This article analyzes the variability and the frozenness of averbal speech formulas in French, such as bonne soirée / journée (lit. ‘nice evening / day’; ‘have a nice evening / day’), à demain / bientôt (lit. ‘to tomorrow / soon’;‘see you tomorrow / soon’); au travail / au lit (lit. ‘to work / bed’; ‘let’s go to work / to bed’). A corpus study is conducted and the properties associated with lexical variability, syntactic variation, and genres are examined. Observations show that numerous common properties can be associated with the patterns we examined, that can be analyzed as constructional phrasemes, defined as associations of syntactic structures defined as structure-meaning associations including fixed and variable elements. A model of constructional phrasemes is then outlined in the framework of the Prefab project, which aims to identify the most productive constructional phrasemes for speech formulae in French. The analysis carried out shows that a degree of frozenness can be associated with the structure of constructional phrasemes themselves (for example, use as a greeting formula of [Adjbon Npériode]), whereas a certain degree of variability can be observed at the lexical level and in terms of gender and register.

This study highlights the linguistic specificities of loanwords and uses a theoretical model based on the assumption that the English lexicon is composed of different subsystems. Loanwords have specific phonological, letter-to-sound, semantic, morphological, segmental and syntactic properties that distinguish them from the items of the other subsystems. The first part of the study lists these properties and illustrates them with examples taken from dictionary data (Wells [2008]). The formation of the loanword subsystem is thus a product of the dynamic interaction between different linguistic systems (Dabouis & P. Fournier [2022]). The emergence of this subsystem reveals that the theoretical framework for loanwords is rooted in the principles of interaction, categorisation and gradience. The second part of the study proposes a conceptual and dynamic representation of the subsystem, mainly grounded in cognitive linguistic theory (Culioli [1990]; [1999a]; [1999b]; Lakoff [1987]; Taylor [1998]). The category of loanwords is represented by a set whose centre concentrates the prototypical and defining parameters of the category, and from which items that fall within this class are identified. A loanword is more or less close to the prototypical form according to the number of parameters it activates. The notions of gradience and boundary are essential in the conceptualisation of the system, because they contribute to defining the concept of loanword as a category based on a continuum. Less-prototypical elements of the subsystem are defined by a lower concentration of defining characteristics and may even shift to other subsystems. The theoretical model proposed in this study is grounded in the principles of prototypicity, gradience, and boundary and provides a dynamic representation of the loanword category, while also addressing the limitations of linguistic categorisation.

I claim that it is the variability of collocations rather than their restrictedness that makes it a challenge for foreign learners to acquire them and for researchers to categorize them (1.1.) and argue that a usage-based approach drawing on notions from functional and cognitive linguistics makes it possible to account for collocations as a dynamic and flexible language resource. The functional and cognitive framework that I propose for the categorization of collocations is one of several lines of investigation within Usage-based linguistics, an approach evolving since the 1980s as a challenge to the top-down approach of structuralist and generative linguists who study the language system separately from language use (Langacker [2000: 91-93], Herbst, Schmid & Faulhaber [2014], Diessel [2017: 1]). I use corpus data in a qualitative approach based on the theory and methodology developed in Poulsen [2022]. Frequency, which is a complex phenomenon (Geeraerts [2017: 158]), is part of the description of the data while function is considered more important for analysis. In the introduction to the first part of this paper, I give a brief example of the usage-based framework that I intend to introduce before giving an overview of previous approaches that include insights converging on a usage-based approach. In the second part, an alternative functional and cognitive approach is outlined, which is subsequently tested, in the third part, by means of an exemplary study of the noun joke. Using examples from the study of joke, I argue that verbs in conventional and entrenched collocations make up a functional domain and can be construed as a lexicogrammatical prototype category of support verbs.

Lexical knowledge development has always been a seriously challenging task for L2 learners throughout their language learning process, and the necessity of identifying the most efficient L2 vocabulary input encoding for long-term retention is a constant controversy in the field. To address this concern, a quasi-experimental research design was incorporated to explore the effects of single- (Pylyshyn [1978]) and dual-coding (Paivio [1971; 1986]) representation of L2 vocabulary in recall and retention of concrete words in two succeeding and distinct phases. In phase one, four distinct groups of male and female language learners (n=80) received target concrete vocabulary instruction within the single-coding framework via both decontextualized (i.e., word-only L2-L1 flashcards—verbal) and semi-contextualized channels (i.e., picture-supported L2 flashcards—visual). In phase two, three distinct groups of participants (n=60) were provided with decontextualized (verbal), semi-contextualized (visual), and dual-coding (i.e., verbal representations paired with pictures and L1 equivalents) instructions of target concrete words. A comparison of pre-test and post-test scores demonstrated that decontextualized instruction was effective for short-term recall of L2 concrete words in both phases; however, it failed to significantly improve long-term retention of L2 concrete words. Conversely, the semi-contextualized encoding channel significantly well-equipped the L2 learners with two distinct sets of retrieval cues, facilitating both immediate recall and later retention of L2 concrete vocabulary. Meanwhile, the immediate and delayed post-test results of learners who received instruction via the dual-coding channel suggested an additive effect of L2 input representation with L1 equivalents and visual imagery. This combination yielded more remarkable and consistent results in boosting language learners’ short- and long-term recall and retention of concrete L2 words. This implies that exclusively focusing on verbal or visual encoding could hinder L2 learners’ ability in building L2 lexicon in an efficient and progressive manner. The findings suggest the significance of integrating both verbal and visual encoding channels for L2 concrete vocabulary instruction, especially in bilingual and foreign language learning contexts.

Multi-word expressions can be defined as combinations of words with a certain degree of idiomaticity. These expressions can become “defixed,” meaning they undergo formal and semantic changes, such as “V for Vendetta” (James McTeigue, 2005) becoming “V pour Vend des tartes” (“V for we sell pies”, the name of a bakery in the town of Berfray, France). In this article, we study the phenomenon of defixation through the concept of “snowclones,” generally defined as partially fixed phrase patterns (such as “X is the new Y”) derived from or alluding to an expression. We seek (i) to determine whether, as the literature on them suggests, snowclones invariably manifest themselves in a single, sTable configuration and (ii) to demonstrate that snowclones should not be considered specific phraseological units, but rather situated within a broader descriptive framework, among other types of defixed expressions.

The aim of the article is to test the hypothesis that go wrong pseudo-clefts might be frozen sequences with specific meaning and to establish the semantics and pragmatics of these structures. The research is based on a corpus of 880 occurrences (487 pseudo-clefts and 393 reversed pseudo-clefts) collected from the COCA, the BNC and English Web 2021 (enTenTen21) available on SketchEngine. It confronts different approaches to pseudo-clefts; in particular, it draws on studies of pseudo-clefts in interaction and conversation arguing that pseudo-clefts are “not lexically open grammatical constructions” (Hopper [2004]) to shed light on the question of whether pseudo-clefts should be considered as derived from more basic counterparts. The article also uses these studies to explore whether the syntactic, semantic and pragmatic characteristics of go wrong pseudo-clefts are specific to those structures or can be accounted for by the features of the category of pseudo-clefts as a whole. Studying go wrong pseudo-clefts both quantitatively and qualitatively, the article shows that they are semantically frozen structures, since they do not feature the same meaning of the go wrong idiom as non-cleft sentences: go wrong pseudo-clefts do not identify the situation that “bec[a]me unpleasant and [was] not a success” (Cambridge dictionary), as is the case in “the plan went wrong” (COCA), but rather are used to establish what the cause of the problem was. Moreover, the fact that pseudo-cleft go wrong is interpreted in the same way irrespective of whether the predicative complement is a clause or a NP shows that the meaning is at least partially determined by the frozen pseudo-cleft structure itself. This is linked to the fact that a negative situation is often preconstructed in the cotext. In fact, the corpus shows that go wrong pseudo-clefts predominantly occur in contexts where the question “what went wrong?” is preconstructed more or less explicitly. The article uses this to explore the potential pragmatic frozenness of go wrong pseudo-clefts. It questions whether these clefts can be considered as pragmatically similar to do / say and especially happen clefts, which represent the great majority of pseudo-clefts used in conversation (Hopper & Thompson [2008]; Koops & Hilpert [2009]) and have been described as allowing the speaker to “delay an assertion for any of a number of pragmatic reasons” (Hopper [2001: 110]), to give speakers time to “prepare the formulation of an utterance”, to inform the co-utterer of an “imminent significant segment” (Hopper [2001: 110-111]) or to frame the portion of discourse that follows as “event, action, and paraphrase” (Hopper & Thompson [2008: 104]). The article shows that although go wrong pseudo-clefts can be used for such pragmatic reasons, their occurrence is very often linked to the speaker’s desire to signal that they are going to / are in the process of answering a question more or less explicitly preconstructed in the immediate cotext.

In Construction Grammar, many units are depicted as semi-schematic objects, with a fixed part and a slot to be filled by another language unit, leading to the utterance of a construct. These slots are highly structured organizations, obeying a Zipf-Mandelbrot distribution over the frequency of the filler units. Through use, it may nonetheless happen that a filler becomes so favored in the construction that the resulting construct becomes entrenched on its own, leading to a frozen unit that no longer results from the paradigmatic variation of a constructional schema. In this paper, I argue that we can empirically detect these entrenchment events, based on two criteria: structurally, from a deviation of the filler’s frequency to the Zipf-Mandelbrot distribution; diachronically, from a significant increase in frequency share over time. I apply these notions to the edge-based constructional system of complex prepositions, showing that on the edge of town has become a frozen construct. The question arises of the slot from which this construct originated. To answer this question, I fully flesh out the system, concluding that the construct has likely become entrenched thanks to its use at the intersection of several constructional schemas.