Abstract
This study investigates facilitative processing of grammatical gender in heritage Spanish speakers whose dominant language is German, using eye-tracking in the Visual World Paradigm. Bilinguals with two gender systems are known to have an integrated mental lexicon with shared gender features that are co-activated during language processing and can result in interference. The present study shows that, despite observed effects of gender congruency with German, heritage speakers were able to use gender information on prenominal articles in Spanish to facilitate lexical retrieval of the target noun. This suggests that processing of gender agreement in the heritage language is resilient to competition from gender in the dominant language during real-time spoken-language comprehension. Moreover, direct comparison with previous results from heritage Spanish speakers in the USA does not show evidence that overall speed of facilitative processing in the heritage language is modulated by the presence or absence of gender in the majority language.
1 Introduction
Heritage speakers (HSs) consistently show non-target-like performance in the production and comprehension of grammatical gender (e.g., for Spanish, Montrul et al., 2014; Montrul et al., 2008; Scontras et al., 2018). Nevertheless, work on language processing has found that HSs’ real-time processing of gender agreement is qualitatively target-like (e.g., Fuchs, 2021a, 2022; Di Pisa et al., 2022; Luque et al., 2023). These findings connect to a broader endeavor of investigating HSs’ anticipatory processing – the ability to use information or features available in the speech stream to pre-activate some aspect of the linguistic representation before encountering the relevant linguistic element (Pickering & Gambi, 2018). Anticipatory processing has been widely studied in monolingual populations and is commonly studied in the literature on second language acquisition (for overview, see Kaan & Grüter, 2021), but it is still very much understudied in heritage populations. The endeavor in heritage linguistics is proceeding along two dimensions, investigating (A) how the type of feature used to pre-activate linguistic representation impacts HSs’ anticipatory processing, and (B) whether overlap – or lack thereof – in this feature between the heritage language and the majority language aids or interferes with anticipatory processing.
In pursuit of (A), researchers have found evidence for HSs’ anticipatory processing using a variety of linguistic elements, including grammatical gender agreement marking (Sekerina, 2015; Fuchs, 2021a, 2022; though see results from a reading study by Parshina et al., 2022), number marking (Parshina et al., 2022), classifiers (Ito et al., 2023), case markers (Karaca et al., 2023; Özsoy et al., 2023; Meir et al., 2020), and verbal lexical semantics (Ito et al., 2023). Regarding (B), too few studies exist to allow a general claim, but preliminary evidence suggests that at least the absence of a feature in the majority language does not prevent anticipatory processing of that feature in the heritage language, e.g., classifiers (Ito et al., 2023) or case (Karaca et al., 2023). Investigations of how overlap in a feature between the heritage and majority languages impacts anticipatory processing are also few. Özsoy et al. (2023) found that HSs of Turkish whose dominant language was German were able to use case marking predictively in their heritage language; Meir et al. (2020) found similar results for HSs of Russian whose majority language was Hebrew. On the other hand, Ito et al. (2023) found that overlap in verbal lexical semantics interfered with anticipatory processing by HSs of Vietnamese whose dominant language was German, specifically when the German translational equivalent verb had different selectional restrictions than the Vietnamese verb being tested.
Question (B) has only partly been addressed for HSs’ anticipatory (or facilitative) processing of grammatical gender. Existing work has investigated HSs whose dominant language was English, which does not have grammatical gender, and has shown that the absence of gender in the majority language does not prohibit facilitative processing of gender agreement in the heritage language (Sekerina, 2015; Fuchs, 2021a, 2022). However, given the lack of work in this domain on dyads in which both languages have a gender system, it is unclear how the presence of gender in the majority language affects facilitative processing patterns. This is additionally motivated by work that shows that, for bilinguals who speak two languages with grammatical gender, the gender systems interact during language processing during production and comprehension (see Sá-Leite et al., 2020 for an overview), and their representation may even be shared in the mental lexicon (Salamoura & Williams, 2007). We thus build on existing work by investigating how the existence of a gender system in the majority language impacts facilitative processing of gender agreement in the heritage language. We specifically investigate whether HSs of Spanish whose dominant language is German are able to efficiently access and integrate gender agreement features on prenominal articles to facilitate lexical retrieval during language comprehension in Spanish. This would constitute evidence for HSs’ ability to process gender agreement in a target-like manner, despite competition from the gender system in the majority language.
2 Background
2.1 Gender in Heritage Spanish
Spanish, the heritage language investigated in the present study, has two gender categories, masculine and feminine; most feminine nouns end in -a, while most masculine nouns end in -o, although there are many exceptions. Various formal and experimental evidence suggests that masculine is the default gender (e.g., Beatty-Martínez et al., 2020; Fuchs et al., 2015; Harris, 1991). Most relevant for this study is that the masculine form of the definite article, which occurs prenominally, is el, while the feminine form is la.
Both monolingual and bilingual Spanish-speaking children appear to be largely error-free in their production of gender agreement of determiners by around age 2;7–3;0, suggesting they have acquired language-specific gender categories and agreement by this age (e.g., Eichler et al., 2013; Kuchenbrandt, 2005; Pérez-Pereira, 1991; Ticio Quesada, 2018). In adulthood, HSs’ comprehension and production of gender agreement is consistently non-target-like (e.g., Alarcón, 2011; Boers et al., 2020; Montrul et al., 2008). However, there is a growing body of evidence that suggests bilinguals’ performance on a linguistic task may be dependent on the modality of the task (e.g., Jegerski & Sekerina, 2020), and indeed, with respect to gender agreement, HSs’ performance seems to be qualitatively more target-like in online tasks, which measure real-time language processing, than in offline tasks, which measure the outcome of a linguistic process and which often, in heritage linguistics, target explicit knowledge and/or require metalinguistic judgment; these may in turn be affected by non-language-internal factors such as linguistic insecurity, yes-bias, or lack of formal education in the language (e.g., Alarcón, 2011; Montrul et al., 2014; for further discussion see Bayram et al., 2020).
2.2 Two Languages, Two Gender Systems
A substantial portion of the literature on heritage languages, including work on grammatical gender, has been focused on those languages that develop in the presence of English as the dominant language. English does not have grammatical gender, and therefore any non-target-like performance in production and/or comprehension of grammatical gender in a given HL could in principle be attributable to the lack of grammatical gender in the majority language. However, the available evidence suggests that HSs are non-target-like in production of gender agreement in their HL even when their dominant language has grammatical gender (Müller, 1990, 1999; Boers et al., 2020; Janssen, 2016; Rodina et al., 2020). Work on the degree to which the properties of the gender system in the majority language can affect bilingual children’s acquisition of the gender system in the HL has found mixed results (Brehmer & Rothweiler, 2012; Cornips & Hulk, 2008; Egger et al., 2018; Eichler et al., 2013; Kaltsa et al., 2019; Meir et al., 2017; Rodina et al., 2020; van der Linden & Hulk, 2009).
For adult bilinguals, evidence from work on language processing shows that the two gender systems in the mental lexicon of a bilingual speaker interact with each other, in that the processing of a noun in one language appears to be modulated by the gender of the translational equivalent in the other language (e.g., Salamoura & Williams, 2007). This effect has been observed in language production in oral picture-naming tasks, in which bilinguals are faster to name objects in their L2 when they are gender-congruent, i.e., the grammatical gender of the label in the L2 is the same as the grammatical gender of the translational equivalent in their L1 (e.g., Costa et al., 2003; Klassen, 2016; Lemhöfer et al., 2008, Exp. 2 & 3; Paolieri et al., 2019; for an overview see Sá-Leite et al., 2020). The effect has also been observed during language comprehension in an L2, wherein L2 speakers are faster to recognize nouns whose gender is congruent across their two languages; recent work additionally suggests the effect is modulated by proficiency in the L2 (Hopp & Lemmerth, 2016; Lemhöfer et al., 2008, Exp. 1; Morales et al., 2016). While most work in this area has focused on adult bilinguals who are late L2 learners of their non-dominant language, recent evidence suggests gender congruency effects may also be observed for HSs, at least in production (Sá-Leite et al., 2023), and for bilingual children during language comprehension (Bosch & Foppolo, 2022, 2023).
2.3 Facilitative Processing of Grammatical Gender
It has robustly been shown that a definite article that agrees with the noun in grammatical gender can serve as a cue in the anticipatory processing of the nominal phrase (e.g., for Spanish, see Grüter et al., 2012; Lew-Williams & Fernald, 2007, 2010; for German, see Hopp, 2013). In relevant studies, participants view two (or more) images in a visual display and are directed to look at a target item by an auditory prompt. In this auditory prompt, there is a linguistic element (e.g., article or adjective) that precedes the target noun and crucially agrees with that target noun in gender. When all images in the display are of the same gender (match condition), the first unambiguous cue to the target item is the onset of the noun in the auditory prompt. But when the target and the distractor(s) are of different grammatical genders (mismatch condition), the gender-marked article constitutes an informative cue to the target item prior to the onset of the noun, as there is only one candidate in the visual display that matches the gender agreement cue. In these conditions, participants have been found to fixate on the target item more rapidly. Assuming an interactive activation model of the lexicon (e.g., McClelland & Rumelhart, 1981; Rumelhart & McClelland, 1982),1 this happens because encountering the gender feature on the prenominal element activates the corresponding abstract gender node in the mental lexicon, which then spreads activation to all connected lemmas; this pre-activates lexical items in that gender category and facilitates retrieval of the target lexical item upon the onset of the noun in the auditory input.
While this effect is robust for monolingual speakers, findings are mixed for L2 learners (Dussias et al., 2013; Grüter et al., 2012; Hopp & Lemmerth, 2016; Lew-Williams & Fernald, 2010). In studies of heritage language processing, the effect has been investigated in work on HSs of Spanish (Fuchs, 2021a), Polish (Fuchs, 2022), and Russian (Sekerina, 2015). In these studies, using the paradigm introduced above, participants were faster to direct their gaze to target items in mismatch conditions than in match conditions, given a gender agreement cue on the prenominal article el or la (for Spanish) or on a color adjective (for Polish and Russian), providing evidence that HSs can use abstract gender information on prenominal agreement markers to facilitate lexical retrieval of the noun. Notably, the majority language of the HS groups in these studies was English, so the mental lexicons of the participants contained gender information from only the heritage language. This suggests that the absence of grammatical gender in the majority language does not prevent facilitative processing of gender agreement in the heritage language, but it is an open question what the impact is of the presence of grammatical gender in the majority language. The question is particularly compelling given independent evidence on bilingual language processing that shows that both languages are typically active during language processing and that abstract gender information is shared across two languages if both have grammatical gender (Salamoura & Williams, 2007; Sá-Leite et al., 2020, 2023). This could lead to interference in facilitative processing in the heritage language, especially since connections in the lexicon may be weaker for lexical items in the non-dominant (heritage) language; this motivates the following research question:
Research Question 1: Can heritage speakers use grammatical gender to facilitate lexical retrieval when their dominant language also has a grammatical gender system?
Given the general goal of assessing how absence versus presence of a particular feature in the majority language affects anticipatory processing of that feature in a heritage language (cf. Section 1), we can also use existing data (Fuchs, 2021a) to make a direct comparison between groups of HSs with different majority languages. This motivates a second research question:
Research Question 2: Do heritage speakers differ in their patterns of facilitative processing of grammatical gender depending on whether their majority language does or does not have grammatical gender?
To pursue these research questions, we investigate facilitative processing of grammatical gender in HSs of Spanish whose majority language is German. We motivate this choice of heritage population in Section 2.4 below, before providing experimental details in Section 3.
2.4 Selecting a Dyad
Spanish was selected as the target heritage language for direct comparison with previous work. A number of considerations motivated the choice of German as the majority language in the dyad. First, German has grammatical gender; nouns in the language belong to one of three gender categories: masculine, feminine, and neuter. Moreover, this gender system is non-overlapping with Spanish (cf. Sabourin & Stowe, 2008; Klassen, 2016).2 Work on Spanish-German bilingual children suggests that the early acquisition of the Spanish gender system is not delayed as compared to monolingual Spanish-speaking children (Kuchenbrandt, 2005). Assuming these bilingual children grow up to be HSs, this suggests that adult HS speakers of Spanish in Germany would have acquired the Spanish gender system in childhood.
Finally, there is evidence that congruency in the syntactic environments in which a feature is marked across two languages modulates a bilingual’s ability to use this feature to anticipate properties of the upcoming input, at least at some level of proficiency (Foucart & Frenck-Mestre, 2011; Hopp & Lemmerth, 2016). German has frequently occurring prenominal definite articles that agree with the head noun in gender (in the nominative case: masculine der, feminine die, neuter das), and monolingual German speakers exhibit facilitative processing of gender information on these articles (Hopp, 2013). Thus, for the HSs investigated in the present study, both the heritage language (Spanish) and the majority language (German) not only have grammatical gender but also are syntactically congruent in having prenominal articles that agree in gender with the head noun.
3 Methodology
3.1 Materials
The language of testing for all materials and tasks was Spanish; all materials were a replication of those employed and reported in Fuchs (2021a). Forty-two images representing picturable and concrete nouns were selected based on their gender in Spanish: 21 masculine and 21 feminine. The results of a norming study ensured minimal dialectal variation in the target items. Items were randomly colored blue, red, or green, with an equal number of images of each color. A complete list of lexical items is provided in the Supplementary Materials. These images were also used in an oral picture-naming task used to assess participants’ familiarity with the lexical items and as a measure of lexical proficiency (Lieberman et al., 2018; Fuchs, 2021a, 2022).
Images were combined into displays with two images each, equidistant from a fixation cross (Fig. 1). The pairing of images was based on the images’ Spanish gender (match, mismatch) and congruency of Spanish and German gender (congruent, incongruent, target-congruent). In match displays, the gender of the Spanish words corresponding to each image was the same; in mismatch displays, the gender of the Spanish words corresponding to each image was different, i.e., one masculine and one feminine item. There were a total of 60 match displays and 42 mismatch displays, for a total of 102 displays. In each display, the lexical items corresponding to the images differed from each other in their first phoneme, in order to ensure that the onset of the lexical item served as an unambiguous cue to the target.
Sample visual display: match (Spanish: casco ‘helmetM’; martillo ‘hammerM’) & congruent (German: Helm ‘helmetM’; Hammer ‘hammerM’)
Citation: Heritage Language Journal 21, 1 (2024) ; 10.1163/15507076-bja10028
Sample visual display: mismatch (Spanish: casco ‘helmetM’; vela ‘candleF’) & congruent (German: Helm ‘helmetM’; Kerze ‘candleF’)
Citation: Heritage Language Journal 21, 1 (2024) ; 10.1163/15507076-bja10028
In order to test for effects of congruency (cf. Sá-Leite et al., 2023), there were three congruency conditions, tracking the relationship between the Spanish label of each image in the display and the German translational equivalent of the given label (Table 1). In the congruent condition (present in both match and mismatch displays), both images were congruent, i.e., the labels for each image in Spanish and German were of the same gender. In the non-congruent condition (also present in both match and mismatch conditions), the labels for both images in the display correspond to different genders in Spanish and German. In the target-congruent condition – which was restricted to match conditions (as in Morales et al., 2016) – the target item was congruent, while the distractor was non-congruent. Of the 102 total displays, 42 were congruent, 42 were non-congruent, and 21 were target-congruent.
Experimental conditions
Citation: Heritage Language Journal 21, 1 (2024) ; 10.1163/15507076-bja10028
Each visual display was paired with an auditory prompt of the form in (1).
Citation: Heritage Language Journal 21, 1 (2024) ; 10.1163/15507076-bja10028
Prompts were first recorded by a male speaker of Spanish born in Colombia who immigrated to the US within one year prior to the recording. From these recordings, a single token of the carrier phrase (Dónde está ‘where is’; length: 1150 ms), a single token of the article el (length: 280 ms), and a single token of the article la (length: 280 ms) were selected and spliced with tokens of each lexical item (average length: 575ms) to prevent any effects of coarticulation that could provide unintended cues to the target item. Each trial began with 800ms of silent looking time, a high-pitched sound that prompted participants to gaze at the fixation cross, then an additional 200ms of silence before the start of the auditory prompt.
3.2 Participants
A total of 54 Spanish-German bilinguals participated in the study: 21 German- dominant adult HSs of Spanish and 33 control adult speakers of Spanish who were late-L2 German learners. All participants were living in Heidelberg or Frankfurt, Germany at the time of data collection. Data collection efforts occurred at two different time periods. Participants received monetary compensation for their time. All participants completed an abbreviated version of the Language Experience and Proficiency Questionnaire (LEAP-Q) (Kaushanskaya et al., 2020; Marian et al., 2007), provided to them either in Spanish or German, based on their stated preference; both language versions of the questionnaire are available in the Supplementary Files. HSs were identified as those who self-reported Spanish to be their first language and were either born in Germany (N=13) or in a Spanish-speaking country but immigrated to Germany in childhood (N=8). Control speakers of Spanish were identified as those who were born in a Spanish-speaking country and immigrated to Germany after age 18 (Table 2).
Selected demographic information of study participants
Citation: Heritage Language Journal 21, 1 (2024) ; 10.1163/15507076-bja10028
3.3 Procedure
Participants were tested individually in a private office space at the University of Heidelberg or the University of Frankfurt. They first completed the LEAP-Q, followed by the oral picture naming task. This task was self-paced and displayed on a laptop, and participant responses were recorded as correct or incorrect by hand by the experimenter. The oral production task was conducted before the main experimental task, as it was used for data cleaning and therefore it was critical that it occur before exposure to the lexical items in the comprehension task (Fuchs, 2021a, 2022; Lieberman et al., 2018).
During the eye-tracking comprehension task, participants were seated facing a 17-inch screen, with their head not stabilized. They received instructions in written and oral form in Spanish. Participants were instructed that they could explore the display freely until the high-pitched tone, at which point they were to direct their gaze to the fixation cross; subsequently they were to direct their gaze to the item they heard in the auditory prompt as rapidly as possible. They completed three practice trials, after which the eyetracker was calibrated using nine points on the screen. During the first data collection effort (N=25), the eyetracker used was a Tobii Pro X3–120. During the second data collection effort (N=29), an SR Research Eyelink Portable Duo was used. Participants were able to take a break after the first block of trials was completed; calibration was repeated before the second block.
4 Analysis & Results
4.1 Oral Production Task
Participants’ performance on the oral production task was taken as a proxy measure of lexical proficiency (Fuchs, 2021a, 2022; Godson, 2003; Polinsky, 2006). The response to each image was checked for (a) the appropriateness of the lexical item as a label for the image (dialectal variants3 of the intended lexical items were considered appropriate at this stage, cf. Section 4.2) and (b) gender agreement on the article. Results of the oral production task are shown in Fig. 2. The control group performed at ceiling (mean accuracy = 0.98, SD=0.02) – as expected for Spanish-dominant participants – while the HSs were much more variable in their performance and, overall, significantly less accurate (mean accuracy = 0.86, SD=0.15) (Wilcoxon Rank-Sum test, p < 0.001).
Accuracy on oral production task, averaged across 42 trials, grouped by participant group. Dots represent outliers. On average, the heritage group knew fewer of the target items than did the control group.
Citation: Heritage Language Journal 21, 1 (2024) ; 10.1163/15507076-bja10028
4.2 Coding and Data Cleaning
All data were cleaned and analyzed with R version 4.2.2 (R Core Team, 2021). To correct for gender assignment, for each participant, trials were excluded from analysis if the participant did not produce the correct label or produced incorrect gender agreement on the article in the oral production task for the target, distractor, or both. At this stage, dialectal variants also resulted in exclusion of the trial from analysis, as for the purpose of measuring speed of lexical retrieval it was important that the participant associated the image with the lexical item that was intended in the experimental design. This procedure removed on average 21% of the trials for the control group (min: 6%, max: 47%) and 39% of the data for the heritage group (min: 10%, max: 87%).
4.3 Eye-tracking Results
Data was prepared using the package eyetrackingR (Dink & Ferguson, 2015) and analyzed using the package clusterperm (Barr, 2022). To analyze the data, permutation-based cluster analyses – non-parametric tests that can track differences in the proportion of fixations between conditions that arise in an interest period – were implemented to identify effects of condition and congruency on the proportion of fixations to the target item over the course of the trial (Maris & Oosternveld, 2007; Ito & Knoeferle, 2022). Data was binned into time bins of 50ms. Separate analyses were conducted for the heritage and control groups.
4.3.1 Control Group
The timecourse of proportion of fixations to the target item in match vs. mismatch conditions for the control group is plotted in Fig. 3; visual inspection of the plot suggests that the proportion of fixations to the target item was greater in mismatch trials than in match trials. This is confirmed by the cluster-based permutation analysis, in which generalized mixed effects models were fitted to the data in each time bin, predicting the empirical logit of the proportion of fixations to the target item in the given bin by condition (sum-coded: match +1, mismatch –1), with by-participant and by-item random intercepts. The data was permuted 1000 times. The analysis found significant clusters for the effect of condition for the period of 3200ms to 4300ms (cluster mass statistic = 1767.0, p < 0.001) and the period of 4700 to 4950ms after the start of the trial (cluster mass statistic = 60.8, p < 0.001). This suggests targets were inspected more when the gender of the target in Spanish was different than the gender of the distractor.
Effect of condition for control group. The analysis found significant clusters for the effect of condition for the control group.
Citation: Heritage Language Journal 21, 1 (2024) ; 10.1163/15507076-bja10028
A similar analysis was conducted to test for the effect of congruency, with separate analyses for match vs. mismatch trials, given that congruency was a three-level variable in match trials (Fig. 4a) but a two-level variable in mismatch trials (Fig. 4b). In match trials, congruency was Helmert-coded to first test for the difference between the congruent condition and the two conditions with non-congruency (non-congruent and target-congruent taken together), and then to test for the difference between the non-congruent and target-congruent conditions. In mismatch trials, congruency was sum-coded (congruent +1, non-congruent –1). For the match trials, the analysis found significant clusters for an effect of congruency between congruent condition trials and the two conditions with one or both non-congruent items at 2000ms to 3600ms from the start of the trial (cluster mass statistic = 585.0, p < 0.001), 3850ms to 4000ms (cluster mass statistic = 25.0, p < 0.001), and 4150ms to 4950ms (cluster mass statistic = 417.0, p < 0.001). The analysis also found a significant cluster for the effect of congruency between the non-congruent and target-congruent conditions for the period of 4600ms to 4700ms (cluster mass statistic = 15.0, p = 0.02). For the mismatch condition trials, significant clusters for the effect of congruency were identified for the periods of 2900ms to 3050ms (cluster mass statistic = 17.8, p = 0.004) and 4500ms to 4950ms (cluster mass statistic = 88.3, p < 0.001). These findings indicate that the proportion of fixations to the target item was modulated by the gender congruency of the items in the display in both match and mismatch conditions.
Congruency effects for control group in match condition trials. The analysis found significant clusters for the contrast between non-congruent and target-congruent conditions (shaded in pink), as well as between these two conditions taken together as compared to the congruent condition trials (shaded in gray).
Citation: Heritage Language Journal 21, 1 (2024) ; 10.1163/15507076-bja10028
Congruency effects for control group in mismatch condition trials. The analysis found significant clusters for the contrast between congruent and non-congruent condition trials (shaded in gray).
Citation: Heritage Language Journal 21, 1 (2024) ; 10.1163/15507076-bja10028
4.3.2 Heritage Group
Visual inspection of the proportion of fixations in Fig. 5 suggests that the proportion of fixations to the target item was greater in mismatch trials than in match trials. This is confirmed by the cluster-based permutation analysis, conducted as in Section 4.3.1, which found significant clusters for the effect of condition for the periods of 3200ms to 3950ms after the start of the trial (cluster mass statistic = 743.0, p < 0.001) and 4650ms to 4750ms (cluster mass statistic = 16.4, p = 0.013). This suggests targets were inspected more by heritage speakers when the gender of the target in Spanish was different than the gender of the distractor.
Effect of condition for heritage group. The analysis found significant clusters for the effect of condition for the heritage group.
Citation: Heritage Language Journal 21, 1 (2024) ; 10.1163/15507076-bja10028
The congruency analysis also proceeded as described in Section 4.3.1 (Figs. 6a & 6b). In match condition trials, the analysis found significant clusters for the effect of congruency between the congruent conditions and the two conditions with one or both non-congruent items for the period of 2000ms to 3600ms (cluster mass statistic = 585.0, p < 0.001), 3850ms to 4000ms (cluster mass statistic = 25.0, p < 0.001), and 4150ms to 4950ms (cluster mass statistic = 417.0, p < 0.001). The analysis also identified a significant cluster for the effect of congruency between the non-congruent and target-congruent condition trials for the period of 4850ms to 4950ms (cluster mass statistic = 19.2, p = 0.003). In mismatch condition trials, the analysis found significant clusters for the effects of congruency (sum-coded: congruent +1, non-congruent –1) for the periods of 3000ms to 3150ms (cluster mass statistic = 33.7, p < 0.001) and 4400ms to 4550ms (cluster mass statistic = 27.4, p < 0.001). These findings indicate that the proportion of fixations to the target item by the heritage group was modulated by the gender congruency of the items in the display in both match and mismatch conditions.
Congruency effects for heritage group in match condition trials. The analysis found significant clusters for the contrast between non-congruent and target-congruent conditions (shaded in pink), as well as between these two conditions taken together as compared to the congruent condition trials (shaded in gray).
Citation: Heritage Language Journal 21, 1 (2024) ; 10.1163/15507076-bja10028
Congruency effects for heritage group in mismatch condition trials. The analysis found significant clusters for the contrast between congruent and non-congruent condition trials (shaded in gray).
Citation: Heritage Language Journal 21, 1 (2024) ; 10.1163/15507076-bja10028
4.4 Comparison Between the Spanish-German and Spanish-English Groups
To pursue Research Question 2, the results of the current study were further compared to the results of Fuchs (2021a), which tested Heritage Spanish speakers in the US and Spanish-dominant control speakers in the US. In line with the analysis presented on the HSs in the US and other previous work on facilitative use of grammatical gender, the dependent variable in this part of the analysis was time of first fixation on the target item (Fuchs, 2021a, 2022; Grüter et al., 2012; Lew-Williams & Fernald, 2007, 2010). This is measured as the time of first recorded fixation on the target item after the onset of the gender-marked article in the auditory stimulus and is considered to be a measure of response time; accordingly, faster times of first fixation indicate faster word recognition. The time of first fixation in each trial was calculated by subtracting 2350ms (the time of the onset of the determiner in each auditory stimulus from the start of the trial plus 200ms of time required to launch a saccade) from the raw time of first fixation. To control for possible systematic bias in the times of first fixation resulting from any implementational differences in the software and/or in the equipment between the two data collection sessions using different eyetrackers for the Spanish-German data (cf. Section 3.3), the times of first fixation of each session were standardized, and the standardized first fixation times were used in the statistical analysis.
The comparison of the Spanish-English and Spanish-German results found a significant interaction between dataset, group, and condition.
Citation: Heritage Language Journal 21, 1 (2024) ; 10.1163/15507076-bja10028
The data cleaning procedures were replicated across the two studies. Outliers outside of two standard deviations from the mean of the first fixation times for the Spanish-German data were excluded from analysis, removing 6.62% of the data for the control group and 5.41% of the data for the heritage group. In total, there were 1241 data points for the Spanish-German heritage group (N = 21) and 2481 data points for the control group (N = 33) included in the analysis. In total, the dataset from Fuchs (2021a) contained 1172 data points for the heritage group (N = 20) and 824 data points for the control group (N = 10). Because the dependent variable (time of first fixation) was standardized for the analysis in the current study, we also standardized the first fixation time in the Spanish-English dataset for direct comparison. The dataset from the current study was coded as Spn-Ger, while the dataset from Fuchs (2021a) was coded as Spn-Eng.
A linear mixed-effects regression model was fit to the data with the R package lmerTest (Kuznetsova et al., 2017), predicting standardized time of first fixation by group (heritage +0.5, control –0.5), condition (match +0.5, mismatch –0.5), dataset (Spn-Eng +0.5, Spn-Ger –0.5), and trial number (centered), as well as their interactions (pairwise, three-way, and four-way). Random effects were grouped by participant and item, including random intercepts and random slopes for all terms in the fixed effects. The method of backward selection was used to select the model that best explained the data (Zuur et al., 2009). The most complex model with full sets of fixed and random effects was fitted. Then, random effects that overfit were removed, and interaction terms in the fixed effects that did not improve the model fit in a likelihood ratio test were discarded.
We report only significant effects that involved the variables of interest to the theoretical questions asked here, but the full report of fixed effects – including control variables – in the selected model is provided in Table 3. The selected model found a main effect of dataset (
Fixed effects of the LMM fitted to compare the Spn-Ger and Spn-Eng results
Citation: Heritage Language Journal 21, 1 (2024) ; 10.1163/15507076-bja10028
More critically, the selected model found significant interactions between group and dataset (
4.4.1 Heritage Groups
A linear mixed-effects model was fitted to the data of both heritage groups, predicting standardized first fixation times by condition, dataset, trial number, as well as their interactions (pairwise and three-way), with by-participant random intercepts and slopes for condition and by-item random intercepts. The model found a significant effect of condition (
Fixed effects of the model fitted to the heritage groups of the two datasets
Citation: Heritage Language Journal 21, 1 (2024) ; 10.1163/15507076-bja10028
The main effect of condition suggests that both heritage groups were faster in the mismatch condition (M = –0.08, SE = 0.03) than in the match condition (M = 0.26, SE = 0.02). The interaction between condition and trial number suggests that the HSs’ overall speed changed at different rates across the trials between the two conditions. Follow-up models found a negative effect of trial number on first fixation times in the mismatch condition (
Critically, the linear mixed-effects model did not find a main effect of dataset or any interaction effects involving it,4 suggesting the HSs in Germany and in the US did not differ significantly in their behavioral patterns.
4.4.2 Spanish-Dominant Control Groups
A linear mixed-effects model was fitted to the data of Spanish-dominant control groups from Germany and from the US, predicting standardized first fixation times by condition, dataset (Spn-Ger, Spn-Eng), trial number, as well as their interactions (pairwise and three-way), with by-participant random intercepts and slopes for condition and by-item random intercepts. The model found significant main effects of condition (
5 Results & Implications
5.1 Facilitative Processing of Gender When the Majority Language Has a Gender System
Research Question 1 (Section 2.3) asked whether HSs can use grammatical gender in their HL to facilitate lexical retrieval, specifically when their dominant language also has a grammatical gender system. The results from HSs of Spanish whose dominant language is German presented in Section 4.3 indicate that participants fixated more on target items in mismatch conditions, similar to the control group of Spanish-dominant late-L2 learners of German. Thus, we conclude that, when gender assignment is controlled for, HSs of Spanish who have a gender system in their dominant language are able to access and deploy grammatical gender agreement features in their heritage language in a target-like manner. This is consistent with previous results for HSs whose dominant language (English) does not have grammatical gender (Sekerina, 2015; Fuchs, 2021a, 2022; but contra the results from eyetracking-in-reading in Parshina et al., 2022).
Independent evidence discussed in Section 2.2 suggests that bilinguals with two gender systems have an integrated lexicon with shared abstract gender information (Salamoura & Williams, 2007; see Sá-Leite et al., 2023 for work specifically on HSs), and so co-activated gender information from the two languages may lead to interference during language processing. We found effects of gender congruency both in match and mismatch conditions for the heritage group, suggesting that gender congruency between Spanish and German modulated their lexical retrieval, including during facilitative processing. This is consistent with results from Sá-Leite et al. (2023) and counter results from Di Pisa et al. (2022); it also contributes evidence that congruency effects can be observed for heritage bilingual speakers during spoken-language comprehension. The analysis also identified effects of congruency for the control group. It should be noted that, given that these were late-L2 learners of German, this constitutes a backward congruency effect, wherein the gender of the speakers’ weaker L2 modulates processing of gender in the stronger L1. Backward congruency effects have been observed more rarely than forward congruency effects in the literature on gender congruency and only for highly proficient L2 speakers (ex. Basnight-Brown & Altarriba, 2007; Duñabeitia et al., 2010). A limitation of our finding is that we did not collect measures of participants’ proficiency in German and therefore cannot comment on the level of L2 German proficiency that allows for backward congruency effects.
The combination of observed effects of condition and congruency for the heritage group suggests that, though heritage speakers are like other bilinguals in terms of having an integrated mental lexicon, connections in the mental lexicon between gender nodes and lemmas in the heritage language are nevertheless robust enough to allow for efficient integration of gender features on agreement marking present in the speech stream, resulting in the facilitation of lexical retrieval of the upcoming noun.
These findings thus add support to previous findings that HSs’ facilitative processing of gender features during spoken language comprehension is target-like when their knowledge of gender assignment is controlled for, contributing evidence that this is the case even in the environment of a majority language with competing gender information. The broader literature on HSs’ target-like processing of gender agreement (Fuchs, 2021a, 2021b; Grüter et al., 2012; Montrul et al., 2014) notes also that with respect to facilitative processing of grammatical gender, HSs align with adult and child monolingual speakers of languages with grammatical gender (e.g., Lew-Williams & Fernald, 2007; Melançon & Shi, 2015; Bosch & Foppolo, 2023). This contrasts with L2 learners, for whom the findings on facilitative processing of grammatical gender have been mixed (cf. Section 2.3). Researchers suggest that the critical contrast between these groups may be early and naturalistic acquisition of the language. This is supported by evidence that children, who acquire gender from the speech stream, initially analyze determiner-noun sequences as monomorphemic – as evidenced by their production of proto-determiners in languages like Spanish (López-Ornat, 1997; Ticio Quesada, 2018) – and only subsequently segment them into an article and noun. It has been argued that this acquisition trajectory, which is shared by heritage and monolingual speakers but not (typical) L2 learners, may give these groups an advantage in developing a target-like ability to efficiently integrate gender agreement information during language comprehension.
5.2 Presence vs. Absence of Gender in the Majority Language: Group Effects
To better understand how absence or presence of gender in the majority language may impact facilitative processing of gender in the heritage language (Research Question 2), we leveraged the fact that the present study was a direct replication of the study in Fuchs (2021a) and directly compared the results from the two studies – one conducted in the US, where the dominant language does not have gender, and the other conducted in Germany, where the dominant language does have gender. The results, presented in Section 4.4, found an interaction between dataset and group, pointing to a difference in the group effect between the two studies. However, follow-up models revealed that this effect was driven not by a difference between the two samples of HSs but rather a difference between the two samples of Spanish-dominant control speakers. We discuss each of these findings in turn.
First, the results did not show a difference between the HSs whose majority language was English and those whose majority language was German. Both groups showed facilitative processing of grammatical gender, and the analysis in Section 4.4 did not find evidence for a significant difference between the two groups. Thus, we do not find evidence that competing gender information from the dominant language constitutes a disadvantage in facilitative processing, relative to a heritage population whose mental lexicon contains only gender information from the heritage language. However, gender in the majority language also does not appear to offer an advantage (cf. Section 2.2), even when gender marking occurs in a syntactically congruent environment (Foucart & Frenck-Mestre, 2011; Hopp & Lemmerth, 2016). Research on German monolingual speakers as well as highly proficient L2 learners of German suggest that these populations use gender information on prenominal articles to facilitate lexical retrieval during spoken-language comprehension (ex. Hopp, 2013; Hopp & Lemmerth, 2016). One can reasonably assume that the German-dominant HSs would pattern with these populations in facilitative processing of grammatical gender on German articles. Given the availability of this processing strategy in their majority language, HSs of Spanish in Germany may have been expected to show more efficient anticipatory processing of gender agreement on prenominal articles in their heritage language as well. We found no evidence for this to be the case, but we interpret these results with caution, as the lack of an interaction effect between (heritage) group and dominant language may be the result of a lack of power. We hope future work in this domain with larger sample sizes will be able to further explore this matter.
The second finding – the difference between the Spanish-dominant control groups – is incidental and secondary to the research questions posed by the present paper. We therefore keep the discussion of this finding speculative and brief. We suggest that there may be differences in the level of exposure these groups have to Spanish in their daily lives (in a metropolitan US city vs. a college town in Germany). This has been known to affect prediction (Ito & Sakai, 2021) and lexical retrieval in an L1, possibly due to the degree to which immersion in the L2 requires daily suppression of the L1 (Wodniecka et al., 2020). It may also be that the two groups have differing rates of first-language attrition (Schmid & Köpke, 2017). Given that we did not collect information about participants’ communities or daily Spanish use, we leave open the question of what the cause of the observed difference between the control groups may be, but we note that the results highlight that, in comparisons of heritage groups with different majority languages, the relevant baseline groups cannot be assumed to be equivalent to each other in their linguistic abilities.
6 Conclusion
The goal of the present study was to investigate the processing of Spanish grammatical gender by HSs of Spanish whose dominant language is German; both Spanish and German have grammatical gender. We tested these HSs’ facilitative processing of gender agreement on prenominal articles during language comprehension in an eye-tracking study using the Visual World Paradigm. Results showed effects of gender congruency that suggest that the HSs’ (and Spanish-dominant controls’) gender systems indeed interact during spoken-language comprehension, consistent with findings from HSs’ language production (Sá-Leite et al., 2023). Still, when gender assignment was controlled for, HSs were able to use gender to facilitate lexical retrieval, fixating more on targets in mismatch than match conditions in a target-like manner. Thus, whereas previous work has found that HSs are able to use gender to facilitate lexical retrieval in a target-like manner when the majority language does not have grammatical gender (Sekerina, 2015; Fuchs, 2021a, 2022), the present study demonstrates that HSs’ facilitative processing of gender agreement remains target-like in the context of competing gender information in the majority language. Considered within the bigger picture of work on anticipatory processing by heritage speakers, this provides supporting evidence that HSs can efficiently make use of abstract syntactic features to pre-activate certain aspects of the linguistic representation during real-time spoken-language comprehension, regardless of the presence or absence of grammatical gender in the dominant language.
Supplementary Materials
Supplementary materials are publicly available online on OSF at https://osf.io/f3knc/?view_only=5c4c0e5dfa97435db2cab9f6e45e4683. These materials include data and code used for analysis, as well as tables and figures showing the output of any post hoc analyses mentioned in the main text.
Acknowledgments
For very helpful feedback and discussion, we are grateful to Maria Polinsky, Holger Hopp, Kevin Ryan, and Jonathan Bobaljik, as well as to the audiences of Heritage Languages Around the World and the 14th International Symposium on Bilingualism. For support during data collection, we are indebted to Francisco Moreno, Esther Rinke, and Héctor Álvarez Mella. All errors are our own.
Notes on Contributors
Zuzanna Fuchs is an Assistant Professor in the Department of Linguistics at the University of Southern California. She is also the director of the Psycholinguistics of Mono- and Multilingualism Lab (PoMMLab) at USC.
Wenqi Zeng is a Ph.D. candidate in Linguistics at the University of Iowa. Her research focuses on between- and within-community differences in the language processing of multilingual speakers.
References
Alarcón, I. (2011). Spanish gender agreement under complete and incomplete acquisition: Early and late bilinguals’ linguistic behavior within the noun phrase. Bilingualism: Language and Cognition, 14(3), 332–350. https://doi.org/10.1017/S1366728910000222.
Barr, D. (2022). clusterperm: Cluster-permutation tests for time-series data from factorial experiments. Retrieved from: https://rdrr.io/github/dalejbarr/clusterperm/.
Basnight-Brown, D. M., & Altarriba, J. (2007). Differences in semantic and translation priming across languages: The role of language direction and language dominance. Memory and Cognition, 35(5), 953–965. https://doi.org/10.3758/BF03193468.
Bayram, F., Rothman, J., Pisa, G., & Slabakova, R. (2020). Current Trends and Emerging Methodologies in Charting Heritage Language Bilingual Grammars. https://doi.org/10.31234/osf.io/pa83g.
Beatty-Martínez, A. L., Bruni, M. R., Bajo, M. T., & Dussias, P. E. (2020). Brain potentials reveal differential processing of masculine and feminine grammatical gender in native Spanish speakers. Psychophysiology, 58(3), 1–19. https://doi.org/10.1111/psyp.13737.
Boers, I., Sterken, B., van Osch, B., Parafita Couto, M. C., Grijzenhout, J., & Tat, D. (2020). Gender in unilingual and mixed speech of Spanish heritage speakers in The Netherlands. Languages, 5(4), 1–35. https://doi.org/10.3390/languages5040068.
Bosch, J., & Foppolo, F. (2022). Predictive processing of grammatical gender in bilingual children: The effect of cross-linguistic incongruency and language dominance. Lingue e Linguaggio, 41(1), 5–27.
Bosch, J. E., & Foppolo, F. (2023). Prediction during spoken language processing in monolingual and multilingual children. Linguistic Approaches to Bilingualism. https://doi.org/10.1075/lab.22099.bos.
Brehmer, B., & Rothweiler, M. (2012). The acquisition of gender agreement marking in Polish. In K. Braunmuller & C. Gabriel (Eds.), Multilingual individuals and multilingual societies (pp. 81–100). Amsterdam: John Benjamins.
Cornips, L., & Hulk, A. (2008). Factors of success and failure in the acquisition of grammatical gender in Dutch. Second Language Research, 24(3), 267–295. https://doi.org/10.1177/0267658308090182.
Costa, A., Kovacic, D., Fedorenko, E., & Caramazza, A. (2003). The gender congruency effect and the selection of freestanding and bound morphemes: Evidence from Croatian. Journal of Experimental Psychology: Learning Memory and Cognition, 29(6), 1270–1282. https://doi.org/10.1037/0278-7393.29.6.1270.
Dink, J. W., & Ferguson, B. (2015). eyetrackingR: An R library for eye-tracking data analysis. Retrieved from http://www.eyetrackingr.com.
Di Pisa, G., Kubota, M., Rothman, J., & Marinis, T. (2022). Effects of markedness in gender processing in Italian as a heritage language: A speed accuracy tradeoff. Frontiers in Psychology, 13. https://doi.org/10.3389/fpsyg.2022.965885.
Duñabeitia, J. A., Perea, M., & Carreiras, M. (2010). Masked translation priming effects with highly proficient simultaneous bilinguals. Experimental Psychology, 57(2), 98–107. https://doi.org/10.1027/1618-3169/a000013.
Dussias, P., Valdés Kroff, J. R., Guzzardo Tamargo, R. E., & Gerfen, C. (2013). When gender and looking go hand in hand. Studies in Second Language Acquisition, 35(2), 353–387. https://doi.org/10.1017/S0272263112000915.
Egger, E., Hulk, A., & Tsimpli, I. M. (2018). Crosslinguistic influence in the discovery of gender: The case of Greek-Dutch bilingual children. Bilingualism, 21(4), 694–709. https://doi.org/10.1017/S1366728917000207.
Eichler, N., Jansen, v., & Müller, N. (2013). Gender acquisition in bilingual children: French – German, Italian – German, Spanish – German and Italian – French. International Journal of Bilingualism, 17(5), 550–572. https://doi.org/10.1177/1367006911435719.
Foucart, A., & Frenck-Mestre, C. (2011). Grammatical gender processing in L2: Electrophysiological evidence of the effect of L1–L2 syntactic similarity. Bilingualism, 14(3), 379–399. https://doi.org/10.1017/S136672891000012X.
Fuchs, Z. (2021a). Facilitative use of grammatical gender in heritage Spanish. Linguistic Approaches to Bilingualism, 12(6), 845–871. https://doi.org/10.1075/lab.20024.fuc.
Fuchs, Z. (2021b). Heritage Spanish in the US: How heritage languages can contribute to disentangling factors driving language development. Estudios Del Observatorio / Observatorio Studies, 2949. https://doi.org/10.15427/or070-04/2021en.
Fuchs, Z. (2022). Eyetracking evidence for heritage speakers’ access to abstract syntactic agreement features in real-time processing. In F. Bayram, M. Kubota, & S. Soares (Eds.), The next phase in heritage language studies: Methodological considerations and advancements [Special Issue]. Frontiers of Psychology, 13. https://doi.org/10.3389/fpsyg.2022.960376.
Fuchs, Z., Polinsky, M., & Scontras, G. (2015). The differential representation of number and gender in Spanish. Linguistic Review, 32(4), 703–737. https://doi.org/10.1515/tlr-2015-0008.
Godson, L. (2003). Phonetics of language attrition: Vowel production and articulatory setting in the speech of Western Armenian heritage speakers. [Unpublished doctoral dissertation]. University of California, San Diego.
Grüter, T., Lew-Williams, C., & Fernald, A. (2012). Grammatical gender in L2: A production or a real-time processing problem? Second Language Research, 28(2), 191–215. https://doi.org/10.1177/0267658312437990.
Harris, J. W. (1991). The exponence of gender in Spanish. Linguistic Inquiry, 22(1), 27–62.
Hopp, H. (2013). Grammatical gender in adult L2 acquisition: Relations between lexical and syntactic variability. Second Language Research, 29(1), 33–56. https://doi.org/10.1177/0267658312461803.
Hopp, H., & Lemmerth, N. (2016). Lexical and syntactic congruency in L2 predictive gender processing. Studies in Second Language Acquisition, 40(1), 171–199.
Ito, A., & Knoeferle, P. (2022). Analysing data from the psycholinguistic visual-world paradigm: Comparison of different analysis methods. Behavior Research Methods. https://doi.org/10.3758/s13428-022-01969-3.
Ito, A., Nguyen, H. T. T., & Knoeferle, P. (2023). German-dominant Vietnamese heritage speakers use semantic constraints of German for anticipation during comprehension in Vietnamese. Bilingualism. https://doi.org/10.1017/S136672892300041X.
Ito, A., & Sakai, H. (2021). Everyday language exposure shapes prediction of specific words in listening comprehension: A Visual World eye-tracking study. Frontiers in Psychology, 12. https://doi.org/10.3389/fpsyg.2021.607474.
Janssen, B. (2016). The acquisition of gender and case in Polish and Russian: A study of monolingual and bilingual children. Uitgeverij Pegasus.
Jegerski, J., & Sekerina, I. A. (2020). The processing of input with differential object marking by heritage Spanish speakers. Bilingualism, 23(2), 274–282. https://doi.org/10.1017/S1366728919000087.
Kaan, E., & Grüter, T. (Eds.). (2021). Prediction in Second Language Processing and Learning. John Benjamins.
Kaltsa, M., Tsimpli, I. M., & Argyri, F. (2019). The development of gender assignment and agreement in English-Greek and German-Greek bilingual children. Linguistic Approaches to Bilingualism, 9(2), 253–288. https://doi.org/10.1075/lab.16033.kal.
Karaca, F., Brouwer, S., Unsworth, S., & Huettig, F. (2023). Morphosyntactic predictive processing in adult heritage speakers: Effects of cue availability and spoken and written language experience. Language, Cognition and Neuroscience, 1–18. https://doi.org/10.1080/23273798.2023.2254424.
Kaushanskaya, M., Blumenfeld, H. K., & Marian, V. (2020). The Language Experience and Proficiency Questionnaire (LEAP-Q): Ten years later. Bilingualism, 23(5), 945–950. https://doi.org/10.1017/S1366728919000038.
Klassen, R. (2016). The representation of asymmetric grammatical gender systems in the bilingual mental lexicon. Probus, 28(1), 9–28. https://doi.org/10.1515/probus-2016-0002.
Kuchenbrandt, I. (2005). Gender acquisition in bilingual Spanish. In J. Cohen, K. T. McAlister, K. Rolstad, & J. MacSwan (Eds.), Proceedings of the 4th International Symposium on Bilingualism (pp. 1252–1263). Cascadilla.
Kuznetsova, A., Brockhoff, P. B., & Christensen, R. H. B. (2017). lmerTest Package: Tests in Linear Mixed Effects Models. Journal of Statistical Software, 82(13), 1–26. https://doi.org/10.18637/jss.v082.i13.
Lemhöfer, K., Spalek, K., & Schriefers, H. (2008). Cross-language effects of grammatical gender in bilingual word recognition and production. Journal of Memory and Language, 59(3), 312–330. https://doi.org/10.1016/j.jml.2008.06.005.
Lew-Williams, C., & Fernald, A. (2007). Young children learning Spanish make rapid use of grammatical gender in spoken word recognition. Psychological Science, 18(3), 193–198.
Lew-Williams, C., & Fernald, A. (2010). Real-time processing of gender-marked articles by native and non-native Spanish speakers. Journal of Memory and Language, 63(4), 447–464. https://doi.org/10.1016/j.jml.2010.07.003.
Lieberman, A., Borovsky, A., & Mayberry, R. (2018). Prediction in a visual language: Real-time sentence processing in American Sign Language across development. Language, Cognition and Neuroscience, 33(4), 387–401. https://doi.org/10.1080/23273798.2017.1411961.
López-Ornat, S. (1997). What lies in between a pre-grammatical and a grammatical representation? Evidence on nominal and verbal form-function mappings in Spanish from 1;7 to 2;1. In A. T. Pérez-Leroux & W. R. Glass (Eds.), Contemporary perspectives on the acquisition of Spanish, Vol. 1. Cascadilla, 3–20.
Luque, A., Rossi, E., Kubota, M., Nakamura, M., Rosales, C., López-Rojas, C., Rodina, Y., & Rothman, J. (2023). Morphological transparency and markedness matter in heritage speaker gender processing: An EEG study. Frontiers in Psychology, 14, 1114464. https://doi.org/10.3389/fpsyg.2023.1114464.
Marian, V., Blumenfeld, H. K., & Kaushanskaya, M. (2007). The Language Experience and Proficiency Questionnaire (LEAP-Q): Assessing language profiles in bilinguals and multilinguals. Journal of Speech Language and Hearing Research, 50(4), 940.
Maris, E., & Oostenveld, R. (2007). Nonparametric statistical testing of EEG- and MEG-data. Journal of Neuroscience Methods, 164(1), 177–190. https://doi.org/10.1016/j.jneumeth.2007.03.024.
McClelland, J. L., & Rumelhart, D. E. (1981). An interactive activation model of context effects in letter perception: I. An account of basic findings. Psychological Review, 88(5), 375–407. https://doi.org/10.1037/0033-295X.88.5.375.
Meir, N., Parshina, O., & Sekerina, I. A. (2020). The interaction of morphological cues in bilingual sentence processing: An eye-tracking study. In M. M. Brown & A. Kohut (Eds.), Proceedings of the 44th Annual Boston University Conference on Language Development (pp. 376–389). Cascadilla.
Meir, N., Walters, J., & Armon-Lotem, S. (2017). Bi-directional cross-linguistic influence in bilingual Russian-Hebrew children. Linguistic Approaches to Bilingualism, 7(5), 514–553. https://doi.org/10.1075/lab.15007.mei.
Melançon, A., & Shi, R. (2015). Representations of abstract grammatical feature agreement in young children. Journal of Child Language 42, 1379–1393. Doi: 10.1017/S0305000914000804.
Montrul, S., Davidson, J., de la Fuente, I., & Foote, R. (2014). Early language experience facilitates the processing of gender agreement in Spanish heritage speakers. Bilingualism: Language and Cognition, 17(1), 118–138. https://doi.org/10.1017/S1366728913000114.
Montrul, S., Foote, R., & Perpiñán, S. (2008). Gender agreement in adult second language learners and Spanish heritage speakers: The effects of age and context of acquisition. Language Learning, 58(3), 503–553. https://doi.org/10.1111/j.1467-9922.2008.00449.x.
Morales, L., Paolieri, D., Dussias, P., Valdés Kroff, J. R., Gerfen, C., & Bajo, M. T. (2016). The gender congruency effect during bilingual spoken-word recognition. Bilingualism, 19(2), 294–310. https://doi.org/10.1017/S1366728915000176.
Müller, N. 1990. Developing two gender assignment systems simultaneously. In J. M. Meisel (Ed.), Two first languages: Early grammatical development in bilingual children (pp. 193–234). Benjamins.
Müller, N. 1999. Gender and number in acquisition. In B. Unterbeck & M. Rissanen (Eds.), Gender in grammar and cognition (pp. 351–399). De Gruyter.
Özsoy, O., Çiçek, B., Özal, Z., Gagarina, N., & Sekerina, I. A. (2023). Turkish-German heritage speakers’ predictive use of case: Webcam-based vs. in-lab eye-tracking. Frontiers in Psychology, 14. https://doi.org/10.3389/fpsyg.2023.1155585.
Paolieri, D., Padilla, F., Koreneva, O., Morales, L., & Macizo, P. (2019). Gender congruency effects in Russian-Spanish and Italian-Spanish bilinguals: The role of language proximity and concreteness of words. Bilingualism, 22(1), 112–129. https://doi.org/10.1017/S1366728917000591.
Parshina, O., Lopukhina, A., & Sekerina, I. A. (2022). Can heritage speakers predict lexical and morphosyntactic information in reading? Languages, 7(1), 60. https://doi.org/10.3390/languages7010060.
Pérez-Pereira, M. (1991). The acquisition of gender: What Spanish children tell us. Journal of Child Language, 18(3), 571–590. https://doi.org/10.1017/S0305000900011259.
Pickering, M. J., & Gambi, C. (2018). Predicting while comprehending language: A theory and review. Psychological Bulletin, 144(10), 1002–1044. https://doi.org/10.1037/bul0000158.
Polinsky, M. (2006). Incomplete acquisition: American Russian. Journal of Slavic Linguistics, 14(2), 191–262.
R Core Team. (2021). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/.
Rodina, Y., Kupisch, T., Meir, N., Mitrofanova, N., Urek, O., & Westergaard, M. (2020). Internal and external factors in heritage language acquisition: Evidence from heritage Russian in Israel, Germany, Norway, Latvia and the United Kingdom. Frontiers in Education, 5(March). https://doi.org/10.3389/feduc.2020.00020.
Rumelhart, D. E., & McClelland, J. L. (1982). An interactive activation model of context effects in letter perception: II. The contextual enhancement effect and some tests and extensions of the model. Psychological Review, 89(1), 60–94. https://doi.org/10.1037/0033-295X.89.1.60.
Sabourin, L., & Stowe, L. (2008). Second language processing: When are first and second languages processed similarly? Second Language Research, 24(3), 397–430. https://doi.org/10.1177/0267658308090186.
Salamoura, A., & Williams, J. (2007). The representation of grammatical gender in the bilingual lexicon: Evidence from Greek and German. Bilingualism, 10(3), 257–275. https://doi.org/10.1017/S1366728907003069.
Sá-Leite, A. R., Flores, C., Eira, C., Haro, J., & Comesaña, M. (2023). Language balance rather than age of acquisition: A study on the cross-linguistic gender congruency effect in Portuguese – German bilinguals. Bilingualism: Language and Cognition, 1–14. doi:10.1017/S1366728923000378.
Sá-Leite, A. R., Luna, K., Fraga, I., & Comesaña, M. (2020). The gender congruency effect across languages in bilinguals: A meta-analysis. Psychonomic Bulletin & Review, 27, 677–693. https://doi.org/10.3758/s13423-019-01702-w.
Schmid, M. S., & Köpke, B. (2017). The relevance of first language attrition to theories of bilingual development. Linguistic Approaches to Bilingualism, 7(6), 637–667. https://doi.org/10.1075/lab.17058.sch.
Scontras, G., Polinsky, M., & Fuchs, Z. (2018). In support of representational economy: Agreement in heritage Spanish. Glossa: A Journal of General Linguistics, 3(1), 1. https://doi.org/10.5334/gjgl.164.
Sekerina, I. A. (2015). Predictions, fast and slow. Linguistic Approaches to Bilingualism, 5(4), 532–536. https://doi.org/10.1075/lab.5.4.16sek.
Ticio Quesada, E. (2018). The emergence and type of functional categories. In The emergence of nominal expressions in Spanish-English early bilinguals: Economy and bilingual first language acquisition (pp. 75–131). John Benjamins.
van der Linden, E., & Hulk, A. (2009). The vulnerability of gender on determiners in L1, 2L1 and L2 acquisition. Bucharest Working Papers in Linguistics, 11(1), 97–107.
Wodniecka, Z., Szewczyk, J., Kałamała, P., Mandera, P., & Durlik, J. (2020). When a second language hits a native language. What ERPs (do and do not) tell us about language retrieval difficulty in bilingual language production. Neuropsychologia, 141. https://doi.org/10.1016/j.neuropsychologia.2020.107390.
Zuur, A., Ieno, E. N., Walker, N. J., Saveliev, A. A., & Smith, G. M. (2009). Mixed Effects Models and Extensions in Ecology with R (1st ed. 2009.). Springer. https://doi.org/10.1007/978-970-387-87458-6.
While the discussion and findings presented in this study do not necessarily depend on the model of the lexicon assumed, for concreteness the assumptions are here briefly spelled out.
Klassen (2016) addresses the matter of what happens when nouns in one of a bilingual’s languages belong to a gender category that does not exist in the bilingual’s other language. In this case, the relevant category is the German neuter.
For example, some participants labeled the image of the strawberry using (la) frutilla – a dialectal variant of (la) fresa, which was the lexical item assumed in the design of the experimental materials.
A regular linear regression model found a significant interaction between dataset and condition. This reflects a numerical difference in times of first fixation in match condition trials between the two heritage groups, wherein the HSs in Germany had numerically slower times of first fixation in the match condition. However, the effect is not significant in the mixed effects model reported in Table 5.
