And obviously age, obviously old people and young people speak differently. (London, Evonne, 33, b. 1975)
Over the past sixty years, quantitative sociolinguistic research has sought to understand the processes underlying language variation and change, amassing a significant body of work whose empirical foundations have established a number of largely consistent broad scale principles governing change across time and space. Across time, an S-shaped curve is observed (Blythe & Croft, Reference Blythe and Croft2012; Denison, Reference Denison and Hickey2003; Kroch, Reference Kroch1989; Labov, Reference Labov1994), where at the beginning of a change only a few speakers use the incoming form or construction, followed by a middle stage when the change gains ground rapidly among the speakers, and a slower finishing phase in which the vast majority of users have adopted the new form. This S-shaped trajectory has been demonstrated for many types of changes: phonetic (Labov, Reference Labov1994:67), lexical (Chambers, Reference Chambers1995), morphosyntactic (Nevalainen, Reference Nevalainen2015), and discourse-pragmatic (Tagliamonte & D'Arcy, Reference Tagliamonte and D'Arcy2007), to the point where it is now regarded as a “template for change” (Chambers, Reference Chambers, Chambers, Trudgill and Schilling-Estes2002:361). Across space, language change is shown to originate in areas of cultural or political dominance (focal areas), subsequently diffusing outward (e.g., Britain, Reference Britain, Ammon, Dittmar, Mattheier and Trudgill2004; Bynon, Reference Bynon1977:214; McMahon, Reference McMahon1994:229; Weinreich, Reference Weinreich1953/1968:155). A number of models exist that predict how changes take place (e.g., Bailey, Wikle, Tillery, & Sand, Reference Bailey, Wikle, Tillery and Sand1993; Bloomfield, Reference Bloomfield1933; Boberg, Reference Boberg2000; Britain, Reference Britain, Chambers, Trudgill and Schilling-Estes2002; Labov, Reference Labov, Britain and Cheshire2003; Trudgill, Reference Trudgill1974b), but despite differences in detail, in “normal” cases of language change (i.e., those which are not brought about by, for example, large scale migration) (e.g., Otheguy, Zentella, & Livert, Reference Otheguy, Zentella and Livert2007), change proceeds sequentially in geographic space. At the same time, in the process of diffusion, linguistic forms undergoing change are said to “mutate en route” (Britain, Reference Britain, Chambers, Trudgill and Schilling-Estes2002:482) to their new homes (see also Labov, Reference Labov2007), developing more localized social and linguistic constraints in the process of change. The evidence arising from studies of linguistic change across time and space has been replicated time and time again, both in the historical record (e.g., Nevalainen & Raumolin-Brunberg, Reference Nevalainen and Raumolin-Brunberg2003) and in the present day (e.g., Bailey et al., Reference Bailey, Wikle, Tillery and Sand1993). Diffusion of innovating forms through time and space follow an orderly sequential pathway of change. These findings provide the empirical foundations for interpreting and explaining how language change evolves in the world. In such cases, the Uniformitarian Principle is invoked where “the forces operating to produce linguistic change today are of the same kind and order of magnitude as those which operated five or ten thousand years ago” (Labov, Reference Labov1972:275) or that “the (global, cross-linguistic) likelihood of any linguistic state of affairs (structure, inventory, process, etc.) has always been roughly the same as it is now” (Lass, Reference Lass1997:28).
However, more recently, Trudgill has speculated that change may not proceed in the same way as other times in human history:
Sociolinguistic typology shows us that there is one very important respect in which the present is not like the past at all. This has to do with the enormously rapid development of transport and communications facilities in the past 150 years – but even more importantly, it has to do with demography, and, as a consequence, social network structure […] the dominant standard modern languages in the world today are likely to be seriously atypical of how languages have been for nearly all of human history. (Reference Trudgill2011:211)
Indeed, increasingly globalized communication networks in the modern world may have profound influences on the traditional patterns of linguistic change through time and space, such that language evolution may be starting to follow quite different pathways. For example, Tagliamonte and D'Arcy's (Reference Tagliamonte, D'Arcy and Rodriguez Louro2016) study of the rise of be like in the quotative system shows that this innovation “increases at the same time, in the same cohort, in all locales” (Tagliamonte, D'Arcy, & Rodriguez Louro, Reference Tagliamonte, D'Arcy and Rodriguez Louro2016:831). Be like displays “no evidence for incremental geographic spread” but instead has “emerged full-blown at the same juncture in time across a great distance” (Tagliamonte, D'Arcy, & Rodriguez Louro, Reference Tagliamonte, D'Arcy and Rodriguez Louro2016:834). In other words, the change accelerated simultaneously in different places. Moreover, in contrast to what is found in cases of diffusion, Tagliamonte et al. (Reference Tagliamonte, D'Arcy and Rodriguez Louro2016:841) found an “attendant suite of parallel internal linguistic constraints” across all communities. There is no evidence for mutation en route because there is no route. According to the authors, “such a development is unprecedented in the recorded history of language change” (Tagliamonte et al., Reference Tagliamonte, D'Arcy and Rodriguez Louro2016:825) and suggest that this may be one of many in a suite of yet uncovered “mega trends” (Tagliamonte & Hudson, Reference Tagliamonte and Hudson1999:168) in language change in the twenty-first century.
A possible kindred of this suite is found within sentential adverbs that encode evidentiality, meanings that allude to an external source of evidence to support the speaker's statement (e.g., Saeed, Reference Saeed2016; Simon-Vandenbergen & Aijmer, Reference Simon-Vandenbergen and Aijmer2007) as in (1):
-
(1)
a. I've done tarot cards. Didn't work, obviously.
b. Years ago, of course, they didn't run cars in the winter.
c. When the mines were going to shut down, naturally the guys—the younger guys—said, “Well, I'm out of here.”
d. If they thought they were fooling the little kids into thinking there was one Mickey Mouse because clearly there wasn't.
e. Well, Mother knew we weren't because evidently you could hear us when we were around.
The use of these adverbs of evidentiality as a cohesive group is said to have increased fourfold since Old English (Swan, Reference Swan1986:2). This is demonstrated in the rising use of obviously and of course in English fiction since 1800, using Google NGrams:Footnote 1
Figure 1 shows a conspicuous rise in use of these adverbs of evidentiality over the past two hundred years in writing (English fiction). However, the trajectory of change may be quite different in spoken data, where, over the course of the twentieth century, the system shows signs of undergoing a dramatic system-internal reorganization. Even the most casual observation suggests that obviously is the adverb of choice in everyday speech: its sheer frequency is notable. It is also singled out for overt (most often negative) commentary in speech, as demonstrated in the extracts from a number of online sources (2–5):
(2) Reddit
Using the word “obviously” unironically is a very quick way to look like a jackass. Obviously….
(3) Manager Tools
Can someone just explain to me why some people, especially managers at meetings, overuse the word obviously.
https://www.manager-tools.com/forums/saying-obviously-too-often
(4) Atomic Object
“Obviously” is a purely destructive word
https://spin.atomicobject.com/2014/09/09/never-use-the-word-obviously/
(5) CV Library
Ten buzzwords not to say if you want a promotion.
1. Obviously A whopping 49.8% of employers said that they were fed up with hearing employees say this annoying buzzword. That's nearly half! While you might think that something is “obvious,” your employee may not agree. Not to mention, it is just plain rude. https://www.cv-library.co.uk/career-advice/work-life/10-buzzwords-not-to-say-if-you want-a-promotion/
As testament to the (at least perceived) ubiquity of this particular evidential adverb, it has even developed a popular clipped variant, as in (6–7), which may be indicative of its status as a stereotype (Labov, Reference Labov2001:196):
(6) The Guardian Newspaper 21 May 2015
As world Scrabble champion, I think new words are obvs lolz
https://www.theguardian.com/commentisfree/2015/may/21/world-scrabble-champion- new-words-lolz-collins-official-words
(7) Urban Dictionary
Obvs
shortened version of the word obviously for the lazy people who refuse to say the whole thing
“Hey are you going to the homecoming dance tomorrow?”
“Uhm. obvs!”
“Cool!”
Figure 1. Frequency of obviously and of course in English fiction using Google NGrams.
The frequent use of obviously in spoken data, coupled with intense social media meta-awareness of this form, points toward a potentially rapid reorganization within the set of adverbs of evidentiality. This offers the analyst an ideal opportunity to test the nature of a recently developing linguistic change and explore the possibility of it being a “mega trend” (Tagliamonte & Hudson, Reference Tagliamonte and Hudson1999:168) or even another “black swan” as anticipated by Tagliamonte et al. (Reference Tagliamonte, D'Arcy and Rodriguez Louro2016:843). Corpora that allow for scrutiny of language patterns in both time and geographic space in the last one hundred years would make it possible to determine what type of change a recently evolving linguistic phenomenon might be. Does it follow the more traditional pathways of language change in a sequential spread through time and space, or is it more in line with a “simultaneous, instantaneous, parallel development, in multiple urban locations, at the same time” (Tagliamonte et al., Reference Tagliamonte, D'Arcy and Rodriguez Louro2016:842)? To this end, this study employs several large archives of informal, vernacular spoken language that represent the key contrasts of geographic location and community size across two large territories (the UK and Ontario, Canada) comprising individuals born from the late 1800s into the early 2000s, the cusp of the twenty-first century. The study targets a linguistic phenomenon known to be in flux across this period of history: sentential adverbs of evidentiality.
SENTENTIAL ADVERBS, AND ADVERBS OF EVIDENTIALITY
Sentential adverbs in general are said to have three characteristics: they are speaker oriented; they have sentential scope; they have clausal properties of a reduced sentence, for example, apparently ‘it is apparent that…’ (Swan, Reference Swan1988:29). Within the broad category of sentential adverbs, a number of semantic subcategories have been proposed (e.g., González-Álvarez, Reference González-Álvarez1996:219–29; Ifantidou, Reference Ifantidou2001:97–9; Saeed, Reference Saeed2016; Swan, Reference Swan1988:30–77), including evaluative adverbs (predictably, remarkably), modal adverbs (certainly, probably), and speech act adverbs (bluntly, precisely).
While there are many different labels for sentential adverbs and highly gradient semantic nuances among them, we focus here on what we refer to as the adverbs of evidentiality, where “evidentiality is first and foremost to do with the indication of source of information, on which basis speakers voice their claims to ‘knowledge’” (Hoye, Reference Hoye and Romero-Trillo2008:157; see also Saeed, Reference Saeed2016). These adverbs are considered part of a larger set of modal adverbs: actually, certainly, clearly, obviously, maybe, perhaps, and probably (Payne, Huddleston, & Pullum, Reference Payne, Huddleston and Pullum2010:71) and comprise some or all of the following set: apparently, clearly, evidently, obviously, and of course.Footnote 2 These adverbs of evidentiality are known to be among the most frequently used in corpus-based studies (Swan, Reference Swan1986, Reference Swan1988) and have a long history in English, as detailed below.
HISTORICAL PERSPECTIVE
The emergence and subsequent development of adverbs of evidentiality have been investigated in-depth in a number of studies (e.g., Lenker, Reference Lenker2010; Lewis, Reference Lewis2002). Although debate remains about when exactly the various adverbs entered the language, the Oxford English Dictionary (OED) provides indicative dates. Naturally, is first citedFootnote 3 in the fifteenth century (8a), with evidently and clearly appearing in the sixteenth century (8b-c). Obviously, which derives from the adverbial use of “in a clear perceptible manner,” is first cited with the evidential meaning of “evidently, plainly, manifestly” in the seventeenth century (8d). Of course, as an adverbial construction, had an earlier use of “in ordinary or due course according to the customary order, as a natural result” that is attested as early as 1542; however, it is first cited with the meaning of “naturally, as will be expected in the circumstances; for obvious reasons, obviously” a full 150 years later, with a first citation in 1823, (8e).
-
(8)
a. c1425 J. Lydgate Troyyes Bk. (Augustus A.iv) ii. 6832 (MED) Naturelly no man schal desyre Of his enmye þe helthe nor welfare.
b. 1532 T. Abell Inuicta Veritas sig. Fv Thies persons opinion & sayng..ys euidently false.
c. 1569 R. Grafton Chron. II. 287 But Britaine was clerely excepted.
d. 1668 M. Hale Pref. Rolle's Abridgm. sig. b ij Other matters more obviously deducible by Argumentation.
e. 1823 J. D. Hunter Mem. Captivity 39 She made some very particular inquiries about my people, which, of course, I was unable to answer.
These timelines and associations establish a history and layering of form and function within this set of evidentials over a number of centuries that sets the context for a number of previous studies, both historical and contemporary.
PREVIOUS RESEARCH
The study of sentential adverbs has been undertaken from a variety of different linguistic perspectives, including semantics (e.g., Bellert, Reference Bellert1977), syntax (e.g., Jackendoff, Reference Jackendoff1972), and pragmatics (e.g., Hoye, Reference Hoye and Romero-Trillo2008). Sentential adverbs are considered extremely complex due to their nuanced meanings expressing distinctions of modality, evaluation, and illocutionary force, making them what some scholars refer to as “the most nebulous and puzzling” lexical category in English, and one that “does not fit the definitions for other word classes” (Quirk, Greenbaum, Leech, & Svartvik, Reference Quirk, Greenbaum, Leech and Svartvik1985:438). Trask (Reference Trask1993:251) further described sentential adverbs as “a lexical item typically having the approximate distribution, and often the form, of an adverb, but behaving not as a modifier of the verb or VP but as an operator upon the entire proposition.” In Jackendoff's (Reference Jackendoff1972:76) framework, sentential adverbs are categorized as “speaker-oriented.” As such, they are generally treated not as truth conditional indicators but instead within the domain of the speech act, making them semantically external to the proposition. In Traugott's (Reference Traugott1989) framework, they are integrally involved in ongoing processes of subjectification in which there is a development toward the encoding of “speaker informativeness about his or her attitude … [and a] strengthening of focus on knowledge, belief, and the speaker's attitude toward the proposition” (Traugott, Reference Traugott1989:49).
Some studies focus on the trajectory of these adverbs in historical written texts (e.g., Lenker, Reference Lenker2010; Swan, Reference Swan1986, Reference Swan1988). Others are based on their use in different written registers, such as scientific writing (e.g., Barbaresi, Reference Barbaresi1987). Many studies tend to concentrate on one adverb or the other, for example, of course (Holmes, Reference Holmes1988; Lewis, Reference Lewis2002; Simon-Vandenbergen & Aijmer, Reference Simon-Vandenbergen and Aijmer2007; Wichmann, Simon-Vandenbergen, & Aijmer, Reference Wichmann, Simon-Vandenbergen, Aijmer, Cuyckens, Davidse and Vandelanotte2010), surely (Downing, Reference Downing2001), obviously versus clearly (Kang, Reference Kang2017). A few studies target multiple adverbs of the same class, for example Suzuki (Reference Suzuki2016). Simon-Vandebergen and Aijmer (Reference Simon-Vandenbergen and Aijmer2007) offered a rare comparison of over twenty adverbs of "certainty” in both spoken and written data.
SENTENTIAL ADVERBS AND EVIDENTIALS IN PARTICULAR
The broad category of sentential adverbs is known to have undergone a dramatic development as a type of adverb across the history of English. For example, working with texts from Old English, Middle English, Modern English (1500–1900), and Present Day English (1900–2000), Swan (Reference Swan1986, Reference Swan1988) showed that “truth intensifiers” truly, indeed, “evaluative adverbs” wonderfully, horribly, and “speech act” adverbs, frankly, briefly, steadily increased from Old English to Present Day English. Of these, the evidentials, what she refers to as “truth intensifiers”—those used by speakers “to convey that he or she has evidence, often perception, or to draw conclusions from evidence” (Mitchell, Reference Mitchell1976; Swan, Reference Swan1988:3)—are the most frequent class, rising from 292 tokens in OE to 1336 in PDE (in texts of similar length) (Swan, Reference Swan1988: Table 2).Footnote 4
A great deal of attention has been directed to the position of the adverb relative to the clause structure as a means of determining function. For example, Aijmer (Reference Aijmer, Nevalainen, Taavitsainen, Pahta and Korhonen2007) claimed that evidential adverbs “can be inserted in practically any position in the sentence”; however, when they become specialized to particular syntactic slots, they may become conventionalized to specific pragmatic functions. More specifically, Aijmer (Reference Aijmer, Nevalainen, Taavitsainen, Pahta and Korhonen2007:65) suggested that clause final obviously is used for “affective meaning and solidarity” (e.g., I'll help you with that, obviously) whereas initial position functioned as “imposition and emphasis” (e.g., Obviously, we don't want to spend too much money).
Given the striking increase in use of evidential adverbs as a type in the history of English, and the possibility of using linguistic criteria to distinguish form and function, our goal in the present study is to contribute to the understanding of this “most nebulous and puzzling” lexical category (Quirk et al., Reference Quirk, Greenbaum, Leech and Svartvik1985:438) and to add new information to the question of types of linguistic change more generally.
DATA AND METHOD
The data come from a unique compendium of vernacular speech data from research projects in the United Kingdom and Canada over the past thirty years.Footnote 5 The first set of corpora was collected between 1997 and 2013 across the UK (Smith, Reference Smith2000 et seq.; Tagliamonte, Reference Tagliamonte1996–1998 et seq.). The UK materials extend from Devon in the southwest to Shetland in the far north. The second set of corpora was collected in Canada between 2001 and 2017 (Tagliamonte, Reference Tagliamonte2003–2006 et seq.). The Canadian materials extend from Toronto in the south to Kapuskasing in the north. Across both countries, the data include both rural and urban communities, from small villages to larger metropoles. Maps indicating the different locations of the communities are shown in Figure 2. All of these materials were collected using standard sociolinguistic fieldwork methods and techniques using the sociolinguistic interview described extensively in the literature (e.g., Labov, Reference Labov1970). In addition to the corpora collected by the authors, we also include two other large corpora. Data from urban London, England come from a series of projects by Cheshire, Kerswill, and associates (Cheshire, Fox, Kerswill, Khan, & Torgersen, Reference Cheshire, Fox, Kerswill, Khan and Torgersen2007–2010; Kerswill & Cheshire, Reference Kerswill and Cheshire2004–2007), and data from The Freiburg English Dialect Corpus (FRED) come from early dialectological surveys or folklore projects in the UK (Anderwald & Wagner, Reference Anderwald, Wagner, Beal, Corrigan and Moisl2005). In terms of corpus size, the Canadian materials comprise approximately fourteen million words, while all the UK materials combined total approximately nine million. To our knowledge, the totality of these materials in a single study represents one of the largest bodies of spoken vernacular English to be subjected to variationist analysis to date.
Figure 2. Locations of the UK and Canadian data.
Circumscribing the variable context
The first challenge for this study was to circumscribe the set of adverbs to be included in the study. We considered the evidence from the OED as well as definitions from earlier research on evidentiality in order to identify the relevant forms. Adverbs of evidentiality are often quite broadly defined and can be “vague and underspecified” and “pragmatically polysemous” (Simon-Vandenbergen & Aijmer, Reference Simon-Vandenbergen, Aijmer, Davidse and Hayvaert2004:28). Furthermore, the meaning of an adverb may differ depending on their syntactic slot in the clause (Aijmer, Reference Aijmer, Nevalainen, Taavitsainen, Pahta and Korhonen2007; Lewis, Reference Lewis2000). This has implications for the present analysis, given that functional equivalence is a key tenet of the variationist paradigm. We thus relied on the concept of “weak complementarity” (Sankoff & Thibault, Reference Sankoff, Thibault, Johns and Strong1981) in justifying our overall method. According to Sankoff and Thibault (Reference Sankoff, Thibault, Johns and Strong1981), the relevant criteria are: (1) that the variants “must serve one or more generally similar discourse functions (Sankoff & Thibault, Reference Sankoff, Thibault, Johns and Strong1981:213); and (2) that “one form appears to be replacing the other, either in time or along some socioeconomic or demographic dimensions in the speech community” (Sankoff & Thibault, Reference Sankoff, Thibault, Johns and Strong1981:215). In our data, the adverbs used for evidential meaning are: obviously, of course (and the variants of of course in spoken data),Footnote 6 clearly, naturally, and evidently, as shown in (1).Footnote 7 Manner adverbials, as in (9), were excluded.
-
(9)
a. Each year we'd put a new building up and it was all naturally ventilated. (UK, M, b. 1967)
b. I can mind on that pretty clearly. (UK, M, b. 1937)
In most cases, it was straightforward to distinguish between evidential and manner uses. However, Simon-Vandenbergen and Aijmer (Reference Simon-Vandenbergen and Aijmer2007:165) noted a number of ambiguous cases in their data, including (10):
-
(10)
a. Mercedes have always been at the forefront of luxury car design and this car is clearly aimed at keeping them there. (Simon-Vandenbergen & Aijmer, Reference Simon-Vandenbergen and Aijmer2007:165)
Such an example could be interpreted as “in a very clear way,” that is, manner adverbial or “it is clear that,” that is, evidential (Simon-Vandenbergen & Aijmer, Reference Simon-Vandenbergen and Aijmer2007). Such ambiguity undoubtedly arises from the development of evidentials from manner adverbs into sentential adverbs, with transition zones resulting in ambiguity (e.g., Buchstaller & Traugott, Reference Buchstaller and Traugott2006; Traugott, Reference Traugott, Beeching and Detges2014). Our data also contained examples that may be ambiguous, as in (11):
-
(11)
a. It was obviously a male-based class. (UK, M, born 1995)
b. If you're true to yourself then you'll naturally end up meeting people that you would like. (CAN, F, born 1984)
In 11b, for example, naturally could be interpreted as “in the natural course of events,” that is, a manner adverbial, or “it is clear that,” that is, evidential. In such cases, we erred on the side of caution and removed ambiguous examples from the dataset. After these exclusions, we were left with 8,089 tokens in the total data set, 3,726 in Canada and 4,363 in the UK.
To study the possibility of linguistic change in progress in these synchronic data, we use two proxies for time depth. On the assumption that language change progresses across space according to the size and location of communities (e.g., Trudgill's [Reference Trudgill1974a] Gravity Model), we make use of urban and more outlying locations in the two countries. For example, in the UK, the small villages and towns in peripheral locations can be expected to lag behind cities such as London. Similarly, in Canada, small villages and towns in peripheral locations can be expected to lag behind Toronto, and this has already been demonstrated in recent work (e.g., Jankowski & Tagliamonte, Reference Jankowski and Tagliamonte2019). On the assumption that language change can be studied among people of different ages (i.e., apparent time [Bailey, Tillery, & Sand, Reference Bailey, Wikle, Tillery and Sand1991]), we will use the individuals of different generations in the communities whose dates of birth span over one hundred years (1890s to 1990s). Finally, we will employ variationist sociolinguistics methods (Labov, Reference Labov1963; Weinreich, Labov, & Herzog, Reference Weinreich, Labov, Herzog, Lehmann and Malkiel1968 et seq.) to examine complex patterns of language use. In this case, we will exploit both the frequency of forms as well as the underlying social and linguistic constraints on their competition. This will expose not only the social and geographic underpinnings of the variation among the adverb cohort but also enable us to infer the type of linguistic change in progress. Layering—the co-existence of multiple forms “within a broad functional domain” (Hopper, Reference Hopper, Traugott and Heine1991:22)—between the main variants will be the foundation of our analysis. We begin with systematic study of the data, using comparison of marginals and cross-tabulation to arrive at a detailed understanding of the patterns inherent in the data set. We employ conditional inference trees (e.g., Strobl, Malley, & Tutz, Reference Strobl, Malley and Tutz2009) to uncover how the most important predictors work together (see Tagliamonte & Baayen, Reference Tagliamonte and Baayen2012). Finally, we put to work the tools of statistical modeling, using a series of mixed effects generalized linear models to evaluate the significance of patterns and interactions across competing contextual constraints (e.g., Baayen, Reference Baayen2008). Interpretation of the results of the statistical models compared across communities and across generations (see, e.g., Tagliamonte, Reference Tagliamonte, Chambers, Trudgill and Schilling-Estes2002) will be used to help us situate the change in time and space.
RESULTS
The question of rapid acceleration of obviously can only be determined by scrutinizing the relative frequencies of the lexical adverbs in the set.Footnote 8
Table 1 shows the contemporary distribution of layering within the adverbs of evidentiality. The overarching pattern is parallel in both locales: of course makes up the lion's share of the data in both varieties (84% in Canada and 79% in the UK). Obviously is the next most frequent form (15% in Canada and 19% in the UK). Use of clearly, evidently, and naturally is very sparse. However, these data comprise speakers of different ages and from different communities within the countries, thus it is important to investigate the use of adverbs of evidentiality along these parameters.
Table 1. Overall distribution of evidential adverbs in Canada and the UK (Total n = 12072)
Through the decades
The next step is to investigate the possibility of generational change, using the individual's date of birth as a proxy, providing a view from apparent time. Figures 3 and 4 provide a striking perspective of the frequency of the main forms by birthdate of the individuals in each country, Canada and the UK.
Figure 3. Overall distribution of evidential adverbs by date of birth in the UK.
Figure 4. Overall distribution of evidential adverbs by date of birth in Canada.
The figures show a distinct trajectory of change in the use of these forms. Naturally, clearly, and evidently are very infrequent, existing at very low rates across the board. The competition in adverbs of evidentiality is centered on obviously and of course. While of course declines, obviously increases incrementally and precipitately through the decades with obviously the clear “winner” among individuals born after the 1980s.Footnote 9 In both the UK and Canada, the use of obviously begins to accelerate in the 1920s, despite the fact that the OED's earliest citation of obviously is about three hundred years earlier, in the mid-1600s (examples 8d-e). Note, however, that, in the UK, the changeover from of course to obviously is virtually categorical, while, in Canada, of course remains a competitor. In other words, obviously accelerates more slowly in Canada. Nevertheless, the trajectory in both countries has visible stages of development: an early phrase up to about 1930, a middle phase of acceleration up to about 1960, then after 1980, and a final phase after 1990.
We can now look in more detail at the progress of this change through the extensive community-based samples within each of the two countries. Figure 5 shows the rise of obviously when the data are divided by the different communities in each region. The frequency of obviously as opposed to all other evidential adverbs (mostly of course variants) is represented by ‘obviously’ on the Y axis and the date of birth of the individuals on the X axis.Footnote 10 Note that we have combined data in geographically contiguous areas where data were sparse: “small north” refers to a collection of rural communities in Northern England, and “small south” refers to communities in more southerly areas of England. “SCOTLAND” combines a number of smaller datasets in that country, with larger datasets (BUCKIE, SHETLAND) kept separate. In the Canadian data, “SMALL NORTH” and “SMALL SOUTH” refer to small communities in northern and southern Ontario. The two peaks in the right-hand figure are the error bars for certain communities where there is scant data in the corresponding decade.
Figure 5. Distribution of evidential adverbs by date of birth in the Canada and the UK.
Figure 5 exposes the rapid, but highly consistent, pattern of rocketing use of obviously across both countries and across different regions and types of communities within each nation. What we might expect to see, if obviously were diffusing through space, is that the large urban cities in Canada and the UK—Toronto and London—would lead in the change, with other areas lagging behind. Instead, Figure 5 shows that these communities move in parallel in their rapid uptake of obviously. These patterns suggest that traditional models of language change via diffusion are not fully explanatory for this linguistic change.
Sex
Figure 6 shows the rise of obviously by gender in Canada and UK, here represented as a contrast between perceived gender, male (“M”) and female (“F”). The proportion of obviously tokens is shown as a binary contrast between obviously and all other forms.
Figure 6. Obviously across gender in Canada and the UK.
In contrast to most studies of linguistic change in progress, which show a strong lead by women—whether change from above or change from below—in both Canada and the UK, males and females rise in tandem in use of obviously with neither group in the lead until a very slight advantage by women in the last two decades of the twentieth century in Canada.
Internal constraints
A common predictor in the advancement of linguistic change in progress, particularly with features in the discourse pragmatic domain, is the position of the form in the clause structure. Indeed, it has been suggested that “everything that happens to the meaning of a gram happens because of the contexts in which it is used” (Bybee, Perkins, & Pagliuca, Reference Bybee, Perkins and Pagliuca1994:297). Accordingly, many corpus-based studies have used this contrast in their analyses (D'Arcy, Reference D'Arcy2017; Denis, Reference Denis2015; Denis & Tagliamonte, Reference Denis, Tagliamonte and Pichler2016; Traugott & Dasher, Reference Traugott and Dasher2002). In order to test the effect of syntactic structure in the progress of change, we distinguish four positions for the evidential adverbs: clause initial (12a), clause final (12b), clause internal (12c), and what we call “standalones” (12d).Footnote 11
-
(12)
a. And obviously, they knew something about medicine, so they were both chosen to go into the medical field.
b. Oh yes, he picked up, yeah, all the Devon words, of course
c. The other excitement, of course, was the spies that were picked
d. Well, yes. Obviously.
Figure 7 partitions the data into these four categories in Canada and the UK: final (“F”), initial (“I”), middle (“M”), and standalone (“S”). Once again, the proportion of obviously tokens is shown as a binary contrast between obviously and all other forms and the date of birth of the individuals, “date_of_birth,” is shown on the X axis.
Figure 7. Obviously by syntactic position in Canada and the UK.
Figure 7 reveals parallel trajectories for the use of obviously across the four syntactic positions where there is sufficient data to tell. The evidential adverb obviously is most frequent in medial position, then final position, then initial position in both regions for most of its trajectory.Footnote 12 Further, this pattern stays constant through most of the S-curve trajectory of the change. Only the standalone cases in the UK look different. However, these contexts are very infrequent in the data (n = 139), with the error bars indicating that the trajectory is not reliable. The main difference between Canada and the UK is that obviously has fully taken over the system in the UK, and all the contexts have converged by the end of the twentieth century.
Statistical modeling
The next step is to seek validation of a systematic change in progress within the linguistic system, using statistical modeling. The patterns visible in the foregoing factor by factor analyses must be assessed for statistical significance when all the relevant factors are considered simultaneously and assessed over and above the effect of the individuals in the sample. The analysis in Table 2a was based on 6,768 observations of the dependent variable from 649 individuals. The analysis in Table 2b was based on 4,851 observations of the dependent variable from 698 individuals. Each of the independent predictors is shown with the number of observations per cell and the frequency of obviously out of the total number of observations. The mixed effects model was performed using the statistical package R using the lme4 package (Baayen, Reference Baayen2008) and the “bobyqa” optimizer with individual included as a random effect. The reference level is set so that the model predicts the incoming favored variant, obviously.
Table 2a. Mixed effects model of the adverbs of evidentiality in Canada
Significance. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Table 2b. Mixed effects model of the adverbs of evidentiality in the UK
Significance codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Even with the large number of individuals and observations in the sample, this was not sufficient to model date of birth as continuous, even when the data from both countries was run together. Any model including date of birth as continuous failed to converge and produced very large eigenvalues. To obviate this issue, we used conditional inference tree technique using the party package in R (R Core Team, 2007) to determine relevant junctures in the date of birth continuum. Despite the slightly different points of acceleration in the two countries (see Figures 2 and 3), the main juncture points in the trajectory of change are comparable except in the later stages of the change as obviously picks up speed. Nevertheless, due to these differences between countries, we model them separately to expose the influence of the main urban center and syntactic position in each local. No model led to a statistically significant effect of gender, so it was removed from the final model. Standalone contexts were removed due to small numbers.
Tables 2a and 2b show that date of birth and syntactic position are significant in both countries and with a parallel incrementation; however, the difference between the main urban center of each region (London and Toronto) compared to the outlying areas is only significant in Ontario, Canada. The extreme acceleration of this change in the last decades of the twentieth century in the UK trumps regional differences. In contrast, in Ontario, the change is moving comparatively slower, and outlying regions lag behind. This geographic pattern is consistent with other current changes in progress in the Canadian context (Franco & Tagliamonte, Reference Franco and Tagliamonte2020; Reference Franco and Tagliamonte2021; Jankowski & Tagliamonte, Reference Jankowski and Tagliamonte2019; Tagliamonte & Jankowski, Reference Tagliamonte and Jankowski2018).
Testing the difference between Canada and the UK with respect to change among the cohort of adverbs used for evidentiality reveals that obviously began to increase in use at approximately the same point in time in both varieties—the early twentieth century (see Figures 3 and 4) and has been accelerating ever since. In the UK, obviously has risen to the point of having taken over the system, and the differences among syntactic contexts has leveled. In Canada, the visible stepwise progress of the constant rate of change is lagging slightly behind (see Figure 7). The greater use of obviously in the UK and a faster acceleration may be due to any number of reasons; however, the internal patterns of use tap into system level mechanisms of language organization. What implications do these results have for understanding each of the foundational hypotheses about the progression of language variation and change reviewed earlier? Further, what do they suggest for continuing to assume that the Uniformitarian Principle holds in the twenty-first century, just as it always has done, and, if not, what are the implications for linguistic change in the future?
DISCUSSION
Over the past sixty years, quantitative sociolinguistic research has sought to understand the processes underlying language change in synchronic data. As noted in the introduction, a number of key principles have emerged from this research that are shared across most, if not all, studies of language change. Specifically, diffusion of innovating forms follows a sequential pathway of change, moving through time and space in an orderly S-shaped curve. More recent work suggests that so-called “mega-trends,” that is, those innovations that are widespread throughout the English-speaking world, may defy those well-established principles in that some changes are starting to happen simultaneously (Tagliamonte et al., Reference Tagliamonte, D'Arcy and Rodriguez Louro2016). This, in turn, may have a fundamental influence on how language change proceeds. Our analysis of the rise of obviously as an adverb of evidentiality across one hundred years in recent history and two major varieties of English allowed us to assess that possibility.
First, we have documented a rapid change taking place in this system. While apparently, evidently, naturally, and clearly were stable over the twentieth century, the focus of change in the system involves competition between of course and obviously, with the latter usurping the former in speech data. Examination of the rise of obviously across the UK and Canada (Figures 3 and 4) revealed that, in two these countries, obviously took hold in the system at approximately the same time. This suggests near simultaneous innovation in two widely separated geographic areas, a change at odds with the classic tenets of diffusion, where innovations move from place to place in an ordered succession. We also note that obviously had entered the grammar around the seventeenth century, making the simultaneous take-up of this form some four hundred years later even more intriguing. Once obviously begins to escalate, it does so very rapidly: within the space of fifty years, it moves from extremely marginal use to the majority variant in both countries. Our analysis further shows that this rapid rate of change continues to (near) completion in the UK, where of course in highly vernacular spoken data is virtually obsolete. Canada looks set to continue the same trajectory, but at a slightly slower pace (Figures 3, 4, 5, 6, and 7).
Examination of the patterned use of obviously across geographic space within each country provides further details of the progress of this change. Figure 5 demonstrated that the S-shaped curve was replicated across all communities within the UK and Canada, and, perhaps more surprisingly, each community, regardless of size or geographic location, moved in lockstep in the rise of obviously. These patterns of use in the descriptive statistics were confirmed in the mixed effects model for the UK (Table 2b) where urban versus outlying was not selected as significant. However, for Canada (Table 2a), the urban/outlying contrast was selected as significant, with obviously in the large urban conurbation of Toronto in advance. These results suggest that this change initiates near simultaneously in both the UK and Canada, but with slightly different trajectories across geographic space once the change is in progress. In Canada, diffusion of this form seems to be following more traditional pathways of sequential spread, as per Trudgill's Gravity Model. In contrast, we find no apparent incremental diffusion within regions in the UK: obviously moves concurrently across the landscape in each region. Differences in “sociolinguistic typology” (Trudgill, Reference Trudgill2011:211), including demography and social network structure, may account for these results.
Analysis of the remaining social and contextual factors showed strikingly parallel constraints on use of this rapidly expanding innovation. In each case, the classic S-shaped curve is visible in the apparent time continuum, confirming that the course of this development follows in the same trajectory as many other linguistic changes in progress. The S-shape curve is also evident by gender, as demonstrated in Figure 6. With regard to Principle 3, that women will lead in the adoption of incoming forms, the social embedding of this change is contrary to expectation: there are no gender leaders. When the data is partitioned by male/female, both groups participate equally.
When we examine linguistic constraints in the progress of change, the mixed effects model showed that, across both countries, syntactic position was significant. Crucially, Figure 7 shows that the rise of obviously moves in tandem across a number of different syntactic slots, both within and between the UK and Canada, lending further support to the fact that obviously advances in parallel across all linguistic contexts. This pattern is indicative of the constant rate effect (Kroch, Reference Kroch1989), which suggests a single underlying change in the linguistic system. Perhaps more importantly, these results are in line with other rapid linguistic developments that show parallel internal linguistic constraints (Tagliamonte et al., Reference Tagliamonte, D'Arcy and Rodriguez Louro2016:841). In addition, we note that, in the UK, where the trajectory of change is more advanced than in Canada, the different syntactic slots merge at the end of the S-shaped curve. This suggests that, regardless of potential meaning differences according to syntactic slot (e.g., Aijmer, Reference Aijmer, Nevalainen, Taavitsainen, Pahta and Korhonen2007), the innovating form infiltrates all contexts as it nears completion.
In sum, the adverbs of evidentiality exhibit all the hallmarks of a systematic change in progress, which embody both traditional models of language change and a novel organizational design: (1) an S-curve developmental trajectory; and (2) a constant rate of contextual constraints; but (3) no binary gender effect; and (4) near simultaneous advancement. The broad geographic and generational perspective from individuals born from the late nineteenth and across the twentieth century permits affirmation of key attributes that are beginning to emerge from the study of accelerating, innovating features in the late twentieth century. They are essentially “off the shelf” changes (Milroy, Reference Milroy, Cain and Russom2007): easily borrowed and receptive to global trends but exhibiting regular and constant patterns as the change progresses. Thus, with respect to the Uniformitarian Principle, in some respects the “present is not like the past at all” (Trudgill Reference Trudgill2011:211). Whether this new combination of characteristics for a linguistic change in progress is due to the nature of variation within this adverb cohort or is the result of recent and ongoing sociocultural change, and that we have actually tapped a new linguistic black swan (Tagliamonte et al., Reference Tagliamonte, D'Arcy and Rodriguez Louro2016), presents an exciting avenue for future research. Emerging models of grammaticalization that offer new methods for documenting generational shifts at key stages of linguistic change (Petré & van de Velde, Reference Petré and van de Velde2018) will be useful in determining whether these types of changes follow the same episodic trajectory. One thing is sure, as we continue to collect representative vernacular speech into the next century, Labov's (Reference Labov1994:11) trope that historical linguistics is the “art of making the best of bad data” and the restriction of synchronic studies to the late nineteenth and twentieth century, will become a thing of the past.
ACKNOWLEDGMENTS
We gratefully acknowledge the support of the Social Sciences and Humanities Research Council of Canada for grants to Tagliamonte and the Economic and Social Science Research Council of the UK for grants to Smith. The extensive data extraction and coding was completed by a skilled team of research assistants in sociolinguistic laboratories in Glasgow and Toronto. We thank Tim Gadanidis, Derek Henderson, and particularly Bridget Jankowski for skilled data wrangling of the extensive corpora. An earlier version of this work was presented at ISLE 6 in London. We thank Peter Trudgill for his helpful comments.
