Introduction
While there has been extensive research on adults’ sociolinguistic variation (Labov, Reference Labov2001), children's sociolinguistic variation is less well-understood; this is particularly the case for sociolinguistic variation involving morphosyntax. Thus far, the few studies investigating children's acquisition of morphosyntactic variation have yielded conflicting results. Some studies suggest that children's variable morphosyntactic patterns match those of their caregivers by age three (e.g., Smith & Durham, Reference Smith and Durham2019), whereas others suggest that acquisition of morphosyntactic variation can take a long time to emerge during childhood (e.g., Shin, Reference Shin2016). This developmental trajectory and the age at which children master a structure varies depending on the structure's complexity and its frequency in the input in addition to its perceived prestige in the community. For example, Smith and Durham (Reference Smith and Durham2019) note that caregivers are reluctant to use stigmatized variants with young children, and instead opt for the socially recognized standard form in Child Directed Speech (CDS). As such, stigmatized morphosyntactic variants may develop later in children's output in comparison to other less salient forms.
This brief research report contributes to the growing field of developmental sociolinguistics by zeroing in on a sociolinguistic variable in Spanish that is well-studied and well-understood among adults, but not yet well-understood in child language. More specifically, the current study investigates Spanish children's variation between the second person singular (2sg) preterit standard forms (e.g., caíste ‘you fell’) and the non-standard forms, which include word-final -s (e.g., caístes). To do so, preterit forms were extracted from corpora of spontaneous speech produced by 78 children in Spain and analyzed for the effects of age, lexical frequency, and the possible influence from Catalan or Galician.
Background
In Spanish, the perfective past tense is often called the ‘preterit’. The standard 2sg preterit inflection is –ste. As observable in Table 1, 2nd personal singular forms generally include the 2sg word-final –s morpheme, making this paradigm different from all other 2sg verbal paradigms and thus an irregular pattern. Moreover, Miller and Schmitt (Reference Miller and Schmitt2012) suggest that Spanish-speaking children associate the 2sg word-final –s morpheme to 2sg subjects by age 3;00.
2sg verb conjugations in simple tenses
The non-standard forms, which include word-final –s (e.g., hablastes, comistes, vivistes), are an example of analogical leveling; speakers regularize forms so that they follow the dominant pattern of the language (Bybee, Reference Bybee2015; Penny, Reference Penny2000). While several studies show that Spanish-speaking adults vary between standard 2sg preterit forms, like caíste, and non-standard forms, like caístes, their use of the non-standard is constrained by social factors such as socioeconomic status and education level. Importantly, non-standard preterit –s is a stigmatized variant in many speech communities (Lipski, Reference Lipski2014; Tinoco, Reference Tinoco2017). For example, Barnes’ (Reference Barnes2012) study, which draws on data from sociolinguistic interviews, television broadcasts, and public events, highlights the importance of speech register, with the non-standard showing a strong correlation with informal speech. This finding aligns with the general association found between informal speech and less-prestigious variants (Smith & Durham, Reference Smith and Durham2019).
Variation between standard and non-standard -s is also influenced by lexical frequency. Both Barnes (Reference Barnes2012) and Escalante (Reference Escalante2015) find that non-standard –s is more prevalent among low-frequency (23%, 48%, respectively) than high-frequency verbs (10%, 8%, respectively) in adult Spanish. Finally, previous studies of adults’ 2sg preterit forms demonstrate varying rates of the non-standard –s across Spanish dialects with its acceptability relating to the speakers’ variety (Aleza & Enguita, Reference Aleza and Enguita2010; Lipski, Reference Lipski2014). While Barnes (Reference Barnes2012) finds rates as low as 13% in an adult Spanish population, Lerner's (Reference Lerner2016) adult Mexican population in the US produces non-standard –s at a rate of 33%. These varying rates will likely impact children's production of the non-standard -s. Indeed, previous research on developmental sociolinguistics suggests that children are more likely to show variation in their output when exposed to it in their input (Smith & Durham, Reference Smith and Durham2019; Miller, Reference Miller2013).
In addition to dialect variation, an open question is whether language contact affects speaker's variation between standard and non-standard –s. Previous research highlights the importance of crosslinguistic influence (CLI) in language contact settings noting that variable morphosyntactic structures are more susceptible to CLI (Silva-Corvalán, Reference Silva-Corvalán2014). CLI can manifest in the form of accelerated or delayed acquisition of a construction or the transfer of an element from one language to another (Paradis & Genesee, Reference Paradis and Genesee1996). Thus, while Barnes (Reference Barnes2012) notes low percentages of the non-standard –s in Peninsular Spanish, Spanish speakers in Galicia and Catalonia may produce more non-standard 2sg preterit forms due to contact with Galician and Catalan, two languages in which the standard 2sg preterit forms include word final -s.
To date, there has been one study of Spanish-speaking children's variation between standard and non-standard -s. In her study of Mexican children from three to eight-years-old, Lerner (Reference Lerner2016) finds that children use the non-standard variant at a rate (61%) that exceeds that of their caregivers. She also finds that both children and adults alike produce higher rates of the non-standard in informal than formal speech. Lerner observes an age effect whereby older children produce more non-standard –s than younger children. These results are consistent with previous research in developmental sociolinguistics showing children use increasingly more non-standard forms as they get older, a finding that has been explained as the result of caregivers’ tendency to use fewer non-standard forms with younger children and more with older children and adults (Smith & Durham, Reference Smith and Durham2019). The increased use of non-standard forms with age could also be the result of increased exposure to more verb forms, including low-frequency forms, as children age. While Lerner (Reference Lerner2016) studies the factors influencing children's 2sg –s variation, she does not analyze the effect of lexical frequency, which is known to condition adults’ variable 2sg preterit production in Spanish. Further, lexical frequency is known to impact child language: children overregularize irregular forms that are low-frequency, whereas high-frequency forms are less likely to undergo overregularization (Clahsen, Aveledo & Roca, Reference Clahsen, Aveledo and Roca2002; Fernández-Dobao and Herschensohn, Reference Fernández-Dobao and Herschensohn2020; Johnson, Reference Johnson and Clark1995). How does lexical frequency impact children's use of the non-standard preterit –s in Spanish? To date, there is no study that has examined this question.
The current study aims to further our understanding of children's acquisition of sociolinguistic variation by analyzing Spanish children's 2sg preterit verb forms and the effects of age, contact setting, and lexical frequency. It is predicted that children, like adults, will produce both ‘caiste’ and ‘caistes’, and that their variation will be impacted by age, language contact, and lexical frequency.
Research questions and hypotheses
The following research questions and their corresponding hypotheses were developed based on previous research:
1) Does age affect children's variation between standard and non-standard 2sg preterit forms?
H1a) Based on Lerner's (Reference Lerner2016) findings that older Spanish-speaking children produce more non-standard –s than younger children, as well as Smith and Durham's (Reference Smith and Durham2019) finding that caregivers use more non-standard forms in CDS as children age, it is predicted that older children will produce higher rates of non-standard –s.
H1b) Based on findings that Spanish-speaking children associate the 2sg word-final –s to 2nd person subjects as early as 3;0 (Miller & Schmitt, Reference Miller and Schmitt2012), it is predicted that younger children will overregularize the preterit forms and produce more non-standard –s.
2) Do Spanish children in language contact settings produce more non-standard –s than those in non-contact settings?
H2a) If language contact influences 2sg preterit variation as has been observed in other variable morphosyntactic structures (Silva-Corvalán, Reference Silva-Corvalán2014), children in Galicia and Catalonia are predicted to transfer the Galician and Catalan 2sg –s morpheme and thus produce higher rates of the non-standard forms in Spanish.
H2b) Given the overall low rates of non-standard –s in Peninsular Spanish (Barnes, Reference Barnes2012), all Peninsular speakers are expected to produce low rates of non-standard –s.
3) Does lexical frequency impact children's production of non-standard –s?
H3a) It is predicted that children will produce non-standard –s more often with low-frequency verbs. This outcome could reflect patterns found in the input since adults produce more non-standard preterit –s with low-frequency verbs (Barnes, Reference Barnes2012; Escalante, Reference Escalante2015) or could be the result of children's own experience with verb forms since children tend to overregularize more with low-frequency verbs (Clahsen et al., Reference Clahsen, Aveledo and Roca2002; Johnson, Reference Johnson and Clark1995).
H3b) A strong tendency to overregularize could override children's lexical learning of frequent forms, in which case children will produce non-standard –s on both high and low-frequency verbs.
Methodology
Participants and data extraction
Data for this study come from 12 Spanish-language corpora from the Child Language Data Exchange System (CHILDES) Talkbank database (MacWhinney, Reference MacWhinney2000). A total of 78 Peninsular Spanish speakers between the ages of 1;10 and 12;4 years were included (see Table 2).
Number of children and tokens of 2sg preterit forms per CHILDES corpus
Note: BecaCESNo = CHILDES Databate BecaCESNO Corpus (no date), Koine = Fernández Pérez (Reference Fox and Weisberg2015), DiezItza = Diez-Itza (Reference Diez-Itza, Ruiz, Sheerin and González-Cascos1995), FernAguado = CHILDES Database FernAguado Corpus (no date), FerFuLice = Liceras et al. (Reference Liceras, Fuertes, Perales, Pérez-Tattam and Spradlin2008), LlinasOjea = CHILDES Database LlinasOjea Corpus (no date), Marrero = Cappelli et al. (Reference Cappelli, Marrero-Aguiar and Albalá1994), OreaPine = Aguado-Orea & Pine (Reference Aguado-Orea and Pine2015).
In the ‘other’ category, four corpora were grouped together due to their low token count of 2sg preterit forms: Aguirre (Reference Aguirre2000; Madrid), Ornat (Reference López Ornat1994; Madrid), Linaza (Reference Linaza, Sebastián and del Barrio1981; Madrid), and Vila (Reference Vila1990; Catalonia). The 12 corpora utilized included longitudinal and cross-sectional data consisting of naturalistic speech between children and caretakers/researchers, child play, and more formal production tasks such as narrating a story. All tokens of standard and non-standard 2sg preterit forms (N = 230) were extracted manually from the CHILDES transcripts. Tokens followed by an alveolar fricative must be excluded from studies of word-final -s, as this phonological context would make it difficult to distinguish between the 2sg preterit –s and the following consonant. However, there were no such contexts in the current data set.
Coding
All 2sg preterit tokens were coded for whether the form was the standard form without –s (1) or the non-standard form with –s (2).
(1) Me metiste aquí dentro. (Alejandro, 4;00; Corpus: FernAgudo)
‘You put me here inside.’
(2) ¡Me pillastes! (Rosa, 7;11; Corpus: BecaCESNo)
‘You found me!’
Furthermore, in order to address the research questions, all tokens were coded for (i) the participant's age, (ii) whether the children were in a language contact setting, and (iii) lexical frequency.
Age
Children ranged from 1;10 to 12;4 years. To address possible age effects on 2sg preterit production, children were split into three groups:
• 1;10-3;11 (N = 46)
• 4;0-6;7 (N = 24)
• 7;1-12;4 (N = 12)
Contact setting
To address language contact effects, children were categorized in two groups: 1) contact and 2) non-contact. Children in the contact group (N = 25) came from Galicia and Catalonia, while the non-contact children (N = 53) came from Madrid, Navarre, Castile-Leon, The Canary Islands, and Asturias. Children were grouped in this manner because both Galician and Catalan, which are official languages alongside Spanish in their respective provinces, mark –s on 2sg preterit forms.
Lexical frequency
Lexical frequency was coded for each 2sg preterit verb form produced by the children by obtaining frequency counts for each token's standard preterit form (-ste) in the Davies’ (Reference Davies2018) Corpus del Español NOW. The most frequent form was viste ‘you saw’ with a total of 46,505 examples in Davies’ corpus. The second most frequent form was fuiste ‘you went’/‘you were’ with 19,870 tokens. There was one token extracted from the CHILDES corpora that did not appear at all in the Davies’ corpus: mancaste ‘you maimed’. The Davies’ corpus is updated with new data monthly; as such, the frequency counts increase quite steadily. Therefore, the coding for frequency was completed within one month to gain an accurate comparison of frequency categories. Following Barnes (Reference Barnes2012), frequency was split into three categories: low, mid, and high (Table 3).
Frequency categories and examples with frequency counts from Davies’ corpus
Low-frequency verbs accounted for 55% (N = 126) of the overall tokens, with mid and high together accounting for the remaining 45% (12%, 33%, respectively). The following examples show children's production of a low-frequency verb (5), a mid-frequency verb (6), and a high-frequency verb (7).
(5) ¿Y con qué la pintastes (.) con boli? (Raquel, f, 3;06; Corpus: BecaCESNo)
‘What did you paint it with (.) with a pen?’
(6) No me lo distes. (Alejandro, m, 4;00; Corpus: FernAguado)
‘You didn't give it to me.’
(7) No viste el otro día esa película de…(Sandra, f, 6;05; Corpus: BecaCESNo)
‘You didn't see that movie the other day about…’
Results
Overall distribution
The overall rate of non-standard was 17% (N = 38) out of a total of 230 2sg preterit forms. Thus, while the non-standard forms were indeed present, the sampled children produced the standard form more frequently. Table 4 demonstrates the children's distribution according to their age group and contact setting and their rates of non-standard –s. Meanwhile, Table 5 presents the distribution of tokens across frequency categories
Distribution of children classified by contact setting and age
Note. While the total number of participants in the current study is 78, there are N = 82 children accounted for in Table 5 as four children produced tokens at different periods spanning the two youngest age groups.
Distribution of tokens across frequency categories
Table 4 demonstrates varying non-standard rates across age groups, with the oldest children in non-contact settings producing an increased rate of non-standard –s (31%). Further, children in the two younger age groups residing in contact areas have comparable non-standard rates (28%, 27%,). At the same time, Table 4 displays a high degree of variance especially within categories with low 2sg preterit token counts (e.g., Contact: 4;0-6;7; Non-contact: 7;1-12;4). Lastly, Table 5 shows a frequency effect whereby low-frequency verbs are produced as the non-standard variant most often, whereas high-frequency verbs almost always occur in their standard form.
To investigate whether the children's variable 2sg preterit production is significantly influenced by (i) age, (ii) contact setting, and (iii) lexical frequency, a mixed-effects binary logistic regression was run in R (R Core Team, 2019) using the glmer function in the lme4 package (Bates, Mächler, Bolker & Walker, Reference Bates, Mächler, Bolker and Walker2015). The regression analysis included data from all 78 children with non-standard vs. standard 2sg preterit forms as the dependent variable with the model set to predict non-standard. The final mixed-effects model (Table 6), which included participant as a random factor to ensure generalizability beyond individual, included the following categorical independent variables: age group, contact setting, and lexical frequency. Multicollinearity was investigated using the vif function in the car package in R (Fox & Weisberg, Reference Fox and Weisberg2019), and revealed no issues; Variance Inflation Factors (VIFs) ranged from 1.00 to 1.09. Further, there were no interactions found between predictor variables – thus, the final model presents main effects only. Reference levels for predictor variables are labeled “Ref”.
Mixed Effects logistic regression analysis, 78 participants, AIC = 191.4
Results from the mixed-effects analysis (Table 6) show that contact setting and lexical frequency significantly predict non-standard –s with children residing in contact areas and low-frequency verbs increasing the likelihood of producing the non-standard –s. Contrarily, there is no significant effect of age on children's 2sg preterit forms.
Discussion
Age results
The first research question explored the relationship between age and children's variable 2sg preterit production. Based on Lerner's (Reference Lerner2016) results and caregiver's tendency to use fewer non-standard forms with younger children (Smith & Durham, Reference Smith and Durham2019), H1a predicted that older children would produce the non-standard more than younger children while H1b predicted that as Spanish-speaking children associate 2sg word-final –s with 2sg subjects as early as 3;0 (Miller & Schmitt, Reference Miller and Schmitt2012), young children would overregularize and produce more non-standard preterit –s. A mixed-effects regression analysis (Table 6) found no significant age effect suggesting that age is not a significant predictor of Spanish children's 2sg preterit variation. However, there were age groups with fewer data than others (see Table 4). For example, the oldest age group included 12 children all from non-contact settings. While this age group was underrepresented in the current study, they notably produced the highest rates of non-standard (31%).
Thus, it is possible that with more data from older participants, any age effects would become more discernible. Indeed, there are reasons to believe that as children grow older, they are exposed to more non-standard –s in the input as adults tend to use more non-standard forms with older children, which is especially true for stigmatized variants (Smith & Durham, Reference Smith and Durham2019). Furthermore, Lerner's (Reference Lerner2016) results demonstrate a developmental trajectory in this same direction with the older children producing more non-standard forms. Therefore, while Spanish children acquire imperfect and preterit tense forms at an early age (Pérez-Pereira, Reference Pérez-Pereira1989), these results could signal a later developmental trajectory of 2sg preterit variation in relation to other sociolinguistic variants.
Contact results
The second research question explored the relationship between contact setting and children's 2sg preterit forms. Hypothesis 2a predicted that if 2sg preterit variation was influenced by language contact as has been observed in other variable morphosyntactic structures (Silva-Corvalán, Reference Silva-Corvalán2014), children in contact settings would transfer the 2sg –s morpheme and produce higher rates of the non-standard forms in Spanish since 2sg preterit forms in Galician and Catalan end in –s. Contrarily, H2b predicted that all Peninsular speakers would produce low rates of the non-standard given the overall low rates in Peninsular Spanish: while Barnes (Reference Barnes2012) found relatively low rates of non-standard -s in Spanish adults (13%), the results in Table 6 demonstrate that Spanish children in contact with Galician and Catalan produce the non-standard –s proportionately more than children in non-contact areas (e.g., Madrid), which corroborates H2a.
These preliminary findings highlight the importance of language contact on the production of 2sg preterit forms in Spanish. Although anecdotal evidence abounds suggesting non-standard –s is frequent in Galicia and Catalonia (e.g., Cousillas, Reference Cousillas2019), we were unable to find empirical studies directly comparing rates of usage in these varieties with other varieties in Spain. Thus, the current study provides empirical support for what has heretofore been a widespread, but understudied phenomenon. The age and contact results taken together suggest that increased exposure to the non-standard –s, whether due to language contact with Galician and Catalan, or due to more exposure to verb forms containing the non-standard –s as children age, leads to higher rates of the non-standard –s in children's speech.
Lexical frequency results
The final research question explored the relationship between lexical frequency and children's 2sg preterit forms. Hypothesis 3a predicted that children would produce non-standard –s more often with low-frequency verbs. In contrast, H3b predicted that a strong tendency to overregularize could override children's lexical learning of frequent forms, in which case children who produce non-standard –s would do so with both high- and low-frequency verbs. The results in Tables 5 and 6 confirm H3a evidencing that low-frequency verbs such as pintaste ‘you painted’ and abriste ‘you opened’ occurred more with the non-standard -s while high-frequency verbs such as viste and fuiste favored the standard forms. These results lend support to the Conserving Effect (Bybee, Reference Bybee2007) suggesting that children learn high-frequency verbs as entrenched forms; the more entrenched a form is in the child's mental repertoire, the less likely it is to be overregularized. These findings further suggest that when children overregularize, they do so by extending a productive morphological pattern to the low-frequency verbs. That is, low-frequency forms like pillaste ‘you caught’ are targeted during the regularization process (Bybee, Reference Bybee2015; Clahsen et al., Reference Clahsen, Aveledo and Roca2002).
The lexical frequency effect is further exemplified by the 38 non-standard tokens, of which 30 were low-frequency verbs, three mid-frequency, and five high-frequency. It is also worth noting that none of these non-standard tokens was a collocation, such as te diste cuenta ‘you realized’. In fact, there was only one lexically bound collocation, tuviste miedo ‘you were afraid’, which occurred once and with tuviste in the standard form. While there were no other lexically bound collocations, the verb viste was frequently followed by que ‘that’ and como ‘how’. Viste in these contexts was never produced as vistes. Similarly, dijiste que ‘you said that’ occurred four times, always with the standard variant (8). Meanwhile, when used without que, dijiste occurred with the non-standard –s on two occasions (9).
(8) No me dijiste que me ibas a llevar. (Luis, m, 6;07; Corpus: BecaCESNo)
‘You didn't tell me that you were going to take me.’
(9) Pero como no me dijistes el otro día. (Virginia, f, 8;09; Corpus: BecaCESNo)
‘But since you didn't tell me the other day.’
While there were insufficient collocations to arrive at firm conclusions, the lack of non-standard –s observed in units like dijiste que supports the idea that highly frequent strings of words are learned as a chunk and therefore resist overregularization of a form internal to the unit (Bybee & Hopper, Reference Bybee and Hopper2001). In this sense, frequent collocations show an effect that is similar to the one found for high-frequency verbs. In summary, the lexical frequency results support the conclusion that high-frequency, entrenched forms are protected from overregularization.
Closing remarks and future research
This preliminary study investigated Spanish children's 2sg preterit variation examining the effects of age, contact setting, and lexical frequency. Results demonstrate that Spanish children, like adults, do in fact vary between non-standard and standard 2sg preterit forms with non-standard –s produced at a rate of 17%. Results from mixed-effects regression analyses indicate that contact setting significantly affects children's non-standard –s production. Specifically, those in contact settings (Galicia, Catalonia) are likelier to produce non-standard –s. A question that remains open for exploration, however, is whether these findings reflect CLI from the children's own L1 (Galician or Catalan) or increased exposure to non-standard –s in the Spanish spoken in contact areas. Additionally, lexical frequency predicts children's 2sg preterit forms: low-frequency verbs are likelier to occur with non-standard –s than high-frequency verbs. Contrarily, mixed-effects results (Table 6) indicate no age effect. While age was not a significant predictor of 2sg preterit variation in the current data set, there were limitations regarding the distribution of participants across age groups and contact settings as the oldest age group included 16 children who all resided in the non-contact group (Table 6). Since contact setting was a significant predictor of non-standard –s, it is important to include data from older children in contact settings for more representative data for contact setting and to better perceive the effects of age. Indeed, based on Lerner (Reference Lerner2016), there is reason to believe that 2sg –s variation is more prevalent among older children.
The current study's findings provide areas of interest for future research – in addition to incorporating more children across age groups, future studies should address the effect of register and the age at which children acquire adult-like constraints on 2sg preterit variation by incorporating a balance between formal and informal speech. Due to the disparity between formal and informal settings, it was difficult to address this question in the current study. The majority of corpora utilized consisted of more informal speech situations such as child play or conversation between family members and researchers. However, one corpus incorporated more formal speech acts such as narrating a story, and there was a preference for the standard form in this setting with all but one token being the standard variant (N = 28).
Furthermore, while the current study did not analyze input effects, this factor merits future attention. This is especially important considering that children's production of the non-standard –s presents an interesting conundrum: it could be attributed to an input effect, but could also be due to children's tendency to regularize irregular forms (Clahsen et al., Reference Clahsen, Aveledo and Roca2002; Johnson, Reference Johnson and Clark1995). Previous research on developmental sociolinguistics showing that children's variation mirrors their input predicts that children exposed to both standard and non-standard 2sg preterit forms in the input will produce both forms and learn the social and linguistic factors that constrain adults’ variation (Lerner, Reference Lerner2016; Miller, Reference Miller2013; Smith & Durham, Reference Smith and Durham2019). While it is expected that children's variation between 2sg preterit forms is influenced by the variation found in their input (e.g., Lerner, Reference Lerner2016) analyses of input were beyond the scope of the current study due to the nature of the data.
One limitation of corpus studies like the one reported here is that 2sg forms are not very frequent, and 2sg past tense forms are even rarer. Although the spontaneous conversations included CDS, the scarcity of 2sg preterit forms directed at the children prevented meaningful analysis of input effects. The lack of input data in the current study leaves us with the question of whether children's non-standard 2sg preterit –s forms are the result of acquisition of sociolinguistic variation – that is, learning from the variation present in the input – or the outcome of the general tendency to regularize irregular forms during childhood. Children's general tendency to regularize may result in the production of non-standard –s even among children who are not exposed to this form in the input, provided they have acquired 2sg –s in other verbal paradigms such as the present simple (e.g., caes ‘you fall’) and imperfect tenses (e.g., caías ‘you were falling’). Future studies that include input data in addition to child language production data will help tease apart the effects of input and regularization on Spanish-speaking children's 2sg preterit variation.
Acknowledgements
I would like to extend my deepest gratitude to Dr. Naomi Shin for all her support throughout this process. Thank you, Naomi. Your guidance and encouragement have been invaluable in helping me prepare this manuscript. I would also like to thank Dr. Eva Rodríguez-González, Dr. Jill Morford, and Dr. Karen Miller for their input and thoughtful comments on this manuscript. Lastly, I would like to thank two anonymous reviewers for their helpful feedback.
