2.1 Intensifiers and language change

The intensifier system has been said to constitute one of the most dynamic and rapidly changing areas of a language (Brinton and Arnovick, Reference Brinton and Arnovick2006). Frequent innovation in the intensifier paradigm has been attributed to speakers’ search for creative means of expression and emphasis (Lorenz, Reference Lorenz, Wischer and Diewald2002: 146). For example, Fagyal et al. (Reference Fagyal, Kibbee and Jenkins2006: 283) describe how the use of novel intensifiers such as furieusement (‘furiously’) was common in the seventeenth-century court system, and particularly among women. Intensifiers have also been observed to undergo cycles of rapid renewal and recycling, as older intensifiers “do not fade away [but rather] stick around for a very long time” (Ito and Tagliamonte, Reference Ito and Tagliamonte2003: 277). An example of such dynamic change is found in D’hondt and Defour’s (Reference D’hondt and Defour2012) diachronic description of vraiment. Using historical corpus data from the Dictionnaire du Moyen Français (DMF) and the Frantext corpora (Table 1), these authors show vraiment to have first been attested as an intensifier in the 16^th century, only to increase in this capacity in the two centuries following and then decrease in this function thereafter. This relatively rapid pattern of fluctuation in the frequency of use of vraiment typifies the trajectory of intensifiers over time.

Table 1.

Diachronic use of vraiment as an intensifier vs. other lexical capacties, e.g. as an adjunct, emphasizer, disjunct (adapted from D’hondt and Defour, Reference D’hondt and Defour2012:177)

That intensifiers develop and change relatively rapidly compared to other parts of the lexicon means that they offer valuable insights into the mechanisms of language change (Ito and Tagliamonte, Reference Ito and Tagliamonte2003). As intensifiers constitute an open grammatical class, new intensifiers may enter into a language at any given time (Partington, Reference Partington, Baker, Francis and Bonelli1993); this occurs via processes of delexicalization, by which words lose their original semantic value and acquire novel lexical meaning (Lorenz, Reference Lorenz, Wischer and Diewald2002). For example, vraiment derives from Old French veraiement by way of the Latin adjective veracus, meaning ‘true, honest’, but has since become delexicalized to include uses irrelevant to truth value (3).

More delexicalized intensifiers tend to be used more frequently and, in turn, the more frequent an intensifier is, the less expressively marked it is (Lorenz, Reference Lorenz, Wischer and Diewald2002: 144–145). It has been argued that the extent to which an intensifier has undergone delexicalization can be observed through analysis of its grammatical patterning. For instance, delexicalization of an intensifier has been claimed to function in relation to the frequency with which it modifies attributive vs. predicative adjectives (Mustanoja, Reference Mustanoja1960: 326–327). Whereas attributive adjectives appear adjacent to the noun they modify (4), predicative adjectives are separated from the modified noun by a verb (5).

Nearly all studies considering attributive vs. predicative position of intensifiers have shown a general preference for intensifiers to appear in predicative vs. attributive contexts (e.g. Ito and Tagliamonte, Reference Ito and Tagliamonte2003; Macaulay, Reference Macaulay2006; Tagliamonte, Reference Tagliamonte2008; Barnfield and Buchstaller, Reference Barnfield and Buchstaller2010; Romero, Reference Romero2012; Tagliamonte, Reference Tagliamonte2016). Nonetheless, this trend has been shown to vary for different intensifiers (e.g. Alshaboul et al., Reference Alshaboul, Jarrah, Alghazo and Al-Shawashreh2022; Stratton and Sundquist, Reference Stratton and Sundquist2022). Intensifiers have also been found to change over time in terms of their preference for modifying predicative vs. attributive adjectives. For instance, Barnfield and Buchstaller (Reference Barnfield and Buchstaller2010) found that newer intensifiers in British English tend to modify predicative adjectives and then spread to attributive positions as they increase in frequency over time, whereas older intensifiers appear relatively more often in attributive contexts.

Past studies have also analyzed intensifiers’ collocational width as an indicator of its degree of delexicalization (e.g. Tagliamonte, Reference Tagliamonte2008). Generally speaking, more delexicalized intensifiers tend to collocate with a wider array of adjectives compared to intensifiers less far along in the delexicalization process (Partington, Reference Partington, Baker, Francis and Bonelli1993: 183). Collocational width is also thought to relate to how recent and/or dynamic an intensifier is, with Tagliamonte (Reference Tagliamonte2008: 376) arguing that “incoming intensifiers can be expected to collocate with a small set of specific lexical items, older ones can be expected to appear widely across a broad range of adjective types and those that are falling away retreat to particular collocations or restricted registers of the language”. Given the extent to which adjectival position and collocational width have been studied in past research as grammatical correlates of intensifier change, these two aspects are likewise examined here as they relate to French intensifiers.

Changes in intensifier use are often studied through apparent time analyses (e.g. Tagliamonte, Reference Tagliamonte2008; Stratton, Reference Stratton2020), which assume that conclusions about diachronic language change in a community can be drawn by comparing linguistic forms used by speakers of different generations at a single point in time (Bailey et al., Reference Bailey, Wikle, Tillery and Sand1991). Such analyses presuppose that language use is largely fixed in adulthood and, thus, that the speech of older speakers is reflective of older linguistic forms and the speech of younger speakers of more recent forms. Furthermore, this view argues that, by comparing older and younger speakers, one can deduce the nature and direction of changes in a language. In contrast, real-time analyses rely on comparisons of data collected from the same speakers or speaker community at two different time periods to draw conclusions about language change (Tillery and Bailey, Reference Tillery and Bailey2003). Two past intensifier studies have incorporated both of these approaches: i) Barnfield and Buchstaller (Reference Barnfield and Buchstaller2010), who look at intensifier use in Tyneside, England, comparing data collected in the 1960s, from 1994–1995, and from 2007–2008, and ii) Fuchs (Reference Fuchs2017), who considers intensifiers in data from Spoken British National Corpus collected in 1994 and in 2014. The present study also incorporates both apparent- and real-time analyses to better understand how, in the case of intensifier use, individual linguistic change over the lifespan may occur alongside larger, generational changes in a community (see Sankoff, Reference Sankoff2019).

2.2 Sociolinguistic factors

Intensifier use has also been shown to co-vary with a number of sociolinguistic factors, among which the most frequently studied are speaker gender and age. Gender has long been evoked as a correlate of intensifier use, with female speakers being claimed to use more intensifiers than male speakers (e.g. Stofel, Reference Stoffel1901; Jesperson, Reference Jespersen1922). Support for this claim has been found in studies on intensifiers in English (e.g. Bradac et al., Reference Bradac, Mulac and Thompson1995; Xiao and Tao, Reference Xiao and Tao2007; Fuchs, Reference Fuchs2017) and other languages such as German (Stratton, Reference Stratton2020) and Norwegian (Stratton and Sundquist, Reference Stratton and Sundquist2022). Female speakers have also long been claimed to lead in the use of fashionable new intensifiers (e.g. Jespersen, Reference Jespersen1922), as previously mentioned to be the case with furieusement in the 17th-century French court system. Nonetheless, this finding is not ubiquitous, as some studies have failed to find a large difference in intensification rates between female and male speakers (e.g. Ito and Tagliamonte, Reference Ito and Tagliamonte2003). Some research has also shown male and female speakers to show differing preferences for individual intensifiers. For instance, Precht (Reference Precht2008) finds that, while the English intensifier so is used more by female speakers, totally shows the opposite pattern in being used more by male speakers. Additionally, this observation extends beyond English-language studies, as gendered preferences for different intensifiers have been reported in Korean (Kwon, Reference Kwon2012) and Ammani Arabic (Alshaboul et al., Reference Alshaboul, Jarrah, Alghazo and Al-Shawashreh2022).

Regarding age, apparent-time analyses have generally shown younger speakers to use higher rates of intensifiers than older speakers (e.g. Stenström et al., Reference Stenström, Hasund and Andersen2002; Ito and Tagliamonte, Reference Ito and Tagliamonte2003; Macaulay, Reference Macaulay2006; Roels and Enghels, Reference Roels and Enghels2020; Stratton, Reference Stratton2020; Stratton and Sundquist, Reference Stratton and Sundquist2022). This trend is due in part to the fact that intensifiers, and particularly newer ones, can serve as shibboleths for indexing group membership and identity of speakers (Lorenz, Reference Lorenz, Wischer and Diewald2002: 143). This has been said to be the case with the French intensifier grave, which is used most often by French-speaking teenagers and young adults in France, Switzerland, and Belgium (Zribi-Hertz, Reference Zribi-Hertz, Dostie and Hadermann2015). Like for gender, speakers of different ages have also been shown to hold varying preferences for different intensifiers (e.g. Tagliamonte, Reference Tagliamonte2008). Some studies have also shown gender and age to be correlated as they relate to intensifier use. For instance, Fuchs (Reference Fuchs2017) finds that younger speakers show larger gender-based differences in intensifier use as compared to older speakers. The fact that gender and age have been shown by past research to be important social factors in studying intensifier variation and change motivates their consideration in the present study.

2.3 Past variationist research on French intensifiers

Although much research has looked at the general semantic properties of intensifiers in French (e.g. Lenepveu, Reference Lenepveu2007; Gaatone, Reference Gaatone2013; Kleiber, Reference Kleiber2013; Labelle, Reference Labelle2022), quantitative data-driven analyses of French intensifiers are scarce. Nonetheless, a few studies have empirically documented the sociolinguistic properties of specific French intensifiers. For example, as previously mentioned, Zribi-Hertz (Reference Zribi-Hertz, Dostie and Hadermann2015) describes the semantic and syntactic properties of the intensifier grave, using data collected from internet sources to explore this intensifier among young European French speakers. Another example comes from, Corminboeuf and Avanzi (Reference Corminboeuf and Avanzi2020), who investigate the regional intensifier monstre, drawing on large-scale, crowdsourced data to situate the use of this intensifier in French-speaking Switzerland and neighboring regions of France, particularly by young speakers.

Other studies have used quantitative methods to compare and contrast the patterning of a range of French intensifiers. The most comprehensive study of this kind is Dostie (Reference Dostie2018), who details the usage-based properties of twelve intensifiers in Quebec French, using examples taken from the contemporary Corpus de français parlé au Québec. In another relevant study from Krištofíková (Reference Krištofíková2012), sixty French speakers from Lyon (comprising young adult, middle-aged and elderly speakers) were asked to report how often they used different intensifiers. Among the fourteen intensifiers considered, seven were found to be used most by the young adult speakers (trop, grave, carrément, cher, méchant, super, mega), whereas five (vachement, très, sacrément, absolument, fort) were found to be used most by the elderly speakers. Participants in the middle-aged group reported lower use than both of the other two groups for méchant, fort, cher, absolument, and grave, and higher overall use than the other two groups for the intensifier hyper. These results are informative in revealing an effect of age-grading in the use of different intensifiers in French. However, because speakers’ self-reports of language use are known to deviate from their spontaneous production patterns (e.g. Trudgill, Reference Trudgill1974), this study is limited in its reliance on self-reported data as a window into French intensifier use.

Several other recent studies have analyzed how variation in the French intensifier system patterns and changes, albeit by drawing from written data sources. For instance, Cartier and Huyghe (Reference Cartier and Huyghe2021) examine change in the use of intensifiers super, hyper, extra, ultra, archi, méga, and supra from 1800 to 2010 using the French Google Books corpus, finding a notable rise in the use of these intensifiers in writing between 1900 and 1970 (see also Kamber and Huyghe, Reference Kamber and Huyghe2023, who analyze intensifiers méga, giga, and hypra in a written corpus of online French collected in 2017). A similar study comes from Hendrikx et al. (Reference Hendrikx, Van Goethem, Meunier and Hiligsmann2017), who use the Frantext coprus to quantify the use of a variety of intensifiers in texts written between 1980 and 1992. The most frequent intensifiers found by these authors are très (accounting for 41.1% of sampled intensifiers), tout Footnote ³ (18.4%), si (12.2%), purement (2.5%), and parfaitement (2%). Nonetheless, these authors find a relatively low intensification rate (i.e. the percentage of adjectival heads that are intensified vs. those that are not) of only 3.9%, which is unsurprising given that intensification is primarily found in spoken language (e.g. Paradis, Reference Paradis1997; D’Arcy, Reference D’Arcy2015). More generally, written and spoken French are typically understood as occupying separate linguistic registers, each characterized by a distinct set of linguistic norms and tendencies (see Batchelor and Offord, Reference Batchelor and Offord2000; Bourns, Reference Bourns2017), thus making comparison of data collected across these registers difficult.

2.4 Present study

From this literature review, it is apparent that past research on French intensifiers has yet to comprehensively document the quantitative patterning of intensifiers in spontaneous, spoken French. Without such research, a fuller understanding of how French intensifiers vary and change, let alone how this variation compares to that found in other languages, cannot be gained. This study therefore addresses this gap by analyzing intensifiers as they occur in a corpus of spoken Hexagonal French (i.e. the French of continental France) across two timepoints over the past half-century. In doing so, this study also contributes to a recent body of research serving to expand the variationist study of intensifiers beyond an Anglocentric purview (e.g. Kwon, Reference Kwon2012; Roels and Enghels, Reference Roels and Enghels2020; Stratton, Reference Stratton2020; Alshaboul et al., Reference Alshaboul, Jarrah, Alghazo and Al-Shawashreh2022; Stratton and Sundquist, Reference Stratton and Sundquist2022). This investigation is underpinned by three general questions:

1. 1. How do intensifiers pattern quantitatively in spontaneous, spoken Hexagonal French?

2. 2. How has the intensifier system in spoken Hexagonal French changed grammatically over the past half-century?

3. 3. What role does speaker gender and age play in this change?

The first question is operationalized through an analysis of intensification rates and proportional distributions of variable intensification use between 1970 and 2010. The second question is explored through an investigation of adjectival function and collocational width as grammatical correlates of intensifier change. Lastly, the third question is examined through an analysis of speaker gender and age as social correlates of overall intensification rate and individual intensifier use across time. Observations from the following confirmatory analyses are expected to align largely with cross-linguistic trends observed in a majority of past variationist intensifier studies. Namely, intensifiers are expected i) to show significant quantitative differences in use over time, ii) to show evidence of grammatical change alongside changes in frequency of use, and iii) to be used most innovatively by younger and female speakers. The following section outlines the methods used for testing these predictions.

3.1 Corpus

Data were taken from the ESLO, a corpus of spoken French collected from habitants of the city of Orléans (about 75 miles south of Paris). The ESLO comprises recorded speech data reccorded four decades apart: the initial ESLO corpus (the ‘ESLO1’) was collected from 1968 to 1974 and contains 274 hours of transcribed speech. Work on the second corpus (the ‘ESLO2’) began in 2008 and currently consists of roughly 432 hours of transcribed speech. The ESLO corpus was selected for use in this study for several reasons. First, it is one of the largest collections of transcribed, spoken French currently available to the public, containing more than 10,000,000 words of spoken French (Baude and Dugua, Reference Baude and Dugua2016), and thus lending itself to relatively large-scale linguistic analyses. Second, the ESLO is exceptionally diverse in its representation of speakers (Blanc and Briggs, Reference Blanc and Biggs1971: 23), allowing for analysis of socially stratified speaker samples. Third, and most importantly for this study, the ESLO is unique among other spoken French corpora in its longitudinal scope (see Fonseca-Greber, Reference Fonseca-Greber and Fonseca-Greber2023: 519–528 for a review of spoken French corpora).

While some recent studies (e.g. Stratton, Reference Stratton2020; Stratton and Sundquist, Reference Stratton and Sundquist2022) have used random sampling of adjectives to study the quantitative distribution of intensifiers in speech corpora, the present study follows more traditional methods (from e.g. Ito and Tagliamonte, Reference Ito and Tagliamonte2003; Tagliamonte, Reference Tagliamonte2008; Barnfield and Buchstaller, Reference Barnfield and Buchstaller2010) by considering all relevant data from a pre-determined sample of speakers. Although this method yields imbalanced representation of tokens for each speaker in the sample, it allows for more in-depth analysis of individual intensifier use. Hence, a sample of forty-eight speakers from the ESLO corpus was selected for analysis, with a balanced representation of female and male speakers across age categories (Table 2). Data were drawn from semi-directed, face-to-face sociolinguistic interviews ranging from 45 to 128 minutes in length and conducted from 1969-1970 (for the ESLO1) and in 2010 (for the ESLO2), with all tokens of possible adjectival intensification coded for each speaker (see the following section). While speakers were selected based on the criterion that they contributed a relatively large amount of interview data to the ESLO corpus, some older speakers had less transcribed data than other speakers but were included nonetheless to allow for a socially balanced sample. Additionally, to allow for balanced representation of age and gender across samples, two researchers were included (speakers ‘ch_OB1’ and ‘ch_CD2’) in the ESLO2 sample. While this potentially risks introducing complicating effects of speech style on intensifier use, these two speakers were found to have intensification rates within the range of other speakers in the ESLO1 and ESLO2 (9% and 18.3%, respectively; see Section 4.1) and were thus retained in the sample.Footnote ⁴

Table 2.

Balanced sample of speakers from the ESLO corpus

Speakers born across a ninety-year period were represented in the study, with ESLO1 speakers born between 1899 and 1950 (mean = 1943) and ESLO2 speakers born between 1930 and 1989 (mean = 1963). Three to five speakers were chosen to represent each decade from the 1900s to the 1980s, determined by their respective birthyears. Exceptionally, temporal overlap in when the corpora were collected meant that eleven speakers born in the 1940s (five from the ESLO1 and six from the ESLO2) were included in the sample. This allowed for examination of how speakers with similar birthyears may show different intensifier use in the ESLO1 vs. the ESLO2, thus potentially reflecting individual linguistic change across the lifespan (see Sankoff, Reference Sankoff2019).

3.2 Data coding

Intensifiers are typically grouped into two categories consisting of ‘amplifiers,’ which “scale upwards from an assumed norm,” and ‘downtoners,’ which “scale downwards from an assumed norm” (Quirk et al., Reference Quirk, Greenbaum, Leech and Svartvik1985: 455). Although downtoners (e.g. c’est assez grave ‘It’s pretty serious’) are sometimes included in analyses of intensifiers (e.g. Stratton, Reference Stratton2020; Stratton and Sundquist, Reference Stratton and Sundquist2022), the present study only considers amplifiers, as they occur more often (D’Arcy, Reference D’Arcy2015) and have been argued to be more linguistically complex than downtoners (Ito and Tagliamonte, Reference Ito and Tagliamonte2003: 258). These claims are supported by the fact that only three downtoners (i.e. assez, plutôt, and légèrement) occur in the ESLO sample, as compared to the twenty-five unique amplifiers found in the same sample (see Dostie, Reference Dostie2018: 185–187 for further description of how downtoners such as assez pattern differently than French amplifiers). Nonetheless, for the purposes of comparability, downtoners are included in analyses relating to overall intensification rate (section 4.1).

Moreover, this study, like most past quantitative studies on intensifiers (e.g. Lorenz, Reference Lorenz, Wischer and Diewald2002; Ito and Tagliamonte, Reference Ito and Tagliamonte2003; Barbieri, Reference Barbieri2008; Fuchs, Reference Fuchs2017; Stratton, Reference Stratton2020), limits its purview to intensifiers in adjectival heads, as more than 70% of intensifiers have been found to occur in this context (see Bäcklund, Reference Bäcklund1973: 269; Rickford et al., Reference Rickford, Wasow, Zwicky and Buchstaller2007: 8 for studies on English intensifiers). To allow for observations about intensification rates across speakers, as well as to increase comparability with previous studies, both intensified and potentially intensifiable (yet non-intensified) adjectives were collected, thus satisfying the Principle of Accountability (Labov, Reference Labov1973). To facilitate this process, transcripts of participants’ interviews were downloaded from the online ESLO website and part-of-speech tagged using the software TagAnt (Anthony, Reference Anthony2022). Words with the part-of-speech tag “_ADJ” were then checked manually to verify that they fit the criteria for being intensifiable. Additionally, many adjectives that were known to be intensifiable were also hand-checked in case they had been erroneously tagged as something other than an adjective (e.g. bien ‘good’ used in an adjectival function). Contexts in which adjectives were deemed non-intensifiable were determined based on criteria of past studies (e.g. Ito and Tagliamonte Reference Ito and Tagliamonte2003); these included adjectives in fixed expressions (6), those in superlative/comparative structures (7), those deemed non-scalable (8), those in contexts in which they functioned grammatically as adverbs (9), those already modified by another adverb grammatically blocking a following intensifier (10), and those not fully realized due to false starts (11):

In terms of intensifiers, those in negative constructions (12) were excluded as it has been argued (by e.g. Ito & Tagliamonte Reference Ito and Tagliamonte2003 ), that they serve an inherently different function (closer to that of downtoners) compared to those with positive polarity (13).

As noted above, intensifiers modifying words other than adjectives or serving grammatical functions other than intensification were excluded (14–15).

Making this latter judgment proved more difficult for some intensifiers than for others. For instance, since the word trop can express either a sense of quantitively excess (16) or of intensity (17) (see Dostie, Reference Dostie2018: 36), only tokens of trop that unambiguously served the latter function were included in analyses.

In several cases, adjectives were deemed intensifiable (19) in contexts in which they would otherwise be non-scalable (20).

The reduplication of intensifiers can also be used in French for stylistic purposes, and is particularly characteristic of casual speech (e.g. Kamber and Huyghe, Reference Kamber and Huyghe2023; Torreira et al., Reference Torreira, Adda-Decker and Ernestus2010). In instances of reduplication in the ESLO sample (21), each intensifier was counted as an individual token. The percentage of reduplicated intensifiers was nearly identical across the samples, with 15.7% (n = 126) of all intensifiers being reduplicated in the ESLO1 compared to 15.9% (n = 121) in the ESLO2.

Lastly, exclamative constructions serving a function of intensification in French (see e.g. Marandin, Reference Marandin2010) were not considering in this study, as their pre-posed syntactic nature made their comparison to adverbial intensifiers challenging (22).

After accounting for these criteria, a total of 8,109 adjectival heads (or 8,463 including downtoners) were coded as intensifiable and subject to further analyses.

3.3 Statistical analysis

To test statistical significance of results relating to intensification rate, a generalized logistic mixed-effects regression model was fit to the data using the lme4 package (Bates et al., Reference Bates, Mächler, Bolker and Walker2017) in R (R Core Team, 2023). This model included as a binary dependent variable the presence vs. absence of an intensifier in each adjectival head (see Fuchs, Reference Fuchs2017; Stratton, Reference Stratton2020), with corpus (ESLO1 vs. ESLO2), speaker gender (female vs. male), and speaker age (centered around zero) tested as fixed factors, and interactions tested between all fixed factors. Speaker and adjective were included as random intercepts to account for variation in intensifier use across individual speakers and lexical items in the sample. Corpus was included as a random slope within speaker and adjective to account for variation in the effect of corpus across these grouping factors.

To test change within the intensifier system, a second generalized logistic mixed-effects regression model was run on only the intensified adjectives in the sample. In this model, the use of très vs. all other intensifiers was tested as a binary dependent variable. This decision was motivated by the fact that très accounts for the vast majority of intensifiers observed in the ESLO (as will be shown in Section 4.1.1), thus making running statistical tests on these intensifiers as individual factors infeasible. Additionally, testing très vs. all other intensifiers allowed for statistical analyses to be run on the overall change in intensifier use over time while still ensuring adequate statistical power. This model also included adjectival function (attributive vs. predicative), corpus, speaker gender, and speaker age as fixed factors, with interactions tested between all these factors, and random intercepts included for speaker and adjective. As in the first model, corpus was initially tested as a random slope within speaker and adjective, but dropped when the model failed to converge. For both logistics regressions, competing models were compared using ANOVAs, with the model of best fit selected according to reduction of the AIC value (Burnham and Anderson, Reference Burnham and Anderson2004). Significance of correlations between variables was tested using Spearman’s rank-correlation (to test ordinal variables in non-linear relationships), via the stat_cor() function in the ggpubr R package (Kassambara, Reference Kassambara2023).

An inherent limitation of studies on spoken corpora is corpus size, an important consideration when interpreting the statistical and descriptive results to follow. Full results and specifications of both mixed-effects regression models can be found in Appendix A. All linguistic data and analysis scripts from this study are available as online supplementary materials at https://osf.io/tkc4r/?view_only=1d6e6f1ab99e40b8b01a4b5381957070.

4.1. Distributional results

Of the sampled intensifiers, 1,921 (or 22.7% including both amplifiers and downtoners) were in fact intensified. Including only amplifiers, 1,567 (or 19.3%) adjectives were found to be intensified. These intensification rates are both lower than those found in past studies on other languages (Figure 1); assuming comparability in the coding schemes across these studies, this finding suggests that intensification is less common in French than in languages such as English, German, and Norwegian. Individual speakers’ intensification rates in the ESLO sample ranged from 8.8% to 39.1% (SD = 6.8%).

Figure 1.

Comparison of intensification rates across intensifier studies (Ito and and Tagliamonte, Reference Ito and Tagliamonte2003; Tagliamonte and Roberts, Reference Tagliamonte and Roberts2005; Tagliamonte, Reference Tagliamonte2008; Stratton, Reference Stratton2020; Stratton and Sundquist, Reference Stratton and Sundquist2022). On the ‘Hexagonal French’ bar, the pink portion shows the intensification rate for only amplifiers, while the pink and red portion together show the intensification rate for both amplifiers and downtoners.

Table 3 shows the rank order (from most to least frequent) of different intensifiers in the ESLO.

Table 3.

Distribution of all amplifiers in ESLO1 and ESLO 2 samples by overall frequency

These results reveal a Zipfian distribution of intensifier types in the ESLO, with the high-frequency intensifier très accounting for more than three-fifths (64.3.%) of all intensifiers in the sample and occurring nearly ten times as often as the second most frequent intensifier vraiment. The preponderance of très, as well as the fact that it holds majority usage in both written and spoken French (cf. Hendrikx et al., Reference Hendrikx, Van Goethem, Meunier and Hiligsmann2017), corroborates the claim that this intensifier is neutral in terms of its indexical stylistic properties (Dostie, Reference Dostie2018: 173). Other relatively frequent intensifiers in this corpus include tout, bien, extrêmement, tout à fait, complètement, and tellement. Notably missing from this sample are intensifiers carrément and énormément, as well as other more recent intensifiers such as grave, cher, méchant, sacrément, and extra (see Krištofíková, 2012), which is perhaps due to the relative low-frequency of these intensifiers and/or the relative formality of the interview setting in the ESLO data considered.

4.1.1 Longitudinal analyses

Comparing intensification rates longitudinally, speakers in the ESLO1 show a slightly higher intensification rate of 21.6%, compared to a rate of 17.4% in the ESLO2. Results from the first mixed-effects regression reveal this difference in intensification rate across the corpora to be statistically significant (β = −0.395, SE = 0.103, p = 0.016). If taken to reflect an overall trend in Hexagonal French towards less intensification over time, this finding would contrast with Fuchs (Reference Fuchs2017), who shows intensification in British English to have increased over time. While it is likely that these results stem from differences in data sources and methodological approaches (e.g. different timespans) between these two studies, these contrastive longitudinal trends in English and French may also reflect more systematic disparities between these two languages.

Differences are also found in the frequency with which individual intensifiers are used at each time point. Table 4 shows for each sample the percentage of total intensifiers represented by each intensifier (e.g. 594 tokens of très out of 805 total intensifiers in the ESLO1 = 73.7%). Such percentages both allow for comparison of the relative frequency of individual intensifiers in each sample and provide a measure of how much each intensifier has changed in frequency across the samples.

Table 4.

Longitudinal change in intensifier use between ESLO1 and ESLO2

The most notable finding among these results is the nearly 20% drop in use of très between the ESLO1 and ESLO2 (Figure 2), which is substantiated by results from the second mixed-effects regression model revealing the use of très to show a significant reduction in use across the two corpora (β = −2.26, SE = 0.546, p <0.001). Although très holds a monopoly in use at both time points, its decrease in use between the corpora coincides with a rise of occurrence in intensifiers such as tout, vraiment, and super, alongside less frequent intensifiers extrêmement and vachement. In particular, vraiment (the second most frequent intensifier at both time points) has more than doubled in its relative frequency between the ESLO1 and ESLO2, and tout (the third most frequent intensifier overall) is used four times as often in the ESLO2 than in the ESLO1. Other intensifiers such as bien, totalement, and trop remain relatively stable in terms of frequency of use between the two samples. Besides très, intensifiers that show a decrease in relative use between the two time points include parfaitement, absolument, tellement, and tout à fait.

Figure 2.

Percentage differentials of ESLO intensifiers between 1969–1970 and 2010.

As shown in Table 4, several intensifiers also show use in one sample but not the other. For instance, fort is used in the ESLO1 but not the ESLO2 sample, while super, vachement, and hyper occur in the ESLO2 yet not in the ESLO1 sample. However, it cannot be claimed that fort had fallen entirely out of use at the time the ESLO2 sample was collected, nor that super, vachement, and hyper were not attested at the time the ESLO1 was collected (cf. Cartier and Huyghe, Reference Cartier and Huyghe2021). Rather, these longitudinal frequencies suggest that these intensifiers have, respectively, fallen and risen in how commonly they are used in spoken French between 1970 and 2010. For example, although vachement does not appear in the ESLO1 sample considered here, eight tokens of this intensifier are found elsewhere in the ESLO1 corpus. Among these intensifiers, super stands out, as it represents 8.8% of all intensifiers in the ESLO2 sample, making it the third most frequent intensifier in the ESLO2 after très and vraiment. This steep rise in use suggests that super has become popular in the interim between when these corpora were collected. This intensifier, alongside other intensifiers hyper, méga, archi, and ultra (the latter three of which appear only once in the sample) are unique for two reasons. First, unlike the other intensifiers considered here, super, hyper, méga, archi, and ultra are typically considered intensifying prefixes, showing properties of polysemy and morphological productivity not shared by adverbal intensifiers (see Amiot, Reference Amiot2004; Izert, Reference Izert2015). Second, these prefixes are also used as intensifiers in modern-day English (Méndez-Naya, 2008: 216), raising the question of to what extent English has influenced the use of these intensifiers in French. While this topic is beyond the scope of this study, it is relevant for a larger understanding of English lexical influence on French (see e.g. Saugera, Reference Saugera2017) and merits further investigation in future research.

In sum, these results show a considerable amount of change in the French intensifier system over the past half-century, thus confirming the dynamic nature of the intensifier system over time. The following sections examines the grammatical and sociolinguistic dimensions of this change.

4.2 Grammatical change

4.2.1 Adjectival function

Consistent with past studies (e.g. Macaulay, Reference Macaulay2006; Tagliamonte, Reference Tagliamonte2008; Barnfield and Buchstaller, Reference Barnfield and Buchstaller2010), the majority of intensifiers in the sample modify predicative adjectives (n = 1,065), with less than a third modifying attributive adjectives (n = 435).Footnote ⁵ Figure 3 shows the extent to which the eight most frequent intensifiers modify predicative vs. attributive adjectives in the ESLO1 and ESLO2 samples (note the range of the y-axis). These results show that, apart from extrêmement (which shows less than 50% predicative use in both samples), the most frequent ESLO intensifiers are overall used more in predicative than in attributive contexts. This finding therefore generally corroborates Ito and Tagliamonte’s (Reference Ito and Tagliamonte2003: 276) claim of “an overriding constraint such that intensifiers are preferred with predicative adjectives […] regardless of the intensifier.”

Figure 3.

Change in adjectival function for the eight most frequent intensifiers in the ESLO sample.

Longitudinally, virtually no difference is found in the overall percentage of intensifiers that appear in attributive vs. predicative contexts between the corpora (71.1% predicative use in the ESLO1 vs. 70.9% in the ESLO2). This result differs from that of Barnfield and Buchstaller (Reference Barnfield and Buchstaller2010), who find an overall shift to increased predicative use over time in intensifiers in British English. However, considerable variation is found when considering the longitudinal patterning of individual intensifiers in the ESLO corpus (Figure 3). Intensifiers showing a fall in predicative use include tellement (by 28.4%), tout à fait (by 42.8%), bien (by 27.2%), and extrêmement (by 5.5%). Conversely, intensifiers showing a rise in predicative use include trop (by 23.3%), tout (by 11.2%), très (by 5%), and vraiment (by 3.4%). These patterns of change are confirmed by a significant two-way interaction found between corpus and adjectival function in the second mixed-effects regression model testing change across intensified adjectives (β = 1.35, SE = 0.532, p = 0.01).

Based on previous findings concerning the relationship between diachronic development and the adjectival function of intensifiers, intensifiers showing a rise in frequency longitudinally were also expected to show more predicative use over time (see Barnfield and Buchstaller, Reference Barnfield and Buchstaller2010). To test this prediction, Figure 4 plots the correlation between change in frequency of the most common intensifiers and change in percentage of predicative use over time. Results show those intensifiers that have fallen in use (tellement, tout à fait, and, to a lesser extent, bien) to also show a decrease in predicative use over time. However, frequent intensifiers that have increased in use over time reveal more varied results, with some showing an increase in predicative use (trop, tout, and, to a lesser extent, vraiment), while others (extrêmement) show a slight decrease in predicative use. Although this positive correlation between change in frequency and percentage of predicative use over time fails to reach statistical significance in a Spearman’s correlation test (R = 0.75, p = 0.066), these results nonetheless corroborate the findings of Barnfield and Buchstaller (Reference Barnfield and Buchstaller2010). It is important to note that the most frequent intensifier très (which has been removed from this analysis as its proportional decrease by 19.5% over time makes it difficult to compare to less frequent intensifiers) shows a slight increase in predicative use from 67.7% in the ESLO1 to 72.7% in the ESLO2. The fact that très deviates from the general pattern seen in Figure 5 lends credence to Tagliamonte’s (Reference Tagliamonte2008: 389) observation that predicative vs. attributive use as an indicator of intensifier development may not apply for high-frequency, older intensifiers (such as very in English).

Figure 4.

Correlation between change in proportional frequency and change in adjectival function for frequent ESLO intensifiers.

Figure 5.

Collocational width of intensifiers (operationalized through type-token ratios) in the ESLO1 and ESLO2 samples. Higher values on the x-axis indicate more lexical diversity. Intensifiers are organized from left to right in order of increasing overall lexical diversity.

Turning lastly to super and hyper, intensifiers found in the ESLO2 but not in the ESLO1 sample, these variants would be expected to seldom modify attributive adjectives given their apparently recent status in French as compared to other intensifiers (see Barnfield and Buchstaller, Reference Barnfield and Buchstaller2010). This prediction is largely borne out, as 80% (13/52) of super tokens and 76.9% (3/13) of hyper tokens appear in predicative contexts, providing support for the claim that, compared to older intensifiers, newer intensifiers tend to modify predicative adjectives more often.

4.2.2 Collocational width

In this section, data are analyzed by tracking the collocational width of intensifiers in the ESLO corpus across time. While past studies (e.g. Ito and Tagliamonte, Reference Ito and Tagliamonte2003; Tagliamonte, Reference Tagliamonte2008) have used closed-set classification schemes to investigate the extent to which intensifiers modify adjectives of different semantic types (adjectives describing, dimension, age, color, etc.), collocational width is operationalized here by quantifying the number of unique adjectives modified by individual intensifiers, without grouping them by semantic type. In addition to avoiding imbalanced distributions of adjectives across different semantic types, this method provides a measure of collocational width independent from researcher judgement in coding. Collocational width was calculated by deriving a type-token ratio (TTR; Templin, Reference Templin1957) of the modified adjectives for each intensifier, allowing comparison across intensifiers despite their differing frequencies in the ESLO. A common measure of lexical diversity, TTRs are calculated by dividing the total number of unique words (i.e. types) by the total number of words (i.e. tokens). This ratio yields a measure from 0, indicating complete repetition of a single type across all tokens, to 1, indicating a different type for each token. In other words, a higher TTR indicates more lexical diversity and greater collocational width.

Figure 5 plots the collocational width of intensifiers appearing ten times or more in the ESLO sample. The most frequent intensifier très has the lowest lexical diversity, meaning that it has the most repetition in terms of its adjectival collocates. 120 (or 11.9%) tokens of très are found modifying adjectival bien, 47 tokens (or 4.7%) modify difficile, and 43 (or 4.3%) tokens modify important. Despite showing a 20% decrease in relative frequency between the ESLO1 and ESLO2, très shows little change over time in terms of its collocational width.

While intensifiers such as trop and bien also show little change in collocational width across time, other intensifiers show more change between the two timepoints. For instance, absolument, tellement, and tout à fait show a higher TTR in the ESLO2, indicating increased lexical diffusion over time. Lexical diffusion is most pronounced for tout à fait: in the ESLO1 sample, tout à fait occurred seven times (or 24.1%) with the adjective différent and five times (or 17.2%) with normal, whereas each token of this intensifier in the ESLO2 sample modifies a unique adjective. Other intensifiers such as tout, vraiment, and, to a lesser extent, extrêmement all show lower TTRs, suggesting more lexical restriction over time. One example of this is seen with tout, which, in the ESLO1, collocates with the adjective petit-e-s five times (or 35.7%), whereas in the ESLO2, it occurs with this same adjective 22 times (or 41.5%).

Regarding those intensifiers occurring only in the ESLO2 sample, super and vachement show relatively low lexical diversity. For instance, ten tokens of super (or 14.9%) are found to modify sympa and eight tokens (or 11.9%) modify cher. For vachement, five tokens of this intensifiers (or 26.3%) are found in collocation with bien. These observations support Tagliamonte’s (Reference Tagliamonte2008: 376) claim that “incoming intensifiers can be expected to collocate with a small set of specific lexical items.” On the other hand, hyper shows higher diversity in terms of its adjectival collocates, possibly indicating that this intensifier is more delexicalized compared to super and vachement and/or that it is presents an exception to the above-stated tendancy for newer intensifiers’ to show restricted collocation patterns.

Longitudinally, intensifiers having decreased in use over time would be expected to show more restrained collocation, whereas those having increased would show wider collocation (cf. Tagliamonte, Reference Tagliamonte2008). This prediction is tested in Figure 6, where the correlation between change in frequency and change in collocational width for the most common intensifiers is plotted. These results reveal a positive correlation between change in use and collocational width, such that those intensifiers that have increased over time (e.g. vraiment, tout, extrêmement) also show an increase in collocational width. On the other hand, intensifiers that have decreased in use (e.g. tout à fait, absolument, tellement) have become more lexically restricted in the range of adjectives they modify. A Spearman’s correlation coefficient reveals frequency of use and collocational width to show a significant, strong positive correlation (R = 0.88, p<0.001) for the intensifiers included in Figure 6. Note that, although très is not included in this figure (as its greater magnitude of change across the corpora again makes its comparison to the other intensifiers difficult), a significant positive correlation is still found when très is included in this rank correlation (R = 0.79, p = 0.004).

Figure 6.

Scatterplot showing correlation between change in frequency of use and collocational width (as a function of the type-token ratio) for individual intensifiers.

Overall, these findings confirm previous claims that decreasing intensifiers tend to show more restrained collocation and that older, frequent intensifiers collocate with a wider range of adjective types. Nonetheless, newer intensifiers such as hyper are not necessarily restricted to collocation with a small set of adjective types, but can show wider application across adjectival types. Taken together, these results indicate a close relationship between collocational width, intensifier frequency, and how far along an intensifier is in the delexicalization process, while also highlighting the complex link between these factors.

The following section explores the role of sociolinguistic factors, namely gender (Section 4.3.1) and age (Section 4.3.2), as relating to change in the French intensifier system.

4.3 Sociolinguistic analyses

4.3.1 Speaker gender

Based on previous literature, female speakers were expected to use more intensification overall compared to male speakers. Figure 7, showing intensification rates in the ESLO1 and ESLO2 samples by speaker gender, reveals that, at both time points, female speakers show numerically higher intensification rates, although this gender difference is relatively larger in the older compared to the more recent corpus. Furthermore, both female and male speakers show on average lower intensification rates in the ELSO2 than in the ESLO1. Nonetheless, the relatively large error bars in Figure 7 indicate much individual variation across male and female speakers at both time points. Statistically, results from the first mixed-effects regression model show speaker age to not be a significant predictor of intensifier rate, neither as an individual predictor (β = −0.0276, SE = 0.266, p = 0.790), nor in interaction with other fixed effects (see Appendix A; Table A1). This result mirrors the finding of Fuchs (Reference Fuchs2017), who shows little change in intensifier rate in British English by speaker gender over time.

Figure 7.

Percentage of intensified adjectives by speaker gender and corpus. Error bars show the standard deviation for each group.

Table 5 shows variation in the use of individual intensifiers between female and male speakers in the ESLO1 and ESLO2 samples. Considering the five most common intensifiers, male and female speakers in the ESLO1 show comparable frequency of use for très, vraiment, bien, and extrêmement, whereas female speakers show relatively higher use of tout. In the ESLO2, however, there is more difference in the quantitative use of these intensifiers between male and female speakers: whereas female speakers use bien more at this time point, male speakers show more use of très, vraiment, and tout. Some gender differences are also found for less common intensifiers. For example, female speakers use tout à fait, tellement, and parfaitement more than male speakers in the ESLO1, yet show similar percentages of use for these intensifiers in the ESLO2 sample. The fact that these three intensifiers show decreased use over time in the ESLO sample may suggest that gender differences tend to become less pronounced for intensifiers on the decline, although more data would be needed to test this speculative claim.

Table 5.

Distribution of intensifiers in the ESLO1 and ESLO2 sample by speaker gender

Considering those intensifier that appear in the ESLO2 but not the ESLO1 sample, the use of vachement and hyper differ little in frequency between the female and male speakers. Conversely, super shows a marked difference in use between genders, with female speakers using this intensifier twice as frequently as male speakers, which may indicate that female speakers are leading the increasing use of this intensifier.

These results show that, despite a numerically small and non-significant difference in overall intensification rates between male and female speakers in the ESLO samples, nuanced differences are found between male and female speakers in the frequency of use of individual intensifiers. This finding thereby mirrors past studies on English (Precht, Reference Precht2008) and other languages (Kwon, Reference Kwon2012; Alshaboul et al., Reference Alshaboul, Jarrah, Alghazo and Al-Shawashreh2022) that have shown the relation between intensifier use and gender to vary depending on the intensifier considered. Taken in sum, these findings provide evidence against the long-standing claim of intensification as primarily a characteristic of female speech (e.g. Jesperson 1922), although female speakers may indeed play a role in advancing the increased use of newer intensifiers such as super. Based on past research, younger speakers would also be expected to be instrumental in this regard, a prediction which is considered in the following section.

4.3.2 Speaker age

Figure 8 shows intensification rates based on speaker age in the ESLO1 and ESLO2 samples. As seen in the left panel of this figure, speakers’ intensification rates are widely distributed, with virtually no correlation between intensification rate and speaker age in the ESLO1 (R = 0.0075, p = 0.97). On the other hand, the right panel shows a clustering of intensification rates below 20% for ESLO2 speakers under the age of 55, with a wider distribution of intensification rates for older speakers. Correlation coefficients reveal a positive, albeit non-significant, correlation between speaker age and intensification rate in the ESLO2 sample (R = 0.23, p = 0.28). Numerically, this result seems to suggest a trend towards consistently lower intensification use overall among younger speakers.

Figure 8.

Intensification rate by speaker age in the ESLO1 and ESLO2.

As shown in section 4.3.1, more nuanced patterns of variation and change can be found when considering intensifiers individually. One of the most striking patterns relating to speaker age and intensification use concerns the majority use of très. Despite the fact that très is used by nearly all speakers in both the ESLO1 and the ESLO2 samples, this intensifier is used markedly less among the youngest ESLO2 than by the other speakers sampled. For the vast majority of ESLO1 speakers and ESLO2 speakers over the age of 40, très makes up at least 50% of all intensifier tokens whereas ESLO2 speakers below the age of 40 show a notably lower percentage of très use. This finding is supported by results from the second mixed-effects regression model, which reveal a significant interaction between corpus and speaker age in the use of très vs. other intensifiers (β = 1.207, SE = .531, p = .023). Figure 9 shows a significant positive correlation between ESLO2 speakers’ proportional use of très and their age (R = 0.48, p<0.001), whereas the correlation coefficient for the ESLO1 reveals a very weak magnitude of associate between these factors (R = 0.12). This is consistent with the findings of Krištofíková (2012), who shows that contemporary younger speakers report using très less than older speakers.

Figure 9.

Percentage of très use out of total number of intensifiers by speaker age and corpus.

This pattern of age-grading seems to suggest a change-in-progress whereby the ubiquitous use of très is being supplanted by a wider array of intensifiers. Extrapolating from this observation, younger ESLO2 speakers would also be expected to be responsible for the rise in intensifiers such as super, tout, vachement, and vraiment, which have shown the greatest increase in use over time among the intensifiers sampled. Figure 10 plots the proportional use of each of these intensifiers in the ESLO2 by speaker age (for speakers who used one of these intensifiers at least once). Correlation coefficients from these plots show a strong negative correlation between proportional use of super and speaker age (R = −0.76), and weak negative correlations for tout (R = −0.11) and vraiment (R = −0.2). Conversely, a weak positive correlation is found between the use of vachement and speaker age (R = 0.22), although this may be due to the fact that relatively few speakers used this intensifier compared to super, tout, and vraiment. These results reveal an age effect to be particularly apparent for super: on average, speakers below 40 show higher rates of use of this intensifier than those over 40. Moreover, the majority of ESLO2 speakers who use super are under 50. Similarly, for vachement, five out of seven ESLO2 speakers who us intensifier are below 35, indicating that this intensifier is particularly prevalent among younger speakers. Lastly, for vraiment, it is shown that speakers who use this intensifier the most are among the youngest speakers sampled in this study. Taken together, these results provide some evidence that it is indeed the youngest speakers in the ESLO2 corpus who are shifting away from a majority use of très in favor of varied intensifier use.

Figure 10.

Percentage of individual ESLO2 speakers’ use of super, tout, vachement, and vraiment (as a percentage out of total intensifiers used) by speaker age.

Lastly, it is worth briefly mentioning how these newer intensifiers may relate to individual language change across the lifespan. Closer examination of intra-speaker variation in intensifier use (full details are provided in Appendix B) reveals that the use of super, hyper, and vachement is attested among ESLO2 speakers born prior to 1950. The fact that these more recent intensifiers are used by the oldest speakers in the ESLO2, yet not by the youngest speakers in the ESLO1 (all of whom were born between 1930 and 1950) may suggest that the oldest ESLO2 speakers have adopted these intensifiers later in life as they have grown more common in spoken French. It should also be noted that the contemporary use of super, hyper, vachement by older speakers is not anomalous, as speakers born from 1941–1948 also report using these intensifiers in Krištofíková (2012). This provides some support for a trajectory of language change whereby older speakers change in the direction of the speech community in their use of variants across time, as captured by apparent-time analyses (see Sankoff, Reference Sankoff2019).