In this section, we will explore two variables that may possibly account for some of the internal variation that has been attested within and across the three varieties and we will briefly discuss other variables that may be worth taking up in future research.
Following Liebermann et al.’s (2007) study, we will first explore the possible effect of frequency on the retention of t forms. Secondly, we will explore to what extent salience in terms of differences between the irregular and the regularised form (in terms of pronunciation and orthography) also leads to retention.
7.1 Frequency
As observed in Section 1, Lieberman et al. (2007) established a clear link between fre- quency and retention or disappearance of irregular verbs: verbs with a low frequency of occurrence are more vulnerable to patterns of regularisation than verbs that occur very frequently. An interesting question that was not explicitly addressed, however, concerns the impact of frequency on the preferences of verbs that are listed as having two different options (either ed or t). Do frequently occurring irregular verbs tend to
Free ebooks ==> www.ebook777.comPatterns of regularisation in British, American and Indian English 361 prefer t endings (even if both options are possible), and less frequently occurring verbs ed endings? Even if the current set of verbs cannot be subdivided into categories that are either very frequent or very infrequent, according to Lieberman et al.’s criteria, it is still interesting to examine whether there is a link between the relative frequency of the examined verbs with respect to one another and their preference for ed forms or t forms. If frequency indeed has an impact on the retention of irregular t forms, one would expect verbs with the lowest frequency to show the highest preference for ed forms and vice versa. In other words, the order of the verbs ranked according to decreasing frequency of occurrence would also have to reflect the order of preference for t endings.
Table 14. Order of frequency (high to low) compared to order of t preference (high to low) in AmE, BrE and IndE
AmE BrE IndE
Freq ↘ Prop t ↘ Freq ↘ Prop t ↘ Freq ↘ Prop t ↘
learn light learn light learn light
burn speed burn kneel burn dwell
dream dwell light dwell light kneel
light kneel dream leap dream speed
spell leap spoil speed spoil burn
lean burn spell spoil spell leap
spoil dream lean spell spill dream
smell smell leap smell lean spell
spill spill spill burn speed smell
leap spell smell spill smell learn
speed spoil speed dream leap spoil
dwell learn kneel learn dwell spill
kneel lean dwell lean kneel lean
Table 14 provides an overview of the frequencies of the verbs in decreasing order of frequency within and across each of the varieties, together with an overview of t proportions, in decreasing order of strength. It clearly shows that there is no corre- spondence between the frequency of the verbs and the preference for t forms. Contrary to expectations, a verb like learn, for instance, which occurs more than 82,000 times in past tense forms in all the data taken together shows a (very) weak preference for t forms compared to other less frequent verbs in all three varieties. Conversely, dwell and kneel, which have the lowest and second lowest frequencies of all verbs (1,096 and
www.ebook777.com
Free ebooks ==> www.ebook777.com
362 Bernard De Clerck & Klaar Vanopstal
906 instances respectively), show a very strong preference for t forms (again contrary to expectations based on the Lieberman study), in all three varieties. The absence of such a clear correlation is further supported by correlation tests (Spearman and Kend- all’s Tau), which do not reveal any correlation patterns in any of the varieties.
In the next section, we will explore salience of change that is involved in the trans- fer from irregular to regular form as one of (many) other variables that may account for the attested proportions and preferences.
7.2 Salience of change
We will start from the intuitive assumption that processes of regularisation which involve more pervasive changes from irregular to regular forms will meet more resis- tance. More specifically, we will assume that verbs whose only change in the regulari- sation processes involves a switch from t to ed in terms of spelling, and from voiceless to voiced ending in pronunciation, will evolve faster (and display more regularised uses) than verbs that are subject to additional changes in the regularisation process, including stem vowel change or changes in orthography (cf. irregular lit vs regular lighted). In other words, we will take salience of change as a variable that may slow down ongoing processes of regularisation: the more the regularised form differs from the original irregular form, the less regularised forms will be attested. As such, we could expect verbs such as learn, burn, spell, spoil, spill, smell and dwell to have evolved more rapidly and to display more ed forms as they do not have vowel change in t forms of their preterite and participle forms. By contrast, dream, light, lean, leap, speed and kneel would be more reluctant to change, as they do have vowel change in the t forms, and show a higher proportion of t forms. In addition, we will postulate that those verbs which also reflect vowel change in their orthographic form (e.g. lit, knelt and sped) evolve more slowly than others such as dreamt and leant.
Figure 7 below provides an overview of the possible correspondence between preferences for the ed or t form, and the presence or absence of vowel change in irreg- ular conjugations for all GloWbE data taken together. The X-axis of the plot chart presents vowel change as a variable. The left side groups verbs that do not have vowel change, the right side groups verbs that do have vowel change. The Y-axis represents the ed/t variable: the bottom of the axis groups verbs that prefer t, the top of the axis groups verbs that prefer ed.
The regression line in Figure 7 clearly goes down (R² = 0.219), indicating that the proportion of ed decreases in verbs with vowel change. The data points in the top left corner of the plot chart show that there is indeed a link between vowel change and ed regularisation patterns: many of the verbs that do not display vowel change occur more frequently with ed forms (i.e. burn, learn, smell, spell, spill and spoil). Note that the strongest correlation between t retention and vowel change is
Free ebooks ==> www.ebook777.comPatterns of regularisation in British, American and Indian English 363
attested for those verbs whose preterite and participle forms also change in terms of orthography, i.e. light, speed and kneel. The more the irregular form deviates from the regular stem+ed pattern, the fewer regularised forms are attested (irrespective of frequency or variety).
Figure 7 also displays a number of interesting outliers. Dwell, for instance, does not have vowel change but strongly prefers t. Dream and lean do have vowel change but prefer ed. The retention of t in the case of dwell may be related to the formal nature of the verb, which may account for its more conservative spelling in the formal contexts that it is associated with.
Language variety: total
VowelChange
PropEd
1,00
,80
,60
,40
,20
,00 ,0
42
47 43
49 45 41 44
R2Linear = 0,219
52 40 48 50 46 51
,2 ,4 ,6 ,8 1,0
Figure 7. Vowel change and t/ed preference
No vowel change Vowel change
40 burn 41 dream
42 dwell 43 kneel
46 learn 44 lean
48 smell 45 leap
50 spell 47 light
51 spill 49 speed
52 spoil
www.ebook777.com
Free ebooks ==> www.ebook777.com
364 Bernard De Clerck & Klaar Vanopstal
Relevant factors for ed preference in dream and lean may be analogy and ambigu- ity avoidance. Consider for example the possible analogical impact of the noun dream (and its pronunciation) on the pronunciation of the simple past and past participle forms. A quick look at the GloWbE corpus reveals that, as a noun, dream occurs three times more frequently than it does as a verb, a difference which may have accelerated the development of the regularised pattern with a similar pronunciation (AmE: 39,844 x as a noun vs 11,112 x as a verb; BrE: 31,288 x as a noun vs 10,731 x as a verb; IndE:
10,922 x as a noun vs 2,688 x as a verb). In the case of lean, ambiguity avoidance with the preterite and participle forms of the verb lend may have fuelled polarisation and a division of labour between t and ed, where ed is now the default form with the preterite and participle of the verb lean. In contrast, the stronger rate of retention of t in leapt may be pronunciation-related: as opposed to dream, learn and lean, whose voiced final consonants allow for voiced pronunciation of ed endings, the voiceless /p/ in leap trig- gers a voiceless /t/ pronunciation of both ed and t. This may lead to a retention of the t form in orthography as a reflection of pronunciation.
Summing up, the degree to which irregular forms differ from their regularised counterparts has an effect on attested proportions: regularised patterns occur more frequently in those cases where the irregular and the regularised form show fewer differences in pronunciation and writing. The outliers, however, also show that other forces are at play which accelerate or slow down regularisation processes.