Abstract
Objectives:
Previous research has found that object marking with the Spanish particle a is variable among heritage Spanish speakers, based on data from offline experimental measures. The online self-paced reading method used for the present study has the advantage of being less metalinguistic, which may be important with heritage speakers. This incremental measure of sentence processing can also examine whether the single letter form a is simply skipped over during reading.
Methodology:
Thirty-two heritage Spanish speakers and 16 later Spanish-English bilinguals participated. Critical stimuli were 20 items testing a marking with direct objects and 20 items testing a marking of indirect objects in ditransitive constructions.
Data and analysis:
The data set included reading times from self-paced reading, accuracy for post-stimulus comprehension questions, and secondary data from an offline acceptability judgment, all of which were analyzed via ANOVAs by subject and by item.
Findings:
Both groups exhibited robust sensitivity to a marking of indirect objects, sensitivity to a marking with inanimate direct objects, and no sensitivity to a marking with animate direct objects.
Originality:
This is the first study to examine the real-time processing of object marking among heritage Spanish speakers. It is the second study to include a comparison group of late Spanish-English bilinguals in the US, as opposed to monolinguals residing abroad.
Significance:
Incremental data from online processing indicate that the visual nonsalience of the marker a is not the sole or primary cause of variability in the marking of animate direct objects, because it was noticed in other written sentential contexts. In addition, the similarity between the two participant groups shows that variability with differential object marking is not limited to heritage speakers, but can also occur among US Spanish users educated abroad, where incomplete acquisition is not a question.
Keywords
Heritage speakers, or early bilingual acquirers of a home or community language that is not used by the majority of the broader society, often show interesting linguistic traits and variation that can arise due to multiple factors such as incomplete acquisition, attrition, and cross-linguistic influence. For instance, heritage bilinguals appear to have special advantages over second language learners when it comes to pronunciation and phonological perception, presumably because of earlier exposure to the language (e.g. Au, Knightly, Jun, & Oh, 2002; Knightly, Jun, Oh, & Au, 2003). At the same time, they can have more limited word knowledge and less efficient lexical access in the minority language than in the dominant language (O’Grady, Schafer, Perla, Lee, & Wieting, 2009). Heritage speakers are also known to differ from monolingual and late bilingual speakers that were raised abroad in language majority contexts with regard to a range of language areas such as inflectional morphology (e.g. Benmamoun, Albirini, Montrul, & Saadah, 2014), morphosyntactic agreement (e.g. Montrul, Davidson, de la Fuente, & Foote, 2013), semantics (e.g. Montrul & Ionin, 2010, 2012), and discourse-level dependencies (e.g. Keating, VanPatten, & Jegerski, 2011).
Thus far, relatively little research has examined online sentence processing behavior among heritage speakers using real-time methods like self-paced reading, eyetracking, and ERPs (event-related potentials). Such measures allow the researcher to record moment-by-moment behavior as participants read through a stimulus sentence, which provides a different perspective on comparison between different populations of language users. For instance, one methodological concern that is specific to the study of heritage language is that heritage bilinguals typically have had limited opportunities for formal study of the home language and therefore tend to have underdeveloped literacy, metalinguistic skills, and formal register, as compared to their own language skills in English. This is different from the case of native speakers raised in a language majority context and formally educated in their native language, or even formally educated second language learners. This important difference between populations of language users raises the question of whether heritage speakers may respond differently to experimental measures than other participant populations to which they are frequently compared. Indeed, there is some empirical evidence that this is the case and that tests and tasks that are more metalinguistic or explicit with regard to target form can be biased against heritage bilinguals and lead to inconsistency of outcome across studies conducted with different tests of language knowledge and skill (e.g. with gender agreement, Montrul, Davidson, de la Fuente, & Foote, 2014; Montrul, Foote, & Perpiñán, 2008). In addition, several researchers have advocated for a shift away from traditional grammaticality judgment measures in heritage language research (Benmamoun, Montrul, & Polinsky, 2010, 2013) and towards less explicit psycholinguistic methods (Bolger & Zapata, 2011), even though only a handful of studies so far have done so.
The present study used the self-paced reading method to examine the processing of the Spanish case marker a on direct and indirect objects by heritage bilinguals. Previous research using offline interpretation and production measures has observed that a marking of animate direct objects is variable among heritage speakers. The results of the current investigation show that there is no online preference for a marking with animate direct objects, even though there are robust effects during the processing of a marking with indirect objects. Thus, despite the visual nonsalience of the single letter form in text, the a marker is not categorically overlooked by heritage speakers in reading, so salience alone cannot explain the variability observed here and in previous research. In addition, the inclusion of a comparison group of late L1 Spanish-L2 English bilinguals (later immigrants to the US) and the overall similarity of this group to the heritage bilingual group suggests that variability with a cannot be attributed to incomplete acquisition of the heritage language, a finding that is consistent with a recent proposal by Montrul (2014; Montrul & Sánchez-Walker, 2013) that intergenerational transmission is a factor in variability with Differential Object Marking in heritage Spanish.
Object marking with the particle a in Spanish
Indirect objects in Spanish ditransitive constructions are marked with a. In examples such as (1) below, in which the ditransitive verb to send requires two arguments, the second argument is marked with a. This marking is consistent and there are no semantic restrictions on the use of a with indirect objects. Still, it is most often seen with animate NPs, since this is the most common type of second argument that occurs with ditransitive predicates. It is also preferred in most dialects to see the indirect object doubled with a preverbal clitic, le in example (1). The a marker closely resembles a preposition in these contexts, as evident in the English gloss in (1), although it also has some properties that suggest it is a case marker. One analysis takes both into account by proposing that it is a preposition that assigns case to the indirect object (Zagona, 2002). The present study does not assume any specific formal account of indirect object marking with a a priori, but if a is more similar to a preposition in this context than with direct objects, that could be a factor in the apparent difference in acquirability between the two uses of a.
(1) La alumna le mandó un ensayo “The student sent the teacher an essay this morning.”
Direct objects in Spanish are also marked with a, but this occurs only under certain conditions and has therefore been dubbed Differential Object Marking (DOM; Bossong, 1991). Spanish has flexible word order, so overt case marking is used primarily on those direct objects that could most easily be confused for subjects, meaning those that are both animate and definite (Aissen, 2003). Although the illustration of DOM in (2) and (3) is straightforward, the system in general is quite complex because a variety of factors can lead to exceptions to this general rule (Aissen, 2003; Torrego, 1998). For the purposes of the present study, only cases in which case marking can be determined on the basis of animacy will be considered.
(2) Animate direct object María conoce “Maria knows the principal of the school.” (3) Inanimate direct object El plomero conoce el edificio de la escuela. “The plumber knows the school building.”
Even though direct and indirect objects are both marked with the same particle, there is theoretical and empirical evidence that suggests that the two types of case marking have important differences. First, as mentioned above, a may be more like a preposition (although still distinct from the homonymous lexical preposition) when used in ditransitives than with direct objects. In addition, work in formal linguistics has suggested that case can be either structural or non-structural (Chomsky, 1981), and this classification has been applied to object marking in Spanish (Torrego, 1998). Finally, recent empirical work examining both types of case marking with monolingual and bilingual speakers of Spanish has found that the marking of indirect objects generally does not vary, but there is considerable variability with DOM among the same language users (Hopp & León Arriaga, 2016; Jegerski, 2015; Montrul, 2014), which further suggests that the two types of a marking have important differences. These differences in the grammatical status of a in the two different linguistic contexts could potentially interact with the low salience of the form, which is discussed in greater detail in the next section.
Spanish case marking among heritage bilinguals in the US
Although it has not been very widely investigated, there is some evidence that the marking of indirect objects is relatively consistent among heritage Spanish speakers. Montrul (2014) found that simultaneous and sequential heritage bilinguals marked around 95% of indirect objects with a and showed a similarly high rate of accuracy in an aural and written interpretation task. Montrul and Bowles (2009) found that advanced and intermediate proficiency heritage speakers rated sentences with unmarked indirect objects as unacceptable, even though their ratings were closer to acceptable than were those of a comparison group of monolingually raised native speakers (2.0 versus 1.2 on a 5-point Likert scale of increasing acceptability), and a lower proficiency heritage speaker group in the same study was undecided in these judgments.
On the other hand, several studies of DOM among heritage speakers have observed that it is quite variable and often quantitatively different from that of monolingual speakers of non-US varieties. Studies of heritage Spanish using oral production measures have observed that inanimate direct objects are consistently unmarked, at rates of 95 to 100%, but animate (and specific) direct objects are marked on average about 70 to 80% of the time (Montrul & Bowles, 2009; Montrul & Sánchez-Walker, 2013), and the rate of DOM can vary with proficiency level from 50 to 90% (Montrul, 2004; Montrul & Bowles, 2009). Acceptability judgments by heritage bilinguals have shown ratings of 3 to 3.5 on a 5-point Likert scale for unmarked animate objects, as compared to ratings around 1.2 by monolingually raised Spanish speakers (Montrul & Bowles, 2009). Montrul (2014) found similar variability as measured by two other offline tasks, with simultaneous and sequential heritage bilinguals – and also adult immigrants to the US exposed to English only after age 18 – all showing around 80% marking of animate objects in written production and 80% accuracy in aural and written interpretation. Thus, variability with DOM in heritage Spanish is a consistent empirical finding, as is the observation that most of the variability occurs in the marking of animate direct objects, even if they are still marked most of the time.
One limitation so far in this line of research on heritage bilinguals is the absence of online measures of the processing of Spanish case marking. Prior investigation of DOM with heritage Spanish speakers has employed a range of measures, including written picture-sentence matching, auditory picture-sentence matching, oral story retelling, oral picture description, and prompted written production, as well as the more traditional acceptability judgments (Montrul, 2004, 2014; Montrul & Bowles, 2009; Montrul & Sánchez-Walker, 2013). Still, no previous study has used an online measure to examine moment-by-moment behavior as it unfolds in real time during the processing of sentences with DOM. Online measures of processing like the self-paced reading used for the present study also have the advantage of being more implicit and less metalinguistic (at least when administered separately from any judgment measures), which could be particularly advantageous in research on heritage speakers, since they typically have less metalinguistic knowledge and experience with explicit language tasks than second language learners and native speakers formally educated in Spanish and this may affect their performance in some experiments (Montrul, 2016; Montrul, Davidson, de la Fuente, & Foote, 2014; Potowski, Jegerski, & Morgan-Short, 2009).
Another advantage of measuring online processing as it occurs incrementally is that the effects of form salience can be examined more directly. Perceptual salience is when a sensory stimulus stands out and is thus more likely to be noticed and processed. In listening comprehension, the Spanish a particle in question is but a single sound that is often contracted with a verb that precedes it or a masculine definite article that follows it, so it has been proposed that the acoustic nonsalience of the form may lead to its being less likely to be acquired in the grammar (Montrul, 2014; Montrul & Sánchez-Walker, 2013). In the comprehension of written language, the a form is probably more easily noticed than in listening because in many cases it is written as a separate word, but it is also frequently contracted with a masculine definite article el that follows it, so it could often be easy to miss (e.g. al presidente vs el presidente).
Although the processing of written input is unlikely to have played a major role in the childhood acquisition of Spanish by heritage speakers, it is relevant to the study of adult heritage speakers because it could affect continued language maintenance and use, particularly in educational settings, and the outcomes of research that uses written materials. Heritage speakers could be especially likely to skip over such a small detail as the a particle in reading, since they often have underdeveloped literacy skills in Spanish (Carreira & Kagan, 2011), but previous studies have only explored this question indirectly. The present study is the first to examine this question of form salience more directly by employing an incremental processing measure to record separate reading time data for individual sentence segments like al presidente and el presidente in order to more closely examine whether the nonsalience of the a form is associated with its not being noticed, and whether this varies according to the linguistic context in which the form occurs (i.e. with direct objects as compared to indirect objects).
The present study
The present study addressed these methodological limitations in previous research by employing the self-paced reading method to investigate the online processing of the case marker a in three linguistic contexts by heritage Spanish bilinguals. More specifically, this investigation sought answers to the following research questions.
Is the online sentence processing behavior of heritage bilinguals and later bilinguals affected by the absence of the particle a with Spanish ditransitives? Do the two groups differ in this regard?
Is the online sentence processing behavior of heritage bilinguals and later bilinguals affected by the superfluous particle a with Spanish inanimate direct objects? Do the two groups differ in this regard?
Is the online sentence processing behavior of heritage bilinguals and later bilinguals affected by the absence of the particle a with Spanish animate direct objects? Do the two groups differ in this regard?
Method
Participants
A total of 48 Spanish-English bilinguals participated in this study. They were all exposed to Mexican Spanish from birth, were fluent in both Spanish and English, and were recruited on the campus of a large public university in Northern Texas. There were two participant groups that differed with regard to the age of onset of bilingualism. The heritage bilingual group comprised typical heritage speakers who were in the US from a very young age, had all been exposed to English by age five, and thus could also be considered early bilinguals or even simultaneous bilinguals. The later bilingual group comprised immigrants who had not begun learning English until ages 12 to 16. The two groups also differed with regard to Spanish proficiency test score, self-ratings for proficiency in Spanish and English, and estimated relative exposure to both languages, as can be seen in Table 1. In addition, as seen in the standard deviations in Table 1, there was variability within each group with regard to these measures and greater variability within the heritage bilingual group with regard to proficiency scores in Spanish (range 25–45 vs 38–49 for late bilinguals) and self-ratings of proficiency in Spanish (range 4–10 vs 8–10 for late bilinguals), which is common with heritage speakers. On the other hand, the late bilinguals had a wider range of estimated current exposure to Spanish relative to English (range 20–97% vs 0–60% for heritage bilinguals).
Language background information for participant groups.
Materials
There were two sets of stimuli for the self-paced reading. Twenty stimuli tested a marking of indirect objects in ditransitive constructions and 20 additional stimuli tested DOM with direct objects. Both types appeared in grammatical and ungrammatical conditions, with longer RTs predicted in the ungrammatical condition on the critical region. All noun phrase objects were masculine and singular, in order to control for segment length across conditions. Examples of the stimuli and phrase-by-phrase segmentation for self-paced reading are given in (4) and (5).
(4) DOM stimulus a. María \ conoce \ al director \ de la escuela. Grammatical b. María \ conoce \ el director \ de la escuela. Ungrammatical “Maria knows the principal of the school.” (5) Indirect object stimulus a. La alumna \ le mandó \ un ensayo \ al maestro \ esta mañana. Grammatical b. La alumna \ le mandó \ un ensayo \ el maestro \ esta mañana. Ungrammatical “The student sent the teacher an essay this morning.”
The 20 DOM items, each presented in grammatical and ungrammatical conditions, were distributed across two lists in a Latin Square design, as were the 20 doublets with ditransitives. No item appeared more than once per list in any condition. The 40 target stimuli were combined with 120 distractors, including 40 stimuli for another experiment, and the presentation order was pseudo-randomized so that no two sentences of the same type appeared in succession.
General Spanish proficiency level was assessed using an abbreviated version of the highest level DELE exam (Cervantes Institute, 2007) that comprised 50 multiple choice and cloze items testing grammar and vocabulary. This provided a rough estimate of proficiency that would probably not be suitable for course placement purposes or for practical skills assessment, but very similar instruments been used quite widely for at least ten years in research on heritage speakers of Spanish, so it allows for basic comparison of participants across published studies. Following Montrul (2005) and others, scores on this test are commonly interpreted as advanced (40–50/50), intermediate (30–40), and low intermediate (20–30). The scores from the heritage bilingual group in the present study (mean 37.4, range 25–45) were consistent with those from previous research on adult heritage Spanish speakers at the university level (group means around 35–40, ranges 18–48; Cuza & Frank, 2011; de Prada Perez & Pascual y Cabo, 2011; Montrul, 2005; Montrul & Foote, 2014; Montrul & Ionin, 2012).
The second assessment of proficiency in Spanish was a series of three self-ratings items asking participants to gage their own competence for speaking, understanding, and reading on an 11-point Likert scale that ranged from 0 “none” to 10 “perfect”.
Procedure
The self-paced reading stimuli were presented with SuperLab (Cedrus Corporation, 1992) in a non-cumulative, left-to-right presentation. Each trial began with a cue symbol that appeared at the location of the first letter of the stimulus. All of the words of the stimulus were masked with dashes, while spaces and punctuation were visible, and the participant pressed a button to make each subsequent phrase appear and the previous phrase disappear. After each stimulus, a binary choice comprehension question appeared on a separate display screen and the participant responded using keys marked “A” and “B”. No feedback was given. Participants were instructed to read at a normal speed, as if reading a newspaper or a book, and they were told that the test targeted reading comprehension. Detailed instructions and ten practice items were presented prior to the experimental block and an optional ten-minute break was offered when the participant had read half of the 160 sentences.
Participants first completed a background questionnaire and the self-paced reading, both via computer, then took the 50-item proficiency test in written format and completed the acceptability judgment via computer. All experimental tasks were completed in a single research session lasting 90 to 120 minutes and participants were paid for their time.
Results
The primary experimental task was the self-paced reading and secondary data came from the timed acceptability judgment. Statistical analyses were conducted on the rating scores from the acceptability judgment, on the RTs from the self-paced reading, and on the accuracy scores from the meaning-based comprehension questions that followed each of the self-paced reading stimulus sentences. All analyses were conducted on mean scores aggregated by subject and by item, with F1 and t1 used to indicate analysis by subject and F2 and t2 indicating analysis by item. Alpha was set at 0.05 and p values of 0.05 to 0.10 were interpreted as marginally significant, as were effects that were significant only by subject or by item but not both.
The results of the timed acceptability judgment are given in Figures 1 and 2. Mean ratings for each of the two stimulus types were submitted to 2 (Group) × 2 (Grammaticality) mixed design ANOVAs. In the analysis by subject and by item, group was a between- and within-subjects factor, respectively, and grammaticality was a within- and between-subjects factor, respectively. Analysis of the ratings of the ditransitive items showed a main effect of grammaticality: F1(1, 46) = 729.732, p < 0.001, F2(1, 8) = 916.415, p <0 .001, a main effect of group that approached significance by subjects only: F1(1, 46) = 2.944, p = 0.093, F2(1, 8) = 3.193, p = 0.112, and no interaction: F1(1, 46) = 1.407, p = 0.242, F2(1, 8) = 1.858, p = 0.210. Thus, grammatical ditransitives were rated as more acceptable than ungrammatical ditransitives in general and the two participant groups were similar in their ratings and in the reliable difference in ratings for grammatical versus ungrammatical stimuli.
Ditransitives: Mean acceptability ratings on a four-point Likert scale.
DOM: Mean acceptability ratings on a four-point Likert scale.
Analysis of the ratings of the sentences with DOM revealed a main effect of grammaticality: F1(1, 46) = 75.841, p < 0.001, F2(1, 8) = 17.015, p = 0.003, no main effect of group: F1(1, 46) = 1.035, p = 0.314, F2(1, 8) = 0.407, p = 0.541, and an interaction that was significant by subjects only: F1(1, 46) = 8.427, p = 0.006, F2(1, 8) = 2.750, p = 0.136. Separate independent samples t-tests were conducted on the ratings from each group in order to probe the potentially significant interaction. These showed that the main effect of grammaticality held in each of the groups separately, both the heritage bilinguals: t1(31) = 5.174, p < 0.001, t2(8) = 2.495, p = 0.037, and the later bilinguals: t1(15) = 6.734, p < 0.001, t2(8) = 3.771, p = 0. 005. Thus, grammatical DOM was rated as more acceptable than ungrammatical DOM in general and the two participant groups were similar in their ratings, but the difference between the ratings for ungrammatical versus grammatical sentences was larger for the later bilinguals than for the heritage bilinguals.
The reaction time data from the self-paced reading task are presented in Tables 2 and 3. For each of the three stimulus types, mean RTs by subject and by item for each stimulus region from the critical region on were submitted to 2 (Group) × 2 (Grammaticality) mixed design ANOVAs, as were the reaction times and accuracy for responses to the post-stimulus comprehension questions. In the analyses by subject and by item, Group was a between- and within-subjects factor, respectively, and Grammaticality was always a within-subjects factor.
Ditransitive stimuli: Mean response times in milliseconds for stimulus regions of interest and post-stimulus comprehension questions and mean proportion of accurate responses (SDs in parenthesis).
DOM stimuli: Mean response times in milliseconds for stimulus regions of interest and post-stimulus comprehension questions and mean proportion of accurate responses (SDs in parenthesis).
The results of the ANOVAs for the ditransitive stimuli are presented in Table 4. The main effect of grammaticality was significant at the location of the marked or unmarked indirect object NP in Region 4 and carried over to Region 5, with RTs longer in the ungrammatical condition. The grammaticality × group interaction was marginally significant by items only at Region 4 and marginally significant at Region 5, so pairwise comparisons were conducted to explore the potential interaction. At Region 4, the effect of grammaticality held its significance in each of the two groups separately: heritage bilinguals, t1(31) = 3.408, p = 0.002, t2(19) = 2.481, p = 0.023, later bilinguals, t1(15) = 4.951, p < .001, t2(19) = 3.841, p = 0. 001. Thus, the potential interaction reflects a difference in the magnitude of the effect of grammaticality between the two groups, in which the RT difference was greater in the later bilinguals than in the heritage bilinguals (229 vs 130 ms, respectively). At Region 5, the effect of grammaticality was only marginally significant in the heritage bilinguals and was still fully significant in the later bilinguals: heritage bilinguals, t1(31) = 1.672, p = 0.105, t2(19) = 1.935, p = 0. 068, later bilinguals, t1(15) = 3.085, p = 0.008, t2(19) = 3.013, p = 0. 007. Thus, although both groups showed the grammaticality effect over Regions 4 and 5, it was more robust in the later bilingual group. Returning to the omnibus ANOVA, the main effect of group was marginally significant by subjects and significant by items for the RTs for the post-stimulus comprehension questions and it was also significant in the accuracy to the comprehension questions. This reflects a pattern in which the heritage bilinguals responded more slowly and less accurately in general than did the later bilinguals. There were no other significant effects or interactions in the RTs for the indirect object stimuli.
Ditransitive stimuli: 2 × 2 ANOVAs*.
*Effects significant at α = 0.05 appear in boldface.
The results of the statistical analyses for the stimuli with inanimate direct objects are presented in Table 5. The main effect of grammaticality was marginally significant only by subjects at Region 3, the region with the case marker, and also at Region 4. Both groups showed longer RTs for ungrammatical stimuli on both regions, but there was apparently too much variability for the differences to reach full significance. The main effect of group was significant only by subjects in the RTs for the comprehension questions, which potentially reflects a tendency of the heritage bilingual group to respond faster to these questions than the later bilingual group (3718 vs 4253 ms, averaged across conditions). There were no other significant effects or interactions in the RTs for the stimuli with inanimate direct objects.
DOM stimuli with inanimates: 2 × 2 ANOVAs*.
*Effects significant at α = 0.05 appear in boldface.
The results of the statistical analyses for the stimuli with animate direct objects are presented in Table 6. The main effect of group was significant at Region 4, the sentence-final region, where the RTs for the heritage bilingual group were overall faster than those for the later bilingual group (1078 vs 1299 ms, averaged across conditions). It appears that the heritage bilinguals did not take as much time for sentence wrap-up with these stimuli, typically evident as longer RTs on the last part of a stimulus sentence. There were no other significant effects or interactions in the RTs for the stimuli with animate direct objects, although there was a numerical difference present only in the later bilingual group across Regions 3 and 4, in which ungrammatical stimuli were associated with longer RTs.
DOM stimuli with animates: 2 × 2 ANOVAs*.
*Effects significant at α = 0.05 appear in boldface.
The results of this experiment can thus be summarized as follows.
Both participant groups showed robust online sensitivity to case marking with ditransitives across two stimulus regions, although the heritage bilingual group was slower and less accurate in the responses to the post-stimulus comprehension questions.
Both groups also showed evidence of sensitivity to DOM with inanimate objects across two sentence regions, but it was generally less robust than with the ditransitives. Variability was greater, so the differences did not reach full significance.
There was no significant online sensitivity to DOM with animates. Nevertheless, the later bilingual group showed a smaller and nonsignificant numerical RT difference in this direction.
In addition, the offline acceptability ratings were similar for both groups for marking of indirect objects in ditransitive constructions, but the preferences of the later bilinguals with DOM were clearer (i.e. more differentiated) than were those of the heritage bilinguals.
Discussion
The first research question pertained to online sensitivity to the particle a on indirect objects in Spanish ditransitive constructions among heritage bilinguals and later bilinguals. Both participant groups showed evidence of processing difficulty upon encountering an unmarked indirect object, which suggests that both types of bilinguals have the relevant grammatical principles and quickly integrate them during online sentence processing. This outcome builds on the offline evidence provided by one previous study (Montrul, 2014), in which heritage bilinguals and later adult immigrant bilinguals were found to be largely similar to monolingual Spanish speakers in Mexico in terms of their written production and comprehension of a marking on indirect objects in ditransitive constructions. The RT effects are also consistent with the self-paced reading data from two monolingual L1 Spanish groups and two high-proficiency L2 Spanish groups in Mexico and Spain reported by Jegerski (2015), and with the eyetracking data from adult immigrant L1 Spanish and advanced proficiency L3 Spanish users in Germany reported by Hopp and León Arriaga (2016). With all six of the aforementioned participant groups tested, there were immediate and robust RT effects for unmarked indirect objects, but no effects on sentence comprehension as measured by post-stimulus questions. Thus, this first part of the results of the present study contributes to a growing body of data showing that the particle a on Spanish indirect objects is integrated immediately during online sentence processing by both monolinguals and bilinguals and that omission of the particle is associated with notable RT effects, but does not appear to affect ultimate comprehension of sentence meaning.
The apparent lack of variation with a marking of Spanish indirect objects among a range of different monolingual and bilingual users stands in contrast with the situation for Spanish DOM with the a particle (discussed below), a finding that has been consistent across studies that have examined both uses of the a marker in Spanish (Hopp & León Arriaga, 2016; Jegerski, 2015; Montrul, 2014). The difference was also present in the offline judgment data from both participant groups in the present study, which showed clearer preferences with indirect objects than with direct objects. The difference in experimental effects associated with the two different types of case marking with a in Spanish could be taken as support for the structural/lexical case contrast proposed in formal linguistics (Chomsky, 1981), if dative marking on indirect objects is seen as a uniform structural operation, whereas selective accusative marking on direct objects involves semantic specifications that vary with each individual item.
The second research question pertained to online sensitivity to a superfluous particle a on inanimate objects in Spanish among heritage bilinguals and later bilinguals. As with the ditransitives, the two participant groups showed similar online sensitivity to the experimental manipulation in the stimuli, but the effects were not as robust. Still, the data suggest that both types of bilinguals were similar in their attempt to integrate the case marker during online sentence processing.
The third and final research question pertained to online sensitivity to a missing particle a on animate objects in Spanish among heritage bilinguals and later bilinguals. Neither participant group showed reliable effects of grammaticality with these stimuli, which suggests that their grammars do not consistently require a marking of animate direct objects or that the marking requirement is not always integrated during online sentence processing. It is not likely that the absence of the a particle was simply not noticed, given that there were robust effects when the a was missing on indirect objects in the ditransitive stimuli, as discussed above. In addition, there was also a hint of the potential for the missing marker to affect processing in the later bilingual group, who showed a numerical trend in the right direction. The results from the two bilingual groups in the present study stand in contrast with those of two monolingual L1 Spanish groups from one prior self-paced reading study (Jegerski, 2015) and of an adult immigrant L1 Spanish group from an eyetracking study (Hopp & León Arriaga, 2016), all three of which showed fully significant RT effects for missing DOM on animate direct objects in stimuli similar to those for the present study. Thus, the two bilingual groups were not reliably different from each other with regard to DOM of animates, but they were different from the monolingual Spanish users from previous research.
Taken together, the results for DOM with animate and inanimate direct objects are largely consistent with the observations of previous research on heritage Spanish using offline measures, which had shown that variability in DOM occurs with animate direct objects but not with inanimate objects. Specifically, heritage Spanish speakers frequently produce unmarked animate direct objects or find them acceptable in judgments, but they typically do not overextend the a marker to inanimate direct objects in production or find such marking acceptable in judgments (Montrul, 2004; Montrul & Sánchez-Walker, 2013). The same asymmetry was present in the results of one prior eyetracking study of the L2 processing of Spanish DOM, which observed that an L2 group had RT effects for the grammaticality of case marking with inanimate direct objects but not with animates (Hopp & León Arriaga, 2016).
Regarding the question of why DOM is variable among heritage Spanish speakers, the present study can speak to the potential roles of incomplete acquisition and cross-linguistic influence. First, the results do present some minimal support for the importance of incomplete acquisition with DOM. Although there were no significant differences between the heritage bilinguals and the later bilinguals in the online data, there was a numerical difference, and the heritage bilingual group also exhibited less determinate offline judgments. This suggests that variability with DOM may be affected by extensive early exposure to English, which can curtail early acquisition of Spanish. Second, there is also evidence that at least some of the variation seen with DOM in heritage Spanish can also occur with other native speakers of Spanish who immigrate later to the US and who therefore would not experience incomplete acquisition in childhood. Although the later bilingual participant group in the present study was not identical to the heritage bilingual group in terms of the online and offline results for DOM, they still did not show the significant online effects that had previously been observed among monolingual L1 Spanish speakers abroad (Jegerski, 2015) and even among adult immigrant L1 Spanish speakers residing in another L2 environment (Hopp & León Arriaga, 2016). This suggests that their Spanish may have changed as a result of residence in the US, despite their arrival in adolescence rather than in early childhood, their high proficiency in Spanish, and that by their own estimation they were more proficient in Spanish and they were using Spanish slightly more than English at the time of testing. This change could be due to cross-linguistic influence at the individual level, resulting from the acquisition and use of English, to cross-linguistic influence already present in the community, if Spanish-speaking immigrants to the US adopt the characteristics of US Spanish, or both. Either way, the outcome of this study is consistent with the picture painted by Montrul (2014; Montrul & Sánchez-Walker, 2013).
The heritage speaker participants in the present study were all very early acquirers of English, or simultaneous bilinguals (age of onset 0–5), but heritage speakers can also be sequential bilinguals (age of onset 6–12). Given the overall similarity between the simultaneous bilingual group and the later bilingual group (age of onset 12–16) in the present study, a prediction that follows is that sequential bilinguals would be similar as well, showing online sensitivity to a marking with indirect objects and inanimate direct objects but not with animate direct objects. Still, this hypothesis would have to be empirically confirmed with future research.
As for the processing of the particle a, this study was the first to examine whether the single letter marker tends to be overlooked by heritage bilinguals during reading. It had previously been suggested that the acoustic nonsalience of a might be an important factor in its variability (Montrul, 2014; Montrul & Sánchez-Walker, 2013). In the present study, a reading-based measure was used to explore the effects of visual nonsalience. The a marker was easy to overlook in all of the written self-paced reading stimuli (i.e. animate direct objects, inanimate direct objects, ditransitives) because it was always contracted with the masculine definite article (i.e. a + el = al) and the stimulus conditions being compared were identical except for the one letter. Minor differences such as el presidente versus al presidente could simply be missed in reading for meaningful comprehension, especially because short words like articles are more likely to be skipped than longer words, which have higher overall fixation rates in eye movement studies. Nevertheless, significant RT differences were seen on regions of interest with small differences like these, including among the heritage speaker group, with inanimate direct objects and with indirect objects. Thus, the marker was noticed and processed most of the time, which suggests that the variability observed with a marking of animate direct objects is not exclusively a salience issue. This is not to say that heritage bilinguals never overlook the particle during reading. Although the self-paced reading measure used in this experiment did not provide separate RTs for the a marker alone, it is likely that it was skipped over some of the time by all of the participants because it is such a short word. The real question would be whether heritage bilinguals miss the particle more than late bilinguals or immersed L1 speakers abroad, but this point could be better investigated with eyetracking and is thus an open question for future research. Hopp and León Arriaga (2016), for instance, found that L2 learners skipped the definite article more often than did native readers (40.8% L2 versus 28.2% L1) and that this was regardless of whether the a case marker was contracted with an article. Another approach to exploring the role of salience more in depth would be to include stimuli in which the a marker stands out more, such as with feminine objects, since there is no contraction in those cases (e.g. a la presidente vs al presidente).
Given that perceptual salience cannot explain why variability is observed with a marking of direct objects but not with indirect objects among Spanish-English bilinguals, it is possible either that the low salience of the a form does not play a role, or that it does but only as part of an interaction with another factor or factors. For instance, as already explained in the background section, the a marker used with indirect objects may be more like a preposition and thus have more lexical content than the a marker used with direct objects. A linguistic form that carries meaning may be more readily acquired or maintained than one that carries out a purely grammatical function. Another potential linguistic difference between the use of a with direct and indirect objects is that direct object marking can be described as non-structural case, but indirect object marking is structural case, as per the dichotomy originally proposed by Chomsky (1981). Structural case should be easier to acquire because only one generalized syntactic principle is applied categorically, whereas non-structural case should be more difficult because it depends on a large number of individual lexical specifications that must be acquired and maintained (e.g. Eisenbeiss, Bartke, & Clahsen, 2006). Looking beyond the linguistic differences between a marking of direct and indirect objects, the two also appear to differ with regard to processing complexity. In both cases, there is the morphosyntactic marking of an object, but with direct objects there is the added layer of the semantics of the object, since marking only occurs with direct objects that are animate and specific. Finally, a marking of direct objects seems to be subject to more variability than with indirect objects in general, not just with US bilinguals or at the present moment in time. Other contexts of variation include contemporary Argentine Spanish, which shows extension of the a marker to inanimate direct objects (Hoff & Díaz-Campos, 2015; Montrul, 2013), and diachronic change, as a-marked animate direct objects have slowly increased in frequency over a period of at least seven centuries (von Heusinger, 2008). In light of this broader picture, it is perhaps not surprising that there is also variability in DOM among Spanish heritage speakers.
One limitation of this investigation is that it used a reading based measure of sentence processing. This did not appear to present any critical methodological issues, as the heritage bilingual participants showed evidence of reasonably fluent reading in Spanish, including significant RT effects that are consistent with those of comparison groups in the present study and in previous research. At the same time, the data from this experiment also suggest that the heritage bilingual group was not quite as fluent as the comparison group of later bilinguals. Evidence for this comes from the self-rating scores for reading ability in Spanish (7.7/9.3 Spanish/English for the heritage bilinguals vs 9.4/8.2 for the later bilinguals on a 0 to 10 Likert scale), as already reported in the methodology section. In addition, a post-hoc analysis of average RTs and comprehension accuracy for the entire self-paced reading experiment, including all 160 total target stimuli and fillers, revealed numerically longer RTs for the heritage bilingual group: 935ms vs 886ms; t(46) = 0.804, p = 0.426, and significantly lower comprehension scores for the heritage bilingual group: 82.5% vs 87.8%; t(46) = 2.085, p = 0.043. Thus, it seems there was a difference in reading fluency, but the difference was not very large and it likely did not affect the results of the experiment. This suggests that self-paced reading can be an appropriate method for research on heritage speakers, at least those with good reading fluency in the heritage language. It also leaves open the possibility that there were more notable differences in literacy skills between the two participant groups, but that these were at a more advanced level than the basic reading of Spanish sentences that was required for the present study.
Another limitation is that, although the pattern of results is consistent with intergenerational transmission playing a role in variability with DOM, this was not directly examined in the present study. Although it seems likely, it is not known for certain whether the participants in the heritage bilingual group were exposed to input in childhood from first generation immigrants like those in the later bilingual group, since there was no known relationship between the individuals in the two groups (cf. Montrul, 2014, for a design that addresses this question more directly).
In conclusion, the present study has contributed to our knowledge of variability in heritage Spanish in the marking of animate direct objects with the a particle, as well as consistency in the marking of indirect objects with a, using online data from a real-time measure of sentence processing. Data from meaning-oriented self-paced reading have provided the first precise RT evidence that the a marker is not simply overlooked by heritage speakers in reading, despite the visual nonsalience of the written form. In addition, the overall similarity of sentence processing by heritage bilinguals and late Spanish-English bilinguals suggests that variability with DOM cannot be explained solely in terms of incomplete acquisition of the heritage language, as it also occurs among US bilinguals who immigrated after formal schooling abroad in Spanish. This finding is consistent with the multifactorial account proposed very recently by Montrul (2014; Montrul & Sánchez-Walker, 2013), in which intergenerational transmission is a factor.
Footnotes
Declaration of conflicting interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The writing of this report was supported by a Humanities Released Time award from the Campus Research Board of the University of Illinois at Urbana-Champaign, for which I am grateful.