Exploring the architecture of variable systems to predict language change
Sali A. Tagliamonte
University of Toronto
In this paper I review the results from three changes in progress that offer insights into whether or not we can predict linguistic change: 1) stative possessive have ➜ have got; 2) future temporal reference will/shall ➜ going to and 3) the innovation quotative be like. I argue that absolute predications about the locus of change in the linguistic system and the precise point in time when a change will actuate are impossible. However, the probability of knowing subsequent stages of development in a current change (e.g. increase in frequency) can be estimated given a longitudinal perspective, a known trajectory of constraints as well as attention to social, cultural and economic conditions.
Quantitative study of language variation and change has been an evolving enterprise for nearly fifty years (e.g. Tagliamonte 2015). Research in this area has amassed extensive evidence from a broad range of communities and innumerable linguistic features. The building evidence supports general principles of linguistic change (Labov 1994, 2001, 2010) and ongoing studies attest to their broad applicability (e.g. Labov 2007; Tagliamonte & D’Arcy 2009). Sanchez-Stockhammer (2014) suggests that “a useful forward step is to ask the question: can we make predictions about future developments?”
In order to predict linguistic change, it would be necessary to understand how it happens in the first place. A certain amount of advancement has been made in this endeavour. Labov (2001) proposed a model of linguistic change based on the logistic function (Verhulst 1845) which predicts an S-curve for linguistic change in progress, as in Figure 1 (Labov 2001: 452, Figure 14.1; Tagliamonte & D’Arcy 2009: 58, Figure 1).
Figure 1. The S-curve of linguistic change.
This model is intended to capture the logistic trajectory of change so as to account for the linear increase in frequency of a new form across adult cohorts. In essence, the model predicts that language changes initially proceed slowly, reach a maximum rate at mid-course, and then slow down as they near completion. This trajectory provides a useful foundation for the study of linguistic change in progress.
This theoretical model has been replicated in many studies. Perhaps the most well-studied example is the development of do in English. In the progression of this change, do (also referred to as ‘periphrastic’ do) came to be required as a ‘dummy’ auxiliary in sentences with a tensed main verb innegative declaratives, as in (1) and questions, as in (2). The new structures with do are shown in the (b) examples.
- a. I know not.
b. I do not know.
- a. Know you the answer?
b. Do you know the answer?
Figure 2. The development of do in the history of English (Ellegård 1953).
Figure 2 shows the trajectory of change of the incoming variants with do. The horizontal axis denotes the date (i.e. the year of the written document) and the vertical axis denotes the percentage of usage of do found in the group of documents consulted for that time period. Kroch (1989: 225) demonstrates that when these S-shaped trajectories are modeled with the logistic function, the slopes are identical. His results “support the hypothesis that the slopes of the curves are underlying the same and that the observed differences among them are random fluctuations”.
In order to facilitate the ability to study language change, Variationist Sociolinguistics made an important step forward beginning in the early 1960’s (e.g. Labov 1963) by introducing the study of linguistic change in synchronic data. The way to do it was to use the apparent time construct. Apparent time refers to a procedure such that generational differences at a single point in time are used to make inferences about how a change may have taken place in the (recent) past (e.g. Bailey 2002; Bailey et al. 1991; Sankoff 2006). Another proxy for the temporal progression of a change can come from dialects arrayed in geographic space on a continuum from peripheral to urban. Differences across dialect can be taken to represent different stages of a change on the assumption that linguistic change is taking place at different rates in each community (e.g. Tagliamonte 2003; Tagliamonte 2006b).
A study illustrating the S-curve of linguistic change in synchronic data in Canadian English is shown in Figure 3 (Chambers 1998).
Figure 3. The rise of snuck in Canadian English – 20th century.
Figure 3 shows the rise of the past tense of the verb ‘to sneak’ in Canadian English. This change is charted in apparent time across six regions of Canada from speakers of different ages in the late 20th century. An S-curve in apparent time is visible.  Individuals over 80 years of age at the far left use the form sneaked, as in (3a). In contrast, the youngest generation on the right hand side use snuck, as in (3b). In fact, there were almost no tokens of sneaked in my archive of spoken contemporary Canadian dialects collected in many different communities between 2003 and 2015. 
- a. Well, we’d a mile and a half to walk and it was actually fun. We sneaked in what we call Little White Lake. (Beatrice Gellert, 70, Almonte) 
b. My mother snuck me over ‘cause my father never liked me eating outside the house. (Lucinda Ciroli, 34, Toronto)
The key fact for the purpose of the present discussion is that many linguistic changes progress with the same trajectory — an S-curve.
Another important observation about change in progress is that, at least in certain types of change, there is a constant rate of change across linguistic environments. In the development of do, for example, Ellegård (1953) categorized various contexts in which do occurs: negative declaratives, questions, negative questions, yes-no questions, etc. However, as Figure 2 shows, these types rise in tandem. Kroch (1989: 238) noticed this pattern and suggested a general observation about change in progress: that it “proceeds at the same rate in all contexts”. Thus, although negative questions in Figure 2 are the most favorable to innovating do, there is no privileged or causal reason for it. Nor is it a reflection of underlying forces pushing the change forward. Instead, different frequencies are the result of “functional effects, discourse and processing” (Kroch 1989: 238). In subsequent studies, Kroch and his associates gathered corroborating information from different linguistic features and different languages and found the same pattern — a constant rate of change across contexts (Kroch & Taylor 1997, 2000; Pintzuk 1996; Pintzuk & Kroch 1989; Pintzuk & Taylor 2006; Santorini 1993). This constant rate in each context is the result of a grammar in which two forms or constructions compete and one wins out, e.g. a structure with do and one without compete until one structure is the only available option. Although environmental constraints may be present as the change progresses, these preferences stay stable over time.
The S-curve and the Constant Rate Effect make fairly substantive cumulative findings for predicting the trajectory and internal constraints of linguistic change in progress. Given these consistent findings and proposals for linguistic change, how can a comparative approach, the quantitative techniques of Variationist Sociolinguistics (Tagliamonte 2012; Weinreich, Labov & Herzog 1968) and building evidence from large-scale corpora contribute to predicting language change?
Figure 4 shows a map of ten fieldwork sites in the UK, from the Shetland Islands to Southwest England. Data from these communities was collected between 1997 and 2007 on research grants designed to document dialect features and track linguistic change in time and space.  The communities are Lerwick, in the Shetland Islands [SHL], Buckie [BCK] in northern Scotland, Cumnock [CMK] in Southwest Scotland, Maryport [MPT] in Northwest England, York [YRK], in Yorkshire, Henfield in Sussex [SSX], Tiverton in Devonshire [DVN], Wincanton [WIN] in Somerset, and Portavogie [PVG] and Culleybackey [CLB] in Northern Ireland.  They come from informal conversations and approximately three million words. Because of the number of different communities, their differences in terms of peripherality and relationship to mainstream developments as well as the cross-generational coverage, these materials give us a rare opportunity to scrutinize linguistic change across dialects and in apparent time. In the next sections, three linguistic changes in progress are scrutinized. Each one operates at a different level of grammar and is proceeding at a different rate.
Figure 4. Map showing the location of 10 research sites in the UK.
2. Stative possession
The forms used to encode stative possessive meaning have been undergoing change for hundreds of years. The oldest variant for this meaning is the main verb ‘have’, with variants have/has as well as their contracted forms, as in (4). The forms with ‘got’, i.e. have/has got arose in approximately 1500 (e.g. Crowell 1959; Tagliamonte, D’Arcy & Jankowski 2010: 4), leading to extensive layering at the individual, community and variety level. A later development is the use of got alone, as in (4d).
- a. Joyce and I just have the one daughter and I ’ve got two granddaughters … (PVG)
b. We always have an advance-party, each troop has about three or four …. It’s got its advantages, that. (MPT)
c. I ’ve got a cousin in New Cumnock that has it and she gets it every month. (CMK)
d. They got [sic] a lovely family bible, you know. Thomas has it. (CLB)
Figure 5 arranges the communities from Figure 4 from north to south in the UK.
Figure 5. Frequency of have got variants across communities.
The incoming variants have got or has got are scant in Northern Scotland and Ireland (BCK, CLB, PVG). They are moderate in England (MPT, WHL, YRK), but in the south (TIV, WIN, HEN) the older forms (have/has) are barely present. The communities reflect a demise of have/has from north to south (Tagliamonte 2013: Figure 7.14). The new forms with got correspondingly increase and the use of got alone is present only in the southwest in Tiverton and Wincantin. In sum, British dialects expose the stages of this ongoing change in geographic relief, i.e. the stages of the linguistic change are parallel to the array of dialects across the country.
The data obtained from some of these communities (e.g. YRK, BCK, WHL) spans the age spectrum from youngest to oldest individuals, permitting assessment of this development in apparent time as illustrated in Figure 6.
Figure 6. Frequency of have got variants by speaker age and community.
The shift to the innovative variants is mirrored within the communities according to speaker age, with younger speakers tending to use the innovative form have got more frequently.
Two linguistic predictors have been shown to condition the change: 1) the nature of the complement and 2) the type of subject. The innovative forms are favoured with concrete complements, as in (5), while the older variant have/has is favoured for abstract complements as in (6) (e.g. Kroch 1989).
- a. She has got near two-thousand pound with her. (CLB)
b. They’ve got a very big boat, like. (PVG)
- a. You have these memories lass. (CMK)
b. I’ve a notion to go to the trip. (CLB)
Similarly, the innovative variants are favoured when the subject is a pronoun, as in (7), while the older variant have/has is favoured when the subject is a noun phrase, as in (8) (e.g. Tagliamonte et al. 2010).
- a. I ’ve got a phone number. (CMK)
b. You’ve got Scotch blood in you. (MPT)
- a. The turbines has a habit o sticking sometimes. (PVG)
b. Every area has different words for things. (PVG)
Leaving aside the differences in frequency of the different variants, how do the community grammars reflect these predictors? Table 1 shows a fixed effects logistic regression of the choice of the innovative have got forms testing the significance and relative strength of each predictor. The corrected mean at the top of the Table shows a measure of the overall probability of have got and the proportion shows the overall frequency. For each of the predictors, type of complement and type of subject, the point form numbers indicate the probability of the have got variants (rather than the have/has variants). Numbers over .50 indicate that these variants are favoured, while those under .50 indicate that the alternative forms (with have/has) are favoured. Where the numbers are not enclosed in square brackets, the effect is significant at the p < .05 level (for further discussion see Tagliamonte 2006a).
Type of complement
Type of Subject
Table 1. Fixed effects logistic regression analysis of have got by linguistic predictors.
Table 1 demonstrates that in every location these constraints are constant regardless of the frequency of the variants. Concrete complements in every case (e.g. money, boats) favour the have got variants, while abstract complements (e.g. memories, notions) disfavor them, occurring more often with have/has. Similarly, for different subject types, pronouns (e.g. you, I) favour the have got variants, while noun phrases (e.g. turbines, every area) disfavour them. The predictors underlying the change have remained constant through to the end point (Tagliamonte 2014). The constancy of these effects across communities confirms the constant rate effect in the progression of change.
In summary, have/has got is reported to be increasing over time in the literature on this topic (Jankowski 2012; Kroch 1989; Tagliamonte et al. 2010). In this study, we observe it increasing across generations and across synchronic British dialects arranged geographically. Moreover, there is a constant rate of the underlying constraints on variation. Synchrony mirrors diachrony. A reasonable prediction is that British dialects will continue to evolve towards incoming have got and that the use of have for possession will fade away. In this case, it is likely that have got will continue to evolve in British dialects according to the extant trajectory.
3. The future temporal reference system
The evolution of going to as a marker of future temporal reference (rather than will or shall)is probably the best-studied case of grammatical change (e.g. Hopper & Traugott 1993). Contemporary dialects of English reflect varying stages in this development and exhibit extensive layering of forms, i.e. alternation between older and newer variants. This is particularly true of British dialects, as in (9).
- a. I says, “It’ll only be six month.” Didn’t know it were gan be six year. (MPT)
b. “How you gonna do that?” “Oh,” said I, “I’ll soon do it.” (CMK)
c. You’ll have to marry Wendy, because otherwise she’s not gonna be able to .. afford to live. (SSX)
Of critical importance for the study of language change in progress is that in some British dialects the frequency of going to remains low. Figure 7 shows the proportion of going to variants out of all future markers. It demonstrates that the going to variant has not yet extensively encroached on the future temporal reference system in these locals, at least not among the elderly speakers (60 years of age and older) represented in the figure. 
Figure 7. Frequency of going to across communities c. 2001–2003 (Tagliamonte 2013).
In North America, going to occurs in over 50% of the future temporal reference sector; however, in the conservative communities in the UK it barely reaches 10% (Tagliamonte, Durham & Smith 2014). The fact that synchronic dialects represent the first half of the S-curve (10% ➜ 50%) means that a cross-dialectal perspective may expose the incremental steps in the grammaticalization process as each predictor evolves via leveling, specialization and reorganization.
Figure 8 displays the frequency of going to as opposed to other future markers (will, shall) in four communities where there is representation of individuals across the age spectrum.
Figure 8. Distribution of going to by speaker age and community.
Once again, the apparent time distributions within communities display an incremental increase in going to (compare Figure 6). In some varieties, like York, there is marked acceleration of going to. This trajectory across age groups is statistically significant in York and Buckie but not Shetland or Wheatley Hill. Note, too, the acceleration among the youngest generation in York.
Nevalainen and Raumolin-Brunberg (2003) argue that a change is incipient when incoming forms represent below 15% of a system and new and vigorous when the change has progressed to 15–36% of the system. Using this convention, the communities under investigation can be partitioned into incipient and vigorous situations.
In addition to the frequency of the incoming future temporal reference form (i.e. going to), linguistic constraints are reported to operate that correlate with different degrees of its grammatical development (e.g. Bybee & Pagliuca 1987). We can use the opportunity of having many different communities with different frequencies of going to to test for differences in the stage of grammaticalization (e.g. Poplack & Tagliamonte 1999; Tagliamonte et al. 2014).
The two main predictors reported in studies of going to are: 1) grammatical person and 2) sentence type. First person (i.e. I/we) is intricately linked to volition, a reading associated with will, as in (10).Going to is correlated with negatives and questions (Berglund 1997; Szmrecsanyi 2003; Torres-Cacoullos & Walker 2009), as in (11).
- a. I’ll ring you up as soon as I get it. (MPT)
b. I’ll catch you tomorrow and I’ll give you all the news. (BCK)
- a. How long you gonna be anyway? (MPT)
b. What are we gonna do here like? (YRK)
Examination of these predictors across communities as in Figure 9a demonstrates that in places where going to represents a very small part of the future temporal reference system, the contexts most highly favouring the innovative form are 1st person negatives and questions, as in (12).
- a. What am I going to do? (CMK)
b. I thought to myself, “I’m not going to go to the shop for bread. (CLB)
Figure 9a. Distribution of going to by grammatical person and sentence type.
In contrast, by the time going to has risen to 20% of the system, this strong effect has begun to level, as in Figure 9b.
Figure 9b. Distribution of going to by grammatical person and sentence type.
In summary, going to is increasing across synchronic British dialects both geographically and with respect to apparent time. However, because going to is only just beginning to penetrate the highly conservative dialects, the early stages of its development are visible. Negatives and questions apparently play a key role at the inception of the change. The use of the new form appears to be pre-disposed in these contexts, but levels as it reaches mid-range in the S-curve. As each predictor evolves via leveling, specialization and semantic bleaching, we can actually see the reorganization of the system. In this case, we can predict how the communities in which going to is incipient will continue to evolve in step with more progressive dialects. Moreover, this leads to a plausible hypothesis, i.e. that data collection from the youngest generation in these communities today would evidence an increase in ‘going to’ and there would be either an ongoing effect of sentence type in the pattern noted above or increasing leveling across contexts.
4. The quotative system
English has a large group of verbs that can be used for direction quotation (a.k.a. verba dicendi). In contemporary varieties say and think are the most frequent forms, but there are innumerable minor variants, e.g. scream, ponder, muse, etc. In recent years a new form has arisen, quotative be like, as in (13). In 1995 direct quotation in narratives of personal experience by Canadian youth was typically variable with rare occurrences of be like, as in (13a). By 2002, this form comprised the vast majority of quotative verbs, as in (13b).
- a. So, I said, “Sure let’s go!” And they were like, “Kim, Kim we’re just joking”. And I said, “Oh no! I’m not.” (TOR, c. 1995)
b. I’m like, “Hi.” I’m like, “Can I please speak with Antonion?” And his mom’s like, “Oh, no. He’s not at home.” (TOR, c. 2002)
In the period between 1995 and 2002, be like increased from 10% to 65% of all quotatives used for direct quotation in Canadian English among speakers under 30. By 2014 it was nearing saturation: quotative verbs other than be like hardly ever occurred anymore, representing only approximately 10% of the system. As be like was nearing the saturation point (i.e. 100% of the system), my students and I conducted fieldwork in two cities where this form had been studied in the 1990’s — York (England) and Toronto (Canada), shown in the map in Figure 10.
Figure 10. Location of communities for be like study in 2014.
The research team recorded narratives of person experience in 2014, just as in 1995 in Ottawa and in 1997 in York (Tagliamonte & Hudson 1999). Then, we undertook a consistent 4-way comparison using the original data files from Tagliamonte & Hudson (1999) and Durham et al. (2012) and coding all the data consistently (Gardner, Denis, Brook & Tagliamonte submitted).
Figure 11 plots the results of a binomial logistic regression, testing the effect of speaker year of birth on the probability of be like in Toronto in blue and in York in red.
Figure 11. Trajectory of change for quotative be like in York and Toronto.
Figure 11 exposes be like as a vigorous and rapid change rising in two S-curves. In both speech communities, be like entered the quotative system at closely consecutive times and then quickly took over as the favoured variant in less than 30 years. Notice the similarity of this curve to the mathematical model above.
This system also has a suite of conditioning predictors, including the nature of the grammatical subject (1st person vs. 3rd person), the tense (past vs. present vs. Historical Present) and the type of quote (direct speech, internal thought, gesture, sound) (e.g. Ferrara & Bell 1995; Tagliamonte & D’Arcy 2004; Tagliamonte & Hudson 1999). While be like can occur in any context, in all my studies, it is favoured in Historical Present contexts. This is visible in (14), where a young man, aged 20 in 2002 tells a story entirely in the Historical Present. It is also favoured with 1st person subjects (14 b, d, f, g, j, k, m) and where the quote records internal thought (14 g, k). Moreover, these constraints have been found to be statistically significant across communities (Buchstaller & D’Arcy 2009; Gardner et al. submitted; Tagliamonte, D’Arcy & Rodriguez Louro under review).
- a. And the girl’s like, “Oh I guess you’re an employee.”
b. So, I’m like, “Yeah sure.”
c. So, she’s just like, “Okay.”
d. So, I’m like, “Okay!”
e. She’s like, “Uh sorry. I think you dropped this.”
f. And I’m like, “Okay.”
g. So, I’m like, “Yes, I fucking just made 20 dollars!”
h. So, anyways, so I ask her if she has any birthday cards
i. And she’s like, “Yeah yeah go upstairs.”
j. So, I’m like, “Okay.”
k. I’m like, “Hmm. I wonder if the same thing works?”
l. And she’s like, “Oh, you’re an employee. Okay.”
m. I’m like, “Yeah, sure!” (TOR)
In summary, be like is increasing across the major varieties of English and with consistent speed and an unmistakable trajectory. This is observable in time, space and across generations. These facts enable the analyst to make a reasonable prediction: other varieties of English will start using be like. Further, where be like takes hold, it is likely to embody similar constraints. At the very least, these predictions enable the analyst to spot deviations and determine what explains them (see discussion Tagliamonte & Denis 2014). It is also likely that a new quotative may evolve. Once a system is saturated, it may only be a matter of time before there is a push towards renewal.
I now return to the central question: can we predict linguistic change? Given my experience in tracking language variation and change across two continents and innumerable community-based data sets, it is clear that once a change gets going, it has a predictable trajectory of development. In terms of frequency, a progression that follows an S-curve is highly foreseeable. Change proceeds regularly according to the logistic model. Some changes proceed according to the constant rate effect, rising constantly across contexts through the course of the change. Other changes proceed according to well-known principles of grammaticalization, extension and levelling. Therefore, if we have sufficient evidence about an ongoing linguistic change, we can make a reasonable prediction about the next steps in its advancement.
Of course, change does not always continue in the same trajectory. Known mechanisms can follow local deviations and external factors can divert a systemic change. As demonstrated in Figure 2 the evolution of do is a virtual model of the S curve. However, there was a visible reversal of direction of change among negative questions and negative declaratives between 1550 and 1600. If we did not have the rest of the curve after this dip in the trajectory, we might think that this change had reversed directions. However the evidence for the rise of do continues to its completion. Warner (2005) demonstrated that the anomaly was caused by register variation. An adverse evaluative effect causes the use of do to plummet in certain styles for a certain period of time. Eventually, however, the change in progress sweeps these cases into the course of change. This provides evidence that stylistic influences do not impede change in progress. Despite hostile stylistic evaluation by certain people in specific high registers for a brief period, do keeps on moving along.
Nevertheless, it appears that some changes can and do reverse direction. Consider the use of have got for stative possessive meaning across the 20th century in British and American plays studied by Jankowski (2005), as shown in the two panels in Figure 12, with British plays on the left and American plays on the right.
Figure 12. Trajectory of variants of stative possessive in British and American English.
Figure 13 plots the frequency of three variants of stative possessiveacross the 20th century — have, have got and got. In British English, as the earlier study of stative possession suggested, the change towards have/has got ascends step by step across the 20th century and have recedes. In the United States however, this form rose to approximately 40% of the system and then it upturned and went the other way, while have, which had earlier been in decline, rises. This occurs among speakers born just after World War 2.
Figure 13. Frequency of have got variants for stative possessive in Canadian English.
A parallel decline in use of have got is visible in Canadian English from the perspective of apparent time in Figure 13. Speakers over 60 years of age (interviewed in 2003–2004) born by the mid 1940’s and earlier use have got (rather than have/has) at approximately the same frequency as the language in plays written in the early 20th century. However, among every age cohort after that, the use of have/has got plummets. The question is: what caused the reversal and the steady late-20th-century progress of have in North America?
In this case, the explanation may lie in overt external pressures that intervened with sufficient impact to divert an already established development (Tagliamonte et al. 2010). In the early 20th century, American grammarians reviled the form got (e.g. Crowell 1959; Rice 1932).In fact, early in the 20th century in North America use of got was considered the fourteenth most frequent error in English (Rice 1932: 292) and people were admonished not to use it. This overt social stigma appears to have been sufficient to lead North Americans to favour have. In Canada, this change has proceeded so far that young people rarely use any other variant for stative possession meaning (see Figure 13), as in (15), both from pre-adolescents.
- a. I have this bumper sticker that says like, “I am Canadian.” (TOR)
b. I don’t have any funny stories. (TOR)
Negative social evaluation has also been put forward as the explanation for the demise of which as a relative pronoun, e.g. “The house which Jack built”.Editing conventions and institutionally backed colloquialization led to the obsolescence of this form (Hinrichs, Szmrecsanyi & Bohmann 2015). Similarly, social and cultural media can elevate and reinforce innovations. For example, I argued in a recent paper that the rise of weird as the favoured adjective of strangeness in North American English was due to the rise of the genre of Weird Fiction in the early 20th century (1923–1954) as well as Weird comic books, promoting use of this adjective by ensuing generations (Tagliamonte & Brooke 2014).
Taken together, these studies of language variation and change lead me to conclude that absolute predications about exactly what will change and when a change will start are impossible. However, the evidence suggests that once a change gets going, it will follow a predictable path unless social influences are strong enough to shock the system into a different trajectory. Of course, predicting language change is not deterministic. Language change is probabilistic just like so many other complex systems. Therefore the success of any predictive model is having extensive information. The chance of knowing the next stages of development in a current linguistic change can be relatively high given the following criteria:
- comparable data
- a longitudinal perspective
- several points in time
- a documented trajectory
Also important is corroborating information from different localities, multiple changes, and due attention to social, cultural and economic factors that may be influencing the linguistic system. Such factors are multiplex, from broad social categories of age, sex and education to the impact of register, style and mode of communication, to cultural conditions such as political movements or government policies, and even the influence of business, industry, advertising and social media. To pursue the goal of predicting linguistic change, I suggest we follow through with replications of existing studies into the next century. Taking care with data selection, coverage, comparability and analytic rigor, we stand a good chance of being able to design studies that will enable us to determine if our predictions about linguistic change are successful. In any case we should be able to catch language change in action, which will make predicting language change a much more successful endeavor in the future.
I gratefully acknowledge the support of the Economic and Social Science Research Council of the United Kingdom for research grants 1997–2003 and the Social Sciences and Humanities Research Council of Canada for research grants 2003 to the present. I also extend deep appreciation to the Killam Trust for Research Fellowship 2013–2015 that enabled me to do the work on this chapter. The broad research program presented here could not have been accomplished without my co-authors on the projects reported in this work: Marisa Brook, Alexandra D’Arcy, Derek Denis, Mercedes Durham, Matt Gardner, Celeste Rodrigues-Louro and Jennifer Smith. I also extend thanks to my two anonymous reviewers, who provided astute and detailed comments on an earlier version of this paper.
 The full historical trajectory of the alternation between sneaked and snuck is not available. Given that the older form is recorded to be snuck, which leveled to sneaked at an earlier point in time, it may be the case that these forms have risen and fallen over a more extended period of time.
 Tagliamonte, Sali A. (2007–2010). Directions of change in Canadian English. Research Grant. Social Sciences and Humanities Research Council of Canada. #410 070 048, Tagliamonte, Sali A. (2010–2013). Transmission and diffusion in Canadian English. Standard Research Grant #410-101-129. Social Sciences and Humanities Research Council of Canada. (SSHRCC), Tagliamonte, Sali A. (2013–2018). Social Determinants of Linguistic Systems. Insight Grant: Social Sciences and Humanities Research Council of Canada.
 Example (3a) comes from a highly conservative dialect region in the Ottawa Valley (Pringle & Padolsky 1983).
 Tagliamonte, Sali A. (2000–2001). Vernacular Roots: A database of British dialects. Research Grant B/RG/AN 6093/APN11081. Arts and Humanities Research Board of the United Kingdom (AHRB).
Tagliamonte, Sali A. (2001–2003). Back to the roots: The legacy of British dialects. Research Grant. Economic and Social Research Council of the United Kingdom (ESRC). #R000239097.
 All the examples in this paper come from the Roots Archive (Tagliamonte 2013) and the Toronto English Archive (Tagliamonte 2003–2006). Acronyms for these communities which appear in the figures are as follows: AYR = Cumnock, Ayrshire, BCK = Buckie, Northern Scotland, CMK = Cumnock, Southwest Scotland, CLB = Culleybackey in Northern Ireland, DVN = Tiverton, Devonshire, MPT = Maryport, Northwest England, PVG = Portavogie, Northern Ireland, SHL = Shetland Islands, SSX = Henfield, Sussex, TOR = Toronto, Canada, WHL = Wheatley Hill, Northern England, WIN = Wincanton, Somerset, YRK = York.
 The portion of the Roots Archive comprising the communities of Cumnock, Maryport, Culleybackey, Portavogie, Henfield, Tiverton and Wincantin is made up of individuals over 60 years of age only.
Bailey, Guy. 2002. “Real and apparent time”. The Handbook of Language Variation and Change, ed. by J.K. Chambers, Peter Trudgill & Natalie Schilling-Estes, 312–332. Malden: Blackwell.
Bailey, Guy, Tom Wikle, Jan Tillery & Lori Sand. 1991. “The apparent time construct”. Language Variation and Change 3(3): 241–264.
Berglund, Ylva. 1997. “Future in Present-Day English: Corpus-based evidence on the rivalry of expressions”. ICAME Journal 21: 7–19.
Buchstaller, Isabelle & Alexandra D’Arcy. 2009. “Localized globalization; a multi-local, multivariate investigation of quotative Be Like”. Journal of Sociolinguistics 13(3): 291–331.
Bybee, Joan L. & William Pagliuca. 1987. “The evolution of future meaning”. Papers from the 7th International Conference on Historical Linguistics, ed. by Anna G. Ramat, Onofrio Carruba & Giuliano Bernini, 107–122. Amsterdam and Philadelphia: John Benjamins.
Chambers, J. K. 1998. “Social embedding of changes in progress”. Journal of English Linguistics 26: 5–36.
Crowell, Thomas L. 1959. “‘Have got’, a pattern preserver”. American Speech 34(2): 280–286.
Durham, Mercedes, Bill Haddican, Eytan Zweig, Daniel Ezra Johnson, Zipporah Baker, David Cockeram, Esther Danks & Louise Tyler. 2012. “Constant linguistic effects in the diffusion of be like”. English Language and Linguistics 40(4): 316–337.
Ellegård, Alvar. 1953. The Auxiliary Do: The Establishment and Regulation of Its Use in English. Stockholm: Almqvist & Wikwell.
Ferrara, Kathleen & Barbara Bell. 1995. “Sociolinguistic variation and discourse function of constructed dialogue introducers: The case of be+like”. American Speech 70(3): 265–289.
Gardner, Matt, Derek Denis, Marisa Brook & Sali A. Tagliamonte. Submitted. “The new global flow of linguistic influence: Be like at the saturation point”. Language Variation and Change.
Hinrichs, Lars, Benedikt Szmrecsanyi & Axel Bohmann. 2015. “Which-hunting and the Standard English relative clause”. Language 91(4).
Hopper, Paul J. & Elizabeth Closs Traugott. 1993. Grammaticalization. Cambridge: Cambridge University Press.
Jankowski, Bridget L. 2005. “We’ve got our own little ways of doing things here: Cross-variety variation, change and divergence in the English stative possessive”. Generals Paper, University of Toronto.
Jankowski, Bridget L. 2012. Cross-Register Language Variation and Change in Canadian English. PhD dissertation, University of Toronto.
Kroch, Anthony S. 1989. “Reflexes of grammar in patterns of language change”. Language Variation and Change 1(3): 199–244.
Kroch, Anthony S. & Ann Taylor. 1997. “Verb movement in Old and Middle English: Dialect variation and language contact”. Parameters of Morphosyntactic Change, ed. by Ans van Kemenade & Nigel Vincant, 297–325. Cambridge: Cambridge University Press.
Kroch, Anthony S. & Ann Taylor. 2000. “Verb-object movement in Early Middle English”. Diachronic Syntax: Models and Mechanisms, ed. by Susan Pintzuk, George Tsoulas & Anthony Warner, 132–163. Oxford: Oxford University Press.
Labov, William. 1963. “The social motivation of a sound change”. Word 19: 273–309.
Labov, William. 1994. Principles of Linguistic Change: Volume 1: Internal Factors. Cambridge & Oxford: Blackwell.
Labov, William. 2001. Principles of Linguistic Change: Volume 2: Social Factors. Malden & Oxford: Blackwell.
Labov, William. 2007. “Transmission and diffusion”. Language 83(2): 344–387.
Labov, William. 2010. Principles of Linguistic Change: Volume 3: Cognitive and Cultural Factors. Malden & Oxford: Wiley Blackwell.
Nevalainen, Terttu & Helena Raumolin-Brunberg. 2003. Historical Sociolinguistics: Language Change in Tudor and Stuart England. London: Pearson.
Pintzuk, Susan. 1996. “Varation and change in Old English clause structure”. Language Variation and Change 7(2): 229–260.
Pintzuk, Susan & Anthony S. Kroch. 1989. “The rightward movement of complements and adjuncts in the Old English of Beowulf”. Language Variation and Change 1(1): 115–143.
Pintzuk, Susan & Ann Taylor. 2006. “The loss of OV order in the history of English”. Blackwell Handbook of the History of English, ed. by Ans van Kemenade & Bettelou Los, 249–278. Oxford & Malden: Blackwell.
Poplack, Shana & Sali A. Tagliamonte. 1999. “The grammaticalization of going to in (African American) English”. Language Variation and Change 11(3): 315–342.
Pringle, Ian & Enoch Padolsky. 1983. “The linguistic survey of the Ottawa Valley”. American Specch 58(4): 325–344.
Rice, Wallace. 1932. “Get and got”. American Speech 7(2): 280–296.
Sanchez-Stockhammer, Christina. 2014. Workshop on ‘Building bridges into the future: Can we predict linguistic change?’ International Society for the Linguistics of English 3 [ISLE 3]. Zurich, Switzerland.
Sankoff, Gillian. 2006. “Apparent time and real time”. Elsevier Encyclopedia of Language and Linguistics, ed. by Keith Brown, 110–116. Amsterdam: Elsevier.
Santorini, Beatrice. 1993. “The rate of phrase structure change in the history of Yiddish”. Language Variation and Change 5(3): 257–283.
Szmrecsanyi, Benedikt. 2003. “‘Be going to’ versus ‘will/shall’: Does syntax matter?” Journal of English Linguistics 31: 295–323.
Tagliamonte, Sali A. 2003. “‘Every place has a different toll’: Determinants of grammatical variation in cross-variety perspective”. Determinants of Grammatical Variation in English, ed. by Günter Rohdenburg & Britta Mondorf, 531–554. Berlin & New York: Mouton de Gruyter.
Tagliamonte, Sali A. 2003–2006. Linguistic Changes in Canada Entering the 21st Century. Research Grant. Social Sciences and Humanities Research Council of Canada (Sshrc). #410-2003-0005. http://individual.utoronto.ca/tagliamonte/.
Tagliamonte, Sali A. 2006a. Analysing Sociolinguistic Variation. Cambridge: Cambridge University Press.
Tagliamonte, Sali A. 2006b. “Historical change in synchronic perspective: The legacy of British dialects”. Handbook on the History of English, ed. by Ans van Kemenade & Bettelou Los, 477–506. Malden & New York: Blackwell.
Tagliamonte, Sali A. 2012. Variationist Sociolinguistics: Change, Observation, Interpretation. Malden & Oxford: Wiley-Blackwell.
Tagliamonte, Sali A. 2013. Roots of English: Exploring the History of Dialects. Cambridge: Cambridge University Press.
Tagliamonte, Sali A. 2014. “System and society in the evolution of change: The view from Canada”. Variability in Current World Englishes, ed. by Eugene Green & Chuck Meyer, 199–238.
Tagliamonte, Sali A. 2015. Making Waves: The History of Variationist Sociolinguistics. Malden & New York: Wiley-Blackwell.
Tagliamonte, Sali A. & Julian Brooke. 2014. “A weird (language) tale: Variation and change in the adjectives of strangeness”. American Speech 89(1): 4–41.
Tagliamonte, Sali A. & Alexandra D’Arcy. 2004. “He’s like; she’s like: The quotative system in Canadian youth”. Journal of Sociolinguistics 8(4): 493–514.
Tagliamonte, Sali A. & Alexandra D’Arcy. 2009. “Peaks beyond phonology: Adolescence, incrementation, and language change”. Language 85(1): 58–108.
Tagliamonte, Sali A., Alexandra D’Arcy & Celeste Rodriguez Louro. Under review. “Outliers, impact and rationalization in linguistic change”.
Tagliamonte, Sali A., Alexandra D’Arcy & Bridget Jankowski. 2010. “Social work and linguistic systems: Marking possession in Canadian English”. Language Variation and Change 22(1): 1–25.
Tagliamonte, Sali A. & Derek Denis. 2014. “Expanding the transmission/diffusion dichotomy: Evidence from Canada”. Language 90(1): 90–136.
Tagliamonte, Sali A., Mercedes Durham & Jennifer Smith. 2014. “Grammaticalization at an early stage: Future ‘be going to’ in conservative British dialects”. English Language and Linguistics 18(1): 75–108.
Tagliamonte, Sali A. & Rachel Hudson. 1999. “Be like et al. beyond America: The quotative system in British and Canadian youth”. Journal of Sociolinguistics 3(2): 147–172.
Torres-Cacoullos, Rena & James A. Walker. 2009. “The present of the English future: Grammatical variation and collocations in discourse”. Language 85(2): 321–354.
Verhulst, Pierre-François. 1845. “Recherches mathématiques sur la loi d’accroissement de la population”. Nouv. mém. de l’Academie Royale des Sci. et Belles-Lettres de Bruxelles 18: 1–45.
Warner, Anthony. 2005. “Why do dove: Evidence for register variation in Early Modern English negatives”. Language Variation and Change 17: 257–280.
Weinreich, Uriel, William Labov & Marvin Herzog. 1968. “Empirical foundations for a theory of language change”. Directions for Historical Linguistics, ed. by Winfred P. Lehmann & Yakov Malkiel, 95–188.