The Subjunctive in Native Spanish

As previously stated, it is generally accepted that, following adverbial conjunctions, the subjunctive is used to denote states and events that have not yet occurred, whereas forms in the indicative mood denote actions that are habitual or have already occurred (Blake, Reference Blake1985; Lee et al., Reference Lee, Young, Bransdorfer and Wolf2005; Montrul & Perpiñán, Reference Montrul and Perpiñán2011). This contrast is shown in examples (1) and (2).

1. (1) Cuando viene Marta, cenamos en el centro.

2. “When Marta comes, we eat dinner downtown.”

3. (2) Cuando venga Marta, cenaremos en el centro.

4. “When Marta comes, we will eat dinner downtown.”

In example (1), Marta’s visiting is depicted as a habitual occurrence, and, thus, the verb form (indicative) implies that she has visited previously and, likely, numerous times. In contrast, in (2), the subjunctive verb form that follows the adverbial conjunction leads to the interpretation that the event has not yet occurred—namely, that Marta has not yet arrived. Although there is often additional information (such as the use of periphrastic or morphological future forms in the main clause) in the context to corroborate the interpretation of these verb forms, NSs are said to be able to accurately interpret such forms without additional cues and, thus, L2 learners of Spanish must also acquire the ability to interpret such forms appropriately (e.g., Blake, Reference Blake1985; Lee et al., Reference Lee, Young, Bransdorfer and Wolf2005).

Despite the general agreement in college textbooks and formal grammars of the Spanish language, the L2 learner faces an additional challenge in that NS use and interpretation of these forms exhibit some variation. In other words, use and interpretation are not categorical but rather exhibit some sociolinguistic variation, and under some circumstances, some NSs might interpret example (1) as denoting an event that has not yet occurred. Whereas a full review of all that has been reported regarding mood use in native Spanish is beyond the scope of the current article, the review that follows provides a sense of what is known about how NSs of Spanish use the subjunctive forms across contexts and speech communities. We begin with four important studies conducted with NSs from Mexico, all of which indicate a certain degree of variability regarding mood. García and Terrell (Reference García, Terrell and Hagiwara1977) collected data in Mexico City via an instrument that contained 25 sentences with comment triggers (e.g., Es triste que “It’s sad that”) that prescriptively require the subjunctive. Overall, the authors found that use of the subjunctive was generally accepted, but participants changed the mood to indicative in 10 cases. Additionally, respondents demonstrated a preference for the indicative in the past (i.e., they preferred preterit forms in past-time contexts). This study served as one of the initial indicators that NSs demonstrate variable mood use and that the indicative is being extended to previously subjunctive contexts. In similar work, Blake (Reference Blake, Lantolf and Stone1981) elicited data from participants from Mexico City using a sentence-completion task. He noted the occurrence of variation in contexts of uncertainty and that, for example, the indicative was at times used following no creer “to not believe,” a prescriptively subjunctive context. Likewise, the participants of Bayerová’s study (Reference Bayerová1994) in Mexico City demonstrated variable mood use in spoken production in that adverbs of uncertainty (such as posiblemente “possibly”) co-occurred with 38 verbs in the indicative and 25 in the subjunctive (the prescriptive expectation). Also, similar to García and Terrell (Reference García, Terrell and Hagiwara1977), there was greater use of the indicative in past-time contexts. Finally, González Salinas (Reference González Salinas2003) examined subjunctive use in Monterrey, Mexico, in nominal, adjectival, and adverbial clauses in language elicited through a sociolinguistic interview and reported greater rates of use of subjunctive forms by females with professional degrees, in accordance with more prescriptive linguistic behavior. Taken together, these studies demonstrate that the use of subjunctive forms varies according to a range of linguistic and social factors.

Turning our attention specifically to studies that focused on mood use in adverbial clauses, four additional investigations merit attention. De la Puente-Schubeck (Reference De la Puente-Schubeck1992) analyzed 30 Spanish NSs in Albuquerque using an informal sociolinguistic interview and a sentence-completion task. According to the GoldVarb analysis, the category adverbial clause trigger was significant, as speakers favored the subjunctive in clauses with antes de que “before,” whereas hasta que “until,” después de que “after,” and cuando “when” disfavored the subjunctive.^{Footnote 3} Additionally, the informal interview disfavored the subjunctive mood, whereas the semiformal sentence-completion task favored it, indicating that use of the subjunctive correlates with higher formality. In fact, this decreased use of the subjunctive in adverbial clauses in New Mexican Spanish is consistent with later work found in Castellón, Spain (Blas Arroyo & Porcar Miralles, Reference Blas Arroyo and Porcar Miralles1997), and Costa Rica (Murillo Medrano, Reference Murillo Medrano1999). Blas Arroyo and Porcar Miralles (Reference Blas Arroyo and Porcar Miralles1997) analyzed adverbial clauses in Castellón via a sentence-completion task, noting that participants preferred the subjunctive with future contexts and that the future indicative was acceptable for participants in the acceptability task in such clauses. This reporting of the acceptability of the indicative in adverbial clauses for events that had not yet occurred is an important precursor to the present study.

Murillo Medrano (Reference Murillo Medrano1999) analyzed the speech of 46 Costa Rican speakers via dyadic conversations, formal interviews, and formal oral discourse. The investigator focused on mood use in adverbial clauses, noting greater rates of use of the subjunctive in clauses with nonspecific time reference, as opposed to clauses with specific time reference. More recently, Isabelli (Reference Isabelli2006) studied mood use in adverbial clauses in Reno, Nevada, finding a rate of about 90% subjunctive use in so-called obligatory contexts for first-, second-, and third-generation Mexican Americans. Once again, the fact that adverbial clauses denoting information that had not yet occurred did not categorically co-occur with the subjunctive is consistent with the previous studies reported across tasks and Spanish-speaking communities. In sum, these studies confirm the general finding reported earlier that subjunctive use is variable and, furthermore, that this variability does occur within adverbial clauses. What has not been investigated extensively is the degree to which individual factors, such as geographic origin, social class, age, gender, and the like, impact variation with the subjunctive. For example, there is a lack of comparative data across geographic regions, and this makes it impossible at present to know whether certain regions are more likely to allow greater variability after adverbial conjunctions than others.

The Acquisition of the Subjunctive in L2 Spanish

We turn now to research on the L2 acquisition of the subjunctive, noting that, just as our knowledge of the variability of subjunctive use among NSs has grown, so too has the degree to which such variability has been incorporated into L2 research. One of the most prolific researchers on the L2 acquisition of the subjunctive in Spanish is Collentine, whose 1995 study was the first to analyze nonnative acquisition of the Spanish subjunctive in terms of both semantic and morphological features. He concluded that intermediate learners were able to generate a high number of nouns-over-verbs in each clause but also demonstrated use of simplified or insufficiently complex clauses. These findings were consistent with earlier work, such as that of Terrell, Baycroft, and Perrone (Reference Terrell, Baycroft, Perrone, VanPatten, Dvorak and Lee1987), who found that learners were not capable of producing the complex structures necessary for subjunctive use. Building on his earlier study, Collentine (Reference Collentine1997) focused on whether the use and interpretation of the subjunctive mood was affected by the regularity of the verb form itself. He found that learners paid greater attention to irregular forms, which in turn interfered with correctly interpreting sentence meaning. Although discussion of more recent studies will demonstrate some methodological improvements that have been made since Collentine’s early studies, the findings that morphological regularity and verb semantics are important factors in subjunctive use and acquisition continue to prove accurate.

Lubbers Quesada (Reference Lubbers Quesada1998) described the Spanish subjunctive in terms of the prototype schema used by NSs and nonnative speakers (NNSs). Her participants were 16 intermediate university learners studying abroad in Querétaro, Mexico. Two sets of oral interviews were conducted—one soon after the students’ arrival and the other 6 weeks later. She noted that, in the first NNS interview, subjunctive marking was quite low (7.5%) and only occurred following cuando “when” and querer que “to want that.” In the second interview, subjunctive use increased to 43% and to a larger range of contexts. Overall, learners (unlike those of Collentine, Reference Collentine1995) were indeed capable of producing complex syntax, and the prototype schema for subjunctive use for these learners was described as [+future], [+desire], and [+irregular]. This schema illustrates a reliance on syntactic, morphological, and semantic features, unlike the NS schema, which does not include verb regularity and instead includes a pragmatic basis, built on whether a speaker chooses to assert information. Lubbers Quesada’s study was an important first step in attempting to see what motivates learners’ selection of the subjunctive, and her consultation of NSs in determining whether or not certain situations enabled subjunctive use was a positive innovation; however, the view of mood selection as correct or incorrect and the limited number of NS participants with which to compare the learners are methodological limitations.

Most recently, a series of studies has examined the acquisition of the subjunctive mood in L2 Spanish while, at the same time, integrating the variability inherent in NS use into the research design. The first of these studies was that of Gudmestad (Reference Gudmestad, Klee and Face2006), whose results showed that, in a preference task, advanced learners selected the subjunctive significantly more frequently than their intermediate counterparts. In the analysis of the linguistic factors that predicted selection of the subjunctive, it was revealed that, at the intermediate level, only the regularity of the verb form was statistically significant, whereas all linguistic features were significant for the advanced group. Despite the many methodological improvements over previous studies, Gudmestad (Reference Gudmestad, Klee and Face2006) does not show whether these results generalize to production tasks.

Geeslin and Gudmestad (Reference Geeslin, Gudmestad, Bruhn de Garavito and Valenzuela2008) analyzed mood contrasts in native and L2 Spanish using two different elicitation tasks (a written contextualized task [WCT] and a sociolinguistic interview). Their participants were 10 NSs from various Spanish-speaking countries and 10 highly advanced NNSs. This is the first study of mood contrast in variable contexts that included NSs among its participants, and, thus, the authors were able to compare NNS subjunctive use directly with that of NSs. There was a significant difference in the frequency of subjunctive selection and use between the two groups on both tasks, as the NSs used or selected the subjunctive more than the NNSs. Likewise, there was a significant difference across tasks for both groups in that all participants chose the subjunctive significantly more frequently on the WCT than they produced it in the interview. The predictive models showed that, for the WCT, semantic category and futurity were significant predictors of subjunctive selection for both groups; additionally, age, gender, proficiency, and study abroad location were significant for the NNSs. On the interview task, semantic category and verb morphology were significant for both groups; futurity, age, proficiency, and country of origin were significant for the NSs, whereas proficiency and study abroad location were significant for the NNSs. On the basis of these findings, the authors concluded that task was a greater source of variability than speaker group.

In a recent, large-scale developmental study conducted using a variationist framework, Gudmestad (Reference Gudmestad, Geeslin and Díaz-Campos2012) analyzed data from 20 NSs and 130 L2 learners of Spanish at five proficiency levels collected through three oral elicitation tasks. Her results demonstrated a nearly linear increase in rates of use of the subjunctive, beginning at 4.1% for the lowest level and rising to 53.8% at the highest level (which was not significantly different from the 52.7% of the NS group). The predictors of use of the subjunctive included only task type at the lowest level and then included semantic category at Level 2, time reference and hypotheticality for Level 3, and, finally, form regularity for Levels 4 and 5 (which matches the NS model). Gudmestad (Reference Gudmestad, Geeslin and Díaz-Campos2012) showed development in three ways: the gradual increase in frequency of use of the subjunctive, the gradual integration of additional nativelike constraints on use into the model, and the increase in the predictive power of these models, which demonstrated a decrease in individual variation as proficiency increased. This research is quite relevant for the current study because it showed the early importance of the semantic category for use of the subjunctive. Moreover, it provides a model of how developmental stages may be identified using response frequencies and linguistic and extralinguistic constraints on these responses, rather than accuracy rates.

Although research on the L2 acquisition of subjunctive mood in contexts in which NS use is variable remains somewhat limited, this body of work is important to the development of several theories of how NS grammars allow such variation. For example, prototype theory (see Adamson, Reference Adamson2009, for a detailed description in the L2 context), states that acquisition involves the development of prototypes that reflect patterns of frequency of use and contain details of the contexts in which a given form is most likely to occur. These contexts can be described in terms of the linguistic constraints described in the research on the L2 subjunctive previously cited. Likewise, connectionism (Bates & MacWhinney, Reference Bates, MacWhinney and MacWhinney1987; Ellis, Reference Ellis, Doughty and Long2003; MacWhinney, Reference MacWhinney, Robinson and Ellis2008) and, more recently, complexity theory (Ellis & Larsen-Freeman, Reference Ellis and Larsen-Freeman2009; Larsen-Freeman, Reference Larsen-Freeman and Atkinson2011) provide detailed accounts of how information in the input—including social information, information regarding frequency of collocations, and linguistic features of contexts of use—can influence developing grammars. Furthermore, these accounts advocate for a model that views grammars as continually evolving according to patterns in the input, even for adult NSs. Following any of these models, our knowledge of acquisition rests on our ability to identify and understand all of the features in the input (e.g., linguistic, discursive, social) that influence language use. The purpose of the current article is not to advocate on behalf of one particular model but rather to provide evidence of the NS variability that exists as well as of how learner grammars change in variable contexts as proficiency increases. In so doing, we demonstrate the need for a model that can account for these patterns and, in this way, contribute directly to theory building for native and L2 grammars.^{Footnote 4}

L2 Spanish Interpretation Studies

Relative to the quantity of research available on production of the subjunctive in L2 Spanish, there are few studies that consider the interpretation of subjunctive forms. Even the studies previously cited that include written preference instruments seek to understand the form that a speaker might choose at the moment of production, rather than to explore the impact of a given form on the interpretation of an utterance, most especially when that interpretation is variable among NSs. This emphasis on production studies at the expense of interpretation studies has been identified as a general weakness in SLA literature (Bardovi-Harlig, Reference Bardovi-Harlig1999) and, thus, is not an isolated issue for mood distinction. The studies that do address interpretation are reviewed here briefly prior to summarizing the goals for the current study. Working in a generative framework, Bruhn de Garavito (Reference Bruhn de Garavito1993) tested whether advanced learners of Spanish observed restrictions of coreferentiality—namely, that the subject of the main clause cannot also be the subject of the subordinate clause in sentences such as Juan quiere que prepare la cena “Juan wants him to prepare dinner.” She found that some participants accepted coreference, with Juan as the subject of the subordinate clause, indicating that they did not have the morphological knowledge necessary to determine a verb’s mood. In contrast, some learners did reject coreference for such sentences, indicating that advanced learners can eventually process complex syntax in a nativelike way.

Subsequent investigations of mood distinction have also considered the syntax-semantics interface in L2 acquisition and, more specifically, have demonstrated that advanced learners can indeed acquire target representations, even when such interpretations were not present in the input (e.g., Borgonovo et al., Reference Borgonovo, Bruhn de Garavito, Prévost, Burgos, Clark-Cotton and Ha2005). These have included different syntactic contexts, such as following negation and in relative clause contexts (e.g., Borgonovo et al., Reference Borgonovo, Bruhn de Garavito, Prévost, Burgos, Clark-Cotton and Ha2005; Borgonovo & Prévost, Reference Borgonovo, Prévost, Beachley, Brown and Conlin2003, Reference Borgonovo and Prévost2005; Bruhn de Garavito, Reference Bruhn de Garavito, Pérez-Leroux and Glass1997). For example, Iverson et al. (Reference Iverson, Kempchinsky and Rothman2008) demonstrated that English-speaking learners of Spanish were able to interpret the subjunctive in relative clauses similarly to NSs in volitional and negated epistemic clauses (e.g., No creo que “I don’t think” + subordinate clause). They found that interface vulnerability was easiest to overcome when specific lexical items (e.g., volitional verbs) triggered the subjunctive. Although these studies do not seek to address contexts in which NSs may demonstrate variability, the findings regarding the effect of particular lexical items are germane to the goals of the present investigation.

Another strand of research has focused on the comprehension and input processing of subjunctive forms in L2 Spanish (Lee, Reference Lee1987; Lee & Rodríguez, Reference Lee, Rodríguez, Glass and Pérez-Leroux1997). Lee (Reference Lee1987) provided empirical evidence of the subjunctive’s low communicative value by demonstrating that learners who had not previously encountered the form showed no significant differences in their comprehension of a passage that contained the subjunctive as compared to learners who had received instruction on the target form. Similarly, Lee and Rodríguez (Reference Lee, Rodríguez, Glass and Pérez-Leroux1997) found that passages containing subjunctive forms were no more difficult to comprehend than those that did not. However, they did argue that subjunctive forms became intake less frequently than indicative forms, meaning that the inclusion of the subjunctive did not interfere with the global meaning of the passage, but it often did not enter into the learners’ intake either. Thus, studies of this type have shown that learners can comprehend utterances that contain subjunctive forms even at lower proficiency levels. Nevertheless, the examples in these studies were not created to test learner interpretation in contexts in which the subjunctive form itself contributed to overall meaning. Thus, they were more likely to contain nominal, rather than adverbial, clauses in their test items.

Finally, Quer (2001) argued that the subjunctive serves as a cue that the interlocutor must adapt his interpretation of an embedded event or state to some alternative model (rather than the default model), inferable from discourse context. He focused on adverbial clauses and reasoned that, despite the subjunctive’s low communicative value, the form played a key role in connoting interlocutors’ cognitive models of the propositional content presented to them—namely, that it indicates whether or not an action has already occurred.

Taking these studies together, we find evidence that the subjunctive forms are less perceptually salient than those associated with other tenses and aspects (e.g., the preterit), that the presence of a subjunctive form at lower proficiency levels has little effect on overall comprehension of a reading passage, and that those contexts that involve semantic and pragmatic (or so-called interface) features are more difficult to acquire. Thus, research on interpretation of subjunctive forms, in both categorical and variable contexts, will benefit from these insights. The current study seeks to expand this body of research by examining in greater depth the role that subjunctive and indicative forms play in variable adverbial contexts.

Summary and Goals for the Present Study

The preceding review has demonstrated a contrast in the L2 variationist literature between the growing interest in the L2 acquisition of the subjunctive in Spanish and the lack of research on the interpretation of this form in variable contexts. Thus, the primary goal of the current study is to begin to address this gap by exploring a context in which the verb form (i.e., subjunctive vs. indicative mood) is thought to contribute in important ways to the interpretation of the utterance. More specifically, although prescriptive accounts may attribute a categorical interpretation for each form in the postadverbial contexts under investigation, in which the subjunctive always means an event has not yet occurred and the indicative forms always mean the event is habitual, sociolinguistic research suggests that this is not the case. Thus, our project also contributes more generally to the growing body of research on the acquisition of L2 variation and demonstrates how an analysis of rates and predictors of use can enhance our understanding of how L2s develop. As stated previously, this line of investigation stands to contribute to several existing and developing theories about how individual speakers—both native and nonnative—acquire, store, and produce language that varies across speech contexts. Furthermore, because we seek to identify the stages of acquisition for interpretation of the subjunctive forms in these contexts, our research provides important knowledge about the path along which learner grammars develop. A final goal of this article is to begin the exploration of a single syntactic-semantic context, given that the literature on the L2 acquisition of the subjunctive to date suggests that semantic category and perhaps even individual lexical items are important predictors of use of the subjunctive forms and, thus, are likely to play a role in the interpretation of these forms as well. In the remainder of this article, we describe in detail the empirical study designed to meet these goals, followed by the results of our analysis and a discussion of the implications of these findings.

Participants

The participants (N = 97) in the current study included NNSs enrolled in three different levels of instruction in Spanish at a large Midwestern university and a group of NSs of Spanish from a variety of countries of origin. The first NNS group consisted of 29 English-speaking learners of Spanish enrolled in a fifth-semester language course. The average age of this group was 19.0 years. The fifth semester was chosen because, in the university’s curriculum, the relationship between adverbs and mood selection is not presented until the fourth semester; thus the students at this level had been exposed to this relationship and were enrolled in the subsequent grammar course. All 29 of the NNSs at this level reported on the language background questionnaire (described in the next subsection) that they had completed the fourth-semester course at the university. The second participant group included 35 English-speaking learners of Spanish enrolled in a 400-level linguistics course. The average age of these participants was 20.4 years. The third and final NNS group was made up of 17 NSs of English who were graduate students in the department of Spanish and Portuguese at the same university and who, in addition to teaching Spanish courses, had also lived or studied abroad for long periods of time. The participants in this group averaged 25.8 years in age and distinguished themselves from the lower two groups in the extent of their experience abroad. Participants from this group had spent time in Spain, Argentina, Mexico, Puerto Rico, Nicaragua, Chile, Guatemala, and Paraguay. Their time abroad ranged from 6 weeks to 3 years (M = 13.6 months). They had studied Spanish formally for a period ranging from 6 to 14 years (M = 11.5 years). The rationale for including such an advanced group in the current study is to be able to determine whether the most advanced learners, those with many years of study and extensive experience abroad, are able to overcome the challenges associated with the interpretation of the subjunctive mentioned in the preceding review.

The final participant group included 16 NSs of Spanish, whose average age was 32.6 years. Seven of the speakers were between the ages of 22 and 29, whereas the other 9 were at least 30 years of age. These speakers came from a variety of countries of origin and had all lived in the United States for at least 1 year. The nations of origin of these participants include Mexico (n = 4), Puerto Rico (n = 4), Venezuela (n = 2), Spain (n = 2), Argentina (n = 2), and Colombia (n = 2). The reason for selecting speakers from various dialectical backgrounds is that the highest proficiency group had also lived in a variety of Spanish-speaking regions, and all three NNS groups had received formal instruction from a diverse group of native and nonnative instructors. Additionally, all but one of the NS participants taught Spanish at the university (the one who did not teach Spanish was a graduate student at the university and had a similar formal educational background to the other NSs). Because the grammatical structure examined in the current study is variable, the NS responses serve as a basis for comparison without requiring the evaluation of accuracy that would be employed if no such variation were demonstrated in NS interpretation of the subjunctive mood. Furthermore, it has been argued extensively that bilingualism, rather than monolingualism, is a reasonable target for L2 learners (Cook, Reference Cook and Han2008; Ortega, Reference Ortega2010) and also that, for language learners with multiple points of contact (e.g., study in several geographic locations, friends and instructors from multiple points of origin), the arbitrary selection of monolinguals from a single geographic point of origin does not produce a meaningful comparison group (Cook, Reference Cook and Han2008; Geeslin, Reference Geeslin2003).

Data Collection

In the current study, each participant completed three activities. Participants spent approximately 5 min on the language background questionnaire, 15 min on the proficiency exam, and 20–25 min on the mood interpretation task. Each of these tasks is described here in greater detail.

Language Proficiency Exam

The language proficiency exam was an independent measure to gauge nonnative proficiency and to confirm the groupings based on the level of enrollment described previously. The exam contained 25 multiple-choice grammatical items in a contextualized story. These discrete-point items correspond to many of the formal properties of Spanish typically included in the language classroom. This instrument was relatively difficult for the lowest proficiency group, and it was also able to distinguish between the other two more advanced NNS groups. Moreover, despite the focus on formal properties of the grammar, some variation among the NSs was also exhibited on this instrument. The results for each participant group are included here to demonstrate that the participant groups were indeed qualitatively different from one another in terms of their level of knowledge of the Spanish language (Table 1). One further notes that the range of scores is lower for the higher proficiency group and that the standard deviation is smaller.

Table 1.

Summary of the results for all participant groups on the proficiency test

A one-way between-subjects ANOVA was run to determine whether there were significant differences among the participant groups according to their proficiency scores. There was a statistically significant difference between groups, F(3, 93) = 79.7, p < .001. A Tukey’s post hoc test revealed that all learner groups were significantly different from one another. The Level 1 group (M = 13.40, SD = 2.96) scored significantly lower on the proficiency test than the Level 2 group (M = 17.83, SD = 3.31, p < .001), the Level 3 group (M = 22.47, SD = 1.33, p < .001), and the NS group (M = 23.44, SD = 1.21, p < .001). The Level 2 group also scored significantly lower than the Level 3 group, p < .001, and the NS group, p < .001. The only comparison that was not significant was that between the Level 3 group and the NS group, p = .717).

Data Coding and Variables

Each completed item on the mood interpretation task constituted a single token, and each was coded for a series of linguistic and extralinguistic variables. The dependent variable in the study was the response for that item (i.e., habitual, not yet occurred, or both possible). The independent linguistic variables (see Table 2) included the mood of the verb form, the regularity of the verbal morphology, the order of the main and dependent clauses, and the adverbial conjunction used. The first independent linguistic variable, mood of the verb form, contains two categories: indicative and subjunctive. As previously mentioned, prescriptively, subjunctive verbs in the adverbial clause indicate that the event has not yet occurred, whereas those in the indicative refer to an event that has already occurred or occurs habitually. The independent variable, regularity of the verbal morphology, also includes two categories: regular and irregular. Regular verbs were defined as any forms for which the indicative and subjunctive stems are the same. An example of a verb with regular morphology can be seen in (3), where the verb visitar has indicative and subjunctive forms that share the same stem (visita “he/she visits [indicative]” / visite “he/she visit [subjunctive]”). Irregular verbs were classified as any verb form whose subjunctive stem differs from its indicative counterpart (e.g., viene “he/she comes [indicative]” / venga “he/she come [subjunctive]”), as in example (4). This follows the work of Gudmestad (Reference Gudmestad, Klee and Face2006) and Geeslin and Gudmestad (Reference Geeslin, Gudmestad, Bruhn de Garavito and Valenzuela2008), who updated Collentine’s (Reference Collentine1995) and Lubbers Quesada’s (1998) original definition of (ir)regularity, to account for the fact that verbs whose stems change in both the indicative and subjunctive are not inherently different for speakers who first encounter the subjunctive. In contrast with Geeslin and Gudmestad (Reference Geeslin, Gudmestad, Bruhn de Garavito and Valenzuela2008), only present tense forms were used in the current study, and, thus, additional categories of this variable were unnecessary.

1. (3) Hablamos del fútbol americano cuando nos visita Marcos.

2. “We talk about football when Marcos visits us.”

3. (4) Salimos del campus cuando venga María.

4. “We are going to leave campus when Maria comes.”

Table 2.

Summary of the coding scheme

The third independent linguistic variable, order of the main and dependent clauses, allows for the codification of the distinction between preposed and postposed main clauses, as illustrated in examples (5) and (6), which prepose and postpose the main clause, respectively. Both examples contain forms in the indicative, but subjunctive forms are also possible in these sentence-level contexts for events that have not yet occurred.

1. (5) Bebemos cerveza cuando llega Juan.

2. “We drink beer when Juan arrives.”

3. (6) Cuando salen los estudiantes, los profesores hablan del fin de semana.

“When the students leave, the professors talk about the weekend.”

Finally, the variable adverbial conjunction classifies the first word or phrase of the subordinate clause. The categories of this variable are cuando “when,” hasta que “until,” and después de que “after.” The linguistic variables and the categories that they each subsume are summarized in Table 2.

Because of the manner in which the task was designed (i.e., creating contexts in which all combinations of the categories of all of the linguistic variables in the study occurred), the coding of each token for these linguistic variables was consistent across participants, with each participant completing one item for each condition. Additionally, the extralinguistic variables of participant age, gender, nation of origin (for the NSs), and nation of study abroad or prior residence (for those NNSs to whom this applied) were also included in the analysis. It will be recalled that our first two levels of learners were essentially all the same age, whereas our Level 3 learners were all at least 22 years of age. Our NSs were later divided as under 30 years of age or 30 years of age or older. A summary of the nations of origin or study abroad and the time spent abroad by the learners was presented previously in the description of the participants, as was a summary of the ages of the participants in each group. Additionally, the individual participant was also coded as an extralinguistic variable to enable us to see whether individual variability within a participant group was selected as a significant predictor of interpretation.

Data Analysis

Following data collection and coding, the distribution of the responses for each participant group was tabulated. Next, individual chi-square tests were run (using SPSS) to determine which of the independent variables were statistically related to the response choice for a given participant group (p < .05). Third, a binomial, step-up/step-down regression was run using SPSS to determine which of the linguistic and extralinguistic independent variables contributed to the prediction of response choice (i.e., interpretation of an event as habitual or having not yet occurred) when all factors were considered simultaneously in a single statistical model. In the regression analysis, the response “both interpretations are possible” has been included with the “not yet occurred” response because the indicative is encroaching on terrain previously occupied by the subjunctive, and, thus, language change is believed to be moving in the direction of allowing an association previously associated only with the subjunctive (Geeslin, Reference Geeslin2003; Silva-Corvalán, Reference Silva-Corvalán1994). In other words, a subjunctive form was believed to be required (categorically) for an interpretation that an event had not yet occurred, and, thus, any indicative context that allows this interpretation is evidence of language change or variability.^{Footnote 6} The final portion of the analysis is an in-depth examination of the distribution of responses across the categories of each of the independent linguistic variables to determine the magnitude and directions of the influence of these variables. This part of the analysis is based on crosstabulations as well as chi-square tests for each variable for each participant group.