2.1. Number of nominal demonstrative terms

Languages vary with respect to the number of nominal demonstrative terms that contrast in function. Many languages have two or three nominal demonstrative terms, while languages with four or more terms are rarer (Coventry et al., Reference Coventry, Gudde, Diessel, Collier, Guijarro-Fuentes and Vulchanova2023; Jungbluth & Da Milano, Reference Jungbluth, Da Milano, Jungbluth and Da Milano2015; Rubio-Fernandez, Reference Rubio-Fernandez2022; Skilton & Peeters, Reference Skilton and Peeters2021; Woensdregt et al., Reference Woensdregt, Jara-Ettinger and Rubio-Fernandez2022). As mentioned above, English has two demonstratives: this and that. Spanish has three: proximal este, medial ese, and distal aquel. Footnote ¹ However, aquel is infrequent in oral discourse (Shin & Vallejos-Yopán, Reference Shin, Vallejos-Yopán, Rojo and Cacoullos2023), and some varieties, like Peruvian Amazonian Spanish, rely mostly on este and ese (Vallejos-Yopán, Reference Vallejos-Yopán2023).

There are also languages with only one nominal demonstrative. In these languages, distance is typically marked by the addition of an adverbial. For example, Present-day French has one nominal demonstrative (ce), and spatial distance is marked by constructions that combine ce with adverbial ci “here” or là “there” (ce-ci “this-here,” ce-là “this-there”) or with pronominal and adverbial forms (celui-ci “this one here,” celui-là “this one there”) (Guillot, Reference Guillot, Jungbluth and Da Milano2015; McCool, Reference McCool1993). Likewise, in German, the archaic distal form jener is no longer used for exophoric reference, and distance is marked by combining the nominal demonstrative with adverbs hier, da, and dort (Coventry et al., Reference Coventry, Gudde, Diessel, Collier, Guijarro-Fuentes and Vulchanova2023; Rauth & Speyer, Reference Rauth and Speyer2021). Similar situations are found in Danish, colloquial Swedish, and Norwegian (Coventry et al., Reference Coventry, Gudde, Diessel, Collier, Guijarro-Fuentes and Vulchanova2023; Lander, Reference Lander2021, p. 17; Vulchanova et al., Reference Vulchanova, Collier, Guijarro-Fuentes and Vulchanov2023).

Shifting distance marking from nominal to adverbial demonstratives also occurs in systems with two demonstrative terms. In spoken Present-day Brazilian Portuguese, the proximal nominal demonstrative este has fallen out of use, yielding a two-term system (esse/aquele). Adverbs frequently combine with esse/aquele to reinforce spatial distinctions (Meira & Guirardello-Damian, Reference Meira, Guirardello-Damian, Levinson, Cutfield, Dunn, Enfield, Meira and Wilkins2018). Similarly, in Northern Italian dialects, which have two nominal demonstratives, spatial adverbs reinforced nominal demonstratives so frequently that, as Ledgeway (Reference Ledgeway, Bárány, Biberauer, Douglas and Vikner2020, 456) puts it, “all deictic force has been transferred to the adverb, reducing the demonstrative to a mere marker of definiteness” (also see Andriani et al., Reference Andriani, Groothuis and Silvestri2020).

Thus, a pathway of change across numerous languages involves the loss of spatial contrast in nominal demonstratives and the use of adverbial demonstratives to mark space. Lander (Reference Lander2021, p. 9) describes the pathway as follows: “When the deictic force of a demonstrative is felt by speakers to be too weak, a particle (e.g., adverb) can be added, which in turn gets weakened over time, necessitating reinforcement once again.” This common pathway of change is important to keep in mind when interpreting findings that point to bilinguals diverging from monolinguals in their use of nominal demonstratives (see Section 2.3), especially since bilinguals may sometimes lead or accelerate language change (Heine & Kuteva, Reference Heine and Kuteva2005; Joseph, Reference Joseph, Mufwene and Escobar2022; Kupisch & Polinsky, Reference Kupisch and Polinsky2022; Silva-Corvalán, Reference Silva-Corvalán1986, Reference Silva-Corvalán1994).

2.2. Functions of demonstratives

A basic function of demonstratives is to encode the distance between the speaker and a referent (Coventry et al., Reference Coventry, Gudde, Diessel, Collier, Guijarro-Fuentes and Vulchanova2023; Diessel, Reference Diessel1999; Diessel & Coventry, Reference Diessel and Coventry2020). Speaker-referent distance marking corresponds especially to the division between the speaker’s peripersonal (reaching space) and non-peripersonal space; proximal demonstratives are generally used for referents in the peripersonal space, whereas demonstrative usage is more variable in the non-peripersonal space (Caldano & Coventry, Reference Caldano and Coventry2019; Shin et al., Reference Shin, Hinojosa-Cantú, Shaffer and Morford2020; Skilton, Reference Skilton2024). Addressee-referent distance as well as pragmatic factors, like joint visual attention, and misalignment in referent identification between the speaker and addressee, have also been shown to shape demonstrative use (Peeters et al., Reference Peeters, Azar, Özyürek, Bello, Guarini, McShane and Scassellati2014, Reference Peeters, Krahmer and Maes2021; Rubio-Fernandez, Reference Rubio-Fernandez2022; Shin et al., Reference Shin, Hinojosa-Cantú, Shaffer and Morford2020; Woensdregt et al., Reference Woensdregt, Jara-Ettinger and Rubio-Fernandez2022). This study sets aside addressee effects and pragmatic factors and focuses on the impact of bilingualism on the use of nominal demonstratives to mark the speaker-referent spatial distance.

2.3. Bilinguals’ nominal demonstratives

To date, research on bilinguals’ demonstrative systems has included Spanish–Norwegian, Catalan–Spanish, and Spanish–English bilinguals. Vulchanova et al. (Reference Vulchanova, Guijarro-Fuentes, Collier and Vulchanov2020) found that Spaniards in Norway produced aquel less often than Spaniards in Spain, which they interpret as evidence of a shift from a three-term to a two-term system. At the same time, the decreased use of aquel was not impacted by Spanish/Norwegian language use or proficiency (Vulchanova et al., Reference Vulchanova, Collier, Guijarro-Fuentes and Vulchanov2023), suggesting that the shift is not necessarily propelled by the structure of Norwegian, but instead may be an example of bilinguals accelerating the common pathway of change in Romance languages (see Section 2.1).

Studies of Catalan–Spanish bilinguals also reveal differences between bilinguals’ and monolinguals’ Spanish. Rubio-Fernandez (Reference Rubio-Fernandez2022) found that Catalan–Spanish bilinguals evinced a weakening in their sensitivity to addressee location effects in Spanish, possibly due to the lack of such effects in Catalan. In another study on Catalan–Spanish bilinguals and Spanish monolinguals, Todisco et al. (Reference Todisco, Guijarro-Fuentes, Gudde and Coventry2024) found no significant effect of addressee location on demonstratives – neither among the bilinguals (in Catalan and Spanish) nor among the monolinguals (in Spanish).Footnote ² At the same time, they found preliminary evidence that, for referents at a middle distance, Catalan–Spanish bilinguals produced Spanish este more often than Spanish monolinguals, who relied more heavily on Spanish ese.

With respect to Spanish–English bilinguals, Shin and Mendieta-Rodríguez (Reference Shin and Mendieta-Rodríguez2024) compared demonstratives produced in Spanish by Spanish-English bilinguals and Spanish-speaking monolinguals. Participants completed a puzzle task designed to elicit demonstratives. Spatial distance was manipulated by means of a cord that was placed 50 cm from the participant, and the participants were instructed not to reach across the cord. The instructions and the presence of the cord helped create a division between the participants’ peripersonal (“near”) space from their non-peripersonal (“far”) space. This division has been shown to strongly condition demonstrative use (Caldano & Coventry, Reference Caldano and Coventry2019; Skilton, Reference Skilton2024). Puzzle pieces were placed on either side of the cord and the experimenter sat across from the participant with the board and pieces between them (see Figure 1). The experimenter then asked 25 “find-it” questions that led the participants to identify specific puzzle pieces: ¿Qué pieza tiene los ojos del dinosaurio rojo? “Which piece has the red dinosaur’s eyes?” Overall, participants relied heavily on esta and esa for puzzle pieces in the near and far space, respectively. However, the bilinguals produced significantly more esta for referents in the far space as compared to the monolinguals, which concords with the extension of proximals among Catalan–Spanish bilinguals (Todisco et al., Reference Todisco, Guijarro-Fuentes, Gudde and Coventry2024). Shin and Mendieta-Rodríguez (Reference Shin and Mendieta-Rodríguez2024) hypothesized that the bilinguals’ increased use of esta could be related to English influence. However, the authors did not test the bilinguals’ demonstrative use in English, nor did they test English monolinguals’ usage, rendering their hypotheses about English influence entirely speculative.

Figure 1. Puzzle task layout.

To summarize, previous research has found that Spanish–English bilinguals’ demonstrative use differs from that of Spanish-speaking monolinguals but has not fully explored explanations for those differences. This article aims to fill this gap and to advance our understanding of whether and how bilingualism shapes demonstrative use by performing comparative analyses between bilingual and monolingual groups (Study 1), as well as examining the effect of language dominance on demonstrative use and linguistic convergence (Study 2).

3.1. Participants

The participants in this study include 20 Spanish-speaking monolingual adults residing in Monterrey, Mexico; 20 English-speaking monolingual adults residing in New Mexico, USA; and 22 Spanish–English bilingual adults residing in New Mexico. All 20 Spanish-speaking monolinguals grew up in Monterrey, Mexico, and were raised speaking only Spanish at home. The English-speaking monolinguals resided in Albuquerque, New Mexico, at the time of the study, and all grew up monolingually in the USA. All monolinguals report speaking only one language at home and with family and friends, and none report fluency in a second language.

All 22 bilinguals in Study 1 resided in the USA at the time of the study. The bilinguals fall under the umbrella of what are typically considered Spanish heritage speakers since they grew up speaking Spanish at home in the USA, where English is societally dominant (Valdés, Reference Valdés2005). All bilinguals were either born in the USA (N = 17) or in Mexico (N = 5). The Mexico-born bilinguals arrived in the USA by age 7;0 (range = 1;5–7;0, mean = 4.9, SD = 2.51). It is common for heritage speakers to primarily speak the home language until they go to school, which means that bilinguals born in the USA and those who arrive at around 5;0 often have similar language experiences in the early years. Table S1 in the Supplementary Material provides more details about the bilinguals’ language histories.

3.2. Puzzle task

Participants completed the 25-piece puzzle task used by Shin and Mendieta-Rodríguez (Reference Shin and Mendieta-Rodríguez2024). As mentioned above, in this task, the participant sits at a table that has an empty puzzle board, the puzzle pieces, and a cord placed 50 cm from the participant (Figure 1). Participants are told not to touch any pieces and not to reach across the cord. The inclusion of the cord thus functions as a barrier that divides the space into a near (peripersonal) and far (non-periperonsal) space. As noted before, this division has been shown to condition demonstrative use (Caldano & Coventry, Reference Caldano and Coventry2019; Skilton, Reference Skilton2024). The near space contains the empty puzzle board and 12 puzzle pieces. The other 13 pieces are placed in the far space. The puzzle task begins with one practice item followed by a series of 25 scripted “find-it” questions like “Which piece has the red dinosaur’s eye?” These questions prompt the participant to identify pieces which the experimenter then places on the puzzle board.Footnote ³

Spanish sessions in Mexico and the USA were conducted by an experimenter whose first and dominant language was Spanish, and English sessions in the USA were conducted by an experimenter whose first and dominant language was English. All sessions were video and audio recorded. For the bilinguals who completed the task in both languages (N = 18), the data collection for each language was carried out on separate days: the Spanish session preceded the English session by an average of 10 days (SD = 9.5).

3.3. Data extraction and coding

All responses to find-it questions with nominal demonstratives were extracted. We excluded tokens that did not have an identifiable referent. We also excluded tokens of aquel/aquella (n = 31) from the Spanish data due to its infrequency. After these exclusions, our datasets included 858 Spanish nominal demonstratives and 1,127 English nominal demonstratives, which were coded for the variables described below. The second author checked all transcriptions and coding for the dependent and independent variables.

We ignored gender in Spanish and collapsed este and esta into one level called esta and ese and esa into one level called esa. The two levels of the dependent variable, demonstrative, are proximal and non-proximal, with proximals comprising esta for the Spanish datasets and this for the English datasets. Likewise, the non-proximals comprise esa for the Spanish datasets and that for the English datasets. Henceforth, we label the proximals esta/this and the non-proximals esa/that.

The puzzle task layout (Figure 1) manipulated speaker-referent distance by placing puzzle pieces in the “near” space (between the barrier and the participant) or in the “far” space (on the other side of the barrier from the participant). This manipulation enabled us to include distance as a binary independent variable (near versus far) in the analyses. We also included a binary independent variable called adjacency to control for how participants may respond to pieces in different locations within the near and far spaces. The puzzle pieces were either “adjacent” (directly above or below the barrier) or “not adjacent” (Figure 1). Participant group – monolingual Spanish, monolingual English, or bilingual – was also included as an independent variable.

3.4. Results

Figure 2 presents a boxplot of the participants’ proportions of esta/this and esa/that by dataset. The language labels “Span” or "Eng” indicate whether the proportions refer to the use of esta versus esa (“Span”) or this versus that (“Eng”). On average, speakers overwhelmingly produced proximal demonstratives for pieces in the near space and decreased their use of proximal demonstratives for pieces in the far space. At the same time, the proportions for each demonstrative type were variable across datasets and individuals, especially in the far space.

Figure 2. Speakers’ proportions of demonstrative forms by distance and dataset.

To measure the impact of speaker-referent distance on demonstrative use, we conducted generalized linear mixed-effects models using lme4 (Bates et al., Reference Bates, Mächler, Bolker and Walker2015) in R Studio Team (2023). In each model, we predict the likelihood of non-proximal demonstratives. Each best-fit model was determined by iteratively eliminating terms that decreased the Akaike Information Criterion (AIC) when eliminated, starting with the one that produced the highest AIC until the lowest possible AIC value was reached.

We first compared the Spanish-speaking monolinguals’ and English-speaking monolinguals’ demonstrative use. The maximal model included an interaction between group and distance and an interaction between distance and adjacency, as well as a random slope that allowed the effect of distance on demonstrative use to vary by participant. The best-fit model retained the interaction between distance and adjacency, as well as group as a main effect. Table 1 reports the best-fit model. Positive estimate (β) values indicate that the likelihood of esa/that increases, whereas negative estimate values indicate that the likelihood of esa/that decreases. Interaction terms are listed with the pertinent levels of the variables involved in the interaction. The values associated with the main effect of any variable in interaction should be understood as the effect of the reference level of the variable (the one not listed for the interaction term).

Table 1. Results of generalized logistic regressions predicting esa/that for Spanish-speaking monolinguals and English-speaking monolinguals

The best-fit model (Table 1) did not include the interaction between group and distance, suggesting that the English-speaking monolinguals were just as likely as the Spanish-speaking monolinguals to produce non-proximal demonstratives in the far space. That is, the impact of speaker-referent distance on demonstrative use was similar across the two monolingual groups (also see Figure 2). The results also indicate that the likelihood of producing non-proximal demonstratives increased when pieces were in the far space, and even more so when the pieces were “far” and not adjacent to the line. While both monolingual speaker groups were similarly impacted by speaker-referent distance, the English-speaking monolinguals were, overall, significantly more likely to produce non-proximal demonstratives as compared to the Spanish-speaking monolinguals.

Next, we compared the bilinguals to each reference lect. That is, we compared the bilinguals and Spanish-speaking monolinguals’ Spanish demonstratives in one model and the bilinguals and English-speaking monolinguals’ English demonstratives in a separate model. The maximal models included an interaction between group and distance and an interaction between distance and adjacency, as well as a random slope that allowed the effect of distance on demonstrative use to vary by participant. Table 2 reports the results of the models, with the Spanish data in the left columns and the English data in the right columns. Empty cells indicate that the term was dropped from the best-fit model. In this and subsequent tables, we also report the percentage of simulations (out of 1,000) in which the results are significant when we randomly sample the dataset with replacement (i.e., bootstrapping) up to the size of the original dataset and run the data through the maximal model (Hay & Foulkes, Reference Hay and Foulkes2016). We offer these values to address a reviewer’s concern about the composition and sample size of the bilingual group. Despite introducing differing amounts of heterogeneity into the dataset with each simulation, the results are consistent across simulations.

Table 2. Results of two generalized logistic regressions predicting esa/that; one for bilinguals’ and Spanish-speaking monolinguals’ Spanish demonstratives (left) and one for bilinguals’ and English-speaking monolinguals’ English demonstratives (right)

The results in Table 2 show a significant interaction between group and distance for the Spanish data; as compared to the Spanish-speaking monolinguals, the bilinguals were less likely to produce esa when referring to pieces in the far space. The results for the English data show a main effect of group, but no interaction with distance. This indicates that the bilinguals were overall less likely to use that than the English-speaking monolinguals, but this difference was not conditioned by speaker-referent distance.

3.5. Study 1: Discussion

Study 1 compared the impact of speaker-referent distance on demonstrative use among Spanish-speaking monolinguals in Mexico, English-speaking monolinguals in the USA, and Spanish–English bilinguals in the USA. The study investigated whether these groups’ demonstrative use is differentially affected by speaker-referent distance (RQ1). Previous research had demonstrated that bilinguals were significantly less likely to produce esa as compared to Spanish-speaking monolinguals (Shin & Mendieta-Rodríguez, Reference Shin and Mendieta-Rodríguez2024), but it was unclear whether their English demonstratives would also differ from those produced by English-speaking monolinguals. We predicted that all speakers would produce more non-proximal demonstratives when referring to pieces in the far space. At the same time, we anticipated that the impact of speaker-referent distance would vary across groups.

Supporting our prediction, all groups produced more non-proximal demonstratives for referents farther from the speaker, which is a common cross-linguistic tendency (Coventry et al., Reference Coventry, Gudde, Diessel, Collier, Guijarro-Fuentes and Vulchanova2023; Diessel, Reference Diessel1999; Diessel & Coventry, Reference Diessel and Coventry2020). The comparison of English- and Spanish-speaking monolinguals revealed no significant difference with respect to the impact of speaker-referent distance on demonstrative use (Table 1 and Figure 2). However, one difference between the monolingual groups emerged: the English-speaking monolinguals were more likely to produce non-proximal demonstratives overall. This reflects usage patterns in conversational English and Spanish. The frequency counts reported in Section 1 demonstrate that the English distal demonstratives that/those are more than twice as frequent as the proximal demonstratives this/these, whereas the Spanish medial demonstratives ese/esa/esos/esas are only 23% more frequent than the proximal demonstratives.Footnote ⁴

The comparisons between the bilinguals and monolinguals showed some differences with respect to speaker-referent distance (Table 2). The bilinguals produced more proximal Spanish demonstratives than the Spanish-speaking monolinguals when referring to pieces in the far space (Figure 2). The bilinguals’ English demonstrative use was similar to that of the English-speaking monolinguals with respect to speaker-referent distance, but the English-speaking monolinguals were overall more likely to produce non-proximal that than the bilinguals.

The differences between the monolinguals’ and bilinguals’ demonstrative use lend support to previous research showing that bilinguals differ from monolinguals in their use of demonstratives (Todisco et al., Reference Todisco, Guijarro-Fuentes, Gudde and Coventry2024; Vulchanova et al., Reference Vulchanova, Guijarro-Fuentes, Collier and Vulchanov2020, Reference Vulchanova, Collier, Guijarro-Fuentes and Vulchanov2023). Like our findings, an increase in Spanish proximal este/esta at the expense of non-proximal ese/esa was also found among Catalan–Spanish bilinguals in Majorca (Todisco et al., Reference Todisco, Guijarro-Fuentes, Gudde and Coventry2024). Interestingly, as mentioned above, Vulchanova et al. (Reference Vulchanova, Guijarro-Fuentes, Collier and Vulchanov2020, Reference Vulchanova, Collier, Guijarro-Fuentes and Vulchanov2023) found that Spanish–Norwegian bilinguals in Norway produced fewer tokens of distal aquel as compared to monolinguals in Spain, providing evidence of a shrinking demonstrative system among the bilinguals. Yet, language usage and proficiency variables did not significantly impact the bilinguals’ use of demonstratives, suggesting that the reduction of aquel was not necessarily shaped by crosslinguistic influence form Norwegian, a language in which adverbials rather than nominal demonstratives encode distance. Likewise, in our study, the bilinguals’ greater reliance on the Spanish proximal esta is not due to the cross-linguistic influence of English. This is evident from the fact that the English-speaking monolinguals were, overall, more likely to produce non-proximals as compared to the other two groups. In other words, the monolingual English data does not point to a preponderance of this. In fact, if English were impacting the bilinguals’ Spanish, we would anticipate an increase in non-proximal esa rather than proximal esta (Shin et al., Reference Shin, Marchesi and Morford2021). Thus, an English-contact explanation cannot explain why the bilinguals produced more esta when referring to pieces in the far space.

To further explore the bilinguals’ nominal demonstrative use, Study 2 addresses the extent to which these bilinguals’ demonstrative use evinces linguistic convergence. Additionally, we consider the impact of language dominance and individuals’ strategies to mark distance to further understand the potential sources of linguistic convergence.

4.1. Research questions

Previous research yields two predictions (P2.1a and P2.1b) related to RQ2.1:

Previous research yields two predictions (P2.2a and P2.2b) related to RQ2.2:

Previous literature yields two competing predictions (P2.3a and P2.3b) associated with RQ2.3:

4.2. Participants

The study includes the same bilingual participants from Study 1, but only those who completed the puzzle task in both Spanish and English (N = 18).

4.3. Puzzle task

Each participant completed the puzzle task described in Study 1 using the same layout and procedure.

4.4. Independent variables

Study 2 tests the effects of three independent variables on bilinguals’ demonstrative use. Like Study 1, we include distance and adjacency. The third independent variable is language dominance, which is based on participants’ responses to the Bilingual Language Profile (BLP; Birdsong et al., Reference Birdsong, Gertken and Amengual2012). The BLP includes questions on language history, language usage, self-rated language proficiency, and language attitudes. To calculate language dominance, sections are scored, weighted, and then combined to obtain a global language score. Then one global language score is subtracted from the other to obtain the language dominance score. In this study, participants responded to the history, usage, and proficiency questions, but not the attitudes questions, which are a slightly less reliable indicator of language dominance (Olson, Reference Olson2023). With three sections included, the potential range for the global language score was 0–163.5. To obtain the language dominance score, the English global language score was subtracted from the Spanish global language score. As such, positive scores indicate Spanish-language dominance, and negative scores indicate English-language dominance. Scores close to zero indicate balanced bilingualism. The average BLP score of the 18 bilingual participants in Study 2 is −29.96 (range = −104.78–5.19, SD = 26.41). In the statistical analysis, we use scaled BLP scores, which are standardized scores around the mean. Table S1 in the Supplementary Material provides more details about the bilinguals’ responses to the BLP questions.

4.5. Data coding

All coding procedures and exclusion criteria followed those described in Study 1. After all exclusions, the dataset for Study 2 included 416 tokens of Spanish nominal demonstratives and 339 tokens of English nominal demonstratives. As in Study 1, all tokens were coded for distance, and the second author checked each video and all coding.

4.6. Results

4.6.1. Results from generalized linear mixed-effect models

We constructed one generalized linear mixed-effects regression model for the bilinguals’ Spanish and English data combined. We addressed our research questions by including a three-way interaction between language session (Spanish versus English), BLP score, and distance. This way, we investigate the effect of BLP score and distance on demonstrative use and examine whether this effect is different in the bilinguals’ Spanish and English. We also included an interaction between adjacency and distance, as was done in Study 1. Random slopes led to model overfitting, so the maximal models only included participant as a random effect. We adopted the same model selection procedure used in Study 1, as well as the same bootstrapping procedure. Table 3 reports the results of the model.

Table 3. Results of generalized logistic regressions predicting esa/that for bilinguals

The results of the best-fit model reveal a significant interaction between BLP score and distance. The direction of the effect indicates that when the bilinguals referred to pieces in the far space, the likelihood of esa/that increased as BLP scores increased. Stated differently, the more balanced the bilingual, the higher the likelihood of using esa/that in the far space. Figure 3 illustrates the interaction between BLP scores and distance by plotting the predicted probabilities of esa/that in the near (solid line and solid circles squares) and far (dotted line and open squares) space by the participants’ BLP scores in both language sessions. On the far left of each panel, the most English-dominant participants infrequently produce esa/that, producing instead primarily esta/this in the far and near spaces. On the far right of each panel, the more balanced bilinguals more frequently produce esa/that in the far space and less frequently produce esa/that in the near space. There was no effect of BLP score for referents in the near space on the likelihood of esa/that, as evidenced by the nonsignificant main effect of BLP score in Table 3. In other words, all bilinguals, regardless of BLP score, primarily produced esta/this when referring to pieces in the near space.

Figure 3. Predicted probabilities of esa/that by BLP score and distance.

The results also reveal an interaction between distance and adjacency of puzzle piece to the barrier. When bilinguals referred to puzzle pieces in the far space that were also not adjacent to the barrier, rendering these pieces even farther from the participant, the likelihood of esa/that significantly increased. The main effects associated with this interaction reiterate that proximal demonstratives were the dominant forms in the near space, and that non-proximal demonstratives were more common in the far space.

Absent from the best-fit model is the “language” variable, as well as interactions between language and BLP scores and between language and distance. The lack of a language effect suggests that the interaction between BLP score and demonstrative use spans across the two languages. Indeed, Figure 3 illustrates the similar effect of BLP score on demonstrative use in both English and Spanish. It seems that individuals adopt patterns that either coincide with tendencies in the reference lects (i.e., producing esa/that in the far space) or contradict these tendencies (i.e., producing esta/this in the far space), and regardless of which pattern they adopt, they apply it across both their languages. Nevertheless in 18.7% of datasets simulated in the bootstrapping procedure, there was a significant effect of the three-way interaction between language, BLP, and distance. It is possible that with more data, a significant three-way interaction would emerge. Thus, we adopted a second type of statistical analysis to further investigate language convergence in the bilinguals’ use of demonstratives.

4.6.2. Results from recursive partitioning models

The second set of analyses was conducted with recursive partitioning methods: a conditional inference tree (CIT) and conditional random forest (CRF) in partykit (Hothorn & Zeileis, Reference Hothorn and Zeileis2015). The goal of the CIT algorithm is to partition the data into maximally homogenous subsets based on the strength of the associations between the predictors and the dependent variable and until partitioning is no longer justified (Levshina, Reference Levshina, Paquot and Gries2020). The results of the CRF yield variable importance scores for each predictor as ascertained from growing numerous CITs. This approach is meant to further explore the three-way interaction and uncover variation among individuals in the present dataset.

Recursive partitioning models are beneficial in cases where the number of predictors is too large for the number of datapoints and when there are complex interactions involved (Levshina, Reference Levshina, Paquot and Gries2020). Both scenarios apply here. More specifically, determining whether individuals’ usage patterns exhibit linguistic convergence involves testing a three-way interaction between participant, distance, and language. Additionally, demonstrative use might be predicted by the joint influence of BLP scores and distance in one language but not the other. We tested for these interactions in the regression model; however, the three-way interaction was not significant. One interpretation is that we have found a case of widespread linguistic convergence wherein “language” is not a significant predictor of bilinguals’ usage patterns. Another interpretation is that testing these interactions via regression models requires a larger dataset. The output of inference trees, which includes “participant” as a variable, also offers an efficient way to group participants with similar demonstrative use and thus provides a principled approach to individual analyses.

In this study, the CIT analysis assesses the likelihood of producing esa/that versus esta/this based on three predictors: distance, language, and participant. If individual participants use different strategies across their languages (i.e., they have a higher likelihood of producing non-proximal demonstratives in the “far” space in one language rather than in both languages), we should see an interaction between language and distance for those participants. By “interaction” we mean that the datapoints associated with these participants end up in separate subsets of the data as partitioned by the algorithm. In contrast, for individuals that use the same strategy across languages, we should not find any interaction with the language variable; these participants’ Spanish and English datapoints should be partitioned into the same subset.

The results of the CIT reveal two significant predictors: distance and participant (Figure 4). The algorithm first splits the data by distance. In the “near” subset (n = 431), no further splitting was justified, which indicates that language and participant do not constrain demonstrative use in the near space: all speakers, in both languages, overwhelmingly produce esta/this (Figure 4, Node 2). In the far space (n = 324), participant, but not language, constrains nominal demonstrative use (Nodes 3–7). The algorithm identifies three groups of participants with distinct demonstrative behavior in the far space (Nodes 5–7). The language variable does not appear in any CIT node, which suggests that the participants who fall into each group behave the same way across their languages. Figure S1 in the Supplementary Material presents each participant’s rates of esa/that and esta/this in the near versus far space and in each language, further demonstrates that most individuals behave similarly across their languages.

Figure 4. Conditional inference tree for Spanish–English bilinguals’ demonstrative use.

The results of the CRF performed in partykit are presented in Figure 5. The higher the variable importance score, the more strongly the predictor impacts demonstrative use. Figure 5 indicates that distance is the strongest tested predictor of demonstrative use, then the individual, and language has a score of 0, indicating that its effect is inconsequential to demonstrative use.Footnote ⁵

Figure 5. Variable Importance Factors extracted from conditional random forest results predicting Spanish–English bilinguals’ demonstrative use.

The CIT and the CRF show that the participants’ use of demonstratives is similar across their languages. At the same time, the CIT results (Figure 4) indicate that there are three groups of participants who vary in their use of demonstratives for referents in the far space. The first group of participants (N = 3; Node 5) followed the typical trend found among monolinguals: they generally produced non-proximal esa and that for puzzle pieces in the far space. As such, we consider these participants to be “conservative” in their nominal demonstrative behavior with respect to distance marking. The participant IDs for these three speakers are provided above Node 5. Figure S1 in the Supplementary Material further confirms that these speakers are conservative across their languages: their esa/that rates are appreciably high in the far space (75% or greater) and low in the near space.

The second group of participants’ (N = 6; Node 6) far-space demonstrative use was highly variable. In English, each speaker produced that in at least 30% of responses, and in Spanish, each participant produced esa in 25% of responses (Figure S1 in the Supplementary Material; refer to Figure 4 Node 6 for participant IDs). Among the participants in this group, some exhibited divergent behavior across their languages, whereas others tended to be similar. For instance, for the far space p251 and p258 produced primarily esa (85%; 75%, respectively) but produced that less frequently (60%; 57%, respectively). By contrast, p253’s rate of that in the far space was high (80%), but her rate of esa was low (30%).

The third and largest group of participants (N = 9; Node 7) primarily produced esta and this for puzzle pieces in the far space, thereby bucking the typical trend of producing non-proximals in the far space. We consider these participants to be “innovative” in their demonstrative use since this is not a strategy typically found among Spanish and English speakers. Moreover, as illustrated in Figure S1 in the Supplementary Material, eight of these nine innovative speakers produced high rates of proximal demonstratives in the far space in both their languages (see Figure 4 above Node 7 for participant IDs). P256 is the only exception; his assignment to the innovative group may be related to his small number of datapoints. To summarize, the results of the recursive partitioning models uncover a large amount of linguistic convergence among individual speakers. Discounting p256, 11/18 bilinguals’ demonstrative use exhibits linguistic convergence. Of these 11 speakers, 8 exhibit innovative behavior (i.e., producing proximal demonstratives for referents in the far space) and 3 exhibit conservative behavior across their languages. Table 4 provides a summary of the far-space proximal demonstrative usage and language dominance information for each of the three groups of bilinguals that were identified by the CIT algorithm.

Table 4. Language dominance information for groups identified via CIT

With respect to the question of whether language dominance impacts convergence (RQ2.3), we performed a correlation between the absolute difference in the rate of esa and that in the far space and BLP scores. We take this absolute difference to be an indicator of how similar bilinguals’ use of nominal demonstratives is across their languages; smaller absolute differences indicate greater similarity. Figure S2 in the Supplementary Material plots the absolute differences by BLP score for each participant. A Spearman’s rank correlation test shows no correlation between the absolute difference in the rate of esa and that in the far space and the participants’ BLP scores [ρ(16) = 897.89, p = .38]. This finding corroborates the results from the group-level regression analysis discussed earlier: language (Spanish versus English) was not a significant predictor of demonstrative use – neither as a main effect nor in interaction with BLP scores. Thus, language dominance predicts which demonstrative form a bilingual is more likely to produce (Figure 3), but it does not predict which types of bilinguals are more likely to exhibit linguistic convergence (Figure S2 in the Supplementary Material).

4.7. Study 2: Discussion

The primary goal of Study 2 was to investigate the extent to which Spanish–English bilinguals’ use of nominal demonstratives is shaped by linguistic convergence at the group (RQ2.1) and at the individual level (RQ2.2), and whether linguistic convergence is impacted by language dominance (RQ2.3). To that end, the study analyzed 18 bilinguals’ use of nominal demonstratives in both Spanish and English. In addition, we conducted group-level and individual analyses of demonstrative use with a range of statistical tests to gain insight into the extent of convergence (RQs 2.1, 2.2) and to examine the impact of language dominance on convergence patterns (RQ2.3). In what follows, we discuss how our findings answer the three research questions.

We found that 11/18 participants’ nominal demonstrative use provides evidence for linguistic convergence, and this linguistic convergence manifested at the group and individual levels in two different trends (RQs 2.1, 2.2). On one hand, some bilinguals’ (N = 3) demonstrative use in both their languages was “conservative.” That is, their usage was similar to the trend found among monolinguals in that they produced non-proximal demonstratives for referents that were farther away (Table 2 and Figure 4, Node 5). The results from the regression model suggest that this behavior was more common among the more balanced bilinguals in the sample (Figure 3). Furthermore, these three “conservative” speakers’ BLP scores were relatively close to 0, indicating balanced bilingualism (Table 4). For these three participants, the relationship between speaker-referent distance and demonstrative forms in one language reinforces the same relationship in the other language. We interpret this as evidence of enhancement or entrenchment of a structural similarity across the bilinguals’ languages (Bullock & Toribio, Reference Bullock and Toribio2004).

Linguistic convergence in other bilinguals’ (N = 8) nominal demonstrative use manifested as producing primarily proximal demonstratives in the far space in both languages, thereby bucking the more typical trend of producing non-proximals for referents that are farther away. The results from the regression model suggest that this behavior is more common among the English-dominant bilinguals in the sample since the likelihood of producing non-proximal demonstratives decreased with increasing English dominance. The fact that these bilinguals’ usage pattern goes against the typical trend in the reference lects and that they apply this pattern across their languages suggests that this “innovative” behavior also illustrates linguistic convergence.

The manifestation of linguistic convergence among the “innovative” bilinguals is unlikely a result of crosslinguistic influence but rather an internally induced enhancement of structural similarities between linguistic systems. This interpretation is supported by the comparisons between the Spanish-speaking monolinguals and English-speaking monolinguals (Study 1). Figure 2 presents all three groups side-by-side and illustrates the bilinguals’ higher rates of proximal demonstratives when identifying referents in the far space as compared to the other groups. That is, the bilinguals’ use of proximals in the far space outstrips that of both monolingual groups. Moreover, the group-level analysis of the bilinguals’ data shows that it is the most English-dominant participants who are most likely to produce proximal demonstratives when referring to pieces in the far space (Table 3). But again, these English-dominant bilinguals’ extension of proximals does not appear to stem from English influence since the English-speaking monolinguals do not exhibit this same trend. In fact, the English-speaking monolinguals produce, overall, more non-proximal demonstratives than the Spanish-speaking monolinguals. An increase in Spanish esta at the expense of esa was also found among Catalan–Spanish bilinguals in Majorca (Todisco et al., Reference Todisco, Guijarro-Fuentes, Gudde and Coventry2024), and an increase in esa at the expense of aquel was found in analyses of Spanish–Norwegian bilinguals in Norway (Vulchanova et al., Reference Vulchanova, Guijarro-Fuentes, Collier and Vulchanov2020, Reference Vulchanova, Collier, Guijarro-Fuentes and Vulchanov2023). Thus, our findings concord with previous research in the sense that the bilinguals extend their use of one demonstrative at the expense of another, and this difference does not appear to be due to cross-linguistic influence.

The finding that the “innovative” bilinguals mainly use proximal demonstratives may initially give the impression that they do not mark speaker-referent distance. Nevertheless, subsequent individual analyses reveal that, instead of varying between nominal demonstratives, the “innovative” bilinguals use other strategies already available in both Spanish and English to indicate speaker-referent distance. The principal structure is a schema that combines a proximal nominal demonstrative and locative language: [DEM.PROX + LOCATION]. Consider p242, who used esta/this categorically. This innovative bilingual most frequently produced esta de aquí “this one here” for near pieces. For pieces in the far space, she used esta de acá for pieces that were adjacent to the barrier and thus closer to herself, and for those that were not adjacent to the barrier and thus farther away, she produced esta de allá. This suggests that, for this speaker, esta de acá means “this one here,” but is used for referents slightly farther away, and esta de allá means “this one there.” Thus, the speaker places the functional load of locating the referent on adverbial demonstratives rather than on nominal demonstratives when speaking Spanish. This same participant used a similar strategy in English. Her two most frequent phrases in the English condition were this one right here, for referents in the near space, and this one over here, for referents in the far space, demonstrating that, just like in Spanish, the speaker placed a greater functional load on locative language rather than on nominal demonstratives to guide the interlocutor to the location of the referent.

Similar to p242, another innovative bilingual, p241, alternated between esta de aquí and esta de acá when referring to near and far pieces, respectively. Deploying the adverbs in a different way, p243 frequently used esta pieza “this piece” for near-space pieces and added adverbial demonstratives when referring to pieces in the far space. Other innovative bilinguals tended to rely on prepositional constructions to guide the experimenter to the relevant puzzle piece. For example, p261 produced just esta for near pieces, but consistently produced esta en el otro lado “this one on the other side” for far pieces. In sum, the "innovative" bilinguals rely less on nominal demonstratives to locate referents in space, deploying other locative constructions, primarily adverbial demonstratives, to mark distance instead.

It is worth noting that the combination of nominal and adverbial demonstratives is also found among the Spanish- and English-speaking monolinguals. However, compared to the monolinguals, the bilinguals produce more constructions that combine nominal demonstratives and adverbials (15% and 25%, respectively). More importantly, when referring to far pieces using a proximal nominal demonstrative, the bilinguals produce non-proximal adverbial demonstratives (e.g., this one there) more often than the monolinguals do (27% and 22%, respectively). Also, when referring to far pieces using a non-proximal demonstrative, the bilinguals produce non-proximal adverbials (e.g., that one there) more often than the monolinguals do (38% and 13%, respectively). Thus, these analyses reveal a case in which bilinguals enhance a structural similarity across their languages. This construction is available in both Spanish and English and, while it is found among monolinguals, bilinguals produce it to a greater extent. In this case, the shared structure is a schema of [DEM.PROX + LOCATION], primarily using a [DEM.PROX + LOC.ADV] construction. This finding, along with findings from the studies on Spanish-Catalan and Spanish–Norwegian bilinguals, shows that Spanish nominal demonstratives undergo changes in bilingual contexts. More specifically, the individual analyses uncover an innovative analytic construction, whereby nominal demonstratives serve as a reference marker and adverbial demonstratives, or other locative constructions, mark distance.

This “innovative” strategy of relying on adverbials rather than nominal demonstratives is found in other languages, too, like Italian (Jungbluth & Da Milano, Reference Jungbluth, Da Milano, Jungbluth and Da Milano2015). Moreover, this “innovative” strategy is more grammaticalized in languages like French and some varieties of German (Coventry et al., Reference Coventry, Gudde, Diessel, Collier, Guijarro-Fuentes and Vulchanova2023; Guillot, Reference Guillot, Jungbluth and Da Milano2015; Rauth & Speyer, Reference Rauth and Speyer2021). In other words, the bilinguals in this study are only innovative in comparison to the Spanish- and English-speaking monolinguals. In fact, the bilinguals’ “innovative” usage reflects a common cross-linguistic diachronic change whereby first the distance meaning of nominal demonstratives is supported by adverbials that reiterate the distance meanings, and then the distance meaning becomes associated with the adverbials rather than the nominal demonstratives. Farther along this path, we may expect that the nominal demonstrative continues to serve only as a marker of definiteness (Ledgeway, Reference Ledgeway, Bárány, Biberauer, Douglas and Vikner2020), while the adverb or other locative language encodes distance. As discussed in Section 2.1, this change is found in diachronic analyses of demonstratives across languages. Again, the bilinguals’ usage patterns are only “innovative” when we compare them to Spanish-speaking monolinguals and English-speaking monolinguals (Figure 2); they are not unusual when we consider diachronic changes in demonstratives in other languages. If anything, the bilinguals’ usage patterns are perhaps further along on a common cycle of language change. If this is the case, this finding aligns with the assertion that heritage speakers (Kupisch & Polinsky, Reference Kupisch and Polinsky2022) and bilinguals in general can accelerate or enhance language change (Heine & Kuteva, Reference Heine and Kuteva2005; Joseph, Reference Joseph, Mufwene and Escobar2022). For example, Silva-Corvalán (Reference Silva-Corvalán1986, Reference Silva-Corvalán1994) found that U.S. Spanish–English bilinguals extended the use of the copula verb estar at the expense of copula verb ser and that this was an internally motivated language change that was accelerated by language contact. In the present study, the bilinguals’ [DEM.PROX + LOCATION] schema is already available in Spanish and English spoken by other groups, including monolinguals. With continued usage, the eventual shift to marking distance by means of the adverbials rather than the nominal demonstratives is likely the next step. Nevertheless, future research is needed to understand the motivations and mechanisms behind this change, as well as document its progress in bilingual and monolingual communities.

Having established that the bilinguals’ usage patterns provide evidence of convergence, we now turn to the question of whether language dominance mediates the extent of linguistic convergence (RQ2.3). We entertained two possible ways in which language dominance may predict convergence. On the one hand, balanced bilinguals might show more convergence due to the tendency to “maximize common ground” (P2.3a; Hawkins & Filipović, Reference Hawkins and Filipović2024). On the other hand, less balanced bilinguals may transfer linguistic features and usage patterns from their more dominant to their less dominant language (P2.3b; Silva-Corvalán & Treffers-Daller Reference Silva-Corvalán, Treffers-Daller, Silva-Corvalán and Treffers-Daller2016; Treffers-Daller, Reference Treffers-Daller2019). However, the results indicate no effect of language dominance since we find cases of convergence among the more balanced bilinguals and the more English-dominant ones. Moreover, there was no correlation between BLP scores on the one hand and the difference in the rate of that and esa on the other (Figure S2 in the Supplementary Material). It is possible that the composition of the bilinguals in our study accounts for the lack of a language dominance effect on linguistic convergence. All the bilinguals, including the more English-dominant ones, grew up speaking Spanish at home, and thus had a strong command of both languages. Perhaps bilinguals with a more uneven command of their languages would exhibit less convergence.

Another factor that may have promoted convergence in this study relates to the linguistic structure at hand. Words with fewer translation equivalents are more closely connected in the bilingual mind (Tokowicz & Kroll, Reference Tokowicz and Kroll2007). Previous studies suggest that bilinguals reduce their use of aquel/aquella (Shin & Mendieta-Rodríguez, Reference Shin and Mendieta-Rodríguez2024), which means they rely primarily on este and ese. As such, the proximal demonstratives this and este/esta are arguably translation equivalents with no competitors. Thus, it is likely that the bilinguals in this study strongly associate this with este/esta and that with ese/esa and subsequently apply usage strategies across these associated lexemes. The lack of semantic competitors coupled with the high frequency of these demonstrative forms across the two languages may further intertwine bilinguals’ usage patterns.