Task Type and Distributional Differences in the Spanish Differential Object Marking of Catalan–Spanish Bilinguals

Abstract

This study examines offline acceptance and online processing of Differential Object Marking (DOM) in the Spanish of Catalan–Spanish bilinguals in Catalonia. Both languages evidence DOM, though prescriptive grammars claim only partial overlap. Empirical research on Catalan DOM within these bilinguals reveals differences in distribution. Based on these factors, along with sustained bilingualism at the community and individual levels, more optionality was predicted for the distribution of Spanish DOM. Results from an offline scalar Acceptability Judgment Task and a Self-Paced Reading Task reveal three important findings. First, each task revealed distinct distributions. Participants aligned more with prescriptive grammars for DOs that are high on the animacy and definiteness scales in the offline task and were more tolerant of greater variability with the same DOs in the online task, possibly indicating weakening of the obligatory DOM constraint in these contexts. Second, geographic area modulated acceptance of the absence of DOM with animate DOs, suggesting microvariation. Third, unmarked inanimate DOs were preferred across both tasks. Overall, the results are interpreted as revealing divergence from prescriptive descriptions of Peninsular Spanish DOM system.

1. Introduction

Differential Object Marking (DOM) is a widely studied linguistic property that necessarily invokes the syntax–semantics interface due to its dependence on the semantic features of the direct object (DO), and which can also invoke the syntax–pragmatics interface with topicalization and dislocation, as is the case in some languages. Though much of the experimental literature on the acquisition of DOM among monolingual children of a variety of languages, including Spanish, reports high levels of convergence as early as 1;6 to 5;0 years of age (see Avram, 2015 and chapters therein for overview; Chung, 2020; Foroodi-Nejad, 2011; Rodríguez-Mondoñedo, 2008; cf. Guijarro-Fuentes et al., 2017; Ketrez, 2015), research on other populations shows comparatively more divergence.

For example, second language (L2) learners of Spanish show errors of omission of DOM in sentence completion tasks (Guijarro-Fuentes, 2012), in narrative tasks (Arechabaleta Regulez & Montrul, 2021), in elicited production tasks (Nediger et al., 2017), and are less sensitive to ungrammatical DOM omission (Hopp & Arriaga, 2016). Proficiency (e.g., Guijarro-Fuentes & Marinis, 2007), the specific featural make-up of the DO (e.g., Sagarra et al., 2020), task type (e.g., Arechabaleta Regulez & Montrul, 2021), and the L1 of participants (e.g., Montrul, 2019) all affect convergence on DOM.

Similarly, heritage speakers of Spanish also omit DOM in oral production tasks (e.g., Montrul, 2004), in story-retelling tasks (e.g., Montrul & Sánchez-Walker, 2013), in sentence completion tasks (e.g., Cuza et al., 2018), and have shown less sensitivity to the absence of DOM in written acceptability tasks (e.g., Montrul & Bowles, 2009) and in self-paced reading tasks (e.g., Jegerski, 2014; cf. Jegerski & Sekerina, 2020). Task type (e.g., Arechabaleta Regulez & Montrul, 2023), verb frequency (e.g., Hur, 2020), and proficiency (e.g., Thane, 2024) are some of the factors that affect the production, comprehension, and processing of DOM in heritage speakers. Importantly, though differences are found across many L2 and heritage speaker studies, some of the same studies and others report convergence.

Likewise, research on early simultaneous bilinguals reveals a medley of outcomes, with some convergence and some divergence from monolinguals across a variety of tasks, language combinations, and ages. Austin (2020) reported that young Basque–Spanish bilinguals (2;01–3;4) use DOM at the same age in each language and that bilingualism in Spanish seems to accelerate the use of DOM in Basque, given that DOM is protracted in monolingual Basque as compared to Spanish. Avram and Tomescu (2020) reported convergence on DOM via examination of longitudinal corpora data in both monolingual Romanian and simultaneous Hungarian–Romanian bilinguals (1;11–2;01). Though the bilinguals were less accurate with uses of DOM dependent upon discourse-pragmatics, DOM appears very early and follows the same trajectory in both groups. Differently, Ticio (2015) documented the emergence of DOM in Spanish in the spontaneous production of simultaneous Spanish–English bilinguals. Compared to monolingual children, the simultaneous bilinguals omitted DOM more, leading the author to conclude that DOM was not acquired by age 3;6 in the simultaneous bilingual group. Bilinguals tested later in life also show differences from monolinguals (e.g., Guijarro-Fuentes & Marinis, 2009), across varieties (Rodríguez-Ordóñez, 2020), and according to syntax and the semantic features of the DO (Bel & Benito, 2024).

Importantly, recent experimental and corpus-based studies on bilinguals in contact varieties have shown that these speakers exhibit varying degrees of optionality with regards to DOM acceptability, processing, and production, and this is particularly true of Catalan–Spanish bilinguals. As described in more detail in Section 2, both Spanish and Catalan exhibit DOM. While some overlap exists between the two systems, prescriptive grammars stipulate a comparatively more restricted distribution of DOM in Catalan. Somewhat contrary to prescriptive grammars, the Catalan of Catalan–Spanish bilinguals reveals more optionality, whereby DOM distribution is wider than predicted, a finding that holds across task type. For example, said optionality in Catalan DOM is found in acceptability and processing tasks (Benito, 2023; Bel & Benito, 2024; Puig-Mayenco & Judy, 2025; Puig-Mayenco et al., 2017, 2018) and in production tasks (Benito, 2017, 2020, 2023; Perpiñán, 2018). Research examining specific regional varieties outside of Catalonia demonstrate that the distribution of Catalan DOM is not a monolith, but rather that some varieties show a wider DOM distribution, while some retain a more restricted one (Pineda, 2023; Zeugin, 2021; see also Boeddu (2020) for Arborense in contact with Standard and Regional Italian and Corral Esteban (2020) for Asturian and Vaqueiru in contact with Spanish).

These differences from monolingual and prescriptive norms, as well as the factors that might modulate them, are the focus of this article. More specifically, and in light of the divergent outcomes reported in different types of bilinguals (e.g., L2 learners, heritage speakers), we examine offline acceptance and online processing of DOM in the Spanish of adult simultaneous Catalan–Spanish bilinguals residing in two geographic regions of central Catalonia. The experimental methodology seeks first to determine the actual distribution of DOM in the Spanish of these Catalan–Spanish bilinguals and then to determine which linguistic and extralinguistic factors modulate its distribution. Based on previous research on Catalan, along with sustained bilingualism at the community and individual levels, more optionality in the distribution of Spanish DOM is predicted than described in prescriptive or monolingual grammars. We hope to make strides towards identifying the source(s) of this optionality, whether it be framed as grammatical instability, the presence of competing systems, or a stable state of variation. Our study also contributes to the experimental studies that examine dialectal variation of DOM distribution, in many cases regional dialects (e.g., Balasch, 2011; Benito, 2023; Bautista-Maldonado & Montrul, 2019; Company Company, 2001, 2002; Dumitrescu, 1997; Judy & Iverson, 2020; Montrul, 2013; Puig-Mayenco & Judy, 2025; Zdrojewski, 2020), but also contact dialects (e.g., Boeddu, 2020; Corral Esteban, 2020; Pineda, 2023; Zeugin, 2021). While the aforementioned studies examining DOM in Catalan–Spanish bilinguals examine the distribution in Catalan, the current study focuses on Spanish DOM distribution and is among the first to do so alongside Benito (2023). As described in subsequent sections, our results point to a DOM distribution unique to this contact variety.

2. Differential Object Marking

Differential Object Marking (DOM) is a linguistic property evidenced in many languages, including several Romance varieties. Though various syntactic accounts of DOM exist, the indisputable facts are that, in languages with DOM, some direct objects (DOs) are marked while others are unmarked. Syntactic accounts vary in terms of their formalization of the triggers of DOM. For example, Bossong’s seminal work (Bossong, 1983/1984) distinguishes two types of semantic features that affect DOM: inherent features, such as animacy, and referential features, such as specificity and definiteness. Many of the world’s languages rely exclusively on referential features, such as specificity in Persian, while others rely on a combination of both, such as animacy and specificity in Spanish. Torrego (1998, 1999) pointed out that both the agentivity of the subject and verbal semantics impact DOM in Spanish. Aissen (2003) later combined and formalized these factors as prominence, creating an animacy and a definiteness scale seen in (1) and (2) below, whereby DOs higher on each scale are more likely to be marked.

(1)

Animacy scale:

human > animate > inanimate

(2)

Definiteness scale:

personal pronoun > proper name > definite NP > non-definite, specific NP > non-definite, non-specific NP

Still, many exceptions have been identified for Spanish DOM. For example, telicity and affectedness of the DO play a role (Laca, 2006; von Heusinger & Kaiser, 2011; see García García (2018) for overview; Romero Heredero & García García, 2023; cf. Romero Heredero (2021) regarding telicity), negative quantifiers always require DOM (Leonetti, 2008), DOM is used to distinguish between an inanimate subject and DOs (Rodríguez-Mondoñedo, 2008; García García, 2018), and both dislocation and individuation affect DOM (see Fábregas (2013), Gerards (2023), Iemmolo (2010), and López (2012) for comprehensive reviews). Some authors have also argued for the presence of DOM with inanimate DOs in certain constructs (e.g., Camacho, 2023; López, 2012; Wall et al., 2025). Wall et al. (2025), for example, examined the distribution of DOM with inanimate DOs across four varieties of Spanish and found that inanimate DOs in canonical transitive SVO sentences and in reversible–symmetrical predicates showed some degree of variation with respect to the acceptance of DOM. Furthermore, recent empirical accounts reveal dialect and speaker variation in largely monolingual Spanish-speaking areas. In Caribbean varieties, variable DOM use was revealed with [+animate; +definite] DOs (Alfaraz, 2011; Bullock & Toribio, 2009). For River Plate Spanish, Bautista-Maldonado and Montrul (2019), Judy and Iverson (2020), and Zdrojewski (2020) independently found that DOM is extended to inanimate DOs for some speakers, even when the subject is animate (see also Camacho, 2023; Wall et al., 2025). Earlier findings from Dumitrescu (1997) showed extension of DOM to topicalized inanimate DOs in Argentine Spanish. Similarly, the corpus work of Company Company (2001, 2002) chronicled extension of DOM to inanimate DOs in Mexican Spanish. Certainly, bilingual and contact varieties exhibit variation from prescriptive uses of DOM, as described above for Catalan in contact with Spanish. It is nonetheless methodologically important to outline the prescriptive distribution of DOM in Spanish and Catalan.

Comparison of Spanish and Catalan DOM

In both Spanish and in Catalan, DOM appears as the prenominal syntactic marker a. In Spanish, it is generally accepted that DOM appears with specific and animate DOs, and more frequently with humans than with animals. Importantly, in canonical SVO sentences with agentive—especially human—subjects and inanimate DOs, DOM is typically prescribed as ungrammatical (Batchelor, 2006; Bradley & Mackenzie, 2004; Butt et al., 2019; von Heusinger & Kaiser, 2007; Real Academia Española, n.d.; Zagona, 2002). In Catalan, prescriptive grammars present a more restricted distribution of DOM. DOM is generally considered to be limited to strong pronouns and some indefinite pronouns (see Bel, 2002), though it is prescribed in contexts of animate dislocated DOs and to clarify subject–object ambiguity (e.g., Fabra, 1918; GIEC, 2016; Solà, 1994). Historically, DOM appears variably in medieval Catalan (Segarra, 1985) and has been treated as either an internal development or as a product of contact with other Romance languages (Moll, 1952; Salvador & Pérez Saldanya, 1993). Contemporary prescriptive descriptions thus frame DOM as marginal and frequently stigmatize its extension as the result of contamination from Spanish.

Though we acknowledge that the actual use of DOM often extends beyond the canonical cases described above in both Spanish and Catalan, the experimental design of this study focuses precisely on canonical cases of DOM where there seems to be a greater degree of agreement as to the presence or absence of DOM. Of experimental value is the fact that Spanish and Catalan are said to differ in all but one of the four canonical cases examined herein. Specifically, we examine DOM with Proper Names, Human (definite) Noun Phrases, Human (indefinite) Noun Phrases, and Inanimate (indefinite) Noun Phrases.

Table 1. Prescriptive distribution of DOM in Spanish and Catalan.

	Proper Name	Human (Def) DP	Human (Indef) DP	Inanimate (Indef) DP
Spanish
Marco ha buscado…	a Judit	a la vecina	a una escritora	*a una tienda
Catalan
En Marc ha buscat…	*a la Judit	*a la veïna	*a una escriptora	*a una botiga
Mark looked for…	Judy	the neighbor	a   writer	a   store

below shows that, for all but the last category, Spanish is said to mark these DOs, while Catalan is said not to.

3. Methodological Design

3.1. Research Questions and Predictions

Given the differences between the distribution of DOM in Spanish and Catalan, the growing body of empirical studies reporting a wider distribution in Catalan, and diverging distribution in several varieties of Spanish, this study explores whether the effects of bilingualism and sustained contact at the individual and community level, as measured by language dominance, affect the distribution of DOM in Spanish. Specifically, the research questions guiding this experimental study are as follows:

RQ1: 

What is the distribution of DOM in the Spanish of Catalan–Spanish bilinguals in Catalonia?

RQ2: 

What linguistic and extralinguistic factors modulate the distribution of Spanish DOM in these bilinguals?

Based on previous research, we predict that Spanish DOM will be less categorical in this group of bilinguals than in prescriptive accounts of Spanish. This optionality would most likely show in a higher-than-expected acceptance and faster-than-expected processing of the unmarked versions of the three animate conditions tested below (Proper Name, Human (definite) Noun Phrase, Human (indefinite) Noun Phrase). This variation may be seen across the participants as a whole or in subsets as a function of extralinguistic factors, such as language dominance and use. Furthermore, since Spanish and Catalan converge on the purported absence of DOM with inanimate DOs, we predict less variation in this condition despite previous empirical research evidencing extension of DOM to inanimate DOs (see Section 2).

3.2. Tasks and Participants

To answer these research questions, participants completed two experimental tasks presented in Gorilla Experiment Builder (Anwyl-Irvine et al., 2020), and one task gathering demographic and language dominance information (Bilingual Language Profile (BLP), Birdsong et al., 2012). Each component was presented in Spanish to maintain Spanish mode. Given that the conditions and tokens are shared across both experimental tasks, they are listed here first for efficiency; then each experimental task is described. Specifically, the experimental tasks each contained 96 total1 tokens, 48 of which constitute the 4 conditions of interest listed below in examples (3)–(9) along with 48 filler tokens. Half of the tokens of interest were presented with DOM and half were presented without. Example tokens for the four conditions and the filler items are shown below. Importantly, participants first completed the online Self-Paced Reading Task (SPR), then the BLP, and lastly the offline Acceptability Judgment Task (AJT); participants finished all tasks between 25–60 min.

(3)

Proper Name (n = 12)

Dolores	ha visitado	(*a)   Nuria	en la salida	del   banco.
Dolores	visited	(*DOM) Nuria	at the exit	of-the bank
Dolores visited Nuria at the bank exit.

(4)

Human (definite) DP (n = 12)

Julia	ha encontrado	(*a)   la secretaria	en la entrada	del  despacho.
Julia	found	(*DOM) the secretary	at the entrance	of-the office
Julia found the secretary at the office entrance.

(5)

Human (indefinite) DP (n = 12)

Raquel	ha visto	(*a)   una cantante	en la  barra	del   restaurant.
Raquel	saw	(*DOM) a   singer	at the bar	of-the restaurant
Raquel saw a singer at the restaurant bar.

(6)

Inanimate (indefinite) DP (n = 12)

Juan	ha buscado	(*a)   una carta	en la  casa	de   la  esquina.
Juan	looked	(*DOM) a letter	at the house	from the corner
Juan looked for a letter at the corner house.

(7)

Grammatical Distractor Token (n = 24)

Javier	ha comprado	una libreta	a la tienda	del  pueblo.
Javier	bought	a   notebook	at the shop	of-the town
Javier bought a notebook at the town shop.

(8)

Ungrammatical (agreement) Distractor Token (n = 12)

Ismael	ha bebido	una cafè	al   bar	de la calle.
Ismael	drank	a  coffee	at-the bar	of-the street
Ismael drank a coffee at the bar on the street.

(9)

Ungrammatical (word order) Distractor Token (n = 12)

Isidro	ha limpiado	sala  una	para el encuentro	de padres.
Isidro	cleaned	room one	for the meeting	of parents
Isidro cleaned a room for the parents’ meeting.

3.2.1. Acceptability Judgment Task

The Acceptability Judgment Task (AJT) contained the conditions and tokens described above. The purpose of this task is to gather participants’ untimed judgments of the presence/absence of DOM in the experimental conditions. In this task, participants read each token in its entirety and recorded how good or bad they felt the token was on a sliding bar, as shown in Figure 1. Practice items were included and Gorilla Experiment Builder produced a different order of tokens for each participant.

3.2.2. Self-Paced Reading Task

The purpose of the Self-Paced Reading Task (SPR) was to measure participants’ reaction times (RTs) to the presence/absence of DOM in the four conditions described above. To reduce priming effects, participants completed the SPR (Just et al., 1982) first, followed by the other components of the experiment. The experimental stimuli were presented in a moving-window, region-by-region fashion according to syntactic segments. Thus, Proper Name tokens, for example, were presented as follows to participants:

(10)

Proper Name token by moving window

Subject	Verb	Direct Object	Prepositional Phrase	Prepositional Phrase
Region 1	Region 2	Critical Region	Spillover region	Wrap-up Region
Victor	ha buscado	(*a)   Sandra	en el centro	del  parque
Victor	looked for	(*DOM) Sandra	at the center	of-the park
Victor looked for Sandra in the middle of the park.

Participants simply pressed the space bar to advance through each token in the region-by-region fashion shown above. At the conclusion of 1/3 of the tokens, participants responded to a true–false comprehension question both to encourage attention to content as well as to ensure response time validity. Practice items were included and Gorilla Experiment Builder produced a different order of tokens for each participant.

3.2.3. Participants

In total, 125 Catalan–Spanish participants completed the experimental tasks described above. Participants hail from central Catalonia. Demographic and language dominance information was gathered via the Bilingual Language Profile (BLP; Birdsong et al., 2012). This questionnaire collects participant-reported demographics, such as age, gender, and education, as well as responses to factors known to influence language dominance, such as language history, use, proficiency, and attitudes. Together, the latter set of questions produces a gradient language dominance score on a scale of −218 to +218. Participants fell on the Catalan-dominant side of the scale, though there is a large amount of variation indicating the inclusion of both Catalan- and Spanish-dominant speakers with diverse lived experiences as regards Catalan–Spanish bilingualism. Relevant information from the BLP is summarized below in

Table 2. Self-reported demographics and composite language dominance score.

	Mean	SD
Women (n = 83), Men (n = 41), Unreported (n = 1)
Metropolitan area (n = 36), Central Catalonia (n = 88); Unreported (n = 1)
Average Age at Time of Testing	34.4	16.3
Age of Acquisition of Spanish (years)	1.46	2.29
Age of Acquisition of Catalan (years)	1.72	3.39
Reported Language History of Spanish (maximum score = 54.5)	37.80	10.80
Reported Language History of Catalan (maximum score = 54.5)	38.94	67.95
Reported Use of Spanish (maximum score = 54.5)	17.38	16.59
Reported Use of Catalan (maximum score = 54.5)	35.42	17.18
Self-reported Proficiency in Spanish (maximum score = 54.5)	49.87	5.60
Self-reported Proficiency in Catalan (maximum score = 54.5)	50.38	4.99
Composite Language Dominance Score (positive scores indicate Catalan-dominance; scale −218–+218)	31.66	67.95

.

4. Results

4.1. Acceptability Judgment Task Results

We first examined participants’ sensitivity to grammaticality in the distractor tokens, which contained clear gender and number agreement errors and word order violations. Participants were excluded from analysis if they did not give a score of 33 or less in the sliding bar for all ungrammatical distractor tokens, as this would suggest insufficient attention or an over-acceptance bias. Based on this criterion, 28 participants were excluded, leaving a final sample of 97 participants. For the remaining participants, ratings indicated high acceptance of grammatical distractors (M = 80.9, SD = 32.3) and low acceptance of ungrammatical distractors (M = 17.7, SD = 17.7). Recall that all scores are on a 1–100 scale.

Next, acceptability ratings for all experimental conditions were examined (see Figure 2 and

Table 3. Outcome of MODEL 1.

Predictor	Estimate	SE	t	p-Value
(Intercept)	76.19	7.21	10.562	<0.001
conditionHuman (def) DP	1.95	9.76	0.15	0.875
conditionHuman (indef) DP	−6.35	9.76	0.85	0.398
conditionInanimate (indef) DP	−1.45	9.76	−6.49	<0.001
markerAbsence	−60.68	9.76	−5.71	<0.001
regionCentral	12.11	3.72	−0.67	0.501
conditionHuman (def) DP * markerAbsence	21.32	13.81	1.53	0.131
conditionHuman (indef): DP * markerAbsence	20.34	13.81	1.54	0.129
conditionInanimate (indef) DP * markerAbsence	12.31	13.81	9.28	<0.001
conditionHuman (def) DP * regionCentral	3.84	3.95	0.67	0.503
conditionHuman (indef): DP * regionCentral	2.30	3.95	0.90	0.365
conditionInanimate (indef) DP * regionCentral	−0.80	3.95	1.61	0.106
markerAbsence * regionCentral	30.56	3.95	2.16	0.030
conditionHuman (def) DP * markerAbsence * regionCentral	−8.07	5.59	1.06	0.285
conditionHuman (indef): DP * markerAbsence * regionCentral	37.45	5.59	−0.41	0.676
conditionInanimate (indef) DP * markerAbsence * regionCentral	−27.95	5.59	−3.05	0.002

for all descriptive statistics). As seen in Figure 2, results for the Proper Name condition show a clear preference for the presence of DOM (M = 74.4, SD = 36.1) over its absence (M = 24.7, SD = 34.2), indicating that DOM marking is strongly favored with this DO type. In the Human (definite) DP condition, the ratings also reveal a preference for the presence of DOM (M = 77.8, SD = 35.2) over its absence (M = 53.6, SD = 41.8), though the difference is smaller, suggesting optionality. For the Human (indefinite) condition, participants again rated sentences with DOM more highly (M = 85.3, SD = 25.3) than those without (M = 55.2, SD = 42.3), though participants show more optionality with the absence of DOM in this token type as with the previous token type. By contrast, in the Inanimate condition, ratings were markedly higher for the absence of DOM (M = 81.8, SD = 31.0) than its presence (M = 15.5, SD = 27.7). Thus, participants reveal a consistent preference for DOM with animate DOs, while showing a dispreference for DOM with the inanimate DO token type, and reveal optional acceptance towards the absence of DOM with indefinite and definite Human DP DOs.

A linear mixed-effects model was fitted in the R environment (R Core Team, 2025) using the lme4 package (Bates et al., 2015) to examine the effects of condition (Proper Name, Human (definite) DP, Human (indefinite) DP, Inanimate (indefinite) DP), marker (presence or absence), and region (metropolitan area vs. central Catalonia) on the acceptance score (1 to 100). The potential effects of age and dominance (a continuous variable ranging from −218 [highly Spanish dominant] to +218 [highly Catalan dominant]) were also explored; however, neither factor improved model fit and were therefore excluded from the final model. Random effects for item and participant were also included. Table 3 below contains the output of the main model; we then explored all relevant comparisons.

The results show several significant effects and interactions. There is a main effect of condition, driven by the Inanimate (indefinite) DP condition, which is rated significantly lower than the Proper Name baseline (p < 0.001). A strong main effect of marker is also found, with sentences containing DOM rated significantly higher than those without (p < 0.001). The effect of region is not significant. Significant interactions emerge between condition and marker, particularly for the Inanimate (indefinite) DP condition (p < 0.001), indicating that the effect of DOM differs across conditions. In this condition, the presence of DOM sharply decreases acceptability, whereas the absence of DOM increases acceptability. Additionally, there is an interaction between marker and region (p = 0.030), which suggests that participants from central Catalonia (outside of the metropolitan area) show slightly less sensitivity to the absence of DOM than those from the metropolitan area. Finally, as illustrated in Figure 3 below, a three-way interaction between condition, marker, and region is significant for the Inanimate (indefinite) DP condition (p < 0.001), suggesting that regional differences modulate a slight overacceptance on the part of participants from central Catalonia for the absence of DOM in this condition.

4.2. Self-Paced Reading Task Results

For the SPR data, participants’ accuracy on the comprehension questions was assessed first to ensure they paid attention throughout the task (M = 89.12, SD = 9.11). A predetermined inclusion threshold of 75% accuracy was applied, and no participants were excluded on this basis. Prior to analysis, outlier trimming was conducted following the procedure outlined by Keating and Jegerski (2015). The lower cutoff was set at 100 ms, with any values below this replaced by 100 ms. The upper cutoff was determined as the mean plus two standard deviations. Overall, 2.47% of the data points were adjusted. Following trimming, the raw reaction times (RTs) were analyzed for both the Critical Region (CR) and the Spillover Region (SO) across conditions (Proper Name, Human (definite) DP, Human (indefinite) DP, Inanimate (indefinite) DP) with and without DOM. The descriptive statistics are presented in

Table 4. Descriptive statistics (mean and SD) for the SPR across the four conditions.

		Critical Region (CR)		Spillover Region (SO)
Condition	Marker	Mean	SD	Mean	SD
Proper Name	presence	788.86	494.52	760.54	463.58
	absence	897.57	569.70	792.44	470.52
Human (def) DP	presence	896.47	555.92	807.36	470.60
	absence	865.61	547.04	827.95	502.00
Human (indef) DP	presence	917.67	537.23	806.91	474.13
	absence	891.22	573.63	838.45	493.64
Inanimate (indef) DP	presence	996.77	630.03	817.24	494.31
	absence	813.49	492.45	784.78	474.66

.

A linear mixed-effects model was fitted in R (R Core Team, 2025) using the lme4 package (Bates et al., 2015) to examine the effects of condition (Proper Name, Human (definite) DP, Human (indefinite) DP, Inanimate (indefinite) DP), marker (presence vs. absence), language dominance (a continuous variable ranging from −218 [highly Spanish dominant] to +218 [highly Catalan dominant]), and geographic area on reaction times (RTs). The final model retained only condition and marker, as the inclusion of additional predictors did not improve model fit. Random intercepts were included for both items and participants.

Table 5. Outcome of Model 2.

Predictor	Estimate	SE	t	p-Value
(Intercept)	774.70	37.45	20.68	<0.001
conditionHuman (def) DP	77.61	32.02	2.44	0.019
conditionHuman (indef) DP	87.59	32.02	2.73	0.009
conditionInanimate (indef) DP	132.31	32.02	4.13	0.001
markerabsence	70.31	32.02	2.19	0.032
conditionHuman (def) DP * markerabsence	−75.84	45.28	−1.67	0.101
conditionHuman (indef) DP * markerabsence	−67.77	45.28	−1.49	0.142
conditionInanimate (indef) DP * markerabsence	−178.18	45.28	−3.93	0.001

below includes the model output.

The main model shows multiple main effects as well as interactions, which were further explored by means of post hoc analyses. Pairwise comparisons were performed using emmeans (Lenth et al., 2018) to examine whether participants were slower with either the presence or absence of DOM within each condition. The comparisons are summarized in

Table 6. Outcome of pairwise comparisons for CR ¹.

Condition	Contrast	Estimate	z-Ratio	p-Value
Proper Name	presence vs. absence	−70.31	−2.20	=0.354
Human (def) DP	presence vs. absence	5.53	0.17	1.00
Human (indef) DP	presence vs. absence	−2.54	−0.08	1.00
Inanimate (indef) DP	presence vs. absence	107.87	3.37	=0.017

¹ No significant interactions obtained for the SO region; as such, and for space reasons, we do not report these results.

and the effect is visualized in Figure 4 below. As shown, only the contrast for the Inanimate (indefinite) DP condition reached significance, with longer reaction times observed in the presence of DOM. No significant differences emerged for Proper Names or Human DPs (definite or indefinite). This pattern suggests that DOM sensitivity is limited to the lower end of the animacy/definiteness scale, where its presence may be unexpected or dispreferred.

5. Discussion

We begin the Section 5 by revisiting the research questions guiding the methodological design of the study, as well as their predictions:

RQ1: 

What is the distribution of DOM in the Spanish of Catalan–Spanish bilinguals in Catalonia?

RQ2: 

What linguistic and extralinguistic factors modulate the distribution of Spanish DOM in these bilinguals?

As outlined in Section 3.1, we predicted that the Catalan–Spanish bilinguals tested herein would provide less clear acceptance and RT data to the experimental conditions based on previous DOM research. Specifically, we predicted higher-than-expected acceptance and faster-than-expected RTs to the absence of DOM in the three animate conditions (Proper Name, Human (definite), Human (indefinite) DP) based on previous literature examining both Spanish and Catalan. As discussed in detail below, the distribution of DOM in the Spanish of our Catalan–Spanish bilingual participants differs from more conservative standard accounts of (Peninsular) Spanish DOM, evidencing more variability with animate DOs, across geographical area, and across task type, but importantly, not as a product of language dominance nor with inanimate DOs of the type tested herein.

We expand on this general summary by first considering the findings from the Acceptability Judgment Task (AJT). Recall that the experimental design included four conditions (Proper Name, Human (definite), Human (indefinite), and Inanimate (indefinite) DP) that fall at different points on both the animacy and definiteness scales referenced in Section 2. Perhaps not surprisingly, participants provided the most clear-cut judgments to the conditions at either end of the scale (i.e., Proper Name and Inanimate (indefinite) DP). First, participants showed categorical acceptance for the presence of DOM with Proper Names, and a simultaneous rejection of the absence of DOM with Proper Names. Though this finding aligns with the predicted distribution according to the animacy and definiteness scales and prescriptive and empirical accounts of Spanish DOM, it contrasts with recent research on the distribution of DOM in Catalan within the same speaker profile and in the same geographic area (Puig-Mayenco & Judy, 2025), where it was found that Catalan–Spanish bilinguals accept both the presence and the absence of DOM in Catalan. This contrast suggests that Catalan allows for greater flexibility and optionality even at the upper end of the scales than the Spanish of Catalonia. We interpret this finding in Spanish, culled from a similar participant population and geographic area, as an indication that Catalan does not seem to directly influence DOM marking in Spanish with Proper Names. That is to say that the fact that our bilingual participants retain categorical DOM in Spanish with Proper Names implies that the Spanish system is (metalinguistically) resistant to apparent CLI from Catalan and that the constraints governing DOM in Spanish remain strongly anchored at the top of the hierarchy, at least for the AJT data.

At the other end of the scales, the AJT results reveal a near-categorical rejection of DOM with Inanimate (indefinite) DPs, and near-categorical acceptance of the absence of DOM with this DO type. This is unsurprising given the animacy and definiteness scales, and general agreement that inanimate DOs tend to be unmarked in most varieties of Spanish (but see Camacho, 2023; García García, 2018; López, 2012; Wall et al., 2025), and in Catalan. Indeed, our inanimate condition served as a conservative control condition, and participants categorically rejected DOM in this context. This fits with the traditional view that, unlike other varieties (largely non-Peninsular; see Bautista-Maldonado & Montrul, 2019; Camacho, 2023; Company Company, 2001, 2002; Dumitrescu, 1997; Judy & Iverson, 2020; Zdrojewski, 2020), DOM with inanimate DOs remains outside the core grammar of Peninsular Spanish (see overviews in Fábregas, 2013; García García, 2018; Wall et al., 2025). At the same time, some research has argued that DOM can appear with inanimate DOs in highly restricted lexical or semantic environments (Camacho, 2023; López, 2012). In particular, Wall et al. (2025) showed that, when verb class and predicate type are carefully controlled, especially in reversible–symmetrical predicates, speakers across several Spanish varieties accept the presence of DOM with inanimate DOs. Importantly, our study did not include these predicate types nor specific syntactic constructions, and thus our findings should be understood as reflecting the canonical, highly dispreferred contexts for DOM with inanimate DOs. Nevertheless, as Wall et al. (2025) demonstrated, future work must systematically manipulate predicate type and syntax to determine whether the Spanish of Catalan–Spanish bilinguals behaves similarly to other contact and non-contact varieties regarding these emerging inanimate DOM patterns. We offer this suggestion for future research.

For the remaining two conditions (Human (definite), Human (indefinite) DP), participants show a high degree of acceptance of the presence of DOM, while simultaneously showing much more variability as regards the absence of DOM. That participants accept DOM with animate DOs in Spanish is expected, which makes their variable acceptance of the absence of DOM of import, given their prominence on the animacy and definiteness scales, prescriptive and empirical accounts of DOM distribution, and given the absence of any mitigating factors (e.g., a relative clause with the subjunctive indicating non-specificity). Since prescriptive accounts of Catalan purport that these animate DOs are unmarked in Catalan, Catalan can reasonably be considered a potential source of this variability. In fact, we initially expected language dominance (LD) to modulate DOM distribution in these bilinguals, based on previous experiments conducted within a similar population and geographic area. Nonetheless, the statistical measures exclude LD as a modulating factor, indicating that dominance in Catalan does not promote more acceptance of the absence of DOM with Human definite and indefinite DPs. Instead, there is an extralinguistic effect of geographical area, whereby participants from the metropolitan area of Barcelona accept the absence of DOM with animate DOs in Spanish slightly less than participants from central Catalonia. We interpret this result as indicating a change (potentially in progress) in the distribution of DOM in Spanish in certain areas of Catalonia, which parallels the effects that are seen for Catalan DOM (Benito, 2023) and suggests microvariation across metropolitan and non-metropolitan areas. We also suggest that these results may indicate an extension of the reported and expected variability seen for Human indefinite DOs (e.g., with a subjunctive relative clause) to Human definite DOs. This finding seems to suggest that participants across different areas of Catalonia are spreading the absence of DOM up the definiteness scale, which is happening asymmetrically across regions.2

The results of the Self-Paced Reading task (SPR) tell a slightly different story from that of the AJT, in that only one significant difference obtained for the pairwise comparison: participants were significantly slower with Inanimate (indefinite) DOs in the presence of DOM. That participants are slower with the presence of DOM in the inanimate condition aligns with the AJT findings, and is entirely expected given the categorical rejection of inanimate DOM in both Spanish and Catalan. It also confirms that the participants’ parser is sensitive to the presence of DOM with inanimate DOs in real time. That said, we might also expect a significant difference for the Proper Names condition given the participants’ categorical acceptance of DOM and rejection of the absence of DOM in the AJT; this finding was not born out. In fact, the SPR results for the three other conditions, all animate, show that speakers process both the presence and absence of DOM with Proper Names, Human (definite) and Human (indefinite) DPs similarly. The lack of significant differences within these three conditions indicates optionality (or different groups within the larger population). We take this optionality to indicate that, from a processing perspective, participants are not sensitive to the absence of DOM with animate DOs, meaning that their grammars allow for animate DOs to be unmarked. Considering the AJT results and the standard distribution, this is somewhat expected for the Human (definite) and Human (indefinite) conditions, as the participants accepted both the absence and presence of DOM. Nevertheless, that no statistically significant difference obtained for Proper Names in the SPR reveals a clear difference between the AJT and SPR results. Recall that, in the AJT, participants categorically rejected the absence of DOM in this condition, maintaining DOM with DOs that are high on the animacy and definiteness scales. However, the absence of DOM with Proper Names did not trigger a slowdown effect in the SPR. Such AJT–SPR asymmetries have been observed in similar and different bilingual and contact varieties (e.g., Arechabaleta Regulez & Montrul, 2023; Puig-Mayenco et al., 2018; Puig-Mayenco & Judy, 2025), where categorical judgments coexist with gradient or optional representations in online processing. In our study, this pattern suggests that the acceptability of the absence of DOM may be spreading up the animacy and definiteness scales to even Proper Names and at the level of processing, as was observed for the Human conditions in the AJT.

The discrepancy across tasks highlights the importance of triangulating different data sources (e.g., offline and online) when attempting to determine the distribution of DOM in bilingual grammars. Judgments alone would suggest a stable, categorical rule for proper names; processing alone would suggest much greater fluidity. The combination of both reveals a more nuanced picture: speakers appear to maintain a prescriptive or metalinguistic commitment to DOM with DOs that are high on the animacy and definiteness scales, while their real-time grammars tolerate greater variability with these same DOs, possibly indicating early stages of reanalysis or weakening of the obligatory DOM constraint in these contexts. Taken together, the SPR data, along with the AJT data, lead to the conclusion that DOM in the Spanish of Catalan–Spanish bilinguals in Catalonia is no longer uniformly obligatory even in contexts where canonical grammars predict categorical marking (with DOs high on the animacy and definiteness scales), revealing a degree of optionality that is not captured by descriptions of standard Peninsular Spanish.

6. Conclusions

In conclusion, our findings support the view that DOM is undergoing linguistic change not only in the Catalan of Catalan–Spanish bilinguals, as argued in the literature (Benito, 2017, 2020, 2023; Bel & Benito, 2024; Perpiñán, 2018; Puig-Mayenco & Judy, 2025; Puig-Mayenco et al., 2017, 2018), but also in their Spanish. Recent work in Catalan shows the existence of optional DOM in specific syntactic and semantic contexts (e.g., Benito, 2023; Puig-Mayenco & Judy, 2025), suggesting an internal shift in the object-marking systems of Catalan. Parallel processes have been documented for Spanish in a range of language contact and heritage settings where DOM displays increased optionality, reduced obligatoriness, or context-dependent variability (see the literature reviewed in Section 1). Our results indicate that similar processes may be shaping the Spanish of Catalonia: the traditional boundary between obligatory and variable DOM appears to be loosening, particularly for DOs in the middle ranges of the animacy and definiteness scales.

There are two potential sources for the observable changes to the DOM system in the Spanish of Catalan–Spanish bilinguals. One possibility is cross-linguistic influence from Catalan, whereby Spanish speakers extend some of Catalan’s greater tolerance for the absence of DOM marking into their Spanish grammars, especially in the two Human conditions tested herein, where variability is most pronounced. The second possibility is that we are observing the emergence of a stable bilingual variety (Perpiñán, 2018), in which at least some speakers have a Spanish grammar that independently licenses the absence of DOM with human DPs, not as the result of direct CLI, but as part of a broader process of reweighting the animacy, definiteness, and discourse-related constraints that regulate DOM in their variety of Spanish. In either case, DOM in the Spanish of Catalonia appears to be diverging from the more conservative Peninsular Spanish system, showing early signs of a restructuring.