Authors: Giada Palmieri, Renato Basso, Júlia Nieto i Bou, Yoad Winter, Joost Zwarts
Categories: Article, Reciprocity, Reflexivity, Romance, Arity-reduction, Lexicon
Source: Natural Language & Linguistic Theory
Authors: Giada Palmieri, Renato Basso, Júlia Nieto i Bou, Yoad Winter, Joost Zwarts
In many languages, reciprocal meanings are expressed either by grammatical means or by using lexical predicates. The grammatical strategy is productive and may involve derivational affixes (Swahili -an) or pronouns (English each other) with transitive forms, whereas lexical reciprocity is expressed by a restricted class of intransitive predicates like English kiss or meet. The situation is more complex in Romance languages, where reciprocal verbal constructions often require a se clitic without a clear separation between transitive and intransitive forms. Addressing this puzzle, we propose that Romance languages involve a grammatical/lexical distinction as in other languages. We show that numerous Romance constructions systematically allow se omission with certain reciprocals, exhibiting parallel properties to those of lexical intransitives in other languages. A similar observation is made in relation to the distinction between grammatical reflexivity (e.g., English oneself) and lexical reflexives (wash, shave). Furthermore, we show that the se requirement may also be relaxed with transitive verbs, when reciprocity or reflexivity is conveyed by an overt reciprocal/reflexive item (e.g., Spanish mutuamente ‘mutually’). The emerging theoretical picture supports an analysis of se as a head projection that licenses arity-reduction, though language-specific conditions allow se omission when arity reduction is achieved by a lexical reciprocal item or by another overt reciprocal element.
The way in which languages express reciprocal meanings received attention both in theoretical linguistics and typological studies (Frajzyngier and Walker 2000; Nedjalkov et al. 2007; König and Gast 2008; Evans 2011). Two reciprocal strategies have been identified cross-linguistically: lexical reciprocity and grammatical reciprocity. Lexical reciprocity is expressed by a restricted class of intransitive predicates without pronominal elements or other productive derivational strategies. In English, lexical reciprocity is often realized with zero morphology (1). This kind of reciprocity is restricted to verb meanings in the conceptual domain of ‘natural reciprocals’ (Kemmer 1993) and is not possible with just any transitive verb (2).1
(2)
Grammatical reciprocity is the productive strategy by which pronouns, adverbs or affixes lead to reciprocal interpretations with all transitive verbs. In English, grammatical reciprocity requires the elements each other or one another (Dalrymple et al. 1998) as in (3) and (4).
(3)Mary and Lisa kissed each other. (4)Mary and Lisa described each other.
Lexical and grammatical reciprocals lead to different interpretations (Kemmer 1993; Carlson 1998, inter alia): the grammatical reciprocal construction in (3) can be interpreted with different unidirectional events (for instance, Mary and Lisa kissing each other consecutively on the forehead), whereas its intransitive counterpart in (1) refers to a single collective event (one mutual simultaneous kiss). This semantic contrast reflects a difference in argument grammatical reciprocals are treated as predicates with two arguments bound by a reciprocity operator, and lexical reciprocals as intransitive predicates with one semantically plural argument (Langendoen and Magloire 2003; Dimitriadis 2008a; Winter 2018).
The lexical strategy is overtly manifested in several languages, using morphological forms that characterize reciprocal meanings. For instance, while lexical reciprocals are realized with zero morphology in English and Dutch (Reinhart and Siloni 2005), in Hebrew they are usually realized in the hitpael template (Doron 2003), in Modern Greek with non-active morphology (Doron and Rappaport Hovav 2009) and in Hungarian with the verbal marker -oz (Rákosi 2008).
In some languages, however, reciprocals do not seem to show any clear distinction between lexical and grammatical processes. This is the case with Romance languages, where the clitic si/se (in its different realizations, henceforth se) is generally required in sentences with transitive verbs that receive a reciprocal interpretation. Such sentences typically receive an additional reflexive interpretation, which is illustrated by the Italian examples in (5).2
(5) From the surface realization of sentences like (5) we cannot determine whether their interpretation is derived using lexical intransitives such as English hug, or using complex transitive constructions with a reciprocal/reflexive operator.
A similar puzzle appears with Romance reflexives. For instance, unlike the English distinction between intransitive wash and grammatical reflexives like wash oneself or describe oneself, Italian generally requires a se clitic. Accordingly, the interpretation of plural sentences like (6) is either reflexive or reciprocal, without a clear distinction between lexical and grammatical reflexive usages of wash:
(6)
The possible existence of lexical reciprocals and reflexives in Romance has been occasionally mentioned in the literature (Labelle 2008; Doron and Rappaport Hovav 2009; Siloni 2012), but little attention has been dedicated to their characterization. Given the lack of morpho-syntactic cues enabling their immediate identification, this is not a straightforward task. Furthermore, due to the uncertain distinction between grammatical and lexical reciprocal/reflexive processes in Romance, the role of the clitic se has been predominantly studied in relation to the grammatical strategy. In this paper we show that the lexical/grammatical distinction systematically appears in Romance, with similar effects to those familiar from other languages. After distinguishing between the lexical and the grammatical reciprocal strategies, we focus on some general aspects of their semantic analysis, and on the contribution of se to these strategies. A parallel analysis, with similar motivations, is provided for the lexical/grammatical distinction with Romance reflexives.
When concentrating on the distinction between lexical and grammatical reciprocity/reflexivity, we substantiate the argument in favor of a class of Romance predicates that have a transitive entry and an intransitive reciprocal/reflexive entry, similarly to other languages. Despite the absence of an overt distinction in Italian finite clauses like (5), we find syntactic environments where predicates from a closed class lead to reciprocal/reflexive interpretations all by themselves, without se or another overt reciprocal/reflexive element. A language that is especially convenient for illustrating this finding is Brazilian Portuguese (BP). In BP, the clitic se is productively associated to reflexivity and reciprocity, similarly to other Romance in (7), se is mandatory for the grammaticality of the clause, which gets a reflexive or a reciprocal reading, just like the Italian example in (5). However, there is a handful of verbs that allow a reciprocal interpretation in their bare intransitive form. One example is abraçar ‘to hug’ in (8): it receives a reciprocal interpretation either with se (8a) or without it (8b).
(7)
(8)
The two configurations with abraçar in (8) differ in two respects. First, the reflexivity/reciprocity polysemy only emerges in (8a), whereas its counterpart without se in (8b) is unambiguously reciprocal. Second, the sentence in (8b) receives an interpretation that is reminiscent of the ‘single-event’ interpretation of English lexical reciprocals as in (1): (8b) describes a situation where Mary and Lisa are involved in a mutual hug, and it would not be true in a scenario where each of the girls hugged the other in a different moment (e.g., while she was asleep).
We propose that reciprocal interpretations without se as in (8b) are due to the lexical meaning of the intransitive verb stem (‘hug’), and that predicates similar to abraçar exist in all Romance languages. We focus primarily on data from BP, Catalan, Italian, and Spanish, showing that the possibility of BP to express reciprocity without se in (8b) is representative of a broader pattern in Romance. In all four Romance languages, we present constructions where verbs with similar meanings to English intransitives like hug lead to reciprocal interpretations without se. The peculiarity of BP within the Romance family is that se may be omitted in finite clauses, whereas in the majority of Romance languages this only happens in some non-finite constructions. We argue that this syntactic variation does not affect the semantic characterization of lexical reciprocal predicates, which is uniform across the Romance languages we studied. Whenever a Romance verb manifests a reciprocal meaning in some construction without se and without any overt reciprocal element, we show that the verb exhibits the semantic characteristics of reciprocal intransitives in English. We reveal a similar pattern in relation to across the four Romance languages that we study, we show that the same constructions that support lexical reciprocity without se also support reflexive meanings with verbs that have similar meanings to English lexical reflexives (shave, bathe).
After presenting evidence for the existence of lexical reciprocity and reflexivity in Romance, we move on to cases of grammatical reciprocity and reflexivity, and to the role of se in their derivation. We show that in the syntactic environments where lexical reciprocity and reflexivity emerge without se, grammatical reciprocity and reflexivity are also possible without se for all transitive verbs, provided that they appear with an overt reciprocal or reflexive operator. For instance, we saw in (7) that BP descrever ‘to describe’ cannot lead to a reciprocal interpretation with zero morphology. Yet, with the element um o outro ‘one another’, this verb does allow se omission, as in the following
(9)
Taken together, these facts lead us to propose that se itself is not an exclusive source of elements such as BP um o outro have the meaning of reciprocal operators, whereas the reciprocal interpretation of lexical verbs such as abraçar ‘to hug’ is due to the inherent intransitive meaning of the verb stem. We will show that analyses of se as a marker of valence-reducing operations (Reinhart and Siloni 2005) or as a reciprocal anaphor (Doron and Rappaport Hovav 2009) fail to account for the appearance of se with predicates that are inherently reciprocal, and to capture instances of grammatical reciprocals without se. By contrast, our findings are consistent with Labelle (2008)’s proposal that se is a functional head projection, more specifically Voice. However, against Labelle, we argue that se never carries itself the meaning of a reciprocal or reflexive operator. We propose that the role of se is purely it is a Voice head that marks reflexive and reciprocal predicates, by requiring that no external argument is introduced in Voice. Whether se is obligatory depends on the syntax of the clause and on the presence of other elements that mark reflexive/reciprocal interpretations. In the absence of a lexical reflexive/reciprocal verb or of an overtly realized reflexive/reciprocal item, se is required to satisfy Condition B, according to which a reflexive predicate must be reflexive-marked (Reinhart and Reuland 1993). Conversely, in the presence of a lexical reflexive/reciprocal verb or of an overtly realized reflexive/reciprocal item, only the syntactic requirements of the construction determine whether se is obligatory (e.g., in Italian finite clauses), optional (BP finite clauses) or disallowed (Italian causative clauses).
The paper is structured as follows. In §2 we provide an overview of previous work on lexical reciprocity in Romance languages. In §3 we describe the distribution of se across different constructions in four Romance languages, and we identify a group of Romance predicates that may receive a reciprocal interpretation without se. In §4 we show that Romance predicates that express reciprocity without se have semantic properties that are cross-linguistically associated with lexical reciprocal nominalization (§4.1), semantic drift (§4.2), pseudo-reciprocal interpretation (§4.3), discontinuous reciprocity (§4.4) and acceptability of singular group NPs (§4.5). In §5 we point out the parallelism with lexical reflexivity, considering predicates that lead to reflexive interpretations without se. In §6 we explore instances of grammatical reciprocity/reflexivity without se. Section §7 presents our analysis of se as a functional head projection, which accounts for its distribution and contribution to arity-reduction. In section §8 we provide general conclusions.
The term ‘naturally reciprocal’ has been used in the typological literature since Lichtenberk (1985) and Kemmer (1993) to refer to predicates that typically denote reciprocal configurations, and that are often realized with morphological markers associated with the middle voice. In Haspelmath (2007) and Knjazev (2007), the additional term ‘lexically reciprocal’ is employed to refer to the sub-group of the ‘naturally reciprocal’ predicates that express reciprocity without any overt marking (e.g., fight or quarrel in English). A more encompassing definition of ‘lexical reciprocals’ is provided by Nedjalkov (2007), who defines them as verbs “whose meanings is not a mere sum of the meaning of the base and the meaning of ‘each other’’’ (p.14).
Following Nedjalkov’s definition of ‘lexical reciprocals’, in this paper we use this term to refer to predicates whose reciprocal interpretation does not arise from a productive morpho-syntactic operation, but from an inherent collective meaning of the verb’s intransitive entry. Thus, predicates like kiss are assumed to have two distinct a transitive entry (10a) and an intransitive, lexical reciprocal, alternate (10b).
(10) Mary kissed Lisa.Mary and Lisa kissed.
Although many lexical reciprocals have a transitive alternate, some of them do not have a corresponding transitive form. For example, the intransitive verb talk does not have a transitive alternate, but an alternate that takes a prepositional complement (11a). The meaning relation between the two talk alternates is parallel to the kiss alternation in (10). Accordingly, we characterize the collective use of intransitive talk in (11b) as a lexical reciprocal.
(11) Mary talked to Lisa.Mary and Lisa talked.
We oppose this notion of ‘lexical reciprocity’ to ‘grammatical reciprocity’: a process whereby reciprocity is derived through a productive strategy, as with each other in English. Unlike lexical reciprocity, grammatical reciprocity in English is possible with all transitive verbs and verbs with prepositional complements, whether they have an intransitive reciprocal alternate (12a,12b) or not (12c).
(12) Mary and Lisa kissed (each other).Mary and Lisa talked (to each other).Mary and Lisa described *(each other).
When contrasting lexical reciprocity with grammatical reciprocity, we rely on the theoretical assumption that reciprocal meanings of verbs are lexically associated with the intransitive entry, without any necessary morpho-syntactic process. This assumption leaves room for the possibility that certain languages possess lexical reciprocals that are, on the surface, less distinctly differentiated from grammatical reciprocals than in English. Specifically, in Romance languages there is no clear morpho-syntactic marking that is reserved to lexical verbs with a transitive alternate usually require the element se in order to get a reciprocal (or reflexive) meaning. This clitic is not restricted to reflexivity and reciprocity, and it is also used to convey other typical functions of middle forms, including unaccusative, impersonal, passive, and subject-experiencer configurations (Cinque 1988; Chierchia 1995; Dobrovie-Sorin 1998; Rivero 2001; d’Alessandro 2008; Dobrovie-Sorin 2017). The role of the se clitic in Romance has been extensively studied in works investigating valence-reducing operations (Grimshaw 1980; Everaert 1986; Reinhart and Reuland 1993; Baauw and Delfitto 2005; Reinhart and Siloni 2005; Doron and Rappaport Hovav 2009; Labelle 2008; Labelle and Doron 2010, inter alia). However, the identification and characterization of lexical reciprocal predicates in Romance has not received much attention. This class of verbs has only been occasionally taken into account in theories of Romance reflexives and reciprocals (generally focusing on the former), and it is often treated as an orthogonal question to the grammatical realization of valence-reducing operations.
Reinhart and Siloni (2005) propose a ‘lexicon-syntax’ parameter, by which arity-reducing operations in any language may apply in the lexicon or in the syntax. Thus, in Reinhart and Siloni’s approach, the distinction between lexical and grammatical reciprocity/reflexivity plays a cross-linguistic role, but they do not elaborate on lexical/grammatical distinctions within one and the same language. Reflexivization is claimed to take place through the bundling operation, that maps an internal θ-role onto the external argument to form a complex θ-role. This operation of arity-reduction is illustrated in (13): reflexivization bundling turns a two-place predicate (with two θ-roles) into a one-place predicate (with one complex θ-role).
(13) In ‘lexicon languages’, such as English, Dutch or Hebrew, reflexivization and reciprocalization are not productive operations, and lexical reflexive and reciprocal predicates are distinguishable from their counterparts with anaphors. Reciprocal predicates formed in the lexicon are characterized by the absence of ambiguity with reflexive interpretations (Reinhart and Siloni 2005), by a ‘single-event’ interpretation, and by the availability of the discontinuous reciprocal construction (Siloni 2012).3 ‘Lexicon languages’ are opposed to ‘syntax languages’, where the reflexive/reciprocal strategy is productive and is assumed to take place in the syntax. This is for instance the case in the Romance here, Reinhart and Siloni (2005) propose that the clitic se is functional for the bundling operation. This element is therefore assumed to operate on the argument structure, and to be insensitive to the semantics of the verb. Within this account it is observed that there may be instances of lexical reciprocals in syntax Siloni (2012) notes that the French verb se battre ‘to quarrel’ displays syntactic and semantic characteristics typical of lexical reciprocals (such as the availability of the discontinuous construction and a ‘single-event’ interpretation). However, no special treatment is reserved to the role of se with respect to the reciprocalization of such se is uniformly analyzed as a marker of valence-reduction, regardless of the verbs it combines with. This leaves a gap in the without more semantic content, this does not account for the differences between verbs like se battre and other reciprocal verbs that are formed in the syntax.
Labelle (2008) proposes an advancement of the bundling theory, providing a unified analysis of se that accounts for cases where this element is responsible for reciprocal interpretations, as well as for cases where reciprocal readings originate elsewhere. Labelle (2008) observes that the French se obligatorily appears with verbs that express reciprocity or reflexivity on their own, such as predicates prefixed with entre- or auto-, respectively. Labelle assumes that entre- and auto- bind the internal and external arguments, yielding a verb entry with a reciprocal/reflexive interpretation. For instance, entreregarder is considered to already denote a mutual configuration, but it nonetheless requires se (14).
(14)
The possibility of se to appear with reciprocal verbs rules out a treatment of this element as an arity-reducing morphological unit. However, se also appears with simple transitive verbs, and in these cases it is considered responsible for the reciprocal interpretations. To address this distribution, Labelle (2008) treats the French se as a functional head projection that introduces the external argument x through the agent role (following Kratzer 1996) and identifies it with the object argument of the predicate P (15).
(15) This treatment is meant to unify cases where se is the source of reciprocity and cases where it is semantically redundant. With transitive verbs, se is needed to express reciprocal interpretations, co-referencing external and internal thematic roles. With predicates prefixed by entre-, where the lexical semantics of the verb already has an agent variable (introduced by the prefix), it is assumed that the external argument introduced by se is identified with the external argument provided by the verb’s entry. In such cases, se does not contribute to the reciprocal interpretation of the constructions, but it is considered obligatory to ensure a coherent interpretation.4 While Labelle (2008) recognizes the possibility of se to combine with predicates that are already reciprocal, this observation only relies on predicates bearing the productive prefix entre-; lexical reciprocal verbs are not considered, and no diagnostics for their identification is provided. One shortcoming of the reliance on grammatical reciprocity is that the analysis is based on the assumption that constructions with se are semantically there are always two distinct thematic roles, that are bound at some stage in the derivation (either by prefixation or by se). Semantically, this view is suitable for grammatical reciprocity, but it is in conflict with more recent observations on the meaning of lexical reciprocals, according to which the ‘single-event’ reading must originate from an intransitive entry (Dimitriadis 2008b; Siloni 2012; Winter 2018).
In Doron and Rappaport Hovav (2009), the distinction between lexical and grammatical reflexive/reciprocal entries is taken as a starting point for the development of a twofold account of the Romance se. Doron and Rappaport Hovav take the reflexive French se as a case study, and propose a syncretism of this element between reflexive morphology and reflexive anaphor. They provide morpho-syntactic arguments for treating the cases in which se combines with transitive predicates as instances of anaphoric binding. However, an analysis as a marker of argument identification is reserved to se when it is associated with lexical reflexive or reciprocal predicates. Such verbs are identified based on the fact that nominals derived from lexical reciprocals have an inherently collective meaning (16) and on the possibility to receive a reciprocal interpretation without se in causative constructions (17).
(16)(17) However, as we will show in this paper, various syntactic constructions in Romance allow grammatical reciprocity without se under certain circumstances. This fact challenges Doron and Rappaport Hovav’s twofold account of se, and will constitute a major element in the evidence that leads us towards an alternative, unified, approach to se.
We conclude that despite the overall agreement on the existence of a class of lexical reciprocals in Romance, there is currently no consensus on general tests for identifying those Romance predicates that have a lexical reciprocal (or reflexive) entry, nor on the theoretical implications of the existence of this class. The works outlined above propose different analyses of se and of its interaction with lexical reciprocity. However, they all agree on the idea that se is semantically responsible for deriving grammatical reciprocal strategies when it combines with verbs that have no reciprocal meaning of their own. In contrast with these previous studies, we will propose a unified treatment of se, where its role is purely syntactic as a marker of arity-reduction, and does not derive reflexive or reciprocal interpretations all by itself.
Across languages of the Romance family we can find predicates that express reciprocity without se; see Godoy (2008) for BP and Vázquez and Fernández-Montraveta (2016) for Spanish. The Italian predicate chiacchierare ‘to chat’ in (18a) receives a collective interpretation in its bare intransitive entry, and cannot combine with the clitic se. This configuration is restricted to verbs that do not have a transitive chiacchierare ‘to chat’ cannot take a direct object (18b).5
(18)
In the absence of any reciprocal marking, the reciprocal interpretation of (18a) must originate from the verb’s entry. Thus, verbs like Italian chiacchierare ‘to chat’ fit our definition of lexical they must be stored in the lexicon with an inherent reciprocal meaning. These verbs also fit within the proposed universal that all languages have predicates that denote mutual configurations by themselves (Haspelmath 2007).
However, it is unclear whether the categorization of lexical reciprocals in Romance may also be extended to predicates with a transitive alternate. In many cases, as in the Italian examples in (19), verbs with a transitive entry (19a) require se in reciprocal sentences (19b), regardless of whether they denote events typically associated to the class of ‘naturally reciprocal’ predicates (‘hug’) or not (‘describe’).
(19) The situation is similar in the vast majority of Romance predicates with a transitive entry consistently require se in finite clauses with a reciprocal interpretation. Despite the ubiquity of this phenomenon, the rest of this section will show that all four Romance languages that we studied exhibit constructions where certain verbs with a transitive alternate do express reciprocity without se. As we will see, these Romance verbs have meanings that are typical of ‘natural reciprocals’ cross-linguistically, and they give rise to similar semantic effects.
In BP finite clauses, most transitive verbs require se for obtaining a reciprocal or reflexive interpretation. When se appears with a transitive verb and a plural subject, BP sentences uniformly have both reciprocal and reflexive interpretations (20). This is a common situation in other Romance languages as well. However, there is a restricted class of BP transitive verbs where the reciprocal interpretation can also emerge without se. For instance, the verb abraçar ‘to hug’ expresses a reciprocal meaning both with se (21a) and without se (21b).6
(20)
(21)
Importantly, the two sentences in (21) differ semantically. The se-clause in (21a) displays the common Romance reflexivity/reciprocity it holds true if each individual in the denotation of the subject hugged the other or hugged herself. The meaning of the bare intransitive in (21b) is more it is only in line with one mutual, collective hug. For instance, unlike (21a), (21b) would not support a scenario with multiple unidirectional hugging events (for instances, where Mary hugs Lisa while Lisa is asleep, and later Lisa hugs Mary while Mary is asleep). Essentially, (21b) only supports a ‘single-event’ interpretation, similarly to its intransitive counterpart in English.
The possible omission of se in BP finite clauses has already been observed in the literature (Nunes 1995; Galves 2001; Cyrino 2007; Carvalho 2018), for example with anticausative verbs (22a), or in constructions that receive a medio-passive (22b) or impersonal reading (22c). To the best of our knowledge, however, the optionality of se has not been previously recognized with reciprocal or reflexive verbs.
(22)
Doron and Rappaport Hovav (2009) notice that in French causative constructions, some verbs with intrinsic reciprocal, reflexive or anticausative interpretations can express these meanings without se – although this element would be required in simple finite clauses. This observation holds in other Romance languages too. In Spanish, Catalan, and BP analytic causatives, se can be used on the embedded verb to express reflexivity or reciprocity. This process is virtually any transitive verb can be embedded in a causative with se, leading to reciprocal or reflexive interpretations. For instance, (23a) holds true if Mary and Lisa described each other or themselves. This is parallel to the situation we have seen in BP finite clauses with se in (20). Spanish analytic causatives allow se to be omitted, but then the direct object is interpreted as the theme of the action denoted by the embedded verb (Guasti 2006; Folli and Harley 2007). We characterize this as a ‘passive’ interpretation. For example, sentence (23b) means that the subject (“I”) caused Mary and Lisa to be described by an unspecified agent.
(23)
Unlike describir in (23b), there are Spanish predicates that allow a reciprocal interpretation in analytic causatives even if se is omitted. Let us consider the Spanish verb abrazar ‘to hug’. In finite clauses, this verb requires se for obtaining reflexive and reciprocal interpretations (24). The verb abrazar also gets a reflexive and a reciprocal reading in causatives with se (25a). However, unlike describir in (23), abrazar retains a reciprocal reading in causatives without se. Thus, sentence (25b) has a passive interpretation similar to (23b), but it also has a reciprocal reading where the subject (“I”) caused Mary and Lisa to be involved in a mutual hug. Note that in the absence of se, no reflexive interpretation emerges in (25b): the sentence is only in line with a passive or a reciprocal interpretation.
(24)
(25)
In contrast to Spanish, Italian causatives do not tolerate se (Zubizarreta 1985; Guasti 2006). With most transitive verbs in Italian, the only possible interpretation of analytic causatives is passive, similarly to most Spanish transitives in causatives without se. For example, sentence (26) is interpreted as claiming that the subject (“I”) caused an unspecified agent to describe Mary and Lisa. By contrast, and similarly to Spanish as well, a restricted set of Italian predicates receive a reciprocal interpretation without se in causatives. For instance, with the verb abbracciare ‘to hug’, sentence (27) receives a reciprocal interpretation (a mutual hug between Mary and Lisa) on top of the canonical passive interpretation (Mary and Lisa being hugged by an unspecified agent).
(26)
(27)
In Spanish and Catalan, another construction reveals the possibility of some verbs to express reciprocity by the absolute construction with participials, which does not allow se in these two languages. When an absolute clause presents a participial followed by an NP, its default interpretation is passive (Hernanz 1991; De Miguel and Lagunilla 2000). For instance, the Catalan example in (28a) states that Teo and Ana left the conference after having been thanked by an unspecified agent. However, with certain verbs, a reciprocal interpretation is available in absolute constructions. Consider for instance the verb abraçar ‘to hug’ in (28b), which has an interpretation where Teo and Ana are hugged by a third party, as well as an interpretation where they are involved in a mutual hug.
(28)
Relying on the data presented in this section, we can characterize three groups of predicates in Romance, summarized in Table 1. Table 1Three classes of Romance verbscombines with sereciprocity by itselfexamplereciprocal intransitive–+‘chat’transitive+–‘describe’reciprocal intransitive / transitive++‘hug’
In the first class we find verbs like ‘chat’ (18) and ‘discuss’. These verbs do not have a transitive entry, they cannot combine with se, and they invariably express reciprocity without any grammatical marking. Similarly to their English counterparts, they get a collective interpretation in their bare intransitive entry.
In the second class there are transitive predicates that combine with se and cannot denote reciprocal interpretations without se or other additional elements. This is the case of verbs like ‘describe’ or ‘thank’ (20,26,28a). We propose that these verbs are unambiguously transitive, hence they can only express reciprocity through a productive grammatical strategy.
Verbs in the third class have a transitive entry that combines with se, but in certain syntactic environments they also get reciprocal readings without se or any other additional element. The constructions where omission of se is allowed, or even obligatory, differ per language (Table 2). Despite this syntactic variation, the meanings of these verbs are remarkably similar to those of lexical reciprocals in other languages; they all fall into Kemmer (1993)’s categorization of ‘naturally reciprocal’ predicates. Furthermore, in the absence of se they all unambiguously lead to the type of ‘single-event’ readings that characterize lexical reciprocals cross-linguistically (§4.3). When reciprocity emerges from the verb alone, the reflexive/reciprocal ambiguity that is typical of Romance se-constructions disappears. A summary of the environments where se can be omitted with these verbs is given in Table 2. Table 2Constructions where reciprocity emerges without seBPItalianSpanishCatalanfinite clauses+–––analytic causatives++++absolutes with participial––++
Based on these observations, we take reciprocal readings without se (and without any other reciprocal marking) to be an indication of lexical reciprocity. We propose that verbs that allow reciprocity with and without se have two a transitive entry and an intransitive entry with a lexical reciprocal meaning. A common example of the kind of meaning that is often conveyed by such verbs is ‘hug’; verbs with this meaning have a transitive alternate and can express reciprocity without se in all four languages studied here (21b,27,25b,28a). (29) provides a more comprehensive list of English translations of lexical reciprocal verbs with a transitive alternate in these four BP, Catalan (c), Italian (i), and Spanish (s).7
(29) **Lexical reciprocals with a transitive ** ‘break up’ (c,i,s); ‘confer’ (bp,c,i,s); ‘date’ (bp,i); ‘greet’ (bp,i); ‘hug’ (bp,c,i,s); ‘kiss’ (bp,c,i,s); ‘know each other’ (i); ‘marry’ (bp,c,i,s); ‘meet’ (bp,c,i,s); ‘run into each other’ (c,i,s).
We should stress that under our assumptions, the possibility of a verb to denote reciprocity by itself can be used as a diagnostic for having an intransitive reciprocal entry only under two (i)Lexical reciprocals with a transitive alterate do require se in many environments other than those in Table 2. For example, the Italian verb abbracciare ‘to hug’ expresses reciprocity without se in causatives (27), but requires se in finite clauses (19b).(ii)As will be discussed in §6, reciprocal interpretations may emerge in the absence of se in constructions as in Table 2, as long as there is an overt reciprocal element. An example are BP finite if the reciprocal pronoun um o outro ‘one another’ is present, se can be omitted with all transitive verbs (e.g., descrever ‘describe’ in (9)). With these caveats, we formally define ‘lexical reciprocals’ in Romance as (30)Romance lexical reciprocals: In a Romance language, we characterize as lexical reciprocals those verbs for which there are syntactic constructions (whose identity is determined by language-specific factors) where a reciprocal interpretation emerges without se or another reciprocity element. With this notion of lexical reciprocals in Romance, the next section demonstrates that these predicates share semantic properties with lexical reciprocals in other languages.
The properties of lexical reciprocal predicates have been explored in many works, typological (Kemmer 1993; Knjazev 2007; Haspelmath 2007), theoretical (Rákosi 2008; Dimitriadis 2008b; Doron and Rappaport Hovav 2009; Siloni 2012; Winter 2018), and experimental (Gleitman et al. 1996; Kruitwagen et al. 2022). In this literature there is an agreement that lexical reciprocals have a different interpretation from their grammatical counterparts, and that they may appear in constructions where grammatical reciprocity is blocked. In this section, we review these properties and show that they consistently appear with the Romance verbs that we characterize as lexical reciprocals. This supports our claim that the ability of a Romance verb to express reciprocity without additional elements reflects the same phenomenon that is cross-linguistically characterized as lexical reciprocity.
One property of the Romance predicates that we characterize as lexical reciprocals is the possibility to form nominals with a reciprocal interpretation. Doron and Rappaport Hovav (2009) notice that certain French verbs that they consider lexical reciprocals can be nominalized and keep a reciprocal interpretation (see (17) in §2). This observation can be extended to other Romance languages. Consider for example the Italian verb incontrare ‘to meet’, characterized as a lexical reciprocal in (29). The nominal derived from this verb has an inherent reciprocal its use in (31) refers to a meeting between Mary and Lisa. (31)
However, verbs often miss nominalized forms. For this reason, nominalizations do not always provide an opportunity to test the verb’s reciprocity. For example, the Italian predicate lasciare ‘to leave/break up’ cannot be nominalized, although it can express reciprocity without se in causatives, and it has a meaning that is cross-linguistically common among lexical reciprocals. Thus, although we adopt Doron and Rappaport Hovav’s proposal that reciprocal nominalization can only appear with lexical reciprocal verbs, we emphasize that this property is not a necessary characteristic of lexical reciprocals under our diagnostics.
A common phenomenon among verbs that we categorize as lexical reciprocals is that they do not fully preserve the meaning of their transitive alternate. For example, the Italian verb trovare has a transitive entry with the meaning ‘to find’ (32), as well as a logically distinct intransitive ‘to have a planned meeting’ as in (33). As with all transitive predicates, the ‘find’ meaning of trovare can receive a reciprocal interpretation through the grammatical strategy as in (34).8
(32)
(33)
(34)
Semantic drift has been observed for verbs with an inherent reciprocal meaning across several languages (Kemmer 1993; Haspelmath 2007; Siloni 2012). This phenomenon results in reciprocal verbs not having a corresponding transitive base; accordingly, such verbs cannot be analyzed as the outcome of a productive strategy where a reciprocity operator applies to a transitive entry. Similarly to nominalizations, we take semantic drift to be an indication of a lexical reciprocal entry, even though we do not expect it to be a necessary characteristic of all verbs of this class.
Cross-linguistically, grammatical and lexical reciprocity lead to different interpretations. In events with two participants, grammatical reciprocals describe two different events, where in each event the same binary relation holds between the participants in a different direction. The resulting reciprocity is the accumulation of these different ‘unidirectional events’. By contrast, lexical reciprocals describe a single collective event (Carlson 1998) that typically involves two ‘unidirectional sub-events’. To illustrate this contrast in English, let us consider the grammatical reciprocal form with each other in (35a). This sentence is in line with an interpretation that involves different kissing events, where Mary kissed Lisa and Lisa kissed Mary (e.g., on the forehead). The two events do not have to be simultaneous or related to some collective act of Mary and Lisa, but they can. By contrast, the lexical reciprocal form in (35b) does not allow two independent unidirectional kisses and it can only describe a single kissing event between the two people (e.g., a romantic kiss on the lips).
(35) Mary and Lisa kissed each other.Mary and Lisa kissed.
Grammatical reciprocity is central to studies that explore the core meanings of reciprocal elements like each other, and their relation with contextual information and predicate concepts (Dalrymple et al. 1998; Beck 2001; Sabato and Winter 2012; Mari 2013; Poortman et al. 2018). We will not delve into all the possible configurations supported by English each other, nor into the contrast between “weak” (=partial) and “strong” (=maximal) reciprocity. For the sake of simplicity, we will restrict our attention to reciprocal configurations involving only two entities. For situations with more than two participants, we assume that se-less Romance sentences are interpreted like the parallel intransitive sentences in English. The exact interpretation of such sentences in relation to weak or strong reciprocity has not been studied in the literature even with respect to English, hence we have to put it aside.9 Our only prediction here is that whatever reciprocity mechanism is uncovered with such reciprocal intransitive sentences in English, it should be uncovered in Romance as well.
Grammatical reciprocal forms with two participants (35a) generally lead to equivalences with a conjunction between two opposite ‘unidirectional’ statements, as in (36).10
(36)x and y kissed each other ⇔ x kissed y and y kissed x By contrast, the interpretation of lexical reciprocal predicates is not exhausted by this equivalence. Winter (2018) illustrates that different lexical reciprocals show different entailments between the collective intransitive form and the two unidirectional statements. Some lexical reciprocals are indeed characterized by a mutual entailment between the collective form and multiple unidirectional relations (37); this equivalence is defined by Winter as plain reciprocity. Winter argues that lexical reciprocals that show this equivalence, e.g., ‘meet’, have a symmetric transitive alternate, as illustrated in (38).
(37)x and y met ⇔ x met y and y met x (38)x met y ⇔ y met x
Many lexical reciprocals do not show plain reciprocity and are not symmetric. For instance, the reciprocal entry of the verb divorce in (39) does not entail two unidirectional a divorce can be initiated by only one individual.11 There are also lexical reciprocals for which the reverse entailment does not hold, i.e., where multiple unidirectional relations do not entail a collective in (40) two unidirectional kisses do not imply the occurrence of a mutual kissing event.12
(39)x and y divorced ⇏ x divorced y and y divorced x (40)x and y kissed ⇍ x kissed y and y kissed x
The lack of entailment relations in (39) and (40) is characteristic of lexical reciprocals whose transitive alternate is not symmetric. We use the term pseudo-reciprocal to encompass the interpretations that characterize the two kinds of lexical plain reciprocity (37) with its characteristic symmetric equivalence (38) and non-plain reciprocity, where this equivalence fails in one of its two directions (39,40). Pseudo-reciprocity allows us to semantically distinguish lexical reciprocity from grammatical reciprocity. Pseudo-reciprocity primarily emerges with lexical reciprocal verbs (e.g., ‘divorce’, ‘break up’, ‘collide’), and not with grammatical constructions like ‘each other’ pronominals. It should be emphasized that in this paper we are not aiming to give any systematic account of the lexical semantic processes that underly pseudo-reciprocity. Rather, lack of plain reciprocity is simply used as a semantic diagnostic for lexical reciprocity. For more ideas on the semantic principles that underly lexical reciprocity, see Dimitriadis (2008b), Winter (2018), Kruitwagen et al. (2022), among others.
As we will now show, Romance languages show interpretative differences between grammatical and lexical reciprocals parallel to other languages. Se-clauses with unambiguously transitive verbs get the same interpretation as English forms with each other. For instance, the reciprocal reading of sentence (41) entails that Mary described Lisa and Lisa described Mary.
(41) Romance lexical reciprocals get pseudo-reciprocal interpretations. This is most easily observed in constructions without se (or other overt reciprocal markers): such cases only show pseudo-reciprocal readings. By contrast, se-clauses are consistently in line with the same range of reciprocal interpretations as the corresponding English clauses with each other. This contrast can be noted with lexical reciprocals like ‘kiss’ that show a non-plain interpretation.13 For instance, the BP example in (42a) and the Spanish example in (43a) have a plain reciprocal interpretation, in line with the equivalence in (36). Both sentences are true if there were at least two unidirectional kisses between the participants. In particular, both sentences are true in a scenario where Mary and Lisa each kissed the other on the forehead in different moments. By contrast, their counterparts without se in (42b) and (43b) cannot get an interpretation where each girl was kissed by the other in a different they necessarily denote a mutual kiss. Note that a scenario with a single, mutual kissing event is also supported by (42a) and (43a): the pseudo-reciprocal reading of the lexical reciprocal predicate remains accessible in the presence of se.14(42)(43)
Furthermore, in (42a) and (43a) the presence of se correlates with the availability of a reflexive interpretation. A reflexive reading is generally dispreferred with verbs that have a lexical reciprocal entry, but it is not logically excluded. A scenario where Mary and Lisa each kissed herself would be supported in (42a) and (43a), but utterly inaccessible in the absence of se (42b,43b).
This evidence illustrates that the only reciprocal interpretation licensed in Romance constructions without se (or another reciprocal marker) is the pseudo-reciprocal reading that is associated with lexical reciprocals in other languages. In contrast, se-clauses featuring lexical reciprocal predicates are ambiguous between three (i) a pseudo-reciprocal reading, due to lexical reciprocity; (ii) a plain-reciprocal reading, due to grammatical reciprocity; (iii) a grammatical reflexive reading.
Our assumption that se-clauses are ambiguous, rather than underspecified, between lexical reciprocal entries and grammatical reciprocal entries is supported by conceptual and empirical considerations. An analysis in terms of underspecification would treat the pseudo-reciprocal and the plain reciprocal readings as two possible senses of the same unambiguous reciprocal construction. This would not explain the different truth conditions of clauses with and without se as in (42) and (43). Sentences without se like (42b) and (43b) do not support plain they unambiguously require the collective (“reciprocal”) activity of the plural agent to occur in one event. This lack of plain reciprocity is cross-linguistically manifested with intransitives like English kiss. When se does appear, as in (42a) and (43a), plain reciprocal situations are allowed. This parallels the interpretation of grammatical reciprocals with each other in English. Thus, it is natural to explain the licensing of these additional situations by our assumption that adding se gives rise to an additional reading of grammatical reciprocity. Our ambiguity proposal is further supported once we look at the zeugma test. Let us consider a scenario where (i) Mary unilaterally hugged Lisa, and Lisa unilaterally hugged Mary in a different moment, (ii) Irene and Bea were involved in a mutual hug, where Irene did not actively wrap her arms around Bea (for instance, because her arms were occupied carrying bags).15 In such a scenario, the sentence in (44) below is not easily accepted.16
(44)
Thus, we argue that pseudo-reciprocal interpretations and plain reciprocal interpretations stem from two distinct entries in the an intransitive reciprocal entry, and a transitive entry, respectively. Crucially, we do not argue that se itself is in §7, we will provide a unified analysis of this element.
Another property that characterizes Romance lexical reciprocals is the availability of the so-called ‘discontinuous reciprocal construction’, a construction where the logical subject of a reciprocal predicate is split into two one part is encoded as syntactic subject, while the other is in a complement introduced by a comitative preposition (45). It has been noted since Kemmer (1993) that in languages with an overt distinction between lexical and grammatical reciprocity, the discontinuous reciprocal construction is restricted to lexical reciprocals.17 In the Greek examples below, discontinuous reciprocity is allowed with the lexical reciprocal ‘kiss’ in (45a), but it is ungrammatical with the productive quantificational strategy in (45b).
(45)
Dealing with Romance languages, Siloni (2012) discusses discontinuous reciprocals in French and Romanian, and proposes that they are restricted to reciprocal verbs formed in the lexicon. For Italian, felicitous instances of the discontinuous reciprocal construction have been noted by Mocciaro (2011), suggesting that they only occur with symmetric verbs like ‘meet’. However, this generalization does not cover all the verbs that show discontinuous reciprocity in Italian. We observe that for many speakers, discontinuous reciprocity is also possible with non-symmetric predicates that belong in the class of lexical reciprocals (46). Discontinuous reciprocity is not unanimously accepted with certain lexical reciprocals, such as baciare ‘to kiss’ or abbracciare ‘to hug’, and some authors have considered it ungrammatical with these verbs (Dimitriadis 2004; Mocciaro 2011). However, cases where also these verbs appear in the discontinuous construction are accepted in spoken language, and despite their marginality in formal registers, attested examples can be readily found (47). By contrast, the discontinuous reciprocal construction is not felicitous with unambiguously transitive predicates (48).
(46)
(47)
(48)
Below we provide more attested examples with lexical reciprocals from Spanish (49), Catalan (50), and BP (51). In these languages too, the discontinuous reciprocal construction is allowed with the verbs that we propose to treat as lexical reciprocals, in line with the cross-linguistic property of this class of verbs. As shown by (51), in BP the omission of se is optional in this construction.18
(49)
(50)
(51)
The vast majority of the examined predicates that can express reciprocity without se allows discontinuous reciprocity. It is hardly surprising that not all of them are unanimously this kind of idiosyncrasy is also found in languages with an overt lexical/grammatical reciprocity distinction. In English, for instance, the non-symmetric transitive verb ‘hug’ does not take a reciprocal ‘with’ (52a), whereas ‘fight’ does (52b), although both verbs have intransitive reciprocal entries.
(52) *Mary hugged with Lisa.Mary fought with Lisa.
Another characteristic of Romance lexical reciprocals is the possibility of expressing reciprocity with morpho-syntactically singular group NPs. These are NPs headed by singular nouns like committee, team, and choir that refer to collections, usually of animate entities. Barker (1992) defines group nouns in English as those nouns that can take a plural but not a singular of-complement, as in (53).
(53)A team of women/*woman.
As noted by Authier and Reed (2018) for French and English, group NPs support some kinds of reciprocal interpretations. In English, morpho-syntactically singular group nouns can act as the subject of lexical reciprocal verbs, allowing an interpretation where the members of the group are mutually involved in the action described by the verb (54a). By contrast, as observed in Barker (1992), English singular group NPs cannot serve as antecedents for each other, as each other is generally incompatible with singular predication (54b).19
(54)
In Romance languages, unambiguously transitive verbs with se do not get a reciprocal interpretation with singular group NPs. The BP example in (55) only has a reflexive interpretation, for instance where some representative member(s) of the team described the team as a whole.
(55) By contrast, the verbs that we characterize as lexical reciprocals can express reciprocity with group NPs and singular agreement. This is shown in the BP example in (56), which is felicitous under the collective reading where the members of the team were involved in a hug. The same holds for Italian (57), Spanish (58), and Catalan (59).
(56)
(57)
(58)
(59)
In this section we have explored a class of Romance predicates that express reciprocity by themselves, with or without the typical clitic se. In a set of language-specific constructions where se may or must be omitted, these verbs express reciprocity without any additional reciprocal element, and without the reflexive/reciprocal polysemy that characterizes se constructions. We propose that this property indicates a class of lexical reciprocal verbs, which other languages exhibit with similar verbal concepts.
In addition to the licensing of se omission, we have demonstrated that Romance lexical reciprocals show other properties that are characteristic of lexical reciprocals in other languages. With or without se, these verbs consistently support a pseudo-reciprocal interpretation, discontinuous reciprocity using ‘with’, and reciprocal readings with singular group subjects. A comprehensive list of lexical reciprocals we identified in the four Romance languages we study is provided in the Appendices, together with more examples of their properties.
In this section, we propose that a parallel lexical/grammatical opposition is observed with respect to Romance reflexives. We identify Romance predicates that receive a reflexive interpretation without se or another reflexive item, and we argue that they have a lexical reflexive entry. This strengthens our claim that the distinction between lexical and grammatical valence-reducing alternations is general in Romance. The observed facts on reflexivity also support the generalization that the reflexivity/reciprocity polysemy only occurs in the presence of se.
We identify lexical reflexive verbs using the same constructions where we observed the emergence of lexical reciprocity without se. These constructions vary per language, as previously summarized in Table 2 (§3.4).
In BP, se can be omitted in finite clauses. In example (60) below, the BP verb depilar ‘remove body hair, shave’20 supports situations where Mary shaved herself, similar to the English translation. We propose that this fact categorizes depilar as a lexical reflexive in BP.
(60)
In Catalan, Spanish, and Italian, similar reflexive interpretations emerge in other environments where se can (or must) be omitted. Consider for instance the Italian verb lavare ‘to wash’ in the analytic causative in (61). Like all Italian causatives with transitive verbs, this sentence has a passive reading, where the subject caused Mary to be washed. Crucially, the verb ‘wash’ in (61) also allows a reflexive reading, where Mary washed herself.
(61) Similar facts hold for analytic causatives in BP, Spanish, and Catalan. Accordingly, we propose that verbs like Italian lavare ‘to wash’ have an intransitive reflexive entry similar to English.
Absolute constructions in Catalan and Spanish also support reflexivity without se. For instance, the Spanish verb afeitar ‘to shave’ in (62) has a reflexive interpretation, on top of the passive interpretation that is standard with transitive verbs in absolute clauses.
(62) The same holds for absolute clauses in Catalan, where the verb afaitar ‘to shave’ (among others) leads to the same array of interpretations as the Spanish version in (62). Accordingly, we propose that verbs like Spanish afeitar and Catalan afaitar have an intransitive reflexive entry.
Semantically, there are two critical facts to be observed with respect to the reflexive interpretations above. First, verbs that allow reflexivity without se (or another reflexive element) support an interpretation that is subtly but critically different from the meaning generated for transitive verbs with reflexive pronouns. Second, in cases of reflexive interpretations without se, no reciprocal meaning emerges.
Considering the first point, let us note a well-known grammatical reflexivity requires identity between two thematic roles. For instance, in (63) the subject is the agent and the patient of the event denoted by the Ali was the entity who shaved/described Ali. It has been noted that instances of grammatical reflexivity are also in line with a so-called ‘proxy’ reading, where the referent of the object is a sufficiently close copy of the referent of the subject, so that it functions as a proxy for it (Jackendoff 1992; Lidz 1997; Reuland 2001). For instance, the English clauses with himself in (63) hold true in a scenario where Ali, in the context of a visit to a wax museum, shaved or described a statue of himself. We refer to the interpretation of grammatical reflexives as plain reflexivity.
(63)Ali shaved himself.Ali described himself. Lexical reflexives are semantically distinguished from such cases of grammatical reflexivity (Doron and Rappaport Hovav 2009; Spathas et al. 2015; Haspelmath 2022). First, across languages, speakers accept lexical reflexives in scenarios where the subject is a willing patient (mentally and consensually involved in the event), whereas the agent physically carrying out the action coincides with a different entity. For instance, the English intransitive form in (64) is acceptable if Ali was the one who performed the act, but also if Ali went to the barbershop for a shave. Importantly, the latter possibility is ruled out for the grammatical reflexive in (63a). Moreover, the ‘proxy’ reading is excluded with lexical reflexive unlike (63a), (64) does not support a reading where Ali shaved a statue of himself. (64)Ali shaved. We refer to this interpretation of lexical reflexives as pseudo-reflexive. We standardly assume that it emerges from the verb’s intransitive entry, and not from any process of argument binding in the syntax.
Grammatical reflexives do not require volition of their agent and patient (which, except in ‘proxy’ readings, are identical).21 For example, in an unfriendly scenario where Ali was forced to shave himself against his will, the grammatical reflexive form in (63a) may be considered true. By contrast, volition from the argument may improve the acceptability of lexical reflexive forms, and make them acceptable in situations where the subject does not correspond to the entity physically performing the action. Thus, in case someone shaved Ali against his will, (64) is deviant. We can summarize these semantic differences between grammatical reflexives and lexical reflexives as *Grammatical reflexives require identity between two arguments of a binary predicate (possibly by proxy), but not all lexical reflexives require strict identity.*Lexical reflexives may require more volition of their argument than parallel grammatical reflexives.
We standardly assume that the identity requirement in grammatical reflexives results from the binding of the reflexive pronominal argument by the other argument. Like in the case of pseudo-reciprocal interpretations, we assume that the reading of lexical reflexives comes from intransitive entries. As we will elaborate in §7, we argue that both lexical reflexives and lexical reciprocals take one argument with a complex thematic role AgPt, where some properties of agents and patients are retained. This treatment encompasses the interpretation of lexical reflexives, where the active role of the agent is not necessarily retained, and of lexical reciprocals, where the active participation of both individuals is not always required. However, we do not delve into the question of how such lexical interpretations are truth-conditionally associated with the stem’s meaning. Rather, we use the semantic properties of English lexical reflexives to study the dissociation between lexical and grammatical processes in Romance.
In parallel to our observations on Romance reciprocals, lexical reflexives behave differently from grammatical reflexives in relation to the realization and the semantic effects of se. For example, with the transitive verb descrever ‘to describe’, the BP example in (65) gets the same interpretation associated with its grammatical reflexive English correlate.
(65) Now let us consider the two sentences in (66) below. The sentence with se in (66a) supports situations where Mary shaved her own self (a scenario that is in line with both plain and pseudo-reflexive readings), but it is also true in two kinds of exceptional situations. First, it is true in a situation where Mary volitionally went to the beautician for depilation; secondly, it is true in a situation where Mary shaved a statue of herself. These two kinds of situation are characteristic of lexical reflexivity and grammatical reflexivity respectively, as the translations of (66a) indicate. By contrast, example (66b) only supports a pseudo-reflexive it is in line with situations where Mary shaved her own self or was volitionally shaved by a beautician, but it excludes the ‘proxy’ reading where Mary shaved a statue of herself.
(66) This pattern is fully parallel to what we observed with lexical and grammatical reciprocals (42).
Our second observation is also parallel to what we have already observed with lexical reciprocity. We saw that clauses with se and plural subjects are systematically ambiguous between reflexive and reciprocal interpretations, whereas clauses without se only get access to the inherent meaning of the verb’s entry. We now observe this also with lexical reflexive verbs. For example, let us consider sentence (67) below.
(67) Just like its singular counterpart in (66a), sentence (67) above can be true in two kinds of exceptional situations. First, Mary and Lisa might have been shaved by someone else, but on their own accord; this is the kind of situation that characterizes the pseudo-reflexive reading. Second, each of Mary and Lisa might have shaved a statue of herself; this kind of situation falls under the plain reflexive reading. Additionally, like all sentences with se and a verb with a transitive entry, (67) also gets a plain reciprocal reading where Mary and Lisa shaved each other. Now let us consider sentence (68) below, where se is
(68) The sentence (68) is true as long as each individual referred to by the subject is a volitional patient of the action described by the verb (e.g., if each of Mary and Lisa shaved herself or went to the beautician). The form in (68) does not support a situation characteristic of grammatical reflexivity where Mary and Lisa each shaved a statue of herself, nor a reciprocal reading where they shaved each other. Accordingly, we characterize this sentence, like its singular counterpart (66b), as having an unambiguously pseudo-reflexive reading.
The data of Sects. 3 to 5 support the following generalizations on se and on its interactions with reflexive and reciprocal (69)SE generalizations – lexical reciprocity and reflexivity: Se-clauses without an additional reciprocal or reflexive item allow both plain reflexivity and plain reciprocity.Certain transitive verbs can appear in sentences without se or any other item expressing reflexivity or reciprocity. The interpretation of such sentences is unambiguously pseudo-reciprocal with some verbs, and unambiguously pseudo-reflexive with others. Accordingly, we refer to these verbs as lexical reciprocals/reflexives.The pseudo-reciprocal/reflexive interpretation of these verbs is also retained when they appear with an overt se, on top of the plain reciprocal and reflexive interpretations that appear with all transitive verbs.
So far, we have proposed that reflexivity and reciprocity can be lexically expressed in Romance languages. This proposal is most easily supported by cases in which these interpretations appear while se is omitted. In this section, we will see that also grammatical arity-reducing operations can take place without se. This phenomenon uniformly occurs with all transitive verbs in environments that allow se omission, provided that they appear together with an overt reflexive/reciprocal element. As we will show, Romance reflexive/reciprocal pronominals and adverbials (like BP si mesmo ‘himself’ or Spanish mutuamente ‘mutually’) disambiguate the interpretation of se-clauses, ridding them of the reflexivity/reciprocity polysemy. Additionally, with lexical reflexives and reciprocals, such elements also remove the lexical pseudo-reflexive or pseudo-reciprocal reading.
The facts presented in this section, together with the generalization on lexical reciprocals and reflexives in (69), will lead to the theoretical picture proposed in Sect. 7: the reflexive and reciprocal items discussed below will be treated as semantic operators, and the clitic se as an arity-reduction marker that licenses reciprocal/reflexive interpretations, but does not carry an independent meaning of its own.
Overt reciprocal elements include adverbials like Italian a vicenda ‘mutually, in turns’ and Spanish mutuamente ‘mutually’, as well as pronominal elements like BP um o outro ‘one another’ and Catalan l’un a l’altre ‘one another’. These elements have three different functions that are relevant for our study.
First, when they appear with se they remove the reflexivity/reciprocity ambiguity, leading to a reciprocal interpretation. For example, the Italian clause in (70) can receive either a reflexive or a reciprocal reading, whereas only the latter is accessible in the presence of a vicenda in (71).
(70)
(71)
Second, overt reciprocal elements also remove the pseudo-reciprocal reading that appears with se and lexical reciprocals, leaving plain reciprocity as the only reading. For example, let us first consider the following example, without any overt reciprocal
(72) In sentence (72), on top of a less accessible plain reflexive reading (72i), the predicate lasciare ‘leave/break up’ leads to two prominent one that is here paraphrased using a reciprocated transitive entry ‘leave’ (72ii), and another using the pseudo-reciprocal meaning ‘break up’ (72iii). The last reading does not entail two unidirectional relations, as the relationship could be unilaterally terminated by one individual. Thus, sentence (72) is considered true in a scenario where Mary terminated the relationship with Lisa, while Lisa was left heartbroken. When we add the element a vicenda, this pseudo-reciprocal reading disappears. Consider for instance what happens when a vicenda is added to sentence (72):
(73) Unlike (72), sentence (73) cannot be accepted if the relationship between Mary and Lisa was unilaterally for the sentence to be true, each of the two people must have left the other. Thus, in (73) the adverbial a vicenda disambiguates (72) and only allows the plain reciprocal reading.
Third, in constructions where se can be omitted (Sect. 3), overt reciprocal expressios derive reciprocity without se for all transitive verbs. For example, in the Italian causative in (74) below, grammatical reciprocity is possible with a vicenda, but without se:
(74) Similarly, BP transitive verbs can lead to reciprocity without se in finite clauses where the adverbial um o outro occurs.22 For example, let us consider sentence (75):
(75) Sentence (75) unambiguously has the plain reciprocal interpretation where Mary described Lisa and Lisa described Mary. Due to the presence of the reciprocal item um o outro, and in contrast with sentence (20) above, the clitic se in (75) is only optional. The same observation extends to Spanish absolute constructions, where mutuamente ‘mutually’ can lead to a reciprocal interpretation without se with any transitive verb. For
(76)
Similar observations hold for reflexive elements like the pronominals BP si mesmo ‘himself’ in BP, si mismo ‘himself’ in Spanish, and si mateix in Catalan.23 First, these items disambiguate se-clauses by eliminating the reciprocal reading. For instance, while the BP sentence (77) shows the familiar reflexivity/reciprocity polysemy, (78) can only be interpreted reflexively.
(77)
(78)
Second, when they appear with lexical reflexive predicates, overt reflexive elements disallow the pseudo-reflexive interpretation. For instance, the BP sentence (79a) (=(66a)) supports a pseudo-reflexive interpretation where Mary was volitionally shaved by a beautician, whereas (79b) requires that Mary shaved herself.
(79)
Finally, overt reflexives are allowed without se in the same environments that allow lexical reciprocity/reflexivity without se. Such constructions receive plain reflexive interpretations. For example, in the si mesmo finite clause in (80) below, se can either appear or not, and the interpretation is plain reflexive either
(80)
We summarize this section using the following generalizations about se and overt reciprocal and reflexive items, which complement the generalizations in (69): (81)SE generalizations – overt reciprocity and reflexivity: Clauses (with or without se) containing a reciprocal (reflexive) item are unambiguously reciprocal (reflexive, respectively).When a lexical reciprocal or reflexive predicate appears with a reciprocal (reflexive) item, it only shows plain reciprocity (reflexivity, respectively), but no pseudo-reciprocity or pseudo-reflexivity.The same environments that support pseudo-reciprocity and pseudo-reflexivity without se also support plain reciprocity (reflexivity) without se, provided that they contain an overt reciprocal (reflexive, respectively) item.
This section proposes a unified analysis of generalizations (69) and (81), focusing on the syntactic-semantic role of se with lexical and grammatical reciprocity/reflexivity. We follow Labelle (2008) in assuming that se is a Voice head projection. However, we diverge from Labelle’s proposal that se has a direct contribution to reflexive or reciprocal meanings as an operator that binds external and internal arguments. Instead, we propose that se is a marker à la Reinhart and Reuland (1993), which marks the VP as reflexive/reciprocal, without providing the reflexive/reciprocal meaning itself. In our analysis, arity-reducing operators may be overt (like the reflexive and reciprocal items discussed in Sect. 6) or they can operate covertly, licensed by se.
Labelle (2008) proposes that French se is responsible for reflexive/reciprocal interpretations when it combines with transitive predicates, but it is semantically redundant with predicates that already have a reflexive/reciprocal reading. To illustrate Labelle’s account, let us consider the following French
(82)
(83) Labelle’s main point about these examples is reminiscent of the facts that we discussed in Sect. 6: French se leads to reflexivity (or reciprocity) with ordinary transitive verbs (82), but it also appears with verbs whose reflexive (or reciprocal) meaning comes from an additional element, e.g., auto- in (83). Labelle’s syntactic-semantic analysis of examples like (82) and (83) is given in (84) and (85) below.24
(84)Following Labelle (2008, (31), p.844):
(85)Labelle (2008, (32), p.844): Labelle’s analysis relies on Kratzer’s (1996) neo-Davidsonian semantics. In Labelle’s use of Kratzer’s proposal, se is a functional head that introduces the external argument (x in the component Agent(e,x)) and binds it to the internal argument of the verb (x in P(e,x)). With transitive verbs as in (84), se introduces the event’s agent as an external argument, binding it to the internal argument. By contrast, verbs that are prefixed by auto- or entre-, like autoanalyzer in (85), already contain an external argument variable in their denotation. In such cases, se has the same meaning as with transitive verbs. However, Labelle’s neo-Davidsonian assumption is that the Agent operator is a (possibly partial) function on events, which maps every event to a unique agent (if any). This assumption makes sure that when the Agent operator is introduced twice for the same event (e.g., by the auto- prefix and by se) it binds the same entity to the external argument. Although se is semantically redundant in such cases, Labelle (2008) notes that it is nonetheless obligatory for the grammaticality of French sentences like (83). Labelle explains this requirement by assuming that se is needed to ensure a coherent it prevents Active Voice from introducing a distinct variable for the subject. The idea in Labelle (2008) is that in the absence of se, Active Voice would occupy the Voice head and it would introduce an external argument which would not be bound to the internal argument of the VP. Labelle’s assumption (which we question below) is that this would lead to an incoherent interpretation where there are two distinct unsaturated external positions for one and the same subject, a situation that would violate the Theta-criterion.
Although Labelle’s analysis is the starting point for our treatment of se, it presents some incompatibilities with the observations in previous sections, and it raises some questions about its own merits. One problem is a problem of generality. Labelle’s analysis relies on the assumption that whenever se appears, the verb has an internal argument. Labelle’s semantic analysis works for verbs like autoanalyzer, which are the outcome of a productive prefixation process. However, the analysis should be extended to capture the semantics of the lexical reciprocals we covered in §4.3 (as well as lexical reflexives §5). As we saw, the inherent reading of many lexical reciprocals and reflexives is not in line with multiple thematic roles. Thus, the pseudo-reciprocal/reflexive interpretations of lexical reciprocals like ‘hug’ or lexical reflexives like ‘wash’ cannot be derived by a standard binding mechanism like the one Labelle proposes for auto- and entre- prefixation. Another problem of generality appears with Labelle’s assumption that se is uniformly needed with transitive verbs to co-index the internal argument and the external argument of VPs. As we saw, there are many examples of transitive verbs in Romance where se does not appear, and a reciprocal or reflexive interpretation emerges due to the presence of another reciprocal/reflexive item. Such cases are not addressed in Labelle’s account, where the presence of se is considered as a ubiquitous requirement to ensure reflexive/reciprocal interpretations. Another problem lies in the semantic motivations for the presence of se. As mentioned earlier, Labelle proposed that se is necessary to prevent Active Voice from occupying the Voice head, which she assumes would result in the introduction of two different external arguments for verbs like autoanalyzer (one introduces by auto-, one by Active Voice). As we mentioned above, the identification of the two agent arguments in Labelle’s analysis (85) is a basic assumption of the neo-Davidsonian approach, where Agent is treated as a function from events to entities. Due to this property of the semantic framework that Labelle relies on, it is unclear what meaning she proposes for Active Voice that would introduce a different agent to the event from the one introduced using auto-.25
To overcome these issues, we propose an alternative explanation for the required presence of se, where it is not directly responsible for reflexive and reciprocal interpretations. We follow Labelle’s proposal that se is a Voice head projection, but we argue that se never contributes to reflexive or reciprocal meanings all by itself. Instead, we propose that se combines with VPs that already have a reflexive/reciprocal either a lexical interpretation due to the intransitive meaning of the verb stem, or a grammatical interpretation derived by a reflexive/reciprocal operator. We propose that such operators can be introduced overtly (e.g., as pronouns or adverbials) or covertly, as operators that are responsible for the semantics of arity-reduction. Overt operators perform the necessary ‘marking’ of a VP headed by a transitive verb as being reflexive/reciprocal, whereas covert operators do not. We propose that it is only in such cases, where such R(eflexive/reciprocal)-marking is missing, that the introduction of se is necessary to make the reflexive/reciprocal semantics correspond with morpho-syntax. We propose that the variability in the presence of se across different Romance languages emerges due to different syntactic restrictions on the presence of this element.
In more detail, we analyze the appearance of se as relying on four different (i)Types: We follow the typed meaning that is assumed by Kratzer and Labelle for Active Voice as restated in (86) an operator that takes predicates over events (type st) and adds to them an external argument, which leads to a predicate of type e(st).26 We propose that Active Voice is ruled out with reflexive/reciprocal VPs due to a simple type mismatch. We assume that such VPs already contain an external argument variable, which is either part of the lexical entry or introduced by a grammatical operator. Thus, reflexive and reciprocal verb entries are assigned the lexical type e(st). This is in contrast to other intransitive verbs that are assumed to be of the neo-Davidsonian type st. This contrast is illustrated in (87) below. The intransitive reflexive/reciprocal entry of verbs like shave/hug is analyzed as a predicate ‘shave1’/‘hug1’ that has an external argument with the thematic role ‘AgPt’. This thematic role shows both agent-like and patient-like semantic properties of the corresponding transitive entry, which is denoted ‘shave2’/‘hug2’. Further, overt reciprocal and reflexive items like si mesmo (BP, ‘himself’, (88)) or um o outro (BP, ‘each other’) turn transitive predicates into predicates with an external argument that bears the thematic roles of both Agent and Patient. For the clitic se we propose the typed function in (89): the (semantically void) identity function on two-place predicates over entities and events. As a result of these types and meanings, lexical reflexives/reciprocals and transitive verbs with a reflexive/reciprocal operator can combine with se, but not with Active Voice. Conversely, non-reflexive/non-reciprocal intransitive verbs like ‘laugh’ can combine with Active Voice but not with se. A major difference between the two kinds of VPs is that with simple intransitives like ‘laugh’, an additional operation is required in order to add their external argument, whereas reflexive/reciprocal VPs contain the external argument as part of the lexical meaning of the VP. (86)(87)(88)(89)(ii)Covert operators: As we mentioned above, we assume that se licenses covert reciprocal and reflexive operators, which have the same meanings as ‘herself’ and ‘each other’ in (88).27(90)(iii)R-marking: Like all overt reflexive and reciprocal items in Romance, se has the function of syntactically ‘marking’ reflexive/reciprocal predicates, in the sense of Reinhart and Reuland’s (1993) analysis of Conditions A and B. We implement Reinhart and Reuland’s proposal as (91)Condition A: An R-marked predicate has a reflexive/reciprocal interpretation.28Condition B: Any reflexive/reciprocal interpretation of a predicate requires R-marking. Specifically, Condition A requires that any transitive verb appearing with se or an overt reflexive/reciprocal item has a reflexive/reciprocal interpretation. We have seen that this requirement systematically holds in Romance languages. Condition B requires that any verb that is interpreted reflexively/reciprocally must be R-marked. Like Reinhart and Reuland, we assume that this R-marking can come from se, from an overt reflexive/reciprocal item, or from a reflexive/reciprocal intransitive in the lexicon. Thus, reflexive/reciprocal intransitives satisfy Condition B even when they are not accompanied by any overt R-marker like se or an overt reflexive/reciprocal item. In the absence of a lexical reflexive/reciprocal entry, reflexive and reciprocal interpretations cannot emerge from the verb alone. Thus, for Condition B to be satisfied in that situation, transitive predicates must be marked by an overt reflexive pronominal and/or by se.(iv)Syntactic construction: We observe that there are syntactic constructions where se may or must be omitted. We take this as a distributional fact about se, and we consider here three types of syntactic environments in Romance in relation to the presence/absence of se with reflexive/reciprocal +SE, where se is obligatory (It/Sp/Cat finite clauses)−SE, where se is disallowed (It causatives, Sp/Cat absolutes)±SE, where se is optional (BP finite clauses, Sp/Cat causatives).We hypothesize that the possible omission of se correlates to the presence/absence of a Voice projection to host this element. For causatives, this is in line with the existing proposal that Romance causative complements do not project functional layers (Ciutescu 2013; Pitteroff and Campanini 2013), and generally lack an external argument (Labelle 2017). For BP finite clauses, a connection between se-omission and absence of a Voice head is proposed in Carvalho (2018). This hypothesis raises questions with respect to syntactic parameters underlying constructions with or without se, and with respect to the cross-linguistic variation within Romance. We defer these questions to future work. For our present purposes, it is sufficient to observe that the requirements for the presence of se vary across languages and constructions. The hypothesis that availability of Voice is the key for this variation is convenient for our present goals, but it requires further study and is not part of our core proposal.
Let us now exemplify the working of this fourfold proposal. We start with intransitive predicates that do not receive a reflexive or reciprocal interpretation, like ‘laugh’ (92). For such verbs, the external argument is introduced in Voice using the denotation of Active Voice, as demonstrated in (93).
(92)(93) By contrast, VPs that are interpreted as reflexive/reciprocal are treated as containing an Agent variable. Let us first consider lexical reflexives/reciprocals as in (94). With VPs headed by such verbs, the external argument variable is already part of the lexical entry, which contains the complex thematic role AgPt (95). The predicate is of type e(st), and it cannot combine with Active Voice. In +SE syntactic environments, the predicate appears with se. This element does not introduce the external argument and does not have any reflexive or reciprocal semantic it merely marks that the VP is reflexive/reciprocal, by stating that no Agent variable is introduced in Voice. Note that because the verb has a lexical reflexive entry, which is assumed to be an R-marker, Condition B is also satisfied without se. The presence of se merely depends on the syntactic requirements of the it is obligatory in +SE constructions, but not in –SE or ±SE environments, like BP finite clauses (94). In either case, the interpretation of lexical reflexives/reciprocals originates from the verb stem.
(94)
(95)
The variant of (94) with se also supports the grammatical strategy, which applies with the transitive meaning of depilar. We propose that this polysemy of depilar leads to structural ambiguity with the se variant of (94), which we examine more closely in (96) below.29 The reading in (96i.) is the same that we analyze in (95) above as stemming from the lexical reflexive entry ‘depilar1’. The reading in (96ii.) is due to the transitive entry ‘depilar2’, here reflexivized using the covert reflexive operator RFL (97). This interpretation requires the presence of se to satisfy Condition B, because R-marking is performed neither by the transitive entry, nor by the covert RFL operator. In this case, se is required for R-marking (although by itself it does not provide the reflexive interpretation), whereas the covert RFL operator is responsible for the reflexive interpretation (but it does not provide R-marking). Thus, although sentence (96) is treated as ambiguous between lexical reflexivity and grammatical reflexivity, our analysis unifies the role of se with the two strategies.
(96)
(97)
Let us now analyze grammatical strategies more closely. Recall that grammatical reflexivity and reciprocity can be realized in three different configurations; they are exemplified in (98) below for BP, using the unambiguously transitive predicate ‘describe’. We propose that plain reflexive/reciprocal interpretations consistently come from reflexive/reciprocal operators that can be realized overtly (e.g., BP si mesmo/um o outro) or covertly (RFL/RCP operators). Overt and covert operators have the same meaning; the only difference is that overt operators can satisfy Condition B, whereas covert operators cannot because they are not morpho-phonologically realized. Thus, overt operators can operate without se, whereas covert operators cannot lead to reflexive/reciprocal readings all by themselves. Both overt and covert reflexive/reciprocal operators introduce the external argument variable and assign it the thematic roles of both Agent and Patient (99). Just like in the case of the lexical strategy, this results in a VP of type e(st) that cannot combine with Active Voice, but can combine with se whenever this element is allowed or required. Once again, se does not have any reflexive semantic content (99).
(98)
(99)
In sentence (98a), all the components of the representation in (99) are overtly realized. The clause in (98b) is compositionally analyzed like (98a), but without se. This is possible because the overt operator already satisfies Condition B, so se can be omitted in –SE and ±SE constructions, like BP finite clauses (98b). By contrast, in (98c) the operator responsible for the reflexive interpretation is covert. For this reason, se is required to satisfy Condition B: the absence of se results in ungrammaticality (98d), or in the unavailability of reflexive/reciprocal readings (e.g., in Italian causatives (26), §3.2).
Let us conclude by discussing the familiar ambiguity between reflexivity and reciprocity in se-clauses. The clitic se licenses both the covert reflexive operator RFL and the covert reciprocal operator RCP. For this reason, se-clauses with a plural subject support both reflexive and reciprocal interpretations, and can only be disambiguated by further contextual information. Furthermore, if the verb has a lexical reflexive/reciprocal entry, the clause receives an additional interpretation coming from the intransitive verb stem. Throughout the paper, we have explored this three-way ambiguity in sentences like (42a) and (67). (67) is repeated below as (100). With our proposed analysis, we can now examine the emergence of these three readings. The pseudo-reflexive reading in (100i.) is due to the inherent meaning of the intransitive ‘depilar1’, derived as in (94). The plain reflexive reading in (100ii.) and the plain reciprocal reading in (100iii.) contain the transitive verb ‘depilar2’, as in (97). The interpretation in (100ii.) is due to the covert reflexive operator RFL, whereas (100iii.) is due to the covert reciprocal operator RCP. As previously seen, only the pseudo-reflexive interpretation in (100i.) survives if se is the interpretations in (100ii.) and (100iii.) require se to license the RFL and RCP operators.
(100)
Reciprocal and reflexive interpretations result from lexical and grammatical strategies that have been argued to exist in all languages. In this paper, we studied the case of Romance languages, where the two strategies are not always morpho-syntactically distinct. In these languages, many environments invariably require the element se for expressing a reciprocal meaning or a reflexive meaning. While this is a considerable obstacle for characterizing reciprocity and reflexivity in Romance semantics, in this paper we have aimed to show that the challenge is not unsurmountable. Romance lexical reciprocals and reflexives can be fruitfully studied based on properties that cross-linguistically characterize this class of predicates. We focused on Italian, Brazilian Portuguese, Spanish, and Catalan. In these four language, we identified a class of verbs that, in constructions that vary between languages, express reciprocity without se or other reciprocal elements, without giving rise to a reflexivity/reciprocity ambiguity. We showed that systematic semantic characteristics of such cases give substantial support to the claim that Romance lexicons contain reciprocal and reflexive verbs, whose meanings are fairly stable across languages.
Moving on to the role of se in the semantic derivation, we pointed out that in the presence of an overt reciprocity/reflexivity operator, se can be omitted in precisely the same environments where it is not required with lexical reciprocals/reflexives. These data go against accounts of se as operating directly on the verbal valency, and support the treatment of se as a functional head projection, along the lines of Labelle (2008). We extended Labelle’s analysis, arguing that se never has any reciprocal or reflexive semantics, although it has a central role in licensing reflexivity and reciprocity in the spirit of Reinhart and Reuland (1993).
The distribution of se clitics varies remarkably across different Romance languages. Yet, we believe that the current work offers a unifying perspective on some of the key challenges by showing the ways in which syntactic projections, semantic types, binding conditions, and covert operators interact in relation to lexical and grammatical functions. Further, we believe that the data we presented and the theoretical perspective we proposed may also prove useful for studying non-Romance languages whose reciprocal markers are comparable to Romance, e.g., German (Everaert 1986; Gast and Haas 2008), Icelandic (Wood 2014), Serbo-Croatian (Marelj 2004), Polish, and Slovenian (Rivero and Sheppard 2003; Wiemer 2007). Future studies may also reveal a contrast between the grammatical and lexical strategies beyond the Indo-European family, in other languages without clear distinctions between these two strategies, such as Lingala and Kanuri (Kemmer 1993). Such a larger language sample may allow us to substantially test the hypothesis (cf. Haspelmath (2007)) that all natural lexicons contain reciprocal and reflexive verbs. Furthermore, the set of concepts that these verbs refer to may possibly be governed by universal principles. Whether these ideas can be substantiated remains to be seen, but we hope that the present paper has highlighted possible directions for future studies.