The i nfluence of Guarani on g ender a greement in Paraguayan and Correntino Spanish: A c ontrastive a nalysis

. In this paper we analyze the influence of Guarani, a genderless language, on patterns of gender agreement in two distinct Spanish varieties, Paraguayan Spanish and Correntino Spanish. Using two unique conversational corpora, we show that, of the two varieties, Paraguayan Spanish more closely resembles Standard Spanish in patterns of gender agreement, but the cases of nonagreement that do occur are found in a wide variety of linguistic contexts, resembling the kinds of transfer effects common in L2 varieties of Spanish in speakers whose L1 is genderless. Correntino Spanish exhibits patterns of nonagreement that are not only more common but are limited in linguistic scope, i.e., only occurring in noun phrases involving feminine nouns with non-article modifiers. We attribute the differences between these two varieties to the differing presence of Guarani in the two regions where they are currently spoken. Paraguayan Spanish, in closer contact with Guarani, is subject to synchronic L2 transfer effects, while Correntino Spanish, a largely monolingual variety in the modern day, is characterized by diachronic L2 transfer effects that have fossilized and come to be found in the speech of monolinguals.

By "Correntino Spanish" we refer to the variety of Spanish unique to the Argentine province of Corrientes, a northeastern province that borders Paraguay, as illustrated below in Figure 1.Correntino Spanish, while exhibiting sufficient Guarani influence to be considered a contact variety, is predominantly spoken by monolinguals in the modern day (Pinta 2023).The influence of Guarani on this variety is mostly historical, being reflective of widespread bilingualism in earlier periods of the province's history.Having acknowledged that, Spanish/Guarani bilingualism is still common in the rural provincial interior, where the local variety of Guarani is still prevalent (Pinta 2022a), and synchronic Guarani influence in these communities is still notable.Guarani speakers constitute likely around 10% of the provincial population (Cerno 2013:35), extremely few of whom are found in urban centers such as the provincial capital of Corrientes City.In this paper we demonstrate that, due to different social contexts of contact, Paraguayan Spanish and Correntino Spanish exhibit different rates and distributions of non-canonical patterns of grammatical gender agreement.Paraguayan Spanish is more variable, providing fewer non-agreeing tokens but with greater breadth of modifier types and more diverse patterns generally, while nonagreement in Correntino Spanish is both more common and more linguistically constrained, occurring only with particular modifiers in noun phrases with feminine nouns.
This paper is structured as follows.Section 2 provides background information, about both the relevant grammatical aspects of Spanish and Guarani, and Section 3 provides the methodological details of this study.In Section 4 we provide the results, with accompanying discussion in Section 5, and we give concluding remarks in Section 6.

Background.
In this section we provide a short summary of the relevant background to our study, including a summary of the grammatical gender systems of Spanish and Guarani in Section 2.1 and an overview of relevant previous studies on contact-induced effects on grammatical gender in other varieties of Spanish in Section 2.2.
2.1.GRAMMATICAL GENDER IN SPANISH AND GUARANI.Spanish has a binary system of grammatical gender in which nouns can be either masculine (commonly morphologically marked with the suffix -o) or feminine (commonly morphologically marked with the suffix -a).As demonstrated in (1), below, modifiers such as determiners or adjectives must agree in gender with the noun they modify.
(1) Spanish a. una casa roja a.F house.F red.F 'a red house' b. un libro rojo a.M book.M red.M 'a red book' Guarani, however, does not express grammatical gender.In (2), below, the forms peteĩ 'a' and pytã 'red' are invariable, and róga 'house' is a genderless noun.
(2) Guarani peteĩ róga pytã a house red 'a red house' Guarani is often described as a genderless language, and this is the case for all pre-contact varieties of Guarani and the overwhelming majority of modern varieties.However, Argentine Guarani, the variety unique to Corrientes, has developed a system of nominal gender as a result of Spanish influence; in this system both the Spanish definite articles el and la and the indefinites un and una have been borrowed and are used in ways that evidence the emergence of underlying nominal gender (Cerno 2010).Even in this variety, however, nouns and adjectives are not morphologically marked for gender, and nominal gender is only recoverable in noun phrases containing articles.
It is worth mentioning that the use of the Spanish feminine definite article la is common in Paraguayan Guarani.However, Thun (2006) and Herring (2012) found that la, despite its use in Spanish as a gendered definite article, does not express grammatical gender, nor does it fulfill a definite article function.Rather, it is used to express pragmatic functions, such as emphasis.
2.2 PREVIOUS STUDIES OF SPANISH GENDER AGREEMENT.Adult Spanish speakers are commonly described as "highly consistent" in their production of grammatical gender agreement (Hawkins and Franceschina 2004: 175), especially in monolingual varieties.Agreement rates "at ceiling" have been demonstrated in experimental tasks involving naturalistic speech (Sagarra and Herschensohn 2010), including in studies surveying speakers from a wide variety of countries (Alarcón 2011), and speech errors involving gender agreement are significantly less common than speech errors involving number agreement (Antón-Méndez et al. 2002;Barber and Carreiras 2005).
However, L2 Spanish varieties spoken by groups whose L1 is an indigenous language of the Americas often exhibit patterns that differ from the system of canonical Spanish gender agreement exemplified above in (1).Contact varieties of Andean Spanish (Sánchez 1996), Yucateco Spanish (Michnowicz 2012), Guatemalan Spanish (Tesoro 2002), and Northern Belizean Spanish (Balam 2014) exhibit patterns of nonagreement in what can be described as nonsystematic L2 transfer effects (see the discussion in Pinta 2022b: 613-617 for further details).The common theme of this body of literature is that nonagreement is most often attested in L2 speakers of Spanish who speak a first language that does not have grammatical gender, and nonagreement -if the patterns are given -is either seemingly random or favors feminine nouns, which are modified via the non-canonical use of masculine modifiers.

Methodology.
The data informing this study come from two corpora.The first corpus, compiled by the first author, consists of 8 hours of recordings from 18 speakers, recorded in Asunción, Paraguay over the course of six months between 2011 and 2012.Recordings were made of sociolinguistic interviews as well as elicitation sessions involving the "frog story" (Berman and Slobin 1994).All Spanish-speaking participants recorded were bilingual, also speaking Guarani.
The second corpus, compiled by the second author, consists of data collected in rural northcentral Corrientes, specifically the departments of San Miguel and General Paz, from 2017-2018.Sociolinguistic interviews were carried out with 24 participants, including both bilingual and monolingual Spanish speakers, producing recordings that total approximately 14 hours (see Pinta 2022b:618-619 for more detailed information about the corpus).
Regarding the data extracted from the interviews in both corpora, an important note about the envelope of variation, i.e., the context in which variation occurs (Cameron and Schwenter 2013: 466), is necessary.In our data, Correntino Spanish variation between canonical agreement and nonagreement only occurs with feminine nouns and non-article modifiers.That is, casa roja 'red house' and casa rojo are both possible, given that casa is a feminine noun, but *libro roja 'red book,' in which a masculine noun occurs with a feminine modifier, is not possible.Additionally, there is no variation in agreement between nouns and articles, but variation does occur between nouns and all non-article modifiers (i.e., adjectives, demonstratives, quantifiers, and possessives).Thus, while casa rojo 'red house' (canonically casa roja), este casa 'this house' (canonically esta casa), muchos casas 'many houses' (canonically muchas casas), and nuestro casa 'our house' (canonically nuestra casa) are all possible, *el casa 'the house' and *un casa 'a house' are not (Pinta 2022b).
In our Paraguayan Spanish data, the envelope of variation is much less restricted.In fact, variation occurs in all possible combinations of noun+modifier, including article+noun combinations, in addition to being unrestricted by the gender of the noun, i.e., masculine nouns can take feminine modifiers and feminine nouns can take masculine modifiers.
Accordingly, the tokens extracted from the Correntino Spanish corpus only consist of grammatically feminine nouns with a non-article modifier marked for morphological gender.An example of a feminine noun with a non-article modifier that is not marked for morphological gender would be casa verde 'green house,' where the adjective verde is unmarked for gender and therefore the noun phrase offers no information about gender agreement.Such cases are excluded from the token count provided here.The tokens extracted from the Paraguayan Spanish corpus, however, include both masculine and feminine nouns and all modifier classes.The only cases excluded would be cases such as casa verde, in which information about agreement is not retrievable.
The data were extracted and coded for both linguistic and social variables.The data from both varieties were coded for animacy, priming, definiteness, adjacency, modifier class, nounmodifier order, and noun morphology, as well as age and sex of the speaker.The Paraguayan Spanish data were also coded for some additional social variables, as discussed in the following section.

Results.
After extracting tokens using the criteria discussed in the previous section, 736 tokens of Correntino Spanish and 1244 tokens of Paraguayan Spanish were produced.We will first describe the patterns in Correntino Spanish, then Paraguayan Spanish, before comparing the two.
4.1 CORRENTINO SPANISH.2 17% (n = 124/736) of the Correntino data demonstrated nonagreement, with token count by speaker varying given that these data were extracted from conversational contexts, where some conversations lasted longer than others and where the occurrence of feminine nouns with relevant modifiers, as discussed in the previous section, were expectedly variable in the frequency of their occurrence.Controlling for token count by speaker produced a slightly higher frequency of nonagreement, at 22%, leaving an agreement rate of 78%.Of the speakers with at least 20 tokens, the lowest rate of agreement was 56% and the highest was 95%.
Notably, all speakers, including self-reported Guarani-dominant bilinguals, Spanishdominant bilinguals, and Spanish monolinguals, exhibited patterns of nonagreement.Pooling all tokens from Guarani-dominant bilinguals (n=228) yields an agreement rate of 76%.Doing the same for Spanish-dominant bilinguals (n=383) yields 78% and for Spanish monolinguals (n=125) 83%.Across the board, despite small differences between the three groups, canonical agreement patterns are more common than nonagreement patterns, and nonagreement occurs at rates -even among Spanish monolinguals -far beyond those of Standard Spanish.
Regarding the linguistic variables, a logistic mixed-effects regression model (see Pinta 2022b:624-626) shows that both modifier class and the distance between the noun and modifier impact the likelihood of agreement.More specifically, adjacent modifiers are more likely to agree with their noun than modifiers with intervening lexical material, i.e., casa roja 'red house' vs. La casa de mi abuelo es roja 'My grandfather's house is red.' Within adjacent modifiers, determiners (again, excluding articles) are more likely to agree with their noun than adjectives, i.e., esta casa 'this house' vs. casa roja 'red house.' Representative examples of nonagreement patterns from the Correntino Spanish corpus are given below in (3).In (3a) we see intervening material between la tonada 'the tune' and distinto 'diferent', with the expected canonical form of the latter being distinta.In (3b) la conversación 'the conversation' is modified using the masculine form of the possessive nuestro instead of nuestra.(3c) illustrates that nonagreement also occurs in contexts of animacy (despite this not being a statistically significant predictor of nonagreement), with the masculine divorciado 'divorced' being used in reference to a woman.
(3) Correntino Spanish a. La tonada que tienen ellos es distinto the.F tune.F that have.3PLthey be.3SGdifferent.M 'They speak with a different intonation' b. la conversación nuestro the.F conversation.F our.M 'the conversation of ours' c.Ella está divorciado she be.3SGdivorced.M 'She is divorced' 4.2 PARAGUAYAN SPANISH.The nonagreement rate in the Paraguayan Spanish data is 2.4% (n=30/1244).Given that the data also was conversational, rates per speaker were variable, as in the Correntino data.Tokens of nonagreement were produced by seven different speakers, accounting for 39% of the 18 total speakers.Two of the seven were men and five were women.The two men were brothers and two of the women were a mother-daughter pair.While five out of the seven speakers identified their native language as Spanish, none of them had natively monolingual Spanish speaking parents.Further information about these seven speakers is given below in Table 1 736), the relatively small number of tokens of nonagreement in the former variety did not allow for the same kinds of statistical inferences to be made with regard to the linguistic factors influencing nonagreement.While recognizing that the nonagreement rate of 2.4% in the Para-guayan Spanish data is a small deviation from the expected "at ceiling" conversational agreement rates found in the previous work mentioned in section 2.2, the patterning of the nonagreement is an equally important consideration in interpreting these data.The Paraguayan Spanish data includes non-agreeing tokens of feminine nouns with masculine modifiers, as exemplified below in (4a), in addition to masculine nouns with feminine modifiers, as exemplified in (4b).Example (4a) has a feminine noun pera 'pear' with a masculine modifier lindo 'pretty, nice'.(4b) demonstrates the use of the feminine definite article la with the masculine noun timbre 'doorbell.'3Notably, of the seven speakers who produced non-agreeing tokens, three produced both possible types: feminine nouns with masculine modifiers and masculine nouns with feminine modifiers.
( Additionally, our Paraguayan Spanish data contains instances such as that in ( 5), in which a feminine plural object pronoun is used to refer to a plural antecedent of mixed genders.The example in ( 5) is ungrammatical in Standard Spanish, as the masculine plural object pronoun los is used in such cases.
( As a reflection of the different social contexts of contact, the Paraguayan data is less easily broken down into Guarani-dominant and Spanish-dominant than the Correntino data.Instead, sociolinguistic data has been determined by a series of questions asked of participants regarding languages spoken at home by the speaker and those they have and currently live with.The majority (n = 980, 77%) of tokens were produced by speakers who identify their native language as Spanish.Meanwhile, only 25.3% (n = 320) of the tokens were produced by speakers who identified their parents' native language as Spanish.That is, 74.2% (n = 924) of the tokens were produced by speakers with parents whose native languages were identified as either Guarani or mixed.This corresponds to 75.6% (n = 940) of the tokens being produced by speakers who indicated that the language spoken at home as a child was either Guarani or mixed.Home languages, however, appear to have shifted, as 63.6% of tokens (n = 804) were produced by speakers who say that the language spoken at home now is Spanish.Spanish is the language spoken by housemates in homes of those who produced 51.6% (n = 653) tokens, with "mixed" being the best option describing the language spoken at home for those producing 45.9% (n = 581).That leaves only 10 tokens (.8%) being produced by speakers who identify Guarani as the language spoken by current housemates.When asked if the participants speak one language better than the other, 62.1% of the tokens were produced by those who answered Spanish.Despite the wording of the question, 30.8% (n = 389) of tokens came from those who said they speak both equally ("mixed"), while only 70 tokens (5.5%) came from those who speak Guarani best.Lastly in response to the question "What language do you use most on a daily basis?"55.3% (n = 700) of the data were produced by participants who answered Spanish, 36.4% (n = 461) answered "mixed", and only 6.6% (n = 83) said Guarani.
4.3 COMPARISON.Comparing the patterns in our Correntino and Paraguayan Spanish data, several notable differences emerge.First, the overall rate of nonagreement is higher in Correntino Spanish (17%) than in Paraguayan Spanish (2.4%), with both being higher than what is expected in conversational monolingual Spanish (as discussed in section 2.2).Perhaps more important than frequency, however, are the linguistic patterns.The Correntino Spanish system has comparatively more systematic, linguistically constrained patterns of nonagreement than the Paraguayan Spanish system.Paraguayan Spanish allows for nonagreement in all modifier types.In contrast, Correntino Spanish only allows nonagreement in non-article modifiers.Additionally, Correntino Spanish restricts nonagreement to cases involving feminine nouns and their modifiers, which can be optionally morphologically masculine; Paraguayan Spanish, by contrast, allows variation among both masculine and feminine nouns and their respective modifiers.Paraguayan Spanish also allows for variation in pronoun gender with respect to the gender of the antecedent.A comparison between these patterns is given in Furthermore, the Correntino Spanish data were categorical in that all speakers produced tokens of nonagreement (despite variation in rates between speakers), whereas in Paraguayan Spanish only 39% (7/18) of speakers produced nonagreement tokens.Lastly, social variables are important in understanding the Paraguayan data, while social variables in Correntino Spanish carried less weight.In the latter case, no social factors can be meaningfully understood as impacting the likelihood of nonagreement, which occurred in speakers of all ages and language backgrounds, including completely monolingual speakers.5. Discussion.The strongest explanation for this subtle morphosyntactic difference between Correntino and Paraguayan Spanish, we argue, is related to the synchronic status of Guarani in the two regions.As mentioned earlier, Guarani has a considerably more ubiquitous presence in Paraguay than in Argentina,4 with 77% or more of Paraguayans speaking Guarani and likely only 10% or so of the Correntino population being considered speakers (see section 1 above).Consequently, Paraguayan Spanish exhibits the kinds of non-systematic patterns in gender agreement variation that are commonly associated with L2 speakers of Spanish whose L1 is a language without grammatical gender, i.e., phenomena considered instances of "interference" or "transfer" from a contact linguistics perspective (Winford 2010) and "errors" from a second language acquisition perspective (Brown 2020).We are careful to note, however, that not all Paraguayan speakers exhibiting such patterns are L2 speakers of Spanish, but the patterns -across-the-board, non-linguistically constrained use of either gender modifier with either gender noun -resemble those expected from L2 speakers who have not fully acquired the gender agreement system of Standard Spanish.The commonplace status of Guarani dominance throughout the population of Paraguay, alongside Guarani monolingualism (still common in rural regions of the country), has allowed for synchronic L2 transfer effects in Paraguayan Spanish, at least among some speakers.
The Correntino Spanish situation, however, cannot be said to be the product of synchronic Guarani influence in the same way given that noun+modifier agreement is still as robust as in Standard Spanish in some linguistic contexts, i.e., in nouns modified by definite or indefinite articles.Through the consistency of noun+article agreement, it is clear that Correntino Spanish speakers are aware of the underlying gender of a noun and are as sensitive to gender agreementat least in this particular linguistic context -as speakers of Standard Spanish.The variation among non-article modifiers suggests that modifiers of this type with masculine morphology are available for use as gender-neutral modifiers.Additionally, the differing treatment of feminine nouns and their modifiers on the one hand and masculine nouns and theirs on the other serves as further evidence of the sensitivity of Correntino Spanish speakers to underlying gender and modifier agreement.This effect in modern Correntino Spanish is undoubtedly related to fossilized diachronic L2 transfer effects from previous stages of this variety's history when it was in a contact situation that more closely resembles the modern Paraguayan situation, i.e., one of pockets of Guarani monolingualism and region-wide presence of Guarani-dominant Spanish speakers.We suggest that modern Correntino Spanish has inherited a system of variable gender agreement attributable to Guarani influence but that this variety is no longer subject to extensive synchronic influence.Were that the case, we would expect the kinds of across-the-board transfer effects that are attested in our Paraguayan Spanish data.The fact that, as described in section 1, most speakers of Correntino Spanish are monolingual, and that Correntino monolinguals exhibit strikingly similar patterns of gender agreement variation as Correntino bilinguals, is further support for this position.

Conclusion.
This study contributes a novel comparison between two similar but distinct varieties of Spanish that have both been affected by long-term contact with Guarani.The patterns identified in Paraguayan Spanish, i.e., variable gender agreement in all linguistic contexts involving nouns and modifiers, evidence a situation in which Guarani still exhibits synchronic influence on Spanish in this region.Contrastingly, the patterns of Correntino Spanish, i.e., variable gender agreement in linguistically constrained contexts but consistent enforcement of gender agreement in others, evidence Guarani influence stemming from L2 transfer effects from an earlier stage of the language's history that have become fossilized in the modern grammar.Despite the many historical and sociolinguistic characteristics that Paraguay and Corrientes have in common, differences in the status of Guarani in the two regions, and therefore in the degree to which it exerts influence on Spanish, have produced subtle differences in the morphosyntax of these two varieties.The case of Paraguay and Corrientes sheds light on the value of contrastive investigations of dialectal differences in contact scenarios involving the same languages but with differing social contexts, adding to the substantial body of literature highlighting the role of social factors in constraining linguistic outcomes in situations of language contact.

Figure 1 .
Figure 1.Map highlighting Paraguay (in red) and Corrientes, Argentina (in light blue) within South America 1 .
He had two kids with her, a little boy and a girl, and he abandoned them too.'

Table 2 (
with Standard Spanish included as a point of comparison).

Table 2 .
Nonagreement in the gender systems of Paraguayan, Correntino, and Standard Spanish in our corpus analyses, with a check indicating that nonagreement occurs in the context specified.