4.1. Past Time Reference and Aspect

It has been observed in a number of creole languages that predicates used to describe past habitual situations are unmarked for past time reference significantly more often than predicates describing perfective situations (for example, Hackert Reference Hackert2004:171 for Bahamian Creole English; Patrick Reference Patrick1999:189 for Jamaican Creole; Winford Reference Winford1992:337 for Trinidadian Creole). The absence of overt past marking on habitual predicates is even hypothesized to be a “cross-linguistically common phenomenon” not restricted to creoles (Hackert Reference Hackert2004:171). Sentences with habitual predicates differ from generic sentences “by their lack of lawlikeness” (Dahl Reference Dahl1985:97), but in other respects they are semantically similar. In fact, according to Krifka et al. (Reference Krifka, Pelletier, Carlson, Meulen, Link, Chierchia, Carlson and Pelletier1995:17–18), habitual situations are by definition generic situations. Hackert (Reference Hackert2004:170) proposes to account for their similar behavior in terms of the similarities between the two situation types: both are nonspecific and involve “induction and generalization.” Given that “generic sentences are [most frequently] expressed with the most unmarked TMA category” (Dahl Reference Dahl1985:100), the semantic similarity between habitual and generic sentences could account for a low rate of past marking of habitual predicates. Indeed, in Negerhollands, habitual predicates are significantly less often overtly marked for past time reference than perfective predicates (see table 6) but the rate of overt past marking is also significantly lower with progressive predicates. Ongoing (progressive) situations generally overlap with adjacent situations, as can be seen in a narrative (Schiffrin Reference Schiffrin1981:50). This is illustrated in 6, where the progressive situation in 6c overlaps with the situations in 6a and 6b (examples from dJdJ:30).

1. (6)

   

2. 

In Labov & Waletzky's (Reference Labov, Waletzky and Helm1967) theory of narrative structure (discussed in section 4.3), clauses describing a situation that overlaps with that of other clauses are called RESTRICTED clauses, because they can be moved and placed elsewhere within a restricted part of the narrative (Fleischman Reference Fleischman1990:160). Depending on the narrative, a restricted clause can take scope over one to, in theory, all but one of the other clauses. There are also FREE clauses, which have scope over all of the narrative and can be placed anywhere without changing the temporal or logical interpretation of the narrative (Fleischman Reference Fleischman1990:166). Habitual predicates are typically found in either free or restricted clauses, while progressive predicates are typically found only in restricted clauses, with scope over a limited part of the narrative. Thus, what habitual and progressive situations have in common is that they can have scope over other parts of the narrative. The habitual and generic situations can thus be extended to include progressives in that all three can have scope over other elements in a narrative, albeit to different degrees (see table 6 below).

In Caribbean English-lexified creoles (CECs), perfective aspect is expressed by unmarked verbs (Winford Reference Winford1993:33–34). Unmarked verbs are typically associated with a past interpretation in CECs, but they are also used to describe generic and habitual situations (Winford Reference Winford1993:35–36). This leads to an interesting paradox: Perfective predicates are typically unmarked in terms of TMA, and habitual situations—which are semantically unlike perfective situations—are typically unmarked in CECs because the unmarked verb is seen as a perfective marker that has been extended in use to describe habitual situations. Nevertheless, as already mentioned above, habitual predicates are found to be more frequently unmarked for past time reference than perfective predicates in at least three CECs, because they are akin to, if not a type of, generic situations.

Table 6.

The distribution of the expressionof past time reference according to aspect.

Unlike the situation in CECs, perfective aspect is not expressed by the most neutral TMA category in Negerhollands. The TMA marker most closely associated with perfective aspect is (h)a, which is most suitably termed a past marker, as it also occurs with past imperfective situations, as in 6c. The fact that (past) habitual situations are more closely associated with the unmarked verb than perfective situations, not only in CECs but also in Negerhollands, supports the idea that habitual situations are associated with the unmarked verb for a different reason than perfective situations are in CECs (see also Winford Reference Winford1993:42–44).

In Negerhollands, habitual predicates may be unmarked in terms of TMA, or be preceded by imperfective aspect marker lo, habitual aspect marker kan, or kaa gwe n. Table 7 shows that only habitual situations expressed by an unmarked verb have a strong preference for zero pasts, which means that the lower rate of overt past marking of habitual predicates is determined not by the presence of the imperfective marker, but rather by its absence. The interpretation of the unmarked verb in Negerhollands is discussed further at the end of section 5.2.

Table 7.

The expression of past time reference in past habitual predicates according to TMA marker.

A difference between the CECs discussed and Negerhollands is that not only past habitual but also past progressive predicates occur less frequently with (h)a in Negerhollands, as table 6 shows. This similar pattern with progressive and habitual predicates can be explained by the way aspect affects SITUATION TYPE, that is, the classification of situations (not verbs) based on their internal temporal features, such as duration, telicity, and stativity (Smith Reference Smith1997:17–20). Contextual information and/or aspect marking can change the interpretation of a situation to such an extent that it acquires semantics not present or, in some cases, even incompatible with its basic situation type.

It cannot be stressed enough that there is an essential difference between aspect (also known as grammatical or viewpoint aspect) discussed in section 2.2, and situation type (also known as lexical aspect). Viewpoint aspect can alter the basic situation type of a situation in such a way that it would have very different internal temporal properties. Smith (Reference Smith1997:48–53) refers to such altered situation types as DERIVED SITUATION TYPES. What habitual and progressive aspects have in common is that they induce a stative interpretation (de Swart Reference Henriette de1998:354, 359), regardless of the internal temporal features of the same situation in an aspectually neutral context.

The induced stative interpretation of habitual and progressive predicates is by no means incompatible with the observation that states can be habitual, progressive, or perfective (Bertinetto Reference Bertinetto1994:395). Habitual and progressive aspects still have their own semantics, and certainly, neither of them has the same semantics as aspectually neutral states. What is essential is the semantic feature that habitual and progressive situations have in common with states, namely, they hold for a time period preceding (and typically also following) the topic time.Footnote ¹⁶ This makes overt expression of past time reference less relevant, just as time reference is not relevant for generic situations, which are cross-linguistically described using the most unmarked TMA category available (Dahl Reference Dahl1985:99–100). The habitual and progressive situations in 7 and 8a, respectively, overlap in duration with other, perfective situations in the narrative. As mentioned earlier, the potential overlap is bigger for the habitual situation in 7 than for the progressive situation in 8a.

1. (7)

   

2. (8)

   

Thus, reduced relevance of past time reference marking can account for the lower frequency of (h)a in past progressive and past habitual predicates. Nevertheless, overt expression of past time reference is not as irrelevant for past habitual and past progressive situations as it is for generic situations. One finds that in the 20th-century Negerhollands folk stories in dJdJ, the habitual and progressive predicates are still more often overtly marked for past time reference than not.

4.2. Past Marking and Stativity

Stativity is a situation type feature recurrently investigated in creole studies with respect to the variable expression of past time reference. STATIVE situations “will continue[,] unless something happens to change that state” (Comrie Reference Comrie1976:49). As such, they are not “temporally bounded” (van Valin Reference van Valin, Mairal and Gil2006:156–157), as 9 shows, where the stative situation must be interpreted as continuing beyond the topic time.

1. (9)

   

Table 8.

The rate of occurring as a stative predicate and the rate of overt past time reference marking on stative predicates.

Previous studies on CECs have shown that any effect exerted by stativity is often actually the result of idiosyncratic preferences of individual verbs—in particular, have—that are highly frequent and occur mainly or only with the stative meaning (see, among others, Patrick Reference Patrick1999:257, Tagliamonte Reference Tagliamonte, Baker and Bruyn1999:225, Poplack & Tagliamonte Reference Poplack and Tagliamonte2001:142, Hackert Reference Hackert2004:165–166). As table 8 above shows, the frequency of the various stative verbs differs considerably in Negerhollands, too: The verbs ha ‘have’ (N=100) and wees ‘be’ (N=86) account for 62% of all stative predicates (N=302). The rate of stative ha (84%) and wees (97%) being preceded by (h)a is considerably higher than the mean rate of overt past marking on stative predicates (78%). Two other notorious stative verbs, weet ‘know’ (52%) and maηkéé ‘want’ (40%), show significantly lower rates of overt past marking, and so do the two modals, kan ‘can’ and ha fo ‘have to’.

The statistical generalized linear mixed-effects model is able to take into account the variability in past time reference marking across the different verbs. However, the effect of stativity remains significant at p=.0169 (see table 9), despite the enormous variance already observed in table 8 (estimated variance: 0.96641, standard deviation: 0.98306, number of groups: 215). The difficulty in investigating the exact effect of stativity in the current dataset is that when the idiosyncratic verbs ha, wees, weet, kan, ha fo, and maηkéé are removed from the analysis, only a handful of infrequent verbs remain. One could remove the factor stativity from the analysis altogether, assuming that its effect actually reflects the past marking rates of individual verbs, as suggested above, which should already be captured by the random-effect predictor individual verb. However, doing so leads to a significantly worse fit of the model (anova: χ²=.021, df=3), since the residual deviance of the model including stativity is lower (1745) than that of the model without (1755). Thus, although I do not have an explanation for the divergent behavior of the six verbs mentioned above, nor am I able to define the exact effect of the predictor stativity, it does seem to have some explanatory value in the model.

Table 9.

Stative situations; the highly frequent verbs ha ‘have’ and wees ‘be’ included.

Thus, given its statistical significance, stativity is included in the final model on which the tables in section 4 are based, unless stated otherwise.

4.3. Past Marking and Narrative Type

The type of narrative has been found to influence the likelihood of occurrence of zero pasts (Hackert Reference Hackert2004:203). In the Negerhollands data, 98% of all narratives are folktales. In Bahamian Creole English, folktales show a lower rate of overt past marking than narratives of personal experience, which is explained assuming that overt past marking becomes redundant once it loses its temporal referential function, as is the case in fiction (Hackert Reference Hackert2004:196). Whatever the cause, this is not supported by the Negerhollands data, where folktales show a very high rate of overt pasts.

A number of studies have shown how different parts of a spoken narrative influence the likelihood of occurrence of zero pasts (for example, Schiffrin Reference Schiffrin1981, Tagliamonte & Poplack Reference Tagliamonte and Poplack1988, Hackert Reference Hackert2004).Footnote ¹⁷ They follow Labov & Waletzky (Reference Labov, Waletzky and Helm1967)'s framework for narratives of personal experience. In these narratives, people talk about situations they have experienced in their own life. Hackert (Reference Hackert2004:195) has shown that this framework is applicable to folktales too. The main building blocks of narratives are NARRATIVE EVENTS presented in NARRATIVE CLAUSES: If switching the order of two clauses results in a change in the interpretation of the order in which the events happened, then the two clauses involved are narrative clauses (Labov Reference Labov and Labov1972a, Labov & Waletzky Reference Labov, Waletzky and Helm1967, both cited in Schiffrin Reference Schiffrin1981:47). A cluster of narrative clauses forms the COMPLICATING ACTION (hence, narrative clauses are alternatively called COMPLICATING ACTION, or CA CLAUSES): This is where the story is told in a sequence of “temporally-ordered narrative events” (Schiffrin Reference Schiffrin1981: 48), as illustrated in 11a–c. ORIENTATION CLAUSES, exemplified in 10a–c, give background information, such as where and when of the narrative events, and describe “the identities of the characters” (Schiffrin Reference Schiffrin1981: 48). Hence, orientation clauses typically describe states and situations that begin before the onset of narrative events and continue into the CA (Schiffrin Reference Schiffrin1981:49). As opposed to CA clauses, orientation clauses are typically not temporally ordered: The order of orientation clauses within one orientation section can be changed without consequences for the in-terpretation of the information provided (Schiffrin Reference Schiffrin1981:49).

1. (10)

   

2. (11)

   

In Negerhollands, the (perfective) CA clause is the context in which past time reference is most frequently overt. For all other languages in table 10, CA clauses allow the highest rate of zero pasts, the exact opposite pattern. The high occurrence of zero pasts (or historical presents) in CA clauses in Bahamian Creole English and Standard English is attributed to the strong disambiguating force of CA clauses: Overt past time reference is redundant and can be dispensed with because CA clauses report events that are asserted to have already happened and thus receive a past (time reference) reading by default in the context of a narrative (Hackert Reference Hackert2004:192). This explanation does not apply to 20th-century Negerhollands. Table 10 makes it clear how strongly Negerhollands deviates from the other creoles in its high rate of overt pasts in CA clauses, where Bahamian Creole English and Jamaican Creole predominantly have zero pasts. Strikingly, the overall rate of overt past marking in Negerhollands surpasses even the rate of overt past marking in Standard English.

Table 10.

Percentage of overt pasts in CA versus orientation clauses.

The lower overt past rate in orientation clauses compared to CA clauses in Negerhollands can be accounted for by the functional similarities between orientation clauses on the one hand and progressives and habituals on the other hand, as discussed in section 4.1: Just like progressives and habituals, orientation clauses assert that the situation holds for a time period preceding and following the topic time, which means that these situations may overlap with situations described in other clauses.

4.4. Marking on the Preceding Verb (Syntactic Priming)

When form or construction X is in variation with form or construction Y, an occurrence of form or construction X increases the likelihood of form or construction X reappearing. This is referred to as the “parallel processing effect” (for instance, Poplack Reference Poplack1979, Scherre & Naro Reference Scherre and Naro1991:23) or the “principle of concord” (Sabino Reference Sabino1986:62). Poplack & Tagliamonte (Reference Poplack and Tagliamonte2001:126–127) interpret it as an effect of “morphological priming.” In cognitive linguistics, terms such as structural priming, syntactic persistence, and syntactic priming are used (Gries Reference Gries2005:265 and the references therein). Following Gries 2006:265, I adopt the latter term. This effect is so widely attested in all domains of language that “it should be considered as a serious candidate for a universal of language use and processing” (Scherre & Naro Reference Scherre and Naro1991:30).

Syntactic priming appears to be a significant factor in many studies on variable expression of past time reference. As an example, in all but one variety of Early African American English, the absence or presence of past time reference marking on a verb is a significant predictor of the occurrence of that same marker on the immediately following verb (Poplack & Tagliamonte Reference Poplack and Tagliamonte2001:126–127). The same effect has been attested for Negerhollands in dJdJ's informant 1's data (Sabino Reference Sabino1986:62). In these studies, the expression of past time reference is compared to that on the verb in a preceding clause within the same sentence.

I have replicated this method of investigating the effect of syntactic priming for all of the informants of dJdJ. In table 11, the rate of occurrence of overt pasts in Negerhollands is compared in two types of environment: in clauses preceded by other clauses (within the same sentence) and in clauses not preceded by other clauses.

Table 11.

Syntactic priming effect on the expression of past time reference in the preceding clause within the same sentence.

The results obtained in the course of the present study of Negerhollands indicate that the frequency of overt pasts in clauses preceded by a clause containing (h)a, as in 12, does not differ from that in initial clauses or noncomplex sentences, with p=.0572.

1. (12)

   

By contrast, the effect of syntactic priming is clearly visible when the preceding or priming clause in a complex sentence is not marked for past time reference, as in 13b: In this context, zero pasts are even more frequent than overt pasts in the target clause (see table 11 above).

1. (13)

   

Apart from clauses within the same sentence, syntactic priming has been attested in different CA clauses in Standard English and Bahamian Creole English (Schiffrin Reference Schiffrin1981:51 and Hackert Reference Hackert2004:190, respectively). However, studies such as Scherre & Naro Reference Scherre and Naro1991 have shown that the syntactic priming effect holds in many different linguistic environments. Therefore, one would expect the effect of syntactic priming not to be confined to narrative clauses. In principle, one would expect any sentence, not only a CA clause, to exert a priming effect onto the following sentence. Table 12 shows that orientation clauses cause as much an effect of syntactic priming as CA clauses. Thus, in the analysis outlined below, I include the factor preceding sentence and examine the syntactic priming effect of the preceding CA and orientation clause.

Table 12.

Syntactic priming effect on the expression of past time reference.Footnote ²⁶

So far, I have shown that the syntactic priming effect holds between clauses within the same sentence (table 11), and also between adjacent sentences, regardless of whether they contain CA or orientation clauses (table 12). The next logical step is to see if the priming effect would surface in a clause that does not immediately follow the “priming” clause. The finding that “priming is in fact long-lasting” (Gries Reference Gries2005: 374) predicts that this is indeed the case. Gries studies the alternation between bare indirect objects and prepositional objects, a variable that, unlike time reference marking, does not necessarily occur in every clause. He finds that the priming effect holds with as many as twenty-five intervening clauses.

In the case of past marking in narratives, one expects the variable to occur in every clause, with the exception of intervening direct and indirect speech, relative clauses, and the occasional clause without absolute past time reference, where the variable may occur as well, but in a different function. Although one can expect the priming effect to hold in nonadjacent clauses, every following clause with a different past marker may override the priming effect of the preceding marking. Thus, one can expect the priming effect of past markers to be more short-term.

I have explored the question of whether the priming effect would hold in nonadjacent clauses by investigating whether there is an interaction between the factors preceding clause, which involves the marker in the preceding clause within the same sentence, as in 12 and 13, and preceding sentence, which involves the marker in a CA or orientation clause within a preceding sentence. As expected, there is indeed an interaction between these two factors.

First, consider the priming effect in the absence of an immediately preceding clause, that is, in adjacent simple sentences and sentence-initial clauses. When a CA or orientation clause within the priming sentence contains no past time reference marking, the occurrence rate of overt past marking within the target sentence decreases significantly, going from 83% to 67%. In contrast, when the priming sentence contains an overt past, 92% of the target sentences contain overt past marking. This is illustrated in 14: the priming CA clause in 14a, Am see, does not contain a past marker. The sentence-initial target clause in 14b, Sini see, is also unmarked for past time reference.

1. (14)

   

Second, when the priming clause within the same sentence does not contain a past time reference marker, the target clause is less likely to contain an overt past (41% overall, that is, the mean of 61%, 26%, and 33% in table 13) than when the priming clause contains (h)a (87%), regardless of the marking in the preceding sentence. Yet when there is no past marker in the preceding sentence either, as in 15c or 15d, the occurrence rate of overt past marking in the target clause drops even lower, to 26%.

1. (15)

   

Third, it seems that when the preceding clause contains (h)a, the target clause is most likely to contain (h)a as well, regardless of the past marking in the preceding sentence. This is illustrated in 16: target clause 16c contains a, and so does preceding clause 16b, despite the fact that the preceding sentence 16a contains a zero past.

1. (16)

   

These observations are summarized in table 13.

Table 13.

The interaction between the effects of the expression of past time reference.

Sabino (Reference Sabino1986:62) explains the variation in past time reference marking in Negerhollands by attributing it to the interaction between the following two principles: First, past time reference is overtly expressed when not redundant (the functional hypothesis; concurring with Graves Reference Graves1977:146–147). Second, in a complex sentence, the expression of past time reference in the first clause affects the expression of past time reference in the following clause (the principle of concord, or syntactic priming, as it is referred to in the current paper).

While the significance of syntactic priming is uncontested, a more nuanced approach to the data reveals that the functional hypothesis cannot in fact account for the high occurrence rate of overt pasts in initial clauses and simple sentences.Footnote ²⁷ First, the functional hypothesis predicts that when time reference is redundant, it is not overtly expressed. Additionally, Sabino (Reference Sabino1986) assumes syntactic priming to set in where possible. However, the presence of a temporal adverbial, which makes an overt past redundant, has no effect on the absence or presence of overt pasts whatsoever, as shown in table 14.

Table 14.

The interaction between temporal adverbials and past time reference marking.

Even in initial clauses and simple sentences, where the functional hypothesis cannot be overruled by the concord effect, the presence of temporal adverbials has no significant effect, as shown in table 15 below.

Another context where past time reference marking can be argued to be redundant is the CA section of a narrative, which has past time reference by default (Tagliamonte & Poplack Reference Tagliamonte and Poplack1993:90). I have shown in section 4.3 that in Negerhollands, the CA section is precisely the context where the rate of overt pasts is the highest. The functional hypothesis cannot even be upheld, when I restrict functional occurrence of (h)a to the first occurrence in a CA section.Footnote ²⁸ The first narrative clause of a CA section is no more likely to contain an overt past than any other narrative clause, as shown in table 16 below.

Table 15.

The interaction between temporal adverbials and past time reference marking in initial clauses.

Table 16.

Expression of past time reference according to position in the CA section.

Thus, it is clear that overt past time reference marking in Negerhollands is not functional, in the sense that it does not serve for temporal disambiguation; neither are past markers absent when past time reference has already been established. Rather, I assume that the high occurrence rate of overt pasts stems from its functional transparency. I propose the rule in 17 for 20th-century Negerhollands.

1. (17)

   

To describe situations in the past, past marker (h)a is preferred.

This rule can account for every occurrence of (h)a whenever the verb has past time reference. When one assumes the overt past to be the default, only the occurrence of unmarked pasts needs to be explained.