Multiword Modifiers in some Romance languages. Semantic formats and syntactic templates

1 Multiword Modifiers^[1]

This analysis focuses on Multiword Modifiers (henceforth MMs), a specific group of Multiword Lexical Units functioning as modifiers of nouns (Multiword Adjectives) and of verbs (Multiword Adverbs). In particular, this investigation will consider MMs showing the syntactic configuration of a prepositional phrase and fulfilling an adjectival function (1), an adverbial function (2) or both (3):

In general, these configurations cannot be attributed tout court to the categories of adjective and adverb. Nevertheless, in specific syntagmatic contexts, some of them can have an adjectival or an adverbial function. I will refer to them as Multiword Adjectives (1), Multiword Adverbs (2) and Mixed Modifiers (3). This phenomenon receives very little attention in grammars and in lexicographical works, especially in the Italian area. However it is worth further investigation, since it is quantitatively important in Romance languages. The table below represents some examples from Italian, French and Spanish.

From the lexicological point of view, some analyses have been driven on both Multiword Adjectives^[3] and Multiword Adverbs^[4] of Italian, French and Spanish. However, they are mostly based on structural properties and do not consider the interaction of syntactic and semantic properties of these units: a simple structural classification does not reveal the intrinsic features of groups of Multiword Modifiers, which are semantically and syntactically similar to each other. This investigation explores the hypothesis that some Multiword units – sharing similar semantic and morpho-syntactic properties – originate from “pre-packaged”^[5] syntactic schemes which organize the semantic information connected to specific configurations.

The aim of this investigation is twofold: on the one hand it aims to suggest the hypothesis that some sequences are formed on the basis of regular patterns and governed by a restricted group of co-selection rules, which can structure and organize the semantic information associated with words. On the other hand this analysis aims at showing that this kind of MLUs represent a broader typological phenomenon, which is massively attested and is quantitatively relevant in Italian as well as in French and Spanish. In this perspective, after a brief excursus on the main properties of Multiword Lexical Units (§2), I will explain the methodology that has been applied to construct a database devoted to analytical research (§3), as well as the general theoretical framework of this investigation (§4) and the formalism used for the representation of semantically related Multiword Modifiers (§5). The last three sections are devoted to the analysis of some semantic relations frequently fulfilled by Multiword Adjectives (§6), Multiword Adverbs (§7) and Mixed Modifiers (§8) of Italian, French and Spanish.

2 The heterogeneous nature of Multiword Lexical Units

Since earlier studies, Multiword Lexical Units have been attributed several types of denominations, definitions and classifications.^[6] The main difference between the diverse theoretical approaches lies on the level of analysis leading to the definition of this phenomenon. In fact, Multiword Lexical Units are mainly identified on the basis of their syntactic properties (e.g. inseparability of constituents, fixed order and presence of dedicated syntactic structures) and semantic properties (e.g. the paradigmatic invariability and non-compositional semantics). Recent studies have shown that, given the heterogeneous nature of Multiword Lexical Units, a description of this phenomenon could only be provided by an approach based on the close relationship between the different levels of analysis.

In this contribution Multiword Lexical Units (henceforth MLUs) will be referred to as to complex lexical items composed of two or more words, syntactically bound. Hence, my theoretical and methodological approaches are based on two main assumptions: on the one hand, MLUs can only be analysed considering the simultaneous interaction of multiple levels of analysis; on the other hand, the notion of MLUs has a scalar nature and has to be modulated along a continuum of fixedness.

MLUs constitute a heterogeneous group, which can be analysed along different dimensions (i–iii) or properties (iv–v), such as:^[7]

1. function;

2. syntactic configurations;

3. morphological, syntactic, semantic and lexical restrictions, semantic idiosyncrasy (stronger or weaker compositionality);

4. fixedness and cohesion;

5. schematicity or degree of lexical specification (more or less lexically specified configurations).

Multiword Lexical Units are subject to different types of restrictions:^[9]morphological, semantic, syntactic and lexical restrictions. However, the extent to which restrictions are enforced may vary according to the type of MLU (Svensson 2004, Lavieu 2005, Piunno 2015).

On the basis of the fixedness of their constituents, MLUs show a lower or a higher degree of syntactic cohesion: the stronger the cohesion between its constituents, the higher the fixedness of a MLU.

MLUs may also show several degrees of lexical specification. Some MLUs are characterized by lexical variation, since their constituents may not be fully lexically specified. Some MLUs allow for the substitution of one of their constituents, even if the paradigmatic choice may be limited to particular semantic fields.

The notion of MLU has therefore a scalar nature and has to be modulated along a continuum of semantic, lexical and syntactic fixedness. On the basis of the interaction between morphological, syntactic, semantic and lexical criteria, I therefore propose a constructional internal grading, which distinguishes between partially fixed MLUs^[10] and totally fixed MLUs^[11] (Piunno 2013, 2015). While totally fixed MLUs do not allow for any variations and are fully lexically specified, partially fixed MLUs are partially open to lexical or syntactic changes (Piunno 2015). Partially fixed MLUs show a lower degree of fixedness and cohesion of their constituents, as well as a lower degree of lexical specification. However, partially fixed MLUs are not equally open to paradigmatic variation: depending on the degree of their syntactic frozenness, MLUs can therefore show a reduced or extended paradigmatic variation.

3 Methodology

As a first step, the extraction of data from an Italian corpus^[12] and the consultation of major Italian lexicographical works^[13] led to the identification of the most productive Italian syntactic configurations of MMs.

Secondly, the collected exemplars of MMs were divided into different classes, depending on their syntactic and semantic properties: exemplars sharing the same syntactic configurations and denoting the same semantics have been grouped into a single “type”.

Thirdly, the semantic relations most frequently performed by MMs have been identified, as for example:

• shape (Section 6.1): the external form or outline of something;

• working principle, means and instrument (Section 6.2): the mechanisms or the means that allow a device to operate;

• purpose (Section 6.3): the reason why something is created;

• suitability (Section 6.3): the appropriateness of something to a particular function;

• manner (Section 7.1–7.2): the way something is done;

• temporal and spatial progression (Section 7.3): the temporal or spatial act of progressing;

• temporary state (Section 8.1): the momentary condition that someone is in at a given time).

At a later stage, nouns involved in MMs have been classified in different semantic classes, on the basis of the classifications made in LexIt and Wordnet.^[14]

Finally, I have made a comparison between patterns of Italian and their equivalents in French and Spanish. Spanish data have mainly been collected from the Corpus del Español Actual (CEA)^[15] and those of French from the Chambers-Rostand du français journalistique corpus^[16]. Combinatory dictionaries^[17], on-line monolingual dictionaries^[18] and internet pages have in some cases been consulted.

4 The organisation of the semantic information

My theoretical framework is in line with the models of Construction Grammar,^[19] but it is particularly bound to the Construction and Category Grammar – CCG (Simone 2006a). My starting assumption is that the transition from the meaning of a MLU to its syntactic form consists of two different levels: the cognitive-conceptual (pre-linguistic) level and the purely linguistic level (Simone 2006a). At the cognitive-conceptual level, specific patterns (the semantic formats) are able to structure and organise the semantic information associated with a word (Simone 2006a). At the linguistic level, semantic formats are connected to particular linguistic structures (syntactic templates), which represent specific meanings. Syntactic templates can therefore identify the semantic templates of different but conceptually similar constructions.

Even if it would be unfeasible to trace the semantic organisation of all MMs – since their semantics is often idiosyncratic –, it is possible to assign a specific range of relational meanings to some syntactic patterns. In fact, syntactic templates give rise to a considerable number of lexemes sharing similar morpho-syntactic characteristics and semantic properties.

On the basis of a shared regular syntactic structure related to particular sets of meanings, it is therefore possible to subdivide MMs into two main groups:

1. semantically unpredictable: prepositional Multiword Adjectives or adverbs not sharing any syntactic and semantic features with other groups of examples; they are mostly idiosyncratic and, as a consequence, their formation is not to be attributed to any general syntactic rule. The syntactic structure and the preposition employed are not a unit of specific semantic information. Examples of semantically unstable MMs are lexemes with metaphorical meanings, such as in erba (‘young’, literally ‘in grass’), al bacio (‘excellent’, literally ‘to the kiss’), in gamba (‘capable’, literally ‘in leg’).

2. semantically predictable: MMs sharing a specific syntactic configuration with other instantiations and creating a single network of semantic relationships. In this case the semantic format and the syntactic template of the prepositional phrase can suggest particular sets of meanings.

5 The representation of Semantic formats

Syntactic templates are linguistic configurations associated with specific semantic formats. Each configuration responds to co-selection constraints and entails a double semantic restriction: on the nouns or verbs that are modified by the prepositional phrase (N₁ or V); and on the noun included in the adjectival (N₂) or adverbial Multiword (N₁).

The representation of the semantic format of a MM includes both the semantic relation it expresses and the syntactic template associated to it, as follows:

Figure 1:

Representation of the Semantic Format

A syntactic configuration comes into being when a) it is instantiated by a high number of exemplars; b) the exemplars constitute a semantic (or morpho-syntactic) “type”, sharing semantic and morpho-syntactic properties. More types are admitted in a configuration, the more schematic and productive such configuration will be (Bybee 1985, Bybee and Thompson 1997, Barðdal 2008).

The next paragraph is dedicated to the most productive syntactic templates of the most representative Italian MMs formats and their equivalents in French and Spanish.

6 Multiword Adjectives

Multiword Adjectives are remarkably spread in Italian, since they show a significant quantitative extension.^[22] Their syntactic configurations can be various,^[23] but the most exploited pattern is [Preposition + Noun].^[24] Even if these configurations cannot be attributed tout court to the category of adjectives in specific syntagmatic contexts, they can have either an attributive or a predicative function (Piunno 2015).

7 Multiword Adverbs

Multiword Adverbs represent a heterogeneous group, since the adverbial function can be notoriously expressed by different lexical elements (such as nominal phrases, prepositional phrases or propositions). As far as their syntactic configurations are concerned, different combinations are possible. However a great number of Italian multiword adverbs have the syntactic configuration of a prepositional phrase.^[27] From a functional point of view, prepositional phrases with an adverbial function can entail different syntactic relations: they can appear as verb modifiers, as adverb modifiers, as adjective modifiers or as sentence modifiers, just as prototypical adverbs do (Piunno, 2015).

8 Mixed Modifiers

Mixed modifiers are MMs having both an adjectival and an adverbial function, depending on the syntagmatic context. Even if not as frequent as the former two groups of MMs, they are an expanding class in contemporary Italian, whose productivity and diffusion are definitely increasing.

9 Conclusion

Predictable MMs are governed by co-selection rules which are strictly connected to the syntactic configuration and to the preposition involved. All Romance languages considered show a regular use of an analytical resource (MMs structured as prepositional phrases) in adjectival or adverbial function. Some of these sequences are system resources, while others are created ad hoc in the discourse.

The Romance languages considered share the same semantic patterns and similar syntactic templates. This happens also in other languages: for example, Multiword Adverbs and Mixed Modifiers also occur in English. However no similar regularities are there in English for some Multiword Adjectives, and in particular for the semantic format of “shape”, which is typically realised as a monorhematic adjective and not as a prepositional phrase (cf. examples in §6.1).

MMs of this kind seem therefore to be a phenomenon which is peculiar to Romance languages (cf. Piunno 2013). The general common feature of Romance languages being the progression to analyticity, the transition from synthetic to analytic clearly affected mechanisms of word formation (cf. Bally 1963 [1950]; Frei 1969 [1929]). It is worth noting here that this kind of mechanisms should be analysed not only from a synchronic point of view, but also from a diachronic one. The diachronic point of view could explain processes of formation of configurations which are no more productive in contemporary languages, but which just remain as isolate instantiations of semantic formats. The analytical process creating MMs from nouns appears already embedded in the first stages of Romance languages (Piunno & Ganfi 2014).^[34]

The peculiarity of Romance languages lies, indeed, in the fact that they have exploited prepositions as resources to create new lexemes (Mejri 2004, 2007). This types of Multiword units may function as derivational suffixes where proper suffixes may not be admitted or may not exist: this is the case of the sequences viaggio per mare 'sea travel’ or albero in fiore ‘tree in blossom’, which have been described in Bally as a general procedure for the creation of new adjectival constructions (1963 [1950]: 293). Such analytical resources are therefore parallel – although only partially overlap – to the derivational system of suffixes. Thus, syntax operates besides morphology in word formations processes (Piunno 2013, 2015): the more a multiword configuration is schematised, the more productive a constructional schema (cf. Piunno and Ganfi 2014). Since frequent patterns are more predictable than others, they may require a little effort to be memorized: as a consequence, the less is the variability, the major the linguistic economy (Frei 1969 [1929]).

This provides relevant information and new research directions for synchronic and diachronic comparative analyses on the syntactic and semantic nature of MMs in the Romance family.