RQ 4. Putting together the quantitative analysis (RQs1 & 2) and the qualitative analysis (RQ3) what does this tell us about the nature of bi-varietal language use in young
5 V ARIATION IN T ENSE A SPECT M ORPHOLOGY : P RELIMINARY CONSIDERATIONS
5.5 Accounting for the variation: Factor groups and factors
5.5.4 Excluded factor groups
One of the outcomes of working with real data is that often we simply do not end up with sufficient tokens to explore all possibilities. A number of factor groups that were coded for were excluded from further quantitative analysis on the grounds of poor distribution in the current sample. These are adverb expression, ‘na’ expression, polarity, sentence type and object number. Their distributions are presented in Table 5-5 and Table 5-6 below. I will briefly comment on the justification for coding each factor group and the implications of their exclusion.
Table 5-5: Distribution of V, Ving and Vbat per excluded factor groups, HOME context
V Ving Vbat Total
% N % N % N Adverb Expression Temporal/Durative adverb 25 4 19 3 56 9 16 Other adverb 41 40 29 28 31 30 98 no adverb 42 234 21 118 36 201 553 Polarity Affirmative 40 256 23 146 37 239 641 Negative 85 22 12 3 4 1 26 ‘Na’ Expression ‘Na’ 38 15 20 8 43 17 40 no ‘na’ 42 263 23 141 36 223 627 Sentence Type Declarative 41 257 23 141 36 227 625 Interrogative 50 21 19 8 31 13 42 Total 42 278 22 149 36 240 66722
Each token was coded for the presence/absence of an adverb that specified the temporal or durative qualities of the verb: for example, olweis ‘always’, oldei ‘always’23. This factor
group was explored with the expectation that, if there were a significant quantity of adverbs in the corpus, they might align semantically with the speculated aspectual contrasts for the verb forms and further support the analysis. Further, as discussed in the introduction to this thesis, adult second language learners are said to go through a period of relying on adverbs to convey tense and aspect contrasts, and this stage
precedes the use of verbal morphology (e.g. Bardovi-Harlig 2000; Klein 1995). However, this is also one of the potential points of difference between learners of a typologically
22 Note that the total token count here [N=667] post-dates the further exclusions discussed in the following chapter
(§6.2)
distinct L2 and learners of a second, closely-related variety, as is the case here. If the latter use their L1 as the model for the L2, then they wouldn’t necessarily go through a phase of not marking tense/aspect through verbal means, rather, they might ‘import’ the L1 system.
Table 5-6: Distribution of V and Ving per excluded factor groups, SCHOOL context
V Ving Total % N % N Adverb Expression Temporal/Durative adverb 50 2 50 2 4 Other adverb 50 11 50 11 22 no adverb 49 145 51 150 295 Polarity Affirmative 50 146 50 147 293 Negative 43 12 57 16 28 ‘Na’ Expression ‘Na’ 67 2 33 1 3 no ‘na’ 49 156 51 162 318 Sentence Type Declarative 50 154 50 153 307 Interrogative 29 4 71 10 14 Object Number24 Singular 44 30 56 38 68 Plural/Mass 58 7 42 5 12 Total 49 158 51 163 32125
Positive evidence of this would be if there were no substantial reliance on
temporal/durative adverbs to convey aspect in the SCHOOL context. The small number of
tokens of temporal adverbs [HOME N=16; SCHOOL N=4], against the significant
proportion of aspectually marked verbs suggests that, indeed, this is the case. The children in this corpus in fact use temporal adverbs rarely in the school data. The results for use of other adverbs [HOME N=98; SCHOOL N=22] demonstrate that this was not
simply reflective of a lack of adverb use as a general trend. However this conclusion is somewhat tentative, particularly since the other lexical strategies that tend to proliferate at this stage (e.g. use of connectives, other temporal sign posts like days of the week and temporal verbs like ‘start’) have not been investigated.
Clause polarity was coded for after the finding by Walker (2000:139) that both simple and progressive present forms were favoured in affirmative contexts. In the present
24 Transitive clauses only; Unexpressed or otherwise uncodable objects [N=21]. not significant χ2(1, N=80)=.8292, p=.362]
25 Note that the total token count here [N=321] post-dates the further exclusions discussed in the following chapter
corpus, clauses are negated with pre-verbal ‘not’ or ‘don't’. Examples of negated clauses from HOME (51) & (52) and SCHOOL (53) & (54) are given below.
(51) Ai nat noimbat dingo. [SJD-048:371 Shamus HOME]
ai nat no-im-bat dingo 1sg.sbj neg know-tr-BAT dingo
‘I don’t know the dingo.’
(52) Ai don weis thiis [SJD-040:212 Tiffany HOME]
ai don weis thiis 1SG.SUB NEG waste DEM
‘I don’t waste these.’
(53) Ethan [teacher], this not working [SJD-059:298 Shamus SCHOOL]
(54) No, we don’t use that one [SJD-039-C: Daniel SCHOOL]
There are actually very few tokens of negative clauses in each context [HOME N=26;
SCHOOL N=28]. It is noticeable that in the HOME context there is a strong correlation
between negative clauses and V forms (22 of 26 (85%) negated clauses). This does not appear to be the case in the SCHOOL context, where incidence of negative clauses per
verb form [V = 43%; Ving = 57%] are roughly the same as the overall distribution of each form [V = 49%; Ving = 51%] i.e. there is not a strong favouring either way. An examination of the source data suggests that this might be a sampling issue: higher raw tokens of negation of Ving forms in the SCHOOL (compared to HOME) can be largely
attributed to the challenges of remote IT provision, since most of the tokens of this combination were from students complaining about computers “not working”. In any case, the low number of tokens in the HOME data prevents further conclusions being
drawn.
It is worth noting that in the SCHOOL data all the Ving forms are negated with ‘not’ and
all the V forms are negated with ‘don’t’, which follows standard English usage (Huddleston & Pullum 2012). There is a strong tendency towards this pattern in the
HOME context too with all Ving and Vbat forms negated with nat ‘not’ and most V
forms negated with don ‘don’t’ (4 tokens of V are negated with nat).
Preliminary descriptive work on Alyawarr English carried out during the life of the ACLA2 project noted the presence of a potential adverb/discourse marker na, as shown
in examples (55) and (56) below. In Kriol it has been described as a discourse particle, indicating sequentiality of events and topic shifts (Schultze-Berndt et al 2013; Graber 1987). Because the function of this form is uncertain in the present data, na expression was initially included as a factor group to see if mutual light could be shed on its scope by any distributional correlations with verb form. The infrequency of this marker unfortunately means that this approach has not been helpful in this regard [HOME
N=40; SCHOOL N=3], though the distributions in the HOME data [V=38%; Ving=20%;
Vbat=43%] do hover around the overall distributions [V=42%; Ving=22%; Vbat=36%] suggesting no effect. I further note that its relative absence in the SCHOOL data suggests
that it may be a L1-only morpheme for the children.
(55) Ai pudim theya na, reken. [SJD-046-A:263 Deanna HOME]
ai pud-im theya na, reken 1SG.SBJ put-TR there NA reckon
‘I put it there now, (I) reckon.’
(56) Am meikimbat bigethan na. [SJD-048:508 Shamus HOME]
am meik-im-bat bigethan na 1SG.SBJ make-TR-BAT biggest_one NA
‘I’m making the biggest one now.’
Tokens were also coded for sentence type—as declarative or interrogative (intonation only or morphologically and/or syntactically organised)—to see if this had any impact on the choice of verb form. Later these two interrogative categories were collapsed into one factor group due to the low number of tokens of each. Examples of interrogative clauses are given for HOME (57) & (58) and SCHOOL (59) & (60) contexts below. As the
results tables show, interrogative clauses occur infrequently in both the HOME [N=42]
and SCHOOL [N=14] data sets, and this precluded further quantitative analysis. As with
na expression, the distributions in the HOME data [V=50%; Ving=19%; Vbat=31%] do
hover around the overall distributions [V=42%; Ving=22%; Vbat=36%] suggesting no effect.
(57) Kenny yu wandim this kain? [SJD-063:288 Simon HOME]
Kenny yu wand-im thiskain name 2SG.SBJ want-TR this_kind
(58) Wat yu kukimbat? [SJD-040:72 Lenora HOME]
wat yu kuk-im-bat what 2SG.SBJ cook-TR-BAT
‘What are you cooking?’
(59) I cut it this? [SJD-060:449 Simon SCHOOL]
(60) Where they are going? [SJD-039-C:640 Tiffany SCHOOL]
As I noted at the beginning of this section, sometimes the data simply won’t allow for testing of every hypothesis. We can however perhaps infer that since temporal/durative adverbs, ‘na’, negative clauses, interrogative clauses, and plural/mass objects occur with such relative infrequency, they are unlikely to be contributing to the variable expression of V/Ving/Vbat. In other words, none of these clause features are likely to be part of the variable grammar of present temporal reference expression in both HOME and SCHOOL
contexts.
5.6 Summary
This chapter has focused on the preliminary elements of a comparative variationist examination of the variation between V, Ving and Vbat. I first examined how these three morphemes operate in potentially related languages: Australian contact languages, SAE and Alyawarr. This yielded the possibility that aspectual semantics and/or
transitivity should be examined carefully in the current data. We then took a brief digression to examine a handful of other verb forms used by the children in present temporal reference clauses, but ultimately not in significant enough numbers to warrant further quantitative analysis. Nevertheless, this showed that Vs is emerging in SCHOOL
contexts, as are Alyawarr-derived inflections in HOME contexts. Further longitudinal
sampling and analysis of these children will need to pay attention to these forms.
Returning to the question of what might govern the choice between verb forms (V, Ving and Vbat), I discussed the motivation for each of the factors that will form the analysis of variation in the following chapter, and also the factors for which this painstaking process ultimately revealed distributions that rule them out from further consideration. Now that we have a set of motivated and distributionally-sound hypotheses, the
quantitative analysis of how these factors contribute to the choice between V, Vbat and Ving can begin. This will be presented in the following chapter.