Navigating Ambiguity: Scope Interpretations in Spanish/English Heritage Bilinguals

Abstract

This study investigates how Mexican Spanish/U.S. English heritage bilinguals process scope ambiguities in sentences containing the existential quantifiers a/una and the universal quantifiers every/cada in English and Spanish. Sentences like ‘A person bought every book’ are syntactically ambiguous in both languages, allowing for multiple possible interpretations. Research suggests that one interpretation is often preferred due to lower cognitive demand, though degree of preference varies across languages. Notably, heritage bilinguals may have distinct interpretation preferences in each language, highlighting the complexity of bilingual processing. Sixty Spanish/English heritage bilinguals (Age M = 25.48, SD = 2.65) completed a timed and graded truth-value judgment task in both languages, along with language proficiency tests. We analyzed interpretation ratings, response times, and potential effects of proficiency. Results reveal nearly identical preferred interpretation ratings (Spanish: M = 4.19, SD = 0.56; English: M = 4.14, SD = 0.66) and response times (Spanish: M = 6.97 s, SD = 2.70; English: M = 6.67 s, SD = 1.80) across languages, with one interpretation consistently favored and associated with faster response times. Language proficiency had no significant impact. Our experimental findings offer new insights into heritage bilinguals’ processing of competing linguistic structures and inform models of bilingual syntax and cognitive flexibility.

1. Introduction

Quantifiers can be considered complex because they often carry obscure meanings. For instance, the English quantifier a and its Spanish counterpart un/una can be ambiguous, as they can refer to either a specific or a non-specific element (for an additional example, see Tunstall, 1998, for a discussion of the nuances between the English quantifiers every and each). Even greater interpretive challenges arise when multiple quantifiers appear in a single sentence and interact with one another, as in sentence (1). The interaction between quantifiers creates a phenomenon known as quantifier scope ambiguity, in which sentences can have more than one interpretation based not only on the semantics of the individual quantifiers but also on their relative syntactic positions in the “internal syntax” (see Section 2 for a more in-depth explanation).

(1)	A person bought every book/Una persona compró cada libro
	(a)	∃ > ∀: There is a single person that bought all existing books
	(b)	∀ > ∃: For each existing book, there is a person that bought it

For example, sentence (1) can be interpreted in two ways. On the one hand (1a), it can be understood that a single person bought all the books in the scenario, whereas in the second interpretation (1b), it could be understood that for each book there is potentially a different person who bought it. These interpretations arise from two alternative positions of the quantifiers in the underlying/internal syntax. In (1a), the existential quantifier (∃; a) takes scope (i.e., occupies a higher syntactic position) over the universal quantifier (∀; every). In (1b), the universal quantifier takes scope over the existential quantifier.

This phenomenon has been the focus of extensive research, with scholars investigating whether scope ambiguities appear across languages and, when multiple interpretations are possible, which interpretations are preferred. In languages like Spanish and English for example, prior research suggests that speakers tend to prefer one interpretation, presumably because it is more accessible and thus easier to process (Anderson, 2004; O’Grady, 2008). However, research also shows that children, bilinguals, and second language speakers interpret scope ambiguities differently than monolinguals, highlighting the importance of considering diverse linguistic experiences when studying scope interpretations (Chung & Shin, 2023; Musolino et al., 2000; Musolino & Lidz, 2003, 2006; Syrett et al., 2017).

Bilinguals, particularly heritage speakers, seem to process scope ambiguities in complex ways, with studies revealing conflicting patterns in how they handle these ambiguities across their two languages. One key finding is that bilinguals tend to simplify complex syntactic structures to reduce cognitive load, leading them to process scope ambiguities similarly in both languages (Scontras et al., 2017; Ronai, 2018). In turn, some research suggests that bilinguals will favor more rigid scope calculations (Ronai, 2018; Scontras et al., 2017), while other studies indicate greater flexibility in their interpretations, with distinct preferences emerging (Chen & Hester, 2024; Chen & Huan, 2023; Shen & Chen, 2022). These patterns highlight the cognitive pressures that bilinguals face when managing competing linguistic systems, which may influence how they process scope ambiguities. However, this variation may stem from differences in the constructions examined and methodological variations across studies.

The goal of the current study is to explore how Mexican Spanish/U.S. English heritage bilingual adults interpret scope-ambiguous sentences in both languages, and to determine whether similar processing strategies and preferences/ratings emerge across both languages. Our initial focus is on the interaction between a/un/una and every/cada, though many other quantifier pairs could also be explored. Unlike much of the existing literature, which has largely focused on the interaction between ambiguous and unambiguous languages, both languages in this study feature quantifier scope ambiguities. This study also aims to examine whether bilinguals show differences in their interpretation preferences depending on language proficiency, and whether response times correlate with the acceptability of these interpretations. By exploring these questions, this study seeks to contribute to our understanding of how heritage bilinguals process scope ambiguities and whether shared cognitive strategies underlie their interpretations across both languages. It also provides insight into the broader implications for bilingual processing models and contributes to the understanding of how linguistic experience shapes the interpretation of ambiguous structures.

2. Quantifier Scope

In general, there seems to be a one-to-one mapping between syntax and semantics, meaning that for each syntactic structure, only one interpretation is possible. For example, sentence (2) is a declarative sentence that has only one possible semantic interpretation: a person named ‘Octavio Paz’ wrote a book titled ‘The Labyrinth of Solitude’. Regardless of whether the sentence is true or not, only this interpretation is derived from its syntactic structure. However, some sentence structures do not follow this one-to-one mapping, where a single syntactic structure can give rise to more than one interpretation. For instance, sentences containing more than one quantifier can be ambiguous in some languages. Consider the English and Spanish sentence (3), which contains the existential quantifier a/una and the universal quantifier every/cada. Because of the interaction of the quantifiers, two possible interpretations,1 (3a) and (3b), can be derived.

(2)	Octavio Paz wrote the book ‘The Labyrinth of Solitude’/Octavio Paz escribió el libro ‘El laberinto de la soledad’
(3)	A woman wrote every book/Una mujer escribió cada libro
	(a)	∃ > ∀: There is a single woman that wrote all existing books
	(b)	∀ > ∃: For each existing book, there is some woman that wrote it

Importantly, quantifier scope ambiguities do not occur in all languages (see Antonyuk, 2019, for an analysis of Russian; Aoun & Li, 1989, for a comparison of Chinese and English; Bobaljik & Wurmbrand, 2012, for a general cross-linguistic overview). For example, in Hungarian, to obtain an interpretation such as (3a), where there is a single woman who wrote every existing book in the scenario, the most accurate formulation of the sentence is that in (4), where egy nő’ (‘a woman’) precedes minden könyvet (‘every book’) in the syntax. By contrast, to get an interpretation equivalent to (3b), the sentence would have to be reformulated like (5), where minden könyvet (‘every book’) precedes egy nő’ (‘a woman’).

(4)	Egy nő          írt                      minden könyvet
	a     woman write.PST.3SG every       book.ACC
	‘A woman wrote every book’
(5)	Minden könyvet     egy nő          írt
	every     book.ACC a     woman write.PST.3SG
	‘A woman wrote every book’

The key distinction between languages that allow quantifier scope ambiguities and those that do not is whether they adhere to the Isomorphic Principle (Aoun & Li, 1989; Huang, 1982). In languages that follow this principle (6), such as Hungarian, the surface order of quantifiers corresponds directly to the internal syntax, which mirrors the structure of the meaning it conveys, leaving little room for scope ambiguities. If English and Spanish strictly followed this principle, only interpretation (3a) would arise from sentence (3), while interpretation (3b) would be implausible from this structure. However, sentence (3) in both Spanish and English permits the two interpretations, (3a) and (3b). This shows that in these languages, quantifiers do not need to be altered in the surface syntax to yield two distinct interpretations. Thus, languages with scope ambiguities violate the Isomorphic Principle.

(6)	Isomorphic Principle (Huang, 1982)
	Suppose A and B are quantifier phrases. Then if A c-commands B at surface structure, A c-commands B at LF

Instead, the ambiguity exemplified in sentences (3) above and (7) below is attributed to a form of syntactic movement known as quantifier raising (see Fox, 2002). Quantifier raising is the optional movement of a quantifier to a different syntactic position at Logical Form2 (LF; May, 1977, 1985, 1989), or the so-called underlying or “internal syntax.” This movement takes place covertly in LF and is not expressed in the surface structure of the sentence. As a result, a single surface syntactic structure or sentence can correspond to two distinct LFs. As such, interpretations (3a) and (7a) are derived from the LF in which no quantifier raising occurs, aligning with the surface structure; hence, this is referred to as the surface scope or surface interpretation. In this case, the existential quantifier a takes scope over the universal quantifier every (∃ > ∀). By contrast, interpretations (3b) and (7b) are derived from the LF in which the universal quantifier every raises to a higher syntactic position, resulting in every taking scope over a (∀ > ∃), even though the surface structure remains unchanged. This interpretation is therefore referred to as the inverse scope or inverse interpretation.

(7)	A man decorated every cake/Un hombre decoró cada pastel
	(a)	Surface scope (∃ > ∀): There is a single man that decorated all existing cakes
	(b)	Inverse scope (∀ > ∃): For each existing cake, there is some man that decorated it

Although there is more than one possible interpretation in the presence of scope ambiguities, research suggests that one of the interpretations tends to be more intuitive or preferred, both offline and online (Anderson, 2004; Crain & Thornton, 1998; Ioup, 1975; Kurtzman & MacDonald, 1993; Musolino, 1998; O’Grady, 2008; Tunstall, 1998). More often than not, the preference is for the interpretation consistent with the surface structure, such as (3a) and (7a). To account for why the surface structure interpretation is preferred, Anderson (2004) proposed the Processing Scope Economy Principle. According to this principle (8), inverse scope involves additional syntactic mechanisms (i.e., movement of the quantifier), making the processing more costly. This has been attested in reading times (Anderson, 2004; O’Grady & Lee, 2006). As a result, inverse scope is often the least preferred interpretation. Along the same line, O’Grady (2008) suggests that the inverse interpretation is less accessible because it places a greater load on working memory.

(8)	Processing Scope Economy Principle (Anderson, 2004)
	The human sentence processing mechanism prefers to compute a scope configuration with the simplest syntactic representation (or derivation). Computing a more complex configuration is possible but incurs a processing cost.

In addition to cognitive mechanisms, the interpretations of scope ambiguities can also be heavily influenced by other linguistic cues. Studies like Ward and Hirschberg (1985) and Kadmon and Roberts (1986) have explored the relationship between prosody and the interpretation of scope-ambiguous sentences, demonstrating that prosodic cues can help disambiguate meaning. More recently Syrett et al. (2014) showed that when speakers provide consistent auditory cues, listeners successfully arrive at the intended meaning. Features such as context and plausibility of the interpretations are also considered possible influencing factors, such that in the absence of prosodic cues, gestures can serve to disambiguate a sentence (Kamiya & Guo, 2024). In written discourse, Apresjan (2019) demonstrates that readers rely on pragmatic cues in the form of lexical markers to infer plausible interpretations. Thus, ambiguities arise from the structure of the utterance, but this body of work highlights the pivotal role that cognitive mechanisms and other linguistic cues can play in the processing of scope ambiguities.

3. Quantifier Scope Ambiguities in Bilingualism Research

Traditional linguistic theory has relied on evidence from native monolingual speakers, but testing bilinguals with diverse linguistic experiences provides valuable insights into language phenomena that may not emerge from monolingual data alone (Benmamoun et al., 2013; Cheng et al., 2021; Satterfield, 1999). In addition, research in bilingualism demonstrates that even when surface structures appear similar, bilinguals do not always perform like monolinguals (Arredondo et al., 2019). Instead, they can exhibit unique linguistic patterns that vary across linguistic interfaces (e.g., Rothman, 2009; Benmamoun et al., 2013, for a review).

Quantification and scope ambiguities are complex interface-based phenomena that remain underexplored in heritage bilingualism research, alongside other phenomena beyond the sentence level (Jegerski & Sekerina, 2021; Polinsky, 2018). Despite this gap, several seminal studies have informed our theoretical understanding of the cognitive and syntactic strategies bilinguals use to navigate quantification. For instance, Barberán Recalde (2019) found that monolingual and bilingual children exhibit similar developmental patterns in acquiring certain quantifiers. The study, which focused on Basque/Spanish bilingual children processing weak quantifiers like algunos and batzuk (‘some’), used a picture-matching task to reveal that bilingual children’s interpretations aligned with those of their monolingual peers and were consistent across languages. Neither the children’s linguistic profiles nor language dominance impacted these interpretations, leading Barberán Recalde (2019) to conclude that bilingual children do not show evidence of convergence or divergence when acquiring weak quantifiers. Instead, results suggest that early bilingual children develop unified conceptual representations of quantifiers across both languages, processing them through underlying concepts rather than lexical items, thus supporting a long line of theories of a shared conceptual system in bilinguals (Francis, 1999).

Looking into adult bilinguals, Sekerina and Sauermann (2015) investigated how heritage Russian/English bilingual adults interpret quantified sentences, focusing on cognitive processing. They used a visual world eye-tracking study to analyze responses to structures like ‘Is every alligator in a bathtub?’. When presented with a scenario in which there were four bathtubs but only three alligators (i.e., the overexhaustive condition), although the sentence is true, some speakers might conclude that the sentence is false. While monolingual adults performed perfectly across conditions, bilinguals made more errors in Russian, their heritage language (20% q-spreading errors), compared to monolingual Russian speakers (5% q-spreading errors). However, these bilinguals looked longer at extra bathtubs during the verb region, indicating processing difficulty. They did not show these errors in their dominant language of English, in which they performed at ceiling accuracy. These results suggest the occurrence of cognitive overload in bilinguals when processing conflicting information in the less-dominant heritage language.

Scontras et al. (2017) further examined quantifier-scope ambiguities among English-dominant, heritage bilinguals of Mandarin Chinese and English, compared to native English and Mandarin monolinguals. Using a gradient truth-value judgment task, they tested sentences like ‘Every shark attacked a pirate,’ which is ambiguous in English but not in Mandarin. Heritage bilinguals rated the inverse scope interpretation in Mandarin sentences higher than monolingual Mandarin speakers and gave lower ratings for the inverse scope in English than monolingual English speakers. While these differences could initially indicate cross-linguistic influence, Scontras et al. (2017) attributed them to a “yes-bias” in heritage speakers, who tend to rate ungrammatical sentences higher. Further analysis of the data indicated that heritage Mandarin speakers do not permit inverse scope in Mandarin and surprisingly, not in English either. The authors interpreted this outcome as evidence for a simplification strategy in heritage speakers’ grammars, whereby the system with less complex scope prevails in both languages. In other words, they suggest that bilinguals, namely heritage speakers, make use of the simpler or more economical linguistic system when managing multiple languages.

Ronai (2018) conducted a follow-up study to Scontras et al. (2017), investigating quantifier-scope ambiguities in English-dominant heritage bilinguals of Hungarian and Hungarian-dominant heritage bilinguals of English. Using a gradient truth-value judgment task similar to that of Scontras et al. (2017), the study found that English-dominant heritage bilinguals of Hungarian did not transfer scope interpretations from English to the scope-rigid Hungarian. However, Hungarian-dominant bilinguals did not accept inverse interpretations in English, indicating that English, their heritage language, became scope-rigid like Hungarian. Taken together, like Scontras et al. (2017), Ronai (2018) argues that due to processing constraints, heritage speakers default to a simplification of the scope system in both languages, favoring the scope-rigid system. Both Ronai (2018) and Scontras et al. (2017) tested proficiency/dominance and found no effect. Additionally, Ronai (2018) tested for directionality and found that neither of these factors are crucial for the outcome of this particular phenomenon.

Unlike the previous bilingual studies, Shen and Chen (2022), Chen and Huan (2023) and Chen and Hester (2024) found that convergence and simplification of scope interpretations is not always the default outcome in heritage grammars. These three studies examined how Chinese/English, Tibetan/Chinese and Korean/Chinese heritage bilinguals (respectively) interpret quantifier-negation constructions (such as ‘All teachers did not use Sandy’s car’) that are ambiguous in English, Tibetan and Korean but not Chinese. The studies all involved a similar sentence-picture truth-value judgment task where participants were presented with a series of sentences and pictures (reflecting either the surface or inverse interpretation) and participants’ task was to determine whether the two matched. In all three studies, the results revealed three distinct groups of bilinguals, regardless of language combination: one group allowing inverse scope in both languages, a second group prohibiting it in both languages, and a third group successfully distinguishing between the two languages’ allowance of inverse scope. Importantly, the results were not language-specific. Post hoc analyses revealed that proficiency levels did not correlate with these groupings. While the studies’ findings do not reveal a clear-cut pattern, Chen and Huan (2023) point out that the results are nevertheless commonly observed in heritage bilingualism research: some heritage speakers attain monolingual-like knowledge of a given phenomenon, and sometimes they find a compromise. Shen and Chen (2022), Chen and Huan (2023) and Chen and Hester (2024) conclude that at least two of the three strategies seem to reflect a pattern of minimizing syntactic differences between the bilinguals’ two languages. Furthermore, having different strategies in the two languages could be seen as cognitively efficient in some way, as it minimizes the need to impose constraints on one language to maintain consistency across both.

Across all studies reviewed, there is consensus that heritage bilinguals may experience cognitive overload or added difficulty when processing conflicting linguistic information in their two languages (Sekerina & Sauermann, 2015). As a result, bilinguals often process scope ambiguities similarly in both languages (Chen & Hester, 2024; Chen & Huan, 2023; Ronai, 2018; Scontras et al., 2017; Shen & Chen, 2022), but different patterns emerge regarding which pattern prevails. Scontras et al. (2017) and Ronai (2018) suggest that because heritage bilinguals tend to simplify syntactic complexities, the more rigid scope system prevails in both languages. However, Shen and Chen (2022), Chen and Huan (2023), and Chen and Hester (2024) found that reducing syntactic complexity across a bilingual’s two languages can lead to a broader range of strategies, which does not always favor the rigid scope system. On the other hand, Barberán Recalde (2019) proposes a shared conceptual system for processing quantifiers, which may explain consistent patterns across both languages. This aligns with the findings of Scontras et al. (2017) and Ronai (2018). The shared conceptual system proposal could also explain why bilinguals sometimes exhibit unique parameters, as partially exemplified by results in Shen and Chen (2022), Chen and Huan (2023), and Chen and Hester (2024). Interestingly, proficiency and dominance seem to have little impact, suggesting quantifier processing might in fact depend more on how they were acquired than on proficiency. Further research is needed to thoroughly investigate these distinct patterns and fully explore the associated hypotheses.

Importantly, while all studies involve some form of scope ambiguities, the first two focus on quantifier-quantifier constructions, whereas the others deal with quantifier-negation sentences, which may limit the comparability of findings. Additionally, methodological differences impact how the results are interpreted and further affect their equivalence. Although all studies use a version of the truth-value judgment task, a commonly employed method for studying this phenomenon (Crain & McKee, 1985), they have slight variations. For instance, Scontras et al. (2017) and Ronai (2018) use a gradient response scale (1–7) to collect judgments, while Shen and Chen (2022), Chen and Huan (2023), and Chen and Hester (2024) only gather yes/no or true/false responses. This difference raises questions about whether the results in these latter studies might have benefited from a more nuanced evaluation system, or the reverse, whether the use of a gradient scale complicates the assessment of this phenomenon.

4. Quantifier Scope in Spanish and English

Quantifier scope ambiguities in English have been extensively documented (Anderson, 2004; Aoun & Li, 1989; Crain & Thornton, 1998; May, 1977, 1985; O’Grady, 2008). In addition to structures like those in the examples above, significant research has also investigated scope ambiguities that arise from the interaction of quantifiers and negation (9). In contrast, research on quantifier scope ambiguities in Spanish has been relatively limited. Although theoretical work supports the existence of these phenomena in Spanish, empirical findings are limited, and their nature and processing in Spanish is not yet fully understood (Fábregas, 2018; Gutiérrez Rexach, 1995; Jaeggli, 1991; Song et al., 2021).

(9)	Lucy didn’t read every book
	(a)	Surface scope (¬ > ∀): Lucy read some of the books, but not all of them
	(b)	Inverse scope (∀ > ¬): Lucy read none of the books

For example, Song et al. (2021) examined the role of conversational goals in interpreting ambiguous quantifier-negation sentences, such as their example sentence (10), across multiple languages, including English and Spanish. Their study found that inverse interpretations are permissible in Spanish and that both languages show similar patterns of preferred interpretation when conversational goals are accounted for. In addition, Barberán Recalde (2017) examined scope ambiguities involving the quantifier todo/a/s (‘all’) and negation in Spanish, as in their example sentence (11). The study found that both adult and child participants generally preferred the “some” interpretation (i.e., the surface interpretation) over the “none” interpretation (i.e., the inverse interpretation), suggesting that Spanish may be less permissive in allowing inverse interpretations. Taken together, the existing research highlights potential differences in interpretive preferences and constraints compared to English but more empirical data are needed to assess the robustness and limits of these interpretations.

(10)	Every butterfly didn’t go to the city/Todas las mariposas no fueron a la ciudad
	(a)	Surface scope (∀ > ¬): None of the butterflies went to the city
	(b)	Inverse scope (¬ > ∀): Some but not all of the butterflies went to the city
(11)	No todas las manzanas están en las cajas (‘Not all apples are in the boxes’)
	(a)	Surface scope (¬ > ∀): Some but not all of the apples are in the boxes
	(b)	Inverse scope (∀ > ¬): None of the apples are in the boxes

Importantly, there is not always a one-to-one correspondence between Spanish and English determiners (Gutiérrez Rexach, 1995), which makes finding equivalent quantifiers challenging. For example, the Spanish quantifier cada overlaps semantically with the English every and each while every could also overlap with the Spanish todo/a/s (Jaeggli, 1991). These mismatches hint that scope ambiguities may manifest differently across the two languages and underscore important methodological considerations.

5. Current Study

5.1. Research Questions and Hypotheses

In this study, we explore how Mexican Spanish/U.S. English heritage adults interpret quantified sentences, such as ‘A person watered every plant.’ Both Spanish and English reportedly have quantifier scope ambiguities but there appear to be differences in preferences, particularly regarding the accessibility of the inverse interpretation in each language. Previous research suggests that bilinguals often process scope ambiguities similarly in both languages, either to reduce syntactic complexity or due to a shared conceptual system (see Section 3). In this study we examine whether similar processing strategies and preferences are reflected across Spanish and English, despite their shared classification as scope ambiguous languages. In addition to interpretation patterns, we examine response times to better understand not just the level of preference, but also whether cognitive load and processing time appear to be similar between the two languages. With these goals in mind, the following research questions are posed:

Research Question 1: Do Mexican Spanish/U.S. English heritage bilinguals show similar preferences for surface and inverse quantifier scope interpretations in each language? Hypothesis: In line with previous research, we hypothesize that surface interpretations will be more acceptable than inverse interpretations in both languages (e.g., Aoun & Li, 1989; Anderson, 2004; Barberán Recalde, 2017; Crain & Thornton, 1998; Fábregas, 2018; O’Grady, 2008; Song et al., 2021).

Research Question 2: Is there a correlation between response time and the acceptability of quantifier scope interpretations? Hypothesis: Based on previous research suggesting that surface interpretations are easier to process (Anderson, 2004; O’Grady & Lee, 2006), we hypothesize that there will be a negative correlation between response time and quantifier scope interpretations, i.e., the more acceptable an interpretation is, the lower the response time.

Research Question 3: Does language proficiency affect the acceptability of quantifier scope interpretations in Mexican Spanish/U.S. English heritage bilinguals? Hypothesis: Following previous research (Chen & Hester, 2024; Chen & Huan, 2023; Ronai, 2018; Scontras et al., 2017; Shen & Chen, 2022), we predict that language proficiency will not significantly affect the results of this study.

5.2. Participants

60 Mexican Spanish/U.S. English heritage bilinguals3 from the Chicagoland area were recruited to participate in the study, by flyers, social media advertisements and word of mouth. To qualify, participants all met a series of linguistic and demographic criteria to control for possible variation. That is, all participants must have been Spanish/English bilinguals that learned both languages before the age of 10 (mean Spanish AoA = 0; mean Eng AoA = 3). To control for dialectal variation in Spanish, all participants were speakers of a Mexican dialect of Spanish. To control for English dialect, all participants must have reported living in Chicago for over 10 years continuously. Participants of various bilingual proficiencies were encouraged to participate. At the time of the experiment, all participants were between 18 and 30 years old (mean age = 25.48, SD = 2.65).

5.3. Methodology

A truth-value judgment task, commonly used in similar studies (e.g., Ronai, 2018; Scontras et al., 2017), was used to collect participants’ intuitions. In this task, 16 experimental stimuli were designed along with 32 fillers, maintaining a 1:2 ratio. The 16 critical items (CIs) consist of ambiguous sentences with the quantifiers a and every or un/una and cada. The CIs all follow the same structure (see Type 1 in

Table 1. Overview of Experimental Stimuli.

Type	Sentence	Ambiguous?	Number	Type
1	A person watered every plant Una persona regó cada planta	Yes	16	CI
2	A person did not do exercise Una persona no hizo ejercicio	No	8	Filler
3	A woman rode a skateboard Una persona montó una patineta	No	8	Filler
4	A man did not peel every orange Una persona no peló cada naranja	Yes	16	Filler

) and are modeled after stimuli used in Scontras et al. (2017), but with the quantifier a in a higher syntactic position than every in the surface structure. The decision to only include the a-every structure stems from the fact that, in this structure, the inverse interpretation does not entail the surface interpretation, whereas in the alternative every-a structure, it does (see Keenan, 2002; Scontras et al., 2017).

In line with previous discussions, we acknowledge that there are often no perfect equivalencies between Spanish and English quantifiers. In Mexican Spanish, the two most plausible translations for every are cada and todo/a/s (Jaeggli, 1991). As discussed in Shi (2024), the difference between the two is that cada tends to favor a distributive interpretation, i.e., a scenario in which each member of a set carries out the action individually (e.g., every person reading a different book), whereas todo/a/s leans more toward a collective interpretation (e.g., every person reading the same book). The English quantifier every is often described as favoring the distributive interpretation (Tunstall, 1998), making cada the semantically closer translation. Even so, prior to creating our stimuli, we consulted four Spanish/English bilingual speakers with profiles similar to the participants, asking them to translate three sample stimulus sentences into Spanish. All translated the quantifier every as cada, though subsequent discussion confirmed that other translations are also possible. Based on their initial responses/intuitions and semantic parallels, we selected cada as the Spanish equivalent for every.

The filler sentences were divided into three conditions: Type 2, Type 3, and Type 4. Type 2 and 3 items were non-ambiguous sentences4 with the quantifier a in the same position as in the CI. In addition, Type 2 sentences also contained negation. Type 4 fillers were a different type of scope-ambiguous sentences, where the ambiguity arose from the interaction between the quantifier every and negation. Additionally, 4 practice items were created resembling the structure of the stimuli. In total, there were 4 practice items and 48 stimuli items, with 32 ambiguous and 16 non-ambiguous sentences (see

Table A1. Sentence Stimuli in English and Spanish.

Type	Item	Sentence
Practice Items	A	A person prepared a salad Una persona preparó una ensalada
	B	A person baked a cake Una persona horneó un pastel
	C	A child drew every drawing Un niño dibujó cada dibujo
	D	A person did not laugh Una persona no se rió
Critical Items: Ambiguous sentences with quantifiers ‘a’ and ‘every’ (Type 1)	1	A person watered every plant Una persona regó cada planta
	2	A person fed every chicken Una persona alimentó a cada gallina
	3	A person carried every plant Una persona cargó cada planta
	4	A person harvested every vegetable Una persona cosechó cada verdura
	5	A woman planted every plant Una mujer plantó cada planta
	6	A person took every pumpkin Una persona se llevó cada calabaza
	7	A woman read every book Una mujer leyó cada libro
	8	A person slept on every pillow Una persona se durmió en cada almohada
	9	A woman used every computer Una mujer usó cada computadora
	10	A woman wrapped every gift Una mujer envolvió cada regalo
	11	A man decorated every cake Un hombre decoró cada pastel
	12	A person drank every juice Una persona se tomó cada jugo
	13	A woman baked every dessert Una mujer horneó cada postre
	14	A woman made every pizza crust Una mujer hizo cada masa de pizza
	15	A person ate every snack Una persona se comió cada botana
	16	A woman cooked every stew Una mujer cocinó cada guisado
Fillers: Unambiguous sentences with negation (Type 2)	17	A person did not do exercise Una persona no hizo ejercicio
	18	A person did not play soccer Una persona no jugó fútbol
	19	A woman did not ride a bike Una mujer no montó una bicicleta
	20	A person did not eat popcorn Una persona no comió palomitas
	21	A woman did not clean the house Una mujer no limpió la casa
	22	A woman did not play guitar Una mujer no tocó la guitarra
	23	A person did not read a book Una persona no leyó un libro
	24	A man did not walk a dog Un hombre no paseó a un perro
Fillers: Unambiguous sentences with quantifier ‘a’ (Type 3)	25	A man watered flowers Un hombre regó las flores
	26	A person fed the pig Una persona alimentó al puerco
	27	A person ate cupcakes Una persona comió pastelitos
	28	A woman rode a skateboard Una mujer montó una patineta
	29	A person ironed the clothes Una persona planchó la ropa
	30	A woman used a desk Una mujer usó un escritorio
	31	A woman baked cookies Una mujer horneó galletas
	32	A person planted carrots Una persona sembró zanahorias
Fillers: Ambiguous sentences with negation (Type 4)	33	A person did not fill every glass Una persona no llenó cada vaso
	34	A person did not slice every pizza Una persona no cortó cada pizza
	35	A person did not break every vase Una persona no quebró cada florero
	36	A man did not peel every orange Un hombre no peló cada naranja
	37	A person did not clean every couch Una persona no limpió cada sofá
	38	A man did not hang every painting Un hombre no colgó cada cuadro
	39	A person did not drink every cup of milk Una persona no se tomó cada vaso de leche
	40	A person did not carry every box Una persona no cargó cada caja
	41	A woman did not cut every flower Una mujer no cortó cada
	42	A man did not smash every can Un hombre no aplastó cada lata
	43	A woman did not fold every towel Una mujer no dobló cada toalla
	44	A man did not vacuum every mat Un hombre no aspiró cada tapete
	45	A woman did not frame every picture Una mujer no enmarcó cada foto
	46	A person did not pick up every book Una persona no recogió cada libro
	47	A woman did not feed every chicken Una mujer no alimentó a cada gallina
	48	A person did not spill every coffee cup Una persona no tumbó cada taza de café

in Appendix A for the complete list of sentence stimuli).

All sentences described familiar scenarios, such as cooking at home, watering plants in a yard or farm, or riding a bike at the park, to ensure contextual accessibility. The 48 stimuli featured a similar distribution of the subjects ‘a woman’, ‘a man’, and ‘a person’. Sentences were initially written in English, then translated into Spanish by a native speaker of Mexican Spanish and reviewed by two additional native speakers for accuracy. All sentence stimuli were recorded in a sound-attenuated booth by a Spanish/English bilingual speaker with a similar background to the participants. To minimize prosodic cues that could influence sentence interpretation (see Section 2), the speaker was trained to maintain a steady intonation. Recordings were monitored and edited using the software Audacity (Audacity Team, 2021) to correct prosodic irregularities and segment the audios.

Using Canva (2023), a graphic design software, two visual representations were created for each stimulus sentence (see

Table A2. Visual Stimuli.

Item	Surface/True Representation	Inverse/False Representation
A		Not Available
B	Not Available
C		Not Available
D	Not Available
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48

in Appendix A for the complete set of visual stimuli). For scope-ambiguous sentences, one image was created for each possible interpretation (see Figure 1). For non-ambiguous sentences, one image accurately represented the sentence, while the other presented an inaccurate version (e.g., a person running, not riding a bike). Both image pairs were made as similar as possible, with only trivial elements related to the interpretation being manipulated. For images requiring more than one person, such as those for the inverse representations of the CIs (see Figure 1), three people were consistently depicted performing the action.

The truth-value judgment task presented participants with auditory sentence–picture pairs (the sentences were never shown in written form). For each pair, participants answered the question “How well does the sentence describe the image?” using a 5-point response scale (1 = Very Poorly, 2 = Somewhat Poorly, 3 = Not Sure, 4 = Somewhat Well, 5 = Very Well). A 5-point response scale was chosen to balance cross-study comparability and methodological considerations. Prior studies on scope ambiguities (e.g., Ronai, 2018; Scontras et al., 2017) commonly use 7-point, end point-labeled scales. At the same time, research shows that 5- and 7-point scales yield highly comparable results (Dawes, 2008), and that fully labeled 5-point scales can be easier for participants to interpret and use (Revilla et al., 2014). Taking into account the format of our task (completed online), a 5-point scale was considered the most practical and reliable option.

Upon entering each trial screen, the auditory sentence played automatically and could not be replayed until 30 seconds passed without a response, after which it replayed once. Simultaneously, the image and response scale were displayed. Participants were instructed to respond as quickly as possible to reflect their initial intuitions. The order of stimuli was randomized, and each sentence was randomly paired with either a surface or inverse visual representation (for ambiguous items), or with faithful or false depictions (for unambiguous fillers). All participants saw the same practice items. The task followed a 2 × 2 factorial design, manipulating two independent variables: language (Spanish vs. English) and interpretation (surface vs. inverse).

In addition to the experimental task, participants completed both the Spanish and English versions of LexTALE, a lexical decision task in which they judged whether each item (Span N = 90; Eng N = 60) was a real word in the target language (Izura et al., 2014; Lemhöfer & Broersma, 2012). The LexTALE was included as a proxy for general language proficiency, supported by previous validation studies (Bonvin et al., 2023). The original stimuli and instructions were used, with one modification: all items were presented aurally rather than in written form. Recordings followed the same procedure as the experimental stimuli and were produced by the same bilingual speaker. Scores were calculated according to Lemhöfer and Broersma’s (2012) guidelines using the following formula: LexTALE score = (total real words correct/real words × 100) + (total nonwords correct/nonwords × 100)/2. Finally, participants also completed the Bilingual Language Profile (BLP; Birdsong et al., 2012), a self-report questionnaire assessing language history, usage, proficiency, and exposure. The BLP generates a language dominance score ranging from −218 to +218, with higher scores indicating English dominance and negative scores indicating Spanish dominance.

5.4. Procedure

All tasks were administered using Qualtrics, an online survey and experimental platform. To capture judgments in both languages, participants were randomly assigned to complete the experimental task in either Spanish (N = 30) or English (N = 30). The experiment was conducted in a single online session, and participants were encouraged to complete it in a calm, quiet environment with minimal distractions. The task order was as follows: participants first completed the truth-value judgment task (in their assigned language), followed by the Spanish LexTALE, the English LexTALE, and finally, the Bilingual Language Profile in their language of choice. The full session lasted approximately 45–60 min, and participants were compensated for their time. This experiment was conducted under the University of Michigan’s IRB protocol HUM00229813.

6. Results

We begin with an analysis of language dominance and proficiency. The average BLP score across all participants was −27.48 (SD = 33.26), indicating a balanced but slightly Spanish-dominant group. Participants who completed the task in English (N = 30) had an average BLP score of −26.09 (SD = 35.86), while those in the Spanish condition (N = 30) averaged −28.87 (SD = 30.99). LexTALE scores, calculated per Lemhöfer & Broersma (2012), revealed slightly higher proficiency in English overall (M = 81.58, SD = 9.97) than in Spanish (M = 72.87, SD = 9.14). The English-task group scored 79.96 (SD = 12.03) on the English LexTALE and 73.54 (SD = 9.60) on the Spanish version. The Spanish-task group scored 83.20 (SD = 7.10) in English and 72.19 (SD = 8.81) in Spanish. While these scores suggest slightly stronger English proficiency, they are relatively close and align with the BLP in indicating balanced bilingualism. Both groups had comparable dominance and proficiency profiles, supporting the validity of cross-group comparisons.

Before conducting further analyses, mean ratings were calculated for each interpretation type (surface vs. inverse) in Spanish and English. Filler Types 2 and 3 included unambiguous sentences such as ‘A person did not do exercise/Una persona no hizo ejercicio’ and ‘A woman rode a skateboard/Una persona montó una patineta,’ respectively. The sentences were paired with images representing either accurate or inaccurate representations of the sentence (e.g., a person riding a bike, not a skateboard). Although labeled for consistency with experimental items in

Table 2. Analysis of Ratings for Filler Items.

Type	Sentence	Spanish Surface	Spanish Inverse	English Surface	English Inverse
2	A person did not do exercise Una persona no hizo ejercicio	4.44, SD = 0.99	1.23, SD = 0.67	4.10, SD = 1.19	1.42, SD = 0.84
3	A woman rode a skateboard Una persona montó una patineta	4.74, SD = 0.56	1.27, SD = 0.46	4.46, SD = 0.89	1.27, SD = 0.59
4	A man did not peel every orange Una persona no peló cada naranja	4.23, SD = 0.79	4.04, SD = 0.81	4.36, SD = 0.56	3.97, SD = 0.89

, ‘surface’ here aligns with accurate or true representations, and ‘inverse’ with inaccurate or false representations. For Type 2, accurate interpretations were rated highly in both Spanish (M = 4.44, SD = 0.99) and English (M = 4.10, SD = 1.19), while inaccurate interpretations were rated much lower (Spanish: M = 1.23, SD = 0.67; English: M = 1.42, SD = 0.84). Similarly, for Type 3, accurate interpretation ratings were high (Spanish: M = 4.74, SD = 0.56; English: M = 4.46, SD = 0.89), and inaccurate ratings were low (Spanish: M = 1.27, SD = 0.46; English: M = 1.27, SD = 0.59). These results confirm that participants understood the task and used the response scale consistently.

Type 4 filler items consisted of ambiguous sentences, such as ‘A man did not peel every orange/Un hombre no peló cada naranja’ where the interaction between the quantifier every and negation created the ambiguity. Ratings here were more balanced: surface interpretations in Spanish and English received mean scores of 4.23 (SD = 0.79) and 4.36 (SD = 0.56), respectively, while inverse interpretations scored 4.04 (SD = 0.81) in Spanish and 3.97 (SD = 0.89) in English. These results (summarized in Table 2) suggest that although surface interpretations were rated slightly higher in this condition, inverse interpretations were also generally considered acceptable in both languages.

Looking at the critical items (e.g., ‘A person watered every plant/Una persona regó cada planta’), participants showed a preference for the surface interpretation in both languages. In Spanish, the surface interpretation received a mean rating of 4.19 (SD = 0.56), compared to 2.92 (SD = 1.12) for the inverse one. Response times were also faster for surface interpretations (M = 6.97 s, SD = 2.70) than for inverse items (M = 8.75 s, SD = 3.04). In English, surface interpretations were rated 4.14 (SD = 0.66), and inverse 3.18 (SD = 0.87), with corresponding response times of 6.67 s (SD = 1.80) and 7.81 s (SD = 2.48). As summarized in

Table 3. Analysis of Rating and Response Time for Critical Items.

Measure	Spanish Surface	Spanish Inverse	English Surface	English Inverse
Rating	4.19, SD = 0.56	2.92, SD = 1.12	4.14, SD = 0.66	3.18, SD = 0.87
Response Time (s)	6.97, SD = 2.70	8.75, SD = 3.04	6.67, SD = 1.80	7.81, SD = 2.48

, these results indicate that surface interpretations are generally preferred and processed more quickly, while inverse interpretations, particularly in Spanish, receive lower ratings and are associated with longer and more variable response times. Mean ratings of individual stimuli items can be found in

Table A3. Mean Ratings of Individual Critical Items in English and Spanish.

Item	Sentence	Surface/True	Inverse/False
1	A person watered every plant Una persona regó cada planta	M = 4.80 SD = 0.41 M = 1.07; SD = 0.26	M = 3.27; SD = 1.33 M = 2.47; SD = 1.60
2	A person fed every chicken Una persona alimentó a cada gallina	M = 4.75; SD = 0.45 M = 4.80; SD = 0.41	M = 3.79; SD = 1.12 M = 2.20; SD = 1.32
3	A person carried every plant Una persona cargó cada planta	M = 3.40; SD = 1.50 M = 4.13; SD = 1.06	M = 2.67; SD = 1.29 M = 3.47; SD = 1.64
4	A person harvested every vegetable Una persona cosechó cada verdura	M = 4.06; SD = 1.29 M = 4.00; SD = 1.20	M = 3.43; SD = 1.40 M = 3.00; SD = 1.41
5	A woman planted every plant Una mujer plantó cada planta	M = 3.88; SD = 1.22 M = 4.60; SD = 0.51	M = 3.62; SD = 1.39 M = 3.20; SD = 1.78
6	A person took every pumpkin Una persona se llevó cada calabaza	M = 4.60; SD = 0.63 M = 4.93; SD = 0.26	M = 2.80; SD = 1.52 M = 2.60; SD = 1.45
7	A woman read every book Una mujer leyó cada libro	M = 3.33; SD = 1.29 M = 3.73; SD = 1.03	M = 3.47; SD = 0.83 M = 2.47; SD = 1.36
8	A person slept on every pillow Una persona se durmió en cada almohada	M = 4.62; SD = 0.87 M = 4.07; SD = 1.49	M = 3.59; SD = 1.42 M = 4.00; SD = 1.31
9	A woman used every computer Una mujer usó cada computadora	M = 4.20; SD = 1.08 M = 3.87; SD = 1.41	M = 2.47; SD = 1.46 M = 2.33; SD = 1.72
10	A woman wrapped every gift Una mujer envolvió cada regalo	M = 4.80; SD = 0.41 M = 4.80; SD = 0.56	M = 2.93; SD = 1.71 M = 2.73; SD = 1.71
11	A man decorated every cake Un hombre decoró cada pastel	M = 4.53; SD = 0.64 M = 4.80; SD = 0.56	M = 3.67; SD = 1.59 M = 3.33; SD = 1.76
12	A person drank every juice Una persona se tomó cada jugo	M = 4.27; SD = 0.88 M = 4.47; SD = 1.06	M = 3.00; SD = 1.51 M = 3.20; SD = 1.90
13	A woman baked every dessert Una mujer horneó cada postre	M = 4.54; SD = 0.52 M = 4.20; SD = 1.21	M = 3.11; SD = 1.37 M = 2.87; SD = 1.68
14	A woman made every pizza crust Una mujer hizo cada masa de pizza	M = 3.93; SD = 1.38 M = 4.60; SD = 0.83	M = 3.13; SD = 1.71 M = 3.13; SD =1.85
15	A person ate every snack Una persona se comió cada botana	M = 3.69; SD = 1.30 M = 4.47; SD = 0.83	M = 4.00; SD = 1.11 M = 3.00; SD = 1.85
16	A woman cooked every stew Una mujer cocinó cada guisado	M = 3.71; SD = 1.20 M = 4.60; SD = 0.63	M = 3.00; SD = 1.55 M = 2.60; SD = 1.59

of Appendix B.

To assess differences in ratings across languages and interpretations, four t-tests were conducted. Two-tailed, paired-sample t-tests were performed to compare surface versus inverse ratings within each language group (Spanish and English). Statistical significance was evaluated at an alpha level of 0.05. Exact p-values are reported, together with effect sizes and 95% confidence intervals. The results indicated that in Spanish, surface ratings were rated significantly higher than inverse ratings, t(29) = 5.75, p < 0.001 ***, with a large effect size (d = 1.43, 95% CI [0.72, 2.14]). Similarly, in English, surface ratings were significantly greater than inverse ratings, t(29) = 4.79, p < 0.001 ***, also with a large effect size (d = 1.24, 95% CI [0.55, 1.92]).

To compare ratings across the Spanish and English groups, two-tailed, independent-sample t-tests assuming equal variances were conducted. These tests revealed no significant differences between the surface ratings in Spanish and English, t(58) = 0.31, p = 0.754, with a negligible effect size (d = 0.08, 95% CI [−0.44, 0.60]). Similarly, no significant difference was found between the inverse ratings in Spanish and English, t(58) = −1.03, p = 0.308, with a small effect size (d = −0.27, 95% CI [−0.78, 0.25]). Taken together, the results (summarized in

Table 4. Results of t-tests Comparing Surface and Inverse Ratings of Critical Items.

Comparison	t(df)	p	d [95% CI]	Interpretation
Within-language (paired)
Spanish: Surface vs. Inverse	5.75 (29)	<0.001 ***	1.43 [0.72, 2.14]	Lg. effect
English: Surface vs. Inverse	4.79 (29)	<0.001 ***	1.24 [0.55, 1.92]	Lg. effect
Across languages (independent)
Surface: Spanish vs. English	0.31 (58)	0.754	0.08 [−0.44, 0.60]	Neg. effect
Inverse: Spanish vs. English	−1.03 (58)	0.308	−0.27 [−0.78, 0.25]	Sm. effect

Note. p < 0.05 = *, p < 0.01 = **, p < 0.001 = ***.

and represented in Figure 2) show that surface interpretations of the critical items received higher acceptability ratings than inverse interpretations in both Spanish and English, and no significant difference was observed across languages.

To examine response time differences, two-tailed, paired-sample t-tests compared surface and inverse interpretations within each language. Results showed significantly longer response times for inverse interpretations than surface interpretations in Spanish, t(29) = −3.57, p = 0.001 ***, with a medium effect size (d = −0.62, 95% CI [−1.00, −0.24]). In English, inverse interpretations were also associated with significantly longer response times than surface interpretations, t(29) = −2.65, p = 0.013 *, with a medium effect size (d = −0.51, 95% CI [−0.93, −0.10]).

To compare response times across the Spanish and English groups, two-tailed, independent-sample t-tests assuming equal variances were conducted. Results revealed no significant differences in surface response times across the two languages, t(58) = 0.50, p = 0.622, with a negligible effect size (d = 0.13, 95% CI [−0.39, 0.65]). Nor were there significant differences in the inverse response times across language groups, t(58) = 1.32, p = 0.191, with a small effect size (d = 0.34, 95% CI [−0.18, 0.86]). These results (summarized in

Table 5. Results of t-tests Comparing Surface and Inverse Response Times of Critical Items.

Comparison	t(df)	p	d [95% CI]	Interpretation
Within-language (paired)
Spanish: Surface vs. Inverse	−3.57 (29)	0.001 ***	−0.62 [−1.00, −0.24]	Med. effect
English: Surface vs. Inverse	−2.65 (29)	0.013 *	0.51 [−0.93, −0.10]	Med. effect
Across languages (independent)
Surface: Spanish vs. English	0.50 (58)	0.622	0.13 [−0.39, 0.65]	Neg. effect
Inverse: Spanish vs. English	1.32 (58)	0.191	0.34 [−0.18, 0.86]	Sm. effect

Note. p < 0.05 = *, p < 0.01 = **, p < 0.001 = ***.

and represented in Figure 3) suggest that inverse interpretations are associated with longer response times and no significant differences are observable across languages.

Pearson’s correlation analyses were conducted to further assess response times and determine whether there is a relationship between how acceptable a sentence is (i.e., higher or lower rating) and the time taken to respond. According to the results, summarized in

Table 6. Correlations Between Ratings and Response Time of Critical Items.

Interpretation Type	r	p	95% CI
Spanish Surface	−0.24	0.201	[−0.55, 0.13]
Spanish Inverse	−0.09	0.624	[−0.44, 0.28]
English Surface	−0.56	0.001 ***	[−0.76, −0.24]
English Inverse	−0.03	0.856	[−0.39, 0.33]

Note. p < 0.05 = *, p < 0.01 = **, p < 0.001 = ***.

, for Spanish surface items, the correlation was negative (i.e., higher rated items are associated with shorter response times) but not significant, r(28) = −0.24, p = 0.201, 95% CI [−0.55, 0.13]. For Spanish inverse items, the correlation was near zero and not significant, r(28) = −0.09, p = 0.624, 95% CI [−0.44, 0.28]. In English, there was a significant negative correlation between surface interpretation ratings and response time, r(28) = −0.56, p = 0.001 ***, 95% CI [−0.76, −0.24]; however, there was no significant correlation between English inverse interpretation ratings and response time, r(28) = −0.03, p = 0.856, 95% CI [−0.39, 0.33]. These correlations are represented in Figure 4.

To further examine the relationship between ratings and response time, correlation analyses were conducted using Type 4 filler sentences (‘A man did not peel every orange/Un hombre no peló cada naranja’), which also involve scope ambiguity. In Spanish, both surface, r(28) = −0.23, p = 0.224, 95% CI [−0.54, 0.14], and inverse, r(28) = −0.15, p = 0.434, 95% CI [−0.48, 0.22], interpretations showed weak, non-significant negative correlations. In contrast, English surface, r(28) = −0.50, p = 0.005 **, 95% CI [−0.73, −0.17], and inverse, r(28) = −0.48, p = 0.007 **, 95% CI [−0.72, −0.15], interpretations showed significant negative correlations, indicating that higher ratings were associated with shorter response times. These results (summarized in

Table 7. Correlations Between Ratings and Response Time of Filler Items.

Interpretation Type	r	p	95% CI
Spanish Surface	−0.23	0.224	[−0.54, 0.14]
Spanish Inverse	−0.15	0.434	[−0.48, 0.22]
English Surface	−0.50	0.005 **	[−0.73, −0.17]
English Inverse	−0.48	0.007 **	[−0.72, −0.15]

Note. p < 0.05 = *, p < 0.01 = **, p < 0.001 = ***.

and represented in Figure 5), along with the findings in Table 6, suggest that a relationship between ratings and response times appears in English but not in Spanish.

Pearson’s correlation analyses were also conducted to examine the relationship between ratings and language proficiency (as measured by LexTALE). No significant correlations were found. When the task was completed in Spanish, Spanish proficiency was not significantly correlated with surface ratings, r(28) = 0.02, p = 0.932, 95% CI [−0.35, 0.37], or inverse ratings, r(28) = 0.17, p = 0.377, 95% CI [−0.21, 0.50]. Similarly, English proficiency showed no significant correlation with surface ratings, r(28) = 0.07, p = 0.716, 95% CI [−0.30, 0.42], or inverse ratings, r(28) = −0.001, p = 0.996, 95% CI [−0.36, 0.36]. When the task was completed in English, Spanish proficiency was not significantly associated with surface ratings, r(28) = 0.19, p = 0.309, 95% CI [−0.18, 0.52], or inverse ratings, r(28) = 0.10, p = 0.599, 95% CI [−0.27, 0.44]. English proficiency also showed no significant correlations with surface ratings, r(28) = −0.05, p = 0.781, 95% CI [−0.41, 0.31], or inverse ratings, r(28) = −0.08, p = 0.677, 95% CI [−0.43, 0.29]. Taken together, these results suggest that language proficiency is not related to how participants rated surface or inverse interpretations in either language.

To corroborate the findings from the t-tests and correlation analyses of the critical items (CI), a linear mixed-effects model (lme4 package, Bates et al., 2015) was fit with acceptability rating as the dependent variable. Fixed effects included language (Spanish vs. English), interpretation type (surface vs. inverse), Spanish proficiency, English proficiency, and response time. Interpretation type was dummy coded with inverse as the reference level, and language with English as the reference level. Subject was included as a random intercept to account for subject-variation (i.e., individual differences in scale use) and avoid the need to use additional data normalization procedures, such as z-scoring (Baayen et al., 2008; Barr et al., 2013). The model yielded a singular fit, with the subject random-intercept variance estimated near zero, indicating minimal baseline variability across participants. The model’s marginal and conditional R² values were identical (R²m = 0.35, R²c = 0.35), meaning that most of the explained variance was attributable to fixed effects.

The analysis revealed a significant main effect of interpretation type, indicating that surface interpretations received higher ratings than inverse interpretations (Estimate = 1.04, SE = 0.16, t(66.18) = 6.62, p < 0.001 ***). Estimated marginal means were 4.13 [3.91, 4.34] for surface interpretations and 3.09 [2.87, 3.31] for inverse interpretations. This contrast corresponds to a large effect size (Cohen’s d = 1.26, 95% CI [0.88, 1.65]). No significant effects were observed for language (Estimate = −0.12, SE = 0.16, t(55.78) = −0.77, p = 0.44), Spanish proficiency (Estimate = −0.01, SE = 0.01, t(55.74) = −0.91, p = 0.37) and English proficiency (Estimate = 0.01, SE = 0.01, t(55.38) = 1.25, p = 0.22). Response time showed a trend toward significance, with shorter response times associated with higher ratings (Estimate = −0.06, SE = 0.03, t(90.79) = −1.85, p = 0.068). Together, these results support the conclusion that interpretation type significantly influences ratings, with surface interpretations consistently preferred across both languages and associated with faster response times.

7. Discussion

7.1. Interpretation of Results

The goal of this study was to examine how Mexican Spanish/U.S. English heritage bilingual adults interpret scope-ambiguous sentences in both languages, and whether similar processing strategies are used. Specifically, we assessed whether bilinguals show consistent interpretation patterns across languages, despite them being scope ambiguous, and how response times relate to these preferences. The results support our hypothesis for Research Question 1 that surface interpretations were preferred over inverse interpretations in both Spanish and English. Surface interpretations received higher ratings (Spanish: M = 4.19, SD = 0.56; English: M = 4.14, SD = 0.66) compared to inverse interpretations (Spanish: M = 2.92, SD = 1.12; English: M = 3.18, SD = 0.87), aligning with previous findings (e.g., Anderson, 2004; Barberán Recalde, 2017; Crain & Thornton, 1998; O’Grady, 2008; Song et al., 2021). T-tests confirmed these differences were significant in both languages (Spanish: t(29) = 5.75, p < 0.001 ***, English t(29) = 4.79, p < 0.001 ***).

Interestingly, no significant difference was found in ratings for surface and inverse interpretations across Spanish and English, aligning with previous findings that bilinguals process scope ambiguities similarly across languages (Ronai, 2018; Scontras et al., 2017). The absence of a significant difference between languages supports the hypothesis that bilinguals may rely on a shared pattern when interpreting scope ambiguities, irrespective of language. However, we cannot conclusively determine whether the similar patterns are due to simplification strategies (Ronai, 2018; Scontras et al., 2017) or a shared conceptual/linguistic system (Barberán Recalde, 2019), or both, given that the two possibilities are not mutually exclusive. Furthermore, drawing from Anderson’s Processing Scope Economy Principle (Anderson, 2004), it is also possible that the observed simplification is a universal processing strategy and not necessarily a bilingual induced effect. In any case, bilingualism offers unique opportunities to explore scenarios, such as the interaction between systems that are more and less ambiguous. This might lead us to attribute the simplification observed in this case to bilingualism, even though it could be a generally preferred processing strategy. More research is needed to explore these possibilities.

For Research Question 2, we hypothesized that response time would negatively correlate with acceptability ratings and the results partially supported this hypothesis. Surface interpretations had shorter response times than inverse ones in both languages (Spanish surface: 6.97 s, SD = 2.70; inverse: 8.75 s, SD = 3.04; English surface: 6.67 s, SD = 1.80; inverse: 7.81 s, SD = 2.48). In addition, response time showed a trend toward significance in the linear mixed-effects model, with shorter response times associated with higher ratings (Estimate = −0.06, SE = 0.03, t(90.79) = −1.85, p = 0.068). However, the only statistically significant correlation related to response times was a significant correlation between surface interpretation ratings and response times in English (r(28) = −0.56, p = 0.001 ***, 95% CI [−0.76, −0.24]). Further insight into this matter comes from filler sentence analyses (Type 4) which were also scope ambiguous constructions. In Spanish, no significant correlations were observed (surface: r(28) = −0.23, p = 0.224, 95% CI [−0.54, 0.14], inverse: r(28) = −0.15, p = 0.434, 95% CI [−0.48, 0.22]); but in English, both surface (r(28) = −0.50, p = 0.005 **, 95% CI [−0.73, −0.17]) and inverse (r(28) = −0.48, p = 0.007 **, 95% CI [−0.72, −0.15]) interpretations showed significant negative correlations. This outcome suggests that the response time-rating relationship was clearer in English and may reflect differences in processing between the two languages. One possibility is that English, the participants’ non-heritage language, involved less cognitive effort, therefore revealing the correlation more clearly, aligning with the findings of Sekerina and Sauermann (2015). Conversely, Spanish may have required more cognitive resources, obscuring this relationship.

However, according to the results from both the BLP and LexTALE, our study participants would be considered “balanced” bilinguals, thus raising several interesting questions and additional possibilities. On the one hand, there is no strong correlation between response times and ratings, and so the correlation found between the English surface interpretation and response times may have been an anomaly that does not warrant further exploration. Another possibility is that participants are in fact English-dominant, but the dominance and proficiency tests used in this study did not accurately capture the participants’ linguistic profiles (as previously explored by Solís-Barroso & Stefanich, 2019). Along this line, bilinguals may exhibit different dominance patterns across different linguistic tasks. For example, they might be dominant in Language X when it comes to lexical processing but might be more cognitively efficient or find it easier to process overall in Language Y. Thus, the extent to which speakers are relatively balanced and have high Spanish proficiency according to lexical decision tasks, does not necessarily negate that they might be more cognitively efficient when processing English overall.

While the current group of participants had relatively high proficiency in Spanish and English, as hypothesized for Research Question 3, language proficiency did not significantly affect scope interpretations. Correlation analyses between ratings and proficiency scores (as measured by LexTALE) showed no significant relationship between proficiency and ratings. Likewise, the linear mixed-effects model revealed no significant effects of these factors. These results corroborate previous work (Chen & Hester, 2024; Chen & Huan, 2023; Ronai, 2018; Scontras et al., 2017; Shen & Chen, 2022) and suggest that proficiency may not strongly influence scope interpretation preferences. The lack of a significant role for proficiency is important, given that proficiency typically plays a crucial role in bilingual language processing (Benmamoun et al., 2013; Grosjean, 2008; Rothman, 2009). An explanation could be that bilinguals may process scope ambiguities based on a unified linguistic system that is more or less stable, and not heavily subject to variations in line with proficiency levels. Studying bilinguals with a wider range of proficiencies could provide more insight into this matter.

7.2. Limitations and Future Research

Moving forward, several adjustments could improve the stimuli and study design. We acknowledge shortcomings in our materials, particularly regarding the translation of every/cada. Although we consulted four bilinguals with the same background as our participants to identify the most appropriate Spanish equivalent of every, further refinement is needed. For example, future work could involve pilot testing with a larger participant pool or consulting corpus data to better support the translation choice. Since every overlaps semantically with both cada and todo/a/s (Jaeggli, 1991), it could also be of interest for future research to compare both of these quantifiers in scope ambiguous sentences (as opposed to our study which only looked at cada) to further understand similarities and differences between the two quantifiers.

We also note a possible limitation in our visual stimuli, specifically in the images associated with the inverse interpretation; that is, situations in which more than one person was performing an action (e.g., a person watering every plant, but for every plant there is a different person watering it). The scenarios designed were not entirely controlled, as they sometimes varied regarding whether each object was acted upon by a different person, or whether sometimes a single person interacted with multiple objects (see Figure 1b for an example). Although the images still represented the intended inverse interpretation, scenario details could have had an effect in how participants interpreted the stimuli. In addition, no piloting was performed with the images therefore we have no data on the impact of this oversight.

In our current study, participants were randomly assigned to complete the experimental task in either Spanish or in English, but not both. In future versions of the task, we suggest that each participant complete the experimental task in both languages, with a pause between the two language sessions. This approach would help ensure that the results are not influenced by the random sampling of two different groups, even if they share similar characteristics. Conducting two separate monolingual sessions could also better control for language mode (Grosjean, 2008). Additionally, having two sessions would allow for more time to administer the experiment and incorporate more comprehensive or even multiple language proficiency assessments to better capture participants’ language experiences.

Variation in how scope ambiguities are interpreted has been observed across different speaker groups, including children (e.g., Musolino et al., 2000; Musolino & Lidz, 2003, 2006), childhood bilinguals (see Section 3), and second language learners (e.g., Chung & Shin, 2023; Syrett et al., 2017). However, additional complexity may also arise from dialectal variation. Most studies have focused on standard varieties of English and Spanish, yet speakers of different regional varieties may interpret scope differently. This is particularly relevant given evidence that prosody and context shape quantifier scope interpretations (see Section 2). Investigating more language pairs and regional varieties will therefore be essential for better understanding this phenomenon in diverse linguistic contexts.

In addition, given that speakers often find alternative ways to make themselves understood and to avoid ambiguities, it may be the case that quantifier scope structures are relatively uncommon. Therefore, examining corpus data could provide insight into how often and how these constructions arise and function in naturalistic contexts. Finally, future research should explore other types of ambiguous constructions to determine whether the observed patterns generalize across a broader range of syntactic contexts, particularly those with stronger interpretive biases.

8. Conclusions

This study offers valuable insights into how Mexican Spanish/U.S. English heritage bilinguals interpret scope-ambiguous sentences. Results demonstrated a consistent preference for surface over inverse interpretations in both languages, with no significant difference in ratings between them, supporting the idea of shared strategies for processing scope ambiguities across both languages (Ronai, 2018; Scontras et al., 2017). However, further research is needed to determine how much of this pattern depends on a shared conceptual/linguistic system and how much depends on simplification strategies, which may or may not be a result of bilingualism. Language proficiency did not significantly influence interpretation preferences, aligning with previous findings that proficiency plays a limited role in quantifier scope processing (Chen & Hester, 2024; Ronai, 2018; Shen & Chen, 2022). Response time results partially back up our predictions: surface interpretations were processed faster overall, but the correlation between response time and ratings was only significant in English. This asymmetry may reflect differing cognitive demands between the two languages. Together, these findings highlight the complexity of bilingual language processing, particularly with regard to how bilinguals interpret scope ambiguities. While the study provides a foundation for understanding bilingual scope interpretation preferences, further exploration is needed to fully understand the factors influencing these processing patterns, leaving a clear path for future work.