The aim of this section is to examine variation in the co-occurrence of the PP and the SP with several contextual factors over time and across varieties. In doing so it reports the results of a ‘variable rule’ analysis, an application of logistic regression now widely used in variationist sociolinguistic research (see Tagliamonte 2006, 2012 for meth- odological details). Included in the analysis were all coded PP and SP tokens, follow- ing the ‘broad’ approach of defining variable contexts as advocated by recent studies on grammaticalising items (Aaron 2010; Copple 2011; Schwenter & Torres Cacoullos 2008). According to this approach, the variable context can be defined broadly, encom- passing not only items that have strict semantic or functional equivalence, but also those that are in competition with each other along paths of grammaticalisation such as that of ‘anterior > past/perfective’. This approach is motivated by two phenomena:
(a) ‘layering’, which is the availability of different forms to serve ‘similar or even iden- tical functions’ (Hopper 1991: 22–24), and (b) ‘retention’ (Bybee & Pagliuca 1987), or ‘persistence’ (Hopper 1991), which is the potential for evolving constructions to embody old and new semantic features at the same time. Methodologically, coding all PP and SP forms is more reliable than coding only for ‘PP contexts’ or contexts where the two verb forms can be ‘used interchangeably’ (cf. Tagliamonte 2000; Davydova 2011). These contexts are very likely to be different for individual researchers, given the attested variability of PP use across varieties of English.
www.ebook777.com
Free ebooks ==> www.ebook777.com
256 Xinyue Yao
The four contextual factors examined in Yao (2014), namely, temporal specifi- cation, negation, transitivity, and telicity, were coded in this study in light of their important roles in the PP’s grammaticalisation. It is hypothesised that, with contin- ued frequency decline and functional change, the English PP will become increasingly favoured by temporally specified, negative, non-transitive and atelic contexts, in com- parison with the SP (see discussion in Section 2).
The clause (containing either the PP or SP form) was selected as the basic unit of analysis. For temporal specification two factor values – unspecified and specified – were identified, depending on whether the clause contains at least one phrasal or clausal adverbial that expresses temporal meanings, such as location (e.g. yesterday, in the past, when I was young), duration (e.g. for five years, over time), frequency (e.g.
always, sometimes) and sequence (e.g. after he came, before I opened the door). Posi- tive clauses were distinguished from negative clauses that contain both verbal and nonverbal negators (e.g. not, no, nothing, nobody). Transitive verb phrases were also distinguished from non-transitive ones based on whether they take objects or not.
The former include monotransitives (e.g. feed (the dog)), ditransitives (e.g. give (him a present)) and complex transitives (e.g. consider (it wrong), put (the toys upstairs)), and the latter include copular verbs (e.g. become (independent), be (in the garden)) and intransitives (e.g. run, smile). In a similar fashion, phrasal verbs were classified into transitives and non-transitives (for example, turn on (the light) was coded as transitive, and turn aside as non-transitive).
For coding the (a)telicity of the situation expressed by the PP or SP, which is a semantic, nonstructural category, a distinction was made between events and atelic sit- uations by applying linguistic tests as discussed in Vendler (1967), Binnick (1991) and many other previous works.3 In identifying the situation expressed by a clause, atten- tion was given only to lexical, ‘central’ clausal elements (Quirk et al. 1985), excluding grammatical elements such as the progressive -ing and ‘peripheral’ or ‘optional’ ele- ments such as temporal adverbials. This approach, employed previously by Davydova (2011) and Yao (2014), enables us to focus on the effects of the inherent temporal structure of a situation which arises from semantics.4
3. For example, events can occur as the complement of the verb finish (Daniel finished painting a picture), and can be modified by in-adverbials (Daniel painted a picture in an hour) but not by for-adverbials (*Daniel painted a picture for an hour). Furthermore, with events, the perfective form (Daniel painted a picture) is true if and only if the corresponding imperfective form (Daniel was painting a picture) is true, but not vice versa.
4. Previous variationist research on the English PP (e.g. Tagliamonte 2000; Davydova 2008, 2011; Van Herk 2008, 2010; Yao 2014) has developed finer coding schemes for the four con- textual variables. These are not pursued in this study given the already complex structure of
Free ebooks ==> www.ebook777.comThe present perfect and the preterite in Australian English 257 A small number of indeterminate cases for which a confident decision regarding telicity could not be made were also found in the data. Example (6) illustrates:
(6) I’ve suddenly taken against the brothers Spang. I didn’t like those two men in hoods. The way the man hit that fat Negro. The boiling mud. I wouldn’t have minded so much if he’d just beaten the jockey up ordinary Cops-and-Robbers stuff. But that mud showed a nasty mind.
(ARCHER 1956flem.f8b) The situation expressed by the clause with showed can be interpreted as either an event (‘reveal, become evident’) or a state (‘represent, embody’). Indeterminate cases of this type, which total 92 in number – or 2.9% – of all coded verb forms, were removed from the variable rule analysis, and their frequencies by variety/time period are presented in Table 4 of the Appendix.
Six independent variable rule analyses were performed on Rbrul (Johnson 2009) using fixed-effect modelling. In what follows we will consider, one by one, the effect of each contextual factor on the choice of the PP over the SP. For simplicity’s sake we focus on main effects only. It should be noted however that there are sig- nificant interactions between the contextual factors, in particular between temporal specification and negation (χ2 ≈ 100.89, p < 0.001), and between transitivity and telicity (χ2 ≈ 710.92, p < 0.001), since transitive clauses tend to encode events, and never functions both as a temporal specifier and as a negator. Table 2 presents the chi-square statistics, with p-values indicating the strength of correlation between two given factors.
Table 2. Correlation between contextual factors
chi-square p-value
specification × negation 100.89 0.000
specification × telicity 5.76 0.016
specification × transitivity 0.18 0.670
telicity × negation 2.02 0.156
telicity × transitivity 710.92 0.000
transitivity × negation 11.26 0.001
the data, with comparisons made between two linguistic variants, three varieties and two time periods.
www.ebook777.com
Free ebooks ==> www.ebook777.com
258 Xinyue Yao
Table 3. The contribution of contextual factors to the choice of the PP over the SP
AusE BrE AmE
1850–99 1950–99 1850–99 1950–99 1850–99 1950–99 Temporal specification
Unspecified .44 .38 .42 .38 .40 .31
Specified .56 .62 .58 .62 .60 .69
Range 12 24 16 24 20 38
Negation
Positive [.49] [.45] [.51] [.44] [.51] .35
Negative [.52] [.55] [.49] [.57] [.49] .65
Range 30
Transitivity
Transitive [.46] [.47] [.54] [.48] [.55] .42
Non-transitive [.54] [.53] [.46] [.52] [.45] .58
Range 16
Telicity
Telic .61 .62 .62 .72 .60 .62
Atelic .39 .38 .38 .28 .40 .38
Range 22 24 24 44 20 24
Let us now consider the results of the variable rule analysis presented in Table 3, along with the underlying frequencies as shown in the Appendix. Significant factor weights are bolded and non-significant ones are given in square brackets. We find that temporal specification and telicity have consistently significant impacts on the choice of the PP over the SP, with comparable magnitudes of effect. By contrast, negation and transitivity generally do not influence the choice.
The finding for temporal specification confirms Miller’s observation (2000, 2004) that the use of the PP is strongly associated with the presence of temporal adverbi- als. Furthermore, it can be suggested that the PP and the SP serve respectively as the marked and unmarked expressions for a situation located in the past, the PP being not only less frequent, but also preferred by (temporally) specified contexts in compari- son with the SP (see Schwenter & Torres Cacoullos 2008 for discussion of the notion of markedness). Comparing the 19th- and 20th-century data, we find that the PP’s markedness strengthens over time, as indicated by a general increase in the magnitude of effect. The increase is most noteworthy in AmE (20 → 38). AusE (12 → 24) and BrE (16 → 24) resemble each other in this respect. Examining the underlying frequen- cies for temporal specification we find a similar picture: the percentage of temporally
Free ebooks ==> www.ebook777.comThe present perfect and the preterite in Australian English 259 specified PPs out of all PPs (of the same variety/period) undergoes an increase from 23% to 29% in AusE, and very similarly, from 22% to 29% in BrE, but from 30% to 43%
in AmE. These results reinforce the findings of previous studies on temporal specifi- cation with the PP in contemporary BrE and AmE such as Seoane & Suárez-Gómez (2013) and Hundt & Smith (2009), in which the percentages range from 27% to 34%.
On the other hand, the degree of temporal specification remains almost invariant over time for the SP in the present study (AusE 15% → 15%, BrE 15% → 13%, AmE 12% → 12%), confirming that the increased use of the PP in temporally specified contexts is the outcome of its grammaticalisation as opposed to long-term stylistic shifts affecting verbal categories in general.5 Note that this increase is only a relative tendency when comparison is made with the SP, and it may not extend to the frequency of temporally specified PPs normalised to text length.
In order to probe the nature of this increase, temporal specifiers for PPs were grouped into six categories based on their semantic features, as shown below:
a. temporal quantifiers (always, never, ever): quantify over the extended-now6 b. recent past: locate the situation in a recent past time (e.g. recently, just) c. duration: specify the duration of the situation (e.g. for two years, all my life) d. since-adverbials: specify the left boundary of the extended-now (e.g. since 1999) e. present-time orientation (e.g. now, so far, yet): express orientation to the ‘present’,
or the deictic time
f. other: temporal specifiers that cannot be grouped into (a)–(e) (e.g. today, before) Table 4 presents the frequencies the six types of temporal specifiers, with per- centages calculated based on the total number of temporal specifiers shown in each column.7 Although the relatively small scale of the dataset does not afford generalisa- tions of diachronic change for each semantic category, it can be seen that the increase of the PP in temporally specified contexts is due, at least in part, to its strengthened association with the three temporal quantifiers, always, never and ever (AusE 25% → 54%, BrE 24% → 28%, AmE 13% → 41%). As has been suggested in Yao (2014), when
5. For example, Biber and Finegan (2001) found a rise in the total number of situation- dependent references – including temporal adverbials – in the fiction section of an earlier version of ARCHER.
6. Note that with always and never, the PP clause typically receives a ‘continuative’ inter- pretation: the time of the situation or its negation is interpreted as coincidental with the extended-now. On the other hand, the interpretation with ever is ‘experiential’, i.e. the situation occurs at least once within the extended-now.
7. These total frequencies are not necessarily the same as the number of temporally specified PPs, since a PP may co-occur with more than one temporal specifier.
www.ebook777.com
Free ebooks ==> www.ebook777.com
260 Xinyue Yao
co-occurring with the PP, these elements signal universal or existential quantification over the extended-now time span, with the PP functioning similarly to a tense marker.
The past situations are viewed from the present perspective not by virtue of the present persistence of their result states, but by virtue of their occurrence or non-occurrence in the extended-now time span.
Table 4. Frequencies of temporal specifiers for the PP (vertical percentages)
AusE BrE AmE
1850–99 1950–99 1850–99 1950–99 1850–99 1950–99 Always/never/ever 12(25%) 14(54%) 8(24%) 8(28%) 4(13%) 13(41%)
Recent past 3(6%) 1(4%) 3(9%) 1(3%) 5(16%) 2(6%)
Duration 8(17%) 2(8%) 2(6%) 3(10%) 8(25%) 5(16%)
Since-adverbials 4(8%) 3(12%) 4(12%) 1(3%) 5(16%) 5(16%)
Present orientation 3(6%) 2(8%) 4(12%) 1(3%) 2(6%) 1(3%)
Other 18(38%) 4(16%) 12(36%) 15(52%) 8(25%) 6(19%)
Total 48(100%) 26(100%) 33(100%) 29(100%) 32(100%) 32(100%)
Consider, next, the results of the variable rule analysis for negation and transitiv- ity, as shown in Table 3. These two factors are similar in that they are only selected as significant in 1950–1999 AmE, where the PP is favoured by negative and non- transitive contexts. The AmE pattern can be seen as another manifestation of the PP’s functional specialisation. As its (relative) frequency declines, the PP becomes more susceptible to the influence of linguistic constraints and its presence is increasingly triggered by contextual cues. As in the case of temporal specification, AusE aligns itself with BrE, showing insensitiveness to the influence of negation and transitivity. If we further examine the underlying frequencies for negation, we find that the percentage of negative PPs out of all PPs undergoes a two-fold increase in AmE (11% → 24%), which is not found in either AusE (13% → 13%) or BrE (13% → 15%). The percentage of negative SPs only shows some insignificant fluctuations over time (AusE 11% → 9%, BrE 12% → 9%, AmE 10% → 8%).
The frequency findings for transitivity are somewhat unexpected. That transitivity is not a significant factor in AusE appears to suggest that the AusE PP is less advanced than its American counterpart in the move from transitive to non-transitive contexts.
Accordingly, we would expect a lower proportion of non-transitives PPs in AusE, at least in the 19th-century data. However, contrary to this expectation, the percentage of non-transitive PPs in AusE is remarkably high for 1850–1899 – even higher than that in AmE (AusE 45%, BrE 33%, AmE 37%) – although it is surpassed by the latter in the 20th century (AusE 43%, BrE 37%, AmE 46%). Interestingly, this finding is in
Free ebooks ==> www.ebook777.comThe present perfect and the preterite in Australian English 261 line with the results of Davydova’s (2008) study of PP marking in earlier Irish English (IrE), which is based on the Hamburg Corpus of Irish Emigrant Letters, a collection of personal correspondence written by Irish emigrants to Australia and America in the 18th and 19th centuries. In Davydova’s data, non-transitive verbs also favour the PP, whereas transitive verbs tend to be in SP form. An Irish-influence explanation is plausible for the high frequency of non-transitive PPs in earlier AusE, given the strong historical ties between the two countries. A sizeable proportion of Australia’s population during colonial times was Irish settlers and convicts who were bilingual or monolingual speakers of IrE. Roughly 20% present-day Australians are estimated to be of Irish ancestry (McGregor 1980). This demographic fact is likely to have linguistic consequences. As Trudgill (1986: 139–141) argues, evidence for the role of IrE in the formation of AusE, albeit small in scale perhaps, is rather robust. It can be seen at least at the lexical-grammatical level in the presence of features such as the non-standard second person plural pronoun youse, the adverbial but (as in I don’t want it but), neg- ative epistemic mustn’t (as in He mustn’t have seen me, as opposed to the standard English variant He can’t have seen me).8
As for the question of why there is an unusually strong link between the PP and non-transitive verbs in earlier IrE, one possible cause is competition with the well- known ‘medial-object perfect’ characteristic of this variety (as in I have my din- ner eaten). The construction has the same form as the source construction of the present-day English perfect and also expresses possessive-resultative meaning in effect. However, the medial-object perfect is not simply a direct retention from Old English. According to Brinton (1994), the input for the Irish medial-object perfect is the standard (Modern) English stative resultative (as in I have (got) my hair cut). The construction emerged well after the ‘have + past participle + object’ word order was established for the perfect, and only gained in frequency and regularity in mainland England towards the late 19th century. Based on a diachronic evaluation of corpus data, Pietsch (2009) showed that the construction was also present in earlier dialects of English spoken in Ireland. It went through a process of grammaticalisation from the late 19th century onwards, which was triggered by the presence of a functionally and formally related construction in Irish (tá mo dhinneár ite agam, literally ‘is my dinner eaten at-me’). A possible consequence of the expansion of the transitive medial-object perfect is that the functionally similar PP is ‘pushed’ towards more non-transitive con- texts, so that there is a strengthened association between non-transitive verbs and the PP in IrE. In light of what is already known of the demographic makeup of 19th cen- tury Australia, a reasonable hypothesis is that this feature has found its way into earlier
8. See Ramson (1966) and Turner (1966) for skeptical views of Irish influence on the formation of AusE.
www.ebook777.com
Free ebooks ==> www.ebook777.com
262 Xinyue Yao
AusE through dialect contact and has stabilised in this variety. This hypothesis should be further investigated in future research.
Finally we consider the results for telicity in Table 3. Telicity tends to have a strong influence on the choice of the PP over the SP. In each variety/time period, the PP is significantly favoured by events and disfavoured by atelic situations, showing that traces of its original meaning are reflected in the linguistic constraints on its current distribution. Somewhat surprisingly, the magnitude of effect strengthens with time, in particular in BrE (BrE 24 → 44, AusE 22 → 24, AmE 20 → 24). Underpinning this finding is the tendency for the percentage of PPs in atelic contexts to decrease in BrE (30% → 19%), stabilise in AusE (34% → 32%), and increase in AmE (26% → 36%).
Although the BrE pattern does not lend itself to a straightforward explanation, it can be suggested that the BrE PP is most conservative insofar as it retains a strong connec- tion with its resultative roots.