4.1 Distributional analysis

Table 2 provides a measure of intensification overall, by combining all intensifiers together in comparison with nonintensified contexts.

Table 2.

Overall distribution of intensification in Toronto English

The table reveals that out of 9,905 adjectival heads 36.1 per cent were intensified. This is considerably higher than British English, 24 per cent (Ito & Tagliamonte Reference Ito and Tagliamonte2003: 265) and American English, 22 per cent (Tagliamonte & Roberts Reference Tagliamonte and Roberts2005: 287). The heightened rate of intensifier use in Toronto is unlikely to be due to genre differences since the British data in Ito & Tagliamonte (Reference Ito and Tagliamonte2003: 265) come from materials comparable to those that constitute the Toronto English Corpus (i.e. spoken vernacular data). Another hypothesis might be that the Toronto data come from a broader range of speakers, many of them well under 30.

Figure 1 tests this possibility by plotting the overall distribution of intensification by speaker age.Footnote ⁵

Figure 1.

Overall distribution of intensification by age

The figure reveals that the frequency of intensification is generally levelled across age groups with only a small difference between the <13-year-olds (38%) or the 17- to 19-year-olds (37%) and the 50+ age group (29%, 27%).Footnote ⁶ Thus, it is necessary to delve deeper into the intensifier system to assess linguistic change. I begin with the nature of this system's lexical inventory.

Table 3 shows the distribution of intensifiers in Toronto English according to lexical items that occurred ten times or more.Footnote ⁷ In interpreting the results of table 3, I will put the results in context with the historical record since this has an important bearing on how they should be interpreted.

Table 3.

Distribution of intensifiers by lexical item (N ≥ 10)

The most frequent intensifier by far is really, which occurs with a full 13 per cent of the intensifiable adjectives in the data. This corroborates Labov's (Reference Labov and Schiffrin1985: 44) observation that really is ‘one of the most frequent markers of intensity in colloquial conversation’ in American English.

Interestingly, the OED does not list attestations of really as an adverb meaning ‘very, extremely’; however, adverbial use of real with this function can be found from the early 1700s, as in (6a–b). It is also found in contemporary varieties in England, as in (6c–d).

1. (6)

   1. (a)

      An Opportunity of doing a real good Office.

      (c.1718, J. Fox, Wanderer No. 17. 116; OED s.v. real adv. B.1)

   2. (b)

      It looks real nice.

      (c.1885, G. Allen, Babylon vi; OED s.v. real adv. B.1)

   3. (c)

      It makes a real funny story.

      (Maryport, Northwest England)

   4. (d)

      We had a real nice hotel. It was cheap, but very nice.

      (York, England)

While the OED suggests that the origins of real in this use are from Scots and that it is ‘not common in standard use in southern England’, in contemporary British English the related form really Footnote ⁸ is reported to be the most common premodifier of adjectives amongst teenagers in London (Strenström Reference Stenström, Hasselgård and Okesfjell1999). In fact, the ratio of adjective intensifier really is much higher in the COLT corpus than in the BNC (Lorenz Reference Lorenz, Wischer and Diewald2002: 153, table 1). The incidence of intensifier really is also high among 17- to 34-year-olds in York (Ito & Tagliamonte Reference Ito and Tagliamonte2003: figure 2) and in contemporary New Zealand English it ranks among the frequent reported intensifiers in 11- to 12-year-olds (Bauer & Bauer Reference Bauer and Bauer2002: 245). Toronto English appears to go along with these recent trends towards frequent use of really.Footnote ⁹

Figure 2.

Frequency of main intensifiers across quantitative studies

In contrast, table 3 shows that the frequency of very is low, occurring with 6.6 per cent of the intensifiable adjectives. The low rate of very is notable since it is thought to be the most completely delexicalized intensifier in contemporary English (Lorenz Reference Lorenz, Wischer and Diewald2002: 146). It is also the oldest of the major forms in this dataset. As discussed earlier, its use as an intensifier can be found as early as the fourteenth century with an upswing in frequency in the sixteenth century continuing into Early Modern English. Very is also often reported to be the most frequent intensifier in contemporary English (Fries Reference Fries1940: 201; Bäcklund Reference Bäcklund1973: 290; Biber, Johansson, Leech, Conrad & Finegan Reference Biber, Johansson, Leech, Conrad and Finegan1999: 565–7); however, it is important to put this into a context of time, variety and genre. While very is the leading intensifier for contemporary spoken British English in York, in northeast England in the late 1990s, i.e. 38.3 per cent (Ito & Tagliamonte Reference Ito and Tagliamonte2003: 266) (see also table 3), an analysis of its distribution across the population by age demonstrated that it was declining rapidly. The comparatively low rate in Toronto corroborates an interpretation of a general twentieth-century waning of very in favour of really.

Perhaps the most interesting observation about table 3 is the comparatively high rate of intensifier so, as in (8a–c). At 6.1 per cent, so vies for position with very at 6.6 per cent, shaded.

1. (7)

   1. (a) He'd get mad at me, so mad at me. (TOR/q)

   2. (b) Like my mind is so focused and it's like actually the only time my mind is so completely focused. (TOR/&)

   3. (c) I watched it and it was so boring. (TOR/s)

However, an intensifier function for so can also be traced back to earlier times in English. How far back is unclear as many of the earliest uses modifying adverbs or predicative adjectives cannot be replaced by very without considerably changing the sense. The earliest unambiguous examples I have found are from the early to mid 1800s, as in (8a–b). Note too that in these transcripts so is italicized, adding further support to the intensifying interpretation.Footnote ¹⁰

1. (8)

   1. (a)

      I'm so afraid you'll catch cold, aunt – have a silk handkerchief to tie round your dear old head.

      (1837, Dickens, Pickwick Papers)

   2. (b)

      If we walked to or from a party, it was because the night was so fine, or the air so refreshing, not because sedan-chairs were expensive.

      (1853, Mrs Gaskell, Cranford)

An unexpected finding in table 3 is the frequency of pretty, as in (9a–c). At 5 per cent it competes robustly with both very and so, shaded.

1. (9)

   1. (a) Girls are pretty smart at my school. (TOR/c)

   2. (b) It was a pretty crazy thing to do in the first place. (TOR/=)

   3. (c) Mr. and Mrs. Price were pretty good stuff to take in a nineteen year old. (TOR/£)

The use of pretty is apparently a North American phenomenon, where it occurs nearly four times more frequently than in British English (see Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 567, table 7.13). However, this is a form that can be tracked back in time as well. First attested as an adverb modifier in the sixteenth century, it too is a form that exhibits a rise in frequency in Early Modern English (Nevalainen Reference Nevalainen, Lindqvist and Mair2004: 21). Intensifying uses from the nineteenth century are easily found, as in (10).

1. (10)

   1. (a)

      Now Miss Jenkyns was daughter of a deceased rector of Cranford; and, on the strength of a number of manuscript sermons, and a pretty good library of divinity, considered herself literary . . .

      (1853, Mrs Gaskell, Cranford)

   2. (b)

      Mr. Winkle, thus admonished, abruptly altered his position, and in so doing, contrived to bring the barrel into pretty smart contact with Mr Weller's head. ‘Hollo!’ said Sam, picking up his hat, which had been knocked off, and rubbing his temple.

      (1837, Dickens, Pickwick Papers)

In the twentieth century pretty has been observed to be among the most frequent intensifiers, but it is unique in that it is able to express ‘moderate degree’ as well as the sense of boosting meaning (Stoffel Reference Stoffel1901: 147–53). It is often not easy to discern whether it is being used as an intensifier or as a downtoner (Nevalainen & Rissanen Reference Nevalainen and Rissanen2002: 369) and many researchers treat it as the latter (Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 567, table 7.13; Nevalainen & Rissanen Reference Nevalainen and Rissanen2002: 369). In the Toronto speech community pretty scales the quality of the adjective upwards although with perceptually less thrust than other intensifiers. As we shall see, perhaps due to this very quality, it plays an important role in the intensifier system.

There may, in fact, be contrasts in intensifier use from one major variety of English to another. Recent research on the intensifier system using the same methodology as that employed in the present study permits such a comparison. Figure 2 shows the frequency of use of very, really and so across British, American and Canadian varieties.Footnote ¹¹

The same three intensifiers are frequent – really, very and so. However, these forms occur at different orders of frequency. In the York data very is the most frequent intensifier followed closely by really while so represents only a small proportion of the intensifiers used, only 10.1 per cent. In the Toronto data, really is the most frequent intensifier while very and so are near parallel in use. One interpretation of the difference across these studies is that the data represent different stages in the cyclic evolution of English intensifiers (see Tagliamonte and Roberts Reference Tagliamonte and Roberts2005). York English represents an earlier stage in which very was dominant, but Toronto English is more advanced along the trajectory of change with really in an advanced position in the system and so rising. The pre-eminence of so stands out in Tagliamonte & Roberts' (Reference Tagliamonte and Roberts2005) study of the American television series Friends: it represents a striking 44.1 per cent of the intensifiers used. The authors argue that this heightened use was indicative (i) of the rise of so as the new favourite intensifier in North America and (ii) that the medium of television was mirroring this trend while at the same time pushing it forward.

In comparing across these studies it is important to consider the contrast in data composition. The York data comprise material from speakers aged 17−90+ collected in 1997 (Ito & Tagliamonte Reference Ito and Tagliamonte2003: 263); the Toronto data comprise material from speakers aged 9−90+ collected in the early 2000s; the Friends data comprise material from six actors in early adulthood (20s to 30s) collected between 1994 and 2002 (Tagliamonte & Roberts Reference Tagliamonte and Roberts2005: 285). Thus another reason why so is conspicuous in the Friends data is because the speakers come from a single age group, i.e. young adults. Moreover, the television scenario in which the actors are performing champions a trendy expressive style. In fact, Tagliamonte & Roberts (Reference Tagliamonte and Roberts2005: 296) suggest that it may herald developments to come.

In earlier research Ito & Tagliamonte (Reference Ito and Tagliamonte2003: 260) noted that the diachronic recycling and renewal of different intensifier forms could be related to their synchronic distribution patterns in apparent time. Thus, for example, older forms, such as very which arose in the fourteenth century, were less frequent than a form like really which arose some 400 years later.

Yet the recycling and renewal process in intensifiers is said to take place quickly. Because the Toronto data comprise input from speakers born as early as the turn of the twentieth century to those born nearing the end, it may be possible to tap into the recycling process by plotting the frequency of individual intensifiers in apparent time. Figure 3 categorizes the speakers of the community into age groupings that have previously been found relevant for linguistic change in community-based corpora (Labov Reference Labov2001). The proportion of use of each of the four main intensifiers out of the total number of intensifiable adjectives is shown for each age grouping.

Figure 3.

Overall distribution of main intensifiers by age

The apparent time perspective in figure 3 reveals a remarkable topography in which the intensifiers and their relationship to the speech community are unveiled. First, it confirms that these four main intensifiers are used by speakers of all ages in the city of Toronto, exposing the inherently layered system. However, there is a tremendous disproportion in the nature of the layering of these forms depending on the age of the speaker. Very is the most frequently used intensifier in the speakers over 50, but declines rapidly thereafter particularly among speakers under 30. In contrast, really increases steeply from oldest to youngest speakers, with a peak among the 20- to 29-year-olds. Recall the distinct hump in figure 1. While so is a minor variant, it exhibits an incremental increase from the 50+ age group to the 9- to 29-year-olds but is most frequent among the 13- to 29-year-olds. A similar pattern is found for pretty. Overall, these trends confirm that the intensifiers in this community are changing at top speed. Indeed, the community patterns naturally into three quite distinct intensifier systems as indicated by the broken vertical lines: the 13- to 29-year-olds have a system which prefers really; the 30- to 49-year-olds have a mixed system in which both really and very are dominant; and among the 50+ age group very is the favoured intensifier.

The first question to consider is: what processes underlie the trends? Is it delexicalization or some other type of grammaticalization phenomenon? The linguistic mechanisms that enable the delexicalization of words that become intensifiers do not happen catastrophically. Instead, this occurs through a step-by-step process, beginning with metaphorical (Peters Reference Peters and Kastovsky1994) or metonymic (Traugott Reference Traugott, Kemenade and Los2006; Paradis this issue) extension from an intensifier's original lexical meaning, to conventionalized uses with particular adjectives, to diffusion across diverse types of adjectives. The panorama of change visible in figure 3 suggests that the processes through which these changes are happening may also be accessible. Therefore the next question to address is whether any of the stages by which this delexicalization is known to happen can be viewed in the distributional differences between the generations in Toronto.

4.1.1 Adjective function

Perhaps the most straightforward delexicalization pathway attested for intensifiers is their patterning according to adjective function (e.g. Mustanoja Reference Mustanoja1960: 326–7). According to some researchers, a later stage in the development of intensifier very was when it came to modify predicative adjectives. Extrapolating from these observations, a comparison of intensifier use between predicative and attributive adjectives could be taken as evidence for the delexicalization process. Higher frequency of use with predicative adjectives or equal distribution across both might reflect a later point in an intensifier's development.

Such hypotheses can be tested by comparing the distribution of intensifiers across types, attributive adjectives, as in (11), and predicative adjectives, as in (12).

1. (11) Attributive:

   1. (a) Yeah, very very inflated beer prices at some places. (TOR/2j)

   2. (b) They didn't feel comfortable with sort of the risk of buying a really big piece, a really big house and property. (TOR/I^)

2. (12) Predicative:

   1. (a) My mom said that it's really dangerous for me. (TOR/I4)

   2. (b) The guys are so different! (TOR/ND)

Figures 4a–c show the distribution of very, really and pretty according to this distinction for each age group.

Figure 4a.

Distribution of very by age and type of predication

Figure 4b.

Distribution of really by type of the modified adjectives

Figure 4c.

Distribution of pretty by type of the modified adjectives

The figures show that these three intensifiers have quite different patterns according to adjective function. Figure 4a reveals that very is much more frequent for predicative adjectives for all the 30+ age groups. But among the younger generations, the pattern is inconsistent. Figure 4b reveals that really is more frequent for predicative adjectives for the under 20s, but from the 20-year-olds on up there is little difference until the 50+ age group. Figure 4c reveals that pretty occurs more often with predicative adjectives, regardless of speaker age. Putting these results together with the frequency distributions in figure 3 reveals that where really and very occur most robustly, their frequency is higher among predicative adjectives. Thus really exhibits a profile that indicates advanced delexicalization among the younger speakers, while very exhibits a profile that indicates advanced delexicalization among the older speakers. In contrast, pretty indicates an advanced profile throughout the speech community.

As to the fourth intensifier under scrutiny, so is virtually restricted to predicative position. The attributive example in (13) is the only token in nearly 10,000 adjectival heads.

1. (13) Attributive so:

   Because that's usually what we get, so hot wings. (TOR/Ss)

The reason for this anomalous behaviour in comparison with the other intensifiers is likely to be due to the origin of so as a comparative conjunction rather than an adjective. However, this asymmetrical distribution means that the only context that permits comparison among really, very, so and pretty as intensifiers is the predicative adjective context. Therefore, in the analyses that follow, I will consider predicative contexts only. In this way I focus on the layering of forms in a context where all of them are possible, thus tapping the mechanisms underlying the use of the current, key, competitors in the contemporary Toronto English intensifier system.

The next question is, what is the underlying basis by which intensifiers diffuse?

4.1.4 Semantic classification

Different types of adjectives may be more or less propitious to innovating forms. In earlier research on contemporary British English, Ito & Tagliamonte (Reference Ito and Tagliamonte2003) operationalized the model provided by Dixon (Reference Dixon1977), which divides adjectives into eight groups, as illustrated in (17)–(24):

1. (17) Dimension (e.g. big, large, little, small, long, short, wide, narrow, thick)

   He's a realbig guy. (TOR/j/M)

2. (18) Physical Property (e.g. hard, soft, heavy, light, rough, smooth, hot, sweet)

   It was a reallyhot day and like on the way there I started to feel really, really weak. (TOR/k)

3. (19) Colour (e.g. black, white, red)

   It was really really reallyred yesterday. (TOR/F)

4. (20) Human Propensity (e.g. jealous, happy, kind, clever, generous, gay, rude)

   My dad was sohappy . . . my mom was Ø happy too. (TOR/F)

5. (21) Age (e.g. new, young, old)

   He was great but he was also veryold. (TOR/)

6. (22) Value (e.g. good, bad, proper, perfect, excellent, delicious, poor)

   It seemed a verypleasant place. (TOR/6)

7. (23) Speed (e.g. fast, quick, slow)

   Like, Canadians are veryslow at recognizing their own people. (TOR/£)

8. (24) Position (e.g. right, left, near, far)

   I actually got to Newfoundland which is sofar away. (TOR/s)

Figures 5a–d depict the distribution of really, so, very and pretty according to these categories in predicative adjective contexts. For this analysis, the community has been categorized into a three-part division for age. These groupings reflect the divisions sketched earlier in figure 3, which showed three distinct intensifier systems in the community.

Figure 5a.

Distribution of really by type of the modified adjectives

Figure 5b.

Distribution of so by type of the modified adjectives

Figure 5c.

Distribution of very by age and type of adjective

Figure 5d.

Distribution of pretty by type of the modified adjectives

Figure 5a displays the use of really. In the 50+ age group it occurs among five adjective types – physical property, human propensity, age, value, and other. Among the 30- to 49-year-olds it occurs across all types. Among the 9- to 29-year-olds it occurs across all types and much more frequently. This shows that really has diffused into more and more adjectival categories from oldest to youngest speakers as well as increasing in frequency in each category.

Figure 5b shows the occurrence of so. The proportions are lower overall and uniformly so. The 50+ age group have minimal occurrence of so and it occurs in only four adjective types: physical property, human propensity, value and other. The 30- to 49-year-olds have increasing as well as broader use. Finally, among the 9- to 29-year-olds so occurs not only the most frequently, but also across all the different adjective types. The same incremental pattern of shift is visible from one generation to the next with one exception: the ‘other’ category which represents a combination of adjectives that do not fit into Dixon's categorization schema. Note, too, that the rise in frequency from middle-aged to youngest speakers is visible but not as strikingly so as with really.

Figure 5c shows the occurrence of very. Here the most frequent use is among the middle-aged and oldest speakers. Interestingly, despite greatly decreased frequency in each category, the youngest generation maintains a diffused pattern for very, testimony to its entrenched status in the language. The 30- to 49-year-olds even have higher rates in some categories (physical property, age, other) than the 50+ age group where very is much more frequent overall.

Figure 5d shows the occurrence of pretty. Like so, it is most diffused among the 9- to 29-year-olds where it is also most frequent for dimension, value and other adjectives.

Taken together, the results from figures 5a–d suggest that the diffusion of an intensifier does not happen at the same time as an increase in use. Indeed, it appears that diffusion predates a surge in frequency (see also Ito & Tagliamonte Reference Ito and Tagliamonte2003). This result lends support to Mair's (Reference Mair2004) idea of a ‘delayed increase of discourse frequency’ in grammaticalization, which predicts that grammaticalization will occur first, then diffusion begins, only subsequently leading to increasing frequency. Thus, unlike some concepts of grammaticalization in which an increase in frequency is attributed to the development of grammatical status (Krug Reference Krug2000), these findings suggest that usage rates continue to rise, indeed even escalate, as a new form diffuses into broader contexts (see also Bybee, Perkins & Pagliuca Reference Bybee, Perkins and Pagliuca1994: 8). Similarly, when an intensifier is waning, as in the case of very in this community, it appears that the use of a form recedes uniformly across the broad contexts of its use.

4.1.5 Emotional value

Intensifier use is also thought to correlate with the emotional value of the modified adjective (e.g. Peters Reference Peters and Kastovsky1994). Indeed, Tagliamonte & Roberts (Reference Tagliamonte and Roberts2005) found that the use of intensifier so was significantly tied to emotional adjectives particularly among female speakers. This hypothesis can be tested by categorizing each of the adjectives in the data according to whether it encodes an emotion, as in (25), as opposed to some physical attribute or other quality, as in (26).

1. (25) Emotion:

   1. (a) I don't know; that's why I'm so happy. (TOR/2a)

   2. (b) 'Cause she's so funny . . . she's so mean . . . and so evil! (TOR/3b)

2. (26) Nonemotion:

   1. (a) I just thought it was like so small. (TOR/TM)

   2. (b) Like her belly's big and my belly's so flat. (TOR/r)

Figures 6a–d show the distribution of each of the main intensifiers according to whether the adjective they modify is emotional or nonemotional. Due to the binary contrast (and sufficient data in each cell) the distributions can be representatively shown for each ten-year age span.

Figure 6a.

Distribution of so by emotional value of the adjective

Figure 6b.

Distribution of really by emotional value of the adjective

Figure 6c.

Distribution of very by emotional value of the adjective

Figure 6d.

Distribution of pretty by emotional value

Figure 6a reveals that there is a near pan-community tendency for so to modify emotional adjectives more than nonemotional adjectives.Footnote ¹⁴ The only exception to this pattern is among the very youngest age groups, the <13-year-olds and the 13- to 16-year-olds, where so is more likely to occur among nonemotional adjectives. Figure 6b reveals that really has no consistent trend of use along the dimension of emotionality of the adjectives: the pattern (where visible) goes one way or the other across the board with the exception of the 17- to 19-year-olds where there is no difference. Note however the heightened tendency for really with emotional adjectives among the 20- to 29-year-olds.Footnote ¹⁵Figure 6c provides yet another perspective. Intensifier very is more frequent among non-emotional adjectives for every age group, even among the young where it is rare. Figure 6d shows that pretty is more frequent with nonemotional adjectives, except for the 20- to 39- year-olds.

Putting these new findings into perspective with the earlier results now exposes how intimately an adjective's emotional value is linked to the type of intensifier that may be used to boost it. The out-going intensifier very is favoured across all age groups for contexts with less emotional impact, exactly as can be predicted by a highly delexicalized form. These results, together with those in figure 5c above, also make clear the extent to which very remains entrenched in these contexts among the middle-aged speakers. In contrast, really shows alternating tendencies by age, suggesting that it holds an intermediate status along the delexicalization pathway within this speech community. Finally, intensifier so reveals its incipient phase by a strong and consistent propensity for use with emotion-laden adjectives among the vast majority of the population (aged 17–60+). This provides support for Peters' (Reference Peters and Kastovsky1994: 274) observation that new intensifiers will appear in contexts with ‘a stronger personal and emotional involvement’. What about pretty? It appears to be new based on distribution by age and diffusion by adjective type, but with respect to the dichotomy emotional/nonemotional it does not exhibit the expected pattern. In fact, quite the contrary: it collocates with nonemotional adjectives. Further understanding of these results will require exploration of the sociolinguistic arena.

4.1.6 Speaker sex

An association of intensifiers with women, particularly incoming intensifiers, is a frequent observation in the literature (Jespersen Reference Jespersen1922: 249–50 citing Lord Chesterfield; Key Reference Key1975: 75). The use of so, in particular, has been associated with women. Lakoff (Reference Lakoff1973: 15), for example, refers to it as a characteristically female intensifier.

Given this background and the findings up to this point, the high frequency of really among the under 30s in figure 3 can be interpreted as a strong indication that this form is a recent favoured intensifier in Toronto. This is supported by the striking correlation of really with emotional adjectives among the 20- to 29-year-olds (see figure 6b). However, as also visible in figure 3, the intensifiers so and pretty are increasing across the same age group (under 30s). Intensifiers that are incrementally waning when viewed across the speech community, such as very, might be predicted to have even distribution between the sexes since very has the profile of a diffused, highly delexicalized intensifier, and, as far as I am aware, no particular association with females. In contrast, the suggestive upswing of so and pretty might lead to the prediction that young women are leading these developments. Figures 7a–c test for these conjectures by plotting the distribution of the main intensifiers by age but now also separating out the male and female speakers.

Figure 7a.

Distribution of very by speaker sex and age

Figure 7b.

Distribution of really by speaker sex and age

Figure 7c.

Distribution of so and pretty by speaker sex and age

Figure 7a shows the use of very. Among the 60+ age group, the females show a slightly elevated use of very, but among the 40- to 59-year-olds, the contrast is considerable. Then, from the 40-year-olds on down, there is minimal to no sex difference. Figure 7b shows the use of really. The general trend is towards a rise in frequency of really from oldest to youngest speakers; however, a number of erratic shifts can be observed. In particular the <13-year-olds, the 17- to 19-year-olds and the 20- to 29-year-old females have lofty rates of really in comparison to the males. Figure 7c shows the use of the two incoming forms so and pretty. The results corroborate the claims in the literature regarding so – the 13- to 29-year-old young women are leaders in its use. However, there is an equally prominent trend among the young men; they are leading in the use of pretty. The unexpected result is the diametrically opposed behaviour of the young people with respect to their choice of so or pretty. Such contrasts are not unprecedented. Male/female differences in intensifier use are well documented. However, these findings provide a clear and astonishing picture of male/female differences. Among the oldest generation there is a sex difference for very. At various points there are extreme sex differences for the use of really. In the youngest generation in Toronto males and females have selected entirely different intensifiers as their favoured new form. In sum, the use of intensifiers by male and female speakers of different ages is intimately tied to the stages of intensifier renewal in the community grammar.

4.2 Multivariate analysis

In this section the intensifier system in Toronto is subjected to a number of analyses. Recall that this speech community is highly differentiated (see figure 3): the 20- to 29-year-olds have a system of intensification where really is the most frequent intensifier. The middle-aged group, the 30- to 49-year-olds, are a watershed generation, where really and very are robust competitors. Among the 50- to 91-year-olds, very is the pre-eminent intensifier. The critical nature of the age differentiation was confirmed by a multiple regression of each of the four main intensifiers (not shown) in which age and sex were both statistically significant for each of the four main intensifiers. This means that an analysis putting all the age groups together will obscure the underlying processes responsible for the choice of individual forms. Therefore, the analyses that follow test these age groups separately. Further, I restrict the analysis to the two intensifiers very and really which occur at sufficient thresholds of frequency for viable statistical modelling.

A number of different lines of evidence can be inferred from the multivariate analyses: statistical significance, relative strength of factors and ranking of factor categories significant (for further discussion, see Tagliamonte Reference Tagliamonte2006a). Higher numbers for factors can be interpreted as favouring the intensifier under investigation whereas lower ones disfavour it. The higher the number, the greater the contribution of that factor to the use of the intensifier. The factors for which the numbers are enclosed in square brackets were considered in the analysis, but found not to be significant.

Each analysis tests the same set of factors each of which was shown earlier to influence the choice of intensifier and has the potential to reveal the status of the intensifier in the grammar. Advanced delexicalization is indicated by: (i) a correlation of the intensifier with predicative adjectives and (ii) a spread of use of intensifiers across adjective types. The adjective type factor group tests diffusion using Dixon's (Reference Dixon1977) schema (seen earlier in a distributional analysis in figures 5a–d) with the added subcategory of emotion.Footnote ¹⁶ Thus, the new schema provides a more nuanced test of the diffusion of intensifiers across adjective types.

Table 4 shows the results of three separate multivariate analyses of the probability for very to be used as an intensifier in three age groupings in Toronto: 9- to 29-year-olds, 30- to 49-year-olds and 50+ age group.

Table 4.

Three multivariate analyses of very by age group in the Toronto English Corpus

The table shows that the effect of adjective type is the strongest constraint in every age group with range values of 35, 42 and 41. The other factors (sex and adjective function) exert a statistically significant effect in certain age groups, but their strength is comparatively weak (the range values are only 10 and 8 respectively). Women aged 30–49 and 50+ favour the use of very. It is used most often with predicative adjectives among the 50+ age group. The factor weights for the different categories of adjective type are more revealing. Very is favoured strongly and with more types of adjectives among the 50+ age group. This age group favours the use of very with adjectives of human propensity and notably emotion, but younger speakers (aged 9–49) disfavour it. Among the 9- to 29-year-olds very has receded to adjectives of physical property, 0.64, and position adjectives, 0.76. There is one common trend across age groups, shaded: the high probability of very in the ‘other’ category: 0.69, 0.71 and 0.70. The fact that very exhibits this statistically significant pan-community correlation with a grab bag of different types of adjectives reveals that its waning frequency has not caused a loss of its delexicalized status.

Table 5 shows the results of three separate multivariate analyses of the probability for really to be used as an intensifier in the same three age groupings: 9- to 29-year-olds, 30- to 49-year-olds and 50+ age group.

Table 5.

Three multivariate analyses of really by age group in the Toronto English Corpus

This table shows somewhat the reverse of the trends shown in table 4. Once again, adjective type is strong, but this time only among the 9- to 29-year-olds and the 50+ age group. Adjective function exerts a much weaker effect. The youngest and oldest groups favour the use of really with predicative adjectives. This indicates advanced delexicalization at both ends of the speech community. The effect of speaker sex reveals another nuance. Not surprisingly, given the distributional data in figure 7b, it fluctuates by age group. Females aged 9–29 favour the use of really but males aged 30–49 favour its use. Among the 50+ age group the sex effect is not significant although a comparable direction of effect among the younger speakers is visible (females slightly favour really).

As in the analysis of very in table 4, the different categories of adjective type reveal age differences. In this case, emotional adjectives favour the use of really at 0.61 in the oldest generation and at 0.56 in the youngest generation. Thus, at both ends of the age spectrum, the emotional value of the adjective triggers the use of really. Further, the oldest generation favours really with the same number of types of adjectives as the youngest generation, although these are not the same.

These findings confirm that really is frequently used and widely diffused across the speech community, thus confirming Lorenz's (Reference Lorenz, Wischer and Diewald2002: 157) prediction that really ‘is the most likely candidate to follow in the wake of very and become the next fully grammaticalised adjective intensifier’. However, despite the general decline of very, both very and really exhibit a comparable state of diffusion across the speech community. The next step is to consider what these patterns and trends mean.