/only allows a coreferential interpretation of the accusative case marked pronoun acceptable in (22) (Binding condition A), as SE does not allow focus, cf.(o) above, and zichzelf would corefer with the local subject, the barber. (22) Peterj liet de barbier; alleen hem^ knippen. [Peter let the barber only him cut.] Peter let the barber cut only him. The focus meaning triggered by only creates a focus frame in the context, containing all situations of Peter letting the barber shave someone else, who is then uniquely identified to be Peter himself. The shaver may again be an implicit argument in Dutch and, if so, SELF is the preferred way to encode coreference with the subject of the main clause, as in (23), whereas SE would be ungrammatical, not allowing focus. If (22) suppresses the barber argument, the pronoun, still in focus, but now a co-argument of let, would again get a preferred disjoint reference reading. The binding conditions prevail over focus structure, just as one expects the strong constraints of syntactic configurationality to overrule the softer constraints of semantic interpretation and information structure. If SELF in (23) is assigned a nonagentive role, this interpretation yields a strong logical equivalence between (23) with PP and one interpretation of (24). (23) Peterj liet alleen {zichzel4J/{*zichiJ knippen (door iemand;). [Peter let only SELF cut by a barber.] Peter let only himself get cut by a barber. (24) Peterj liet iemand; alleen hem^ knippen. [Peter let a barber only him cut.] Peter let a barber cut only him. Comparing (23) and (24) in economy, (23) encodes efficiently the focus meaning with reflexive, and does not require the agent to be expressed overtly in PP. Lacking the agentive PP, (23) creates an ambiguity, assigning SELF either agent or patient role of knippen. In contrast (24) requires an overt agent of the non-finite predicate and it is threefold ambiguous, between (i) interpreting the pronoun to refer to Peter, (ii) to a deictically determined referent, or (iii) to a referent introduced by prior discourse. The table in Figure 1 summarizes the results of the observations of (i7)-(24), where the NPs are indexed as follows: i=John, j = the barber, k = (reflexive) pronoun; +NPj means that the NP is overtly expressed, and 8 Jack Hoeksema pointed out to me that the focus context would require a coreferential nonreflexive pronoun. His remark led to a significant clarification of the OT analysis. Alice G. B. ter Meulen 273 (17) (i*j) (18) (i*j)-NPj> (i^j) + NPj (19) (i=k),(i*j)-NPj> (i=k),(i*j) + NPj (20) (i=k), (k=patient) NPj > (i=k), (k=patient) + NPj > (i=k), (k=agent) (21) (i * k), (k=patient), (i *j) NPj > (i*k),(k=agent), (i (22) (i*j),(i*k)-NPj> (i*j), (i=k) + NPj > ( j ) , ( ) j (23) (i*j), (i=k), (k=patient) NPj > (i*j),(i=k)(k=agent)-NPj> (i*j), (i=k), (k=patient) + NPj (24) (i*j),(i=k)> Figure I (Non)coreference conditions rankings in —NPj that the NP remains implicit, idem for +/—PP. The symbol > indicates the preference order, where the left argument is preferred over the right argument. The indentation aids in making the levels of preferences transparent. The table does not provide much analytical insight yet into the semantic forces at work in arriving at these interpretations. Nor does it separate the speaker's economic interests in production from those of the consumer of the information. In a traditional or in a dynamic model theoretic semantics coreference conditions would either be imposed by constraints over possible continuations of given assignment functions and meaning-postulates, or, as in DRT, in the design of the representation by the construction rules. To offer a more insightful explanation into the data and speaker and hearer preference rankings requires an approach in which these speaker and hearer constraints are fully explicit and for each possible interpretation of a given form it is determined how it strikes an optimal balance between them. It is beyond the scope of the present paper to do this in all requisite detail, but some first steps of such an OT account are outlined in the next section. 5 OPTIMIZING CONSTRAINT SATISFACTION In Optimality Theory (17) is considered simplest and most economical, even though it does not make it clear whether John is cut voluntarily or against his wishes. Lacking explicit information about who is in control, the 274 Optimal Reflexivity in Dutch hearer will assume that the subject has the control over the described situation, as stated in default constraint Ci. Ci . The subject normally controls the action described by the predicate. This constraint Ci cuts to both sides, as the speaker uses it in economizing on his words and utterance effort in suppressing control information when the agent is in control, and the hearer derives from it the information that the controller is the agent, unless there is information to the contrary. This can be made explicit in the speaker constraint SCi and hearer constraint HCi below. SCi. If the agent is in control, no need to say so. CCi. If there is no control information, the controller is the agent. The hearer would have to continue after (17) with (25), if he wanted to find out whether Peter consented to being subjected to the action. (25) Vond Peter dat goed? [Found Peter that good?] Did Peter consent to it? To continue with (25) after (19) would be markedly redundant, OF at best demonstrate ignorance of the meaning of let, for let already assigns the control role to its subject. The default constraint that the subject controls the action, SCi, is flouted by (19), and it also adds expenses on the part of the speaker for having to use two predicates instead of just one. But (19) saves the hearer from having to ask (25) and the speaker again from answering it. The semantic content of a light verb assigns another thematic role to its subject, besides the role it was assigned in situ by the non-finite predicate. If a light verb is analysed as a relation between its external argument and an event interpreting the descriptive predicate in its complement, this inviolable constraint is captured as lexical meaning-postulate in (26). (26) If P is an element in the set of light verbs {HAVE, MAKE, GET, LET, . . .}, then VP VQ Vx Vy Ve Ve'[(P(x,e) & Q(e', . . . ,y, .. .) & e = e' & y = x) =>• th(x) ^ t/i(y)] The constraint SC2 that uttering fewer words is preferable for a speaker is another soft constraint, for it is meaningfully flouted in (19). The hearer's 9 The constraint that fewer words are preferable to more is a very soft constraint on the economy considerations of the producer (speaker), as effort in pronunciation is a relatively cheap commodity. Of course, saying nothing would be most economical, but that limit is ordinarily eliminated by the overarching human desire to be understood by others. Alice G. B. ter Meulen 275 corresponding constraint HC2 requires him to rely on default rules whenever different interpretations are possible, a meta-constraint that leads him to use CCi for the interpretation of (17). SC2. Using words costs effort (the more complex they are, the more they cost). HC2. Use default constraints when ambiguities arise. Together with the referential hierarchy and the strong configurational Binding condition B (co-arguments of a predicate that are not reflexive marked must be interpreted as disjoint in reference), we derive that (17) is optimized for a situation where the barber controls his cutting someone else, called Peter. In comparison, to describe the same situation (18) without agentive PP gains over (17) in SC2, and still supports SCi. The implicit indefinite argument controls the action, and requires a novel referent, so must be interpreted as disjoint. If (18) without PP is continued with a definite description referring to the barber, whose presupposition is accommodated by resolving it with the inferred existence of the agent argument, the information that the cutter was a barber is recovered, as in (27). (27) The barber used a razorblade. If (18) is used with PP, the same information can be expressed by continuing with (28). (28) He used a razorblade. Perhaps simplistically counting words, (18) —PP + (27) is optimized for SC2, counting — 1 (drop 3 words, add 2), and (18) + PP + (28) is less optimal, counting +4 (add 3 words and add 1). But in the referential hierarchy (27) with a full referring expression is lower down or less efficient than (28) with only a pronoun, so (27) demands more mental processing effort of the hearer, i.e. making an inference and accommodating a presupposition and identifying the referents of both as the same. Referring to someone, i.e. updating the common ground with a new referent, lightens the hearer's load, when that referent is referred to again in subsequent discourse. For the hearer (18) + PP + (28) would be more optimal than (18) — PP + (27), if the pronoun in (28) could not be interpreted otherwise. But, besides deictically, the pronoun can also be interpreted as coreferential with Peter, even though conflicting descriptive information would result. That prevents the hearer from chosing that interpretation, as the over10 The conflict arises as an incompatibility between the meaning of 'let' and the meaning of the descriptive predicates, whereas using a razorblade is a specification of the way in which Peter was cut 276 Optimal Reflexivity in Dutch arching constraint tells him to avoid contradictions, (HC3), assuming that the speaker wants to be consistent, (SC3). SC3 Be consistent and coherent. HC3 Avoid contradictions. Although (19) with PP may be most informative, for it entails both (17) and (18), in OT terms (18) is optimal, unless John controls the action. Pronouns with overtly asserted referential antecedents are optimal for the consumer to create coreference chains in discourse, in comparison to definite NPs with implicit arguments as antecedents. The latter are optimal for the speaker in economizing on words, i.e. someone who weighs SC2 more heavily. In section 6 below it will be argued, however, that quantified antecedents may reverse this OT ranking for creating discourse cohesion. In ranking (19) with (2o)-(24), (19) economizes in using SE instead of SELF in (20), as morphologically words are more expensive for the producer (SC2), and (19) wins over (20) in being higher in the referential hierarchy for the consumer. Also (19) entails (20) and (21), and not vice versa, as in (20) and (21) SELF or the accusative pronoun may be assigned the agentive role, but in (19) SE cannot support that role. Here OT assesses (19) as the winner, and any model theoretic account would consider (19) to have the strongest truth functional content as well. Finally, in comparing the three focus constructions (22)-(24), (23) is optimal in using SELF, since SE is unacceptable here, with the agent argument in optional PP, referring disjointly as it is indefinite. In contrast, (22) and (24) use a pronoun, that may, but need not corefer with the subject, so they cannot entail (23). 6 CONSTRAINING QUANTIFICATION With a quantificational antecedent the economic advantages of SEreflexives become even more apparent in the Dutch data, when constrasted to the fully acceptable bound pronouns in their English translations. Negative quantifiers (left and right decreasing) in subject position cannot bind non-reflexive pronouns in non-finite complements of light verb constructions, cf. (37). (29) a. Een barbier knipte iedere student. A barber cut every student. b. Iedere student werd geknipt (door een barbier). Every student was cut (by a barber). Alice G. B. ter Meulen 277 (30) Iedere student liet zich knippen (door een barbier). [Every student let SE cut (by a barber).] Every student let himself get cut (by a barber). (31) Iedere student liet een barbier hem knippen. [Every student let a barber him cut.] Every student let a barber cut him. The active (29a) and passive (29b) are parallel to (17) and (18), although the definite was replaced by an indefinite here to address scope issues. Scope interpretation is facilitated in (29) by intonation, where H*LL% (prosodic pitch contour H(igh) + stress L(ow) L(ow) boundary) on een barbier would change its meaning into the focus meaning the cardinal one barber. The OT advantage of (30) is first of all seen in C2, when reassigning the control to the student argument with the light verb construction. Making the barber argument implicit in (30), it is existentially quantified and must remain in the scope of the quantified subject, that now binds the SE reflexive. Accordingly, (30) without PP is optimal, as it avoids the ambiguity between the asserted NPs, present in (29a) and in (29b) with PP. As no argument within the nuclear scope of the quantificational subject would bind pronouns in subsequent discourse, any NP anaphoric to it must accommodate its presupposition within that domain. If a definite description is interpreted as continuing reference to the implicit agent of the cutting, its presupposition may be locally accommodated within the nuclear scope of the non-finite predicate of the preceding sentence, as in (32). This is strongly supported by the descriptive content, as using a razorblade specifies the way in which the students were cut. The quantified subject of (30) thus extends its subordination to the following sentence, even though ordinarily quantifiers have no dynamic effect in discourse and restrain their binding potential to sentence internal dependencies. (32) De barbier gebruikte een scheermes. The barber used a razorblade. Of course, (32) also allows for a globally accommodated existential presupposition, introducing one barber cutting all the students. But this would require considerably more processing effort from the consumer, as he would also have to force wide scope to the existentially quantified implicit argument to make it corefer with the accommodated presupposition introducing the barber as the one doing the cutting. From OT perspective the same information could much more economically be expressed using (33). (33) a. Een barbier (H*LL%) knipte iedere student. Hij gebruikte een scheermes. b. One barber cut every student. He used a razorblade. 278 Optimal Reflexivity in Dutch Asserting a referential NP optimizes discourse coreference, whereas accommodating presuppositions of anaphorically interpreted definite descriptions is relatively more costly. Since (33) is optimal for a wide scope barber, (30) — PP continued with (32) is optimal for a narrow scope interpretation of the barber. (34) * Iedere student; liet een barbier hern, knippen. Every student let a barber him cut. Every student let a barber cut him. When a non-reflexive pronoun is used in Dutch as in (34), it is considerably harder, if not outright unacceptable, to interpret it as bound by the quantificational subject, than as coreferential with a proper name. This is in accordance with the predictions of the referential hierarchy. In (34) the pronoun is preferred to refer to a referent available in the situation of use or to one available from memory, as (30) is already optimal in using a reflexive pronoun and making the agent of the predicate implicit, confining it to nuclear scope. In English translation of (34), however, the bound interpretation of the pronoun is certainly acceptable and even preferred over non-local interpretations. This is a sharp contrast with the economical advantages of the Dutch SE-reflexive. When the quantifier is negative, decreasing in both the denotation of the noun and the predicate, the reflexive is the only option in a light verb construction (36), as the negative quantifier in (37) does not bind the pronoun in Dutch, though the English translation is fine. (35) a. Een barbier knipte geen student. A barber cut no student. b. Geen student werd geknipt (door een barbier). No student was cut (by a barber). (36) Geen student liet zich knippen (door een barbier). [No student let SE cut (by a barber).] No student let himself get cut (by a barber). (37) *Geen student; liet een barbier hem; knippen. [No student let a barber him cut.] No student let a barber cut him. The referential hierarchy assigns (36) a higher value than (37), which matches the Dutch data better than the English ones, as the English translation of (36) takes a toll on word count (C2) over (37). English, lacking SE reflexives, makes word count outweigh the referential hierarchy. Alice G. B. ter Meulen 279