Verbless Sentences in Spoken Italian

Introduction

This book is addressed to linguists who deal with spoken corpora, discourse and conversational analysis, syntax and semantics. The work presents the analysis of the expression of the core arguments Subject (S) and Object (O), to which I also refer with the term Grammatical Relations (GRs), in verbless sentences (Vls) with at least one NP, in spoken Italian, such as the those in bold in examples (1) and (2).

1. (1) [VoLIP RA4]¹
   • [Two friends are on a train and comment on the accent of a vendor selling tapes (speaker E)]

1. (2) [VoLIP RA1]
   • [The speaker is telling her friend that her ex-boyfriend is not arriving by train as agreed but warns with an answering machine message that he has changed his mind]

In (1) we have the speaker A referring to the vendor E passing by on a train who pronounces cassette (‘tapes’) with a Roman accent, which speaker B repeats in the next turn. He uses a Vl that is constructed with a AP (carino, ‘cute’), a discourse marker (vero ‘right’), and an NP (l’accento, ‘the accent’). In (2) the speaker uses a Vl to introduce a reported speech in the telephone recording which consists of an NP on which a PP depends (messaggio della segreteria ‘voicemail message’).

The study of the expression of GRs in Vls is conducted on corpora of spontaneous dialogic and monologic speech, as these constructions are very frequent in speech communicative mode (Voghera 2017). Over the past twenty years, corpus-based studies have shown that in speech samplings of Romance languages such as Italian, French, Spanish, and Portuguese, Vls constitute around 30 % of sentences, and in more informal contexts they reach almost 45 % (Cresti & Moneglia 2005). In English, in dialogic texts, Vls constitute more than a third of syntactically autonomous structures (38.6 %) (Biber et al. 1999: 1071). Moreover, in elicited dialogic speech, such as map-task dialogues, Italian, English, and Spanish, Vls are 75 % of the total autonomous structures (Voghera et al. 2010). Speakers frequently resort to Vls because they are better adapted to the short time frames of message planning and processing, the brevity of turns, and the speed of exchange of turns, especially in dialogue (Voghera 1992, 2017).

The frequency of Vls in spontaneous spoken texts, in particular in dialogic ones, is explained by the optimality of these structures to the semiotic conditions of speech mode. In fact, Vls are among the functional syntactic correlates of speech mode, that is, structures that occur most frequently in spoken texts because they are most efficient for speakers (Voghera 2017; Voghera et al. 2022). Speech mode refers to the set of semiotic conditions that interact simultaneously and find their most prototypical realization in spontaneous face-to-face conversations: that is, the co-presence of participants, the use of phonic-auditory and visual-gestural channels, and the synchrony of production and reception times. These conditions lead to linguistic choices that occur more frequently not because of a particular register (i.e., structures that correlate with dimensions of sociolinguistic variation), but because they are more efficient. Vls are more optimal at conveying messages in the short timeframes punctuated by the simultaneity of planning and production (Hopper 2011; Voghera 2017; Voghera et al. 2022).

In addition to their frequency, speech studies have drawn attention to the problematic nature of defining Vls according to the criteria of traditional grammar. The question that most linguists have asked is whether these constructions can be called sentences. Giving an answer to this question has often been problematic, because traditionally, only Vls in which it is possible to identify a subject and a predicate are understood as a sentence, because they are close to the verbal sentence model. Among the various types of Vls, in fact, only the verbless sentences that Meillet (1906) defines as pure can be more easily described through the canons imposed by the verbal sentence model (Lefeuvre 1999; Mortara Garavelli 1971). They are however, only 10 % of the structures produced in both speech and writing (Fiorentino 2004; Cresti & Moneglia 2005; Voghera et al. 2010). Consequently, to look for the verbal sentence model in Vls is to leave out of the sentence definition most Vls that are neither predicative nor dirhematic (Biber et al. 1999; Cresti & Moneglia 2005; Giordano & Voghera 2009; Voghera et al. 2010). For this reason, Vls that deviate from the sentence model have been described only through pragmatic criteria as utterances. According to traditional approach, they would not belong to the syntactic domain that describes the verbal sentence instead.

In this work the analysis of the expression of GRs was conducted on predicative and non-predicative Vls. Contrary to expectations both non-dirhematic and non-predicative constructions proved to be of great interest because GRs could also be traced in the non-predicative Vls.

The analysis I present here is based on a theoretical approach that I have called non-verbocentric, because I do not consider Vls as structures derived from a verbal sentence or implying an unspoken verbal form. This approach harkens back to studies that do not necessarily refer to the verbal sentence model for the definition of Vls (De Mauro 1974; Voghera 1992; Barton & Progovac 2005; Progovac 2006). In this work, in fact, I consider the Vls (predicative and non-predicative) as sentences, because they are syntactically autonomous (Jespersen 1924; Bloomfield 1936; Voghera 1992; Giordano & Voghera 2009) and doubly articulated constructions (De Mauro 1974), which could be predicative (De Mauro & Thornton 1985), as (1), or non-predicative, as (2). In speech they may coincide with autonomous turns or their parts and present a sentential prosodic form (Giordano & Voghera 2009).

In the same way as verbal sentences, Vls can be predicative and can express values of grammatical categories. In fact, according to this theoretical perspective predication or other values such as tense and person do not belong exclusively to the verb (Hjelmeslev 1948). The use of the verbless label in my work does not imply that verb is a trait per se relevant to the definition of a construction (Lefeuvre 1999; Graffi 2001; Giordano & Voghera 2009; Tanguy 2009; Voghera et al. 2010), but I use it because in the linguistic literature, the label verbless is now a common expression to refer to very different structures, which share the characteristic of having no finite verbal form and being syntactically autonomous and prosodically defined.

It is evident that the problematicity of defining Vls is due to the inadequacy of syntactic criteria to the analysis of speech (Voghera 1992), which are mostly forged on the written because traditionally the language model under consideration belongs to it. The analysis of the expression of GRs in Vls is part of an epistemological choice that starts from the idea that the study of speech makes it possible to analyze phenomena belonging to different descriptive levels (phonological, morphological, syntactic, etc.) that in other modalities are not frequent and to observe their systematicity and functioning within the grammar of the language. Here the grammar is not conceived as a priori descriptive model that guides metalinguistic reflection, but is understood as the set of regularities that emerge from the concrete use of the language. Language usage embeds the set of possible choices that speakers of a language make and that enable intercomprehension. In other words, they are the possible linguistic structures, which in potentia can be realized in different forms, adapted to the different enunciative situations: different texts and different situations can express different forms of grammaticality. Grammaticality is equivalent to what on the individual level is called implicit grammar, that is, the processes that speakers unconsciously enact (Voghera et al. 2022). The description of Vls therefore wants to take into account the regularities their use brings out, not their greater or lesser possibility of belonging to a predetermined descriptive model.

The main purpose of this study is to show through a corpus-driven analysis of Vls, how in structures where the verb is not present or even implied, it is possible to talk about GRs.

Traditionally, the GRs of S and O are defined exclusively on the basis of their relationship to the verb. For example, S is the element with which the verb agrees, O is the element on which the action represented by a transitive verb is reflected. These definitions have been challenged by different authors (Keenan 1976; Comrie 1989; Givón 1997), who consider GRs multifactorial concepts that can be described in prototypical rather than categorical terms. According to this approach, the GRs are manifested not only through the agreement but also through semantic, morphological characteristics, which belong to the NPs, and pragmatic and semantic characteristics that concern the whole sentence. These characteristics are grouped into coding properties, i.e., morpho-syntactic coding marks, such as case, agreement control, position in the sentence, and functional properties, i.e., syntactic mechanisms such as coreference, the possibility of admitting passive diathesis, etc. (Givón 1997). Underlying this approach is the assumption that the process of categorization of the linguistic facts follows the same path we use to describe the world (Rosch 1973), i.e., data observation and identification of the perceived salient features through which we construct a representative model of reality: a prototype. Similarly to the phenomena of reality, for linguistic description, categories must also be defined from the data and the identification of their perceived salient features, which construct a representative model of them. Admitting that they can be conceived of as a prototype, rather than as a closed set of defining features, makes it possible to consider their realization as more or less close to a model and to capture aspects that otherwise would find no place in the syntactic treatment.

Thanks to the contribution of the theory of prototypicality (Rosch 1978; Givón 1997), GRs, especially S, are considered in gradual terms, that is, they have multiple encoding properties (case morphology, verbal agreement, and constituent order) that may or may not be co-present depending on the greater or lesser prototypicality of the argument they encode. In other words, these theories have shown how the semantic plane (with which argument description is associated) and the formal plane (morphological-syntactic encoding) function simultaneously. As we will see, contrary to what has been argued in syntax works (Mel’čhuk 1988), lexical semantics contributes to the definition of GRs and is not a separate descriptive level. Nouns express semantic, lexical, and discursive features that help define GRs, as other studies on verbal sentence have already implicitly recognized (Chafe 1976; Keenan 1976; Hopper & Thompson 1984; Givón 1997). Speakers tend to associate more morpho-syntactic encoding marks of a GR with those NPs that have its prototypical semantic features and are close to the model of its most representative semantic role. Conversely, they tend to encode a GR with fewer morpho-syntactic marks when it deviates from the prototype of its semantic role.

The lack of a formal encoding trait, therefore, does not rule out the possibility of considering an NP as a realization of a given GR, but instead indicates the non-prototypicality of its argument. Even if NPs do not exhibit all the encoding properties of a GR, they can retain their semantic properties and exhibit the behavioral properties of that relation. As a result, there can be NPs that are not encoded as S or as O, but undergo syntactic processes that confirm their grammatical nature of S or O. For example, in the case of S, an NP may not be in the nominative case, where intended, but has some syntactic coreference features of its prototypical semantic role, i.e., that of Agent.

Referring to this approach, I assume that GRs are not defined only on the basis of the finite verbal form, but also on semantic, pragmatic factors that belong to the constituents and to the constructions in which they appear (Keenan 1976). The starting point of my research is to study how the marks that in verbal sentences allow the identification of S and O, (i.e., person, which in turn includes number and, in some cases, gender) can be manifested in Vls by the lexical semantic and coding features of NPs (Corbett 2001) and by the presence of participants of communicative situation. Typological studies have shown that the category of person can manifest itself through different forms and that in verbal sentences the agreement of person tends to correlate with semantic features that belong to nouns, such as precisely the animacy and the definiteness of their referents. Considering the NPs that make up Vls, I analyzed, therefore, the animacy and the definiteness of referents expressed by nouns and pronouns in relation to properties of the sentences, such as modality and predicativity. My analysis considers Vls consisting of at least one NP, because the expression of S and O is closely related to the nominal category.

In Vls person is not expressed through verbal affixes but is interpreted equally by speakers, thanks to the contextual and co-textual mechanisms that come into play in the construction of the reference (Siewierska 2004), as it happens in structures that Siewierska (2004) defines zero forms marks of person, i.e., structures where the category of person is not expressed by linguistic marks, because its referent has a high degree of accessibility and tends to be associated to speaker or listener present in the communicative situation. Languages show that the less formalized person marks are, the more their referents are accessible in the context and co-text. For example, the desinences of verb forms that are realizations of person with less formalization, compared to free personal pronouns, tend to agree with S, because it is the syntactic function that realizes highly accessible referents more than the others. Thus the absence of person marks leads interlocutors to interpret the speaker and addressee as the only possible referents in this category, because they are the ones most accessible to the enunciative situation. A clear example may be the infinitive form used to express the negative imperative (e.g., Non urlare ‘do not shout’). In this case using a verbal form that does not have the person mark, the addressee is interpreted as the referent of the unspoken person, as it is also the most accessible in the enunciative situation. In languages like Japanese Vls are systematic and also fall squarely among the zero forms marks of person (Siewierska 2004: 22).

From my analysis it emerged that the absence of the verb form, therefore, does not equate to the absence of GRs, but to a different way of encoding them. GRs can occur in Vls and that their expression has traits in common with the non-canonical encoding of GRs studied in verbal sentences (Aikhenvald et al. 2001). S and O manifest themselves in a non-prototypical way in Vls and present the features of non-canonically encoded core arguments of some verbal sentences e.g., passives, in which S does not present all its protoypical coding and functional properties and it has a low degree of agentivity (Aikhenvald et al. 2001; Benedetti 2013; Holvoet 2015).