Priming as a Diagnostic of Grammatical Constructions: Second-Person Singular in Chilean Spanish

Abstract

Structural priming has been described as a measure of association between constructions. Here, we apply priming as a diagnostic to assess the status of the Chilean second-person singular (2sg) voseo, which exists in variation with the more standard tuteo. Despite being the majority variant in informal interactions, Chileans are reported to have little metalinguistic awareness of voseo and they avoid the vos pronoun, in some cases using the tú pronoun with voseo verb forms, leading to proposals that tuteo and voseo are conflated into a single mixed form. The patterning for priming, however, indicates otherwise. Analyses of some 2000 2sg familiar tokens from a corpus of conversational Chilean Spanish reveal that a previous tuteo or voseo favors the repetition of that same form, indicating that speakers do treat these forms as distinct. We also observe that invariable forms with historically tuteo morphology are associated with neither voseo nor tuteo, while the invariable voseo discourse marker cachái ‘you know’ retains a weak association with voseo. Furthermore, while tuteo is favored with a tú subject pronoun, this effect does not override the priming effect, evidence that, even with a tú pronoun, voseo and tuteo are distinct constructions in speakers’ representations.

1. Introduction

According to a usage-based perspective, grammar is shaped by experience with language, through the repetition and conventionalization of frequently occurring patterns of language use (Bybee 2006, pp. 712–14). That is, speakers create utterances based on a probabilistic grammar that is conditioned by usage. While usage effects can be more global, impacting language change over time, here we examine the local usage effect of structural priming, seen in the tendency to repeat a structure that has been used previously in the same discourse. Specifically, we address questions of grammatical identity in the context of a change in progress in the second-person singular (2sg) in Chilean Spanish, using priming as a means of gauging speaker associations between constructions as evidenced in actual language use.

1.1. Metalinguistic Awareness of voseo in Chile

Almost all varieties of Spanish have a distinction between second-person “familiar” and “polite” forms, realized in the singular by tú and usted in Spain and across much of Latin America (the well-known T-V distinction described in the seminal work by Brown and Gilman 1960). In several Latin American varieties, including Chilean Spanish, there are two 2sg familiar forms, with tú existing alongside vos. This pronominal variation is accompanied by variation in the verbal paradigms, in what is known as tuteo and voseo. An example is given in (1), where in the one utterance and with the same verb (tener ‘have’), the speaker first uses a tuteo verb form and then a voseo verb form (captured with subscripts in the translation), both with unexpressed subjects (captured with a Ø subscript in the translation).

(1) cuando tengas la posibilidad de conocer,

  tenís que ir.

  ‘When you_-Ø have_-tuteo the chance to get to know ((the place)),

  you_-Ø have_-voseo to go.’

(Memories; 475–476; Trinidad)1

It is widely understood that in late nineteenth-century Chile, voseo was a rural phenomenon and a mark of lower socioeconomic status, while tuteo was the preferred form of the educated classes and the urban elite (Lenz [1891] 1940, p. 263). Over 100 years ago, the renowned grammarian Andrés Bello rallied strongly against voseo in Chile, famously describing it as “una vulgaridad que debe evitarse” ‘a vulgarity that must be avoided’ (Bello [1891] 1981, p. 76). By the second half of the twentieth century, a change had been observed in use of the 2sg familiar forms whereby voseo verb forms had expanded to the speech of all social classes, including in urban centers, in some instances even coming to occur with a tú pronoun (e.g., Torrejón 1986). Recent variationist work has shown the change to be considerably advanced, with rates of voseo lying between 70% and 90% today in informal spoken Chilean Spanish between people who are familiar with each other (Callaghan 2020, p. 180; Fernández-Mallat 2018, p. 73; Rivadeneira Valenzuela 2016, p. 101). Despite such high rates in informal interactions, voseo is said to retain an element of stigmatization (e.g., Fernández-Mallat 2018, p. 66). This is seen in a number of ways: rates of voseo are lower in more formal interactions or between people who do not know each other (e.g., Fernández-Mallat 2018, p. 66); voseo is not taught in Chilean schools (e.g., Rivadeneira Valenzuela 2016, p. 90); and in writing, it is largely restricted to interactional genres, such as text messaging and online chats (Cautín-Epifani 2015). This is quite distinct from the situation in neighboring Argentina, where voseo is used as a standard form, though with distinct morphology from that used in Chile (Lipski 1994, p. 141).

Chilean speakers appear to have little awareness of their use of voseo verb forms. For example, responses to direct questions about voseo indicate that many Chileans either do not know what it is, often assuming voseo to refer solely to the pronoun vos, not the verb form (Huerta Imposti 2011–2012, p. 52; Hummel 2010, p. 112; Stevenson 2007, pp. 163–67), or consider it to be a feature of Argentinian, not Chilean, Spanish (Hummel 2010, p. 110; Stevenson 2007, p. 172; Torrejón 1986, p. 682). While the voseo verb forms are in wide use in Chile, the vos pronoun is very rare (Callaghan 2020, p. 133; Fernández-Mallat 2018, p. 71; Helincks 2012, p. 193; Rivadeneira Valenzuela 2016, pp. 99–100), so rare that, in some cases, the voseo verb form occurs with a tú pronoun, as in (2). This phenomenon, known as voseo mixto verbal ‘mixed verbal voseo’ (henceforth, “mixed voseo”) or voseo culto ‘educated voseo’ (Torrejón 1986), has been said to contribute to “invisibilization” of voseo (Huerta Imposti 2011–2012, p. 54), and to the understanding by some Chileans that their use of voseo is in fact tuteo (Carricaburo 1997, p. 34). Some respondents deny using voseo even while using it extensively (and, in some cases, categorically) in the very same discussion (Hummel 2010, pp. 111–12; Stevenson 2007, p. 165).

(2) Pero tú tenís que trabajar po.

  ‘But you_-tú have_-voseo to work, of course.’

(Cousins; 1245; Carmen)

More indirect research methods provide a similar picture. For example, in one written survey in which participants were asked to select the verb form they would use from sentences illustrating different options (e.g., ¿cómo estás/estái/está? ‘how are you?’ with tuteo, voseo and the usted form, respectively), reported rates fell drastically short of rates of use in spontaneous speech of the same participants. The overall reported rate of voseo (vs. tuteo and usted forms) was 10% in the survey data, compared with an observed rate of 54% in the speech data (Bishop and Michnowicz 2010, pp. 419, 424). It appears that, when responding to direct or indirect questions around the 2sg familiar forms, Chileans lean towards what they perceive to be the “correct” form (Bishop and Michnowicz 2010, p. 425). This supports an understanding of voseo in terms of covert prestige, as a widely used non-standard form, with a degree of stigma attached to its use (e.g., Fernández-Mallat 2018, p. 76; Stevenson 2007, p. 228).

Given the context of rapid language change, lack of metalinguistic awareness about voseo, mixing of the paradigms, and stigmatization, the question arises of whether Chilean Spanish speakers separate tuteo and voseo in their grammars, or whether the two paradigms have lost their analyzability and become conflated into a single “mixed voseo”. Some forty years ago this idea was put forward, and it was proposed that mixed voseo might come to replace tuteo as the “norma universal de tratamiento de los chilenos cultos en situaciones informales y familiares” ‘universal standard form of address for educated Chileans in informal and familiar situations’ (Torrejón 1986, p. 682).

Nevertheless, there is also some evidence from patterns of spoken language use that Chileans are not oblivious to the existence of voseo. For example, one study of the spoken Spanish of Chilean migrants in Montreal found higher rates of voseo in interactions with other Chileans compared with conversations with non-Chileans (Fernández-Mallat 2011, p. 47). In this paper, we address the question of Chileans’ awareness of voseo by turning to structural priming.

1.2. Structural Priming in Variation and Change

Structural priming, also known as persistence, perseveration, or linguistic parallelism, is the favoring of a previously produced structure over an alternative that has not been recently used. For example, there is a tendency to repeat the same form for future temporal reference in English, such that speakers are more likely to choose be going to over will if there is an instance of be going to in the previous discourse (e.g., Szmrecsanyi 2005). Priming was first observed in spontaneous speech data in early variationist sociolinguistic work, for example for clitic pronouns in Montreal French (on, tu/vous, nous) (Sankoff and Laberge 1978); expression of plural -s in the Spanish noun phrase in Puerto Rican Spanish ([s] vs. [h]) (Poplack 1980); and the agentless passive vs. active in English (Weiner and Labov 1983). It has since been observed in laboratory settings, for example for the double object vs. prepositional dative construction, and agentless passive vs. active clauses (e.g., Bock and Griffin 2000; Loebell and Bock 2003), as well as in corpus linguistics, in both speech (e.g., Gries 2005; Jaeger and Snider 2007) and writing (e.g., Szmrecsanyi 2005). The robustness of structural priming effects in naturalistic discourse and laboratory contexts provides overwhelming evidence that “morphological and syntactic variation is controlled by a tendency to preserve parallel structures” (Labov 1994, p. 550).

This preservation of parallel structures has also been reported for the Spanish 2sg forms, whereby tuteo tends to follow tuteo, and voseo tends to follow voseo. This is illustrated in example (3), where the speakers produce a string of tuteo, marked in bold, before a switch to voseo, underlined. Priming was found to be the strongest internal factor conditioning the choice between voseo and tuteo in sociolinguistic interview data from Uruguayan Spanish (Carvalho 2010, p. 17). Further, in a corpus of conversational Spanish from Santiago de Chile, it was observed that the use of tuteo was categorical closely following another tuteo, and that there was a greater chance of using voseo in the context of a preceding voseo (Fernández-Mallat 2018, pp. 78–79).

Given the robustness of priming as a factor conditioning variation, this patterning may seem entirely unsurprising. However, priming is not a foregone conclusion for the 2sg familiar forms in Chile, where Chileans claim not to use the voseo form, and mix it with a tú pronoun. If Chileans genuinely do not perceive voseo forms to be distinct from tuteo, then we would not expect a previous tuteo to favor a subsequent tuteo over voseo.

(3) Tatiana: cuando tú te dejas de acordar.

      Ahí realmente mueres.

  Viviana: … Sí=,

      pero t- --

      mient- --

      si tú quieres estar con alguien.

      Estás todo el rato.

  Tatiana: [Con el alma].

  Viviana: [Porque estái pens]ando.

  Tatiana: Sí po.

  Viviana: Cierto.

  Tatiana: ‘when you_-tú stop_-tuteo remembering.

      then you_-Ø really die_-tuteo.’

  Viviana: ‘… Yes,

      but --

      while- --

      if you_-tú want_-tuteo to be with someone.

      you_-Ø are_-tuteo ((with them)) all the time.’

  Tatiana: ‘[With your soul].’

  Viviana: ‘[Because you_-Ø are_-voseo think]ing.’

  Tatiana: ‘Yeah.’

  Viviana: ‘Exactly.’

(Barbecue; 1040–1050; Tatiana, Viviana)

Of particular relevance here are the observations that structural priming occurs between related constructions, and that the strength of priming varies according to the degree of relatedness. For example, structural priming is enhanced by repetition in the prime and target, across verbs (Pickering and Branigan 1998, p. 641; Rosemeyer and Schwenter 2017, p. 27; Szmrecsanyi 2005, p. 133), TAM (Travis 2007, p. 126) or referents (Torres Cacoullos and Travis 2018, p. 91). Similarly, though it occurs both across and within languages, it is stronger within the one language (Torres Cacoullos and Travis 2018, p. 200; Travis et al. 2017, pp. 287–88). Conversely, priming has been shown to be attenuated, or absent, between more weakly related constructions. For example, English future be going to (vs. will) is primed by previous instances of future be going to; it is also primed by instances of motion verb go, but to a lesser degree (Szmrecsanyi 2005, p. 139). Research in grammaticalization has reported a diminishing of priming as one form gains autonomy from its source construction. The Spanish progressive estar ‘be’ + verb-gerund construction was primed by estar + adjective and estar + locative forms when the construction first emerged and was still closely related to these source estar constructions. But the priming weakened over time, and was eventually lost, as the progressive construction grammaticalized and gained autonomy from other estar constructions (Torres Cacoullos 2015). Finally, variable velar vs. alveolar realizations of English -ing in both nominals (e.g., ceiling and morning) and gerunds (e.g., working and jumping) are subject to priming, but only within the one morphological category, such that, for example, working primes jumping, but ceiling does not (Tamminga 2016, pp. 348–49). Tamminga regards this lack of priming as an indication that “the prime and target are not representationally related” (2016, p. 337).

From a usage-based perspective, we interpret structural priming as a kind of local entrenchment, whereby the presence of an analogous form in the preceding discourse activates, or provides a temporary boost in the strength of, its mental representation, promoting subsequent use (cf. De Smet 2016, pp. 86–87). Priming thus only occurs across constructions which share a common mental representation, and provides evidence that speakers perceive two distinct instantiations to be related, that is, exemplars of the same construction. For voseo and tuteo, this means that if they are conflated, the occurrence of a previous voseo or tuteo form should not impact the subsequent choice a speaker makes. However, if the rate of voseo or tuteo rises in the presence of a preceding voseo or tuteo, respectively, then this would suggest that speakers associate one instance of tuteo more closely with another instance of tuteo than with an instance of voseo, and vice versa; that is, they distinguish between the two paradigms.

Our analyses reveal that priming does play a strong role in conditioning the tuteo/voseo variation, leading us to propose that [Verb-tuteo] and [Verb-voseo] exist as two independent constructions. Having established this, we can then use priming as a diagnostic to assess the status of different kinds of constructions, as more or less strongly associated with tuteo or voseo in speakers’ mental representations. We consider two types of non-variable forms, the first a schematic construction deriving from tuteo (the Imperative and the Preterit TAMs), and the second a lexically specific construction derived from voseo (the highly frequent discourse marker cachái ‘you know’). Evidence that these non-variable forms have attained some autonomy from their source constructions would be seen if tuteo is not favored in the context of a previous Imperative or Preterit, and if voseo is not favored in the context of a previous cachái. We then turn to the status of “mixed voseo” (the use of a tú pronoun with a voseo verb form) to ask whether this combination might be less sensitive to voseo > voseo priming, and thus less strongly associated with the voseo form than is voseo without a tú pronoun.

We begin in Section 2 with an overview of the corpus used for this research, and in Section 3 describe the nature of the Chilean tuteo/voseo variation. Section 4 then presents the results for priming, before summing up our findings regarding the mental representations of voseo in Chilean Spanish in Section 5.

2. Corpus of Conversational Santiago Spanish

While priming has attracted great attention in the lab, given the informal nature of voseo, its association with speech, and its stigmatization, the most reliable data in which to accurately observe its patterning is spontaneous speech, for which we use the Corpus of Conversational Santiago Spanish (CCSS) (cf. Callaghan 2020).

The CCSS comprises recordings of 36 residents of Santiago made in 2014 and 2015 by the first author. Participants were recruited from the Province of Santiago de Chile, an area enclosed by mountains to the east, west and north. They were all born and raised in Chile with Chilean parents, were residing in Santiago at the time of the recording, and had not lived outside Santiago for an extended period of time in the last five years—that is, all were santiaguinos, or capitalinos, as people from Santiago are known in Chile.

The interactional nature of the second person and the association of voseo with informal speech between people who know each other render conversation the most appropriate genre in which to study its use. Thus, all recordings are of conversations between friends, family members and colleagues, that took place in the home, car or workplace of one of the participants, that is, the kind of interaction they would be likely to have ordinarily. Participants were told that the focus of the study was the relationship between language and society in Chile, but no further information about any linguistic features of interest was given. They were not assigned topics to discuss, but rather were free to talk about anything that they deemed appropriate. A range of different topics came up, including football, movies, travel, parties, mutual friends, school, work, children, debt collectors, the police, and so on.

Out of a total of some 50 recordings comprising 36 h of speech, 17 were selected for inclusion in the corpus. These were primarily two-party conversations (n = 12), but there were also four three-party and one four-party conversation. Recordings were made by 13 research assistants, each of whom was a participant in one or two conversations. The recordings were an average of 50 min long (ranging from 21 to 82 min), and an average of 30 min were transcribed from each, generally starting from some 15 min into the conversation. This provides a total of nine hours of speech and approximately 110,000 words for analysis.

Table 1. Corpus of Conversational Santiago Spanish: participant distribution by age group and gender.

	Males	Females
>35 (36–62 years)	4	13
≤35 (20–35 years)	8	11
	12	24

presents the breakdown of the participants for age and gender. Participants range from 20 to 62 years of age, split in this table at 35 years, giving a relatively even distribution in each age group. They were of varying social backgrounds: 22 of the 36 have university education; they have a wide range of occupations, working in areas such as education, health care, management, construction, hospitality, and transport; and they come from different comunas, regions of Santiago that have been used to determine socioeconomic class (e.g., Stevenson 2007, p. 233).2

The data were transcribed by five trained Chilean Research Assistants, who produced time-aligned orthographic transcriptions in F4 (Dresing and Pehl 2015). The transcription method followed the principles outlined in Du Bois et al. (1993), breaking the speech into Intonation Units (IUs), prosodic units generally of two to five words, and rarely more than one clause (though often less, e.g., for single-word IUs consisting of backchannels, response tokens, or discourse markers, such as sí ‘yes’, cierto ‘exactly’, or cachái ‘you know’). A set of orthographic conventions was established, particularly important for the feature under study which occurs predominantly in informal speech and is consequently highly variable in orthographic representation. This is particularly so for syllable-final /s/, which is variably aspirated or elided in Chilean Spanish (cf. Rivadeneira Valenzuela 2009, pp. 119–20). Thus, pronoun vos is spelt as vo, voh, bo, or boh; this was standardized to vos for the transcriptions here. In addition, we follow standard spelling conventions in writing voseo forms ending in /ay/ without final /s/ (-ái or -ai) and those ending in /is/ with final /s/ (-ís) (see Table 2).

This conversational corpus, collected by community members and transcribed following defined protocols, provides ideal data to probe the tuteo/voseo variation in Chile, and in particular, the impact that priming has on its use.

3. The Chilean Second-Person Singular

Chilean Spanish is characterized by the co-existence of two 2sg familiar forms, both in the pronouns tú and vos, and in the verb forms, tuteo and voseo, illustrated in (4) and (5), respectively. These forms can be contrasted with usted, the so-called “polite” 2sg form (which takes a third-person singular verb form). There are just 55 occurrences of usted verb forms in CCSS, accounting for two percent of all 2sg verbs, almost all of which are either in quoted speech or are to older speakers. Thus, we consider usted to be outside the variable context, which we define as second-person singular familiar verbal forms.

(4) Tú tienes ahorros.

  ‘You _-tú have_-tuteo savings.’

(Savings; 1302; Carolina)

(5) o sea vos no tenís ningún concierto.

  ‘In other words you_-vos have_-voseo no concert.’

(Football; 947; Matías)

Chilean Spanish has distinct voseo forms throughout most of the verbal paradigm as seen in Table 2, which presents the morphology of the set of TAMs for verbs ending in -ar on the left, and -er or -ir on the right (illustrated for -er, which is identical to -ir for voseo, though not for tuteo). As can be seen, there are distinct forms for tuteo and voseo for all but the Imperative, Preterit and Synthetic Future, which we discuss below.

Tuteo and voseo differ in their morpho-phonological shape—tuteo is most characterized by /as/ and /es/ endings, and voseo by /ay/ and /is/ endings. This is likely to promote priming within the paradigms, as shared shape renders individual instances of tuteo or of voseo obviously similar, even for those ignorant of their grammatical status. The voseo > voseo priming is not, however, dependent on shared morpho-phonological shape, as we show below (Section 4.1).

Table 2. Chilean 2sg verb morphology by tense-aspect-mood (Adapted from Torrejón 1986, p. 678).

	−AR Verbs		−ER/−IR Verbs
	tuteo	voseo	tuteo	voseo
Present indicative	lleg- as	lleg- ái	com- es	com- ís
Imperfect indicative	lleg- abas	lleg- abai	com- ías	com- íai
Conditional	lleg- arías	lleg- aríai	com- erías	com- eríai
Present subjunctive	lleg- ues	lleg- uís	com- as	com- ái
Imperfect subjunctive	lleg- aras	lleg- arai	com- ieras	com- ierai
Imperative	lleg- a		com- e
Preterit	lleg- aste		com- iste
Synthetic Future	lleg- arás		com- erás

Table 2. Chilean 2sg verb morphology by tense-aspect-mood (Adapted from Torrejón 1986, p. 678).

	−AR Verbs		−ER/−IR Verbs
	tuteo	voseo	tuteo	voseo
Present indicative	lleg- as	lleg- ái	com- es	com- ís
Imperfect indicative	lleg- abas	lleg- abai	com- ías	com- íai
Conditional	lleg- arías	lleg- aríai	com- erías	com- eríai
Present subjunctive	lleg- ues	lleg- uís	com- as	com- ái
Imperfect subjunctive	lleg- aras	lleg- arai	com- ieras	com- ierai
Imperative	lleg- a		com- e
Preterit	lleg- aste		com- iste
Synthetic Future	lleg- arás		com- erás

Historically, the Imperative, Preterit and Synthetic Future had distinct tuteo and voseo forms (e.g., Torrejón 1986, p. 679), but today, only what were the tuteo forms remain in wide use, while the voseo forms are vanishingly rare. In the CCSS, of 560 Imperative tokens, there are no distinct voseo forms; of 167 2sg familiar Preterit tokens, there is only one token of the -stes form, which is purported to be historically voseo (Rivadeneira Valenzuela 2009; Torrejón 2010)3; and the Synthetic Future does not occur, as the Analytic Future is the preferred form (n = 86). As invariable forms, these TAMs fall outside the variable context and must be excluded from analysis (as has been done in previous variationist studies of voseo, e.g., Fernández-Mallat 2018, p. 100; Rivadeneira Valenzuela 2016, p. 70).

However, here our interest is in how speakers classify these forms. Do they remain associated with their tuteo origins, as an instance of a general [Verb-tuteo] construction, or are they genuinely neutral, as syncretic forms that collapse the tuteo/voseo distinction? As the tú pronoun occurs with both tuteo and voseo, presence of a subject pronoun is not a reliable indication of the form of these verbs, and furthermore, subjects are most commonly left unexpressed, as in example (6) in the Imperative, and (7) in the Preterit. Thus, we appeal to priming to establish the status of these forms: if they indeed share a mental representation with the tuteo forms, then we would expect them to favor a subsequent tuteo; the lack of such a priming effect would suggest that this association has been lost.

(6) Consíguete un pololo.

  ‘Get_{-imperative-fixed} yourself a boyfriend.’

(Cousins; 1180; Emilia)

(7) te pusiste huevón ya,

  ‘you-_Ø already turned_{-preterit-fixed} into an idiot,’

(Takeaway; 4; Claudia)

Another non-variable form is the lexically specific, fossilized construction cachái, the voseo form of the verb cachar ‘to understand, get’ (possibly from English ‘to catch’, Rivadeneira Valenzuela 2016, p. 100). In the Present Indicative, cachar occurs virtually categorically in the voseo form in the CCSS (with just one cachas in contrast to 360 cachái), consistent with what has been reported in other studies (e.g., Fernández-Mallat 2018, p. 70; Rivadeneira Valenzuela 2016, p. 100; San Martín Núñez 2011, pp. 159–62). Like the invariable TAMs above, then, cachái falls outside the variable context. But the question remains, however, of whether speakers perceive cachái to be an instance of the [Verb-voseo] construction. It is used primarily as a discourse marker, akin to English ‘you know’, as in (8) (322/360). It also occurs as a full verb (n = 28), for example, with a direct object, meaning to perceive or realize something (cf. Urzúa-Carmona 2006, pp. 100–5), as in (9), and with a variably expressed pronoun (nearly always tú, 10/11 instances).

(8) tú enrolái como si fueran más delgados,

  cachái?

  ‘you_-tú roll_-voseo ((cigarettes)) as if they were thinner,

  you_-Ø know_-cachái?’

(A bit of everything; 401–402; Andrea)

(9) y ahí cachái cómo es el nivel de ellos po huevón.

  ‘And that’s when you_-Ø realize_-cachái what level that they’re at man.’

(Barcelona; 925; José)

Cachái has emerged within the last fifty years or so, in parallel with the reported rise of voseo (e.g., San Martín Núñez 2011, pp. 159–62). It is highly frequent in the speech of young Chileans today, accounting for 94% of all instances of the verb cachar in the CCSS, and around 25% of all 2sg familiar forms produced by speakers of 35 years and younger (of a total of 1242 tokens).4 For one speaker (Marcela, 31 years old; 123 tokens), cachái constitutes 55% of her 2sg familiar tokens. Though cachái does occur in discourse of older speakers, it is much rarer, accounting for just 3% of all 2sg familiar forms (of a total of 904).

Do speakers consider cachái to be a voseo form, that is, a manifestation of the [Verb-voseo] construction, or are they unaware of its structure? Although it has a similar phonological shape to the voseo Present Indicative, with the /ay/ ending, there is no tuteo form to contrast it with, and it has been proposed by some authors that it has lost its status as a verb (e.g., Gille 2015; Mondaca Becerra et al. 2015; Rivadeneira Valenzuela 2016, p. 100). Here, too, priming provides insight into the status of this form: if it is recognized as a voseo form, then it should prime a subsequent voseo; the lack of such an effect would suggest that it has become fully autonomous and is no longer attached to the voseo paradigm.

A final consideration is the impact of a pronoun. As already noted, Spanish has variable subject expression, and thus subject pronouns may be expressed (as tú or vos in this case) or they may be left unexpressed. The vos pronoun, however, shows very minimal use. The CCSS presents a total of 304 tokens of the tú subject pronoun, but just 10 of the vos subject pronoun.5 And while the vos subject pronoun does not occur with tuteo, the tú pronoun occurs with both tuteo and voseo verb forms, resulting in the occurrence of both [tú + Verb-tuteo] and [tú + Verb-voseo] constructions. The overwhelmingly favored form with both tuteo and voseo (accounting for 80% of tokens occurring in the variable context) is with an unexpressed subject. While the use of voseo with no pronominal subject has been termed crypto-voseo, in the sense that the very existence of a parallel voseo paradigm is concealed by the absence of the vos pronoun (Lipski 1994, p. 143), it has also been proposed that the “underlying” pronoun in such a context is “always tú, as a neutral unmarked form” (Rivadeneira Valenzuela 2016, p. 93). Here again, we use priming to ascertain the status of “mixed voseo” [tú + Verb-voseo], and the impact the presence of a tú pronoun has on the representation of a voseo verb form.

We have established, then, that instances of the Imperative, Preterit and cachái fall outside the variable context, but that tokens with a tú pronoun do exhibit variability and thus can be included. Figure 1 depicts the distribution in the data across the non-variable and variable contexts. What is of particular note here is that the non-variable contexts make up over one half of all 2sg familiar instances—the Imperative and Preterit together represent 35%, and cachái a further 17%. Within the variable context, the rate of tuteo is 20%, in contrast with 80% voseo. What factors impact this variation?

Change over Time

The rate of tuteo has declined in recent Chilean history, as voseo has grown in use. This is evident in the data under study here in comparisons across speakers of different ages. As has been robustly demonstrated in sociolinguistic research, the relative stability of patterns of speech in adults allows for comparisons across age groups to serve as a proxy for language use in different time periods, and differences as indicative of change in apparent time (Sankoff 2006).

The rates of tuteo vs. voseo by age are presented in Figure 2, which illustrates an overall drop in tuteo for younger speakers, supporting the change over time towards voseo. Comparing the age groups in Table 1, the rate of use drops from 32% tuteo (138/430) for the older speakers to just 12% (77/629) for the younger speakers.

We also see here a large amount of variability even for speakers of similar ages. This is partly due to a range of conditioning factors, including social (gender and socioeconomic class) and linguistic factors (previous realization, subject pronoun expression, discourse type, clause type, morphological class of verb). Examination of the impact of these factors by Callaghan (2020) revealed wholesale change in the nature of this variation, in that not only has the rate of tuteo dropped over time but the conditioning has also changed. Older and younger speakers differ in terms of the set of significant predictors (e.g., discourse type is significant for older speakers only, for whom tuteo is disfavored with generic subjects), and in the relative strength of the effect those factors have (e.g., clause type has a stronger effect in younger than older speakers, with a greater disfavoring of tuteo in questions). In some cases, they differ in the direction of effect (e.g., tuteo is favored by higher social classes for the older age group, but by lower social classes for the younger age group) (cf. Callaghan 2020, p. 195).

One predictor that remains stable across speakers of different ages is previous realization, as a manifestation of priming, which furthermore has the greatest magnitude of effect for both age groups (and overall, see Figure 6 below, Section 4.4). We will therefore now turn to consider this effect in more detail, paying particular attention to what it reveals about speaker awareness of tuteo/voseo variation.

4. Priming as a Diagnostic of Associations between tuteo and voseo

Considering priming to be a measure of speaker associations between forms, priming effects can shed light on whether speakers distinguish between the tuteo and voseo paradigms, or whether, as a result of mixing, stigmatization and “invisibilization”, they have become conflated in speakers’ variable grammars. We address this question in three ways. We first consider the effect of a variable tuteo or voseo form in the preceding discourse. We then examine the effect of the presence of a non-variable 2sg familiar form, namely verbs occurring in TAMs that are historically tuteo (in the data here, the Imperative and Preterit), and the fossilized voseo form cachái. For our third set of analyses, we consider the effect of the “mixed voseo” and whether the presence of a tú pronoun weakens the association between voseo forms.

To provide an overall picture of the nature of the priming effect, we describe the general trends through descriptive statistics and visualizations, before testing for statistical significance in those trends with generalized linear mixed effects models and a random forest analysis. For these priming analyses, we exclude tokens for which the previous form or the status of the pronoun (as tú or unexpressed) could not be coded due, for example, to unclear speech (n = 43); seven instances of a vos pronoun in the variable context were also excluded, leaving a total of 1009 tokens for analysis.

4.1. The Impact of the Form of the Previous Realization

We begin by examining the role that the form of previous realization plays in conditioning the tuteo/voseo variation. If speakers recognize and negotiate two separate paradigms, then this should be evident in the patterns resulting from priming: a previous voseo should favor the repetition of a subsequent voseo form, and a previous tuteo should favor a subsequent tuteo. An absence of priming would support the interpretation that the two paradigms are collapsed, and that speakers do not distinguish between forms.

The sample was coded for the form of the most recent 2sg familiar verb occurring within the previous five IUs, generally, around 20 words. As priming is known to weaken with distance (e.g., Szmrecsanyi 2005, p. 139; Travis 2007, pp. 119–21), this relatively close measure maximizes the possibility of observing an effect. Previous 2sg familiar forms by both the interlocutor and the same speaker were included. Of the 356 instances of a previous tuteo or voseo, 69 were produced by the interlocutor. These tokens maintain a similar priming effect to those produced by the same speaker, and thus we do not separate out same speaker- from interlocutor-produced primes for the purposes of the analyses.6

Five possible previous realizations were coded: a previous tuteo, as in (10) (where the prime is underlined, and the target in bold); previous voseo (11); previous invariable Imperative or Preterit forms (12); previous cachái (13); and no other 2sg familiar token within the previous five IUs.

(10) no andabas conmigo cuando te perdías.

  ‘You_-Ø weren’t_-tuteo with me when you_-Ø got lost_-tuteo.’

(Memories; 1261; Trinidad)

(11) Sabís la cagada que le vai a dejar a tu amigo huevona?

  ‘Do you_-Ø know_-voseo the problems you_-Ø are going_-voseo to leave your friend idiot?’

(Police; 484; Matilda)

(12) Métete cuando --

  cuando quieras.

  You_-Ø join in_{- imperative-fixed} when --

  whenever you_-Ø want_-tuteo.

(Barbeque; 1430–1431; Viviana)

(13) Cristián: yo pierdo tiempo con esas huevadas,

      cachái?

  Claudia: perdís tiempo.

  Cristián: ‘I waste time on that shit,

      you_-Ø know_-cachái?’

  Claudia: ‘You_-Ø waste_-voseo time.’

(Takeaway; 221–223; Cristián/Claudia)

Table 3. Rates and relative frequency of tuteo vs. voseo by previous realization.

	tuteo		voseo		Total
	%	N	%	N	N
Previous tuteo	71%	48	29%	20	68
Previous voseo	8%	22	92%	266	288
No 2sg familiar token in preceding 5 IUs	23%	110	77%	375	485
Previous imperative/preterit fixed form	16%	19	84%	102	121
Previous cachái	13%	6	87%	41	47
Total	20%	205	80%	804	1009

shows the distribution of the data according to these five previous contexts, and the rate of voseo vs. tuteo in each, which is also depicted in Figure 3. We see in Table 3 that one-third of the tokens have a tuteo or voseo in the previous five IUs (n = 356), and just under one-half have no 2sg familiar token within the previous discourse, providing sufficient data for meaningful comparisons. A small proportion occurred in the context of a previous Imperative or Preterit, or of a previous cachái, which we will consider in the following section. For now, we concentrate on the context with a previous (variable) voseo or tuteo, as compared with no previous mention, the two left-most columns and the right-most column in Figure 3.

As can be seen, the rate of tuteo is overwhelmingly highest with a tuteo in the previous discourse, at 71%, and lowest in the context of a previous voseo, at just 8%; in the absence of a prime (with no previous 2sg familiar form), the rate is in between, at 23%. This clearly shows a priming effect, that will be corroborated statistically in Section 4.4 below—tuteo is favored in the context of a previous tuteo, and is disfavored in its absence, particularly in the context of a previous voseo.

We are able to verify that the favoring of tuteo following a preceding tuteo, and voseo following a preceding voseo, is a real priming effect, and not simply the inevitable repetition of the same form by speakers with very high rates of one or the other form by comparing the impact of the previous realization across speakers with different baseline rates of tuteo (cf. Sankoff and Laberge 1978; Torres Cacoullos and Travis 2018, p. 90). Speakers were binned into four groups according to their overall rate of tuteo, separated by 20 percentage points. It was not possible to compare the speakers individually, as few produce enough tokens to reliably identify individual differences—the 36 speakers produce an average of 28 tokens each, with a median of 24, and a range from three to 89. Only five speakers produce over 50 tokens, seven between 30 and 50, and 14 under 20. Fifteen speakers (10 of whom produce under 30 tokens) do not use tuteo at all in the data, and they are excluded from these comparisons.7

Figure 4 shows the rate of tuteo for these groups of speakers when there is no 2sg form in the previous five IUs (the solid, darkest line), in the context of a previous tuteo (the line marked with circles) and a previous voseo (marked with squares). Here we see that, for both low and high users of tuteo, their rate of tuteo is higher when there is a tuteo in the preceding environment and their rate of voseo is higher when there is a voseo in the preceding environment; that is, priming holds independently of the rate of tuteo. That priming is not limited to any individual speakers is confirmed in the mixed-effect models reported below (Section 4.4), where the priming effect emerges with speaker run as a random effect.

A further observation from Figure 4 is that that the strongest tuteo > tuteo priming is found among those with lower overall rates of tuteo: for speakers with rates of tuteo of under 50% (the first two bins), the rate of tuteo in the presence of a previous tuteo deviates more from that in the absence of a prime than it does for speakers with rates of over 50%. Conversely, the strongest voseo > voseo priming occurs amongst those speakers with high baseline rates of tuteo (the last bin). Such patterning is consistent with what has been reported in the priming literature, whereby less frequent variants tend to exert a stronger priming effect, under what has been described as “surprisal” (Jaeger and Snider 2007). Thus, while in general, speakers are sensitive to the form of a previous mention, given the predominance of voseo, those with lower rates of tuteo are even more so. This is corroborated by the findings across age group: younger speakers, who tend to have lower rates of tuteo (as seen in Figure 2), exhibit a stronger tuteo > tuteo priming effect than older speakers (Callaghan 2020, p. 202). From a usage-based perspective, we might interpret this in terms of an unexpected instance being more salient (Jaeger and Weatherholtz 2016), and therefore more readily retrieved and reused in the subsequent discourse.

The priming observed suggests that speakers recognize the [Verb-voseo] construction as distinct from the [Verb-tuteo] construction. But to establish this, we must also demonstrate that priming holds across different morpho-phonological realizations of voseo, and is not limited to instances that share the same form. If priming were reliant on similarity of shape, then it should be manifestly stronger when the prime and target share the same form (/ay/ > /ay/ or /is/ > /is/), as in (14) below, than when they do not (/ay/ > /is/ or /is/ > /ay/) as in (11) above. However, although the rate of voseo (vs. tuteo) is marginally higher when the voseo prime and target take the same shape than when they do not (88%, 168/190 vs. 82%, 98/120), this is not significant (p = 0.13, Fisher’s Exact Test).8 That different voseo morpho-phonological endings prime each other is evidence for voseo as an abstract category in speakers’ mental representations. The priming here is thus functioning at a schematic level while being marginally strengthened when morphological specificity is shared (cf. Rosemeyer and Schwenter 2017, p. 30; Travis et al. 2017, p. 294).

(14) llegái al aeropuerto,

   se te suben dos personas,

   y las vai a dejar a Santiago,

   ‘you_-Ø arrive_-voseo at the airport,

   Two people get in,

   and you_-Ø go_-voseo to drop them in Santiago,’

(Back to Santiago; 76–78; Ernesto)

4.2. The Status of Non-Variable Constructions: Imperative, Preterit and cachái

Having established tuteo > tuteo and voseo > voseo priming, we can now use priming to test the status of the non-variable constructions, and whether or not they retain a link with their respective tuteo/voseo origins. The question we ask is: do speakers associate the historically tuteo TAMs (here, the Imperative and Preterit) with the [Verb-tuteo] construction and the voseo-based cachái with the [Verb-voseo] construction? Evidence of this would be found in a rate of tuteo in the context of a previous Imperative or Preterit parallel or similar to that in the context of a previous tuteo, or at least a higher rate than when there is no prime. Similarly, for cachái, evidence would be a rate of voseo in the context of a previous cachái parallel or similar to that in the context of a previous voseo and higher than that in the absence of a prime.

As we see in Table 3 and Figure 3 above, the rate of tuteo in the context of a previous Imperative or Preterit is just 16%. This is less than one quarter of that in the context of a previous tuteo (71%), and thus these are clearly treated quite distinctly. It is only slightly lower than in the absence of a prime (23%), and we will see below that this difference is not significant (Table 5, Section 4.4). This indicates that the Imperative and Preterit have shed their historical associations with tuteo, and function as syncretic, neutral forms.

For cachái, the rate of tuteo following a previous cachái (13%) is in between that when there is no prime (23%), and when there is a previous voseo (8%). The statistical models we report on below (Section 4.4) indicate that, in the data overall, the rate of tuteo in the context of a previous cachái is not significantly different from that when there is no prime (p = 0.44), suggesting that cachái may be a neutral form. We also find that the impact of a previous cachái is significantly different from that of a previous voseo in the data overall (p = 0.02). However, this difference is not significant for the young speakers, who are the main users of cachái (p = 0.09), indicating that for them at least cachái retains a degree of association with the voseo paradigm.

The classification of these forms is not a minor issue for the analyst, as they account for approximately one-half of all 2sg familiar tokens, as we saw in Figure 1. The patterning we have observed for priming validates the exclusion of both syncretic forms and cachái from the tuteo/voseo variation, and illustrates how priming can help illuminate what counts as an instance of a construction, and so what falls in, and outside of, the variable context (cf. Tamminga 2016).

4.3. The Status of “Mixed voseo”

The usage data thus far demonstrate strong awareness of distinct voseo vs. tuteo paradigms, despite metalinguistic commentary to the contrary. A remaining question is the status of mixed voseo, that is, a voseo verb form with a tú pronoun, seen in (15) (and in (2) above), which we might expect to be less associated with voseo than instances that occur without a tú pronoun. If this is the case, then there should be a weaker priming effect when the target verb occurs with a tú pronoun than without.

(15) Tú creís que va a entender eso el niño?

   ‘Do you_-tú think_-voseo the kid is going to understand that?’

(Family; 361; Sara)

For a first view, we look to patterns of subject expression. While theoretically there are three options for 2sg subject realization (tú, vos, Ø), due to the rarity of the vos pronoun (just 10 tokens in the CCSS), only two are fully exploited by speakers, tú or an unexpressed subject. Subject expression can therefore be used as a proxy for the strength of association between the pronoun and verb: if speakers do not distinguish between the two paradigms, then we would expect the rate of occurrence with a tú pronoun to be similar across voseo and tuteo.

Table 4. Rates of subject expression (tú vs. Ø) for tuteo vs. voseo.

	% tú	N	% Ø	N
Tuteo	31%	63	69%	142
Voseo	18%	147	82%	657
Total	21%	210	79%	799

compares the rates of subject pronoun expression with each verb form. As can be seen, there is a higher rate of pronominal expression with tuteo than voseo (31% vs. 18%), indicating that, though speakers do use a tú pronoun with voseo verb forms, they are less likely to do so than with a tuteo verb form. The favoring effect of a tú pronoun on rate of tuteo also emerges as significant in the statistical model (see Table 5, Section 4.4). Furthermore, mixed voseo accounts for under 15% of the data (147/1009), suggesting that this is a relatively minor phenomenon, despite the reported predominance of this form in the literature (e.g., Torrejón 1986, p. 682).

It is important to note that multiple factors condition patterns of subject expression (cf., Torres Cacoullos and Travis 2018, chp. 5), such that care needs to be taken in comparing overall rate differences. A preponderance of tokens in contexts favorable to pronominal expression (e.g., non-coreferential contexts) may result in a higher rate of expression. Such analyses are beyond the scope of this paper, and we now turn to priming as a further measure of the status of mixed voseo.

Figure 5 shows the rate of tuteo across three priming contexts (previous tuteo, no 2sg familiar token in the previous five IUs, previous voseo), with a tú pronoun (darker column, on the left of each pair) and with no subject pronoun (lighter column on the right). First, note that, in the absence of a prime, we observe the predicted effect: a higher rate of tuteo with a tú pronoun than with no pronoun. But we also observe a nearly identical priming effect with and without a tú pronoun: with a tú pronoun, the rate of tuteo rises from 13% with a voseo previous mention to 33% with no previous mention, to 71% with a tuteo previous mention; with no subject pronoun, it rises from 7% to 19% to 70%. That is, whether or not the target occurs with a tú pronoun, speakers are sensitive to the overall tendency to repeat the previous form.

It is particularly telling that, in the context of a previous tuteo, the presence of a pronoun has no impact on the rate of tuteo; that is, the previous tuteo overrides the pronoun effect. Although the presence of a tú pronoun does impact the rate of tuteo in the context of a previous voseo, the effect of a previous voseo is still felt in this context, where the rate of tuteo is barely one-third that in the absence of a prime (13% vs. 33%). The favoring of tuteo by a tú pronoun is confirmed in the statistical models presented below (Table 5, Section 4.4), which also reveal that there is no significant interaction between the presence of a pronoun and the previous realization, indicating that each holds independently of the other.

The avoidance of the vos pronoun, the use of a tú pronoun with a voseo verb form, and reports that Chileans claim not to use voseo because they associate it exclusively with the pronoun may seem to support an interpretation of pronouns as more salient in speakers’ minds than verb forms. This is precisely what has been proposed in the literature, in accordance with the notion that the greater saliency of the vos pronoun as a lexical item has contributed to it becoming stigmatized, and therefore avoided, while voseo verb forms fly under the radar, allowing for their expansion in use (Bertolotti 2015, p. 19; Huerta Imposti 2011–2012, p. 52; Hummel 2010, p. 112; Stevenson 2007, p. 167). This is not, however, what we observe in usage patterns, where priming of verb forms is upheld even in the presence of a tú pronoun. Though mixed voseo has been used as evidence of the loss of awareness of the voseo paradigm (Carricaburo 1997, p. 34; Huerta Imposti 2011–2012, p. 54), what the priming data show is that even mixed voseo is treated by speakers as an instance of the [Verb-voseo] construction.

These results also allow us to respond to the suggestion that there exists an underlying tú pronoun for all instances of 2sg familiar verbs with an unexpressed subject (e.g., Rivadeneira Valenzuela 2016, p. 93). The priming patterns we observe do not support this interpretation, but rather indicate that speakers create their utterances with tuteo and voseo, and with or without pronouns, based on probabilistic usage-based factors, influenced by, among other things, what precedes in the discourse.

4.4. Statistical Analyses

Having observed the general patterns, we now test the statistical significance of these differences, using generalized linear mixed effects models with the glmer() function in R (Bates et al. 2019; R Development Core Team 2019). Models were fit with verb form (tuteo/voseo) as the dependent variable, and previous realization, presence of a tú pronoun, and age as independent variables. For previous realization, the absence of a prime was used as the reference level; an unexpressed subject was the reference level for pronoun; and age was modeled continuously. Three- and two-way interactions between previous realization, pronoun and age were tested, and then pruned from the model, as none were significant. To take into account differences between speakers and verb types, speaker and verb were included as random intercepts. Table 5 presents the final model summary.

Table 5. Output of a generalized linear mixed effects model predicting tuteo.

	Estimate	Std. Error	Z	p
(Intercept)	5.86752	1.16632	5.031	<0.001
Previous tuteo	−1.53467	0.33822	−4.538	<0.001
Previous voseo	0.87474	0.28351	3.085	<0.01
Previous syncretic	0.37496	0.32207	1.164	=0.24
Previous cachái	−0.40438	0.52527	−0.770	=0.44
Pronoun tú	−0.90796	0.23260	−3.903	<0.001
Age	−0.09075	0.02627	−3.454	<0.001

Overall tuteo 20% (205/1009); negative coefficients are associated with a higher rate of tuteo. For 36 speakers, variance = 2.7 (SD = 1.6) and for 134 verb types, variance = 0.1 (SD = 0.32).

Table 5. Output of a generalized linear mixed effects model predicting tuteo.

	Estimate	Std. Error	Z	p
(Intercept)	5.86752	1.16632	5.031	<0.001
Previous tuteo	−1.53467	0.33822	−4.538	<0.001
Previous voseo	0.87474	0.28351	3.085	<0.01
Previous syncretic	0.37496	0.32207	1.164	=0.24
Previous cachái	−0.40438	0.52527	−0.770	=0.44
Pronoun tú	−0.90796	0.23260	−3.903	<0.001
Age	−0.09075	0.02627	−3.454	<0.001

Overall tuteo 20% (205/1009); negative coefficients are associated with a higher rate of tuteo. For 36 speakers, variance = 2.7 (SD = 1.6) and for 134 verb types, variance = 0.1 (SD = 0.32).

First, this model corroborates the overall priming effect we saw in Table 3 and Figure 3: compared with when there is no prime in the previous discourse, the rate of tuteo is significantly lower in the context of a previous voseo and significantly higher in the context of a previous tuteo. Thus, tuteo primes a subsequent tuteo and voseo primes a subsequent voseo, confirming that these speakers do recognize a relationship between distinct verbs produced in the voseo vs. tuteo forms. Despite a lack of explicit metalinguistic awareness and mixing of voseo verb forms with a tú pronoun, in actual usage, speakers keep the two paradigms separate.

What of previous syncretic forms and cachái? Table 5 indicates that neither is significantly different from contexts where there is no prime, in accordance with our observations above. Releveled models run with a previous tuteo as the reference level indicate that we are significantly less likely to get tuteo following syncretic forms than following tuteo (β = 1.91, p < 0.001), supporting the notion that these TAMs have lost their association with tuteo. Cachái is slightly different. Although overall, we are more likely to get a subsequent voseo following another voseo than following cachái (β = 1.28, p < 0.05), the effect is not significant for the younger speakers, the main cachái users (β = 1.19, p < 0.09), suggesting that the association between cachái and voseo has not been entirely lost. That is, cachái may not be a central exemplar of the [Verb-voseo] construction, but it does retain a link to it.

The model in Table 5 also indicates that the priming effect is not an artefact of speakers with very high rates of voseo, as the impact of a previous voseo or tuteo holds even when we take account of individual speakers’ preference by including a random intercept for speaker in the model. To further test this, we ran another model identical to the above, but with the addition of an interaction between rate of tuteo and previous realization, and no interaction was found between a previous voseo and rate of tuteo (β = 1.40, p = 0.24), nor a previous tuteo and rate of tuteo (β = 2.23, p = 0.10).

Finally, the model in Table 5 also tests the impact of a tú pronoun, which favors tuteo, as we saw in Table 4 and Figure 5. The model shows that the impact of a previous tuteo is stronger than that of a tú pronoun (with z scores of −4.54 and −3.90 respectively). The same is not so for a previous voseo, which has a slightly lower z score than that of the pronoun (3.10). Nevertheless, that voseo priming holds in the presence of a tú pronoun is evident in the fact that an interaction between the presence of a pronoun and previous voseo fails to reach significance (β = 0.11, p = 0.87)—priming and pronoun presence are therefore indeed independent effects.

4.5. The Effect of Priming in Conjunction with Other Conditioning Factors

We mentioned above (Section 3.1) that multiple factors condition the choice of tuteo over voseo. Up to now, we have focused on a subset of these that are directly relevant to priming. Prior work conducting regression analyses that include a full set of linguistic and social predictors has found priming not only to operate alongside these other factors, but to be the strongest predictor of this variation (Callaghan 2020, p. 197). In order to confirm this here, we conduct one final set of analyses, including the identical dataset to that employed above (n = 1009), and adding three new predictors: discourse type (reported speech, generic subjects, specific subjects); clause type (questions, main clause declaratives, subordinate clauses), and gender.

To determine the impact of priming while taking into account a full range of predictors, we conduct a conditional random forest analysis. Random forests are built from multiple conditional inference trees, a statistical approach that makes recursive binary splits in the data, according to the strength of the predictors in each subsequent subset of the data. Random forests measure the overall importance of each predictor included in the model by averaging the results across multiple conditional inference trees (here, 1000), each based on a randomly generated subset of the data (Tagliamonte and Baayen 2012, pp. 159–60). Such a model “ensures that the evaluation of a variable’s importance takes into consideration its behavior in relation to other variables in its ranking” (Schnell and Barth 2018, p. 64). The result of this analysis, obtained using the ctree() function from the “party” package for R (Hothorn et al. 2006), is presented in Figure 6.

As can be seen, the strongest effect is that of individual speaker, as is common in such analyses (Tagliamonte and Baayen 2012, p. 162).9 Priming, however, is the next strongest predictor, and is substantially stronger than any other. Age, pronoun and discourse type are ranked next, followed by gender with a marginal effect, and clause type is found not to be worth further consideration. Thus, priming is not only upheld when considering a full set of predictors but it exerts the strongest effect.

5. Conclusions

We have sought here to gain access to information about speaker representations through the study of actual usage. While perception tasks and surveys are often employed to extract attitudinal information, speakers’ own intuitions and judgements do not necessarily produce reliable data for stigmatized variables (Sankoff 1988, pp. 145–46), as has been shown to be the case for voseo use (Bishop and Michnowicz 2010). Here we have demonstrated how spontaneous speech data can offer insights into speakers’ implicit understanding of constructions, examined in terms of degrees of association: because priming occurs across related constructions, the relative strength of priming provides an indication of the strength of the relationship.

The fact that speakers’ choice between voseo and tuteo is influenced by the previous 2sg familiar form used in the discourse provides strong evidence that they do distinguish between these paradigms; that is, they recognize, and keep separate, [Verb-tuteo] and [Verb-voseo] constructions. For the non-variable forms, however, we found different results. The Imperative and Preterit, despite deriving historically from tuteo, appear to no longer be associated with it; when occurring in the previous discourse, their impact is no different from that of the absence of a prime, and it is significantly different from that of a previous tuteo. Cachái, on the other hand, appears not to have wholly lost its association with the [Verb-voseo] construction, as, for young speakers at least (the main cachái users), the impact of a previous cachái in the discourse is not significantly different from that of a previous voseo. This illustrates that mental representations can change over time, and that non-variable, or fossilized, forms may not be homogeneous in their degree of autonomy from their source constructions (cf. Bybee 2006, p. 715).

As we have seen, voseo and tuteo take distinct morpho-phonological forms (e.g., estás vs. estái ‘you are_-tuteo/voseo’, tienes vs. tenís ‘you have_-tuteo/voseo’), and it is highly likely that this aids in keeping these paradigms apart. But we have also seen that voseo > voseo priming is not dependent on repetition of the same morpho-phonological shape, demonstrating the existence of a schematic [Verb-voseo] construction. Furthermore, the priming of voseo verb forms is retained even with a tú pronoun in the so-called mixed voseo. Thus, though overt expression of a tú pronoun favors tuteo over voseo, it does not override the impact of the previous form, indicating that [tú + Verb-voseo] is still an exemplar of the more general [Verb-voseo] construction.

We can capture the mental representations evidenced here with the set of associations depicted in Figure 7, with [Ø + Verb-voseo] (the majority variant) at its center: the tú pronoun is associated with both tuteo and voseo, but more strongly with the former; the vos pronoun, though minimally used, is associated with voseo verb forms; the syncretic Imperative and Preterit forms are associated with neither construction; and cachái remains (weakly) associated with voseo.

This picture does not seem to be consistent with Chileans’ metalinguistic awareness of these forms, which revolves around the vos pronoun, the subject of much commentary and stigmatization, while the tuteo/voseo verb forms pass unnoticed. Nor is it wholly consistent with much of the literature on this topic, that has assumed the existence of an underlying form for each syncretic form, and for instances with an unexpressed pronoun. The results presented here are, however, consistent with a usage-based understanding of grammar, according to which speakers do not construct their speech around abstract underlying forms, but rather as part of their constantly unfolding experience with language. Further, this can function at a highly local level, where one factor constraining variant choice is the form that has been used previously, as evidenced in structural priming.