Doing Romance Linguistics: A Multilingual Acquisition Perspective

. Doing Romance Linguistics: A Multilingual Acquisition Perspective.


Introduction
As a scholar of language acquisition in multilingual contexts, I have been involved with Romance linguistics since the inception of my career.Grounded in linguistic theory and analyses, my research investigates the linguistic knowledge, behavior and processing of groups of speakers and language learners using experimental psycholinguistic methods.Although I am mainly known for my contributions to the acquisition of Spanish (Montrul 2004a), my scholarship has included other languages (Turkish, Hindi, Arabic, Korean) and the Romance languages (French, Brazilian Portuguese, Italian and Romanian).The focus of my training as a PhD student at McGill University in Montreal was generative linguistics applied to second language acquisition.I was studying bilingualism and living in a bilingual and multilingual city.Being a native speaker of Spanish with some knowledge of French made my experience in Montreal very enjoyable and academically relevant: I was able to include French in my early scholarship as well.Over the years, I have made contributions to Romance linguistics in three broad areas: second language acquisition, bilingualism and heritage languages, and L3 acquisition.

Second Language Acquisition
My early work in second language acquisition was part of the vibrant research program on whether and how Universal Grammar constrains second language acquisition past puberty, and by focusing on properties of Spanish and French not studied before, my work expanded our collective understanding of the formal nature of interlanguage grammars at different stages of development (from initial state to ultimate attainment).Not only did this research contribute new data from Romance languages to theoretical models of how speakers mentally represent knowledge of a second language, but it also provided valid linguistic data from real speakers to understand the human language faculty in general, at a time when a researcher's own intuitions and those of a few neighboring native speakers was the leading empirical paradigm.I particularly enjoyed working on two exciting projects: my dissertation on the L2 acquisition of argument structure in Spanish, Turkish and English and my post-dissertation work on the L2 acquisition of tense and aspect in Spanish.My dissertation and related publications (Montrul 2000;2001a,b,c,d) focused on the L2 acquisition of transitive and intransitive verbs, and on the use of the clitic se as a marker of telicity and intransitivity (El viento quebró la rama 'the wind broke the branch', La rama se quebró 'the branch broke').Several competing theories of L1 transfer at initial stage of L2 acquisition were being debated at the time, none of which were fully upheld by my empirical findings.The results of my crosslinguistic studies showed that full L1 transfer was operative at the level of morphology (omission or oversuppliance of causative or inchoative morphology) but not at the level of argument structure in the same individuals, supporting the modularity of linguistic knowledge and the localized operation of L1 transfer.
The acquisition of tense and aspect was a topic of heightened interest and intense investigation in the late 90s as well.In a very productive collaboration with Roumyana Slabakova, we studied the morphological and interpretive properties of the preterite and imperfect past tenses in L2 Spanish by L1 English speakers from intermediate to near-native levels of proficiency in Spanish (Montrul & Slabakova 2003).Because we approached this topic from new syntactic treatments of tense and aspect, and this theoretical perspective differed radically from the functionalist Aspect Hypothesis assumed in most acquisition studies of aspect at that time, we developed a novel research methodology that elicited semantic judgments and aspectual entailments to make claims about linguistic competence.Our findings engaged with different interpretations of the Aspect Hypothesis on the one hand, and contributed to debates on critical periods regarding the L2 acquisition of morphology and semantics, on the other.But before this work on tense and aspect in Spanish, I also conducted research on French speakers learning Spanish and on the second language acquisition of French.
One of the main questions in second language acquisition research is whether native language transfer (L1 transfer) plays a deterministic role in L2 learning.In the 1990s, several theoretical models of the initial state in second language acquisition proliferated within the generative framework, models that sought to explain the extent to which the linguistic representation of the native language, in its entirety or partially, is involved in the L2 acquisition process.One way to approach this question is to test L2 learners of different L1 backgrounds acquiring the same target language, especially when the L1s and L2 involved differ in important linguistic respects.Montrul (1999) investigated the acquisition of the functional category AgrIOP (indirect object agreement phrase), following Sportiche's (1996) analysis of Romance clitics, by English-L1 and French-L1 learners of L2 Spanish.Because Spanish and French have dative clitics, the assumption was that Spanish and French have AgrIOP and English does not, I asked whether dative clitics were the specific triggers for the projection of the AgrIOP functional category, and whether L2 learners were aware of the syntactic consequences of the availability of this projection in Spanish.This study was inspired by Lightfoot's (1991) account of the demise of the dative case system in the history of English (see also Montrul 1997), by which a change in morphology triggered the emergence and loss of syntactic structures.When structural dative case (evident in morphological endings of nouns) was lost in Middle English, structures such as Exceptional Case Marking (ECM) (Mary believes John to be a good friend.),preposition stranding (What is this book about?), prepositional passives (This bed was slept in.), double object constructions (John gave Mary a present) and indirect passives (Mary was given a present.)emerged.In Spanish, where there is morphological dative case as instantiated in clitics, all the translations of the English sentences are ungrammatical (*María cree Juan ser un buen amigo.*¿Qué es el libro sobre?*Esta cama fue dormida en.*Juan dio María un regalo.*María fue dada un regalo.).
Assuming the Full Transfer/Full Access Hypothesis (Schwartz & Sprouse 1996), I hypothesized that L1 French learners should have no problem acquiring all the properties related to AgrIOP in Spanish, since they presumably have this functional category in their French grammars.Consequently, they should reject all the equivalents of the English sentences in Spanish.They might have trouble accepting clitic-doubled indirect objects though, since French has dative clitics but does not allow clitic doubling.The L1 English learners need to learn that the dative clitics are the overt manifestation of AgrIOP in Spanish.If dative clitics are the triggering evidence, then they would realize that the counterparts of the English sentences are ungrammatical in Spanish.Overall results of a written grammaticality judgment task revealed no significant differences between the native Spanish control group and the French L1 speakers, while the English L1 group differed from both the L1 French learners and the Spanish controls.Although the English L1 learners produced dative clitics in an elicited production task, they were confused with respect to clitic doubling with indirect objects and dative experiencers.Overall, the findings supported the Full Transfer/Full Access Hypothesis.
Another early study in collaboration Joyce Bruhn de Garavito (Bruhn-Garavito & Montrul 1996) was inspired by Kayne's (1991) microparametric account of clitic placement in Romance languages.This one was a bidirectional study on the L2 acquisition of object clitic positions by L1 French L2 Spanish learners and by L1 Spanish L2 French learners at intermediate levels of development.In Kayne's (1991) analysis, the position of the verb has interesting consequences not only for the position of clitics and clitic-climbing, but also for that of adverbs and quantifiers, as well as infinitives in si clauses.Some examples contrasting Spanish and French are given in (1-4): (1) a. Lui parler serait une erreur.him.DAT to.speak would.bean error 'To speak to him would be an error.b.María no sabe si ir al cine.'Maria doesn't know whether to go to the movies.' We reasoned that learning the position of clitic pronouns with tensed verbs and infinitives should not be a difficult task for the L2 learner, given that clitic pronouns are very common in the input and they are a key topic in instructed Spanish and French.The key question was whether the other less obvious grammatical properties related to the parameter, such as the position of adverbs and quantifiers with respect to infinitives, and the fact that infinitives can occur in si clauses in Spanish, would also be acquired once clitic placement was established.
For both the Spanish and the French experiments there were two participant groups, intermediate L2 learners and a native speaker control group.The participants completed a proficiency test, a written elicited production on pronouns, and a grammaticality judgment task with sentences manipulating clitics with finite verbs, with infinitives, in clitic-climbing constructions, infinitives with adverbs and quantifiers, and in si clauses.
The two studies found that the learners had acquired where to place accusative clitics with finite verbs in the two languages, yet more difficulty arose in restructuring contexts (examples (2)): in both studies, the learners were aware that French and Spanish place clitics in different positions, the French learners of Spanish did not accept clitic climbing (disallowed in French), and the Spanishspeaking learners of L2 French incorrectly accepted clitic climbing in French 50% of the time.In the two studies, a few learners behaved in a manner consistent with having reset the clitic placement parameter; while other learners only controlled the position of clitics, but not other associated properties.Since these were intermediate-level learners, one can assume that it takes some time before all the characteristics associated with the parameter are in place.Alternatively, as Rizzi's (2004) work on the structure of the left periphery later uncovered, infinitival sentences with si may not be related to movement of the non-finite verb, as originally proposed by Kayne (1991), but to the structure of the left-periphery.As shy graduate students, Joyce and I did not have the audacity to make bold claims, but the results of this humble bidirectional study exemplify how experimental data from second language acquisition with several speakers could be valid to confirm or falsify specific syntactic analyses.

Bilingualism and Language Change
Transitioning from a bilingual city in Canada, where French and Spanish enjoyed similar status and vitality to the United States, where I had the opportunity to know and teach Spanish heritage speakers in my first faculty position in SUNY, Albany motivated me to understand why the outcomes of bilingualism for early bilinguals are so different in the two countries.Since the early 2000s I have been studying the linguistic development of minority languages in heritage speakers (2 nd generation immigrants) to understand language change and language attrition across generations on the one hand, and differences and similarities between heritage speakers and second language learners, on the other.My work uncovered the linguistic development of several aspects of Spanish heritage speakers' morphosyntax and semantics (tense and aspect, subjunctive, gender agreement, articles and genericity, clitics and topicalizations, null and over subjects, whquestions), elucidated revealing differences and similarities between heritage speakers and second language learners in their linguistic knowledge as a function of experience with spoken and written language, and highlighted the impact of dominant language transfer and age effects in the acquisition, maintenance and loss of the heritage language at the individual level (Montrul 2002, Montrul 2004b, Montrul 2010, Montrul;Foote & Perpiñán 2008;Montrul & Ionin 2010, 2012;Montrul & Perpiñán 2011).My studies of Korean, Arabic, Hindi and Turkish as heritage languages (Coşkun-Kunduz & Montrul 2022; Kim et al. 2009;Benmamoun et al. 2014;Montrul et al. 2012) confirmed many of the patterns and conclusions I have reached on the basis of Spanish as a heritage language.
Intrigued by the findings of Montrul (2004b) and Montrul & Bowles (2009) on the potential erosion of Differential Object Marking (DOM) in Spanish heritage speakers, the search for the root of the apparent convergence with English, a language that does not mark DOM on animate, specific objects like Spanish, led me to question whether the acoustic salience of the marker (the preposition a) as a possible leading factor in the language change observed.Another possibility is that DOM is a vulnerable interface property, since it is triggered and influenced by semantic, syntactic and discourse factors.To test the hypotheses that the erosion of DOM in Spanish is related to the acoustic salience of the marker or to the fact that DOM is an interface phenomenon, finding languages with a more acoustically salient DOM marker was critical.In 2009 I started a project comparing Spanish, Hindi and Romanian as heritage languages, reported in Montrul (2022).The DOM marker in Romanian is pe, and in Hindi it is -ko.
The studies examined knowledge of DOM in oral production, written production, aural and written comprehension and bimodal grammaticality judgements in five groups of speakers in each language (Spanish, Hindi and Romanian): young adult simultaneous bilingual heritage speakers, sequential bilingual heritage speakers, age and SES matched young native speakers in the homeland, first generation immigrants in the United States (the parental generation) and age and SES-matched older native speakers in the homeland.Older and younger native speakers in the homeland (Mexico, India, Romania) were included to ascertain in apparent time potential language change with respect to DOM in the homeland varieties.Comparing these two groups with the adult immigrants allowed to establish whether the immigrant group has undergone L1 attrition.Finally, comparison of the heritage speakers and the adult immigrants informs continuation or disruption in language transmission, especially if the heritage speakers show innovation that are not attested in the immigrant group and that cannot be attributed to direct parental transmission.
The collective results the three heritage languages presented in Montrul (2022) showed that there is high variability in the use of DOM in obligatory contexts, in all the second-generation immigrant groups, not just in Spanish.Therefore, this is evidence that DOM is vulnerable to change in heritage language grammars when the contact majority language, in this case English, does not have DOM.However, revealing differences in the extent of variability in the use of obligatory DOM with animate, specific direct objects were uncovered in the three languages: the variability in DOM expression turned out to be more widespread in Spanish than in Hindi and Romanian because in Spanish it also affected the firstgeneration immigrant population, not just the second-generation heritage speakers.Perhaps counterintuitively, the heritage language with the highest number of speakers in the United States and the most vitality of the three studied-Spanishexhibits the more extensive variability and even change.Several linguistic factors may contribute to the more extensive erosion of DOM in Spanish than in Hindi and Romanian, such as structural differences between the three languages with respect to acoustic salience of the markers, the availability of determiners, and the availability of clitic doubling, as well as situational factors related to the vitality of the languages and patterns of language use in the three immigrant communities.
At a very general level, the results can be captured the Linguistic Niche Hypothesis (Dale & Lupyan 2012;Lupyan & Dale 2010), which relates structural change and situational factors by taking into account the number and variety of speakers who use the language and contribute to its morphological simplification.Thus, Spanish shows more erosion of DOM than Hindi and Romanian, because it has more speakers from a variety of linguistic backgrounds and profiles, including more L2 learners of Spanish, than Romanian and Hindi.Comparatively, Romanian heritage speakers, who showed relatively more language maintenance than the other groups, belong to a speech community in the United States that is much smaller and more homogeneous.Instead of adopting an overly general hypothesis of linguistic and social structure, I opted for a more fine-grained analysis that links language acquisition, language attrition, and diachronic language change, an analysis that elucidates the potential relationship between the language of the heritage speakers and the language of the first-generation immigrants, who are often the heritage speakers' main source of input.Based on analyses of the participant's linguistic background questionnaires and follow-up studies of mothers and children reported in Montrul (2022), I argued that the structural differences between heritage language grammars and baseline grammars (the parental generation) point to interruption in the transmission process.This is not a case where the children go beyond the parental input, just the opposite: they do not fully replicate the parental input and show changes not seen in the parental input.There is a break between the language of the parental generation and the language of the heritage speakers in DOM use, and this transmission "failure" is more evident in Hindi than in Romanian.Of all the groups, the Romanians were the ones who seemed to preserve the language best.The Spanish heritage speakers and adult immigrants omit DOM to a certain extent, but these are due to individual developments in their grammars resulting from restricted input and language contact, rather than from the adult immigrants directly transmitting innovative patterns to the heritage speakers.As recent studies of child heritage speakers and their parents show (Cuza et al. 2019;Daskalaki et al. 2021;Coşkun Kunduz & Montrul 2022), the developmental schedule of heritage language acquisition and adult L1 attrition do not match to support such relationship.The results of this crosslinguistic transgenerational study have implications for our theoretical linguistic models of native speaker knowledge, and for understanding the mechanisms of language acquisition, transmission, and diachronic language change.More broadly, they have implications for language policies and the education of minority language speakers in the United States.

Third Language Acquisition
An important development in the field of second language acquisition at the turn of the century was the need to distinguish between the acquisition of a true second language and the acquisition of languages learned by bilinguals and multilinguals; that is, learners who already know more than one language.All instances of nonnative acquisition beyond the L1 have been subsumed under L2 acquisition, until research by Cenoz (2001) and Leung (2006Leung ( , 2007)), among others, made clear that interlanguage development of an L2 is not necessarily identical to that of an L3, especially regarding potential sources of transfer.With the emergence of the field of L3 acquisition, defined as the sequential acquisition of another language beyond a second language, the role of prior language knowledge has become even more nuanced.Not only do researchers continue to ask whether and when transfer will take place in interlanguage development, but they additionally ponder on the specific interplay between the L1, the L2, and the L3, and the potential sources of transfer from any of the previously known languages in the initial state and subsequent development.Does transfer in L3 acquisition come exclusively from the L1 as in L2 acquisition?Does it come exclusively from the L2?Does it come from both languages, or from "none"?A highly influential model has been the Typological Primacy Model (TPM) (Rothman 2011), according to which the language ultimately selected for transfer at the initial stage is the one identified by the learner as being more typologically similar to the L3.
Keeping these conceptual and methodological issues in mind, Montrul, Dias & Santos (2011) contributed both to the field of Portuguese linguistics and to the emerging field of L3 acquisition, by investigating the potential source of transfer and the structural relationship between the languages involved in the L3 acquisition of Brazilian Portuguese (BP) in two independent studies.The two studies involved different participants who belong to the following language groups: Group 1: Native speakers of Spanish who speak English as a second language and are learning Brazilian Portuguese (L1 Spanish-L2 English-L3 Brazilian Portuguese) Group 2: Native speakers of English who speak Spanish as a second language and are learning Brazilian Portuguese (L1 English-L2 Spanish-L3 Brazilian Portuguese) Group 3: Brazilian Portuguese native speakers The syntactic focus was object clitic pronouns and related properties of clitic placement and object expression.English does not have object clitics, while both Spanish and BP are Romance languages with clitic pronouns.However, the actual use of clitics in contemporary BP, and particularly 3 rd person clitics, is lower than in Spanish.Furthermore, clitic rates and clitic positions in contemporary BP differ as well in spoken and written varieties of the language due to ongoing diachronic change.Portuguese allows null objects in definite and indefinite contexts, whereas only indefinite null objects are felicitous in Spanish.Finally, Spanish has DOM (animate, specific direct objects are preceded by the preposition a), whereas Brazilian Portuguese and English do not.Examples (5) to (10) illustrate critical differences between the languages and Table 1  The combined results of the two studies showed that the two groups of L3 learners in each study converged on the spoken (study 1) and written (study 2) grammars of the BP native speakers in several areas, and that the acquisition of these aspects of BP is not apparently very problematic.The L3 learners seemed to have acquired that BP has clitic pronouns and null objects, and that preverbal and postverbal clitics are acceptable with finite verbs.At the same time, the L3 learners in the two studies were not native-like in all respects, and there were also many areas where they did not converge on the grammar of BP.
In study 1, some individuals in the two groups of L3 learners produced a measurable rate of errors with clitic doubling, clitic climbing and DOM, all properties of Spanish that are not possible in contemporary BP.Furthermore, the two groups almost exclusively produced postverbal clitics with infinitives, the Spanish option.Study 2, which focused on clitics in written language, confirmed some of the findings in Study 1.The two L3 groups were more accepting of postverbal than of preverbal clitics with non-finite verbs, and quite accepting of clitic climbing-the Spanish option-which is not quite acceptable in BP.These patterns of results also support the structural similarity hypothesis, since both the Spanish-L1 and the English-L1 learners made errors that can be attributed to transfer from Spanish.Montrul, Dias & Santos (2009) reported similar results with the that-t effect (a pro-drop-related property) in BP.
A more recent debate in L3 acquisition is whether transfer operates as a whole (wholesale transfer) or only specific properties of the language transfer at different stages of interlanguage development (piecemeal transfer).Wholesale transfer implies that the entirety of a grammar, except for the phonology, is transferred all at once at the initial state, as occurs in the Full Transfer/Full Access Model (Schwartz & Sprouse 1996) for L2 acquisition and its extension to L3 acquisition (Schwartz & Sprouse 2021).In the Full/Transfer Full Access Model, transfer is most prominent at initial stages of development, and this initial transfer may or may not be overcome throughout development in terms of convergence towards a target-like grammar.
In property-by-property transfer, in contrast, specific linguistic properties are transferred at different times throughout the course of acquisition.For example, if an L1 Spanish learner of L2 English exhibited wholesale transfer at the initial state, their Spanish grammar in its entirety transfers all at once and becomes the structural representation for interlanguage English (the L2); this transfer would include all syntactic and morphological properties of their Spanish grammar.L1 influence may or may not persist beyond the initial state, depending on the nature of the input and the nature of a given linguistic property.In the case of propertyby-property transfer, however, the same learner may transfer, for example, the prodrop property of Spanish at one point, while transferring adjective placement at a different point in development.Thus, different linguistic properties may transfer at different developmental stages (initial, intermediate, advanced).Gallardo & Montrul (in press) examined the predictions of the Linguistic Proximity model (Westergaard 2021) in the L2 and L3 acquisition of Italian by Spanish-and English-speaking L2 and L3 learners.The structures of interest were dative experiencer psych verbs and animate specific direct objects.Crucially, dative experiencer psych verbs from the piacere/gustar 'like' class are similar in Italian and Spanish in that they cannot be passivized (*Maria è piaciuta da Francesco/*María es gustada por Francisco.);these same verbs have nominative experiencers and can be passivized in English (Mary is liked by Francisco).Italian and English both lack DOM (*Francesco vede a Maria/*Francisco sees DOM Maria"), which is instantiated in Spanish (Francisco ve a Maria), such that objects which are both animate and specific are morphologically marked.Results from a written acceptability judgment task found that the L3 groups transferred from distinct previously acquired languages, either their L1 or L2 (English and Spanish), to different extents for the two main properties tested, supporting the validity of the Linguistic Proximity Model and property-by-property transfer as a post initial state theoretical model of L3 interlanguage development.While the English-speaking groups did not transfer passivized psych verbs, they did show higher acceptability of DOM, as shown in Figure 1.
The L2 group rated grammatical sentences higher than the two L3 groups, which is to be expected as there is no structural analog to DOM in English.In the statistical analyses, the L2 group was more accepting of the grammatical No DOM sentences (Francesco vede Maria) than the SPAN-L3 group and the ENG-L3.With the ungrammatical *DOM sentences, the L2 group was less accepting of these sentences (and closer to the ratings of the native Italian Control group) than the two L3 groups, who did not differ significantly from each other in their ratings.Thus, our study finds evidence that transfer in L3 Italian, with respect to the properties tested, comes from the two previously acquired languages and takes place property-by-property.

Summary
Our minds are prepared to handle more than one linguistic system from birth.How languages interact in the minds of speakers and how they manifest themselves structurally along the lifetime, continues to fascinate me.Throughout my career, my scholarship has been driven by questions about the nature of language and how different languages are acquired by different groups of speakers of different ages in monolingual, bilingual and multilingual environments.I have engaged with Romance linguistics to address critical questions about the nature of multilingual knowledge, language transfer, language dominance, restructuring and transmission, and to bring the study of heritage languages (Romance or others) to the center of linguistic research, by investigating these varieties through the same theoretical lenses and experimental research methodologies that we use to study standard languages.I am convinced that as we uncover how languages are learned, maintained and transmitted, lost and regained, data from bilingual and multilingual speakers are valid to test and falsify linguistic hypotheses and analyses about the nature and structure of human language, even if these were dismissed when I started in the field.And this is how I do Romance linguistics.

Figure 1 :
Figure 1: Mean Acceptability Ratings on Unmarked and Marked Animate Objects (Gallardo & Montrul in press)

Table 1 .
Cross-linguistic differences in object expression between Brazilian Portuguese, Spanish and English.