Statistical results

The following statistical analyses on the presence or absence of past tense and plural marking in Colloquial Singapore English are computed using Rbrul (version 3.1.3). Rbrul combines the strengths of Goldvarb and R, thereby allowing users to not only connect to the wider community of quantitative linguists who use SPSS and R, but also allows users to incorporate random effects into their statistical modelling. Incorporating random effects in the statistical model helps to solve the problem of overestimating the significance of predictors (Johnson 2009).

Statistical results for past tense marking on weak verbs

Using Rbrul (version 3.1.3), a mixed-effects logistic regression with individual speaker as a random effect was run on the data of weak verbs. Including individual speakers as a random effect means that the model will take into account how some individuals might favor or disfavor a particular linguistic form, over and above what the social and linguistic predictors in the statistical model would predict. The full additive model containing the social factors of ‘age’, 'education’, ‘ethnicity’, ‘gender’, ‘attitude toward English’, ‘English dominance’, and linguistic factors of‘grammatical aspect’, ‘lexical aspect’, ‘priming’, ‘preceding phonological environment’, ‘following phonological environment' is then computed to determine which factors are statistically significant. An additive model simply means that no interactions were included in the statistical model. A total of 1034 tokens from 24 speakers were analyzed and Table 4.1 summarizes the relative weights and p values of the full additive model.

As Table 4.1 shows, the following factors of ‘preceding phonological environment’, ‘English dominance', ‘ethnicity’, ‘attitude toward English’, ‘education’, 'gender’, ‘lexical aspect’, ‘grammatical aspect’, and ‘priming’ were found to be statistically significant predictors (< p = 0.05) of past tense marking in Colloquial Singapore English. Only two factors, ‘age’ and ‘following phonological environment’ were not found to be statistically significant. Furthermore, from the range of relative weights, a ranking of the relative importance of the statistically significant predictors was obtained: ‘preceding phonological environment’ > ‘English dominance’ > ‘ethnicity’ > ‘attitude toward English' > ‘education’ > ‘gender’ > ‘lexical aspect' > ‘grammatical aspect’ > ‘priming’. The relative importance of the predictors is determined by the size of the range. The bigger the range, the more effect a predictor has on the outcome. The range is calculated by deducting the lowest weight from the highest weight. In what follows, we shall look at each statistically significant predictor in greater detail by looking at how the percentages of morphological marking change according to different social and linguistic predictors.

According to the Rbrul analysis, preceding phonological environment is the top linguistic predictor of the presence or absence of past tense marking on weak verbs (range = 66). As seen in Figure 4.1, past tense is marked most frequently when the preceding phonological environment is a vowel (see Example (17a)), at 49.8% of the time, and least marked when the preceding phonological environment is a consonant cluster (see Example (17b)), at 9.3% of the time. Past tense is marked 21.0% of the time when the preceding phonological environment is a single consonant (see Example (17c)).

• (17a) (I) stayed back to look after my mom.
• (Tamil male, 37 years old)

Figure 4.1 The effect of preceding phonological environment on past-tense marking (weak verbs)

• (17b) I was quite happy and I ran, ran, I ran onto a track then I jump.
• (Chinese male, 24 years old)
• (17c) We join OCBC (Oversea-Chinese Banking Corporation), UOB (United

Overseas Bank), and Cheng Kah bank, and another overseas bank.

(Chinese male, 65 years old)

According to the Rbrul analysis, English dominance is the top social predictor of the presence or absence of past-tense marking on weak verbs (range = 55). As seen in Figure 4.2, past tense is marked most frequently by English dominant speakers, at 38.2% of the time, followed by balanced bilinguals, at 27.2% of the time, and finally non-English dominant speakers, at 7.6% of the time.

According to the Rbrul analysis, ethnicity is the second most important social predictor of the presence or absence of past-tense marking on weak verbs (range = 51). As seen in Figure 4.3, past tense is marked most frequently by Tamil speakers, at 53.2% of the time, followed by Malay speakers, at 25.0% of the time, and finally Chinese speakers, at 16.8% of the time. This suggests the importance of crosslinguistic influence in explaining the way in which Singaporeans vary in terms of their use of past-tense morphology.

Figure 4.2 The effect of English dominance on past-tense marking (weak verbs)

5 8 Missing you - Past tense and plural marking

Figure 4.4 The effect of attitude toward English on past-tense marking (weak verbs)

According to the Rbrul analysis, attitude toward English is a statistically significant social predictor of the presence or absence of past-tense marking on weak verbs (range = 29). As seen in Figure 4.4, past tense is marked more frequently by speakers with a favorable attitude toward English, at 31.7% of the time, followed by speakers with a neutral attimde toward English, at 20.1% of the time. Note that since there are no speakers with an unfavorable attimde toward English, there are only two categories in Figure 4.4.

According to the Rbrul analysis, education is a statistically significant social predictor of the presence or absence of past-tense marking on weak verbs (range = 26). As seen in Figure 4.5, past tense is marked most frequently by speakers who hold a degree, at 35.6% of the time, followed by speakers who do not hold a degree, at 20.1% of the time.

According to the Rbrul analysis, gender is a statistically significant social predictor of the presence or absence of past-tense marking on weak verbs (range = 23). As seen in Figure 4.6, past tense is marked most frequently by female speakers, at 34.1% of the time, followed by male speakers, at 19.5% of the time.

According to the Rbrul analysis, lexical aspect is the second most important linguistic predictor of the presence or absence of past-tense marking on weak verbs (range = 20). As seen in Figure 4.7, past tense is marked most frequently

Figure 4.5 The effect of education on past-tense marking (weak verbs)

Figure 4.7 The effect of lexical aspect on past-tense marking (weak verbs)

for state-type verbs, at 40.8% of the time (see Example (18a)), followed by achievement-type verbs (see Example (18b)), at 33.7% of the time, and finally accomplishment-type and activity-type verbs (see Examples (18c) and (18d) respectively), both at 22.0% of the time.

• (18a) I lived nearby, it was just like ten minutes’ walk.
• (Malay female, 30 years old)
• (18b) So I retired when I was fifty.
• (Tamil female, 65 years old)
• (18c) Yah, so they went on top of it, they climb on top of it and they fell in.
• (Malay female, 30 years old)
• (18d) We play with sand, running around in the carpark.
• (Chinese female, 57 years old)

According to the Rbrul analysis, grammatical aspect is a statistically significant linguistic predictor of the presence or absence of past-tense marking on weak verbs (range = 12). As seen in Figure 4.8, past tense is marked most frequently

Figure 4.S The effect of grammatical aspect on past-tense marking (weak verbs)

in perfective contexts (see Example (19a)), at 31.3% of the time, and less so in imperfective contexts (see Example (19b)), at 22.7% of the time.

• (19a) So when I landed, I slipped and I landed on my back.
• (Chinese male, 24 years old)
• (19b) Ah, sometimes is we watch movies, yah.
• (Malay female, 39 years old)

According to the Rbrul analysis, priming is a statistically significant linguistic predictor of the presence or absence of past-tense marking on weak verbs (range = 13). As seen in Figure 4.9, when past tense is primed or when a previous past verb is marked for past, past tense will be marked more frequently, at 42.8% of the time (see Example (20a)). On the other hand, when past tense is not primed or when a previous past verb is unmarked for past, past tense will be marked less frequently, at 19.0% of the time (see Example (20b)).

• (20a) Because of language policy, they eliminated all the dialect broadcast media, yeah, so radio and TV, they wiped it out and there was only like Chinese broadcast media left.
• (Chinese male, 25 years old)

Figure 4.9 The effect of priming on past-tense marking (weak verbs)

(20b) Cause he hang his pants behind (the) door, so the money drop behind it. (Malay female, 30 years old)