My Work on Guyanese Creole

Most of the examples I have published are from African-American Vernacular English, especially re-recordings of Foxy Boston and Tinky Gates (Rickford and McNair-Knox 1994; Rickford and Price 2013). I will come back to these below but I have also done several re-recordings with Guyanese Creole data (e.g., Rickford 1987). For instance, Cane Walk pandit Ustad’s use of the first-person subject pronoun mi (as in mi tel ii “I told him”) was 39% (29/74) at a peer group party in his home surrounded by friends and relatives. This was significantly different (p = .0001, chi-square) both from his usage in all four recordings with me, two spontaneous, two formal (9%, 28/313), and from his usage in an expatriate re-interview with two White Englishmen and an American (Rickford 1979: 458; Rickford 2014b), in which he did not use a single token of the basilectal creole variant (0/30).5

The Foxy Boston and Tinky Gates Series of Interviews

This kind of interviewer (and situation) effect is reminiscent of the comparison in Rickford and McNair-Knox (1994) in which we focused on two interviews (interviews III and IV) in a series of interviews with Foxy Boston, an African-American teenager in East Palo Alto (EPA), California. Interview

III, the third to be done with Foxy by EPA community member and scholar Faye McNair-Knox, also African-American, yielded high percentages of key AAVE variables, like possessive -s absence, third-singular present -s absence, copula is/are absence, and invariant habitual be.6 In contrast Interview

IV, done by an unfamiliar White graduate student just eight months later, yielded significantly lower percentages for the last three of these variables, as shown in Table 3.1.

The theme of stylistic variation across successive interviews, and its implications for panel studies of change, is taken up in Rickford and Price (2013) with additional data on Foxy Boston and her EPA friend Tinky Gates, both re-recorded in their mid-thirties, as mothers, after two

Table 3.1 Foxy’s usage of five AAVE variables with two different interviewers (adapted from Rickford and McNair-Knox 1994: 247). *** indicates whether usage is significantly different or not (n.s.)


Possessive -s absence

Plural -s


Third singular present -s absence

Copula /auxiliary is + are absence

Invariant habitual be









241 per hr





50% n.s.

0% n.s.



78 per hr***

or more recordings as teenagers. Tinky’s data are more straightforward than Foxy’s, so let us begin with it (see Table 3.2).

As we note (Rickford and Price 2013: 161), Tinky shows a “linear [and significant] diminution [...] in vernacular usage over [the] three time points”, which we associate with age-grading rather than community change. Our reasoning is based on Alim’s (2004) study of the same neighborhood, which shows that community norms for copula absence in 2010, i.e., approximately at the time of recording of the third interview by RaShida, remain high, essentially unchanged from what we obtained in the same community in the 1980s.

Assessing the significance of the diminution of Tinky’s and Foxy’s invariant habitual be usage between their earliest and latest recordings, including the intermediate year 1992 for good measure, is more complicated than using a chi-square or Fisher’s exact test, and I have not addressed this issue in previous publications. But the statistical significance of these decreases in the rate of production of invariant be across time can be assessed by a heuristic argument that was suggested by my colleague, Ewart Thomas (personal communication).

If it is assumed that tokens are produced independently and at a constant rate, a, during an interview, then it is well-known that the number, Nt, of tokens produced in an interview of length, t, has the Poisson distribution with a mean of a*t, and a standard deviation of {a*t)u2. Using the normal approximation to the Poisson, we can say also that the observed rate, a (= Nt/t), has a mean of a, and a standard deviation of (a/t)x/1. In other words, the margin-of-error in estimating a is approximately ±2(a/t)xa, which can be approximated by substituting a for a. The confidence intervals shown in Table 3.3 were derived by using

Table 3.2 Tinky Gates’ vernacular usage across her multiple interviews (Rickford and Price 2013: 162). All interviews were conducted at Tinky’s home









Interviewer and interviewee

Faye and RaShida

Faye and RaShida


third present -s absence


80% *

57% *

Is + are absence


70% *


Invariant habitual be / hour




*Tinky’s (1992) usage of third- present -s absence and is + are absence differs significantly from her 1987 usage (tested with a two-tailed Fisher’s exact test). Also, her 2006 usage of these two features differs from her 1987 and her 1992 usage.

Table 3.3 Variation in the sample size on which Tinky and Foxy’s invariant habitual be usage is calculated across three comparable dates (data drawn from Rickford and Price 2013: 162, 165)



Number of tokens, N,

Length, t, of interview (hrs)

Observed rate, a = N/t

Approx, margin of error, 2(a/t)V2

Approx, confidence interval for a







(14.1, 25.1)






(4.2, 13.8)






(1.2, 5.6)







(81.0, 113.0)






(62.1, 93.9)






(6.2, 13.8)

the margin-of-error, ±2(d/t)vl. For each participant, the confidence interval for the rate of production in the earliest period (1986 or 1987) does not overlap with that from the latest period (2006 or 2008, respectively), and the interval for the intermediate period (1992) has a small overlap with the other two intervals. This provides strong evidence of a marked decline in the rate of production across the three periods, both for Tinky (in Table 3.2) and Foxy (from Table 3.4, below).

Two other points should be made about Tinky’s data, and they apply as well to Foxy’s comparable data, to which we will soon turn. The first is that these data are not subject to the weaknesses of the ‘small n’ and ‘gap effects’ to which Cukor-Avila and Bailey (2018) point as potential weaknesses of panel studies, especially those that involve longitudinal studies of small numbers of individuals. The small n effect is “an artifact of small numbers of tokens” (Cukor-Avila and Bailey 2018: 188) that inflates the appearance of variability. The authors suggest that to ensure greater reliability and stability, informants should have “at least 60 third-singular tokens, and at least 120 deletable copula tokens” (Ibid.). Tinky’s 56 third-singular tokens in 1987 come close to this minimum, but her 128 and 201 in 1992 and 2006 are more than two and three times this minimum, respectively. This is even more the case for Tinky’s deletable copula tokens - 256, 286, and 464 in 1987, 1992, and 2006, respectively - or two to nearly four times the minimum they suggest for the copula. The gap effect they discuss is a lapse of several years between successive interviews, which “tend to diminish any familiarity that has been established if that time gap also reflects a parallel break in contact between fieldworkers and interviewees” and can produce a decline in the use of vernacular variants (Cukor-Avila and Bailey 2018: 204). While their point is theoretically valid, it does not apply to the longitudinal data in Rickford and Price (2013) - despite the fact that our study is included in their list of studies exhibiting this limitation (Cukor-Avila and Bailey 2018: 204). The reason? The field-worker in our case (RaShida Knox) is not an outsider to the community, who visits from time to time, but an insider who went to school with the interviewees, remained in the community, and stayed in touch with the informants during their young adult years.

The second point to be made is that Tinky also shows interesting, sometimes significant variation in her use of third-singular present tense -s absence and copula absence by topic within individual interviews. Since we have concentrated on situational switching so far (i.e., variation in interlocutor, place, type of event, see Blom and Gumperz 1992) rather than the metaphorical switching represented by topic change (Ibid.), it is worth remembering that this kind of variability exists too. More specifically, in her intermediate, 1992 interview, Tinky’s is + are copula absence varies significantly (chi-square p = .015) from 48% (14/29) in Topic B, where she talks about herself, her boyfriend, mom, and drugs,













Interviewer and interviewee

Faye and RaShida

Faye and RaShida

Faye (with RaShida and another teenager present)

Beth (unfamiliar white Stanford grad student)

RaShida (occasional interruptions of Foxy’s daughters)

third present -s absence






is + are absence







habitual be/ hour






to 74% (75/101) in Topic H, where she talks about her annoying grandmother and family. Moreover, Topic H is close to the end of the hour and a half interview, while Topic В is close to the beginning, before the interlocutors are fully warmed up. And 42% of the Topic H copula tokens occur in embodied quotations, where Tinky is quoting vernacular family members and friends directly; only 3% of the copula tokens in topic В are of this type, and we show (Rickford and Price 2013: 163) that embodied quotations tend to show higher rates of AAVE use.8

Table 3.4 (from Rickford and Price 2013: 165) shows Foxy Boston's variability across the five interviews in which she was recorded between 1986 and 2008. The key difference between these data and those in Table 3.1 is that Foxy does not show the steady diminution in her vernacular usage from top to bottom (or younger to older) that Tinky does. In particular, she has an intermediate decrease in her is/are copula absence from 90% to 36% between 1986 and 1988, then an increase to 70% in 1990, followed by a fall to 40% in 1988, and a final fall to 35% in 2008. But this micro-variation, as pointed out by Rickford and Price (2013: 164-169), has several lessons to teach us in studying longitudinal panel studies of change.

These studies have several lessons to teach us in studying longitudinal panel studies of change. The first lesson we have emphasized that we need multiple recordings of individuals (ideally multiple recordings per individual even in the same time period) in order to have meaningful and reliable panel studies. A second lesson Foxy’s data provides is that we need more than one linguistic variable. The third singular present tense -s absence evidence for age-grading change in Foxy’s language use between ages 13 and 34 is clearer than is the evidence from her is/are copula absence. When we contrast her copula absence in the early years (1988 and 1992) with her copula absence in 2008, the differences are not statistically significant. But the same comparison for third-present -s IS statistically significant, and of course, using data from the later years (1986 and 1990), even more so.

The third lesson we noted in the 2013 paper is that, just in contrasting Tinky and Foxy, we can see clearly that individuals differ, with Foxy being more of a stylistic chameleon, “appearing in different linguistic guises depending on the personas she considers most appropriate to project to the people around her and the situations she’s in” (Rickford and Price 2013: 166), and Tinky is much less so. This, as we noted there, may relate to differences in personality and mood, elements that Fischer (1958) identified as sources of linguistic variability, but ones that few modern sociolinguists have engaged with. Particularly Eckert’s (2018) work on stylistic variation has great implications for panel research, revealing the manifold ways in which speakers recruit variation for moment-by-moment meaning-making (see also Podesva 2007). However, a few researchers (Buchstaller 2015; Tetreault 2018) have

86 John R. Rickford

begun to address the specific issue of how to incorporate an interactional approach into the methodology of panel studies of change.

The fourth lesson we have learned is that variation in the language use of speakers from one recording year or month to the next is not random, but can indeed be explained, as we attempted to do in our 2013 paper (Rickford and Price 2013: 167-169) by noting relevant details of the recording situation in each case. These include, for instance, where Foxy was at the time in terms of her networks of friends, the school or summer programs and activities she was involved with, who she was trying to impress, and what kind of “persona” she was projecting at the time. It is valuable to have larger numbers of speakers, perhaps more representative of cross-sections of the community, in our panel studies to advance our understanding not only of which types of life stages can have measurable consequences in linguistic behaviour but also the longterm trajectories of variation across the lifespan as individuals manoeuvre age-specific and interactionally relevant speaker identities.

In the spirit of Labov (1972: 97), who noted that alternative methods often have complementary strengths and weaknesses, it is important to remember that close longitudinal studies of individuals, repeatedly sampled, offer valuable perspectives on the reasons for style-switching and for age-grading or community change over time that might otherwise remain opaque, for instance in studies of larger groups, conducted from a distance (see Wagner and Tagliamonte 2018).

< Prev   CONTENTS   Source   Next >