[The four papers largely review the same data, which I gathered - both in sociolinguistic interviews, using Jamaican Patwa, and also in more formal interviews and test environments, using Standard Jamaican English - from residents of an urban neighborhood of Kingston, Jamaica (pseudonym 'Veeton') in 1989-90. These data are only briefly summarized in the present paper, which sketches patterns across a range of individual speakers; they are more fully analyzed in my forthcoming book Urban Jamaican Creole (Patrick 1999). Here I take for granted a number of points I have argued for there and elsewhere.]

(1) The classic variationist notion of a speech community requires:

• that a social community exist,
• that its members share patterns of language use, and crucially,
• that they share norms for social evaluation of speech.

The creole continuum in its original formulation (DeCamp 1971, Bickerton 1973) was held to challenge this notion of the speech community. In fact, Guy's 1980 article was a direct response to Bickerton's, which challenged the whole enterprise of grouping individual speakers together. Bickerton charged that such aggregates were in general an illusion and that the individual norms they represented would, if investigated, prove to be an unruly lot with no real coherence; Guy, looking at (TD)-deletion, showed that this need not be true. But back to the creole continuum...

Grammatical rules in a continuum situation, it was said, differ significantly between individual speakers in such a way that they cannot be resolved into discrete social or geographical dialect groupings, because the differences refuse to bundle together and require a large variety of norms to account for all the individual behaviors. At the same time, the contrast between the extreme ends of the continuum was taken to be too great to allow for shared patterns of use or (presumably) for a unified evaluative mechanism. In the 'polylectal' model offered by Bickerton and C.-J. Bailey (1973), there is a linear ordering of varieties-but it is not predicated upon the social stratification of speakers. In fact, DeCamp (1971) and Bickerton (1980) stress that social characteristics of speakers should not be used a priori to distinguish varieties.

(2) In the classic creole continuum:

• Patterns shown by individual speakers cannot be resolved into a few discrete social or geographical dialect groupings;
• The contrast between the extremes of the continuum is too great to allow for shared patterns of use...
• ...or a unified set of evaluative norms; and
• Varieties/speakers can be linearly-ordered along a purely linguistic dimension, and...
• ...correlated with (one or more) social dimensions.

The two larger problems here - the grammatical structure of mesolectal speech and the socio-linguistic structure of the urban creole speech community - can be viewed independently for analytical purposes, though I believe they are ultimately inseparable. The first three points have been recast by John Rickford (1987) as the criterion of (non-)discreteness: basically, that variation across speakers is very fine-grained and covers a lot of ground, so that the traditional solution of consolidating them into distinct varieties, or dialects united under a single umbrella, is unworkable - in part because of the structural distance between the most standard (or acrolectal) and deepest creole (or basilectal) grammars. This is the topic of the present paper.

The fourth and fifth criteria are rephrased by Rickford as the requirement of unidimensionality: the variation found across the continuum can be expressed along a single dimension. Elsewhere I have argued that such a dimension cannot be purely linguistic in nature, but must be social or sociolinguistic; and that the number of dimensions required to model variation must be an empirical and historically contingent matter, not a theoretical requirement. This leaves only discreteness or continuity as the essential characteristic of a creole continuum (though if it could not be correlated with, and explained by, one or more social dimensions, the construct would be of little use). I will not pursue the correlation issue here, as it is the subject of the fourth paper.

The question before us, then, is whether the creole continuum can be maintained as a descriptively useful concept, and with what modifications. Considering the urban Jamaican speech of Veeton, what conclusions about discreteness can we draw? Are there distinct varieties? a true continuum? or what Bailey (1974) has called a 'gradatum' - a situation where there may be continuous variation within a wide mesolect, but sharp boundaries between it and the acrolect and basilect? (This distinction is implicit in Bailey's 1974 discussion; Fasold (1990:196) raises it explicitly.) Or is some other solution required?

This immediate descriptive problem aside, I would like to show that the gap which earlier researchers believed to exist between complex creole communities and variationist methods of analysis can, and must in fact, be bridged. It has been demonstrated (Patrick 1996) that the urban mesolect consists of a mixed, variable grammar featuring the systematic presence and integration of English forms and rules, distinguished by their level and patterns of variability. It is also independently necessary (Patrick 1998) that the concept of the speech community be expanded to allow for the multiple, mutually-opposing sets of norms that characteristically co-exist for (synchronically-) related varieties in a creole continuum (as well as other socially complex communities, cf. Santa Ana and Parodi 1998).

In this paper I continue the project of reconciling creole speech data and practices with the principles of language variation and change, this time updating and modifying the notion of the continuum. This is in answer both to the relative neglect of the challenge that creole data pose to variationists (or their assumption that such data must constitute a fundamentally different sort of sociolinguistic structure, cf. Guy 1980), as well as in answer to some creolists' rejection of both the continuum and/or variationist analysis as incapable of modelling the richness and heterogeneity of Caribbean sociolinguistic structures (e.g. Carrington 1993, Mufwene 1994, Winford 1988, 1993).

The continuum was first proposed by David DeCamp (1971) to give a unified account of the Jamaican sociolinguistic situation, which he found to be poorly modelled by structural dialectology, diglossia, and other dichotomous descriptive methods of the time. Any empirical solution must be based on some selection of particular linguistic phenomena; and DeCamp's answer, affirming the continuum, was exemplified by his analysis of a variety of unrelated lexical items and surface phonological variables.

But the case for the independence of Jamaican Creole from English rests on the analysis of deeper grammatical properties involving central elements of the verb phrase, such as Tense-Mood-Aspect marking systems. DeCamp's data illustrated his belief in continuity, but it did not constitute a test of the continuum model - he did not conduct a sustained examination of appropriate features.

Accordingly, in the present analysis I include two types of past-reference marking - a creole-like one with preverbal 'Did' and 'Neva', and an English-like one using verb inflection - as well as a related phonological variable, (TD)-deletion - simplification of final consonant clusters ending in an alveolar stop - which is motivated quite independently of past-marking but intersects with it. The other phonological element analyzed, (KYA) or palatalization of velar initials before low vowels, has no connection with the syntax, and is thus similar to the surface-level variation of DeCamp's variables.The idea is that the variation profile that emerges will tell us more about the grammar than DeCamp's study did - if continuity indeed characterizes it, it will not simply be because we are looking at unrelated elements which have no reason to pattern together.

Figures 1 and 2 show the behavior of individuals in both conversational and formal-test situations. All speakers recorded several hours of sociolinguistic interviews; later, they also performed English-to-Creole and Creole-to-English translation tasks, in which I presented them with tape-recorded stimuli - five or six sentences, one at a time, which formed a brief narrative set in the past - and asked them to translate each sentence into the other variety (if it was English, translate into Patwa; if it was Patwa, then into English). Each task was loaded with opportunities for speakers to utter variant forms of the linguistic variables I was interested in, namely (TD)-deletion and (Past)-marking.

I have already (Patrick 1998) used these data to illustrate the simultaneous existence of opposed evaluative norms - some unified and some dichotomous - within the same speech community. Here we are more interested in whether the individual variation is or is not continuous: do speakers fall into subgroups exhibiting fine or sharp stratification? Where quantitative differences in interview speech appear, do they coincide with contrasting targets or stereotypes? And so forth.

(Some social characteristics of the speakers are given in Table 3, for reference. Note that speakers in Figs. 1-2 are stratified by status, low at the left and high at the right. Note also that not all of the 16 speakers listed in Table 3 appear for all variables; see Patrick 1999 for details.)

The interview data for (TD) - i.e. the diamonds in Fig. 1 - show continuous variation with fine stratification. There are no sharp breaks, and there is regular decline in (TD)-absence as status rises (i.e. from left to right). There appears to be a single norm, in the sense that in interview speech the level of this variable runs closer to the "Patwa" target than the "English" one; while this is attenuated for the highest-status speakers, Roxy and Rose, the difference is gradual and not dramatic. In general, cluster-simplification is very high - always over 50%, and at or above 75% for most speakers - and for the lowest-status speaker, Dinah, it approaches 100%.

The interview data for (Past)-marking in Fig. 2 also show fairly continuous variation. The range covers most of the spectrum, from 6% to 89%. The sharper gradient suggests that this gramm-atical variable may be a more salient, if still quantitative, marker of social status. Stratification into three groups is apparent (let's call them Low, Mid and High): the Mid group contains two speakers, Opal and Matty. What really distinguishes the mid- and high-status speakers in the sample from the low group, however, is that their conversational usage, instead of reflecting their "Patwa" norm, is closely identified instead with their "English" target. (Here we're looking at speakers from Opal over to Roxy.)

The third variable (KYA), palatalization of velar initials in words such as /gyaadn/ 'garden' and /kyap/ 'cap', shows a completely discontinuous picture. Figure 3 demonstrates that the variation falls largely along age-group lines.

For pre-verbal past-marking with 'Did' there's also a qualitative distinction: though all speakers recognize and properly interpret the form in translation tasks from Patwa, only older people use it in conversation regularly. Of 100 forms found in some 2,000 past-reference clauses, 96 were used by four speakers over 45; the fifth older speaker used it twice (this was Rose, who showed the prestige pattern for KYA), and two younger speakers used it once each. This can be seen in Table 1 below, in the column labeled 'Did' (more on this table later).

The same sharp divide can be seen with negative past marker 'Neva'. The form is used by all speakers, but only a few - Roxy, Rose, Noel and Olive - use it like the familiar English adverb, independent of the {-ed} tense inflection on the verb. For all other speakers, 'Neva' is a classic creole pre-verbal tense marker, and the following verb never carries the redundant {-ed} inflection.

So it is sometimes possible, on purely linguistic grounds, to consolidate speakers into subgroups that contrast with each other but show internal agreement on key properties - quantitative norms, as well as underlying principles and patterns of linguistic organization. But what are we to make of the fact that these subgroups differ from variable to variable?

Figure 1 gave no grounds for distinguishing subgroups at all, but rather resembled DeCamp's (1971) illustration of a classic continuum; Figure 2 suggested three subgroups, finely stratified and correlated with status; while Figure 3 clearly requires 2 subgroups, largely distinguished on the basis of age, but falling into groupings incompatible with those suggested by Figure 2. One assumes that with these data we have not yet reached the limits of variability.

These results raise several questions. First:

(6a) Does the hierarchy or ranking of speakers fluctuate across subgroups?

If so, what explains the relative position of speakers?

(These questions concern unidimensionality and the sociolinguistic structure of the speech community, which are the subject of the fourth talk - for the present, I'll stick with the issue of discreteness: the organization of the grammar.)

Secondly,

(6b) Do the subgroups constitute distinct varieties?

Do grouping patterns persist across variables, i.e. across the grammar?

To answer that, it will be helpful to compare the patterns across all variables at a glance. In Table 1 I have only represented the AR environment of (KYA), because there's no variation in other ones; and I've split (Past) into its two components, classic pre-verbal creole marker 'Did', versus English-like inflection with {-ed}. (The data for negative past 'Neva' do not appear on this page; see above for them.)

Speakers who only appear once are given in italics; I represent categorical patterns in conversational data by enclosing groups of speakers who exhibit them in a single box. All speakers not so enclosed are distinguished, and vertically ordered, by their probability of use of the variable. Columns are arranged so that usage approximating the standard appears at the top. Small spaces, such as between Olive and Matty for the {-ed} variable, represent smaller quantitative discontinuities; large spaces represent greater quantitative and/or qualitative ones.

Remember that all these speakers were drawn from the middle levels of Jamaican society, and that their speech is also intermediate - so we want to pay attention to the extremes. Are there clearly distinct speakers: those consistently among the most standard, who should belong to the acrolect, or those consistently most creole-like who should be classed with the basilect?

Rose and Roxy are the most consistently standard, but their performance in interview speech is never very far above that of their near neighbors. In Figures 1 and 2 it's clear that they share identical target norms for 'Patwa' and 'English' with other speakers. It might be just possible to differentiate Roxy from the rest for Past inflection with {-ed}: she has above 80% inflection in Past clauses, produces the form categorically in the 'English' test situation, and - though you can't see it here - she inflects Irregular verbs at 96%. But this is the only variable for which she might be singled out; on the rest she seems simply to be at the top end of this mesolectal sample.

At the bottom end, the situation is quite different. Compare Dinah and Mina, the most nonstandard speakers in Figs. 1 and 2. Dinah shows nearly complete absence of (TD) in interviews and utterly misses the 'English' target - the other speakers are variable here too, but she shows 80% absence while they are all under 50%. For {-ed} things are even clearer: these two speakers have similar, and very low, rates of inflection in conversation, but like everyone else Mina clearly distinguishes the 'English' and 'Patwa' stereotypes, while only Dinah has no clue - she cannot inflect regular verbs even in the test situation.

Dinah shares the traditional, non-prestigious pattern for (KYA), and the creole use of 'Did' as a pre-verbal TMA marker, with other speakers. For the latter, note that all users of this form (and all who use 'Neva' in a similar way) are subject to the classic constraints of anteriority and punctuality identified by Bickerton (1975) - see Table 2.

What this asymmetry shows us is that traditional creole elements and constraints do indeed intrude into the mesolect, though they are not used actively by all mesolectal speakers and tend to be restricted to those who are older and/or of lower status. But if the occurrence of basilectal elements cannot reliably identify a speaker, the absence of English forms, constraints and targets does - and by this criterion Dinah is clearly not a mesolectal speaker, but belongs to the basilect.

Generalizing from our findings about individual variation here, and putting them together with earlier work on the nature of the mesolect and the evaluative mechanisms at work in constituting the creole speech community, I will now go back and answer the earlier question about discrete-ness and the continuum. Since the variable presence and systematic integration of English forms and rules defines the mesolect, there appears to be no clear dividing line in the grammar between the mesolect and the acrolect - a speaker like Roxy may reach the categorical level in one variable while not differing essentially from her peers. But the absence of such knowledge boldly marks off basilectal speakers.

Thus the situation in Jamaica most closely resembles Bailey's notion of gradatum: continuous variation within a wide mesolect, but a sharp boundary on the lower end, between it and the basilect. The extent and type of grammatical variation we've seen is indeed greater than what we're accustomed to in classic standard and dialect situations, but it does not require an unmanageable number of lects and it does fall into a limited number of shared patterns of use. Taking into account these modifications and others noted earlier, I conclude that the revised notion of the creole continuum is a rich, flexible tool capable of accommodating variationist methods of analysis and modelling the linguistic variation in Jamaican speech .