Marking and breaking phraseology in English and Polish: a comparative corpus-informed study

1 Introduction

One of the challenges faced by many sub-disciplines of linguistics, such as phraseology, lexicography, natural language processing or machine translation, is to devise better methods of identification, extraction and recording of phraseological expressions (henceforth PEs) found in texts. Generally, the criteria used to extract instances of linguistic prefabrication from texts are varied, as they include formal-linguistic, frequency-driven or distributional and psycholinguistic ones, which also correspond with various perspectives on formulaic language (Forsyth and Grabowski 2015; Gałkowski 2006; Nelson 2018; Pęzik 2018; Schmitt and Carter 2004; Trklja and Grabowski 2021; Wray 2002, 2008; Wood 2015). Formal-linguistic criteria (or, in other words, language system-oriented ones) are grounded in linguistic theory and they include phonological, morphological, syntactic, semantic and pragmatic criteria while frequency-driven criteria, grounded in language use and distribution of linguistic items (hence, text-oriented or language use-oriented criteria), include raw and normalized frequency of use, frequency distributions, measures of lexical attraction/repulsion or collocational strength etc. In studies dealing with processing, storage and retrieval of formulaic language, researchers use a wide range of text processing metrics, e.g., reaction times, eye-tracking methods etc., which are grounded in cognitive linguistics and psychology.^[1] Given the vast array of extraction criteria, Schmitt and Carter (2004: 3) argue that PEs should be treated in an inclusive way so that multiple types of linguistic units fall under a cover-all term formulaicity.

In practice, the quest for PEs involves the application of selected operationalized criteria to text fragments or to word combinations (contiguous or non-contiguous ones) that are candidates for PEs. However, it is also possible to undertake an attempt at identification of PEs using other linguistic items found in co-text or punctuation marks found in near proximity to instances of PEs, which can be treated as potential signals of prefabricated language. Such a method may be treated as a complementary one with respect to system-oriented, usage-oriented or cognitive-based criteria mentioned above. Thus, in this paper, we focus on the ways formulaic language is marked or signalled in texts.

According to Čermák (2005), Chlebda (2010), Rozumko (2011), Ruiz-Gurillo (2015), Rojo (2019), there are many orthographic and linguistic devices that can function as textual indicators of prefabricated expressions. More specifically, Chlebda (2010: 21–27) argues that textual operators signalling the use of prefabricated language in Polish can encompass punctuation marks (e.g., quotation marks introducing quotations or winged words) as well as conventionalized routine formulas, e.g., że tak powiem ‘so to speak’, by tak rzec ‘so to say’, jak to się mówi ‘as it were’, jak mawiają ‘as they say’, jak to się zwykło określać ‘as it used to be called’, jak to się przyjęło nazywać ‘as it used to be called’, tak zwany ‘so called’, przysłowiowy ‘the proverbial n’ studied by Pęzik (2018),^[2] jak mawiają w ‘as they say in’, jak mawia klasyk ‘in the words of’, jak powiedział znany [ktoś] ‘as somebody famous once said’, co określano mianem ‘as it used to be called’, jak mówi mądrość ludowa ‘popular wisdom has it that’, jak głosi znane porzekadło ‘as the (well-known) adage goes’, zgodnie ze starą maksymą ‘as an old saying goes’ etc.). Likewise, Čermák (2005: 58) claims that many idioms or phrases are signalled “by certain phrases or words” that may be referred to as introducers (e.g., so to speak in English or abych tak řekl in Czech). In a similar vein, Rozumko (2011: 316–317) explores so-called proverb introducers, also called metalinguistic tags, in English (e.g., according to the proverb, as the old saying goes, they say) and Polish (e.g., jak mówi przysłowie, zgodnie z przysłowiem, zgodnie ze starym powiedzeniem), which are “words and phrases used by speakers to signal that they are going to use a proverb or that they have just used it” (ibid.), while Rojo (2019) refers to linguistic items that are used to introduce phraseological units as metatextual indicators (e.g., as the saying goes). Such expressions are typically fixed, socially conditioned and their primary textual function is to introduce or demarcate various types of PEs in speech or writing, and help identify those narration fragments that constitute the common knowledge of a sender and addressee as well as of their discourse communities (Chlebda 2010: 21). It seems, however, that we do not have exhaustive knowledge as to how such textual indicators of PEs are used in discourse, notably with respect to discoursal functions other than marking instances of formulaic language.

Hence, in this cross-linguistic corpus-informed study, we undertake an attempt at investigating recurrent phrases whose main textual function is to signal or mark the use of prefabricated idiomatic expressions (idioms, proverbs, sayings, clichés, winged words etc.), and that is why we call such items ‘phraseology markers’ (Pęzik 2018), henceforth PMs. Furthermore, we look at the same expressions used to achieve the opposite objective of marking compositional wordings. We also assume that the patterns of use of PMs may provide new insights into the variability of PEs, which are often “transformed, compressed and partly edited, in order to be marked as prefabricated” (Pęzik 2018: 214). In other words, we believe that apart from marking canonical forms of PEs, PMs may be also used to mark novel, innovative uses of PEs where syntactic and semantic properties of multi-word expressions conflict with more general usage norms. From a different perspective, such conscious or subconscious uses where the fixed, canonical form of PEs is broken, and their new variants are employed in certain contexts may be construed as phraseological errors (Bąba 1989, 2009). However, Andrejewicz (2015: 48–49), who conducted a corpus-informed study of the use of selected PEs in Polish, argues that their use in texts, which very often deviates from the canonical forms recorded in dictionaries, should not be automatically considered as erroneous but rather as an alternative way of using PEs, underscoring such features of PEs as the lack of their absolute fixedness or their elastic stability (cf. Chlebda 2003: 181). This observation also aligns with the Theory of Norms and Exploitations (TNE) put forward by Hanks (2013) whereby there is a set of rules that govern normal and conventional language use and a second-order set of rules that govern exploitation of those usage norms and conventions resulting in incremental innovation and, eventually, in language change (ibid., 411).

Since phraseology and linguistic prefabrication is also grounded in culture, that is, they are treated by many researchers as repositories of cultural or ethnolinguistic knowledge (e.g., Sandomirska 2000; Bartmiński 2002; Wierzbicka 2015 ^[3]), it is possible to study the use of PMs cross-linguistically. Thus, we focus on English and Polish languages, as we believe that the findings of cross-linguistic studies of PMs may be useful for foreign or second language learners in their pursuit of fluent and idiomatic language use. Furthermore, the identification of prefabricated language through PMs may also supplement other methods (formal, distributional etc.) used to extract PEs and formulas from texts. Also, the linguistic material obtained thanks to the PMs may facilitate development of more comprehensive descriptions of phraseological profiles of individual speakers or discourse communities. Finally, we believe the PMs may also come in useful as a complementary method for measuring the degree of formulaicity or phraseological prefabrication in texts.^[4]

2 Phraseology markers and their functions

In the context of this study, the term ‘phraseology markers’ (PMs) can be related to broader concepts of ‘discourse markers’ (e.g., Schiffrin 1987/1996) or ‘pragmatic markers’ (e.g., Fraser 1996), which have been defined in various ways and explored from multiple perspectives (Lenk 1998; Ranger 2018). For example, Fraser (1999: 931–932) argues that discourse markers include “discourse connectives, discourse operators, pragmatic connectives, sentence connectives, and cue phrases”, a large heterogeneous functional class of linguistic items responsible for the flow of discourse (in terms of coherence, cohesion etc.). In other words, discourse markers are “linguistic expressions of varying length which carry pragmatic and/or propositional meaning, occur in both speech and writing, and facilitate rather than disrupt discourse” (Siepmann 2005: 43). Some researchers investigate discourse markers (also called ‘discourse particles’ or ‘discourse-linking devices’) primarily but not exclusively in spoken language and focus on routine speech formulas performing conversational management functions (rather than contributing to propositional content of utterances), e.g., helping express speaker’s stance and shape the relationships between the speaker and listener, among others (Lenk 1998: 37; Stede and Schmitz 2000: 125). In that respect, what we call in this paper as PMs are in fact linguistic expressions that are attested primarily in spoken language and that facilitate management of conversations and dialogue flow, notably by explicitly and purposefully signalling the use of PEs.

Although phraseological competence, measured by knowledge and use of PEs, varies between language speakers, there is also cumulative shared knowledge of a given discourse community regarding how to use language in particular communicative situations. In other words, every use of a PE arises from individual experience of members of discourse communities as well as from the common knowledge base of the discourse community, an idea which is also reflected in the lexical priming theory (Hoey 2005, 2007). To put it simply, language speakers use PEs to a varying degree yet when they do so, they also appeal to the common knowledge base of their interlocutors in the hope that they also share a basic intuition about idiomaticity and prefabrication in language use. According to Pęzik (2018: 214), there are situations, however, when language speakers purposefully mark the use of prefabricated or formulaic expressions, which may be due to, for example, their lack of confidence about the conventionality of a PE,^[5] a peculiar stylistic use of language or a specific pragmatic effect (an appeal to conventional wisdom, irony, humour, emphasis, recollection from memory, filling a pause during a conversation etc.). The linguistic items which are recalled from collective memory of language users and whose function is to explicitly mark the use of PEs (encapsulating the whole variety of prefabricated phrases, routine formulas, proverbs, sayings, clichés etc.) in speech or writing are referred to in this paper as PMs.

Those PMs which happen to be multiword units can be positioned within many typologies of phraseological items developed in Polish, Russian or British research tradition. For example, PMs can be assigned to a peculiar subclass of speech formulas called commentaries (Baranov and Dobrovolskij 2008), which are used to comment on various aspects of communicative situations, e.g., дай бог памяти ‘God give (me) memory’ (Polish: jeśli mnie pamięć nie myli/zawodzi, English: if my memory serves me well/right/correctly). Furthermore, a custom-designed class of such items may be found within a broad concept of ‘reproducts’ (Chlebda 2010), where the textual operators signalling the use of prefabricated expressions are called phraseological indicators ‘wskaźniki frazeologii’ (p. 21). As such, PMs can be also considered as a peculiar subcategory of pragmatemes or routine formulas, which comprise “conventional expressions used in recurrent situations” primarily in spoken language (Szerszunowicz 2020: 173), and which “tend to be neglected in phraseological analyses” (ibid.). According to Mel’cuk and Milicevic (2020: 112), pragmatemes are in fact clichés with fixed meaning and fixed form, which are restricted in terms of their use by a speech act situation (e.g., Hold the line, please; Sincerely yours). As for PMs, their form is also fixed (e.g., so to speak, as it were), yet their use is not restricted to a particular extralinguistic situation (e.g., answering a telephone call). On the contrary, PMs may be used at any point during a conversation and their pragmatic meanings may also vary from context to context, e.g., appealing to conventional wisdom, signalling irony, softening a message, emphasizing metaphorical language use in a particular communicative situation.

The motivation behind exploring PMs in greater detail is manifold. Apart from providing insights into language users’ intuitions regarding the use of prefabricated expressions, the exploration of PMs may help us understand the ways prefabricated language is marked in texts. For example, we may attempt to determine whether and to what extent quotations or winged words are typically accompanied by PMs. We may also endeavour to find out whether there are any other types of PEs (idioms, routine formulas, sayings etc.) signalled by individual PMs. Is it the case that PMs are only used to mark formulaic and prefabricated expressions, or do they perform any other textual functions? These are only a few problems that remain largely underexplored with respect to the use and discoursal functions of PMs.

3 Methodology

Broadly speaking, we adopt a corpus-based perspective in this study focusing on the exploration of the use of PMs. More precisely, we preselect a set of textual operators described as PMs in specialised literature (Čermák 2005; Chlebda 2010; Rojo 2019; Rozumko 2011; Ruiz-Gurillo 2015; Pęzik 2018) and, by consulting large language corpora of English and Polish, we attempt to, first, describe frequency distributions and pervasiveness of PMs in texts, and second, to verify what types of PEs (fixed expressions, idioms, proverbs, sayings, quotations, winged words etc.) are introduced in texts by the PMs under scrutiny. It is hoped that the present study will enable us to identify and describe patterns or tendencies of PMs’ use as labels foregrounding various types of PEs across modalities (spoken and written). Particularly, we will focus on spoken language as we believe that PMs are found in transcripts of conversations in a broader spectrum of functions than in written texts. This is simply because spoken language use happens in real time and it is therefore subject to temporal constraints, unlike the production of most genres of written texts, which can be ‘edited out’ as redundant. Similarly, Rojo (2019) argues that knowledge of how metatextual indicators (i.e. PMs) are employed to signal PEs may be of utmost importance when immediacy is a crucial factor, which is the case of interpreting or conversational language use (or spoken language in general). In what follows, we describe in greater detail the units of analysis, research material (study corpora), research questions and hypotheses as well as the tools and study stages.

4 Results

In this section, we will present the results of quantitative and qualitative analysis of the PMs under scrutiny, followed by a case study related to the use of the English PM ‘so to speak’.

5 Discussion and conclusions

In this paper, we attempted, first, to describe the use of semantically and pragmatically similar, according to dictionary definitions, recurrent phrases labelled under an umbrella term ‘phraseology markers’ (PMs), and second, to measure the PMs’ capacity to demarcate prefabricated idiomatic expressions in texts.

The study findings revealed that apart from marking the use of metaphorical or unusual phrasings in particular situations of language use, the investigated PMs perform important roles with respect to marking prefabricated expressions in spoken language, be it English and Polish. More precisely, we found that the PMs under scrutiny perform two opposing functions with respect to the use of formulaic language. On the one hand, they are used to mark conventional, prefabricated expressions that contribute to formulaicity of speaker’s utterances. On the other hand, we identified contexts where the same PMs are used to mark expressions or phrases that represent unusual, unconventional, idiosyncratic phrasings unattested or rarely used in native texts. We believe that some of those phrases represent so-called phraseological innovations^[15] (Bąba 1989) as they do not conform with the established – at a given point in time – phraseological norm (ibid.). Based on our findings, we put forward a hypothesis that breaking the fixed or stereotypical form of formulaic language and/or conventions of its use may be facilitated by certain phrases that in this paper (and in specialized literature, cf. Section 1 and 2) are referred to as ‘phraseology markers’. Hence, paradoxically, the phrasings usually referred to as phraseology markers occasionally perform the function of ‘phraseology breakers’, where the fixed or canonical form of a phraseological unit is modified yet it is still possible for speakers or writers to intuitively trace back to its very source, i.e. the canonical form. In addition, those phrasings may also be used to signal novel and idiosyncratic, unconventional word combinations, rarely or never attested in language corpora, and hence, we may also refer to them as ‘novelty markers‘. In short, the findings showed that phrasings, such as so to say, so to speak, as it were etc., are pragmatically multifunctional as in different contexts of use they may perform the roles of ‘phraseology markers’ (i.e. marking phraseological prefabrication or idiomaticity), ‘phraseology breakers’ (i.e. marking phraseological innovations or variability) and ‘novelty markers’ (i.e. marking highly-individualized, idiosyncratic phrasings). This concise typology of expressions used to explicitly mark linguistic units as either formulaic or compositional is outlined in Figure 1, where phraseology breakers are shown as a subtype of phraseology markers.

Figure 1:

A conceptual outline of devices used to explicitly mark formulaic and compositional expressions.

Based on the results, we also hypothesize that ‘breaking phraseology’ or’ breaking formulaicity’, as facilitated by the PMs, should not be treated as an instance of habitual and unconscious language use; neither should it be treated as an optional, conscious and regular language use that can account for speaker’s or writer’s idiolectal style. Conversely, we put forward the claim that ‘breaking formulaicity’, through phrases such as so to say, so to speak, as it were etc., is an optional and conscious choice on the part of language speakers yet, at the same time, it represents irregular and idiosyncratic language use (i.e. the means of signalling one-off occurrences of linguistic novelty or idiosyncrasy). With respect to such idiosyncratic language use, Taylor (2014: 278–279) argues that instead of drawing a binary distinction between regular and irregular language use, we should consider it as a cline between creativity and innovative extensions, the latter standing for those cases where syntactic and semantic properties of a linguistic expression, including prefabricated phrases, conflict with more general usage norms. Thus, we assume that such innovative extensions may also encompass situations when formulaic language is used in atypical ways, e.g., when its fixed form is broken, and new variants of phraseological expressions (with novel internal variation) are used, and that such atypical forms can be signalled explicitly using phraseology breakers. In a similar vein, Miller (2014: 19) argues that creativity is the ability to “break formulaicity and re-form patterns” of language use, and that is precisely what we deal with when we inspect how phraseology markers/breakers are used in texts. More precisely, the phrases such as że tak powiem and so to speak have the capacity to mark the use of novel and idiosyncratic word combinations, which include one-off occurrences specific to situations of language use, often hardly attested in reference corpora and not recorded as lexical entries in dictionaries of contemporary English and Polish. Such lexical items, however, may provide valuable data for upgrading lexicographic resources with novel phraseological innovations.

Next, taking the English phrase so to speak as a case in point, we focused on its pragmatic function of phraseology marker and attempted to measure how much prefabricated language that English phrase marks when used in the Spoken component of the COCA 2020 corpus. It was found to be in the range of 14–25 % of its total number of occurrences. On the one hand, we can argue that phraseology markers such as so to speak can, in general, be treated as complementary methods of identification of idiomatic expressions in texts, and that their potential in that respect is still underexplored. As a matter of fact, one of the important discourse functions of such items, which is not explicated in dictionaries of contemporary English and Polish, is to mark instances of prefabricated language, which can be used with various purposes in mind (for emphasis, irony, unusual stylistic effect etc.) across various contexts of use. This finding may be particularly useful for foreign or second language learners as it may help them improve fluency and naturalness of their spoken language. On the other hand, the finding that the phrase so to speak marks prefabricated language only in the range of 14–25 % of its occurrences in the COCA 2020 Spoken corpus implies that its function of a phraseology marker is not a central one and that in the range of 75–86 % of its occurrences it may perform the function of phraseology breaker or novelty marker, among other more fine-grained discoursal functions. This claim, however, should be further verified in future studies.

This study has its limitations which primarily pertain to methodology. First, we investigated only a small subjective selection of phraseology markers and compared their use and discourse functions in various yet not fully comparable corpora of English and Polish, including general-language and spoken corpora. Second, an attempt at measuring the amount of prefabricated language signalled by phraseology markers was made taking only one item, the English phrase so to speak, as a case in point, which leaves us with the question whether our observations should be directly transferred to Polish phraseology markers. It therefore seems that future studies should be conducted using more balanced and comparable corpora (e.g., register-specific) where more units of analysis could be explored in greater detail in order to arrive at more precise estimates and more comprehensive descriptive findings. In a similar vein, it may be worthwhile exploring in greater detail the extent to which phraseology markers, similar to the ones studied in this paper, signal prefabricated language across different languages, also across various text types, genres and modalities (spoken and written). Also, it might be interesting to study sentence-positions of phraseology markers across languages as they may precede, occur inside or follow phraseologisms or phraseological innovations. Finally, more sophisticated statistical methods (e.g., regression models) may be employed to determine contextual conditions impacting the ways speakers or writers typically use phraseology markers. We hope, however, that our study is a good starting point for such future research.