Janayna Carvalho (Universidade de São Paulo) & Ana Paula Scher (Universidade de São Paulo)

Refining the structure of unmarked middles: evidence from Brazilian Portuguese i.Introduction and claim: Lekakou (2005) argues that the middle interpretation is achieved by generalizations over a property of the patient/theme argument. This definition tells apart middles and habitual sentences. Whereas middles are generalizations over inherent properties (no actual event needs to take place), habitual sentences are generalizations over repeated events ((1) vs. (2)). The fact that middles are compatible with modal semantics and habitual sentences are not can be implemented with a generic operator (GEN) above VP in middles, in the same way modals are analyzed (cf.Lekakou 2005; Bresnan 1993). 1. This book reads easily→ This book can be easily read. 2. John works every day ≠ John can work every day. Fábregas and Putnam (2014) claim that the operator above VP in middle sentences blocks the existential closure by T. That is why middles are not instantiated in a particular time interval. Thus, if a language introduces event variables into the derivation to form middles, an alternative element must bind the variable, otherwise the derivation crashes. In the absence of a time interval, expressions that point to the presence of a participant are eligible to this task. We will claim that languages that generate unmarked middles can either introduce or not event variables into the derivation. On the basis of a contrast between the properties of Brazilian Portuguese unmarked middles (BPM) and English middles (EnM), we will show that several properties follow from the fact that EnM have this variable whereas BPM do not. ii. Similarities between BPM and EnM: According to Ackema and Schorlemmer (2006) / AS, there are at least two type of middles. Type I middles Type II middles English –This book reads easily French – Ce livre se lit facilement. Only agentive verbs Experiencer and stative verbs as well Do not accept subject-oriented expressions Accept subject-orientedexpressions Table 1: properties of middles contrasted

(3) and (4) demonstrate that the loss of se in BP has made middles in this language pattern with type I middles. 3. Matemática (*se) sabe bem (quando se estuda muito). Math se knows well (when se studies a lot). ‘One knows math well when he/she studies a lot.’ 4.Receita de bolo (*se) prepara com atenção. (Pacheco 2008) Recipe of cake se prepare with attention. iii. Differences between BPM and EnM: In spite of the fact that BPM and EnM fall inside the same category according to AS’s typology, it is not possible to say that they have the same structure. We point four factors showing that the agent interpretation can be recovered in EnM, but not in BPM. The first factor to take into account is that, in the rare cases in which English uses two different verbs for intransitive and transitive examples, the transitive form is used in the middle sentence. 5. Chickens kill easily. (Keyser and Roeper 1984) This is evidence in favor of the approaches that argue for an empty Voice projection in EnM (e.g. Schäfer 2008, Bruening 2012). In BP, however, the unaccusative variant (morrer) is preferred over the transitive variant (kill) for the middle interpretation. If the transitive variant is employed, as in (7), the most likely reading will be ‘chickens habitually kills someone/something’, an object drop reading. 6. Frango morre fácil. (middle reading) 7. Frango mata fácil. (object drop reading) Chicken die easy. Chicken kill easily. ‘Chicken dies easily.’ Chickens kill (people) easily. In addition, several apparent transitive verbs participate in possessor raising structures in BP. Crucially, in this type of construction, the possessor lands in a position external to vP when

moving to T. This movement is forbidden if this position is already taken by the external argument. Hence, this construction is compatible with unaccusative verbs (Lunguinho 2006; Rodrigues 2010). The fact that it is possible to introduce the agent of the action as an adjunct demonstrates that there is only one semantic participant in the event, although there are two syntactic DPs. 7. A Maria cortou as unhas com a melhor manicure da cidade. The Maria cut the nails with the best naildresser in the town. ‘Maria had her nails cut by the best naildresser in town.’ Since a considerable amount of verbs that participate in middle formation allow possessor raising structures (operar – operate, extrair – extract, vender – sell, encerar – wax, etc), the middle formation in BP is argued to be built from the unaccusative version of verbs. The second factor is the licensing of gerundive sentences in EnMs (8), and the third is the licensing of anaphoric elements (9). Importantly, (8) and (9) do not have possible counterparts in BPM. 8. Most physics books read poorly even after reading them several times. (Stroik 1995) 9. Books about oneself never read poorly. (Stroik 1995) These facts were part of a debate about whether the adjunct gerundives and the anaphora elements are licensed logophorically or whether these elements point to the presence of a PRO in the structure of EnMs. Be that as it may, those sentences can be used as evidence for the existence of an element in EnM that allows the attachment of anaphors and gerundive sentences without an overt agent.The fourth factor is the adverb modification. At first glance, adverbial modification in EnM and BPM seem to be alike. However, whereas middles in English must be modified by an adverb ending in –ly, there is an alternation between-mente adverbs and the adverbial use of adjectives in BP (10), with no impact in interpretation (cf. Lima 2010). The adverbial use of the adjective (fácil in (10)) is argued to be a very low type of modification applied to the VP-level (cf. Leung 2006; Sedrins & Lima 2013) and it is incompatible with se-middles (Rodrigues 1998), where the agent is syntactically represented. 10. Chão de escritório limpa fácil/facilmente Floor of office clean easy/easily ‘The [type of] floor in offices cleans easily’ In contrast, in EnM, adverbs that modify the way agents act, as comfortably in This chair seats comfortably, and poorly in (9), are licit. BPM do not license adverbs that modify the action performed by an agent. This suggests that adverbs in EnM attach higher than in BPM. A supporting fact for this assumption is the licensing of for-PPs in EnM. As Stroik (2006) demonstrates, the sentence No latin text translates easily for me is compatible with I generally do not translate any Latin text easily, which shows that for-PPs have external argument-like properties and the middle adverb is modifying this potential subject. iv. Summary: The overall picture emerging from this comparison shows that, despite the absence of overt external argument in EnM, this element is semantically retrieved. One way to account for that is to argue for the mandatory presence of an event variable in EnM, but not in BPM. As middles are not instantiated in a time interval, an alternative way to quantify over the event variable is syntactically building a potential agent. Therefore, all properties from EnM follow from the necessity to bind the event variable in English, which creates an agent interpretation. In BP, properties such as the low adverbial modification and the unaccusative basis follow from the lack of this variable.These properties give us at least two different structures for unmarked middles. 11. GEN [VoicePVoice[VP V DP]] = English 12. GEN [VP V+ DP]] = Middles in BP Selected references: Ackema, P. &Schoorlemmer, M. (2005). Middles. In: The Blackwell companion tosyntax● Fábregas, A., & Putnam, M. (2014). Middle voice structures with and without agents.The Linguistic Review,31(2). ● Lekakou, M. (2005).In the middle, somewhat elevated. PhD Dissertation.