Fill in the Blank: English Morphosyntax Production in Matched Bilingual Groups

Abstract

Purpose:

This study compared English grammatical performance of bilingual school-age children who spoke either Spanish or Vietnamese at home, focusing on their first-language influence on the acquisition of 13 English grammatical forms.

Method:

Scores from 30 children on a cloze task were analyzed for accuracy, developmental patterns, and error types. Pairwise matching was used to match individual Vietnamese and Spanish bilinguals on age and English experience. Caregivers rated language ability and daily use of the home language and English.

Results:

Higher grammatical accuracy across groups was associated with older age and higher caregiver ratings of children's language ability. Groups showed similar patterns on some grammatical forms (e.g., higher accuracy on plurals than irregular past), consistent with previously reported developmental patterns of monolingual English-speaking children. Differences in other patterns were related to first-language influence. For example, phonological constraints of Vietnamese (e.g., no word-final /s/), led to lower accuracy in English forms requiring such sounds.

Conclusions:

English grammatical performance in bilinguals can reflect both general developmental patterns and first-language influence. As bilingual children may rely on linguistic cues specific to their first language, classroom instruction can be adjusted to be more culturally and linguistically responsive.

Grammatical morphemes in English are acquired in an expected sequence, as outlined in Brown's (1973) influential study, which identified 14 forms progressing across five stages of development. Among monolingual children, grammatical accuracy increases with experience, developing in stages as grammatical structures are consistently mastered over time (Berko, 1958). In contrast, second-language learners demonstrate greater variability in morpheme acquisition, influenced by factors such as age of acquisition, exposure, and the typological relationship between their first language (L1) and second language (L2; Birdsong, 2018; Bosch et al., 2019; Singleton & Leśniewska, 2024).

Given the significant proportion of bilingual children in the U.S. public schools (National Center for Education Statistics, 2024), understanding how bilinguals acquire English grammar is crucial for developing effective educational practices. Unlike monolinguals, bilinguals may not follow the same acquisition patterns. The Unified Model (UM; MacWhinney, 2018) suggests that bilinguals transfer knowledge from L1 to L2 based on the reliability and storage ease of linguistic cues. Greater similarity between L1 and L2 grammatical structures (cue convergence) and the ability to chunk and recognize patterns enhance the likelihood of accurate L2 production. Cues are the information that learners use to connect form and meaning. Within language, cue reliability (e.g., clarity of mapping and ease of storage) and cue strength (e.g., frequency, perceptual silence) facilitate learning. Cue convergence builds on this principle. When cues align across languages, transfer is facilitated, supporting the assertion that anything transferable will indeed transfer. In contrast, cue divergence creates learning challenges, leading to reduced or incomplete transfer between languages.

This process of transferring L1 knowledge to L2 can result in either positive or negative transfer. Positive transfer occurs when similarities between L1 and L2 facilitate learning, leading to correct language production. Negative transfer happens when differences between the languages lead to errors (Treffers-Daller & Sakel, 2012). The degree of cue convergence or divergence between L1 and L2 significantly influences the likelihood of positive or negative transfer. When cues align closely (cue convergence), positive transfer is more likely, facilitating smoother L2 acquisition. When cues differ markedly (cue divergence), negative transfer is more likely, potentially interfering with the learning process.

The present study examines the acquisition of English grammatical forms in two distinct language pairs: Spanish–English (SE) and Vietnamese–English (VE). While English L2 acquisition generally mirrors the milestones of English L1 acquisition, our investigation uncovers nuanced differences attributed to L1 influence on production. Notably, we explored the acquisition patterns in the context of our target bilingual groups, recognizing the typological distinctiveness of Vietnamese and Spanish.

Monolingual Versus Bilingual Grammatical Acquisition

English grammar acquisition follows a predictable order (Brown, 1973), and this sequence, which progresses with increased mean length of utterance, has been supported by studies emphasizing the role of cue reliability—how consistently grammatical cues such as morphemes (e.g., −ing) predict patterns in sentence structure (MacWhinney, 2018). For example, the present progressive (−ing), with a one-to-one form–function mapping, is acquired early, whereas articles (“a,” “the”) with multiple functions and forms appear later (Brown, 1973). These developmental patterns, observed in monolinguals, serve as a baseline for examining L2 learners.

Unlike monolinguals, bilinguals demonstrate variability in morpheme acquisition due to factors such as typological differences between L1 and L2, exposure, and age of acquisition (Davies et al., 2020; Xu Rattanasone & Demuth, 2023). Prior work emphasizes the need to examine cue convergence and divergence when evaluating bilingual grammar acquisition. For instance, SE bilinguals, whose L1 shares typological similarities with English, show high accuracy on grammatical cues with overlapping structures, such as plural –s and present progressive, but struggle with syntactic cues such as prepositions due to cross-linguistic differences in form–function mapping (e.g., Spanish en corresponds to both “in” and “on” in English; Padilla, 1978). As children use their linguistic cues to extract meaning from Spanish to English, this difference in cue reliability of prepositions can constrain transfer of L1 knowledge to the L2.

Typological differences are even more pronounced for VE bilinguals, as Vietnamese does not use inflectional morphology. Research shows that Vietnamese learners show lower accuracy in producing English tense markers and plural forms (Dam et al., 2020). Vietnamese learners often omit English morphemes such as regular past tense –ed and third-person singular –s, reflecting greater cue divergence between these languages. Similarly, Mandarin–English bilinguals exhibit patterns where cue divergence results in lower accuracy for certain morphemes, such as regular past tense, compared to forms with higher overlap between L1 and L2 structures (Xu Rattanasone & Demuth, 2023).

Exposure also plays a critical role in bilingual language acquisition. Studies show that greater exposure to English correlates with higher morpheme accuracy, irrespective of L1 typology (Bedore et al., 2018; Xu Rattanasone & Demuth, 2023). For instance, Mandarin–English bilingual preschoolers with increased exposure to English in early child care centers demonstrated better acquisition of grammatical cues such as progressive and plural forms compared to those with limited English input, although they continued to lag behind monolingual peers in mastery of past and present tense morphemes (Xu Rattanasone & Kim, 2024). These findings underscore the role of both language input and the typological distance between L1 and English in shaping grammatical development.

Patterns of English exposure also differentially affect grammatical accuracy across different forms. For example, in an elicited production task, children grouped by their English exposure patterns (High English, Balanced, and High Spanish) showed similar performance for forms with the highest and lowest accuracy and variable performance for forms in between. Across all groups, the copula consistently showed the highest accuracy, while irregular past remained the most challenging form (Bedore et al., 2018). Some forms showed wide variability in performance across exposure groups such as third-person singular (89%, 72%, and 54% across exposure groups) and negatives (94%, 78%, and 59% across groups).

Collectively, previous studies suggest that typological similarity and consistent exposure to English are critical factors in facilitating bilingual children's mastery of English grammar. However, differences in English morphosyntactic development among bilinguals can be difficult to compare given the different methodologies used across studies. The current study examines two L1s from different language families that have a degree of overlap with the L2 (SE) or less overlap between languages (VE). This study extends the literature that is largely based on monolingual–bilingual comparisons to a bilingual–bilingual comparison. The use of the same task across two different bilingual groups allows for systematic comparisons of English L2 acquisition. If bilinguals who speak distinct L1s follow the same pattern of L2 morpheme development, this would be another indication of the ubiquity of English morphological development.

Two Distinct First Languages: Vietnamese and Spanish

This study examined bilingual children speaking SE or VE to compare their performance on English grammatical forms at the sentence level. Spanish and Vietnamese were chosen for their prominence and typological differences, as they rank among the top 5 languages spoken by U.S. school-age children learning English as an L2 (Office of English Language Acquisition, 2023). Their distinct linguistic features also provide insight into how L1 and English interact based on cue convergence and divergence.

Spanish is an inflected language, marking grammatical functions such as tense, number, and gender by modifying word forms (e.g., adjectives agree with the noun's gender: zapato amarillo “yellow shoe,” camisa amarilla “yellow shirt”). Masculine nouns in Spanish typically end in –o (e.g., zapato “shoe”), while feminine nouns often end in –a (e.g., camisa “shirt”), and this distinction influences the agreement of associated adjectives.

English and Spanish show a high degree of cue convergence and overlap in several grammatical forms (e.g., the copula “is red”/es rojo; the present progressive “are running”/están corriendo), although divergence appears in less common structures (e.g., possessive marked with ‘s in English but a preposition in Spanish: “mom's purse”/la bolsa de mamá). Phonologically, Spanish and English share final consonants (e.g., plural /s/), but English's final consonant clusters (e.g., “shirts”) are unshared.

Vietnamese, a non-inflected language, conveys grammatical meaning through word order, context, and additional words rather than modifying word forms. For example, plurality is marked with quantifiers such as hai (“two”) in hai căn nhà (“two houses”), while classifiers specify a variety of noun properties including animacy (con, e.g., con gấu “bear”) and inanimacy (cái, e.g., cái ghế “chair”; for review, see Pham & Kohnert, 2008).

Vietnamese phonology lacks consonant clusters and limits word-final sounds to three stops (/p, t, k/) and three nasals (/m, n, ŋ/), unlike English, where grammatical markers such as plural /s/ and possessive forms rely on these features (Phạm & McLeod, 2019). This mismatch highlights significant cue divergence between Vietnamese and English, as Vietnamese uses word order and lexical items rather than bound morphemes to encode grammatical meaning. These linguistic differences can complicate English acquisition for VE bilinguals, particularly for grammatical features involving final consonants or bound morphemes absent in Vietnamese. Table 1 shows the 13 English target forms examined in the current study, 11 of which exist in Spanish and four in Vietnamese.

Table 1.

Grammatical forms and exemplars across languages.

Grammatical form	English	Spanish	Vietnamese
Plurals	kids	niños	n/a
Articles	a dress	un vestido	n/a
Negation	doesn't/does not like cookies	no le gustan las galletas	không thích bánh
Preposition	on the floor	al suelo	trên sàn
Relative clause	that is red	que es rojo	mà màu đỏ
Regular past tense	walked	caminó	n/a
Question inversion	Are you sick?	n/a	n/a
Possessive	teacher's glasses	n/a	n/a
Passive	is played by the boy	es jugado por el niño	được chơi bởi cậu bé
Irregular past tense	fed the fish	le dio de comer el pez	n/a
Copula	are red	son rojos	n/a
Present progressive	are reading	están leyendo	n/a
Third person singular	drinks water	bebe agua	n/a

Study Purpose and Research Questions

The present study examines English L2 grammatical performance in two distinct bilingual groups. We matched individual participants from Spanish- and Vietnamese-speaking backgrounds to ensure balanced comparisons. Pairwise matching provides stricter control over confounding factors by aligning participants on key criteria, allowing for more precise comparisons of language performance (c.f. Czapka et al., 2020). This method improves study power, especially in small sample sizes (Hope, 2015). We address two research questions:

• 1)  Do the bilingual groups perform similarly on English morpheme accuracy?

• 2)  Do the bilingual groups produce similar errors?

Similar performance on English grammatical forms by two bilingual groups with distinct L1s may reflect developmental patterns where certain forms are easier to acquire. For instance, studies have shown that both monolingual and bilingual children tend to acquire simpler and more frequent morphemes such as copula and present progressive earlier than irregular past tense or third-person singular (Davies et al., 2019; Xu Rattanasone & Kim, 2024). This aligns with findings from monolingual data (Brown, 1973) and bilingual studies that emphasize the role of linguistic cues and exposure in shaping acquisition trajectories (Xu Rattanasone & Demuth, 2023).

Conversely, group differences in English grammatical form accuracy may reflect the influence of the distinct L1s, Spanish and Vietnamese. If children's L1 contributes to group differences, two factors might explain the patterns: (a) forms with high cue convergence between L1 and L2 (e.g., SE overlap for plural –s) are easier to acquire and (b) forms requiring new learning due to cue divergence (e.g., tense markers in VE) present greater challenges (Xu Rattanasone & Kim, 2024). As shown in Table 1, Spanish grammar overlaps with 11 of the 13 English forms targeted in the present study, whereas Vietnamese grammar overlaps with only four forms. Thus, we anticipate higher accuracy among the Spanish L1 speakers compared to Vietnamese L1 speakers due to the higher degree of overlap between Spanish and English.

Beyond accuracy, group differences in the types of errors children produce can also shed light on L1 influence. Recent studies have shown that children with isolating L1s such as Mandarin or Vietnamese often omit English inflections due to typological mismatches, whereas speakers of inflected languages such as Spanish exhibit different error patterns, such as overgeneralization (Xu Rattanasone et al., 2024). By analyzing errors, we can identify how bilinguals navigate cues from both L1 and L2. Additionally, understanding the differences in grammatical acquisition between the two groups will help educators tailor interventions that support the unique language needs of bilingual students, promote equitable access to education resources, and improve classroom language outcomes.

Method

Data for this study were obtained from an ongoing multisite research project that seeks to compare the language performance of bilingual children in the United States who speak English as the language of the larger society and either Spanish or Vietnamese at home. Following the policies of the funding agency, all sites relied on the human subjects protection approval from institutional review board at the primary site. Caregivers provided written consent to participate and for their children to participate. Children 7 years and older provided additional written assent to participate in the study. A team of bilingual research assistants, each fluent in English and one additional language (Spanish or Vietnamese), conducted individualized direct measures and caregiver interviews to ensure accuracy.

Recruitment and Study Participation

This study utilized both in-person and virtual methods for participant recruitment. For in-person task administration, participants were recruited from school districts in Southern California, specifically targeting those with a substantial number of children whose home language was reported as Vietnamese or Spanish. Caregivers provided written consent, and children participated in the study in a quiet area of their school setting. For virtual task administration, participants were recruited nationally through social media and advertisements placed in local venues frequented by caregivers. Caregivers expressed interest using an online form to verify their child's age (4–8 years old) and languages (Vietnamese or Spanish at home) and provided written consent electronically. Children participated remotely via Zoom and Qualtrics Survey software. During virtual administration, caregivers were present to help children stay focused and resolve any technical issues, but they did not assist children in answering test prompts. Virtual and in-person procedures were designed to be similar in administration to ensure consistency across recruitment methods (Pratt et al., 2022). Of the total sample (N = 30), 20% completed tasks virtually and 80% in person.

Participant Selection and Matching Criteria

Two caregiver report measures, the Inventory to Assess Language Knowledge (ITALK) and Bilingual Input Output Survey (BIOS), were administered during a parent interview to gather information on each child's language proficiency, exposure, and history of use. The ITALK, a component of the Bilingual English-Spanish Assessment (BESA; Peña et al., 2018), is designed to elicit parent reports of speech and language abilities in English and Spanish. Parents are asked to rate five areas of speech and language (pronunciation, comprehension, grammar, vocabulary, and sentence length) on a 5-point scale for each language. These ratings are then averaged, resulting in scores ranging from 1 and 5 for each area and language. Scores of the two languages are then compared and the higher score is retained (i.e., better language; maximum score of 5).

The second caregiver report measure, the BIOS (Peña et al., 2018), documents current exposure in terms of hour-by-hour input and output in each language, as well as year-by-year cumulative exposure. In this measure, caregivers are requested to provide information on language input and output for seventeen 1-hr intervals (between 7 a.m. and 11 p.m.) for both a typical weekday and a typical weekend day. For each 1-hr interval, caregivers are asked what language(s) the child hears and uses: Spanish, English, or both, as well as their communication partners during each activity. Each language response is coded (English = 0, both = 1, Spanish = 2), summed, and projected for a full 7-day week. An input–output composite percentage was calculated using the weighted input and output average of each language to combine total language exposure and use. Percentages for English and the home language are inversely related. For example, a child with a BIOS English input–output score of 70 indicates that the child is exposed to and uses English 70% of the time in a typical week and Spanish the other 30%. A translated and adapted Vietnamese version of the BIOS (Pham & Tipton, 2018) was used for Vietnamese families and retained the same scoring system.

Pairwise matching was carried out among the larger pool of participants to select 30 individuals and form two groups with similar background variables. The matching process involved pairing 15 SE bilinguals with 15 VE bilinguals, based on the following criteria:

• (i)  Chronological age difference within 5 months.

• (ii)  Current English experience, as measured by an English input–output score from the BIOS within 15%.

• (iii)  Language ability, as measured by ITALK better language average score within 0.6.

Descriptive statistics of the participants are shown in Table 2. No significant group differences were found for sex, χ²(1, n = 15) = 0, p = 1, age, t(28) = 0.07, p = .94, BIOS English input–output composite score, t(28) = 0.64, p = .53, BIOS home language score, t(28) = −0.64, p = .53, and ITALK better language score, t(28) = −0.74, p = .46. No significant group differences were found for age of first English exposure at school, t(26) = 1.94, p = .06. Vietnamese speakers (M = 3.6, SD = 0.83) and Spanish speakers (M = 2.85, SD = 1.21) were exposed to English at school at relatively similar ages. However, SE bilinguals had greater exposure to English at home compared to VE bilinguals, χ²(1, N = 30) = 10.91, p ≤ .001. Twenty-two out of 30 total participants were indicated by their caregivers as being exposed to English by a family member at home. Participants exposed to English at home are composed of seven children (47%) from the VE group and 15 children (100%) from the SE group.

Table 2.

Participant demographics and language background measures.

	SE bilinguals	VE bilinguals	Total	p value
n	15	15	30
Sex (%)				1.0
Girls	9 (60%)	9 (60%)	18 (60%)
Boys	6 (40%)	6 (40%)	12 (40%)
Age in months	87.07 (9.21)	87.33 (10.58)	87.2 (9.73)	.94
Min	72	72	72
Max	100	105	105
BIOS English	67.73 (14.16)	70.87 (12.62)	69.3 (13.27)	.53
Min	50.78	52.86	50.78
Max	98.7	100	100
BIOS home language	32.27 (14.16)	29.13 (12.62)	30.7 (13.27)	.53
Min	1.3	0	0
Max	49.22	47.14	49.22
English retell	0.44 (0.12)	0.41 (0.19)	0.42 (0.16)	.65
Min	0.24	0.06	0.06
Max	0.71	0.71	0.71
ITALK	4.43 (0.57)	4.29 (0.46)	4.36 (0.51)	.46
Min	2.8	3.2	2.8
Max	5	5	5
Monomorphemic correct	14.87 (0.35)	13.67 (2.66)	14.27 (1.96)	.10
Min	14	5	5
Max	15	15	15
Monomorphemic omitted	0.07 (0.26)	1.27 (2.69)	0.67 (1.97)	.10
Min	0	0	0
Max	1	10	10

Note. Means (SDs), minimum and maximum for SE (Spanish–English) bilinguals and VE (Vietnamese–English) bilinguals. Values for sex represent the number of participants and corresponding percentages in parentheses. BIOS = Bilingual Input Output Survey; ITALK = Inventory to Assess Language Knowledge.

Characterization of Participants' Speech and Language

In addition to the matching criteria (age, BIOS, ITALK), participants completed two English expressive measures to characterize their speech and language skills: a monomorphemic task and a story retell task. The monomorphemic task verified participants' ability to produce the speech sounds needed for some of the target English grammatical morphemes. The monomorphemic task required participants to repeat 15 words to document the production of final consonant sounds (i.e., /s/, /z/, /t/, /d/, /nd/, /ks/). Participants were shown a picture on the computer screen and played an audio recording of its label which the participant was prompted to repeat.

As shown in Table 2, there were no significant group differences in accuracy on the monomorphemic task, t(28) = −1.73, p = .1. Further analysis of monomorphemic performance found that despite Vietnamese speakers (M = 1.27) omitting final sounds more than Spanish speakers (M = 0.07), there was no significant group difference in final sound omission, t(28) = 1.72, p = .1.

The story retell task was used as a direct measure of oral narrative production in English as narrative assessment has been found to reliably assess bilingual children's ability to use vocabulary and grammar, and express and organize complex thoughts (Miller et al., 2006). In this study, story retell was administered following the standardized English protocol of the Multilingual Assessment Instrument for Narratives (MAIN; Gagarina et al., 2019), an assessment tool used to assess oral narrative skills in children from 3 to 10 years old that has been used with monolingual and bilingual children in over 25 languages. The present study focused on the MAIN's measure of narrative macrostructure (Section A), which evaluates the overall organization and coherence of a story. Macrostructure includes key elements such as story structure (e.g., setting, resolution), which are critical for assessing how well children can organize and express ideas in English. This measure served as an index of children's overall English proficiency.

During administration of the MAIN, participants were shown six pictures of a story on a computer screen while listening to a recorded story model. When the story finished, participants were prompted to retell the story in their own words using the same pictures on the screen. Responses were scored according to the MAIN manual for story structure (Section A). Each story is composed of a setting and three episodes. The setting includes up of two situational elements, time, and place, while each episode is composed of five elements: the initiating event, goal, attempt, outcome, and reaction. Each element is worth 1 point (maximum of 17). Values for English retell in Table 2 represent the total proportion correct out of 17 points. Descriptively, the SE group (M = 0.44, SD = 0.12) seemed to achieve a higher average score on their English narratives than the VE group (M = 0.41, SD = 0.19); however, this difference was not statistically significant, t(28) = −0.47, p = .65.

Morphosyntax Assessment

To assess the accuracy of 13 English grammatical forms, participants completed a computer-based cloze task for each target form. These 13 grammatical forms were selected to test a wide range of grammatical markers and are based on previous studies of bilingual performance on English grammar (e.g., Bedore et al., 2018; Bland-Stewart & Fitzgerald, 2001; Davison & Hammer, 2012). Additionally, many of the target forms are based on Brown's (1973) morphemes, which has established data on age of acquisition for monolingual speakers (Balason & Dollaghan, 2002) and has led to further study of morpheme accuracy among bilinguals. The present study is part of a larger project exploring accurate diagnosis of developmental language disorder (DLD) in English language learners. Eleven forms are shared between Brown's 14 morphemes and our targets: plurals, articles, prepositions, regular past tense, possessive, irregular past tense, copula, present progressive, and third person singular. Three of Brown's 14 morphemes are not included (uncontractible auxiliary, contractible auxiliary, third person irregular) and four targets are added (negation, relative clause, question inversion, passive). All targets in this study have been identified as potential clinical markers for DLD among SE bilinguals (Bedore et al., 2018), and the present study extends findings to VE bilinguals.

This morphosyntax task involved participants listening to a model sentence while viewing a picture and then being prompted to complete a phrase with missing words, while observing the corresponding pictures on the screen. Prior to administering test items, instructions were read aloud to each child, followed by two demonstration items. For example, they would hear, “Here, there is one kid. But here, there are three ___ (e.g., kids).” Corrective feedback was given only during demonstration phase if the child produced an incorrect response, with the correct answer repeated until the child accurately repeated it. This ensured that participants fully understood the task before beginning the test items.

All target forms were elicited in a similar manner, except for Question Inversion. On items of question inversion, participants listened to a sentence describing a corresponding picture and were prompted to create a question about a hypothetical situation. For instance, the child would hear, “The nurse wants to know if Maria is feeling sick. Pretend you are the nurse. What will you ask Maria?” and the child would respond with, “Are you feeling sick?” or a similar response.

For each grammatical form, the child was given two practice items followed by seven to eight test items. Research assistants recorded the child's responses, which were scored as correct or incorrect. The percentage of correct responses was then computed for each target form. Full credit (score of 1) was given for responses that matched the target forms as well as for responses that indicated partial knowledge of the target form. For example, in the regular past cloze task, participants received the prompt, “Here Nam chases a ball. Yesterday he did it, too. Yesterday, he ___ [chased a ball].” A response such as “catched a ball” would be considered correct, as it demonstrates a grasp of the grammatical rule governing the addition of –ed to verbs. Another example of this overgeneralization of the past tense –ed marker was found in the responses “sweated” and “jogsed” for the prompt, “Here Juan jogs. Yesterday he did it, too. Yesterday he ___ [jogged].”

Results

Relative Accuracy by Group

Our initial research question examined whether the bilingual groups performed similarly on English morpheme accuracy. Figure 1 plots the average proportion correct of all 13 cloze forms for each group. From initial observation, the SE group achieved higher scores on nearly all forms compared to the VE group. When considering 0.80 accuracy as achieving mastery, the VE group mastered only one form, copula. In comparison, the SE group achieved mastery for seven forms: plural, negation, preposition, possessive, passive, copula, and third person singular.

Figure 1.

Proportion of correct English cloze items by language group. The line at 0.80 denotes the threshold for mastery, indicating that a form is considered acquired by achieving 80% accuracy or higher. Forms marked by an asterisk represent statistically significant differences in performance between VE (Vietnamese–English) and SE (Spanish–English) groups. Legend: light gray = Spanish–English, dark gray = Vietnamese–English.

Independent samples t tests revealed a significant difference in the average scores for five forms in which the SE group performed higher than the VE group: plural (d = 0.26), regular past tense (d = 0.35), possessive (d = 0.29), irregular past tense (d = 0.30), and third person singular (d = 0.31).

Accuracy rankings of forms by each group are shown in Table 3. The Spearman's rho correlation between the rankings of the SE and VE groups indicated a moderate, positive correlation, r(13) = .47, although it was not statistically significant (p = .11). This suggests that while the groups showed some alignment in their rankings, the relationship was not strong enough to reach statistical significance. Notably, SE and VE groups were identical in their most and least accurate forms, copula and irregular past, respectively, with some variability in rankings between the groups.

Table 3.

Group differences on cloze tasks.

Grammatical form	SE bilinguals	Rank	VE bilinguals	Rank
Copula	0.98	1	0.97	1
Possessive*	0.94	2	0.60	8
Plural*	0.92	3	0.69	7
Negation	0.87	4	0.72	5
Passive	0.87	4	0.70	6
Third person singular*	0.83	6	0.53	11
Preposition	0.82	7	0.74	4
Article	0.79	8	0.75	2
Relative clause	0.79	8	0.56	10
Regular past tense*	0.76	10	0.48	12
Question inversion	0.68	11	0.75	2
Present progressive	0.64	12	0.57	9
Irregular past tense*	0.45	13	0.22	13

Note. Mean accuracy of each form is displayed for the SE (Spanish–English) bilinguals and VE (Vietnamese–English) bilinguals. Ranking from most to least accurate is displayed separately for each group.

^*Indicates a between group difference as a significance level of p < .05.

Error Type by Group

The second research question examines whether the bilingual groups produced similar types of errors. To address this question, we analyzed error patterns for cloze tasks that were produced by at least 25% of the participants in each language group (see Table 4 below). For the Spanish-speaking children, errors made by over half of the group included article omission, article definitiveness, regular past tense omission, pronoun omission with relative clauses, irregular past tense substitution, and third person present omission. For the Vietnamese-speaking students, errors made by over half of the group included plural marker omission, article definitiveness, article omission, regular past tense omission, possessive marker omission, irregular past tense substitution, and third person present omission.

Table 4.

Error patterns.

Form	Pattern	Target	Child response	Number of participants
				SE bilinguals	VE bilinguals
Pluralsb	Omission	socks	sock	2	12
Articlesa,b	Definitiveness	the moon	a moon	9	8
	Omission	a pig	pig	9	7
Negation	Number error	doesn't	don't	1	5
Preposition	Substitution	on, under, to, next to	in	4	5
Relative clause	Substitution	that has books	with books	7	6
	Omission	that has ducks	has ducks	1	6
Regular pastb	Omission	chased	chase	5	11
	Commission	yawned	yawns	1	7
Question inversion	Inversion error	can/will you help	if she can help	5	0
Possessiveb	Omission	nurse's bag	nurse bag	1	12
Passive	Substitution (preposition)	was opened by	was opened from	6	0
	Substitution (tense)	is cleaned by	clean by	0	4
Irregular pasta,b	Substitution	rang	ring	13	14
	Overgeneralization	caught	catched	5	3
Copula	Omission	are red	red	2	1
Present progressive	Omission	are laughing	laughing	5	7
Third person singulara,b	Omission	locks	lock	8	10

Note. Number of participants' columns represent the number of children (out of 15) in the SE (Spanish–English) group and VE (Vietnamese–English) group that committed the error.

^aPattern shown by more than half of the SE bilinguals (n ≥ 8).

^bPattern shown by more than half of the VE bilinguals (n ≥ 8).

In the five forms in which the VE group performed significantly lower than the SE group (plural, regular past tense, possessive, irregular past tense, and third person singular), most errors involved omission of the final consonant for target words. This included omitting the final /s/ in plural, possessive, and third person singular forms as well as omitting the final /d/ and /t/ sounds in regular past tense forms. Accuracy for the VE group ranged from 22% to 69% for these five forms, compared to SE group which ranged from 45% to 94%.

Discussion

This study compared English grammatical performance of typically developing bilingual children who spoke either Spanish or Vietnamese at home. All participants completed the same production task in order to make direct comparisons across two distinct bilingual groups. We used pairwise matching of individual participants to reduce the effects of background variables and allow for direct comparison of task scores and performance.

Our first research question asked whether the bilingual groups performed similarly on English morpheme accuracy. Both bilingual groups demonstrated high accuracy for copulas (e.g., “are red”), with over 85% correct responses, indicating that this particular grammatical structure was relatively easy to acquire (see Table 3). Similarly, both groups showed low accuracy for irregular past tense (e.g., “rang”), with less than 50% correct, reflecting its well-documented difficulty for both bilinguals and monolinguals. These findings generally align with previous research confirming the sequence of mastery outlined in Brown's (1973) morphemes in monolinguals and bilinguals (e.g., Bedore et al., 2018; Padilla, 1978), at least for forms with the highest and lowest accuracy.

However, the two bilingual groups differed in accuracy on forms such as possessive (ranked 2 and 8 among Spanish and Vietnamese bilinguals, respectively) and prepositions (ranked 7 and 4 among Spanish and Vietnamese bilinguals). As shown in Figure 1, there were group differences in accuracy for five grammatical forms (plurals, regular past, possessive, irregular past, and third person singular) in which Spanish bilinguals were more accurate than Vietnamese bilinguals. Although we carefully matched individual participants on current English exposure, we noted that there were group differences in cumulative exposure in that the Spanish bilinguals had earlier exposure to English than the Vietnamese bilinguals. Thus, although these group differences in L2 morpheme accuracy are primarily attributed to L1 typological differences here, cumulative exposure may also be influencing results (e.g., Xu Rattanasone & Demuth, 2023). These findings collectively underscore the complexity of bilingual morphosyntactic development, and the need to consider bilingual context and language exposure when evaluating language acquisition patterns. This also highlights the importance of designing language instruction and assessment tools that are sensitive to the specific L1 of the bilingual children.

The second research question examined error types across groups. Although average accuracy of a grammatical form may be relatively high (as discussed in Research Question 1), accuracy may still vary across items, and individual children will produce errors on occasion. Here, we highlight the error patterns shown in Table 4 that were exhibited by more than half of participants in each bilingual group (denoted with superscripts in the table).

We first focus on patterns exhibited by many children across groups (see Table 4; see forms with superscripts a and b) as these errors could reflect developmental effects, that is, errors common across L2 learners despite L1 typological differences. The grammatical form that was the most difficult for children was irregular past. Nearly all participants in both groups made substitution errors, replacing the target irregular past form with its present tense counterpart (e.g., ring instead of rang). Additionally, one third of the Spanish bilinguals and slightly fewer from the Vietnamese bilinguals demonstrated overgeneralization errors, applying regular past tense rules to irregular verbs (e.g., catched instead of caught). Because irregular past does not follow a consistent grammatical pattern, children acquire individual items with increased exposure and practice. Thus, difficulty with irregular past may be developmental in that accuracy increases with exposure (e.g., Xu Rattanasone & Demuth, 2023).

Errors with third person singular were also exhibited by many individual participants across groups. Over half of the participants in both groups dropped word final /s/ in their productions (e.g., lock instead of locks). Notably, at the group level, third person singular was produced with accuracy by the Spanish bilinguals (83%) but not the Vietnamese bilinguals (53%). Nonetheless, when children in each group did make errors, their error patterns were similar. This may be an example of how L1 typology contributes to lower accuracy among the Vietnamese bilinguals, whereas developmental effects influence the types of errors L2 learners make despite L1 typology.

Finally, many children across groups exhibited errors with articles. More than half of Spanish and Vietnamese bilinguals demonstrated definitiveness errors in article production, replacing the target definite article the with the nondefinite article a (e.g., the moon vs. a moon). Additionally, over half of the Spanish bilinguals and nearly half of the Vietnamese bilinguals omitted the target article entirely in their productions (e.g., pig instead of a pig). These articles could reflect how definiteness is marked in Spanish and Vietnamese compared to English. Spanish requires articles in some contexts where English does not (Snape et al., 2013), which can lead to overgeneralization or omission when adapting to English productions. In contrast, Vietnamese does not use articles at all, relying on classifiers and contextual information to indicate definiteness (Pham & Simpson, 2024). At the group level, articles were produced with similar accuracy, 79% and 75% by Spanish and Vietnamese bilinguals, respectively.

We now turn our attention to error patterns in Table 4 that were exhibited by many children in one group but not the other. These patterns may be more specific to L1 typology. The Vietnamese bilingual group demonstrated significantly more omissions of the plural marker –s and regular past tense marker –ed compared to the Spanish bilinguals. This highlights the strong influence of L1 for Vietnamese bilinguals, particularly in cases where English forms required a change in the final sound of the target word. This result likely reflects Vietnamese's phonological constraint of limited word final endings. Notably, commission errors of regular past tense in which the –ed was replaced with –s (yawns for yawned) primarily occurred with Vietnamese bilinguals. This error type suggests an awareness of some type of word ending and an attempt to fill in this slot. This notion aligns with the UM's concept of ease of storage (MacWhinney, 2018), where linguistic forms that are more regular and frequent in their patterns are easier to encode and store in memory. The cognitive effort needed to store and retrieve new phonological forms likely affects the Vietnamese bilingual group's ability to store these patterns effectively.

Finally, although it did not rise to the level of half of the group, two error patterns are noted to occur in Spanish bilinguals but not in Vietnamese bilinguals: errors with question inversion and passives. In question inversion, one third of Spanish bilinguals failed to invert the subject and verb (e.g., if she can help instead of can you help). This likely resulted from their attempt to complete the prompt, “What will she ask him?” as “[she will ask him] if she can help,” suggesting an embedded response rather than an error. Unlike Spanish, which requires subject–verb inversion (Cuza, 2013), Vietnamese does not use inversion in question formation, instead utilizing question words at the beginning or end of sentences (Thuan & Bruening, 2013). This difference in grammatical structures may explain why Spanish bilinguals, but not Vietnamese bilinguals, exhibited these specific inversion errors.

In passives, two separate patterns were noted: preposition substitution in Spanish bilinguals and tense substitution in Vietnamese bilinguals. Six participants from the Spanish bilingual group frequently substituted prepositions (e.g., was opened from instead of was opened by), likely influenced by the Spanish preposition de which translates to English as both “from” and “by.” This similarity contributed to inaccuracies in English preposition usage, mirroring errors noted by Bland-Stewart and Fitzgerald (2001). Four participants from the Vietnamese bilingual group often replaced the past participle with the present tense in passives (e.g., clean by instead of is cleaned by), omitting the word final –ed in a pattern similar to that seen with regular past tense verbs.

To summarize, L1 influence was evident in specific forms. For example, Vietnamese bilinguals exhibited a high frequency of omission errors for forms requiring word-final sounds, such as plurals (sock for socks) and possessives (nurse bag for nurse's bag; see Table 4). These errors align with Vietnamese' limited phonetic inventory of word-final consonants, highlighting cue divergence between L1 and L2. In contrast, Spanish bilinguals showed preposition substitution errors (e.g., was opened from for was opened by), likely due to Spanish's use of de to indicate both meanings. These substitutions highlight how L1 transfer and cue divergence can shape morphosyntactic acquisition in bilingual learners.

These observed differences reflect broader processes of cue convergence and divergence (MacWhinney, 2018). Monolingual Spanish speakers tend to rely on nonverb agreement cues (e.g., Reyes & Hernandez, 2006), monolingual English speakers on word order (e.g., Bates & MacWhinney, 1989), and bilingual VE speakers on word order and animacy (Pham & Kohnert, 2010). Bilingual learners navigate these competing cues, which can explain variations in their English productions. Cue competition involves these cues vying for influence in sentence processing, with the most reliable cues typically winning. Over time, cue convergence enables learners to gradually adjust their reliance on these cues to match adult speech, increasing the strength of reliable cues through continued exposure and interaction. These findings underscore the importance of explicitly teaching bilingual students the most reliable cues for challenging grammatical forms, enabling more effective language processing in education settings.

Study Limitations

The current study has several limitations that warrant consideration. First, the sample sizes for both Vietnamese (n = 15) and Spanish (n = 15) participants represent only a small subset of bilingual children in the United States. Although small sample sizes can reduce a study's statistical power, it is important to note that even in such cases, meaningful and statistically significant results can be obtained if the study is well controlled. As per Hope's (2015) suggestion, we strived to minimize uncontrolled variance, given the challenges in recruiting larger groups of school-age SE and VE participants. Nevertheless, future research should aim to include larger samples of Spanish- and Vietnamese-speaking individuals for more robust insights.

Second, our study focused solely on speakers of two languages. Expanding the study to incorporate a broader array of language groups, especially those with no bound inflectional morphology such as Vietnamese, would allow us to determine whether the observed difficulties in producing English grammatical forms characterized by word-final sounds are common in other non-inflected languages. Third, some children may have had trouble understanding the instructions for certain cloze task items, particularly for question inversion and negation. For question inversion, children were required to generate a new sentence, which differed from the phrase-completion format of previously administered tasks and may have caused confusion. Similarly, for negation, some children produced affirmative sentences instead of using the target negation form to modify the example. These issues could reflect a lack of clarity in task instructions or the cognitive demands of shifting between task formats. Although these limitations do not undermine the overall validity of the task, they highlight the need for clearer pretask explanations and practice examples to fully understand the task requirements.

Finally, we acknowledge that our assessment of grammatical forms was conducted within the context of structured cloze procedure. The advantage of a cloze procedure is to elicit a wide variety of grammatical forms that may not spontaneously occur otherwise. A disadvantage to cloze procedures is their decontextualized nature, in which the unfamiliarity of certain elicitation procedures may underrepresent children's language ability. Thus, results from a cloze procedure could complement or contrast with findings from contextualized language samples such as spoken narratives.

Concluding Remarks

The purpose of this study was to compare English grammatical performance of typically developing VE and SE bilingual children using pairwise matching to control for background variables. Both groups were tested on 13 English grammatical forms in a cloze task. High accuracy rates (85%+) were achieved for copulas and lower accuracy (< 50%) were achieved for irregular past tense, reflecting developmental patterns similar to monolingual English-speaking children and L2 learners. VE children exhibited lower accuracy in forms requiring word-final sounds due to phonological constraints in Vietnamese, while SE children displayed more preposition-related errors. These findings emphasize the importance of considering first-language influence when designing culturally responsive language instruction and assessments. This foundational data can be used to inform the creation of tailored evaluation tools suitable for a diverse range of English language learners and can reduce misdiagnosis of language disorders. Educators can also use normative data to better tailor instruction and intervention strategies that meet the linguistic needs of bilingual students.

Data Availability Statement

The data generated and/or analyzed are available from the corresponding author upon reasonable request.

Funding Statement

This research was supported by and all authors received funding from the National Institute on Deafness and Communication Disorders, National Institutes of Health Grant R01DC018329 for this work (PI: Elizabeth D. Peña).