5.1 Introduction

Discourse-pragmatic change is perhaps the least well-studied type of change in quantitative approaches to language variation and change. This is undoubtedly due in part to the early debates questioning the extension of the variationist enterprise to levels of grammar above phonology (see, for example, Lavandera Reference Lavandera1978; Rickford Reference Rickford, Fasold and Shuy1975; Sankoff Reference Sankoff, Bailey and Shuy1973; see Waters [Chapter 2] for an overview). More recently, these earlier concerns have resurfaced as cautionary notes on the use of quantitative methods in the study of discourse-pragmatic variation. There is also a concomitant unease that current generalisations about the nature of change in phonology, morphology and syntax, including a constant rate of change (Kroch Reference Kroch1989), layering effects, phonetic reduction/erosion and analogical levelling (Hopper Reference Hopper, Traugott and Heine1991; Hopper and Traugott Reference Hopper and Traugott2003), will apply in the same way to discourse-pragmatic phenomena (see, for example, Pichler Reference Pichler2013: 13). Yet numerous studies over the past decade have determinedly opened up this area of grammar to quantitative investigation. Among the most prominently studied features is the general extender (GE) system (see, for example, Cheshire Reference Cheshire2007; Denis Reference Denis2011; Dines Reference Dines1980; Overstreet Reference Overstreet1999; Pichler and Levey Reference Pichler and Levey2011; Tagliamonte and Denis Reference Tagliamonte and Denis2010; Winter and Norrby Reference Winter, Norrby and Henderson2000). Given the extensive knowledge base that is developing for this area of grammar, the time is ripe to offer continued breadth to the emerging contemporary situation (see Pichler and Levey Reference Pichler and Levey2011).

GEs can be identified using a combined structural and functional approach (see also Denis and Tagliamonte Chapter 4; Pichler Chapter 3; Waters Chapter 2). Structurally, they are semi-fixed constructions that share a common schematic pattern (Pichler and Levey Reference Pichler and Levey2011: 448) which typically involves the following items: a connector (i.e., and, or), a generic (e.g., stuff, thing) or indefinite pronoun (e.g., something, everything, anything), and an optional comparative (e.g., like that), as in the examples in (1). The GE system also includes a number of expressions whose structural configuration is more fixed, as in (2a), as well as the occasional idiosyncratic variants, as in (2b).

(1)

1. a.

   Dad always had the garden. We grew our own vegetables, potatoes and onion-sets and cabbages, sprouts or something like that, yes.

   (Grace Kenway, 74, MPT)Footnote ¹

2. b.

   See Aunty Jessie, she made wee sheets and wee pillowcases for the cot and everything.

   (Esther Hamilton, 88, CMK)

3. c.

   We were told there was no saluting or medals or anything.

   (Roger Bishop, 88, MPT)

4. d.

   You go back thirty years and look at all the old maps and all. Portavogie was only put down as a headland or something.

   (Robin Mawhinney, 55, PVG)

5. e.

   Then the harvest time you know where they cut the corn and that.

   (Hugh Keane, 84, CLB)

6. f.

   I seen there they brought some nettles over and things.

   (Rob Paisley, 78, CLB)

(2)

1. a.

   I says, well I would like some nice gifts to take back … I’ll take this, that, blah, blah, blah.

   (Heather Baker, 67, CMK)

2. b.

   Kerry would send the wee one up with his box of chocolates and what have you for Gran.

   (Iain Ferguson, 72, CMK)

A typical GE will occur at the end of a phrase and will often evoke a general category of similar objects that the speaker has in mind (Dubois Reference Dubois1992: 198). For example, the use of or something like that in (1a) calls to mind different types of vegetables that would grow in a garden. The GE and everything in (1b) refers to a group of attributes relating to the things that make up a child’s bedding. The GE in (1c) refers to the paraphernalia suitable for a military parade, that in (1d) to historical materials that one might find in an archive, that in (1e) to the things that are cut at harvest time, and that in (1f) to any of a variety of herbs. Thus, speakers use a GE in order ‘to suggest the multitude of possible elements of the set’ that they are thinking or talking about (Dubois Reference Dubois1992: 182). As the research on GEs has expanded, the boundaries of what can be treated as a GE has extended to encompass additional forms and in some cases additional functions. As we shall see, however, the core remains fairly constant.

Research shows that the use of GEs is conditioned by social factors such as age, sex, education and socio-economic class (Dubois Reference Dubois1992; Stubbe and Holmes Reference Stubbe and Holmes1995). The distribution of GEs has also been found to be constrained by the semantic-syntactic characteristics of the generic/pro-form they contain (Aijmer Reference Aijmer and Togeby1985; Cheshire Reference Cheshire2007; Dines Reference Dines1980; Overstreet and Yule Reference Overstreet and Yule1997). Other research suggests GEs encode various interactional functions (Aijmer Reference Aijmer and Togeby1985), marking politeness (Overstreet and Yule Reference Overstreet and Yule1997; Winter and Norrby Reference Winter, Norrby and Henderson2000), topic shift and speaker exchange (Dubois Reference Dubois1992).

A rich aspect of the research on GEs is the implications they have for theories of grammaticalisation. Early research on GEs suggested that they were grammaticalising in some varieties of contemporary English (see, for example, Aijmer Reference Aijmer2002; Brinton Reference Brinton1996). Grammaticalisation is change that involves grammatical development. Originally, Meillet (Reference Meillet1912) defined grammaticalisation as ‘the change of an autonomous word into a grammatical element’; however, later research demonstrated that not only lexical items, but also grammatical forms and even phrases can develop along a pathway of grammaticalisation (Traugott Reference Traugott, Lehmann and Malkiel1982). Furthermore, the processes by which items develop grammatical meanings are multiplex and in some cases antithetic. For example, grammaticalisation is associated with reduction in some models (Heine et al. Reference Heine, Claudi and Hünnemeyer1991; Lehmann Reference Lehmann1982) but with expansion in more recent formulations (see, for example, Himmelmann Reference Himmelmann, Bisang, Himmelmann and Wiemer2004; Traugott and Trousdale Reference Traugott, Trousdale, Traugott and Trousdale2010). These hypotheses are particularly germane to GEs since the processes invoked to explain their development include phonetic reduction or clipping, decategorialisation and semantic-pragmatic change (see, for example, Cheshire Reference Cheshire2007), all among the suite of mechanisms that have been discussed in the grammaticalisation literature. In the analyses that follow, I will test the operation of some of these mechanisms.

The originality of the present analysis lies in the viewpoint it offers from conservative dialects. The argumentation and interpretation is built on the foundations of distributional findings presented in Tagliamonte (Reference Tagliamonte2013: 172–84),Footnote ² based on a unique archive of language materials from south-west Scotland, north-west England and Northern Ireland. In contrast, virtually all of the research on GEs has been on contemporary data, much of it from mainstream communities and contemporary youth. Further, claims in the literature that implicate the diachronic development of GEs have never been fully explored (see, however, Carroll Reference Carroll2008). Perhaps one of the most compelling claims is that GEs develop from long forms to short forms, paralleling linguistic change in other areas of the grammar. Using the apparent-time construct (Bailey et al. Reference Bailey, Wikle, Tillery and Sand1991), recent quantitative research (Denis Reference Denis2011; Pichler and Levey Reference Pichler and Levey2011; Tagliamonte and Denis Reference Tagliamonte and Denis2010) took up the challenge of testing this claim by studying the distribution of long and short GE variants in synchronic datasets. The consistent finding arising from these investigations is that there is no relationship between speaker age and the length of GEs. Indeed, it appears that short GE variants such as and stuff and long GE variants such as and stuff like that exist in relative stasis in contemporary varieties. Thus, if the short variants did indeed derive from the longer variants, they must have done so before most of the speakers interviewed for the apparent-time studies quoted above were born. Of course, the real problem is the availability of comparable time-depth to test this hypothesis. Appropriate data, i.e., spoken, interactional and vernacular, from earlier points in time does not exist. Spoken-like materials in drama, diaries and court reports provide some evidence for GEs at earlier periods (see, for example, Culpeper and Kytö Reference Culpeper and Kytö2010) but they reflect a qualitatively different register. There is no perfect solution to this problem given the impossibility of bona fide spoken historical data from earlier than the twentieth century. However, there is another worthwhile data source that may shed light on the diachronic development of GEs: conservative British dialects. Peripheral dialects are known to reflect earlier stages in the evolution of English and thus make possible the analysis of the antecedents of contemporary GE variants. If so, they can offer insight into the evolution of GEs to complement the synchronic corpora studied by Cheshire (Reference Cheshire2007), Pichler and Levey (Reference Pichler and Levey2011) and Tagliamonte and Denis (Reference Tagliamonte and Denis2010).

In what follows, I will first describe in more detail the nature of the data under investigation in this chapter. In Sections 5.3 and 5.4, I will provide a brief historical and synchronic backdrop for the study of GEs. The methods of data analysis employed will be outlined in Section 5.5. The results are then presented in terms of distributions, Section 5.6, and constraints, Section 5.7. I will end with a discussion of the results in Section 5.8 and my conclusions in Section 5.9.

5.2 The data

An informative construct that has been successful in recovering earlier stages in the development of linguistic variables is the relic area (Poplack and Tagliamonte Reference Poplack and Tagliamonte1991: 97; Tagliamonte and Smith Reference Tagliamonte, Smith and Poplack2000: 141). Such areas, because of their peripheral geographic location and/or isolated socio-political circumstances, tend to preserve older features of a language (Anttila Reference Anttila1989: 294; Hock Reference Hock1986: 442). Among the more recent criteria for establishing the status of a variety as ‘peripheral’ are geographic location, historical continuity and group identity (Wolfram and Thomas Reference Wolfram and Thomas2002: 26–36). The present study is based on the Roots Archive (Tagliamonte Reference Tagliamonte2013), a database which comprises four such peripheral communities. Their location is shown by the grey circles on the map in Figure 5.1: Maryport in north-west England, Cumnock in south-west Scotland, and Cullybackey and Portavogie in Northern Ireland. In addition, the white circles on the map in Figure 5.1 show northern England locations where GEs have been studied recently: Kingston-upon-Hull (henceforth Hull) (Cheshire Reference Cheshire2007), the city of York (Denis Reference Denis2011) and Berwick-upon-Tweed (Pichler and Levey Reference Pichler and Levey2011).

Figure 5.1 Location of targeted peripheral communities (Maryport, Cumnock, Cullybackey, Portavogie – grey circles) and locations of previous GE research (Kingston-upon-Hull, York, Berwick-upon-Tweed – white circles)

Contains OS data © Crown copyright and database right (2014)

In the early 2000s, over 100 individuals among the oldest living generation in each of the peripheral communities were interviewed using sociolinguistic fieldwork techniques (Labov Reference Labov1972c; Schilling Reference Schilling2013; Tagliamonte 2006). These audio-recorded conversations contain a wealth of cultural information, local traditions, narratives of personal experience, local gossip and informal discussions, reflecting the typical discourse found in each community. GEs are a vibrant component of the vernacular in these communities, as evident in (3):

(3)

1. a.

   When I worked there, I worked in the stores department and one of my jobs at that particular time was clock cards and things like that. Pay documents all this, making them out. Anybody that was entitled till extra money or owt like that.

   (Jack Dobson, 66, MPT)

2. b.

   Well, she took natural sciences. It covers a wide area. I think she wants to work with people like you know like sort of these trouble shooters people and that you know.

   (Lucy Fisher, 73, MPT)

Most other research on GEs in the UK is based in urban settings, for example London (Stenström and Andersen Reference Stenström, Andersen, Percy, Meyer and Lancashire1996). More generally, research on GEs has in large part been conducted on other major varieties of English, including Australian English (Norrby and Winter Reference Norrby, Winter and Allen2002) and Canadian English (Tagliamonte and Denis Reference Tagliamonte and Denis2010). One of the compelling findings arising from the literature on GEs is the extent to which varieties of English share the same variants. From England to Canada to Australia, virtually the same inventory of GE variants has been reported (at least with respect to the most frequent variants), suggesting longitudinal continuity of GEs. Yet this remains open to investigation. Pichler and Levey (Reference Pichler and Levey2011: 464) call for an extension of current variationist work on GEs to studies of real time and most especially to deeper time-depth and consideration of the earlier antecedents of the contemporary GE system. Exploration of the conservative dialects in the Roots Archive offers the possibility to gain access to a window on the past through which an earlier stage in GE variation and change may become visible, allowing us to probe the formal consistency of the GE system as well as hypotheses about the development of short GE variants from long GE variants. The question is: how do the GEs in these relic communities compare with the extant body of knowledge of this linguistic phenomenon?

5.3 Historical perspective

GEs can be traced far back in the history of English, at least to the fourteenth century (Poutsma Reference Poutsma1926: 914). The earliest documented forms comprise fixed expressions such as and such, and so forth, and so on, and what not. Carroll (Reference Carroll2008: 16) provides an example of and so forth from Chaucer, 1390. Many of these early forms are found in the Roots Archive, as shown in (4).

(4)

1. a.

   He was doing the lambing and such.

   (Elspeth Ferguson, 70, CMK)

2. b.

   And they sold fruit and so forth.

   (Charlie Baxter, 67, CMK)

3. c.

   Saddie McCaine would sell broken biscuits, chocolates, broken chocolates and so on.

   (Fergus Bell, 90, CMK)

4. d.

   My mother went to see my cousin in Workington. They were confectioners you know sort of they had a café and what not.

   (Miriam White, 81, MPT)

GEs tend to comprise a core set of generic nouns and pronouns, including thing(s), stuff, anything, something, everything and nothing. Dialectal variants of the generic pronouns also occur such as summat in (5a) and owt in (5b).

(5)

1. a.

   She’s a black belt and first dan or summat.

   (John Edwards, 70, MPT)

2. b.

   Cos he’s not a fellow to mix with other men or owt like that or do anything or owt, isn’t Bob.

   (Grace Kenway, 74, MPT)

GEs with a generic but no connector occur in the Roots Archive as well, consistent with reports from nearby Hull and Berwick-upon-Tweed (Cheshire Reference Cheshire2007; Pichler and Levey Reference Pichler and Levey2011), as in (6).

(6)

1. a.

   That’s right, aye, you’d the place to yourself kind of thing.

   (Dan James, 64, PVG)

2. b.

   I’ve a feeling it was summat like er- one of the H’s, Hunslett, Huddersfield something like that.

   (Jack Dobson, 66, MPT)

Examination of the historical record shows that GEs with and developed first and those with or arose somewhat later. The earliest GE with or, or something, is attested in the early 1800s; or whatever is attested in the early 1900s (Tagliamonte and Denis Reference Tagliamonte and Denis2010). This suggests that there is incremental historical layering in this system over time that may be corroborated by the Roots Archive.

5.4 Synchronic perspective

GEs have been studied in contemporary data from the following countries: England, Scotland and Ireland (Cheshire Reference Cheshire2007; Denis Reference Denis2011; Levey Reference Levey2012; Macaulay Reference Macaulay1991; O’Keeffe Reference O’Keeffe2004; Pichler and Levey Reference Pichler and Levey2011); the United States and Canada (Overstreet Reference Overstreet1999; Overstreet and Yule Reference Overstreet and Yule1997; Tagliamonte and Denis Reference Tagliamonte and Denis2010; Wagner et al. Chapter 9); and Australia and New Zealand (Britain and Sudbury Reference Britain, Sudbury, Jones and Esch2002; Dines Reference Dines1980; Norrby and Winter Reference Norrby, Winter and Allen2002; Stubbe and Holmes Reference Stubbe and Holmes1995). As mentioned earlier, a common set of variants are reported, including: or something (like that), and everything (like that), and things (like that), and stuff (like that). Some GE types or categoriesFootnote ³ (e.g., those containing stuff and whatever) are more frequent in large urban centres such as Toronto, Canada (Tagliamonte and Denis Reference Tagliamonte and Denis2010) or London, England (Stenström et al. Reference Stenström, Andersen and Hasund2002). On the whole, however, there is remarkable consistency in the inventory of main forms across widely divergent geographic locales, social groups and registers. Another overarching characterisation is that GEs are typical of the vernacular.

A frequent observation is that GEs are rising in frequency, with a greater usage among younger individuals and in particular females (Denis Reference Denis2011; Stubbe and Holmes Reference Stubbe and Holmes1995; Tagliamonte and Denis Reference Tagliamonte and Denis2010). Further, in part due to the wide-ranging variety of GE variants, from almost literary (and conservative) (e.g., and so forth) to highly vernacular and nonstandard (e.g., and that), different GE variants are often associated with varying levels of formality, register or style (see, however, Wagner et al. Chapter 9). Certain GE variants are also often associated with class and other socio-economic indicators. For example, GEs such as and things are associated with middle-class individuals while others such as and that or and that lot are associated with working-class individuals (Cheshire Reference Cheshire2007: 165; Stenström and Andersen Reference Stenström, Andersen, Percy, Meyer and Lancashire1996: 102).

Further, GEs are claimed to have a range of interpersonal and textual functions (see, for example, Cheshire [Reference Cheshire2007] for a recent overview). Among other things, they are claimed to mark in-group affiliation and rapport (see, for example, Youssef [Reference Youssef1993] on an ting ‘and things’ in Trinidad Creole). The in-group nature of GEs is evident in (7a) and (7b) where one must know what a ‘bran tub’ and ‘cobbles’ are to interpret the reference set of the GE.Footnote ⁴ Because cultural knowledge is required to interpret many GEs, their use is thought to signal that speakers perceive an in-group relationship to their listeners.

(7)

1. a.

   Aye, and at Haleve time it was real smashing at Haleve time. Aye, they had the- the bran tub and all this sort of stuff at Haleve.

   (Bob Cottell, 85, CLB)

2. b.

   And a-fore they knew where they were at there was cobbles and all sorts flying at windows.

   (Keith Price, 89, MPT)

The semantic-syntactic properties of the GE generic (e.g., stuff, things) have been shown to influence the choice of GE variant. Grammatically, a form such as stuff refers to mass nouns and a form such as thing(s) to count nouns, as in (8a)–(8b). However, researchers have noticed that GE generics do not always match the semantic-syntactic properties of their antecedent nouns, as in (8c)–(8e). Further, note that it is much more difficult to retrieve a set for the referent pumps in (8e) than for the parsley and leeks and kale in (8b). What does one sink with a pump? This is not part of the knowledge set of most people.

(8)

1. a.

   When they made anything new, they started to make oh meats and stuff like that, all kinds of things in the factory, all chutneys and things.

   (Helen Philips, 79, MPT)

2. b.

   My granny would go out and get parsley and leeks and kale and things out the garden for soup.

   (Esther Hamilton, 88, CMK)

3. c.

   Oh aye, that’s the sort of things I used to give them too, you know, clothes and things.

   (Nancy Taylor, 78, CMK)

4. d.

   It’s a shame when it’s stuff like- even affects like pigeons and stuff.

   (John Edwards, 70, MPT)

5. e.

   Did they not sink pumps and stuff?

   (Kate Devoy, 62, PVG)

More critically, it is sometimes the case that GEs have no clear-cut reference set (Winter and Norrby Reference Winter, Norrby and Henderson2000: 4). What should be inferred by and that in the examples in (9)? What else might be involved in being ‘sent back out’ to work in (9a) and what is involved in being ‘a big friend of Josh’ in (9b)? These examples demonstrate that some GEs are quite vague, not always functioning to generalise to a set (Aijmer Reference Aijmer and Togeby1985). If not, they may have other functions.

(9)

1. a.

   There was nae unions or protective clothing then, they just … and I mean she had to work because her mother made sure she worked. She never missed a day. You know she was Irish. Go to work, that’s it. Cut hand or no she was sent back out the next morning and that.

   (Graeme Nesbitt, 69, CMK)

2. b.

   Pickering chemist, he was a big friend of Josh and that.

   (Keith Price, 89, MPT)

Changes to the syntagmatic length of GEs may be a reflection of shifting function. For example, and things like that may evolve into and things as part of the grammaticalisation process whereby clipping or shortening occurs as the original form expands its functional domain. This is, of course, not a discrete stepwise development but part of a suite of shifts that evolve in the milieu of synchronic variation. In this case, we can conceive of GEs starting out as a generalisation to a set. At a later stage, the forms may not generalise to a set at all. In such a process, the form and that in (9) could be a development. However, researchers are discovering a multitude of forms within each GE type (e.g., and all things like that, and things like that, and things, and thing) and a broader set of contexts (i.e., extension beyond set-marking) (see, for example, Carroll Reference Carroll2008: 14; Pichler and Levey Reference Pichler and Levey2011). The question is: what are the characteristics and patterns of GE variant types and forms that obtain in the peripheral northern communities under investigation?

5.5 Methodology

Discourse-pragmatic variables are notoriously difficult to study from a quantitative perspective because they comprise many different variants with potentially diverse functions (see Waters Chapter 2). To approach the GE system scientifically, I used the combined structural and functional approach employed in early variationist studies of GEs (Dubois Reference Dubois1992) and other discourse-pragmatic variables (Vincent Reference Vincent1992; Vincent and Sankoff Reference Vincent and Sankoff1992), also consistent with Pichler’s (Reference Pichler2010, Reference Pichler2013) more recent methods for studying discourse-pragmatic variation. First, forms and structural configurations are identified. Quantitative research on GEs, which often includes the full inventory of GEs in the data under investigation, provides a precedent. These procedures are aided by the well-known GE template: (connector) (modifier) (generic noun/pro-form) (similative) (deictic) (Pichler and Levey Reference Pichler and Levey2011: 448). With known structural and functional possibilities in mind, additional forms can be identified and included. A total of 1662 GE tokens were accumulated across the four communities presented in the Roots Archive. Each token was then coded so as to identify the generic/pro-form (e.g., thing, stuff, something) and to probe two salient claims in the literature for grammaticalisation of GEs: the relative length of the construction (e.g., and things vs. and things like that) tests for the relationship between grammaticalisation and the syntagmatic length of the GE; the co-occurrence of GEs with other discourse-pragmatic variables in the same clause (e.g., ken, you know, I mean) tests whether there is a relationship between GE use and support from other discourse-pragmatic variables.

5.6 Distributional analysis

Table 5.1 provides a breakdown of the inventory of the GE types in the Roots Archive. The table documents all the types in the data by grouping the GE variants into categories according to the type of generic/pro-form they contain.Footnote ⁵ The most frequent form of each category is displayed first, followed by a combination of all the other variants in the category with that generic type. There were 88 different forms in all. As with many linguistic variables, several main forms dominate the system while many others occur only once or twice. The distribution reveals that nearly a full third of all GEs in the data are some combination of and (all) (that), representing 38.1% of all GEs. Variants with something or thing(s) make up the bulk of the remainder, followed by everything, anything and stuff variants. The large catch-all category ‘other’ at 12.3% comprises innumerable idiosyncratic forms along with a group of fixed expressions which include and whatnot, and such, and all the rest etc., each one alone well under 2% of the data.

Table 5.2 provides a breakdown of the most frequent fixed GEs, which were grouped with idiosyncratic forms as ‘other’ in Table 5.1. These are constructions that do not entirely fit the structural characteristics described earlier, but perform the same range of functions and occur in the same syntactic environments as the more prototypical and less formulaic GE variants in Table 5.1 (see also Pichler and Levey Reference Pichler and Levey2011: 448).

In sum, the Roots Archive is consistent with previous research on GEs in its rich variation of forms. However, the particular GE inventory in these communities contrasts from existing reports outside the northern climes of the United Kingdom, particularly that reported for Reading in the south of England where thing(s)-variants dominate, and Toronto in Canada where stuff-variants dominate. In the Roots Archive, variants with the generic stuff are rare (2%, N = 34/1662). While stuff-variants were also infrequent in Berwick-upon-Tweed (6%, N = 50/783; Pichler and Levey Reference Pichler and Levey2011), the rate in the current data is even lower, perhaps approaching the rate of working-class speech in Hull (Cheshire Reference Cheshire2007: 164, Table 4). Second, the frequency in the Roots Archive of the category of short forms, i.e., and that, and all, and all that, is relatively high. These variants have been reported in earlier research in Scotland (Macaulay Reference Macaulay and Görlach1985) and both Pichler and Levey (Reference Pichler and Levey2011) and Cheshire (Reference Cheshire2007: 164) report these to be frequent among working-class speakers in Berwick-upon-Tweed (34%, N = 268/783) and Hull. Moreover, recent work reports them to be frequent among working-class children in London (Levey Reference Levey2012). In the Roots communities, the rate of and that is only 19% overall, compared to 31% in Berwick-upon-Tweed (Pichler and Levey Reference Pichler and Levey2011). However, when and that, and all and and all that are taken together in the Roots Archive, they represent nearly 40% of the GEs in the data, i.e., their combined rate in the Roots Archive is higher than that in Pichler and Levey’s (Reference Pichler and Levey2011) data from Berwick-upon Tweed, as mentioned earlier. Given that the data under investigation were collected in relic dialect areas and from conservative speakers, the overwhelming presence of these short variants, which are thought to be a later development, is notable.

The next step is to consider whether the Roots Archive communities are differentiated with respect to this overall picture in type/variant distribution. Figure 5.2 shows the distribution of the major GE types by community. It confirms that the communities have a parallel pattern with respect to their inventory of GE types. And (all) (that) (light grey columns) is the prevailing category in nearly every community. Indeed, in Cullybackey and Maryport, it is the dominant category by a wide margin. The GE category with something (dark grey columns) is most robust in Cumnock and Portavogie whereas in Cullybackey and Maryport it is less frequent. The rate of use of forms in the ‘other’ category (nonce forms that are also reported elsewhere, e.g., and so on, and whatnot) is relatively stable throughout. None of the communities show more than a few tokens of stuff- or whatever-variants. Yet GEs with stuff and whatever are reported as frequent in North America (Tagliamonte and Denis Reference Tagliamonte and Denis2010) and to be rising in frequency in apparent time in Berwick-upon-Tweed (Pichler and Levey Reference Pichler and Levey2011: 454, Figure 2) and York (Denis Reference Denis2011). These discrepancies between the conservative dialects in the Roots Archive and the dialects from two larger centres in north-east England support the hypothesis that there is change in the composition of variants within the GE system over time but no change in terms of grammatical development. If it is simply the lexical components that are changing, this may reflect renewal rather than the evolution of lexical items into new grammatical functions.

Figure 5.2 Distribution of GE types by community

Further support for this interpretation comes from a fortuitous comparison that can be made between individuals of different ages in the corpora. As it happens, two of the fieldworkers were in their twenties, i.e., considerably younger than the individuals they were interviewing who were all over the age of sixty.Footnote ⁶ Thus, their use of GEs within the interview context offers a unique generational perspective on usage.Footnote ⁷ Figure 5.3 shows the distribution of the major GE categories but this time contrasts the elderly participants with these interviewers. The results expose some pointed apparent-time differences in GE use. GE use among the elderly community members is dominated by the and (all) (that) category, forms rarely used by the interviewers. In contrast, the interviewers exhibit dramatic use of stuff-variants, a GE type rarely used by the interviewees. These discrepancies between the conservative dialect speakers and the younger interviewers provide a suggestive reflection of the type of change the GE system may have undergone over the twentieth century, namely lexical replacement of: thing > stuff. At the same time, it is interesting to note the relatively parallel frequencies of the something/anything-types across the two groups. This shows that while some parts of the GE system are changing, other areas remain stable, at least across the (apparent-time) span of these materials. While other factors could be contributing to the differential use of GE variants among these two groups (e.g., contrasting roles in the interview context, gender asymmetry), it is striking that the very differences visible here in frequency and proportion of GE variants across elderly interviewees and young interviewers is also consistent with the patterns reported for older vs. younger individuals in urban locales elsewhere, either in North America (Tagliamonte and Denis Reference Tagliamonte and Denis2010) or in more urban settings relatively nearby in the United Kingdom (Denis Reference Denis2011; Pichler and Levey Reference Pichler and Levey2011), namely more stuff-variants among youth and more and (all) (that)-variants among elderly. These results directly impact the hypothesis stated earlier regarding the purported nature of the GE system. The hypothesis is that conservative dialects can be taken as a cautious representation of an earlier stage in the development of this system. If they are, then that system was no more grammaticalised than the GE systems reported in any other study. What contrasts is the composition of the system. These conservative dialects are dominated by forms of the type and (all) (that).

Figure 5.3 Distribution of GE types comparing elderly interviewees and younger interviewers

5.7 Constraints analysis

A central issue that emerges is whether or not GEs have been subject to grammatical change or lexical replacement (see also Denis and Tagliamonte Chapter 4). Two of the main indicators of grammaticalisation are: (i) a change from longer GEs to shorter ones, and (ii) a change from co-occurrence of GEs with other discourse features to independent usage. The key evidence for lexical replacement is marked change in the use of the generic but no shifts in the contextual grammatical constraints. In the analyses that follow, I subject the data to an analysis of these constraints in order to test these potential indicators.

5.7.1 Length of the GE

Cheshire’s (Reference Cheshire2007) research carried out in the 2000s on young people’s GE usage in three English towns found a high frequency of and that. Following Aijmer (Reference Aijmer2002: 227), Erman (Reference Erman, Moen, Simonsen and Lødrup1995: 145) and others, Cheshire (Reference Cheshire2007: 167) reasoned that this was the result of a gradual loss of the longer GE variants because of reduction via clipping. In this hypothesis, longer GEs such as and things like that would have been present at earlier stages in the development of the GE system. Then, these would have gradually evolved into shorter GEs to arrive at a situation of frequent use of and things. If this trajectory is accurate, we would expect a greater number of longer GEs than shorter GEs in the Roots Archive data which represent an earlier stage in the development of GEs than the data analysed by Cheshire (Reference Cheshire2007).

Figure 5.4 tests this possibility by comparing the frequency of short and long GEs in each of the Roots Archive communities. The results provide a striking demonstration that the shortest GEs are predominant across all communities. Interestingly, the longer GEs show the highest frequency in Maryport. This is the locale that has previously exhibited the most advanced profile among the Roots Archive communities (see, for example, Tagliamonte Reference Tagliamonte2013). If anything, this suggests that the longer GEs, rather than the short ones, are developmentally advanced.

Figure 5.4 Frequency of length of GE by community

Figure 5.4 groups all of the GEs together. However, Figures 5.5–5.8 split the data so that the cross-community distribution of long vs. short variants can be viewed separately for the highest-frequency GE categories (see Table 5.1). Figures 5.5–5.8 confirm that the long vs. short phenomenon reported in the overall perspective in Figure 5.4 accurately reflects the situation for each GE type and for each Roots community. There is no evidence for a trajectory of change from long to short. This means that the GE system in these varieties shows no evidence of an earlier stage in which longer GEs were present. Instead, just like more mainstream varieties (e.g., Toronto, York, London), they simply have more two-word variants.

Figure 5.5 Proportion of long vs. short GE variants with pro-form ‘something’

Figure 5.6 Proportion of long vs. short GE variants with generic ‘thing(s)’

Figure 5.7 Proportion of long vs. short GE variants with pro-form ‘everything’

Figure 5.8 Proportion of long vs. short GE variants with generic ‘stuff’

5.7.4 What of ‘and that’?

Given the notable differences across communities based on the type of GE, it will be instructive to determine how the major GE type in these data, and (all) (that), distributes across communities when the individual variants are viewed independently. The form and that is the pivotal form proffered as evidence that the GE system is changing over time. The argument is that it is the most advanced in terms of grammaticalisation based on tests for frequency, length and co-occurrence with discourse-pragmatic features (see, for example, Cheshire [Reference Cheshire2007] based on earlier research by Aijmer [Reference Aijmer2002]; Brinton [Reference Brinton1996]). Figure 5.9 plots the frequency of each of the variants of this type, i.e., and all, and that and and all that. The results reveal that the distribution of these forms is highly differentiated from one community to the next. Maryport favours the and that-variant, Portavogie favours and all that and Cumnock and Cullybackey favour and all. The results for Maryport, in north-west England, show remarkable correspondence with working-class individuals in the communities of Hull (Cheshire Reference Cheshire2007) and Berwick-upon-Tweed (Pichler and Levey Reference Pichler and Levey2011), also both located in northern England. The fact that, in addition, these data comprise a majority of working-class speakers corroborates the hypothesis that and that is a working-class variant (Cheshire Reference Cheshire2007: 164). Further, correspondence across these northern communities suggests that and that may well be a more general northern working-class feature.

Figure 5.9 Distribution of and all, and that, and all that by community

A final question to consider is whether and that, as in (10), is old or new. In working-class speech in the early 2000s in Reading, Milton Keynes and Hull in the United Kingdom, and that is reported to be ‘firmly entrenched’ (Cheshire Reference Cheshire2007: 165). Quantitative investigation in Berwick-upon-Tweed (Pichler and Levey Reference Pichler and Levey2011) allows a consistent comparison with the Roots Archive data, as in Figure 5.10. This demonstrates that the oldest living generation in the small north-east England town of Maryport also use and that to a relatively high degree, patterning along with Berwick-upon-Tweed. This distributional pattern suggests that and that is not an advancing GE in England but rather the retention of a conservative northern English feature generally, as in (10a)–(10b). In sum, the short GE and that is not a development, i.e., a reflex of ongoing grammaticalisation, but it is a retention.

(10)

1. a.

   We had greenhouses like and green tomatoes and that you know.

   (Phil Stevenson, 84, MPT)

2. b.

   And then there were maybe a wee dance and that.

   (Angus Milroy, 66, CMK)

Figure 5.10 Proportion of and that in the Roots Archive compared to Berwick-upon-Tweed (as reported in Pichler and Levey Reference Pichler and Levey2011)

5.8 Discussion

The study of GEs in the Roots Archive has offered a new perspective on the GE system in relatively conservative dialects among elderly speakers from the general vantage point of northern England, south-west Scotland and Northern Ireland. GE use in these conservative dialect data is an indication into a possible earlier stage in the GE system. Given the competing hypotheses regarding the status of the GE system in contemporary varieties in England (and elsewhere), this is an important addition to the current body of knowledge. Both frequency and patterns of use combine to show the inner workings of the system across dialects. While there is very little use of stuff-variants, a type that is reported to be accelerating in North America as well as in the northern English city of York (Denis Reference Denis2011), other short forms are prevalent, especially and that, and all and and all that.

Comparing these conservative northern English dialects with earlier analyses of Hull (Cheshire Reference Cheshire2007), Berwick-upon-Tweed (Pichler and Levey Reference Pichler and Levey2011) and in some cases York (Denis Reference Denis2011) reveals that both short and long forms are part of the GE inventory, regardless of place or nature of the locality or period represented by the data. Thus, while some previous research (see, for example, Aijmer Reference Aijmer2002; Brinton Reference Brinton1996) has built a picture of the GE system as one that involves increasing grammatical development of forms (e.g., and things like that > and things) as they grammaticalise over time, the results presented here continue to corroborate the alternative perspective (Pichler and Levey Reference Pichler and Levey2011; Tagliamonte and Denis Reference Tagliamonte and Denis2010), namely that there is no evidence for ongoing grammaticalisation.

Indeed, it seems likely that the English GE system has had a varied inventory of forms over an extended period of time. Perusal of the Oxford English Dictionary shows that most of the currently available variants have been attested for centuries. At the same time, it is becoming apparent that the generic component of this system is undergoing shift (i.e., that > stuff). Conservative dialect speakers in the four isolated communities in the Roots Archive as well as working-class speakers in Berwick-upon-Tweed and Hull retain a healthy proportion of variants with and (all) (that) and show scant incursion of stuff-variants.

Another attribute of the GE system is its marked dialectal dimension. The cross-dialect comparison has exposed the local aspects of this area of grammar. While a core inventory of forms is stable both across these conservative dialects and across studies (e.g., or something, and things, and everything), the individual communities in the Roots Archive each privilege one of the other forms in their inventory: and all, and all that, and that (Figure 5.9). These results suggest that further research on different varieties will offer additional new perspectives on this grammatical system. Which parts of the system are universal (in the sense that they occur consistently across varieties) and which are local (in the sense that they are unique to particular communities), and what is the linguistic mechanism underlying the difference between these types across locales?

5.9 Conclusion

The use of relic dialects as a proxy for diachronic data in the study of discourse-pragmatic change introduces a novel application of research practice that extends the time-depth for discourse-pragmatic change research. As far as methods themselves are concerned, none of the analyses I have implemented here are particularly new. They follow in line with a long-standing tradition in variationist sociolinguistics to approach linguistic phenomena systematically, employing methods that are both accountable and replicable (Labov Reference Labov1972b). The extension of quantitative methods to features beyond phonology has been going on since the early 1970s, beginning with Sankoff’s (Reference Sankoff, Bailey and Shuy1973) research on tense-aspect phenomena, indefinites and complementisers, then extending into broader areas of more quintessential discourse-pragmatic phenomena, including punctors (Vincent and Sankoff Reference Vincent and Sankoff1992), exemplification (Vincent Reference Vincent1992) and, of course, GEs (Dubois Reference Dubois1992). Within the same time-frame, many developments were concurrently underway in the investigation of the pragmatic and interactional functions of these features (see, for example, Aijmer Reference Aijmer and Togeby1985; Schiffrin Reference Schiffrin1987; Stenström Reference Stenström, Jucker and Ziv1998). What is new in more recent approaches to discourse-pragmatics is to follow through on: (i) Lavandera’s (Reference Lavandera1978: 181) suggestion to ‘relax the condition that the referential meaning must be the same for all the alternants and substitute for it a condition of functional comparability’; and (ii) Sankoff and Brown’s (Reference Sankoff and Brown1976: 650) method to include the grammatical function of a construction as well as the ‘the uses to which it is put’. In other words, variationist methods are employed but are guided by structural, functional and interactional information (see, for example, Cheshire Reference Cheshire2007; Pichler Reference Pichler2013, Chapter 3; Tagliamonte and Denis Reference Tagliamonte and Denis2010, Chapter 4; Wagner et al. Chapter 9). In essence, this practice leads to a synthesis of information that can then be used for delimiting the grammatical dimensions of the variable set of which the discourse-pragmatic variants are a part. These developments tap the best of both worlds (quantitative and qualitative) and together serve to advance the host of discourse-pragmatic features available for study. The next step will be to ensure that a consistent methodological practice is applied rigorously across datasets so that cross-variety comparisons and diachronic pathways can be meaningfully explored.

With regard to more theoretical issues, these findings offer an illuminating window into discourse-pragmatic change. First, the building evidence suggests that the nature of changes in the GE system is not a singular or linear grammaticalisation pathway as often documented in the literature. There is no apparent influence of the Constant Rate Effect (Kroch Reference Kroch1989), the well-attested pattern of parallel development of constraint effects over time. While the GE inventory seems to exhibit a cohesive set of forms and patterns, their compositional inventory varies and local (idiosyncratic) forms appear. Nor is analogical levelling, the foundational process underlying grammatical change (Kurylowicz Reference Kurylowicz1949), visible in the evolution of GEs either singularly or as a group, at least not insofar as I have been able to test for this here.Footnote ⁹ At the same time, the GE system has evidently changed dramatically over the past century as new components have emerged (stuff) and taken a prominent position within the system, at least in some locales and populations. As research on discourse-pragmatic features continues, a more comprehensive inventory of variables, time points and types of data will enrich this building picture of the nature of change in features such as GEs and their like.

6.1 Introduction

Labov’s (Reference Labov1972c: 98) ‘cumulative principle’ which states that ‘the more that is known about a language, the more we can find out about it’ crystallises in a plethora of studies on direct speech- and thought-encoding across Englishes, exemplified in (1)–(4).Footnote ¹

1. (1)

   I walked down the hill and as I was going down the hill I thought, ‘By Jove, I might be able to get into the Flying Corps. Wouldn’t it be lovely?’

   (HB-M-1889)

1. (2)

   Dad had no time for religion. He reckons it was just a big, bloody business robbing all the poor people! And he says, ‘Well, that’s her thing. She can go for her life on that so long as it doesn’t interfere with me!’

   (FB-M-1922)

1. (3)

   So you’re thinking positive and the next thing you know, they find out you’ve got a record or something. Ø, ‘Tough luck Charlie. You don’t get the job.’

   (TM-M-1970)

1. (4)

   She’s like, ‘Oh!’ and then goes off into yelling, you know, and you can’t have a decent argument any more.

   (AM-M-1970)

Quotation plays a key role in how narratives are constructed and researchers have investigated the use of quotative verbs in some depth. A wealth of mostly synchronic research into the quotative system of predominantly inner-circle Englishes has noted the emergence of innovative forms such as be like in (4) to introduce speech, sound, thought and gesture (see, inter alia, Baird Reference Baird2001; Barbieri Reference Barbieri2007; Blyth et al. Reference Blyth, Recktenwald and Wang1990; Buchstaller Reference Buchstaller2006a; Buchstaller and D’Arcy Reference Buchstaller and D’Arcy2009; Cukor-Avila Reference Cukor-Avila2002; D’Arcy Reference D’Arcy2004, Reference D’Arcy2010; Dailey-O’Cain Reference Dailey-O’Cain2000; Ferrara and Bell Reference Ferrara and Bell1995; Johnstone Reference Johnstone1987; Macaulay Reference Macaulay2001; Romaine and Lange Reference Romaine and Lange1991; Tagliamonte and D’Arcy Reference Tagliamonte and D’Arcy2004; Tagliamonte and Hudson Reference Tagliamonte and Hudson1999; Terraschke Reference Terraschke, de Beuzeville and Peters2010a; Winter Reference Winter2002). Crucially, quotative be like is not a fleeting innovation but is instead strongly entrenched in the quotative system of many contemporary varieties of English (Buchstaller Reference Buchstaller2006a; Cukor-Avila Reference Cukor-Avila2002; Levey Chapter 7; Tagliamonte and D’Arcy Reference Tagliamonte and D’Arcy2007). In my recent study of Australian English (AusE) (Rodríguez Louro Reference Rodríguez Louro2013), I have reported an adolescent peak in the use of be like, an indication of a change in progress (Labov Reference Labov2001b: 455). If be like usage in AusE is a recent innovation (as my earlier findings suggest), two questions remain: (i) what was the AusE quotative system like before the incursion of quotative be like?; and (ii) given that self-revelation has emerged across varieties of English as a discourse mode where speakers imbue their narratives with reports of their inner thoughts, attitudes and feelings (Buchstaller and D’Arcy Reference Buchstaller and D’Arcy2009; D’Arcy Reference D’Arcy2012; Ferrara and Bell Reference Ferrara and Bell1995; Tagliamonte and D’Arcy Reference Tagliamonte and D’Arcy2007), is the incursion of be like motivated by an increase in self-revelation in earlier AusE narratives?

The evolution of the English quotative system has received less attention than its synchronic variation. Notable exceptions are listed in Table 6.1. Buchstaller (Reference Buchstaller2011) and D’Arcy (Reference D’Arcy2012) performed variationist analyses on the diachrony of quotation in north-east England and New Zealand, respectively, in order to examine the quotative system of these varieties before the rise of be like. They document little variation in earlier quotative systems, with say uncontested well into the twentieth century and think emerging as an encoder of thought and attitude before the ingress of be like in the second half of the twentieth century. Crucially, both studies show marked restructuring of the quotative system over time. As D’Arcy (Reference D’Arcy2012: 360) notes, the contemporary quotative system of English results from historical evolution and is not the ‘reflex of recent lexical innovations and incursions’, a finding also underscored in the present study and only made possible through the examination of longitudinal data. Moreover, the rise of self-revelation in narrative mentioned earlier is viewed as one of the main reasons motivating the expansion of the English quotative system. Ferrara and Bell (Reference Ferrara and Bell1995: 283) observe that internal state reporting in their American English data is to be expected given Carbaugh’s (Reference Carbaugh1988) socio-cultural observation that American narrative tends towards, as they put it, the ‘lionisation of self-revelation as a preferred cultural mode’. The increase of self-revelation in narrative – and the concomitant rise in the use of quoted material – has been documented beyond American English to include Canadian, British and New Zealand English (NZE) (Buchstaller and D’Arcy Reference Buchstaller and D’Arcy2009; D’Arcy Reference D’Arcy2012; Tagliamonte and D’Arcy Reference Tagliamonte and D’Arcy2007). The question remains as to whether these changes are also attested in earlier AusE.

Drawing on spontaneous narratives of personal experience collected in Perth, Western Australia, between 1963 and 2011, this chapter offers the first longitudinal study of quotatives in AusE, a variety severely underdocumented in the variationist literature (Schneider Reference Schneider and Mesthrie2012: 349).Footnote ² A singular real-time depth of forty-eight years and a sample of speakers aged thirty-five to ninety-three show that the Australian quotative system has undergone dramatic changes from the nineteenth to the twenty-first century and that – in line with D’Arcy’s (Reference D’Arcy2012) findings for NZE – the quotative system was in flux well before the arrival of be like. These findings contribute an original perspective to current knowledge about the expansion of the quotative system in AusE. They suggest that the functional niche of internal thought- and attitude-encoding was already in place by the time be like entered the system in the late twentieth century, which, in turn, may explain why be like’s expansion was so rapid. The analysis also reveals variability in the use of quotative frames across time: a highly invariable system dominated by say giving way to patterned variability in how quotative introducers, grammatical person and tense are deployed in narratives. Moreover, variation in AusE quotatives across time is socially constrained. These results are important in highlighting the complex nature of discourse-pragmatic variation and change.

The chapter is organised as follows. The data and methods used in this study are presented in Section 6.2, followed by the presentation of the longitudinal results in Section 6.3. The chapter ends with a discussion and conclusion in Sections 6.4 and 6.5.

6.2 Data and methods

6.3 Results

6.3.1 Overall distribution of variants

To examine the distribution of quotative variants before the onset of be like and to establish whether the AusE quotative system has changed across time, let us first consider in Figure 6.1 the overall usage frequency of quotative frames in the West AusE speaker sample introduced in Table 6.2.Footnote ⁵ Figure 6.1 shows that, across time periods and in line with previous research (Blyth et al. Reference Blyth, Recktenwald and Wang1990; Buchstaller Reference Buchstaller2011; Buchstaller and D’Arcy Reference Buchstaller and D’Arcy2009; Cukor-Avila Reference Cukor-Avila2002; D’Arcy Reference D’Arcy2004, Reference D’Arcy2012; Ferrara and Bell Reference Ferrara and Bell1995; Tagliamonte and Hudson Reference Tagliamonte and Hudson1999; Tannen Reference Tannen and Coulmas1986), say is the main quotative variant used. These results are noticeably different from those reported in recent comparable research into the contemporary AusE quotative system. Rodríguez Louro (Reference Rodríguez Louro2013) reports that be like is the most widely used quotative variant among Australian speakers aged eleven to twenty-six (born between 1985 and 2000), accounting for 80% of all quotatives used by speakers in this age cohort. The results presented in Figure 6.1 illustrate a vastly dissimilar system for AusE speakers born between 1870 and 1980, with quotative say dominating across cohorts and time periods.

Figure 6.1 Distribution of quotative variants used by adult Anglo-Australians born between 1870 and 1980

Example (24) illustrates typical uses of quotative say in late nineteenth century Anglo-AusE. A usage unattested in the later AusE data is that of say after the quotation (see also (17) in Section 6.2.2.2). This usage is most frequent in the speech of those born between 1870 and 1890 where it accounts for 11% (N = 42/392) of all quotatives used (see D’Arcy [Reference D’Arcy2012] for similar findings in NZE). As noted in Section 6.2.2.2, because post-posed syntactic position is restricted to say, these tokens were not included in the multivariate analysis presented in Section 6.3.2. However, a brief note on their usage is in order. With time, and in line with D’Arcy’s (Reference D’Arcy2012) findings for NZE, quotative post-posing has become obsolete and is absent from contemporary conversational AusE. We can witness a steady decrease in its use from the late nineteenth to the mid-twentieth century: 11% (1870–1890) > 4% (1920–1930) > 0% (1950–1960). This development suggests an important change in how direct speech is framed in AusE narratives.

1. (24)

   On this particular day he said, ‘We are going to have company drill.’ He said, ‘I’m going to make a few purposeful mistakes and I want you to tell me when it’s over what mistakes I’ve made.’ Well, he made some palpable errors and I said, ‘Oh, there’s another mistake, Captain Thompson.’ He said, ‘And what was that?’ ‘Well,’ I said, ‘The company was in such and such a position and you gave a certain order which was quite incorrect.’ ‘Oh,’ he said, ‘Was it?’ So I thought, ‘I put my foot in it!’

   (JT-M-1886)

6.4 Discussion

In line with previous research across varieties of English, the development of the AusE quotative system is such that say as the canonical introducer of direct speech by third parties is slowly but steadily replaced by a system featuring a wealth of quotative frames encompassing zero, think, go and – to a lesser extent – be like. The systems of those born between 1870 and 1980 reflect a trajectory of change which speaks of overarching shifts in how the variable grammar operates. Studying the quotative system as a whole allows for a nuanced understanding not only of why say shifts its sociolinguistic constraints with the passing of time but also of how these shifts ultimately influence the rise and fall of other quotative frames, notably the incomer be like, and their sociolinguistic conditioning. What are some of the forces that triggered these changes?

Close inspection of various developments in the data examined in Section 6.3 suggests a threefold answer. Firstly, the ingress of quotatives other than say is linked to a steady increase in the expression of internal thought. Secondly, not only does storytelling increasingly feature more internal thought expression, the way quotatives are deployed is crucially intertwined with the organisation of quoted content in narrative discourse. Thirdly, changes in the AusE quotative system –in terms of both the number of quotative variants used and the pragmatic contexts of their usage – are socially embedded. Each of these hypotheses is explored in detail here.

Regarding the first hypothesis, the data analysed in this chapter suggest that the decrease of say and gradual incursion of other variants, notably be like, in contemporary AusE (Rodríguez Louro Reference Rodríguez Louro2013) is motivated by the increase in self-revelation in narratives referred to in Section 6.1. Figure 6.2 depicts the relative frequency of quotative frame-introduced direct speech vs. quotative frame-introduced internal thought among Anglo-Australians born between 1870 and 1980. The results reveal that the changes in quotative variant choice and preference were accompanied by changes in the type of content introduced by quotative frames, namely a rise in the encoding of internal thought across time. Once negligible, thought-reporting represents a sizeable portion of direct quotation among those born post-1950 (in line with D’Arcy Reference D’Arcy2012: 367, fn. 18). The finding that internal thought-reporting was already in place when be like entered the AusE system is important given that be like is believed to have entered the English quotative system as an introducer of internal thought (Buchstaller Reference Buchstaller2014; Butters Reference Butters1982; see also D’Arcy Reference D’Arcy2012: 360). As such, the contemporary AusE quotative system is the result of rich historical evolution.

Figure 6.2 Relative frequency of quotative frame-introduced direct speech and internal thought among adult Anglo-Australians born between 1870 and 1980

Regarding the second hypothesis, the construction of narrative discourse is important to the development of the (AusE) quotative system. Two linguistic variables are intrinsically related to the organisation of complicating narrative clauses (in the sense of Labov and Waletzky Reference Labov, Waletzky and Helm1967): grammatical person and tense (Levey Reference Levey2006b). Examination of their effect on say usage in Section 6.3 revealed key differences between the speakers born before and after the mid-twentieth century, especially those born after 1970. Once overwhelmingly associated with third-person grammatical subjects, say is favoured with first-person subjects among those born after 1970; once restricted to the past tense, the grammars of speakers born after 1970 show a more complex tense-marking system, and say occurs in non-past tense contexts. The implications of these findings are best understood by focusing on how information is organised in narrative discourse. Rodríguez Louro and Ritz (Reference Rodríguez Louro and Ritz2014) analysed tense variation in the complicating clauses of personal narratives (the ‘then, what happened?’ portion of narratives, Labov Reference Labov1972a: 359) and established the following pattern: in contemporary AusE narrative discourse, the first person and the past tense introduce information pertaining to the narrator while the third person and the historical present encode situations relating to individuals other than the narrator. This trend is also evident in the longitudinal dataset examined in this study. Examples (31) and (32) contain vastly differently constructed narratives by two Anglo-Australian males: (31) was produced by a speaker born in 1922 where past tense-marked first- and third-person say is used to organise the narrative; (32) was produced by a 1970-born speaker where first person-marked say encodes quoted speech by the narrator, and go in the historical present introduces quoted speech by a third party. In this sense, AusE say does not only decrease in overall usage frequency across time periods (see Section 6.3) but, importantly, it narrows its pragmatic scope to encode first person-marked speech in the narratives of those born post-1970. These changes affect the usage and frequency of say’s competitor variants.

1. (31)

   Well, I used to help a chap named John Edmondson, and he used to go up to the house to get his shovel and axe. He said, ‘I’ll be down shortly,’ and probably about quarter of an hour or half an hour later he’d come down, and I’d be digging away there. But I just didn’t like it, especially when I got down towards the box. I was frightened I’d go through the box or something. So I said to Will Smith, he was the superintendent at the time … He said to me one morning that I’d be going on the digging. I said, ‘Oh no, Will, I don’t like the digging. I don’t want to go on.’ So I went off and he came out and saw me later on and said to me, ‘Why wouldn’t you go on the digging?’ I said, ‘Well, I don’t like it, Will.’ And he said, ‘Oh well, you’re no good to me.’ And I said, ‘Oh well, I don’t care, Will.’ So he went off. He didn’t seem to be too concerned really, and I didn’t have to go digging, except if they were busy I went and helped them to fill in the graves, because everybody, everybody, had to go and fill in the graves.

   (ES-M-1925)

1. (32)

   And I went to the boss and I said, ‘I’d like to see the boss, please.’ I said, ‘I just want to see him. Now.’ And she goes, ‘Okay, hang on.’ I’m in there, and he goes, ‘What can I do for you?’ I said, ‘I resign.’ He goes, ‘What?’ And I said, ‘I resign.’ And he goes, ‘Why?’ And I said, ‘Because that bitch down there is giving me the shits, going off my face.’ And he goes, ‘Would you like to reconsider your decision here?’ and I said, ‘No, I wouldn’t.’

   (TM-M-1970)

Given these findings, the importance of discourse context in the development of the English quotative system cannot be underestimated. One of the most salient syntactic contrasts attested in narrative discourse includes the use of direct speech. The more dramatic the narrative, the more likely the storyteller is to rely on direct speech reporting where ‘the reporter-speaker plays the role of the reported/original speaker’ (Li Reference Li and Coulmas1986: 38). The results of the present and previous longitudinal studies of English quotation (Buchstaller Reference Buchstaller2011; D’Arcy Reference D’Arcy2012) show important shifts in the type of quoted content introduced by speakers (e.g., speech vs. thought). The findings presented here offer a further contribution by considering discourse organisation more generally, addressing the important question of how narrative structure itself constrains quotative usage (see Rodríguez Louro Reference Rodríguez Louro2015).

The third and final hypothesis explored here concerns the social embedding of linguistic change in the English quotative system. As shown in Table 6.7, professional speakers born between 1870 and 1950 show a slight preference for say over other quotative variants. Women in this study also tend to prefer say over other quotatives, a (non-significant) trend operative until 1970. These results are in line with D’Arcy’s (Reference D’Arcy2012: 354) findings for NZE where women born between 1890 and 1935 significantly favour say over other quotative verbs. They support the view that innovation in the English quotative system was brought about by males, with females ‘catching up’ later on. However, given the non-significant findings across time periods, further research is needed. Sex and occupation interact in important ways. Firstly, although non-significant, Anglo-Australian women born before 1960 prefer say, the unmarked or ‘standard’ quotative variant (see D’Arcy Reference D’Arcy2012: 367, fn. 13) (see Table 6.7). Coincidentally, the occupation constraint whereby professionals (significantly) favour say disappears among those born before 1950. These findings indicate a shift in the social evaluation of quotative variants in early Anglo-AusE: once a prestigious form preferred by professionals and women, in the middle to late twentieth century say is no longer the sole favoured quotative frame in the system. Secondly, as noted in Section 6.3.2, non-professional males born in the early twentieth century contribute to variability in the AusE quotative system through their use of the zero variant. This, too, is partly in line with D’Arcy’s (Reference D’Arcy2012: 354) findings for NZE where men born between 1890 and 1935 (occupation unspecified) are responsible for an increase in the use of the zero variant. These trends suggest that what starts off as a virtually invariant system with say encoding direct speech gradually expands to include further quotative introducers (e.g., zero), and males play a central role in this expansion. The role of the non-professional male in the evolution of AusE should not be underestimated. Kiesling (Reference Kiesling and Hickey2004: 422) notes that the language of the predominantly working-class, male population of early colonial Australia constituted a source of covert prestige, a trend likely to have continued well into the late nineteenth century and, arguably, beyond. These findings are especially noteworthy when compared to Buchstaller’s (Reference Buchstaller2014: 169) claims that middle-class Tyneside English speakers in her 1960s/1970s data were responsible for the expansion of the quotative system, i.e., the use of quotative variants other than say. In this study, occupation proved useful in the identification of change in progress, especially in analysing quotation as a primordial feature of narrative.

6.5 Conclusion

Drawing on a previously unexplored corpus of archival and recently collected naturalistic interaction, this longitudinal study has established that quotation in Anglo-AusE has undergone significant changes in a period of some one hundred years. The results are in line with what has been shown for NZE (D’Arcy Reference D’Arcy2012), offering support for the claim that ‘there is a strong linguistic connection between New Zealand and Australia [and that] varieties with similar structural features may be linked by history even if separated by distance’ (Kiesling Reference Kiesling and Hickey2004: 432). The parallels arising from this research are noteworthy in that they indicate similarities in how discourse-pragmatic change operates across geographically distant speech communities.

The longitudinal nature of the original data investigated here, spanning forty-eight years of data collection and including speakers born between 1870 and 1980, has allowed us to trace the trajectory of change in the Australian quotative system, identifying shifting forms and revealing new insights into the AusE quotative system before the encroachment of be like late in the twentieth century. This longitudinal perspective is important to understanding whether what may be seen as recent innovations (e.g., the upsurge of be like) are rooted in the historical evolution of the system as a whole. Such historical evolution is best illustrated here in the rise of internal thought-encoding across time. The findings reported in this chapter underscore the importance of discourse context in understanding language change.

7.1 Introduction

Much of the foundational research on children’s acquisition of sociolinguistic patterns of variation and change has paid specific attention to phonological variables (see, for example, Foulkes et al. Reference Foulkes, Docherty and Watt2005; Roberts Reference Roberts1996, Reference Roberts1997), with an expanding number of studies addressing morpho-syntactic variables (see, for example, Kovac and Adamson Reference Kovac, Adamson, Sankoff and Cedergren1981; Miller and Schmitt Reference Miller and Schmitt2010; Smith et al. Reference Smith, Durham and Fortune2007; Youssef Reference Youssef1999). While cumulative advances in developmental sociolinguistics are refining our understanding of how children acquire variables situated in different grammatical domains (see, for example, Smith et al. Reference Smith, Durham and Richards2013), this line of inquiry remains hindered by the dearth of accountable quantitative studies of the acquisition of discourse-pragmatic variation.

An understanding of how children acquire, replicate and employ community-based patterns of discourse-pragmatic variation poses a number of methodological challenges. A first requirement is the availability of a large compendium of naturally occurring child language as well as a suitable adult control variety with which to compare child and adult usage patterns and to track children’s acquisition of community-based norms. A second requirement calls for access to an analytical approach capable of teasing apart the multiple factors that condition the acquisition process. Among the myriad factors to be considered are the age of the child (Kerswill Reference Kerswill1996); the type of variable being acquired (Smith et al. Reference Smith, Durham and Richards2013); its relative stability or engagement in change (Cameron Reference Cameron2005); the complexity of its sociolinguistic conditioning (Meyerhoff and Schleef Reference Meyerhoff, Schleef and Lawson2013; Roberts Reference Roberts, Chambers and Schilling2013); as well as the possible existence of dialect-specific constraints on variation (Rutter Reference Rutter2013). In view of the multiple factors that constrain variable use, many of which differ from one variable to another, it seems reasonable to hypothesise that the acquisition of discourse-pragmatic variation will not necessarily parallel that of phonological or morpho-syntactic variation. The investigation of this hypothesis constitutes a major impetus for the research described in this chapter. An additional motivation for the work reported here comes from the limited number of studies addressing the acquisition of changes in progress, as opposed to stable variation (Rutter Reference Rutter2013).

Focusing on the quotative system, I examine maturational constraints on children’s acquisition of discourse-pragmatic variation and situate children’s usage patterns with respect to baseline community norms. The quotative system is an ideal site to test the hypothesis described earlier because it typically accommodates multiple variants, and it is widely acknowledged to be the locus of rapid change and reorganisation (see, for example, Buchstaller Reference Buchstaller2004, Reference Buchstaller2011; D’Arcy Reference D’Arcy2012; Ferrara and Bell Reference Ferrara and Bell1995; Rodríguez Louro Chapter 6; Tagliamonte and D’Arcy Reference Tagliamonte and D’Arcy2007). In the course of the past three decades, many varieties of English have witnessed the exponential rise of quotative be like, reported to be one of the most aggressive changes in the recent history of the language (Labov, p.c., 2000, cited in Cukor-Avila Reference Cukor-Avila2002: 21–2). A cursory glance at the strategies used by Canadian children (1)–(6) and adults (7)–(12) to report speech and thought confirms that both speaker groups make use of this innovative quotative alongside a number of competing variants, including the null or zero variant represented by Ø in (5):

The spread of quotative be like throughout mainstream urban varieties of Canadian English is reported to exhibit distinctive age-correlated patterns of use: older adolescents are spearheading this change, with preadolescents using the incoming form less frequently (Tagliamonte and D’Arcy Reference Tagliamonte and D’Arcy2009: 82–3).Footnote ² Preadolescents’ lower rate of use is generally explained by appealing to the fact that children typically have had less time to accrue changes in progress. As children mature and their vernacular stabilises in late adolescence, they are expected to contribute to the ‘logistic incrementation’ of linguistic change by advancing the use of innovative features (Labov Reference Labov2007). The question of how children acquire and propagate ongoing changes merits further research, not least because this issue remains ‘a problem that lies at the forefront of […] efforts to understand the mechanism of linguistic change’ (Labov Reference Labov2012: 276).

Linguistic change can be tracked developmentally by comparing the stages that children pass through during their acquisition of a particular feature (Romaine Reference Romaine, Breivik and Jahr1989: 61). As I will show, it is particularly instructive to pursue such comparisons using narrowly defined age ranges capable of revealing differences which may be concealed by more broadly defined age cohorts (Llamas Reference Llamas, Llamas, Mullany and Stockwell2007: 75). Combining a developmental perspective with apparent- and real-time components to assess what has changed, I focus on: (i) children’s acquisition of social and linguistic constraints on quotative variation; (ii) the extent to which these constraints parallel those operating on the same variable in the corresponding adult community; and (iii) the role of children in sustaining and propagating change in the quotative system. In order to address these objectives, I make use of a number of complementary data sources. These include synchronic corpora of child and adult speech supplemented by real-time data, affording a diachronic perspective on the rise of be like and its progressive infiltration of the quotative system.

In the following sections, I first review in Section 7.2 the literature addressing children’s acquisition of variation and the emergence in childhood of strategies for reporting speech and thought. In Section 7.3, I supply information about the data on which this study is based before describing the theoretical framework as well as the methodological procedures underpinning the analysis. After presenting the results of the study in Section 7.4, I examine in Section 7.5 their contribution to addressing the research objectives before offering my conclusions in Section 7.6.

7.2 Acquiring sociolinguistic competence

7.3 Data and method

7.4 Results

7.5 Discussion

Innovative in the research reported here is the application of the variationist framework to tracking discourse-pragmatic change in later childhood, illuminating the transition between different stages of the grammar within a narrowly defined age range. A key component of this investigation involves the judicious use of a number of complementary data sources to track change in real and apparent time, and to situate children’s usage in relation to a commensurate adult control variety. A lynchpin of the methodological framework underpinning this investigation is the importance attached to variable structure, rather than rates of variant occurrence, as the primary heuristic for comparing the underlying grammar across different speaker groups. This approach is additionally informed by the recognition that an adequate understanding of how change is acquired can only be achieved when the incoming variant, be like, is situated in relation to the larger system of which it is part (Poplack and Malvar Reference Poplack and Malvar2007).

A major finding to emerge from the comparative component of this research is that although the primary exponents of the quotative system are essentially the same in the preadolescent and adult data, their rates of occurrence and associated conditioning differ in a number of important respects. An especially intriguing pattern concerns the existence of a sex effect associated with be like in the youngest age group, superficially mirroring a similar (though stronger) effect in the young adult data. At first sight, it is tempting to construe this effect in the eight- to nine-year-olds as an instantiation of the classic pattern of male retreat from female-led changes (Labov Reference Labov2001b), but the neutralisation of the same effect in the eleven-to-twelve age group militates against this interpretation. A more likely explanation is developmental: the eight- to nine-year-old males lag behind the females in acquiring be like but males subsequently increase their use of this variant in later childhood (see Table 7.5). What is striking in these data is the lability of sex effects: it is not just the contrast between the eight- to nine-year-old males and females in their rates of be like that is remarkable but also the rate increase that emerges when the males in the youngest age group are compared with males in the eleven-to-twelve cohort. Furthermore, differences between these two age groups in overall rates of be like are at odds with the claim that ‘the more diffused be like is, the more likely it is to differentiate male and female speech’ (Tagliamonte and Hudson Reference Tagliamonte and Hudson1999: 167). These discrepancies indicate that there is no constant relationship between the frequency of be like and sex-differentiated patterns of variation, at least when child and adult populations are compared, nor is there any simple correlation between the emergence of a sex effect in preadolescence and an equivalent effect in the corresponding adult community.

Among the major assets of the methodological approach adopted in this study are the refined measures it contributes to elucidating the extent to which preadolescents engage in and propagate discourse-pragmatic change. Comparison of constraint hierarchies across the preadolescent age cohorts provides key information in this regard. It is evident that the youngest children examined here have acquired more than just the form be like, as they have also internalised a complex set of linguistic constraints governing its use. Furthermore, not only are the eight- to nine-year-olds contributing to the incrementation of change, they are also extending its scope and specificity (Labov Reference Labov2007: 346). This information can be inferred from measures of grammaticalisation operationalised in this study. These reveal that the youngest children, along with their older peers, are advancing the use of be like in specific environments (i.e., temporal reference and grammatical person) associated with the progression of this variant along the cline of grammaticalisation (Tagliamonte and D’Arcy Reference Tagliamonte and D’Arcy2004, Reference Tagliamonte and D’Arcy2007). Such findings are not altogether unexpected considering that by the age of eight, restructuring of the vernacular in favour of the linguistic norms of the ascendant peer-based social order is liable to be well underway.

However, on other key measures, i.e., content of the quote and mimesis, the linguistic behaviour of the youngest children sets them apart from the older children who are more closely aligned with the adults, as indicated by correspondences in variant rates as well as the linguistic conditioning of variant choice. An important inference to be drawn from differences in quantitative patterning between the preadolescent age groups is that the grammar associated with the quotative system has not stabilised by the age of eight and remains malleable beyond this age, as demonstrated by the results of this investigation. This inference bolsters the general conclusion that the acquisition of adult-like patterns of quotative variation (and change) is a developmentally protracted process. The prolonged nature of this process points to fundamental differences between the acquisition of discourse-pragmatic variation and phonological variation. Whereas children are claimed to acquire the bulk of the phonology of their local variety by the age of six or seven (Kerswill Reference Kerswill1996: 190), the acquisition schedule for discourse-pragmatic variation extends well into later childhood. Among the possible explanations for this difference is the complexity of the syntax-discourse interface, reported to be the locus of learnability issues in language acquisition (Sorace Reference Sorace2004). Because successful use of quotative markers draws on different areas of linguistic competence transecting discourse and syntax (Meyerhoff and Schleef Reference Meyerhoff, Schleef and Lawson2013), integration of this knowledge may not be fully internalised before adolescence. Whatever the precise explanation, the hypothesis enunciated at the outset of this study, namely, that the acquisition of discourse-pragmatic variation will not necessarily resemble that of other variable components of the grammar, is sustained rather than refuted by the results reported here.

7.6 Conclusion

The findings uncovered in this study generate a number of additional questions for future research. An outstanding issue concerns the origins of the complex set of conditions on be like instantiated by the children in the present investigation. If, in the absence of a parental model, be like is acquired from the peer group and older children, when is this process initiated? At what age is be like first acquired and how is it conditioned? Resolution of these issues can only be achieved by extending the focus of the investigation to early childhood in order to probe the developmental roots of quotative variation (Levey and McIntyre Reference Levey and McIntyreMS). Another key, though understudied, question concerns the impact of inter-dialectal differences on the acquisition of change. In other communities where be like has made less dramatic inroads into the quotative system, earlier research found that neither the form nor its associated structure had been acquired by preadolescents to any appreciable extent (Levey Reference Levey2007). This raises the critical question of how quantitatively robust (and socially salient) a change needs to be in the speech community before children can begin to acquire it. It is not entirely clear to what extent constraints on incipient changes, as opposed to those operating on mid-range and advanced ones, are accessible to young children (see Cameron Reference Cameron2005). Engagement with these issues is a prerequisite for gaining new insights into children’s participation in and propagation of community-based change, with a view to acquiring a deeper understanding of how children ‘identify the newer pattern in the community system that they are learning, adopt that pattern and then move further in that direction’ (Labov Reference Labov2012: 276). The empirically accountable framework adopted in this chapter is intended as a contribution to addressing these issues.