Bilingual Children's Grammatical Productivity From Spontaneous Language Samples in English and Spanish

Abstract

Purpose:

Much work has been done to characterize the grammatical productivity of English-speaking monolingual children and the English skills of bilingual children. However, less work has been done to characterize productivity of bilingual children in other languages. This study investigates the clinical utility of a Spanish grammatical productivity measure to address the need for linguistically responsive measures in bilingual language assessment.

Method:

Spontaneous language samples were collected for 23 typically developing (TD) preschoolers and nine preschoolers with developmental language disorder (DLD). In Spanish, children's productive use of definite articles, indefinite articles, and direct object clitics was calculated to derive group averages. In English, children's productive use of tense and agreement morphemes was examined.

Results:

In Spanish and in English, TD children demonstrated higher productivity than children with DLD.

Conclusion:

In addition to English productivity, children's Spanish grammatical productivity is a promising clinical tool to be used in the assessment of Spanish–English bilingual children.

Current practices in the assessment of bilingual children's language abilities highlight the clinical utility of language sample analysis (LSA; Arias & Friberg, 2017; Castilla-Earls et al., 2020; Ebert, 2020; Simon-Cereijido & Gutiérrez-Clellen, 2009). Clinical utility, in the context of language samples, refers to the value of information derived from these measures with respect to decision making from the beginning (i.e., identification) through the end (i.e., outcome measures) of service delivery (Costanza-Smith, 2010). LSA offers unique insight into children's language production in naturalistic contexts. As a result, language samples can provide useful clinical information for the assessment of language disorders, especially for bilingual children (Bedore et al., 2010; Ebert, 2020). LSA in the first language (L1) and second language (L2) offers insight into their grammatical skills in both of their languages. LSA measures that have been identified as impactful in clinical decision making include mean length of utterance in words (MLUw), number of different words (NDW), and percent accuracy of grammatical forms (Ebert, 2020; Ebert & Pham, 2017; Pham, 2016; Simon-Cereijido & Gutiérrez-Clellen, 2009).

Recent developments in LSA have highlighted the utility of another measure that may be derived from language samples: grammatical productivity. Productivity is a measure of the number of sufficiently unique productions of a certain grammatical structure (Hadley & Short, 2005). Previous research has shown that grammatical productivity mitigates the effect of rote, memorized forms on measures of morphosyntactic performance (Fitzgerald et al., 2012). In monolingual populations, grammatical productivity has differentiated typically developing (TD) children from their peers with developmental language disorder (DLD; Gladfelter & Leonard, 2013; Pham et al., 2023). Additionally, work in bilingual populations has established that bilingual children's English (L2) productivity also yields group differences between TD children and their peers with low-average language abilities (Potapova et al., 2018; see also Potapova & Pruitt-Lord, 2019). The current study aims to extend analyses of bilingual children's grammatical productivity to their L1. Specifically, we examine Spanish–English bilingual children's Spanish and English productivity of grammatical forms in TD children and their peers with DLD.

Grammatical Productivity in Monolingual Children

For English-speaking monolingual toddlers and preschoolers, much work has been done to characterize the grammatical productivity of their emerging grammatical structures. Decades of research on morphosyntactic development in English-speaking children has established that tense and agreement morphemes (i.e., auxiliary BE [he is jumping], copula BE [he was hungry], auxiliary DO [he does like apples], third-person singular –s [he sleeps], and past tense –ed [he jumped]; see Rice & Wexler, 1996) are clinical markers of language disorders for English-speaking children. Hadley and Short (2005) first proposed the use of grammatical productivity to characterize the language abilities of young children aged 2;0–3;0 (years;months) using these tense and agreement morphemes. Given the particular grammatical deficits that children with DLD demonstrate, Hadley and Short (2005) examined the onset of grammatical structures and their productive use in a sample of children with low-average language abilities or those at risk for specific language impairment (SLI). Importantly, these productivity measures require sufficiently unique productions (i.e., bound morphemes such as past tense –ed must use different verbs to credit for productive use; see Hadley & Short, 2005) and prevent giving credit for rote forms. Participants in this study included 20 two-year-old children who were either at risk for being diagnosed with SLI or who scored in the low-average range on several standardized assessments and demonstrated low MLUw. Results showed that children's scores on onset measures (e.g., tense marker total, total productivity score) correlated with traditional language ability measures (e.g., MLU, Index of Productive Syntax) and these measures differentiated between those with low-average language and those at risk for SLI. Moreover, children's productivity score total (the number of sufficiently unique uses of tense and agreement morphemes with a limit of five instances per category) was most strongly correlated with mastery (percent accuracy) of tense and agreement morphemes. Given these results, the authors of this study recommended the use of productivity scores to characterize children's skills. Overall, the results of this study demonstrate that children's productive use of grammatical structures differentiates children based on their language ability.

Gladfelter and Leonard (2013) extended this work and examined the use of productivity in differentiating children with SLI and their TD peers in a sample of older, preschool-aged children. This study examined productivity (the number of sufficiently unique uses of tense and agreement morphemes with a limit of five instances per category), a tense marker total (a measure of different surface forms of tense and agreement morphemes), and the finite verb morphology composite (FMVC; a measure of children's correct use of morphemes over the number of obligatory contexts) with monolingual, English-speaking children aged 4;0–5;6. Participants in this study were divided by language ability (TD vs. SLI groups) and by age (children 4;0–4;6 and 5;0–5;6). Results showed that all three measures demonstrated group differences between TD children and their peers with DLD. However, acceptable sensitivity (proportion of true positives correctly identified) and specificity (proportion of true negative correctly identified) levels differed by age group. For productivity scores, sensitivity and specificity were 67% and 87% for the younger age group, respectively, and 85% and 80% for the older group. For the tense marker total, sensitivity and specificity were 83% and 87% for the younger age group, respectively, and 77% and 80% for the older group. For the FMVC, sensitivity and specificity were both 100% for the younger age group and 92% and 93% for the older group. Importantly, this study highlighted how measures of productivity allow for clear clinical applications, as they demonstrate which morphemes may be particularly difficult for an individual child. Given that these morphemes are often targets for preschool language intervention, productivity also has potential for progress monitoring measures (Gallagher & Hoover, 2020).

For speech-language pathologists working in schools, monitoring progress through language sample measures is a potential pathway to integrate assessment outcomes with intervention (Goffman & Leonard, 2000). For example, Gallagher and Hoover (2020) demonstrated how language sample measures such as productivity scores can be used to monitor progress throughout intervention in a sample of English-speaking monolingual children. Moreover, the authors highlight that using measures such as productivity may better capture finer grained changes in children's language skills than more global measures such as MLU. More research is needed on how to best integrate LSA into progress monitoring practice for speech-language pathologists (SLPs). For productivity, these studies show that productivity can offer adequate to good diagnostic accuracy and allow for clinical applications beyond diagnosis (e.g., goal-setting, progress monitoring).

Furthermore, measures such as productivity and the tense marker total protect against inflated scores that (percent) accuracy is particularly susceptible to. Indeed, inflated scores for young children have yielded a particular pattern where children demonstrate a higher percent accuracy initially and then lower scores at later ages. This pattern of apparent backtracking (Fitzgerald et al., 2012) is due to children's productions of rote, memorized, and high-frequency combinations (e.g., I'm, he's, it's) that overcredit children's accuracy of these tense and agreement morphemes (Wilson, 2003). Thus, productivity mitigates the inflating effect of these forms that are learned as lexically specific constructions (Wilson, 2003).

Recent work has examined productivity in other languages. Pham et al. (2023) investigated clinical markers of DLD in monolingual Vietnamese-speaking children and examined children's production of classifiers (i.e., grammatical structures used to categorize nouns). Ten TD Vietnamese-speaking children in kindergarten were matched with peers with DLD, and these participants completed developmentally appropriate language samples. Results from this study showed that children with DLD produced a limited set of classifiers, while children with TD produced a greater diversity of classifiers and in higher amounts. The results of this study suggest that children's accuracy, diversity, and productivity of classifiers, a salient grammatical feature of Vietnamese, is a clinically impactful measure that demonstrates differences between TD and DLD groups. Importantly, this work also demonstrates that children's productive use of grammar is a useful measure in languages other than English. In addition, this work demonstrates that productivity is a measure that must take into account clinical markers that are specific to a language.

English Grammatical Productivity in Bilingual Children

To our knowledge, only two studies have examined Spanish–English bilingual children's productivity of English tense and agreement morphemes. Potapova et al. (2018) examined the language samples of 74 preschool-aged TD bilinguals and the language samples of 19 peers with low language abilities using an English accuracy measure, productivity scores, and the tense marker total. Similar to studies done with English monolingual children, this study examined bilingual children's use of auxiliary BE, copula BE, auxiliary DO, third-person singular –s, and past tense –ed. Analyses of language samples collected at the beginning and the end of the school year showed that bilingual children's productivity and tense marker total was positively correlated to other language sample measures, MLUw and NDW. Conversely, children's accuracy of tense and agreement morphemes was not significantly correlated to the other language sample measures. Productivity scores and the tense marker total increased over time for both TD children and their peers with low language abilities. However, 42% of TD children and 50% of children with low language abilities demonstrated lower accuracy across time, potentially demonstrating a pattern of backtracking in this sample of children (Fitzgerald et al., 2012). Results from this study also showed group differences in children's tense marker total and productivity scores. Importantly, productivity measures provided unique information on children's language abilities in this study. Global measures, such as MLUw and NDW, average children's performance across all complete and intelligible utterances, while productivity measures offer insights into patterns of how children produce specific grammatical morphemes. This approach accounts for children's emerging language skills while mitigating inflation of scores. Taken together, these results suggest that productivity may capture unique information about bilingual children's language development in the L2 that is not otherwise captured by accuracy alone.

Potapova and Pruitt-Lord (2019) examined the relative use of English tense and agreement morphemes in Spanish–English bilingual preschoolers. Previous research has established that English monolingual children may demonstrate a particular pattern of emergence of tense and agreement morphemes such that copula BE was the most frequently used and productive of the tense agreement morphemes, followed by third-person singular –s, past tense –ed, and auxiliary DO (Fitzgerald et al., 2012). The same study revealed that auxiliary BE is the least frequently used and least productive of these morphemes. Additionally, this pattern has also been observed for older, preschool-aged children (Gladfelter & Leonard, 2013). Potapova and Pruitt-Lord (2019) examined whether or not bilingual children with TD and children with low language skills demonstrated this same pattern of English tense and agreement morpheme acquisition. Results from this study demonstrated that bilingual children demonstrated a different pattern of emergence, such that copula BE remained the most productive, followed by auxiliary BE, third-person singular –s, past tense –ed, and auxiliary DO. These results suggest that bilingual children demonstrate a different pattern of use of tense and agreement morphemes in English than do their monolingual peers. As such, further investigation into how bilingual children demonstrate productive use of grammatical features of their languages is merited. Specifically, productivity measures provide new and useful methods of describing broad patterns of language development. Thus, it is critical to investigate this measure cross-linguistically.

Spanish Grammatical Development in Bilingual Children

Recently, considerable work has been done to characterize the development of Spanish in Spanish–English bilingual children (Baron et al., 2018; Castilla-Earls et al., 2021; De Anda et al., 2023; Hernandez et al., 2024). Over the last decade, much attention has been given to examining how morphosyntactic ability can differentiate TD bilingual children from their peers with DLD (Castilla-Earls et al., 2021, 2023). This research has resulted in advancements in LSA for bilingual children and the norming of standardized assessments with bilingual populations (Peña et al., 2014). Given these advances, patterns in the types of Spanish grammatical structures that are developmentally and clinically significant for Spanish–English bilingual children begin to emerge.

Spanish-speaking children are tasked with acquiring features of both verb morphology and grammatical markers related to nouns (vs. in English). For example, articles in Spanish are required to agree with the noun they modify in gender (masculine/feminine) and number (singular/plural). Verbs in Spanish inflect for person, number, tense, aspect, and mood. Baron et al. (2018) examined grammatical development in a sample of 228 Spanish–English bilingual children aged 4;0–7;6. Results from this study showed that early acquired grammatical features included imperfect (e.g., ella jugaba [she played]), plural –s (e.g., gatos [cats]), singular articles (e.g., el niño [the boy]), and conjunctions (e.g., y [and]), which all children produced with high accuracy during an elicitation task. Plural articles (e.g., los niños [the boys]) and verb conjugations varied across the bilingual sample, such that children who were Spanish-dominant demonstrated higher accuracy of these grammatical features. Finally, prepositions (e.g., en [in/on]), direct object clitics (e.g., lo [it]), and the subjunctive mood (e.g., Espero que tengas un buen día [I hope you have-subjunctive a good day]) were the least accurate grammatical features in this sample of Spanish–English bilingual children across both English-dominant and Spanish-dominant groups (see Baron et al., 2018). In summary, this study suggested that Spanish grammatical features follow a difficulty hierarchy for Spanish–English bilingual children and that use of articles and direct object pronouns varies in difficulty for this population.

The clinical markers of DLD in Spanish-speaking children have been well studied, including in Spanish–English bilingual children in the United States. Importantly, these studies have focused on accuracy measures, which involve accounting for the number of correct and incorrect productions of a certain grammatical structure. In contrast, measures of productivity disregard incorrect productions in favor of focusing on the depth and breadth of correct productions. Moreover, much of the work that has been done in characterizing children's production of grammatical structures in Spanish has used elicitation tasks (e.g., cloze, sentence repetition). In contrast to language samples, which require children to spontaneously produce language, these elicitation tasks are highly structured in favor of experimental control. Productivity measures aim to account for the full range of children's productions of grammatical structures and thus require a more naturalistic measure such as language sampling.

In one study, Castilla-Earls et al. (2023) examined the accuracy of Spanish articles, direct object clitics, verbs, adjectives, plurals, and the subjunctive mood in children with and without DLD. Using an elicitation task that had previously been used with Spanish–English bilingual children (Castilla-Earls et al., 2020, 2021), results from this study showed that children aged 4;0–8;11 were more accurate in their productions of articles and verbs than their productions of clitics and the subjunctive. Furthermore, children without DLD outperformed their peers with DLD across all grammatical structures, and this pattern continued as children aged over a 2-year period. Thus, for Spanish-speaking children, accuracy on verb phrases and grammatical markers related to nouns are measures that have potential for clinical use. Importantly, neither group of children (aged between 4;0 and 8;11) reached 100% accuracy on any grammatical structure, suggesting that these bilingual children were still in the process of mastering these grammatical features in Spanish. In summary, much work has been done to characterize the accuracy of bilingual children's productions of grammar. However, questions remain on productivity-based measures of children's grammatical skills, as these measures may better capture children's early development. Prior studies have primarily focused on accuracy measures, and investigation of productivity measures would go beyond what these studies have reported by examining the breadth and depth of children's productions.

The Current Study

Given the previous literature that has established productivity as a clinically impactful measure for young children speaking English, the current study expands on this by examining Spanish–English bilingual children's productivity in Spanish. This study expands upon findings that suggest that both noun and verb phrases are developmentally and clinically impactful for Spanish-speaking children by focusing on children's productions of noun morphology as an initial investigation into accurately characterizing children's productive use of grammatical structures. This study focuses on articles and direct object clitics for several different reasons. First, these grammatical structures are among the first grammatical structures to appear in children's language in Spanish (Baron et al., 2018). Second, these grammatical structures are highly frequent in young children's language (Bedore & Leonard, 2001; Restrepo & Gutiérrez-Clellen, 2001), and previous work has suggested that these grammatical structures demonstrate the greatest group differences between TD children and their peers with DLD when using accuracy measures (Bedore & Leonard, 2005). Finally, investigations of children's accuracy of Spanish grammatical structures have primarily focused on samples of children ranging from preschool to school aged. Thus, questions remain on the performance of younger children who demonstrate emerging language skills. The current study will first establish patterns of grammatical productivity in bilingual children's noun morphology as a critical first step. Specifically, we examine children's productivity of definite articles (i.e., el, la, los, las), indefinite articles (i.e., un, una, unos, unas), and direct object clitics (i.e., lo, la, los, las), as previous research has suggested that these grammatical features yield group differences between TD children and children with DLD (Castilla-Earls et al., 2021). We examine the productivity of these grammatical features in two groups of children, those with and without DLD. We ask the following:

1. Does Spanish–English bilingual children's Spanish grammatical productivity …

   • a.correlate to their Spanish language sample measures (MLUw and NDW)?
   • b.yield group differences between TD children and children with DLD?

2. Does Spanish–English bilingual children's English grammatical productivity …

   • a.correlate to their English language sample measures (MLUw and NDW)?
   • b.yield group differences between TD and children with DLD?

Method

Participants

The current study was part of a larger community-based study. This study was approved by the institutional review board of the authors' university (Project Number: HS-2022-0188). Written consent was obtained from all caregivers of children, and caregivers were informed of the voluntary nature of the study. Participants included 32 Spanish–English bilingual children attending a local preschool. A total of 17 girls were included in the study, 14 in the TD group and three in the DLD group. Children in this sample were identified as bilingual based on caregivers reporting either Spanish or Spanish and English as their home language(s). As part of the assessment battery, participants completed language samples in Spanish and English. In addition, caregivers completed questionnaires to provide demographic information, including their child's exposure to languages spoken at home, age, and maternal education. Children in this sample were 51.78 months old on average (SD = 4.16, range: 42–59) and were exposed to Spanish 71% of the day (SD = 17.75%, range: 40%–100%). Finally, teachers of the children in this study provided subjective ratings of children's language development across domains using the Inventory to Assess Language Knowledge (Peña et al., 2014). Children's demographic information is listed in Table 1 by language ability group.

Table 1.

Participant demographics.

Group	Age			Percent Spanish heard			BESA Language Index			Maternal education (years)			Teacher ratings
	M	SD	Range	M	SD	Range	M	SD	Range	M	SD	Range	M	SD	Range
TD (n = 23)	52.13	4.68	40–59	73.05	17.47	50–100	96.71*	10.46	76–114	10.86	2.68	6–16	4.44*	0.44	3.40–5
DLD (n = 9)	50.88	2.37	48–56	65.89	18.43	40–100	78.71*	9.92	64–93	11.22	4.09	6–16	3.16*	0.77	1.60–4.40

Note. Age is presented in months. “Percent Spanish heard” is the caregiver reported percentage of time the child hears Spanish in an average week. Bilingual English–Spanish Assessment (BESA) Language Index is the child's highest standard score between Spanish and English tests. Four data points are missing from BESA Language Index (two typically developing [TD] and two developmental language disorder [DLD]). One data point is missing in maternal education in the TD group. Two data points are missing in percent Spanish heard (two TD).

^* p < .05 (comparisons between TD and DLD groups).

Language ability groups were determined based on parent-reported language concern, teacher-reported language concern, and scores on the Bilingual English–Spanish Assessment (BESA), administered in both Spanish and English (Peña et al., 2014). TD children in this sample had at least two of the following: no parent-reported concern, no teacher-reported concern, and/or a BESA best language standard score within normal limits (using age-based cutoff scores). Children with DLD in this sample had at least two of the following: parent-reported concern, teacher-reported concern, and/or a best language standard score below the age-based cutoff score on the BESA. These inclusionary criteria were based on previous work that has underscored the importance of converging evidence—or using multiple assessment tools and sources—to characterize bilingual children's language abilities (Castilla-Earls et al., 2020). The BESA is a standardized assessment designed to assess the language abilities of Spanish–English bilingual children between the ages of 4;0 and 6;11. In the current study, the BESA was administered by trained research assistants, and different language subsections were administered on separate days. Group comparisons revealed that TD children demonstrated significantly higher BESA Language Index scores and teacher ratings.

Transcription and Coding Procedures

Language samples were collected in Spanish and English by trained undergraduate and graduate research assistants on separate days, with distinct materials for each language (cf. Escobedo et al., 2023; see the Appendix). Each language sample was elicited following a set protocol. In each language, the language samples contained a play-based portion and a picture description task. Play-based elicitation is specific to young children and a recommended elicitation method for preschool-aged children (Pezold et al., 2020). In the picture description task, the examiner provided a model story for one picture and children were asked to produce a unique story for two subsequent pictures (Arwood, 1985; see the Appendix). Narratives (including picture descriptions) are another recommended elicitation context that is appropriate for preschool-aged children (Pezold et al., 2020). All language samples were between 15 and 20 min in length. Digital audio recordings of the language samples were transcribed and coded using the Systemic Analysis of Language Transcripts (SALT; Miller & Iglesias, 2012) software and SALT conventions (e.g., exclusion of mazes and filler words from analysis). Research assistants fluent in both Spanish and English elicited, transcribed, and coded Spanish language samples. Research assistants fluent in English elicited, transcribed, and coded English language samples. Code-switching was marked in SALT with a word level code, and utterances with code-switching were not excluded from MLUw and NDW analyses.

Scoring

Spanish language samples were additionally coded for correct use, incorrect use, and omission of definite articles, indefinite articles, and direct object clitics. Aligning with the procedure outlined by Hadley and Short (2005), productive instances of these grammatical features were determined based on the production of these grammatical features in sufficiently unique contexts.

Children's use of these forms was analyzed in SALT. Word codes were used to identify incorrect use of these grammatical structures (i.e., gender and/or number errors, “las niño,” the [plural, feminine] boy [singular, masculine]) or omission of these features in obligatory contexts (i.e., “__ niño fue a su casa,” __ boy [masculine, singular] went to his house). Children's correct productions were then compiled, and a point was awarded to each correct and sufficiently unique use of definite articles, indefinite articles, and direct object clitics. For each of the three categories (definite article, indefinite article, and direct object clitics), children's maximum score was capped at 5, following the procedures outlined by Hadley and Short (2005). Thus, children could receive a score between 0 and 5 for their sufficiently unique productions of definite articles, indefinite articles, or direct object clitics. Thus, a child could receive a maximum score of 15 for their total productivity score in Spanish (see Table 2).

Table 2.

Examples of Spanish productivity scoring.

Definite articles
	Child 1	Child 2
Example utterances	El ranchero se fue La gallina se escondió El caballo tiene hambre	El niño tiene sueño El niño se va El niño no se fue
Productivity score:	3	1
Indefinite articles
	Child 1	Child 2
Example utterances	Quiero una fresa Un gato se atoró Unas niñas eran chiquitas Ayer vi una vaca	Un perro se fue Un perro se cayó Un niño trabajó
Productivity score:	4	2
Direct object clitics
	Child 1	Child 2
Example utterances	Dámelo La vi ayer Lo tiene en su mochila	Lo agarré Lo guardé
Productivity score:	3	2

Note. Bold = unique instance.

Definite Articles

For definite articles, sufficiently different contexts were unique nouns that followed articles. In Spanish, the four definite articles examined were “el,” “la,” “los,” and “las,” as these are productions of articles with masculine-singular, feminine-singular, masculine-plural, and feminine-plural nouns, respectively. As an example, if a child produced the following utterances: “el ranchero se fue,” the [masculine, singular] farmer [masculine, singular] left, “la gallina se escondió,” the [feminine, singular] hen [feminine, singular] hid, and “el caballo tiene hambre,” the [masculine, singular] horse [masculine, singular] is hungry, the child would be credited with two distinct productions of the definite article “el” and one distinct production of the definite article “la.” Conversely, if a child produced article–noun pairs “el niño tiene sueño,” the [masculine, singular] boy [masculine, singular] is sleepy, “el niño se va,” the [masculine, singular] boy [masculine, singular] leaves, and “el niño no se fue,” the [masculine, singular] boy [masculine, singular] did not leave, the child would be credited with a singular production of a unique article and noun for the definite article “el” (see Table 2). This process was repeated for the other definite articles. As productivity is a measure of children's sufficiently unique productions of grammatical features, diminutive and augmentative forms of nouns were not counted separately from their base forms (e.g., el niñito was not credited separately from el niño).

Indefinite Articles

Scoring of indefinite articles was similar. We examined children's productions of “un,” “una,” “unos,” and “unas,” which preceded masculine-singular, feminine-singular, masculine-plural, and feminine-plural nouns, respectively. As an example, if a child produced the following utterances: “Quiero una fresa,” I want a [feminine, singular] strawberry [feminine, singular], “un gato se atoró,” a [masculine, singular] cat [masculine, singular] got stuck, “Unas niñas eran chiquitas,” some [feminine, plural] girls [feminine, plural] were small, and “ayer vi una vaca,” yesterday, I saw a [feminine, singular] cow [feminine, singular], the child would be credited with four sufficiently unique productions of indefinite articles. Conversely, if a child produced the following utterances: “un perro se fue,” a [masculine, singular] dog [masculine, singular] left, “un perro se cayó,” a [masculine, singular] dog [masculine, singular] fell, and “un niño trabajó” a [masculine, singular] boy [masculine, singular] worked, they would be credited with two instances of unique productions—one for “un perro” and another for “un niño” (see Table 2).

Direct Object Clitics

Similar to articles, direct object clitics were examined in masculine-singular (lo), feminine-singular (la), masculine-plural (los), and feminine-plural (las) categories. Both proclitic (i.e., appearing before the verb) and enclitic (i.e., attached to the verb) productions were examined in this study. For example, if a child produced the following utterances: “dámelo,” give it [masculine, singular] to me, “la vi ayer,” I saw it [feminine, singular] yesterday, and “lo tiene en su mochila,” he has it [masculine, singular] in his backpack, the child would be credited with three different productions of direct object clitics. If a child produced the following utterances: “lo agarré,” I got it [masculine, singular], and “lo guardé,” I put it [masculine, singular] away, they would be credited with two sufficiently different productions of direct object clitics.

English language samples followed the procedure outlined by Hadley and Short (2005) and measured children's unique productions of auxiliary BE, copula BE, auxiliary DO, third-person singular –s, and past tense –ed. For English, children's productive use of verb morphology was examined. For example, a child who produced the utterances “the mom eats,” “the dad eats,” and “the boy sings” would be credited with two productive instances of the third-person singular –s, one point for “eats” and one point for “sings” (see Hadley & Short, 2005, for further details on English productivity). This resulted in a maximum total productivity score of 25 across all grammatical features, with each category having a maximum score of 5.

In addition to productivity, MLUw, NDW, number of total words (NTW), number of complete and intelligible utterances, and the number of utterances with code-switching were calculated using SALT (see Table 3). TD children demonstrated an average Spanish MLUw of 3.08 and average Spanish NDW of 119.04. TD children also demonstrated an average English MLUw of 3.09 and average English NDW of 137.09. For NTW, TD children demonstrated an average of 312.04 in Spanish and 374.13 in English. TD children produced an average of 102.22 complete and intelligible utterances in Spanish and 120 complete and intelligible utterances in English. Finally, TD children produced an average of 14.26 utterances with code-switching in Spanish and 6.74 utterances with code-switching in English. Children with DLD demonstrated an average Spanish MLUw of 1.98 and an average Spanish NDW of 62.11. Children with DLD also demonstrated an average English MLUw of 2.11 and an average English NDW of 68.89. For NTW, children with DLD demonstrated an average of 136.11 in Spanish and 156.89 in English. Children with DLD produced an average of 63.33 complete and intelligible utterances in Spanish and 71.44 complete and intelligible utterances in English. Finally, children with DLD produced an average of 7.22 utterances with code-switching in Spanish and six utterances with code-switching in English. Relative to their peers with DLD, independent-samples Mann–Whitney U tests demonstrated that TD children showed significantly higher Spanish MLUw (p < .001), Spanish NDW (p < .001), Spanish NTW (p = .006), Spanish complete and intelligible utterances (p < .017), English MLUw (p = .002), English NDW (p < .001), English NTW (p < .001), and English complete and intelligible utterances (p = .003). No significant differences were found for the number of codeswitches in English and Spanish language samples.

Table 3.

Language sample measures by group and language.

	MLUw				NDW				NTW				Complete and intelligible utterances				Utterances with code-switching
	Spanish		English		Spanish		English		Spanish		English		Spanish		English		Spanish		English
Group	M	SD	M	SD	M	SD	M	SD	M	SD	M	SD	M	SD	M	SD	M	SD	M	SD
TD (n = 23)	3.08*	0.66	3.09*	0.54	119.04*	41.36	137.09*	55.50	312.04*	167.30	374.13*	148.49	102.22*	42.88	120*	38.81	14.26	17.95	6.74	14.39
DLD (n = 9)	1.98*	0.43	2.11*	0.81	62.11*	31.20	68.89*	38.46	136.11*	106.96	156.89*	111.40	63.33*	35.34	71.44*	31.48	7.22	5.93	6	11.28

Note. MLUw = mean length of utterance in words; NDW = number of different words; NTW = number of total words; TD = typically developing; DLD = developmental language disorder.

^* p < .05 (comparisons between TD and DLD groups).

Reliability

After a research assistant independently transcribed a language sample, a different research assistant reviewed the transcript while listening to the language sample audio file. A consensus procedure was utilized to resolve discrepancies. After language samples were transcribed, a research assistant independently coded bound and unbound morphemes. A second research assistant reviewed the coded transcript, and discrepancies were resolved via consensus. Productivity was calculated from these completed language samples. After productivity was calculated by a research assistant, a second research assistant independently scored productivity for 20% of samples. Reliability was calculated by comparing independent rater's productivity scores and deriving a percentage. For Spanish productivity, interrater reliability was 92% on average. For English productivity, interrater reliability was 94% on average. The first author resolved any discrepancies in both English and Spanish language samples.

Analyses

We first examined patterns at the group level to fully describe children's Spanish productivity, including analysis of group differences using the nonparametric Mann–Whitney U test because of differences in n between the TD and DLD groups. To compare the Spanish productivity measure to other language sample measures, Pearson's correlations between children's Spanish productivity score and their Spanish MLUw and NDW were conducted. We conclude with an examination of children's English productivity scores, also examining group differences using the nonparametric Mann–Whitney U test and Pearson's correlations between English productivity and English MLUw and NDW.

Results

To describe patterns observed in the Spanish language samples of Spanish–English bilingual children, we first examined children's Spanish productivity across grammatical categories. TD children showed statistically higher productivity in all three grammatical targets than children with DLD (see Table 4). Additionally, both groups demonstrated a wide range of productivity in each grammatical category, ranging from 0 to 5. Children's productivity in definite articles, indefinite articles, and direct object clitics was summed to calculate a total productivity score for each child. Average total productivity scores were 9.91 (SD = 4.68) for TD children and 4.89 (SD = 4.19) for children with DLD.

Table 4.

Spanish productivity scores across categories.

Group	Definite articles			Indefinite articles			Direct object clitics			Total productivity
	M	SD	Range	M	SD	Range	M	SD	Range	M	SD	Range
TD (n = 23)	4.30*	1.63	0–5	3.13*	2.01	0–5	2.48*	1.97	0–5	9.91*	4.68	0–15
DLD (n = 9)	2.78*	1.92	0–5	1.44*	1.81	0–5	0.67*	1.00	0–3	4.89*	4.19	0–12

Note. The max score for each category (definite articles, indefinite articles, and direct object clitics) was 5 and was summed to result in a max score of 15 for total productivity. TD = typically developing; DLD = developmental language disorder.

^* p < .05 (comparisons between TD and DLD groups).

Correlations Between Spanish Productivity and Other Language Sample Measures

To determine if bilingual children's productive use of Spanish grammatical structures was related to other established measures of LSA, we examined correlations between Spanish total productivity scores and Spanish MLUw and Spanish NDW. Results revealed strong positive correlations between Spanish productivity and other established language sample measures. Spanish productivity was significantly and positively correlated with both MLUw (r = .64, p < .001) and NDW (r = .66, p < .001). Figures 1 and 2 illustrate the relationship between Spanish productivity and these language sample measures.

Figure 1.

Correlation between Spanish productivity and Spanish mean length of utterance in words (MLUw). TD = typically developing; DLD = developmental language disorder.

Figure 2.

Correlation between Spanish productivity and Spanish number of different words (NDW). TD = typically developing; DLD = developmental language disorder.

Spanish Productivity Across Groups

Our results revealed that there were differences in the distribution of Spanish productivity between the TD and DLD groups. Adding across all grammatical categories, the average productivity in Spanish was 9.91 (SD = 4.68) for TD children. For children in the DLD group, the average productivity was 4.89 (SD = 4.19). In other words, TD children demonstrated on average 9.91 productive instances of definite articles, indefinite articles, and direct object clitics in their Spanish language samples. Children with DLD demonstrated an average of 4.89 productive instances of definite articles, indefinite articles, and direct object clitics across their Spanish language samples. Summaries of average productivity by group are available in Table 4. Additionally, we examined the group differences between TD children and DLD children using the nonparametric Mann–Whitney U test. Results from this analysis showed that children in the TD group (mean rank = 19.11) produced significantly more productive instances in Spanish than did their peers in the DLD group (mean rank = 9.83), p = .010.

Correlations Between English Productivity and Other Language Sample Measures

Similar to the correlational analyses examining the relationship between productivity and other language sample measures in Spanish, results revealed strong positive correlations between English productivity and other English language sample measures. English productivity was significantly correlated with both MLUw (r = .75, p < .001) and NDW (r = .69, p < .001). Figures 3 and 4 illustrate the relationship between English productivity and these language sample measures.

Figure 3.

Correlation between English productivity and English mean length of utterance in words (MLUw). TD = typically developing; DLD = developmental language disorder.

Figure 4.

Correlation between English productivity and English number of different words (NDW). TD = typically developing; DLD = developmental language disorder.

English Productivity Across Groups

Our results revealed that there were differences in the distribution of English productivity between the TD and DLD groups. As expected, TD children's average productivity per grammatical structure was higher than it was for the children in the DLD group (see Table 5). However, no statistically significant differences were found in individual grammatical categories. Children's productive use of past tense –ed, third-person –s, auxiliary BE, auxiliary DO, and copula BE were summed to calculate a total productivity score for each child. The average total productivity in English was 6.65 (SD = 4.98) for TD children. For children in the DLD group, the average total productivity was 2.22 (SD = 2.28). Average total English productivity by group is provided in Table 5. Additionally, we examined the group differences for total productivity between TD children and children with DLD using the nonparametric Mann–Whitney U test. Results from this analysis showed that children in the TD group (mean rank = 18.85) produced significantly more productive instances in English than did their peers in the DLD group (mean rank = 10.50), p = .013.

Table 5.

English productivity scores across categories.

Group	Past tense –ed			Third-person –s			Auxiliary BE			Copula BE			Auxiliary DO			Total productivity
	M	SD	Rge	M	SD	Rge	M	SD	Rge	M	SD	Rge	M	SD	Rge	M	SD	Rge
TD (n = 23)	0.87	1.06	0–4	1.17	1.55	0–5	1.65	1.75	0–5	2.04	1.41	0–5	1.00	1.41	0–5	*6.65	4.98	0–15
DLD (n = 9)	0.56	0.88	0–2	0	0	0	0.22	0.44	0–1	0.89	0.44	0–1	0.22	0.44	0–1	*2.22	2.28	0–5

Note. Maximum English productivity score = 25. Rge = range; TD = typically developing; DLD = developmental language disorder.

^* p < .05 (comparisons between TD and DLD groups).

Discussion

The aim of the present study was to capture Spanish–English bilingual children's productive use of Spanish and English grammatical markers. Our study built on previous work by piloting a measure of Spanish grammatical productivity that measured children's unique productions of definite articles, indefinite articles, and direct object clitics. In addition to verb morphology, these Spanish grammatical structures have previously been identified as clinically significant grammatical markers for Spanish-speaking children (Castilla-Earls et al., 2021). To our knowledge, this study is the first to examine the utility of a Spanish productivity measure for Spanish–English bilingual children.

Spanish Productivity

The results from the current study suggest that the Spanish–English bilingual children in this study indeed produced productive instances of definite articles, indefinite articles, and direct object clitics. Furthermore, group differences favoring TD over DLD were observed. These results align with previous work that has examined the clinical utility of the same markers to differentiate between TD Spanish-speaking children and their peers with DLD (Castilla-Earls et al., 2021). Findings from previous research demonstrate that children with DLD are less accurate in their productions of these grammatical markers than those of their TD counterparts in elicitation tasks. Our study contributes to the literature on the clinical utility of these grammatical markers by examining children's productive use of them in spontaneous language samples. We hypothesized that TD children would demonstrate more productive uses of these grammatical markers than would children with DLD. Our results suggest that TD children indeed demonstrate more productive uses of definite articles, indefinite articles, and direct object clitics than do children with DLD. Our results suggest that children used these grammatical markers in a variety of contexts and were able to generate various noun–object combinations and utterances that contained direct object clitics. This pattern was also observed when collapsing across grammatical categories, as the average Spanish productivity score in the TD group was higher than that of the DLD group.

As this measure is intended to be used in LSA, our first research question examined the relationship between this Spanish productivity measure and other established language sample measures. We hypothesized that children's Spanish productivity would be highly correlated with children's Spanish MLUw and Spanish NDW. Our results suggest that children's Spanish productivity and their other language sample measures in Spanish are indeed correlated. As expected, children who demonstrated high productivity in Spanish also demonstrated longer utterances and a higher NDW in their language samples. These results suggest that Spanish productivity captures information on children's emerging language skills in Spanish (i.e., depth and breadth of production of early developing grammatical structures).

English Productivity

The children in this sample were identified as Spanish–English bilingual children, as they were children whose caregivers reported Spanish or Spanish and English as their home languages. Thus, in order to capture children's bilingual skills, this study examined children's productive use of English grammatical markers. First, we hypothesized that TD children would outperform their peers with DLD in their productive use of English grammatical markers, as has been found in previous research. Our results suggest that TD children indeed produced more productive instances of English grammatical markers when compared to children with DLD. Importantly, these results align with findings from previous studies from both English-speaking monolingual populations (Gladfelter & Leonard, 2013) and Spanish–English bilingual children tested in English (Potapova et al., 2018). Thus, our results corroborate these findings and suggest that English productivity scores from language samples result in differences between TD children and children with DLD. In addition, children's English scores were significantly correlated with their average MLUw and NDW, mirroring the results of Spanish productivity.

Clinical Implications

Results from both English and Spanish suggest that measures of grammatical productivity hold potential clinical utility for differentiating between typical development and DLD in Spanish–English bilingual children. Results from English-speaking monolingual children have demonstrated that children with language disorders demonstrate limited productive uses of English grammatical structures (Gladfelter & Leonard, 2013), and the children in this sample demonstrated a similar pattern. Importantly, children's Spanish productivity also demonstrated the same pattern, suggesting that children with DLD produce a relatively limited number of unique definite articles, indefinite articles, and direct object clitics. English and Spanish productivity scores have clinical promise for characterizing children's language abilities in both of their languages, which is a critical aspect of evidence-based practice. As best practice for bilingual language assessment includes assessment in both of a child's languages, productivity scores present a unique opportunity to characterize children's language abilities with a parallel measure across both of their languages. This dual language assessment holds potential to inform assessment that will ultimately support both the L1 and L2. More work is needed to fully characterize bilingual children's productive use of Spanish grammatical markers in both typical and atypical language development, including future analysis of sensitivity and specificity of this measure. Future work will investigate how productivity scores in two languages can be used together to inform assessment for Spanish–English bilingual children.

Limitations

The results of the current study were based on a small sample of Spanish–English bilingual children. Recognizing that bilingualism is a spectrum, future research should consider recruiting a larger sample of bilingual children in order to capture the language experiences of a range of bilingual children, including those with language disorders. One important variable to consider in future work is the effect of children's amount of language exposure, as previous work has determined that exposure has an effect on children's performance on language tasks (Pratt et al., 2021). For productivity, exposure may affect bilingual children's production of grammatical structures included in productivity analyses in both English and Spanish. Additionally, other factors such as age and age of acquisition have consistently demonstrated effects on bilingual children's language performance (Paradis, 2023). This is pertinent to productivity-based scoring approaches, which may hold potential clinical utility in processes such as progress monitoring. Additionally, recent work has suggested that even within DLD groups, children demonstrate a range of language abilities and, thus, DLD should be viewed as a spectrum (Lancaster & Camarata, 2019). From an assessment perspective, this suggests that measures such as Spanish productivity must take the range of language abilities that children with DLD exhibit into consideration. Additionally, previous work that has identified clinically relevant grammatical markers in Spanish-speaking children has suggested that morphosyntactic features of both noun and verb phrases are critical for characterizing young children's language abilities. The current study limited analysis to children's production of articles and direct object clitics. Future research should extend this measure of capturing children's productive use of Spanish grammatical markers to verb morphology.

Additionally, previous work has noted that language sampling is often viewed as time consuming or inaccessible by school-based SLPs (Pavelko et al., 2016). Nevertheless, language sampling offers valuable data in the assessment of bilingual children. Heilmann (2010) suggests that even though the collection and transcription of language samples may seem time intensive, this process is no less time intensive than administering, scoring, and interpreting a standardized assessment. In terms of accessibility, open-source software such as Sampling Utterances and Grammatical Analysis Revised may be a useful resource for clinicians. Additionally, LSA is well established as a measure that limits bias in the assessment of Spanish–English bilingual children (Castilla-Earls et al., 2020; Gutiérrez-Clellen & Simon-Cereijido, 2009; Ortiz et al., 2024). Future research should examine how productivity measures can be used with methods other than language sampling, such as probe tasks and/or in real time.

Conclusions

The current study aimed to examine young Spanish–English bilingual children's productive use of Spanish (L1) grammatical structures and its relationship with other language sample measures. The results of this study demonstrated that children's Spanish grammatical productivity yielded group differences between TD children and children with DLD. Additionally, this measure was positively correlated with other language sample measures. Finally, the results of this study corroborated previous findings that demonstrated that children's English productivity also yields group differences between TD children and children with DLD. Overall, productivity is a clinically useful tool that holds potential to capture children's emerging language skills, which could involve collaboration with other professionals and children's caregivers. Additionally, productivity highlights children's strengths, as it focuses on their productions of grammatical forms rather than their errors. Ultimately, this study contributes to the literature by piloting this measure with Spanish–English bilingual children and serves as a first step toward using this measure to improve bilingual language assessment practices.

Data Availability Statement

The data generated and/or analyzed during the current study are available from the corresponding author on reasonable request.

Appendix

Spanish and English Language Samples

Spanish language sample: play sample and picture description	English language sample: play sample and picture description
Spanish play sample (examples)	English play sample (examples)
Set 1: Toy barn Toy farm animals Toy farmers × 2 Set 2: Plastic breakfast foods and utensils Toy family Play prompts: Examiners were asked to follow the child's lead, use self-talk strategies, and prompts to elicit contextualized and decontextualized language. Example prompts: ¿Qué estará haciendo el granjero? Dime qué va a hacer la familia de comer. Creo que los animales tienen hambre.	Set 1: Toy garage/gas station Toy car × 2 Toy drivers × 2 Set 2: Plastic lunch foods and utensils Toy family Play prompts: Examiners were asked to follow the child's lead, use self-talk strategies, and prompts to elicit contextualized and decontextualized language. Example prompts: I wonder what the driver is doing with … Tell me what the family is going to make for breakfast. I think the baby is hungry.
Spanish Picture Description (Arwood, 1985)	English Picture Description (Arwood, 1985)
Family with cat in a tree (modeled by examiner, see below) Family at the mall Children pulling wagon Story model (example): Había una vez una niñita que se llamaba Sara. A Sara siempre le encantaba jugar afuera después de la escuela. Le gustaba correr y jugar con su pelota. Su hermano mayor, Juan, siempre le decía que jugara on cuidado. Pues un día Sara quería a alguien con quien jugar. Entonces sacó a sus tres gatitos. Mira, un gatito se subió al árbol y se atoró allí arriba. Juan salió y proclamó a sara “te dije que jugaras con cuidado” Sara era muy pequeña y no pudo subir al árbol. Entonces llegó el vecino, el señor Tomás a ayudarlos. Él les dijo “yo tengo una escalera y yo puedo subir al árbol.” Entonces se subió hasta la cima del árbol y rescató al gatito. Sara estaba tan contenta y dijo “muchas gracias señor Tomás seré más cuidadosa la próxima vez.” Entonces Sara nunca volvió a sacar a sus gatitos. Fin. Story retell prompts: The examiner begins with “Te voy a contar un cuento de este dibujo. Después, tú me cuentas un cuento de estos dibujos” and then models a story. Example prompts: Dime qué pasó aquí. ¿Qué pasó cuando … ? ¿Qué más pasó?	Family at grocery store (modeled by examiner, see below) Family washing a car Children playing basketball Story model (example): Once upon a time there were two little boys named Billy and Sam. Billy and Sam liked to go shopping with their mom at the grocery store. One day at the store, they saw a big, tall stack of soda cans. Billy and Sam thought “I sure would like a soda!” Billy said to Sam “I can't reach the ones at the top. Let's take some sodas from the bottom of the stack.” So Billy and Sam pulled out some cans of soda from the bottom of the stack. As soon as the boys pulled out their sodas, the entire stack fell down! Mr. Henry, who owned the store, was very angry and said “you boys ruined my stack of sodas!” The boys said “we're really sorry Mr. Henry, we'll help you clean up the cans.” So Billy and Sam helped Mr. Henry stack up all the sodas again and then he wasn't angry anymore. Mr. Henry said “I'll see you tomorrow. Let's be more careful next time.” The end. Story retell prompts: The examiner begins with “I'm going to tell you a story about this picture. When I'm done, I want you to tell me a story about these” and then models a story. Example prompts: Tell me what's happened over here. I wonder what happened when … What else happened?

Funding Statement

The current study was supported in part by an F31 grant from Eunice Kennedy Shriver National Institute of Child Health and Human Development (F31HD111303, Principal Investigator: Alicia G. Escobedo). Additional research support was provided through a Price Philanthropies community grant awarded to Sonja Pruitt-Lord.