Introduction
English intensifiers are a mainstay of the variationist canon, as they are frequent, easy to spot, and prone to rapid change (Barnfield & Buchstaller, Reference Barnfield and Buchstaller2010; Brinton & Arnovick, Reference Brinton and Arnovick2006:441; Tagliamonte, Reference Tagliamonte2008). They may be localized to a region or a social network (Aijmer, Reference Aijmer, Brezina, Love and Aijmer2018b; Barnfield & Buchstaller, Reference Barnfield and Buchstaller2010; Bauer & Bauer, Reference Bauer and Bauer2002; Macaulay, Reference Macaulay2006; Mustanoja, Reference Mustanoja1960; Peters, Reference Peters and Kastovsky1994; Rickford et al., Reference Rickford, Wasow, Zwicky and Buchstaller2007; Waksler, Reference Waksler, Baumgarten, Bois and House2012), show a sensitivity to style and register (Aijmer, Reference Aijmer, Hoffmann, Sand, Arndt-Lappe and Dillmann2018a; Beltrama & Staum Casasanto, Reference Beltrama and Staum Casasanto2017; Brown & Tagliamonte, Reference Brown and Tagliamonte2012; Waksler, Reference Waksler, Baumgarten, Bois and House2012; Xiao & Tao, Reference Xiao and Tao2007), and/or produce short-term fads (Barnfield & Buchstaller, Reference Barnfield and Buchstaller2010; Macaulay, Reference Macaulay2006). Any variety of English will have a large assortment of intensifiers (Peters, Reference Peters and Kastovsky1994:271; Tagliamonte, Reference Tagliamonte2008:390), but there is no guarantee of cross-varietal parallelism in either the inventory of intensifier forms or the constraints operating on them, or both (Barnfield & Buchstaller, Reference Barnfield and Buchstaller2010; Bauer & Bauer, Reference Bauer and Bauer2002; Tagliamonte, Reference Tagliamonte2008). They therefore encapsulate the dynamic but patterned nature of human language in general (Weinreich et al., Reference Weinreich, Labov, Herzog, Lehmann and Malkiel1968).
We harness these properties of English intensifiers to run a real-time, cross-modal study of language change embedded in hands-on pedagogy. For each of seven sequential years (2016 through 2022), we trained upper-year undergraduates at the University of Toronto taking LIN351 (Sociolinguistic Patterns in Language) in conducting sociolinguistic interviews, following Labov (Reference Labov1970, Reference Labov1971), as well as Poplack (Reference Poplack, Fasold and Schiffrin1989), Tagliamonte and Hudson (Reference Tagliamonte and Hudson1999), Van Herk (Reference Van Herk2008), Denis et al. (Reference Denis, Gardner, Brook and Tagliamonte2019), and Gardner et al. (Reference Gardner, Denis, Brook and Tagliamonte2021). Since intensifiers are highly accessible to undergraduates and tend to engage their interest, at least two previous class projects in sociolinguistics have focused on this variable (Van Herk, Reference Van Herk2008; Vaughn et al., Reference Vaughn, Kendall and Gunter2018). We combine this premise with a longitudinal element, as well as a cross-register design. Each student enrolled in LIN351 did two separate interviews with a single person, one orally (i.e., spoken) and one using computer-mediated communication (CMC). In the cases where individual students granted permission, we retained their data in the Sociolinguistics/Language Variation and Change Laboratory at the University of Toronto. The combined dataset—the LIN351 21st Century Corpus—affords a high-resolution look at linguistic change as it unfolds in real time in English, both spoken and online.
This paper is organized as follows. We first review major findings about English-language intensifiers, with a focus on the varieties of modern English where intensifiers are best studied (British and Canadian English). After that, we describe our methodology and the pedagogical components of the data collection, followed by the hypotheses for the real-time perspective. This is followed by the data analysis and interpretation, then the conclusion.
Background: Intensification in English
Adverbial intensifiers that modify adjectives can be divided into amplifiers and downtoners (Bauer & Bauer, Reference Bauer and Bauer2002; Bolinger, Reference Bolinger1972; D’Arcy, Reference D’Arcy2015; Ito & Tagliamonte, Reference Ito and Tagliamonte2003; Quirk et al., Reference Quirk, Greenbaum, Leech and Svartvik1985:445-446, 589-591; Stoffel, Reference Stoffel1901; see also Labov, Reference Labov and Schiffrin1984b). Amplifiers increase the degree of the modified adjective, while downtoners decrease it.Footnote 1 Amplifiers can be divided into maximizers, which denote the largest possible extent (entirely, absolutely, and completely) (Claridge et al., Reference Claridge, Jonsson and Kytö2021), and boosters, which increase the degree only somewhat (very, really, and so) (Barnfield & Buchstaller, Reference Barnfield and Buchstaller2010:256; Ito & Tagliamonte, Reference Ito and Tagliamonte2003:258).
Diachronically, intensifiers undergo quick turnover (Barnfield & Buchstaller, Reference Barnfield and Buchstaller2010; Bauer & Bauer, Reference Bauer and Bauer2002; Ito & Tagliamonte, Reference Ito and Tagliamonte2003:257; Méndez-Naya, Reference Méndez-Naya2003; Peters, Reference Peters and Kastovsky1994:269; Quirk et al., Reference Quirk, Greenbaum, Leech and Svartvik1985:590; Stoffel, Reference Stoffel1901). Since novelty helps convey amplification, new intensifiers come along regularly (Ito & Tagliamonte, Reference Ito and Tagliamonte2003:257; Peters, Reference Peters and Kastovsky1994:271; Tagliamonte, Reference Tagliamonte2008:391). Conversely, established amplifiers wear out and their meanings become eroded (D’Arcy, Reference D’Arcy2015; Ito & Tagliamonte, Reference Ito and Tagliamonte2003:257; Stoffel, Reference Stoffel1901:2; Tagliamonte, Reference Tagliamonte2008:391). Amplifiers that have become over-familiar are often replaced, either with new intensifiers or with recycled older ones that have lived on at low frequencies (Barnfield & Buchstaller, Reference Barnfield and Buchstaller2010; Buchstaller & Traugott, Reference Buchstaller and Traugott2006; Stenström, Reference Stenström and Kirk2000; Tagliamonte, Reference Tagliamonte2008). This means that synchronically, intensifier systems show layering of newer and older forms (Barnfield & Buchstaller, Reference Barnfield and Buchstaller2010; D’Arcy, Reference D’Arcy2015; Ito & Tagliamonte, Reference Ito and Tagliamonte2003; Tagliamonte, Reference Tagliamonte2008).
Historically, the intensifier very is attested as far back as the late 14th century (Breban & Davidse, Reference Breban and Davidse2016:238); once it became established, it dominated the English intensifier system until the middle of the 20th century (Barnfield & Buchstaller, Reference Barnfield and Buchstaller2010; Ito & Tagliamonte, Reference Ito and Tagliamonte2003:260; Palacios Martínez & Núñez Pertejo, Reference Palacios Martínez and Núñez Pertejo2012). Since then, across varieties of English, very has declined while really has increased. This trajectory has been reported in both northern and southern England (Barnfield & Buchstaller, Reference Barnfield and Buchstaller2010; Ito & Tagliamonte, Reference Ito and Tagliamonte2003; Lorenz, Reference Lorenz, Wischer and Diewald2002; Palacios Martínez & Núñez Pertejo, Reference Palacios Martínez and Núñez Pertejo2012), as well as New Zealand (Bauer & Bauer, Reference Bauer and Bauer2002; D’Arcy, Reference D’Arcy2015) and the province of Ontario, Canada (Tagliamonte, Reference Tagliamonte2008; Tagliamonte & Denis, Reference Tagliamonte and Denis2014).
Intensifiers are known to be sensitive to register and are usually found to be more characteristic of speech than writing (Biber et al., Reference Biber, Johansson, Leech, Conrad and Finegan1999:564-569; D’Arcy, Reference D’Arcy2015:451; Labov, Reference Labov and Schiffrin1984b:61; Xiao & Tao, Reference Xiao and Tao2007). Sub-registers of written language may also diverge in terms of how intensifiers pattern (Xiao & Tao, Reference Xiao and Tao2007:247), especially in CMC (Tagliamonte, Reference Tagliamonte2016).
Many studies of intensifiers focus on the language of teenagers or young adults (Bauer & Bauer, Reference Bauer and Bauer2002; Macaulay, Reference Macaulay2006; Palacios Martínez & Núñez Pertejo, Reference Palacios Martínez and Núñez Pertejo2012; Paradis, Reference Paradis and Kirk2000; Stenström, Reference Stenström, Hasselgård and Oksefjell1999, Reference Stenström and Kirk2000; Tagliamonte, Reference Tagliamonte2016), and some studies of British English have found that young people had a higher rate of overall intensification (Ito & Tagliamonte, Reference Ito and Tagliamonte2003:265; Palacios Martínez & Núñez Pertejo, Reference Palacios Martínez and Núñez Pertejo2012; Xiao & Tao, Reference Xiao and Tao2007:253). Amplifiers have also long been associated with women. As Ito and Tagliamonte (Reference Ito and Tagliamonte2003:260) pointed out, Stoffel (Reference Stoffel1901) attributed the use of amplification largely to women; Jespersen (Reference Jespersen1922:250) proposed that women lead changes in amplifiers; and Lakoff (Reference Lakoff1973, Reference Lakoff1975) suggested that intensifiers are a cornerstone of women’s usage along with hedging (though see D’Arcy, Reference D’Arcy2015:464-465). Empirical results are mixed. Some studies of contemporary British English have found that women use more intensifiers than men (Fuchs, Reference Fuchs2017; Stenström, Reference Stenström, Hasselgård and Oksefjell1999), but Ito and Tagliamonte (Reference Ito and Tagliamonte2003) reported no gender effect in the spoken English of the city of York. Cross-register studies may identify an interaction between gender and register; for instance, Xiao and Tao (Reference Xiao and Tao2007) found that in the British National Corpus, women used more intensifiers than men in writing, while in speech there was no effect.
Our study, analyzing the outcome of an iterative pedagogical exercise, serves two objectives. The first is pedagogical and will be described in the methodology section. The second has to do with the analysis of the accrued data, focusing on the intensifiers of young adults in and around Toronto and the social and/or linguistic factors that affect the rates of use of leading intensifiers. While we also take gender and register (speech or instant messaging [IM]) into account, the primary aim of the empirical study is to capture rapid change unfolding across a narrow slice of real time. If intensifiers change very quickly, we can reasonably expect to find a change in progress in seven years of data. As Barnfield and Buchstaller (Reference Barnfield and Buchstaller2010:255) highlighted, two studies in the literature have uncovered change in intensifiers over a mere eight-year span: Tagliamonte and Roberts (Reference Tagliamonte and Roberts2005) found the use of so on the television sitcom Friends correlating with the show’s popularity, while Macaulay (Reference Macaulay2006)’s young adults in Glasgow retreated from the intensifier dead in favor of pure. Therefore, our goal from the outset of the project in 2016 was to catch change unfolding among intensifiers in Ontario English. Specific hypotheses follow after the description of our methodology.
Methodology
The pedagogical design
The data in this study were collected by a subset of students enrolled in Linguistics 351 (LIN351: Sociolinguistic Patterns in Language) between 2016 and 2022 at the University of Toronto.Footnote 2 LIN351 introduces variationist theory and practice and is aimed at upper-year undergraduates who have already completed an introductory sociolinguistics class. Three of the four homework assignments that we designed for LIN351 train the students in hands-on variationist methods: conducting sociolinguistic interviews, transcribing, extracting, and coding.Footnote 3 Each year, we began by emphasizing ethical necessities (see also Schilling, Reference Schilling2013; Tagliamonte, Reference Tagliamonte2006; Van Herk, Reference Van Herk2008) and having students practice interview techniques (Labov, Reference Labov, Baugh and Sherzer1984a). After that, every student found one person to interview twice, once via speech (the SP interview) and once via IM (the IM interview). The interviews were each required to be 30 minutes or longer. To avoid a conflation of register (SP or IM) and the temporal order of the interviews, we counterbalanced the order such that half of the class would conduct the SP interview before the IM one, and the other half would do the opposite.Footnote 4 Either way, to mitigate the chances that the second interview would merely rehash the first, we asked the students to conduct the interviews at least 48 hours apart.
The SP interview was meant to be conducted face-to-face, with the other completed online using synchronous IM.Footnote 5 The SP interview was required to be audio-recorded using a common media filetype (e.g., .mp3 or .wav); the IM interview could be recorded on any hardware/software, but students were informed that the full transcript would have to be copied and pasted into a basic text document (.txt).
All of the students were required to sign a standardized consent form, which they submitted in advance and which we checked before allowing them to proceed with data collection. After the interviews, each student submitted their sound recordings and text records. For the SP interviews, they transcribed 20 minutes from any portion of text other than the beginning. For the IM interviews, they were required to add line numbers. Students also submitted interview reports (with demographic information and descriptions of how the interviews had gone) and lists of pseudonyms. Finally, the students extracted and coded the intensifiable adjectives in their data and used a Microsoft Excel template to code each token for the linguistic and social factors that had been discussed in class and in tutorials with the teaching assistants (TAs). For the final homework assignment each year, the teaching team pooled the data from that year only, conducted some basic distributional analyses, and asked the students to examine the tables and graphs and do some analysis and interpretation of their collective findings.
Adjustments and limitations
Minor changes in method occurred over the duration of the project. The first year focused on L1 Canadian English speakers, but after that, the sample was expanded to include both L1 and L2 speakers of English from a larger mix of nationalities and origins. Given this change, we altered the interview report to collect more information about the linguistic background of the individuals. For the first five years, we tried to balance the sample, at least in terms of binary gender categories.
The final two years of data collection (2021 and 2022) were complicated by the COVID-19 pandemic. In 2021, LIN351 was taught entirely online; in 2022, it was offered in a hybrid fashion, with in-person lectures and online TA-led tutorials. Students in 2021 and 2022 were permitted to conduct their spoken interviews either in person or via Zoom. We also allowed electronic submission of all of the paperwork. Given a higher rate of interviewee withdrawals, we also relaxed attempts at balancing the sample design by binary gender.
We refer to the whole set of data collected by the students between 2016 and 2022 as the LIN351 21st Century Corpus. The amount of data from each year varies for two independent reasons. One is that enrollment numbers in LIN351 fluctuated from year to year; we had everything from a few dozen to nearly 100 students. The other is that the proportion of students who granted permission for their interview materials to be archived also changed, especially as the ethics review process by the Delegated Ethics Review Committee (DERC) became more stringent over the years and shifted toward an opt-in model rather than an opt-out one. Regardless, we archived only data from students who gave us permission to do so.
Subsample of the LIN351 corpus used in the present study
Given our goal of tapping linguistic change in real time, we focused on individuals with common social characteristics: between the ages of 16 and 23 years, L1 English speakers, and either born and raised in Canada or brought there before the age of 6 years (see Chambers, Reference Chambers1992; Tagliamonte & Molfenter, Reference Tagliamonte and Molfenter2007). Given these parameters, the data for analysis comprises 88 individuals. Per self-reported labels for gender, 52 are female, 34 male, and 2 nonbinary/other. In practice, almost all the individuals turned out to have spent most or all of their lives in the province of Ontario, but a small proportion have lived elsewhere in Canada.
Table 1 shows the 88 individuals in the subsample divided by year of birth (YOB), binned into three categories: 1993 through 1996, 1997 through 1999, and 2000 through 2003.
Table 1. Subsample of the LIN351 21st Century Corpus: individuals (n = 88) by year of birth
Hypotheses
The youngest individuals in the Toronto English Archive (TEA, collected between 2002 and 2006) as investigated by Tagliamonte (Reference Tagliamonte2008) were born in 1990 and 1991. They have an intensifier system dominated by the incoming really, which accounts for more than 20% of intensifiable contexts. Very, having declined in apparent time, joins so and pretty at rates below 5% each. The oldest individuals in our study were born in 1993, so we anticipate that they will show similar rates of these four intensifiers.Footnote 6
We expect to catch evidence of a new incoming variant in the intensifier systems of the younger individuals in our study. Which variant might this be? The leading candidate is so. In the first TEA corpus collected between 2002 and 2006, so was secondary but increasing, with a female lead (Tagliamonte, Reference Tagliamonte2008). Just a few years later, in casual CMC registers among local teenagers between 2009 and 2010, “[b]y far the most common intensifier is so (13.7%) [while] really, pretty, and especially very occur at much lower frequencies” (Tagliamonte, Reference Tagliamonte2016:20).
Another contender for an incoming intensifier in Ontario English is super. In the TEA, it is just one of “innumerable rare forms” (Tagliamonte, Reference Tagliamonte2008:390; see also Waksler, Reference Waksler, Baumgarten, Bois and House2012). Aijmer (Reference Aijmer, Brezina, Love and Aijmer2018b:75) suggested that super “is quickly and dramatically increasing in frequency” in both American and British English—though Vaughn et al. (Reference Vaughn, Kendall and Gunter2018:306), using perceptual data from undergraduates in the United States, cautioned that super might be passé there.
Therefore, if we find an incoming so in the LIN351 21st Century Corpus, we expect the change to be led by women; if the incoming form is any other intensifier, we cannot straightforwardly predict a gender effect. While women lead most linguistic changes (Labov, Reference Labov1990, Reference Labov2001:284, 306-309), it is unclear whether this would be true of the intensifier system of Ontario English. Evidence for a female lead associated with the major innovation really is only partial (Tagliamonte, Reference Tagliamonte2008) and may be limited to non-narrative contexts of conversation (Brown & Tagliamonte, Reference Brown and Tagliamonte2012).
In terms of register (SP versus IM), the design of the methodology allows us to test whether, for this variable, informal CMC keeps pace with changes to the spoken language, as predicted by Tagliamonte (Reference Tagliamonte2016:5). As long as IM is functioning as a casual register, we expect the IM intensifiers to act like SP ones. While linguistic change generally emerges in spoken (or signed) registers before formal writing (Pintzuk, Reference Pintzuk, Brian and Richard2003), we anticipate that colloquial CMC will behave in tandem with speech.
Extraction and coding
Following Ito and Tagliamonte (Reference Ito and Tagliamonte2003) and Tagliamonte (Reference Tagliamonte2008, Reference Tagliamonte2016), we extracted all potentially intensifiable adjectives, whether or not they had an intensifier modifying them. Adjectives without intensifiers are plentiful, which means that proportions of any given intensifier rarely exceed 25% of the whole variable context. While some studies opt to exclude the bare adjectives and focus on intensified contexts only (Barnfield & Buchstaller, Reference Barnfield and Buchstaller2010; Rickford et al., Reference Rickford, Wasow, Zwicky and Buchstaller2007; Van Herk et al., Reference Van Herk, Thorburn and Buchstaller2015), we include the unintensified adjectives for the sake of comparability with the earlier studies of Ontario English (Tagliamonte, Reference Tagliamonte2008, Reference Tagliamonte2016; Tagliamonte & Denis, Reference Tagliamonte and Denis2014). This consistency in methodology means that we can compare the overall intensification rate over time and establish whether it is stable or changing.
As noted, the pedagogical design had the students extracting and coding tokens from their own data. However, putting them together resulted in a patchwork of hundreds of slightly divergent sets of choices. To ensure a streamlined methodology, we used a script written in Python (version 3.9.1) employing the spaCy package (Honnibal & Montani, Reference Honnibal and Montani2017) for part-of-speech tagging that extracted all of the adjectives in the LIN351 corpus. All intensifiers—amplifiers and downtoners alike—were included across both attributive and predicative contexts. Adjacent words and phrases likely to be intensifiers were tentatively labeled as such by default; otherwise, the tokens were coded Ø. Finally, we filtered the hits to reflect the subsample of individuals, checked the tokens one by one, removed non-intensifiable contexts (as per Ito & Tagliamonte, Reference Ito and Tagliamonte2003), ensured that intensifiers and zeroes had all been accurately identified, and finished coding the linguistic and social factors manually (most of which go unaddressed here because they did not turn out to be relevant to the variation or to the main findings). This methodology yielded 2,509 tokens of intensifiable adjectives from 88 individuals interviewed by LIN351 students between 2016 and 2022.
Results and discussion
Overall intensification rate
Nine hundred ten tokens were intensified for an overall intensification rate of 36.3%. While this is similar to the 36.1% amplification rate reported by Tagliamonte (Reference Tagliamonte2008:367) for individuals in Toronto from 2002 to 2004, there are two points of divergence. One is that we have included downtoners, which represented 3.2% of the tokens, while Tagliamonte (Reference Tagliamonte2008) found a similarly low rate of downtowners and grouped them with non-intensified contexts. The other difference is that the 2008 study is based on speech alone, rather than both speech and CMC. Our SP data on their own yield an intensification rate of 38.0%—though this still includes downtoners. Regardless, an overall intensification rate in this range supports the notion of a baseline for Canadian English that is higher than that for British English, as represented by York at 24% (Ito & Tagliamonte, Reference Ito and Tagliamonte2003).
Intensification rate by social factors
Table 2 differentiates the intensification rate by gender and register. There are main effects for both, but no interaction. As per some earlier reports, women use more intensifiers than men (Fuchs, Reference Fuchs2017; Jespersen, Reference Jespersen1922; Stoffel, Reference Stoffel1901; Stratton & Sundquist, Reference Stratton and Sundquist2022). There are more intensifiers in speech than in IM; this matches the findings of cross-register studies of intensifiers by Biber et al. (Reference Biber, Johansson, Leech, Conrad and Finegan1999) and Xiao and Tao (Reference Xiao and Tao2007).
Table 2. Intensification rate by binary gender and register (note: this excludes two nonbinary individuals due to small token counts [n = 13 combined])
As diachronic stability in the intensification rate is unlikely (Barnfield & Buchstaller, Reference Barnfield and Buchstaller2010; D’Arcy, Reference D’Arcy2015; Ito & Tagliamonte, Reference Ito and Tagliamonte2003; Tagliamonte & Denis, Reference Tagliamonte and Denis2014), Table 3 investigates intensifier use by the YOB of the individuals in the data. The trajectory shows a small decrease from the first cohort to the subsequent two, though combining the subsequent two to check for a statistically significant change returns only a marginal result (Pearson’s χ2: p = 0.059, df = 1; n = 2509). There is no clear evidence of a change in the intensification rate, especially if age-graded behavior in Toronto leads people in their twenties to use more overt intensifiers than everyone else (Tagliamonte, Reference Tagliamonte2008:367).
Table 3. Intensification by year of birth (binned)
With the assumption of a stable intensification rate, we proceed to an analysis of the intensifier variants themselves.
Distribution of intensifiers
Table A1 in the Appendix contains a list of the 56 distinct lexical items that serve as intensifiers in the data. Nearly half of these (n = 26) appear only once each in the LIN351 subsample. Table 4 separates the variants represented by 10 or more tokens apiece (as per Tagliamonte, Reference Tagliamonte2008:368). In descending order, the top five are really, so, very, pretty, and super.
Table 4. Individual intensifiers with ≥10 tokens each, with the two additional options (rarer intensifiers and zero tokens) included for context
The most prevalent intensifier in the LIN351 subsample data is really, just as it was in the TEA (Tagliamonte, Reference Tagliamonte2008:368). However, its rate is unexpectedly low. In the TEA, really reached 13.0% (Tagliamonte, Reference Tagliamonte2008:368). In a set of smaller towns of southeastern Ontario, it was “approximately 14%” (Tagliamonte & Denis, Reference Tagliamonte and Denis2014:114). This included both the young people and their older counterparts who had much less really. In this subsequent study with young adults alone, really is less frequent than anticipated. However, the relative order of really, so, and pretty is the same as in the youngest individuals in the TEA (Tagliamonte, Reference Tagliamonte2008:372). Very is ranked higher than pretty rather than below it. Super is fifth at 2.2%. It was so much lower in the TEA, collected between 2002 and 2006, that it was combined into the “other” category, which even all together represented only 1.4% of the data (Tagliamonte, Reference Tagliamonte2008:368, Table 3).
Diachrony
Figure 1 breaks down the distribution of intensifiers by binned YOB.
Figure 1. Rates of the top five intensifiers by year of birth.
The oldest cohort has the intensifier system that we expected based on the prior studies of Toronto English (Tagliamonte, Reference Tagliamonte2008, Reference Tagliamonte2016): really is the most prevalent, so and pretty are also well established, and very is low frequency. Contrary to expectation, neither so nor super is increasing by YOB. Rather, there is a net decline among all the top intensifiers except very. Within the space of 11 years of birth, very becomes the most prevalent intensifier in the system. The individuals born between 2000 and 2003 are using it even more than they use really. The intensifier so falls the most dramatically, and super is stagnant in fifth place.
We had expected to find a new incoming form among intensifiers in seven years of real-time data collection, but it is curious that very would be the only intensifier increasing in frequency. Very is an older and more conservative variant (Aijmer, Reference Aijmer, Brezina, Love and Aijmer2018b; Barnfield & Buchstaller, Reference Barnfield and Buchstaller2010:263; D’Arcy, Reference D’Arcy2015; Fries, Reference Fries1940:204-205; Tagliamonte, Reference Tagliamonte2008; Xiao & Tao, Reference Xiao and Tao2007:259). As of 2009-2010, it was the only intensifier that Tagliamonte (Reference Tagliamonte2016) found in academic essays written by young adults; in CMC registers from the same individuals, very is a distant fourth after so, really, and pretty—consistent with being at the end of a long downward trend. What could explain the shift to higher rates of very, of all things, in Figure 1?
Before we can confidently conclude that this is a change in progress, we consider and evaluate two alternative explanations that could have been caused by pandemic-induced changes to our methodology in 2021 and 2022. One has to do with gender. After 2020, given the additional practical challenges, we no longer aimed for a balanced gender sample in the individuals interviewed by our students. If (1) very is an older, more-standard variant, if (2) women generally use higher rates of more-standard variants (Labov, Reference Labov1990), and if (3) the LIN351 sample became increasingly female, this could account for the increasing use of very without an appeal to linguistic change. The other alternative explanation has to do with how the spoken interviews were conducted. While videoconferencing does not automatically undermine the informal nature of a spoken sociolinguistic interview (Gardner & Kostadinova, Reference Gardner and Kostadinova2024), the presence of a formal variant, that is, very, might signal nothing more than an effect of subtle modality change.
Examination of the top five intensifier variants by the individual’s gender, in Table 5, suggests that women use slightly more very than men do.
Table 5. Rates of the top five intensifiers by gender (n = 2509)
However, analysis of the gender distribution by year of interview (not shown in Table 5) reveals that the proportion of women hovers around 50% and does not increase after 2020. Therefore, the rise of very in Toronto between 2016 and 2022 cannot be attributed to the possibility of more female interviewees later in the project.
Similarly, in 2021, when nearly every LIN351 student did a spoken interview via videoconferencing rather than in person, the rate of very is—if anything—relatively low (5.6%). Although the proportion of very is highest in 2022, only five out of 88 individuals in the subsample were interviewed that year. We thus conclude that neither the gender balance of the sample nor the introduction of videoconferencing can account for the increase in very in our data over time.
One subtler alternative explanation remains with respect to the IM data. The original methodology for data collection was better-suited to 2016 than to 2022, given the pace of technological change and how it affects communicative norms (see Tagliamonte, Reference Tagliamonte2016:8-9). The original setup specified that the IM interview had to be conducted through keyboards attached to computers. This directive proved to be more unusual in 2022 than it had been in 2016; six years later, it was less common for students to have an IM conversation this way. Several LIN351 students in the later years reported that they rarely/never talked to their friends via anything like the IM setup we had in mind. This risked making the task somewhat or highly artificial, even for pairs of good friends—possibly undermining access to the vernacular. For this reason, toward the end of the seven-year period, the later instructor (Brook) told the students that they were welcome to message each other with smaller devices and/or using their thumbs if they wanted, as long as they had an easy way of extracting the transcript at the end. It remains possible that what the 2021-2022 students ended up doing for the IM condition was not a good match for anything they were already doing with their friends. However, this possibility is unlikely to be the explanation for the higher rates of very. Although register mismatch could have amplified rates of very in IM selectively, Table 6 shows that the rates in the SP data from the same set of 88 individuals are higher still.
Table 6. Rates of the top five intensifiers by register (n = 2509)
Multivariate analysis
Table 7 shows the results of a binomial mixed-effects logistic regression conducted in R (version 4.2.1, R Core Team, 2022) with very treated as the predicted variant versus all the other alternatives (including zeroes). We include gender (for individuals who reported a binary gender identity), sexuality (given the hint of a queer lead in the proportions), binned YOB, and register.
Table 7. Mixed-effects logistic regression of the effects of gender (excluding tokens from nonbinary individuals), sexuality (excluding tokens from individuals with unknown orientations), YOB (binned into three groups), and register (n = 2285)
** indicates “highly significant” (p < 0.01),
*** indicates “very highly significant” (p < 0.001).
The number of tokens of the predicted variant very is limited (n = 145/2285 intensifiable adjectives), and only one of the factors attained significance: binned YOB. The individuals in the oldest group (those born between 1993 and 1996) have significantly less use of very than the individuals in the middle group (born between 1997 and 1999). The youngest group exceeds the very rate of the middle group, though only slightly and not enough to cause an additional significant increase.
Given statistical verification and the rejection of three alternative explanations for the apparent increase in very by YOB, we conclude that this is a genuine case of linguistic change, with very as the incoming intensifier. The hints of leads by women and/or queer individuals do not attain significance, nor does the higher proportion in speech than in IM.
Discussion
A resurgent very
While we anticipated finding an incoming intensifier, the innovation that we discovered is neither so nor super, but arguably the least likely contender: very. This variant has been a prominent intensifier in English for more than 500 years (Breban & Davidse, Reference Breban and Davidse2016:228; Ito & Tagliamonte, Reference Ito and Tagliamonte2003:265; Mustanoja, Reference Mustanoja1960:327). It spent most of that time as the majority form (D’Arcy, Reference D’Arcy2015:485), but toward the end of the 20th century, very declined in the vernacular in favor of really across several varieties of English, including British, New Zealand, and Canadian (Barnfield & Buchstaller, Reference Barnfield and Buchstaller2010; Bauer & Bauer, Reference Bauer and Bauer2002; D’Arcy, Reference D’Arcy2015; Ito & Tagliamonte, Reference Ito and Tagliamonte2003; Lorenz, Reference Lorenz, Wischer and Diewald2002; Palacios Martínez & Núñez Pertejo, Reference Palacios Martínez and Núñez Pertejo2012; Tagliamonte, Reference Tagliamonte2008; Tagliamonte & Denis, Reference Tagliamonte and Denis2014). By the beginning of the 21st century, very was “out-going” in Toronto (Tagliamonte, Reference Tagliamonte2008:382), “secondary” in southern England (Palacios Martínez & Núñez Pertejo, Reference Palacios Martínez and Núñez Pertejo2012:791), and “old-fashioned” in British and American Englishes alike (Aijmer, Reference Aijmer, Brezina, Love and Aijmer2018b:61). As recently as 2009-2010, in the written English of young people in Toronto, very was characteristic only of formal academic prose, and rare in casual registers of CMC (Tagliamonte, Reference Tagliamonte2016:22). In other words, only a few years ago, in young adults in the same city, very was an old prestige variant—anything but vernacular. Regardless, our results find very returning to wider use in young Canadian adults who are mostly from Ontario. From data collected between 2016 and 2022, our youngest cohort (born between 2000 and 2003) uses very more than they use really in both speech and IM. We have considered, but ruled out, three alternative accounts of a rise in very. This, in conjunction with statistical verification of the increase between our oldest cohort (born in 1993 through 1996) and middle cohort (1997 through 1999), has led us to interpret the YOB effect as a linguistic change. Still, such a thing seems surprising. Could very be the incoming dominant intensifier, after so recently seeming destined for obsolescence?
If the individuals in our data are representative of Canadian English, and if Canadian English is patterning as per American English in this respect, then complementary evidence for the notion of a resurgent very comes from the results of Vaughn et al. (Reference Vaughn, Kendall and Gunter2018). Like Van Herk (Reference Van Herk2008) and this study, the authors focused on intensifiers as an accessible, interesting case study for undergraduates diving into the study of language variation and change. Unlike Van Herk (Reference Van Herk2008) and the methodology employed here, Vaughn et al. (Reference Vaughn, Kendall and Gunter2018) probed the possible social meanings of intensifiers in English in the United States. The results for very, reproduced in Table 8, are dichotomous. Respondents across age groups report that very sounds the most “intelligent” of very, really, real, super, and a bare adjective—exactly as could be expected of an old prestige form. Likewise, older respondents (those 60 to 81 years old at the time of the study) position very as the least “hip/trendy,” the least “cool,” and moderately old-fashioned. However, “listeners’ associations … have almost completely swapped over apparent time” with respect to the perception of very (Vaughn et al., Reference Vaughn, Kendall and Gunter2018:306). Among young-adult respondents, very places first for “hip/trendy,” “friendly,” and “Millennial;” second for “annoying” (after really); and an astonishing last for “old-fashioned.”
Table 8. Ranking of five intensification options by social meaning in two groups of respondents, from Vaughn et al. (Reference Vaughn, Kendall and Gunter2018:306, Table 1)
Vaughn et al. (Reference Vaughn, Kendall and Gunter2018) examined perception rather than production, and American rather than Canadian English. However, we suggest that their findings help to explain ours, and vice versa. While very retains its connotation of “intelligence” from earlier times, a resurgence in use among young people would go hand-in-hand with connotations of being “hip/trendy” and “cool.” Beltrama and Staum Casasanto (Reference Beltrama and Staum Casasanto2017) found exactly this correlation for the intensifier totally in perceptual data from American English, such that linguistic contexts more characteristic of younger speakers also sound younger to onlookers.
Accounting for very on the upswing
The associations of very with formality, standardness, and the written word endured until recently. In young adults in Toronto, as recently as 2009-2010, very was seldom found outside formal writing (Tagliamonte, Reference Tagliamonte2016). Of course, quick change in the intensifier system is commonly observed (Barnfield & Buchstaller, Reference Barnfield and Buchstaller2010; Macaulay, Reference Macaulay2006; Peters, Reference Peters and Kastovsky1994:269; Quirk et al., Reference Quirk, Greenbaum, Leech and Svartvik1985:590; Tagliamonte & Roberts, Reference Tagliamonte and Roberts2005), and intensifiers are easily recycled (Bolinger, Reference Bolinger1972:18; Buchstaller, Reference Buchstaller2006; Ito & Tagliamonte, Reference Ito and Tagliamonte2003; Stenström, Reference Stenström and Kirk2000; Tagliamonte, Reference Tagliamonte2008:362). This was the motivation for the selection of intensifiers as the focus of this compressed and granular real-time study. However, examples of rapidly incoming intensifiers—such as well in British English (Aijmer, Reference Aijmer, Brezina, Love and Aijmer2018b, Reference Aijmer2021; Stenström, Reference Stenström and Kirk2000)—tend to come from the vernacular end of the system rather than the conservative end. In this case, intensifier recycling has unexpectedly seized upon an old prestige form. It is almost as if young Canadian English users had enthusiastically re-adopted the obsolescent whom (see Bohmann et al., Reference Bohmann, Bohmann and Hinrichs2021) as a way of signaling coolness and hipness. Can recycling be convincingly linked to the revival of very?
In classic intensifier recycling, once-prominent variants become low in frequency, “tend[ing] to remain in the reservoir of forms that users may deploy to intensify” (Barnfield & Buchstaller, Reference Barnfield and Buchstaller2010:281). It must be the case that, just as newly popular intensifiers become victims of their own success by wearing out, untrendy intensifiers may quickly regain novelty and become freshly attractive, even to young adults looking for new vernacular options. As per Beltrama and Staum Casasanto (Reference Beltrama and Staum Casasanto2017:178), “marked variants tend to be particularly salient carriers of social meanings across phonological, morphosyntactic and semantic types of variation.” More specifically, Van Herk and Childs (Reference Van Herk, Childs, Cacoullos, Dion and Lapierre2015) suggested that when a variant has been declining, it might suddenly acquire salience and the ability to pick up social meaning. While nothing guarantees immortality for that given intensifier,Footnote 7 in this case, a restricted very must have become marked enough to catch the collective attention of young adults.
Arguably, the larger mystery in our findings is why so is not the incoming form, after it seemed poised to move into this role in Toronto (Tagliamonte, Reference Tagliamonte2008, Reference Tagliamonte2016). Relative to very, the intensifier so may have disadvantages both linguistic and social. Linguistically, so originated as a postverbal comparative structure and still lacks a productive attributive use, as in *a so exciting day (Bauer & Bauer, Reference Bauer and Bauer2002; Tagliamonte, Reference Tagliamonte2008:374). In the data for this study, so is only ever found predicatively (e.g., the day was so exciting).Footnote 8 This means that so cannot compete as an intensifier of attributive adjectives.
Socially, so and pretty are variants that Tagliamonte (Reference Tagliamonte2008) found to have opposite gender associations in the TEA, with so being favored by women and pretty by men. Figure 2 splits the proportions of the top five intensifiers by binary gender and register, showing a straightforward replication. The gender connotations of so and pretty in Toronto are stable and enduring; they affect both speech and IM.
Figure 2. Rates of top five intensifiers by register and two gender categories (n = 2496).
This means that among the top five intensifiers, if the older incoming form really is losing its novelty and strength as an intensifier, very may be the most obvious gender-neutral alternative. In the last few years, Western societal discourse has come to recognize the complexity and nuance of gender (see Airton, Reference Airton2019) and inspired large numbers of conversations about inclusivity and language (e.g., Conrod, Reference Conrod2019; Konnelly & Cowper, Reference Konnelly and Cowper2020), possibly giving an advantage to forms with gender-neutral connotations. For instance, in a study of changing attitudes toward y’all in the United States, McCurdy (Reference McCurdy2022) found insufficient explanatory power in the linguistic/functional argument—that English benefits from a distinction between singular and plural second-person pronouns. He proposed that the key incentive was instead the clearly gender-neutral nature of y’all, relative to you guys, in American English. Among groups such as northerners who have little historical precedent for the use of y’all, its adoption can signal progressivism and/or queerness.Footnote 9
Another possibility, not mutually exclusive with the first, is that very has caught the edge of what Crystal (Reference Crystal2008:147) referred to as “comic archaism.” Aijmer (Reference Aijmer, Hoffmann, Sand, Arndt-Lappe and Dillmann2018a:75) reported online metacommentary suggesting that very has “a kind of anachronistic charm.” Across variables, minor variants from the periphery of the feature pool are easily used in language play (Bohmann et al., Reference Bohmann, Bohmann and Hinrichs2021; Brook & Blamire, Reference Brook and Blamire2023; Crystal, Reference Crystal1998:147, Reference Crystal2008:147; van Compernolle, Reference van Compernolle2008:331). Given that intensifiers are a showcase for linguistic creativity (Ito & Tagliamonte, Reference Ito and Tagliamonte2003), we should expect intensifier systems to attract playful language of all kinds (see Brook & Blamire, Reference Brook and Blamire2023:521-522).
Conclusion
For seven consecutive years, we used the teaching/research enterprise to conduct a pedagogical project in an undergraduate course to train students in sociolinguistic interviewing techniques. After 2022, we pooled the data that we were granted permission to use in ongoing research and used it to look for real-time change in English-language intensifiers: a linguistic system known to be the site of rapid, creative layering of multiple forms over time. The findings support our prediction that we would discover an incoming form—though we did not suspect this would be the established prestige form very. However improbable this outcome seems, no alternative interpretation could account for this development. Moreover, statistical modeling confirmed a significant increase in the rate of very between the oldest cohort (born 1993 through 1996) and the younger two cohorts (born 1997 through 1999 and 2000 through 2003). We conclude that in Ontario English, very has been enthusiastically recycled by young adults. This finding is consistent with recent trends in American English, where very is reported to have the social meaning of young, hip, cool, and definitely not old-fashioned (Vaughn et al., Reference Vaughn, Kendall and Gunter2018).
To account for these findings, we have suggested that for young adults, very has become just marked enough (as per Van Herk & Childs, Reference Van Herk, Childs, Cacoullos, Dion and Lapierre2015) to have renewed appeal as an intensifier. Elements that may support the re-emergence of very include: (1) very has a linguistic advantage over the syntactically restricted so; (2) it is more gender-neutral in Ontario English than either so or pretty; and (3) it might signal a facetious hyper-standardness to young people (Aijmer, Reference Aijmer, Hoffmann, Sand, Arndt-Lappe and Dillmann2018a:75).
Given the nature of the data, our findings also have implications for broader issues in the field of variationist sociolinguistics. Most notably, the amount of coherence in the data speaks to the ability of naturalistic data to reflect linguistic change straightforwardly (and to the underlying orderly quality of those changes, as per Weinreich et al., Reference Weinreich, Labov, Herzog, Lehmann and Malkiel1968). Despite being distributed among hundreds of different LIN351 students, two separate instructors, multiple TAs, and timepoints before and during the COVID-19 pandemic, systematic patterning emerges from a robust vernacular.
The revival of very, odd though it seems, emphasizes the extent to which constant change is an integral part of the intensifier system (Bolinger, Reference Bolinger1972:18; Buchstaller, Reference Buchstaller2006; Ito & Tagliamonte, Reference Ito and Tagliamonte2003; Stenström, Reference Stenström and Kirk2000; Tagliamonte, Reference Tagliamonte2008:362). Only a short period off-stage as a minor intensifier was enough for it to be repurposed, upcycled, and imbued with new social meanings (Vaughn et al., Reference Vaughn, Kendall and Gunter2018). In contemporary English, then, very is dead—long live very.
Acknowledgements
We gratefully acknowledge the support of the Social Sciences and Humanities Research Council of Canada to the third author (Tagliamonte, Reference Tagliamonte2018–Reference Tagliamonte2024). We thank the students of LIN351: Sociolinguistic Patterns in Language between 2016 and 2022 for their participation in creating the dataset this research is based on; research assistant Anissa Baird for the Python script; Bridget Jankowski for data management; Erin Hall, Lex Konnelly, Katharina Pabst, Pocholo Umbal, Fiona Wilson, and Kaleigh Woolford for assistance with LIN351 and input on the project; Laurestine Bradford, Matt Hunt Gardner, Lex Konnelly, and Bryce McCleary for helpful discussion; and our audiences at Change and Variation in Canada 11 (St. John’s, Newfoundland and Labrador, Canada, 2019) and the Annual Meeting of the American Dialect Society (New Orleans, LA, USA, 2020).
Competing interests
The authors declare none.
Appendix
Table A1. Distribution of individual intensifiers in our data
Open access
