Rethinking lexical semantic fields: relevance and local holism

1 Introduction

Upon first encountering Göran Sonesson’s work, a striking parallel^[1] emerged between his perspective of a phenomenological cognitive semiotics and the semiotic and semantic assumptions underlying the lexical-semantic theory that I had previously attempted to outline, rethinking the notion of “lexical field,” and drawing on the Humboldtian and Saussurean legacy as reworked by Tullio De Mauro^[2] and other eminent structuralist scholars.

As is well known, in recent decades, semiotics and semantics, especially from the cognitive area, typically exhibit an aversion to structural semantics, and they often fail to acknowledge the epistemological and methodological contributions that have shaped twentieth-century semiotic and linguistic research. Although the fundamental concepts underlying cognitive approaches are often derived from these lines of thought, (a vulgate version of) structural semantics is habitually condemned on the basis of two prejudicial stereotypes: the autonomy principle (with a subsequent trauma of immanence) and a “disembodied” and/or “acontextual” bent (with a subsequent anxiety to root all knowledge in the body and/or in the environment).

The definition of the linguistic unit in terms of value^[3] (Saussure 1959 [1916]: 114–117) that emerges from the relations between signs within the system to which they belong, apparently indifferent to the life of the speakers (i.e., their body type, non-linguistic cognitive faculties, environment, and social and cultural contexts), is taken as a postulate that hinders an evaluation of the multiple doors that structuralist thinkers have left open to a semantic theory based on more “ecological” principles.

In contrast to this double stereotype, as will be shown, questions concerning the interplay between language and cognition are extensively discussed in European structural semantics. For example, lexical field theory is not only concerned with intralinguistic lexical organization, but also offers a model of lexical competence that might be disengaged from the Saussurean vulgata^[4] and extended to broader semiotic competence. As will be seen, a semiotic semantics such as the one outlined by De Mauro, albeit in the vein of structuralism, does not contradict basic assumptions about the embodiment of cognition and, in part, of language, nor does it deny, rather it tirelessly emphasizes, the situatedness of semiotic activity.

Still, the effort to define lexical meaning and its cognitive and even pragmatic aspects,^[5] as well as to adapt the original complex formulation of lexical field theory to current concerns, requires a considerable investment of energy, since it often involves translating contested notions rooted in the history of linguistic ideas by shifting from one paradigm to another. Moreover, within the field of cognitive semantics, the lack of attention to the history of linguistic ideas as well as the absence of adequate metalinguistic considerations necessitate a more substantial exertion in order to first clarify their epistemological scope.

In this respect, as will be shown, Sonesson’s reflections provide a valuable beginning for considering the potential for dialogue between some outcomes of the structural tradition and some current cognitive-phenomenological approaches, both because a phenomenological substratum is already intertwined at least with some branches of structuralism (cf. De Palo 2016, 2022; Sonesson 2021), and also because the Swedish scholar examines the thought of classical authors who developed the basic conceptual tools of semiotics, linguistics, and semantics, in the light of which he frames and critiques current movements. Accordingly, the convergences between a “heretic” (at least with respect to the vulgata) structural approach and Sonesson’s phenomenological cognitive semiotics suggest a favorable context for a rethinking of lexical field theory.

In order to present my arguments in an orderly fashion, in Section 2, I will briefly illustrate the notion of pertinence (relevance) as it matured in structural linguistics, from phonology to Prieto (1975), who, among structuralists, made perhaps the most remarkable extension of Saussurean premises to get to a comprehensive theory of human knowledge (Fadda 2015: 98). Prieto’s insights have influenced De Mauro’s semiotics, which, for the purposes of this paper, provides a more coherent framework for combining the premises of a Saussurean semiotic semantics (De Mauro 2019: 65) and the phenomenological approach. More space will, therefore, be devoted to his thought, as it is the background against which I have been able to review some of the theoretical limitations of the early lexical field theory. In Section 3, the notion of the lexical field is introduced, with an examination of some aspects of the Humboldtian and Saussurean upbringing from which it developed. It is no coincidence that it is precisely in the work of Coseriu – a proponent of a Humboldtian structuralism (Hassler 2015: 21) – that some Humboldtian, Saussurean and Neo-Humboldtian traces gain the necessary coherence to question the nature of linguistic competence from the perspective of a linguistics and a philosophy of language conscious of their own tasks (Coseriu 2019; cf. Diodato 2021, 2022). In Section 4, I will offer a concise discussion of the notion of encyclopedia in order to finally arrive at the perspective of local holism underlying the notion of lexical-semantic field. Conclusively, in the last section, I will reframe the previously explored notions with respect to the phenomenological perspective outlined by Sonesson, aiming to demonstrate their mutual consistency, which will be pointed out throughout the paper.

2 Pertinence/relevance

In the field of structural linguistics, the concept of relevance (pertinence), which can be traced back to Trubetzkoy, Martinet, Hjelmslev, and Prieto, has been developed primarily in relation to that of function. It is a fundamental principle of the Saussurean epistemology, consequent on the claim that langue is a “system of classification” (Saussure 1959 [1916]: 120). Accordingly, a linguistic unit is identified on the basis of one or more features (called, precisely, pertinent traits) that distinguish it from other units within the same class or system.

This notion of pertinence has served, as is well known, as a theoretical basis of structural phonology. Thus, a phoneme is defined as an abstract linguistic unit (type) distinct from its correlative phones (tokens), which is classified as such because it possesses at least one feature by which it is opposed to any other phoneme. In a class, free variations are permitted until they do not affect the functional level. The identification of the phoneme is contingent upon the commutation principle: when a change on the plane of the signifier leads to a change on the plane of the signified, two distinct phonemes are found.

The phonological model has been further developed – potentially in excess of Saussure’s indications – up to the Hjelmslevian hypothesis of the symmetry of the sign. Even on the plane of content,^[6] it is possible to identify minimal elements that the Danish linguist called content figurae, displaying a functional value (Hjelmslev 1963 [1943]: 46–47).

Although structural semantics has mostly applied this principle to langue as a semiotic object without any ontological commitment,^[7] most second-generation cognitivists have equated this method with the so-called classical theory of categorization, a model (erroneously) attributed to Aristotle and sometimes Locke, according to which the (pertinent) traits by which a category is identified are abstract, arbitrary, primitive, universal, and (occasionally) innate^[8] (Taylor 1995 [1989]: 34–35; cf. also Lakoff 1987; Lakoff and Johnson 1980; Taylor 1999).

In particular, the debate has focused on the criticism of the principle of arbitrariness, also misunderstanding the meaning of “abstract” when referring to linguistic and semiotic categories. In brief, in the framework of categorization theory, (radical) arbitrariness has been understood as trivial objectivism,^[9] conventionalism^[10] or nominalism.^[11] Among other idiosyncrasies, this line of criticism has led to a confusion between “theory” (linguistics, semiotics) and “object” (language and languages), a distinction upon which any science rests, as remarked, among others, by Coseriu (cf. Diodato 2022: 115–116).

In order to examine some of these contentious issues, I will consider a very broad sense of pertinence that structuralist linguists and semioticians have outlined on the basis of the other Saussurean epistemological principle, according to which in linguistics “it is the viewpoint that creates the object” (Saussure 1959 [1916]: 8). Martinet’s vivid metaphor illuminates this point: “The weight of the material used in the facade of a building has repercussions on the structure of this building, and yet the grain of this material, its external aspect, its aesthetic qualities insofar as they do not cause a variation in weight are non pertinent from the point of view of the architectural structure” (Martinet 1965: 292, emphasis mine).

Thus intended, the pertinence of linguistic units is contingent upon the perspective from which they are observed. Rather than being entities in and of themselves, linguistic units (and signs in general) are the result of a particular mode of perceiving and conceptualizing reality. This insight was the foundation upon which Prieto attempted to build his theory of semiotics as a theory of knowledge (showing no little similarity to Peirce’s; cf. Fadda 2015: 106). Prieto’s words further clarify this notion of pertinence and leave no room for equating the structuralist model with the “infamous” classical model (at least regarding the supposed “objectivity” of the pertinent traits).

As he puts it, “since pertinence never comes from the object, there is no knowledge of material reality that is ‘objective’ in the sense that it can passively stand before the object and reflect it as it is or, in any case, that owes nothing to the object.” The operation of pertinentization is never abstract (even though it produces abstract categories or classes); rather, given that “pertinence is determined by a subject who is always a social subject, it follows that we cannot have a knowledge of material reality that is socially neutral” (Prieto 1975: 149, translation and emphasis mine).

As Sonesson (2018) also observes, the theoretical implications of this deeper sense of pertinence have not yet been fully explored. I believe that this depends on the misunderstanding of the principle of radical arbitrariness, which Sonesson himself views with suspicion, giving some credence to the vulgata. By contrast, maintaining that the activity of pertinentization is radically arbitrary does not entail, as will be seen, that it is primarily (if not exclusively) linguistic,^[12] nor that it is carried out by a disembodied langue and not by flesh-and-blood speakers. So, in order to get to the bottom of the notions of pertinence and radical arbitrariness, it is worth introducing the basic principles of De Mauro’s semiotic and semantic theory, which takes its starting point from an “authentic” Saussure, demonstrating, among other things, that the principle of arbitrariness and the so-called classical model are indeed far apart.

3 Fields

The concept of “field” gained great prominence in last-century linguistics, although it remained ambiguous and loosely delimited. It has been a “migrant notion” (De Palo 2019), with numerous attempts at definition proliferating across several research domains between the nineteenth and twentieth centuries. The principal philosophical interest of a field approach lies in its contiguity with the concept of totality, that is, “the mereological relation between the whole and the parts, subject and object, continuum and discrete, in the frame of the traditional dualism between the supporters of matter and friends of ideas” (De Palo 2019: 201, cf. Cassirer 1923). This is evidenced by the fact that the notion of field crossed a complex path involving the Gestalt school and phenomenology, as well as rising structuralism, where it was naturally related to the notion of “organism,” i.e., the broad idea of a “mutual interdependence between functions and structures, with the consequence that any change to any of these parts inevitably affects the others” (De Palo 2019: 201).

On the side of linguistic theories, Ščur (1978: 9) notes that in the early twentieth century the term was used to describe a multitude of distinct phenomena.^[30] During that time, the number of disciplines in which it has been employed and the number of definitions it has received have been steadily increasing, corroborating its “pandemic” nature (Ščur 1978; cf. also Herbermann 1995). Since I cannot examine the full extent of the field concept in linguistics, I will focus on lexical field theory, which most closely followed the Saussurean definition of langue along a Humboldtian theoretical line.

4 From encyclopedia to local holism: fields and frames

Generally exemplified by the dichotomy between dictionary and encyclopedia suggested by Eco (1984), structural semantics is based on the belief that a purely linguistic dimension of meaning can be discriminated, i.e., that a dictionary-type model is adequate for lexical analysis, and that, likewise, an autonomous lexical competence can be isolated within a broader communicative competence (this idea implies a kind of modularity of communicative competence, an intriguing issue that, however, must be left out of this essay).

This structuralist tenet has been accompanied by a long-standing epistemological debate, which has explored its limitations while highlighting its methodological necessity.^[35] Conversely, while not elaborating on Eco’s distinction, but rejecting a Chomskyan-like demarcation between syntactic, semantic and pragmatic competence, cognitive semanticists have challenged this principle in the name of a non-autonomy of language (and linguistics). Instead of fields, they have advocated the notion of frame precisely because it appears more appropriate to capture the encyclopedic dimension of meaning.^[36] However, in most of its uses, the notion of frame is allied with a conceptual reductionism that treats language as nomenclature. Although cognitive theories give different weight to this attitude, it is generally argued that concepts are structured through embodied and inherently experiential processes, while language represents, if anything, the tip of the iceberg of conceptualization processes, reducing the cognitive function of natural-historical languages to zero.

On a more critical theoretical front and against the background of Eco’s distinction, Violi (2015: 89) assesses the “state of the art” of lexical semantics arguing that it “seems to be trapped in a paradoxical situation: on the one hand, it seems natural to associate lexical items with stable meanings; on the other hand, whenever one tries to describe these meanings in a generalized and principled form there seems to be no viable way out.” Indeed, given the sign’s oscillation between the poles of determinacy and indeterminacy, “semantic models seem always either too rich and constrictive, or too poor and simplistic” (Violi 2015: 89).

As has been mentioned, lexical field theory has been judged as one of the richest and most constrictive semantic models.^[37] In fact, being entirely and exhaustively structured as mosaics, fields have been interpreted as abstract, dead, artificial constructs, unable to capture the living structuring of the lexicon, which more closely resembles a loose and flexible cobweb. However, that lexical systems are (sufficiently) stable both synchronically and diachronically is incontrovertible evidence for speakers, to the point that claiming that fields are too constrictive to capture the indeterminacy of meaning and its contextual variability does not entail going to the opposite extreme, that is, the conclusion that lexicon shows no systematic organization. Speakers, as De Mauro (1999 [1965]: 18) suggests, have the intuitive certainty that the hic et nunc determination of meaning requires not only the ability to grasp significant contextual clues, but also to rely on a sedimented semiotic system that allows them to move through the encyclopedia. If the Neo-Humboldtian conception seemed to lead to the impossibility of explaining lexical competence in ordinary cases where speakers do not have knowledge of the whole field and its structure (holism problem), the local holism hypothesis (Kittay and Lehrer 1992) may prove to be a decisive step in the direction of a theory of lexicon that considers the structure of fields as loose and context-sensitive as much as stable enough to allow mutual understanding.

One might, however, observe that a certain degree of richness and constrictiveness is necessary for synchronic or diachronic lexical analysis, unless one admits that there exists an inalterable semantic core (comparable to a “literal meaning”) behind the variability of communicative practices. If, on the other hand, it is disclosed that the meaning of a word is determined by an agent in the encounter between code and praxis, then the hypothesis of a core meaning merely falls. Borrowing Bühler’s (1934) terminology, a linguistic sign always occurs in a deictic field, assuming one determined meaning and not another, because speakers share the same code (that is, symbolic field). Consequently, when, due to a competence deficit or any other contextual or other interference, a pertinent meaning cannot be identified, it is always the common ground of langue that allows speakers to jump the ditch, thanks to the resource of metalinguisticity.

Commenting on the encyclopedic drift of current lexical semantics, Violi correctly notes that the Encyclopedia – which Eco himself intended as a “semiotic postulate” – is not, and cannot be, a lexical model. What the notion originally emphasizes is that “lexical meanings may be connected to a highly complex knowledge background” (Violi 2015: 89). If a dictionarial model might work for the description of single texts, the encyclopedia must be correctly understood as a model to demonstrate that sign meaning is adaptable according to specific uses. In line with De Mauro’s principle of the sign’s indeterminacy, encyclopedia refers to “a general semantic potential of words” that “cannot be fully described, but only locally reconstructed” (Violi 2015: 89–90). Here is the task of lexical semantics: to reconstruct the Encyclopedia locally, finding models to counterbalance the richness and constrictiveness of some structural (and even cognitive) methods and the looseness of most pragmatic approaches.

In an attempt to rethink lexical fields precisely to arrive at such a counterbalance, Kittay assumes a structural definition (indeed, largely borrowed from Lyons):^[38] “Semantic fields are clusterings of lexicalized concepts. Moreover, semantic fields do not only group together semantically close terms, they also encode the differentiations that individuate concepts and terms” (Kittay and Lehrer 1992: 229–230, emphasis mine). The author addresses the question of the nature of fields not in terms of dictionary or encyclopedic models, but in relation to the debate between internalism and externalism: are fields individual constructs, therefore located in the speaker’s head, or structures provided by the language, therefore located somewhere outside the head? According to her, “the answer is Yes! and No!” (Kittay and Lehrer 1992: 245).

On the one hand “to understand a language each member of the language community must have some grasp of the semantic fields to which terms in the sentences of the utters/hearers belong – as well as the understanding that the field is so articulated for other members of the linguistic community,” so that content must be understood as wide content. On the other hand, one cannot deny that “the available distinctions in the fields – the contrasts and affinities that mark out relations among terms and concepts – are individualistically available,” meaning that content must be also narrowly understood (Kittay and Lehrer 1992: 245).

The resulting hypothesis of local holism (Kittay and Lehrer 1992: 242) can be read, in short, as the attempt to find a place for lexical semantics somewhere between the individual mind and the external world. Local holism thus converges on a weak notion of langue such as that which emerges from the reflections of De Mauro and Sonesson.

This view challenges both radical externalism or contextualism, which leads to the evaporation of the notion of code (or “scheme of interpretation”), and a hypostatization of langue, which downplays the connection between semiotic activities and external reality, between language and history, between language and the subjects without whom it would have no life.

Consistent with the above, a weak conception of langue/code calls into question the very possibility that the encyclopedia can represent a model for both lexical analysis and lexical competence. In fact, while the encyclopedia contains the general semantic potential of signs, the communicative act also requires an (at least provisional) formalization, i.e., an ability to delimit and articulate the lexical field on the basis of the norms sedimented in the community. Ultimately, lexical competence bears on the ability to continuously shift from narrow to wide content, i.e., from in-head to out-of-head contents. If, out of context, a word refers to the intricate cobweb of virtual relations with other words, in context – that is, in ordinary communication processes – the shear of the dictionary comes into play. As Sonesson (2021: 117) puts it, “the concrete situation serves to prune the wild wood of the encyclopedia into the semblance of a Porphyrian tree.”

5 A phenomenological lexical semantics?

Faced with the alternatives of internalism and externalism and of dictionary and encyclopedia, current cognitive lexical semantics complains of three discomforts.

The first might be called the trauma of immanence that cognitivists attribute to the alleged subordination of structural models to a classical theory of categorization. The broad structuralist debate which I have sketched shows how unjustified this trauma is, since the models of lexical analysis proposed by structural semantics (and the notion of field, specifically) start precisely from a conception of linguistic activity as energeia, in order to pinpoint a method of analyzing language also as product (ergon). The idea that the meaning of a word depends on intrasystemic relations is accompanied, as already argued, by an incessant epistemological reflection on its own limits, given the interaction between langage, langue and parole when moving from the descriptions of linguistic products to the lexical competence of the individual speaker. As De Mauro argues, maintaining a weak conception of langue, the play of signs is not confined to the inanimate linguistic system: the langue vivante is deeply rooted in the historical community and ultimately in the individual mind, which is shaped by its social situatedness and biological substratum.

The second discomfort consists of the anxiety to treat language as parole, that is, language as a plurality of weakly or unstructured linguistic acts, entrusting the determination of meaning (almost) entirely to contextual factors.^[39] This radical externalist, pragmatic, or enactivist drift risks sweeping away the semiotic dimension underlying the processes of categorization and communication (cf. Diodato 2020). Precisely with regard to contextualist drifts, Violi (2015: 92) correctly notes that the notion of context itself would require foundational work. A kind of “black hole” for lexical semantics, “context” can range from the simple linguistic environment (cotext) to the virtually infinite set of all the elements that can affect utterances; consequently, a lexical semantics that relies too much on it implies that the semantic potential of a word can be confined only or predominantly outside of language. Furthermore, in cognitive trends there is also a risk of conceptualizing context, which would lead to the same dead ends as traditional lexical semantics (cf. Diodato 2024), bridled by the difficulty of discerning between what is known and what is about to be known in the current linguistic act or will be known in the linguistic acts to come.

The anxiety to treat language as parole must be distinguished from a third inclination of cognitive linguistics and semantics to root all types of knowledge, even the cultural-historical knowledge typically encoded in language(s) through the two filters of lexicon and grammar, in the pre-semiotic dimension, or even in the labyrinth of neural structures. In this regard, cognitive semiotics, as a “transdisciplinary field focusing on the multifaceted phenomenon of meaning” (Zlatev 2015: 1,043), can provide a basis on which to rework the limitations of both structural semantics (too much langue, or codes in general) and cognitive semantics (little or no codes), without throwing the baby out with the bathwater (cf. Sonesson 2007).

In support of what has been argued so far, Sonesson’s rethinking of the notion of relevance may offer the thread to stitch up a notion of lexical field within the framework of a weak conception of langue that nonetheless does not give in to the impossibility of identifying lexical structures, albeit recognizing their inextricable relationship, on the one hand, with bio-cognitive bases and, on the other, with context (from the hic et nunc communicative situation to the socio-cultural background).

Shaping a cognitive semiotic framework that critically intersects basic assumptions of structural semantics with phenomenology, Sonesson proposes a notion of relevance – or better, “systems of relevancies” – drawn from “the phenomenological tradition stemming from Schütz and Gurwitsch, including its antecedents in Husserl’s works” (Sonesson 2018: 22).

With questions similar to those answered by De Mauro in the wake of Saussure, he then asks whether systems of relevancies: (1) imply (a1) that the creation of meaning is always situated in a particular context or (b1) that a system of relevance is embedded in the typical structure of the world, which is normally taken for granted; (2) (a2) are the result of the operations of an innate mental module or (b2) presuppose a socially distributed type of cognition; (3) (a3) play a role in the representation of something new, or whether (b3) the main contribution of these systems is to create the background of knowledge in which something new can emerge.

Consistent with the arguments above, these questions can be (at least provisionally) answered as follows: as lexical structures that encode experiences pertinent to human societies, given the bodily and cognitive constraints of the users (De Mauro’s material arbitrariness), fields formalize a system of relevancies (De Mauro’s formal arbitrariness) rooted in the typical structure of the taken-for-granted world. As their structure is susceptible to semantic indeterminacy, that is, indefinitely extensible or restrictable, fields regulate the creation of meaning as situated in a particular context. More in general, they plausibly represent ultimate outcome of a kind of social-distributed cognition, which is not restricted to the operations of an innate mental module. Ultimately, as their structure is embedded in langue, they play a role in both reinforcing the norms for the use/application of words as well as in offering the conditions for the breaking of norms, allowing for linguistic change (cf. Blomberg and Zlatev 2021: 43). Lexical fields are, then, not mummified mosaics of words, but “structures of the Lifeworld … as part and parcel of the Homeworld” (Sonesson 2018: 48).