"Subject pronoun expression in oral Mexican Spanish" (Yolanda Lastra & Pedro Martín Butragueño, in Subject pronoun expression in Spanish. A Cross-Dialectal Perspective. Ed. Ana M.

SUBJECT PRONOUN EXPRESSION IN ORAL MEXICAN SPANISH

Yolanda Lastra UNIVERSIDAD NACIONAL AUTÓNOMA DE MÉXICO - EL COLEGIO DE MÉXICO

Pedro Martín Butragueño EL COLEGIO DE MÉXICO

1. Introduction Previous works and theoretical framework1 The analysis of presence vs. absence of Spanish subject personal pronouns (SPPs) rests on a solid background (cf. Flores-Ferrán 2007a). The first works —Barrenechea & Alonso (1977) on Buenos Aires Spanish, and especially Silva-Corvalán (1982) on Mexican-American at West Los Angeles (WLA), Morales (1982) on Puerto Rico, Bentivoglio (1980, 1987) on the expression on first person subject in Caracas— put under study some of the variables commonly considered later in the majority of analyses: co-reference, ambiguity, verbal type, among others. Also, from this moment on the use of VARBRUL programs begins, so that it is possible to establish comparisons between frequencies and log-probabilities from the beginning of the eighties. Hochberg (1986) on data from Puerto Rico shapes the clearest defense of the Functional Compensation Hypothesis (FCH): null SPPs rises when deletion of final /s/ and /n/ provokes higher ambiguity. The series of papers published by Cameron (1993, 1995, 1996, 1997) explores the value of constraints through the comparison among different dialects, especially the Puerto Rico and Madrid varieties, questioning —among other things— the role of the functional perspective, given the similar weight of ambiguity in dialects with and without deletion. The Constant Rate Hypothesis (Kroch 1989a, 1989b; Cameron 1993, 319; Martín Butragueño 1999) predicts that the strength of the linguistic constraints remains constant, in a way that is not possible to extrapolate causes from contextual effects; the causes would underlay more abstract levels. Following Morales (1989), for Cameron (id.) that cause could be in a tendency to a consistent SVO order in Caribbean data. This point becomes a question of interest in the examination of pre-V or post-V position of overt SPPs (see below). In our opinion, Cameron's work opens, a second stage in the study of SPPs, characterized by the detailed discussion on constraints (Cameron 1993, Silva-Corvalán 1997) and the extension of analyses to many different varieties. So that, Travis (2005, 2007) studies the effect of priming; Travis & Torres Cacoullos examine the cognitive, constructional and mechanical effects on yo; Cameron (1995) discusses the scope of switch reference; Erker & Guy (2012) consider the enhancing effect of lexical frequency; Shin & Otheguy (2013, 431) propose a ceiling effect ("past a certain occurrence rate, the reference-tracking mechanism embodied in Spanish subject pronoun subject is likely to become inoperative"), and a women effect in contact situations (p. 432). This second moment culminates with the book by Otheguy & Zentella (2012), and their thorough analysis of language contact and dialect contact in the Hispanic population at New York City (NYC), taking into account a main division between Caribbean and Mainland dialects (see also !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!1 We would like to thank Poncho Aispuro's search of exempla in CSCM, as well as the observations by different members of the Grupo de investigación sociolingüística (cf. <http://lef.colmex.mx>), at the meeting of June 13, 2013. We are also very grateful to our editors, especially to Rafael Orozco, whose observations have been fundamental in improving a number of aspects in the analysis, and to two anonymous evaluators. Of course, remaining mistakes are ours.

! 2

Cameron & Flores-Ferrán 2003, Ortiz 2009, Claes 2011), on highland vs. lowland varieties (Michnowicz, this volume). For a long time, discussion has privileged two types of data: 1) data from Caribbean populations, surely by the exploration of functional hypothesis and the high rates of overt SPPs; 2) data from language in contact situations, especially those derived from contact with English (see additionally Amastae et al. 2000, Hurtado 2005, Flores-Ferrán 2007b, Bayley et al. 2012), because of the strong evidence of the contact between pro-drop and non pro-drop languages on overt/null SPPs. By contrast, the set of studies in monolingual and Mainland dialects has been less abundant, relatively speaking. In fact, it is surprising that in some cases data from immigrants in US cities have been presented as representative of the original populations (especially in the first stage of studies on SPPs). In the particular case of Mexican data, the abundance of work in contact settings, in contrast to the scarcity of material in monolingual settings is remarkable (but see Cantero Sandoval 1978, 1986, Montes Miró 1986, and partially Amastae et al. 2000). Social factors have had limited relevance in the results reported in the majority of cases, especially in monolingual settings. In the case of multilectal and multilingual settings, nevertheless, things are different, as Otheguy & Zentella (2012) and Shin & Otheguy (2013, on the effect of social class and gender in NYC) reveal. Our monolingual data do offer some statistically significant evidence of social variation, linked to age (and partially to educational level and gender), despite our hypothesis in (1b). On the other hand, it is increasingly clear that Spanish is divided in two areas with respect to the expression of SPPs: the Caribbean area (with higher rates of overt SPPs) and the rest (or Mainland, following Otheguy & Zentella 2012, with lesser rates), including of course Mexican Spanish. Our theoretical framework in syntactic change and variation is grammaticalization (Company 2012), sustained in a dynamic and realistic point of view about linguistic and sociolinguistic facts (Martín Butragueño & Vázquez Laslop 2002, Martín Butragueño 2014). Motivations and goal Our motivations are three: a) This study is part of a series devoted to explore, describe, and analyze syntactic variation in Mexico City (MC) Spanish (such as Lastra & Martín Butragueño 2010 on future tense, or 2012 on subjunctive mood); our proposal is to offer a contemporary point of view on spoken Mexican Spanish; b) the analysis of SPPs in Mexican Spanish has not been considered in enough detail. What we know today is derived mostly from contact situations; a monolingual point of reference is necessary for future studies (as well as analyses of contact between Spanish and indigenous languages, see Michnowicz in this volume); c) the analysis of SPPs in Spanish, with a background of thirty years of studies, permits the design of detailed analysis and careful comparison among different varieties. Few problems inside Spanish syntax suggest the discussion of transmission and diffusion (in the sense of Labov 2010) of a linguistic pattern as SPPs expression. Our goal is to test the next two hypotheses: Hypotheses Two hypotheses are considered through the analyses and discussion:

! 3

(1) a. The Constant Rate Hypothesis (CRH, Cameron 1993) is also effective in MC data. b. We do not expect remarkable social variation in MC data about SPPs (in view of the

results of the great majority of previous works in monolingual settings). Structure of the chapter In section 2 we expose the main methodological lines; in section 3, overall results and the effects of internal and social factors are discussed; section 4 is dedicated to the discussion of hypotheses in the light of results. Finally, the paper ends with some brief conclusions. 2. Methodology Informants, interviews, and the corpus Our data are a sub-sample of the Sociolinguistic Corpus of Mexico City (Corpus sociolingüístico de la ciudad de México, CSCM, see Martín Butragueño & Lastra 2011, 2012, and forthcoming, for a detailed discussion of methodology). The CSCM was collected between 1997 and 2007 and consists of recorded interviews and linguistic and sociolinguistic questionnaires applied to more than 320 speakers (approximately 500 hrs. of recording) in the Metropolitan area of MC (inhabited by twenty million speakers). The Corpus is divided in a number of sections: non-immigrants, immigrants, children and adolescents, marginal people, and recordings in groups. The nucleus of individual face-to-face interviews is in general the history of collaborators' lives; speakers are located and interviewed by a social networks method, that has allowed to document the speech of people of all kind of social origins, in their own settings (home or work), throughout different neighborhoods (traditional or modern; of low, medium or high social class), and to obtain a reasonable quantity of spontaneous material, with structure of recorded conversation (Silva-Corvalán 2001) in many fragments.2 Our sub-sample for this study consists of 18 speakers from the non-immigrant module, organized by educational level (low= six years of school or less; medium= between seven and twelve years of school; high= more than thirteen years —but usually sixteen or more); age (1, 20-34 years; 2, 35-54 years; 3, 55 and more years); and gender (male and female). The size of the sample is not very different with respect to other studies: e.g., Silva-Corvalán (1982) had interviewed seventy Mexican-Americans of WLA (the term "refers both to a person born in Mexico who have resided in the United States for a minimum of ten years and to persons of Mexican ancestry born and brought up in the United States", p. 97), but the analyzed data in the paper "included 338 expressed subjects out of a total of 795 tokens" (p. 102); besides, "the sample usually included, but was not restricted to, one narrative. [...] The sample for each speaker should contain a minimum of twenty expressed subjects of which a minimum of ten should be nonpronominal" (p. 103). Cameron (1993) analyzes 10 speakers from San Juan and 10 from Madrid. Otheguy & Zentella work with 140 speakers (2012, 77-80); from them, 68 are Mainlanders, and 23 Mexicans (pp. 29-30), and from those 23, 19 are Latin American raised (LAR), and 4 NYC raised (NYR). Among overall newcomers LAR in their sample, there are 20 Mainland

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!2 A great number of studies have been realized or are in progress with the CSCM: books and theses, chapters and articles about segmental phonology and intonation, syntax and morphology, discourse and conversational analysis, variation and change in apparent and real time, by a growing set of researchers, in and out of Mexico (more information in <http://lef.colmex.mx>).

! 4

newcomers (37, 39); only 6 are newcomers (Shin & Otheguy 2013, 434).3 Shin (2012, 132) records 13 children (2 from Querétaro, and 11 from Oaxaca). Erker & Guy (2012, 533) choose 12 speakers from Otheguy-Zentella corpus of NYC (six of Mexican and six of Dominican origin). Michnowicz (this volume) selects for Yucatan Spanish (YS) 20 speakers, with 10 bilingual and 10 monolingual speakers. Selection of data First, we collected 200 utterances including subject slots + FVs from the beginning of each interview (i.e., 3600 utterances). Second, we selected the cases with variable alternation between [Ø] and a SPP, so that our sample is formed by 2040 data (56.7% of the original collection).

Excluded utterances are lexical subjects (662/1560), only [Ø]/non-SPPs alternations (316/1560), impersonal constructions (159/1560), sentential subjects (48/1560), relative clauses with relativized subjects (192/1560), idiomatic phrases (31/1560), impersonal 3rd person pl (151/1560) and inclusive 1st person pl (1/1560). There are a number of cases where decision about choice as [Ø]/SPP is not obvious. In (2), it would be possible to analyze mi hermana in two ways: as a subject with intermediate material between this FN and its Vs (and then tiene, and especially va a cumplir and es la más grande would not be eligible, due to the presence of a lexical subject); or as a theme out of the predication (and then even va a cumplir and es la más grande would be eligible). In this example, the presence of ahorita, however, suggests mi hermana is a theme to the last V, given the unusual character of Mi hermana ahorita es la más grande, while the problem remains in Mi hermana ahorita tiene…, Mi hermana ahorita va a cumplir, where it is perfectly possible as a subject. (2) [sí porque mi] hermana ahorita <~orita> la más grande/ tiene/ [Ø] creo que [Ø] va a cumplir

cuarenta y siete o cuarenta y ocho [Ø] no sé bien/ [Ø] es la más grande (ME-308-12M-07, turn 127).

In example (3), even though nosotros as a subject seems to be the natural choice here (considering ninguno de as a Det), it has to be observed that a sentence like Ninguno logramos terminar nuestros estudios would be perfectly possible; however, under the second interpretation the example would be outside the study. (3) como familia/ fue muy carente/ [Ø] éramos muchos hermanos/ no logramos terminar ninguno

de nosotros nuestros estudios hasta después de adultos (ME-283-23M-06, turn 82). If lexical subjects are the vast majority of ineligible cases, the second group is only [Ø]/non-SPP alternations, as in (4). The third important group is relative clause with relativized subject; while we have considered some examples in which relative and personal pronoun coexist (as in 30), in general these data are not eligible (5). Occasionally, it is not simple to decide which is the subject; in (6), the subject of quieres seems to be a non-specific 2nd person sg, but the subject can also be a relative que, co-referent of una

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!3 Mexican newcomers constitute the group closer to our data. Furthermore, our 18 speakers are from MC, while NYC Mexicans in Otheguy & Zentella (2012) are from Mexico as a country.

! 5

persona. The problem is that in the first analysis, the example is eligible, but under the second, it is not. (4) entonces/ pues mi mamá dijo que/ que no/ que [Ø] no era posible// y yo/ pues siempre este/

tratando de bailar [Ø]/ y tratando de hacer [Ø] cosas en la danza/ por supuesto que [Ø] me encantaba y [Ø] me sentía/ la sílfide/ así [Ø] me sentía este (ME-259-32M-05, turn 5).

(5) ¿por qué desdichado?/ porque [Ø] abre/ no es que la droga en sí sea desdichada/ no/ para nada// cada quien// mide las cosas/ hay quien muere por droga/ hay quien muere por comer// hay quien muere por tomar mucho café (ME-294-33H-07, turn 109).

(6) cuando eres una persona que <~que:>/ que [Ø] las quieres cómo les llaman/ tranzar// pues <~pos> no te conviene tener un puesto fijo porque pues <~ps>/ a la semana siguiente te van a buscar y [Ø] no vas a estar (ME-312-12H-07, turn 135).

Another excluded group is 3rd person pl with impersonal sense (as in Tocaron a la puerta). The main problem is to decide when an example is really a type of impersonal-like construction whose subject is not overt. While in (7) it is possible to suppose a co-referent more or less diffuse (ella and mi papá), (8) does not seem to have any co-referent. (7) ella muy alegre/ muy alegre <…>/ era una persona que se dedicaba a lavar y a planchar para/

colaborar con el gasto de mi papá <~papá:>/ nosotros también pequeños también nos pusieron a trabajar muy pequeños (ME-283-23M-06, turn 90).

(8) ah sí <~sí:>/ me <~m:e> asaltaron dos veces ahí (ME-305-11H-07, turn 226). Other cases that fall outside are impersonal constructions (9), sentential subjects (10), idiomatic phrases (11) and inclusive 1st person pl (12). (9) [Ø] la cargas en tu mochila/ hay un operativo/ pues <~pus> [Ø] me dedico a la cháchara/ [Ø]

enseñas tu credencial (ME-312-12H-07, turn 103). (10) y un barrio se componía de/ pues <~pus> de puestos jacales/ y vivía uno en un jacal bueno un

un <~un:>/ una casa hecha de madera pues/ ¿no?/ y ya el que tenía una finca como ésta/ pues <~pus> ya tenía un/ por eso [Ø] eran colonias/ pero barrio/ barrio son/ Tepito Lagunilla y todas esas/ muy/ y el Distrito/ llegaba nomás hasta una calle que se llama/ que se llama este <~este:>/ ¿cómo se llama esta calle? / Balbuena <~balgüena> pues (ME-314-13H-07, turn 56).

(11) entonces <~entós> por ejemplo todo/ toda esa mm// eh/ ¡creación!/ de su biografía/ pues <~pus> es muy interesante/ ¿no?/ pero/ digamos/ e-/ el tener seguridad sobre esa <~esa:>/ ¡invención! pues/ pues no la hay/ ¿no? (ME-264-33M-05, turn 15).

(12) y no hay esa relación médico paciente/ en donde [Ø] te tienes que ganar al paciente y decirle “oiga [Ø]// pues <~pus> vamos a ver qué tiene”/ explicarles por lo menos/ ellos tienen todo el derecho de saber qué es lo que les está pasando (ME-252-31M-05, turn 21).

Thinking of the near future, it could be of interest to consider overall variable subjects (as in some early works), and not only SPPs. There would be at least two arguments that would support this point of view: i) the counterpart of null subject is many times not only an overt SPP, because it is in fact possible to produce a full (lexical or sentential) subject; ii) there are data where the alternative to null subject cannot be an overt SPP, but a lexical subject or another sort of pronoun (only in a very formal style, solutions such as ello would be adequate). In example (13), the textual alternation is produced between él and su esposa (i.e., between Pr and NP), but the slot of él could have been filled with zero or with mi hermano (i.e., Pr, Ø, and NP); notice even that the character of él as

! 6

contrastive focus does not prevent the possibility of Ø. Probably, the contrast would then be produced between trabaja and trabaja su esposa, and the ambiguity of [Ø] trabaja, given the proximity of mi mamá, would be solved immediately, with the presence of su esposa. Examples like that show it would be useful to consider the treatment of subject as a more general variable in future research. (13) ah pues por ejemplo/ está el ejemplo de mi hermano/ en que mi mamá cuida a los dos niños/ o

sea él trabaja y trabaja su esposa// sí ese es un ejemplo (risa) se puede decir (ME-307-11M-07, turn 242).

The alternation between a null subject and a non-SPP subject is not a minor problem

(314 in 3600 utterances, near 10%). Example (4, above) shows two cases, when the female speaker refers to her mother’s negative in indirect speech, and when she remembers her love to dance. There would be different alternative solutions: mi deseo, aquello, la danza, etc. We have not counted these cases because the SPP alternative, with ello, could only appear in a very formal style. In fact, this neuter personal pronoun never appears in our sample. Internal variables While designing our study, we took into account: i) the observations about variables in key studies; ii) the “Guía de codificación” by Bentivoglio, Ortiz & Silva-Corvalán (2011); iii) the “Coding manual” by Otheguy, Zentella & Heidrick (2012). After elaborating a series of previous models and proofs, the next reflections are our perspective about the question of how to establish the linguistic independent variables and about other linguistic questions considered apart (frequency of verbs, position of overt SPPs):

a) Person and number according to the verb has been a powerful (frequently the most important) factor group in many studies of SPP, including Otheguy, Zentella & Livert (2007) and Otheguy & Zentella (2012). Some contributions (e.g. Shin 2012) prefer to categorize the variable considering pronominal person and number. In alternative models of our data, the statistical difference between the verbal and pronominal point of view about person and number was inconsequential. We also processed the variable according to every individual pronoun, but the categorical character of uno (always overt) and the low figures of overt usted (4 cases) and ustedes (2) produced statistical instability. Finally, the analysis opted by verbal person and number for two reasons: a) ease in the comparisons; b) role of FLEX node as a rector of the concordance. b) Specificity, understanding the property in this way: "un argumento es específico cuando hace referencia a un ser, real o imaginario, que se considera identificable, al menos para el emisor. En caso contrario se dice que es inespecífico" (RAE-ASALE 2009, 15.9b, 1134). The factor group was non-significant in the preliminary overall statistical models. There seems to be an interaction with the grammatical person and number group and specificity was discarded in the process of refining the analysis. Some reflections about (non-) specificity in overt SPPs are included below, in the discussion of results. c) Verbal mood and syntactic domain. Verbal mood was reduced to two factors, indicative and non-indicative inflectional forms (non-inflectional forms are not considered in the study; see below, "other questions", for some comments). In a previous analysis, non-indicative mood has been divided in subjunctive and imperative, but imperative always appeared with null subjects. Our expectation was that the higher autonomy of indicative

! 7

propitiated overt SPPs. In that sense, we also included in the preliminary statistical models a factor group that considered syntactic domain: if the case was in a main, coordinate, or subordinate clause (cf. Otheguy & Zentella 2012). However, verbal mood and syntactic domain interact, due to the behavior of imperative and the common (but not exclusive) presence of subjunctive in subordinate domains. The more significant results for verbal mood led to the rejection of the syntactic domain. d) Verbal tense, progressive aspect, and verbal morphology. These three variables were conceived to characterize different grammatical dimensions of FV, and were included in our database as independent. Nevertheless, the relatively reduced number of progressive constructions (with respect to the second variable), and compound and periphrastic forms (with respect to the third one), besides the non-significant role of both groups in step by step regression analysis, led us not to include them in the final regression calculus. Regarding verbal tense, factors were collapsed into only four variants, equally after a number of log-p proofs: Present, Preterite, Co-preterite, and "others", due to the dispersion of data in the other verbal tenses. e) Verbal class. In this point, we follow the suggested classification of Bentivoglio, Ortiz & Silva-Corvalán (2011): mental processes verbs (27.3% of overt SPPs), dicendi verbs (23.6%), state verbs (22.4%), and activity verbs (18.0%). However, the factor group was excluded from the step by step log-p model as non-significant. The factor group was also non-significant in YS (Michnowicz, this volume). After that moment, we decided to adopt a different way, exploring the behavior of more frequent verbs, those 53 verbs with more than 10 occurrences in the initial collection; see discussion below and data in the Appendix (for the importance of frequency in the study of SPPs, cf. Erker & Guy 2012). f) Ambiguity. It was studied as morphological ambiguity (Silva-Corvalán 1982, 109; 2001, 154-169). Textual ambiguity is scarce in our data, while some examples appear (see below). The factor group has been commonly present in earlier studies about SPPs (Silva-Corvalán 1982, Bentivoglio 1987); beyond functional hypothesis by Hochberg (1986) in Puerto Rican data, a compensation due to morphological ambiguity seems to be active in conservative varieties, such as the Central Mexican dialect. Morphological ambiguity appears between 1st and 3rd sg forms in Co-preterite, Conditional, and subjunctive tenses (Bentivoglio, Ortiz, & Silva-Corvalán 2011), and also between usted/él and ustedes/ellos. g) Enunciative type. The variable is present in Bentivoglio, Ortiz, & Silva-Corvalán 2011), as “clausal type”: affirmative, negative, wh-interrogative, and absolute interrogative. In the first approaches we added negative-interrogative type to the factors, and considered them from an enunciative perspective. The relatively low number of the majority of types (except for affirmative) and the instability of the models led us to collapse the group into two members only: affirmative and non-affirmative utterances. h) Co-reference and turn of speaking as a domain. Co-reference as a variable has been studied from the first contributions about SPPs, showing a remarkable role in the discursive distribution of personal pronouns (e.g. Silva-Corvalán 1982, 104-105). The general principle rests on the idea that it will be more probable to mention the subject when it has had a change in the referent, i.e., in order to establish the new topic. The way to study co-reference can be more or less detailed, employing thorough measures of distance between co-referent items, or limiting considerations to continuity vs. change of subject. After exploring the different possibilities, we adopted an intermediate way, using three factors: without co-reference or with distant co-reference (the subject is new, or its co-

! 8

reference is before the previous main sentence); object (the close co-referent is not a subject); subject (the close co-referent is a subject).4 The hypothesis, of course, is that closeness and grammatical coincidence disfavor overt SPPs. This idea was complemented considering the turn of speaking as a domain (i.e., if the co-referent was in the same turn or in a previous one, implying change of turn), but this factor group was non-significant in the log-p calculus. i) Textual genre. The common topical discursive structure of some genres can favor a textual recurrence of the same subject in fragments of the interview (cf. Travis 2007, Martín Butragueño 2010). After considering at first more detailed characterizations, which produced instabilities in the model, we opted for a four factor group, taking into account a set of traditional genres: argumentation, narrative, description, and dialogue. j) Style. Considering the usual structure of a sociolinguistic interview, it seemed a question of interest to distinguish an initial section (more interview-like, with very defined roles and numerous dialogic pairs) vs. a more advanced section (more conversation-like, with more diffused roles in the relation interviewer/interviewee and longer fragments in many turns). In search of a simple method, we decided to divide samples by informant in two halves: first part style and second part style. Even though recognizing its unrefined nature, this factor group was significant, as can be seen below. k) Pronoun position. The position of subjects was already examined in early works, such as Silva-Corvalán (1982). The question becomes especially important after Cameron's prediction:

If dialect A has a lower relative rate of subject pronoun expression than dialect B, then we would find a higher relative rate of postposed subjects (i.e., subject-verb inversion) in dialect A than in dialect B. [...] If dialect B has a higher rate of subject pronoun expression than dialect A, then we would find a lower rate of postposed subjects (i.e., subject-verb inversion) in dialect B than in dialect A (1993, 321).

In that sense, we have considered overt SPPs position, and carried out a log-p analysis about the factor groups that constrain post-V location. Social variables The limited significance of social variables in the distribution of overt SPPs is a recurrent observation in Spanish linguistic literature (but see Orozco & Guy 2008, Shin & Otheguy 2013). While the CSCM has considered a great number of social, reticular and individual variables, in this contribution we only employ three basic social variables:

a) Educational level. Three levels were considered: i) low (0-6 schooling years); ii) medium (7-12 years); iii) high (more than 13 —usually 16 or more). There are two arguments for this distribution: first, it reflects social reality and corresponds to identifiable social groups in their way of life in many aspects (Lastra & Martín Butragueño 2000); second, it is the same segmentation employed in all PRESEEA corpora (2011) and it has showed its productivity in a vast number of linguistic studies (included many realized with CSCM data). Besides, educational level has been one of the most reliable variables in order to stratify society in the Hispanic world.

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!4 Non-application cases (e.g. at the beginning of the interview) were excluded.

! 9

b) Age. Again, our study distinguishes three variants: a) 20-34 years; b) 35-54; c) 55 and more. It is also the division employed in PRESEEA corpora and corresponds to social divisions: accommodation in society; individual fullness and social establishment; individual maturity and retirement. Of course, age is the key to study variation and change in apparent time.

c) Gender. The division between male and female is critical in any research on variation and change, due to differences regarding linguistic behavior, prestige and leadership. 3. Results Overall results The overall rate of overt SPPs in MC is one of the lowest percentages shown in Table 1 (21.7%, 443/2040).

Table 1. Rates of overt SPPs, in percentage % Mexican Spanish (Mexico City, this work) 21.7

WLA Mexican-Americans (Silva-Corvalán 1982, 102, 118)5 42.5 San Juan, Puerto Rico (Cameron 1993, 306) 44.8

Madrid, Spain (Cameron 1993, 306) 20.9 LA Mexican-Americans (Silva-Corvalán 1994, 153, Group 1, 118/360) 32.8

(Silva-Corvalán 1994, 153, Group 2, 163/586) 27.8 (Silva-Corvalán 1994, 153, Group 3, 198/758)6 26.1

Northern California Mexican immigrant and Chicano children (Bayley & Pease Álvarez 1996) 20.0 Mexican immigrants in New Jersey (Flores Ferrán 2007b) 24.0

Barranquilla, Colombia (Orozco & Guy 2008, 72) 35.7 Castañer, Puerto Rico (Homlquist 2012, 208) 28.0

San Antonio Mexican-background speakers (Bayley et al. 2012, 57) 27.0 NYC Latino Communities (Otheguy & Zentella 2012, 69) 34.0

NYC Mexicans (Otheguy & Zentella 2012, 72) 22.0 Mexico children from Querétaro and Oaxaca (Shin 2012, 134) 6.3

NYC Mexican newcomers (Shin & Otheguy 2013, 434) 18.7 Yucatan Spanish overall (Michnowicz, this volume) 19.7

Monolingual Yucatan Spanish (Michnowicz, this volume) 16.0 MC results are very similar to NYC Mexicans (Otheguy & Zentella 2012, 72), with 22.0% of overt SPPs, as well as NYC Mexican newcomers (6 speakers) (Otheguy & Zentella corpus, apud Shin 2012, 134; Shin & Otheguy 2013, 434), 18.7%, and Madrid data (Cameron 1993, 306), 20.9%. Figures in Bayley & Pease-Álvarez (20%) are very similar to ours, but Flores-Ferrán (2007b) and Bayley et al. (2012) show rates slightly higher or more clearly higher (24%, 27%) to those of MC. Overall percentages are slightly lesser in Yucatan Spanish (19.7%), but openly lesser among monolingual speakers (16%, vs. 23.5% !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!5 See the characteristics of this sample above. The percentage results from "338 expressed subjects out of a total of 795 tokens" (1982, 102), but it is not obvious if the author also considered non-pronominal subjects (103). Furthermore, the sample included a number of requisites about the proportion of expressed subjects and non-pronominal subjects (103). This set of features makes it difficult to establish a comparison with our data. 6 "Group 1 includes speakers born in Mexico, who immigrated to the USA after the age of eleven. [...] Group 2 encompasses speakers born in the USA or those who have immigrated from Mexico before the age of six. Group 3 also comprises speakers born in the USA; in addition, at least one parent responds to the definition of those in Group 2" (Silva-Corvalán 1994, 15-16).

! 10

among bilinguals, see Michnowicz, this volume). Our results are very different from Caribbean data: 44.8% of overt SPPs in San Juan (Cameron 1993, 306), 35.7% in Barranquilla (Orozco & Guy 2008, 72) (but see 28% in Castañer, Holquimst 2012, 208). Silva-Corvalán's analyses about Los Angeles have special characteristics, in methodological (see note 4) and descriptive terms, given the focus in language contact. Nevertheless, her 1994 analysis, with focal and contrastive pronouns excluded, reduce total percentages to 27% for Group 1, 21% for Group 2, and 22% for Group 3 (1994, 156), figures close to ours. Shin's study (2012) refers to first language acquisition. MC data clearly belong to Mainland varieties (in the sense of Otheguy & Zentella 2012). We prefer, however, to speak about conservative vs. innovative varieties (more neutral terms); MC is, therefore, a conservative variety as far as SPP expression is concerned. Internal Factors The statistical model for linguistic factor groups appears in table 2 (data processed with Goldvarb Lion, see Sankoff, Tagliamonte & Smith 2012, Tagliamonte 2006, 2012).

Table 2. Linguistic distribution of overt SPPs in MC, ordered by range Factors F/N =% P

a) Grammatical person and number 1st sg 3rd sg 2nd sg 3rd pl 1st pl

Range= 32

232/939 122/450 49/296 21/176 19/179

24.7 27.1 16.6 11.9 10.6

.58

.57

.40

.33

.26

b) Co-reference Change in reference

Object Subject

Range= 31

182/556 69/303

192/1181

32.7 22.8 16.3

.70

.54

.39

c) Verbal mood Indicative

Non-indicative Range= 29

433/1916 10/124

22.6 8.1

.52

.23

d) Enunciative type Affirmative

Non-affirmative Range= 23

410/1741 33/299

23.5 11.0

.54

.31

e) Textual genre Argumentation

Description Dialogue Narrative

Range= 23

105/366 70/257 76/449

192/968

28.7 27.2 16.9 19.8

.66

.55

.48

.43

f) Verbal Tense

! 11

Co-preterite Present

Preterite Other tenses

Range= 20

153/488 187/888 86/556 17/108

31.4 21.1 15.5 15.7

.60

.50

.44

.40

g) Ambiguity Ambiguous

Non-ambiguous Range= 13

127/352 316/1688

36.1 18.7

.61

.48

h) Style Second part

First part Range= 8

236/1016 207/1024

23.2 20.2

.54

.46

Log likelihood= -938.190, significance= .018, input= .181 Null SPPs are always the predominant solution under any circumstance (at least 2/3 of data, but sometimes occupy 90% of the distribution). Having this in mind, the significant factor groups establish which factors relatively favor the occurrence of overt SPPs.7 a) Grammatical person and number. This factor group is the most significant in MC. 1st and 3rd sg propitiate overt SPPs (p= .58, .57), virtually in the same proportion. After them, the three other possibilities disfavor overt SPPs. Clearly, singular favors overt solution more than plural. This aspect is recurrent in studies about SPPs: Cameron reports 50% for singular of overt SPPs in San Juan and 26% in Madrid, but 19% and 7% for the plural number, respectively (1993, 306), and points to a convincing explanation: "If we conceive of plural subjects as sets, we find that discourse is typically structured so that the great majority of plural subjects occur in contexts where their set members are either explicitly or inferably present within the immediately preceding discourse. Such contexts favor null subject expression. Therefore, plural subjects are frequently null overall" (1993, 328, n. 2). In YS, singular also favors overt SPPs more than plural (19.5% vs. 10%, Michnowicz, this volume).

Results about overt SPPs are similar enough to the production by Mexican newcomers in New York City (Otheguy-Zentella corpus, apud Shin 2012, 135): 1st sg, 24.7% in MC, 25.8% in NYC; 3rd sg (27.1% vs. 22.9%); 2nd sg (16.6% vs. 9.8%); 3rd pl (11.9% vs. 7.6%); 1st pl (10.6% vs. 4.3%). The main coincidence appears between the more favoring forms to the overt solution (first and third singular persons); the low level of other persons in NYC is remarkable, especially in the plural. MC data are relatively more balanced, even though the general pattern is coincident. In Querétaro, Quesada & Blackwell (2009, 128) report 35.% (in 289 contexts) for 1st sg, and Blackwell & Quesada (2012, 157) mention 11.6% for 3rd person (only in 69 contexts). In Los Angeles, Silva-Corvalán offers for type of subject 42%-28%-34% (Groups 1-2-3) for 1st sg; 28%-25%-

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!7 There are some similarities between resulting hierarchy of linguistic groups in MC and linguistic hierarchy in NYC immigrant newcomers (Otheguy & Zentella 2012, 160). In both models person and reference are the first and second significant factor groups, tense appears among significant variables, and overall specificity (here) and definiteness (Otheguy & Zentella, but see n. 11 in p. 159 about the construction of the variable) have a limited role. In Yucatan Spanish, the linguistic constraint hierarchy is person/number, TMA, co-reference, reflexivity (tables 4, 5 and 6, in Michnowicz, this volume).

! 12

10% for 2nd sg, 24%-31%-24% for 3rd sg; 15%-15%-6% for 1st pl; 25%-16%-13% for 3rd pl; and 59%-51%-44% for other pronouns (1994, 163). LA data are not coincident with our results —more similar to NYC or Madrid—, especially if we consider LA Group 1. In monolingual YS, rates are similar in 1st sg (21%), 3rd sg (20%) and 3rd pl (10%). The pronoun uno is not really a variable form, at least in our data, as in (14); una would be rarely expected in Mexican Spanish materials. A vast majority of uno is clearly non-specific (at least 23/27, 85.1%), or supports a non-specific reading. Among variable items, usted shows the highest individual frequency (26.7%), maybe due to its potential marking of social relations (Serrano 2012); example (15) shows the presence of usted (the interviewer), uno (the female speaker as a member of a group) and yo (the speaker as an individual). The second highest frequency corresponds to the pronoun yo. It is clear that overt yo appears more times that could be expected, probably due to the fact that the informant tends to mark those situations which affect him (as in 16); see Travis & Torres Cacoullos (2012) for cognitive and mechanical effects. Regarding él (or ella), it is likely that the favoring weight is associated with the nature of the recorded interviews, generally with only one interviewer and one informant; some cases (as 17) specify the referent even though it would not be strictly necessary to do so. In other cases, the alternation among él/Ø (and even a lexical subject) is an extraordinary material to build a vivid narrative as in (18), where the speaker combines narrative fragments with reported speech in her story.

While tú does not favor the presence of the subject, it shows some interesting features. It could be thought that the low presence is linked again to the dual nature of the interview, and it is probably that this is partly true. However, the specific/non-specific values deserve comment. Specificity was non-significant in the overall model, but it did in a more restricted model, with only two groups: grammatical person and number, and specificity (log likelihood= -1037.612, significance= .000, input= .207). In that case, non-specific overt SPPs received p= .75, vs. p= .47 in specific cases (non-specific= 383/1805, 21.2%; specific= 63/239, 26.4%). The absence of specificity in the overall model has to do with an interaction with grammatical person and number. Specific overt tú reaches 14% (14/103), and non-specific tú rises to 18.1% (35/193); in monolingual YS, rates are divergent: 32% for definite, and 8% for indefinite 2nd sg (Michnowicz, this volume). If there were 63 cases of non-specific overt SPPs in MC, 35 of them (55.6%) corresponded to 2nd sg person. In fact, 65.2% (193/296) of overall tú (overt or null) was non-specific (vs. a rate of 11.7%, 239/2044, of non-specifity in the overall sample). In contrast with 16.6% for overall tú in MC data, Cameron (1993, 325) shows 60% in Puerto Rico and 25% in Madrid. MC figures are similar to Madrid with non-specific tú (19% in Madrid and 69% in San Juan vs. 18% in MC), but MC is very different from Madrid and San Juan in the specific tú rate: only 14% in MC vs. 40% in Madrid and 48% in San Juan (Cameron 1996, 89). If we compare probabilities, however, our data turn out to be closer to Puerto Rico than to Madrid: Caribbean data show .72 for non-specific tú and .86 for non-specific uno (respectively .50 and .72 in Madrid), vs. .75 of overall non-specificity in MC; regarding specific tú, Puerto Rico offers .51 (.72 in Madrid), vs. .47 of overall specificity in MC. These facts will be reconsidered below, in the light of CRH.

In (19), an example is presented in which the speaker is addressing a specific second person, but eventually he changes to a non-specific tú (Tú te encontrabas todavía canales); probably this change provokes the explicit presence of the pronominal form in that moment (notice that it is not absolutely necessary). A second skew, already mentioned, is provoked

! 13

by uno, whose readings are or clearly non-specific or point in that direction; a very few set of non-specific data are linked to other forms. (14) ¿por qué?/ porque realmente pues <~pus> yo sé que <~que:>/ teniendo novio siempre uno se

enamora y/ ¡siempre termina uno juntada! ¡o casada! o equis/ ¿no? (ME-307-11M-07, turn 158).

(15) entonces pues <~pus> ya [Ø] me dijo “no/ pues <~pus> vente [Ø]/ que [Ø] vamos a ver unos muchachos/ que quién sabe qué que sabe cuándo”/ y pues yo como ¡cordero!/ usted sabe que/ uno como ni estudió ni nada/ pues yo como el cordero me fui// entonces esos muchachos con los que anduve pues <~pus>/ me violaron// y pues <~pus> ya [Ø] me quedé embarazada (ME-313-13M-07, turn 13).

(16) cerca de donde/ de donde yo viví/ de donde yo nací/ yo nací en/ en la delegación de Tláhuac/ en la colonia Nopalera (ME-259-32M-05, turn 5).

(17) hicieron/ le construyeron al maestro un salón/ [que por supuesto] fue un súper/ apoyo/ y y bueno/ a na-/ prácticamente a nadie le daban ese tipo de concesiones/ [pero él ganaba siempre] los distritales/ entonces <~entós> como [Ø] ganaba los distritales con su grupo de danza/ porque [Ø] era muy bueno/ pues le dieron el apoyo/ él duró/ treinta años dando clases en la Secretaría de Educación Pública (ME-259-32M-05, turn 61).

(18) sí/ “[Ø] me va a hacer feliz”/ y ella comoquiera me/ me/ pues <~pus> me <~me:>/ cómo se dijera me <~me:>/ me ¿cómo se dijera?/ [Ø] platicó con el muchacho/ que era mi es-/ que fue mi esposo/ [Ø] dijo “no” [Ø] dice “pues <~pus> si [Ø] no te obede-/ si [Ø] no te obedece/ pues <~pus> ahí <~ai> [Ø] le das hasta que/ [Ø] entienda”/ dijo la señora que en paz descanse/ a la mejor [Ø] ya se murió/ “ahí <~ai> [Ø] le das” [Ø] dice este/ “[Ø] no te dejes/ si” [Ø] dice “y si [Ø] no te obedece [Ø] pégale/ [Ø] hazla/ [Ø] hazle lo que [Ø] puedas” [Ø] dice/ y pues <~pus> sí/ yo creo que sí él obedeció/ pues yo ahí <~ai> estaba ya/ [Ø] me golpeaba <~golpiaba>/ [Ø] me trataba mal/ entonces [Ø] me embaracé de <~de::>/ de un niño/ que ahora <~ora> es el/ mi hijo <~mijo> eh mi hijo <~mijo> mayor (ME-313-13M-07, turn 23).

(19) [Ø] acuérdate que en las litografías eh/ del pasado// ya no mucha distancia/ hace cien años/ por decirte algo/ que suena muy lejano/ pero que históricamente/ es un segundo// hace cien años/ tú te encontrabas todavía/ canales// con/ piraguas de verduras en// en Jamaica (ME-294-33H-07, turn 39).

b) Co-reference. The second significant factor group is the kind of co-reference (range= 31, very close to grammatical person and number). Factors involved are ‘change in reference’, a label that includes absence of previous reference, but also those cases with a reference at medium or long distance, or with only a contextual reference (p= .70, 182/556, 32.7% of overt SPPs); data with co-reference at a short distance (in the domain of the previous main sentence) are divided into two factors: subject (p= .39, 192/1181, 16.3%), and object (any other function developed by the co-referent element) (p= .54, 69/303, 22.8%). Results, as many other studies have shown (e.g. Silva-Corvalán 1982, 104; 1994, 157-158), coincide with a general hypothesis: the absence of co-reference and non-short distance of previous reference favor —in terms of probabilistic weights and frequencies— overt SPPs; when the nearby co-reference is established with an object, presence is still favored, but only slightly; finally, if the co-referent is a subject (i.e., same referent, same function), presence is disfavored. Probabilistic weights are very similar to those obtained in other studies. For instance, in WLA Silva-Corvalán (1982, 104) shows 53% of expressed subjects with switch reference, and 25% with same reference (p= .34 for switch reference and .66 for same reference in the case of absence of the subject, p. 113). By Groups of speakers (pronouns and NPs included), Group 1 shows 51% of switch and 25% of same

! 14

reference; Group 2, 58% and 17%; and Group 3, 54% and 20%. By functions, LA speakers offer 57% of switch reference with all arguments, and 20% of co-reference with subjects (all types of subjects included, Silva-Corvalán 1994, 157; in MC, only for Ø/SPP subjects, there are 22.8% of same reference with objects, and 16.3% with subjects). Cameron (1993, 315) discovers differences in frequency but sameness in probability between San Juan (switch reference, 66%, .65; same reference, 35%, .35) and Madrid (switch, 38%, .66; same, 14%, .35). Our data are compatible in terms of probability with San Juan and Madrid, but very similar to Madrid in frequency results. Orozco & Guy find .61 to complete change in reference (44% of overt SPPs), .50 to partial change (co-reference with object, 34%) and .35 to continuity (co-reference with subject, 23%) (2008, 74) —again, MC data are relatively similar in weights, but more distant in frequencies. In YS, switch reference (understood as complete switch and co-object) obtained 22.8%, and same reference (co-subject), 16.7% (Michnowicz, this volume), relatively similar to MC data (29.0% and 16.3%, grouped in a similar way). In monolingual YS, rates are 21% for complete switch, 11% for no switch, and 10% for co-object. A complementary question has to do with the overwhelming number of co-referent subjects in our (and probably in any) sample: only the subjects as co-referents constitute 57.8% of the cases, as it is to be expected due to the role of subject in topical discursive continuity. Under those circumstances, the low level of subject as a co-referent in overt SPPs is reinforced: as the non-marked solution, the greater the role of subjects (i.e., the stronger the topical cohesion) in the line of discourse, its overt expression is less needed. Discursive interpretation, in any case, is conditioned by co-reference, but sometimes also by textual ambiguity. Even though the majority of cases can be easily interpreted, there are a number where it is necessary to employ all tools the interlocutor can put into practice, including general knowledge of the situation. So that, in example (20) it is crucial to know that the speaker is referring to a female writer’s life, especially when she begins to introduce the role of the writer’s mother. (20) entonces/ ellai es de una// familia/ en donde el matriarcado se hace efectivo// su madrej una/

pues <~pus> muj-/ una mujerj/// digamos eh// no común en su… […] para su época// eh mm/ hijaj de un <~un:>/ señor que/ pues <~pus> laj quiere complacer y <~y:> [Ø]j está segura de que/ de que suj hijai/ pues está bien que [Ø]i haga lo que le gusta/ ¿no?// ¡y <~y::>!/ ellaj tiene varios hijos/ [Ø] no me acuerdo si [Ø] son siete pero/ de distinto padre todos// y ellaj ¡los registra!/ como hijos suyosj/ naturales/ con toda la libertad/ con toda la/ tranquilidad// y <~y::>/ ¡pues eso implica! tener ya un <~un:>/ un modelo/ especial/ ¿no?/ parece que [Ø]i tiene ¡muy buena relación! con ellaj// aunque <~aunque:>/ se percibe por sui poesía/ y por lo que ellai dice de ellaj// que [Ø]i no tuvo una// relación ¡física! cercana (ME-264-33M-05, turns 1-3).

If (20) is read without external knowledge, several referents could be ambiguous, particularly from ella tiene varios hijos to the end of the example. From only a syntactic point of view, it is not possible to establish who the mother of varios hijos is, or who is the referent in lo que ella dice de ella (consider an interpretation where the writer is the mother). Discursive resources can help partially in the task, nevertheless. In one interpretation, the co-referent of ella in ella tiene varios hijos is Ø in está bien que Ø haga lo que le gusta, based on the argument of short distance between first and second subject; however, we do not expect (unless a focal construction would have appeared) the presence of personal pronoun in the second occurrence, precisely because of the short distance

! 15

between them. After hearing the passage of the interview, it is evident that there is not a prosodic peak. So, the alternative interpretation would be that ella is the same person initially introduced as su madre. But this, of course, is only an interpretation suggested afterwards by the assertion that ¡pues eso implica! tener ya un <~un:>/ un modelo/ especial (under the belief that one can expect a mother to be a model for a child, more than the other way around).

It should not be believed that co-reference is a difficult task with only obscure results. In fact, there are many cases easily interpreted in spite of their complexity. (21) is an example where a medium age and low educational level female speaker introduces five referents in a brief fragment: mi hermana la más chica, yo, mi mamá, mi sobrina, and a non-specific referent (the subject in nunca me han enseñado la tal carta and the IO in no quiero echarles bronca). It may be interesting to signal that even though this last reference is ambiguous, syntactically speaking, to say the least, it does not represent any problem from a discursive perspective, because it is obvious that the informant is speaking about her family, and more or less specifically about her sister and her niece. Notice that she employs third person plural in negative comments about her family, apparently in order to avoid concrete references.

(21) porque mi hermana la más chica/ este/ según bueno [Ø] no sé ¿no?/ porque/ yo no sé de carta

que haya dejado mi mamá/ hasta ahorita <~orita> no/ mi mamá ya va a cumplir más <de> dos años/ que falleció/ y nunca [Ø] me han enseñado la tal carta/ que dejó mi mamá/ que según [Ø] dejó una carta que luego apa-/ fueron tres cartas [Ø] no sé/ según el departamento [Ø] se lo dejó a mi sobrina/ a la hija de mi hermana la más chica/ ¡según!/ mh/ y yo le ayudé a pagar a mi mamá letras de o sea/ rentas de la casa predios luz todo/ [Ø] no sé la verdad/ y yo como no quiero echarles bronca de que/ “ay que tú ¿por qué?”/ ay mejor [Ø] me quedo así mira mejor [Ø] rento/ y [Ø] ni me busco problemas (ME-308-12M-07, turn 185).

c) Verbal mood. The imperative mood had a categorical behavior in our data, and never presented overt SPPs. In fact, according to the results obtained, the indicative mood has no effective role in the model (p= .52); so, the main assertion is that subjunctive does not favor overt SPPs (p= .23 in non-indicative mood). This is consistent with the usual subordinate function of subjunctive and, in consequence, the likely short distance co-reference with previous elements. It is also a fact that a vast majority of cases is in indicative mood (1916, vs. only 124 in a non-indicative mood), so that it is likely that indicative forms predominate in all the structures. That situation permits to prove that the structural effect does not explain all the differences among moods. If we consider only indicative and subjunctive moods, indicative forms present a virtually equal number of overt SPPs in main clauses (24%, 220/917) than in subordinate clauses (25%, 111/452); in coordinate clauses, however, the rate decreases to 19% (105/551). Overt SPPs with subjunctive forms are always lesser in all types of clause: 10% in main clauses (1/10), 13% in subordinate clauses (8/62), and 8% in coordinate clauses (1/12). Even though coordination showed the lowest rate with both moods, syntactic domain was not significant as a factor group in the overall model. Nevertheless, a very restricted model with only mood and domain shows significance with both of them (log likelihood= -1060.107, significance= .042, input= .214). In this model, indicative and non-indicative have the same probabilistic weights than in the overall calculus. Regarding domain, subordinate and main clauses slightly favor overt SPPs (p= .53 and .51, respectively), and coordinate clauses

! 16

disfavor also slightly (p= .45), in consonance with frequencies. Maybe the corollary with respect to the relation between mood and domain is that subjunctive mood always disfavors overt SPPs, but indicative mood disfavors overt SPPs specifically in coordinate clauses (cf. Otheguy & Zentella 2012, 164, with main clause as top constraint, and coordinate clause as bottom constraint). d) Enunciative type. The enunciative type is the fourth independent variable selected by the regression calculus. A clear difference between affirmative forms and, on the other hand, negative and interrogative forms, seems to exist. As probabilistic weight reveals (.54, 23.5%, 410/1741), it is not that the affirmative type favors the presence of the subject particularly; rather, what is happening is that non-affirmative cases do not favor overt SPPs (.31, 11%, 33/299). While (22) shows a certain accumulation of overt subjects, it does not mean that this is the most common situation. By contrast, negative examples as (23), even with double negation, are infrequent, and it is really more frequent to find cases such as (24), where we can see a series of negative statements; this serial fact could explain the low presence of subjects with negatives, due to the effect of nearby reference. Regarding interrogative type, we have re-examined each case one by one and the scarcity of real questions (as 25) is worth noticing; most of them are similar to tags (¿Qué te puedo decir?, ¿Cómo te dire?, ¿Qué te dire?, ¿Cómo ves?, ¿Qué crees?, etc.), and many of the others are confirmative (26) and phatic questions (¿Sí me entiendes?).8 (22) yo creo mi mamá ya no me quería tener ahí no sé qué pero a <~a:> la/ mi hermana mi m- la

mayor/ este [Ø] le pidió que [Ø] me llevara a Pachuca a trabajar/ y yo pues ahí yo <~yo::> eh la persona con la que [Ø] trabajaba/ este no me daba <~daba:> bien de comer/ [Ø] no me daba este/ jabón para ¡bañarme!/ yo estaba muy sucia y <~y:>/ pues <~pus>/ ella salía/ [Ø] no sé dónde/ porque [Ø] casi nunca estaba ahí/ yo me quedaba sola y/ pues yo sufría mucho (ME-313-13M-07, turn 13).

(23) y pues <~pus> yo la verdad/ yo tampoco no me aguanté o sea/ [Ø] estaba/ como tú ahorita (ME-252-31M-05, turn 7)

(24) [Ø] empecé a ter-/ bueno/ [Ø] empecé otra vez la carrera// pues sí/ sí [Ø] la terminé/ ya/ ahí en la Sep/ [Ø] ya no pude aguantar o sea/ ya era así de que/ [Ø] ya no quería ir// [Ø] ya no quería/ o sea nomás [Ø] me acord-/ “¡chin!/ [Ø] ya me tengo que ir a trabajar”// no/ [Ø] no quería ir/ a veces pues <~pus> (risa) [Ø] ya ni iba/ ¿no?/ mejor [Ø] me iba a mi casa (ME-257-32H-05, turn 40).

(25) ¿nunca has ido [Ø] a Mamá Rumba? (ME-265-21M-06, turn 117). (26) ¿hasta qué año estudié [Ø]?/ hasta primero de primaria (ME-307-11M-07, turn 34). e) Textual genre. Textual variable appears fifth in the model. Argumentation (including explanation) is the highest factor (p= .66, 28.7%, 105/366) for overt SPPs, because of the necessity of fixing points of view and underlining one's own and others' opinion (as in 27). Regarding description (p= .55, 27.2%, 70/257), it favors slightly overt SPPs. Dialogic and narrative genres do not favor overt solutions (p= .48, 16.9%, 76/449 and .43, 19.8%, 192/968). It is likely that a common thread is running through textual genre, verbal tense, and verbal function. As Silva-Corvalán (1997) has observed, a correspondence exists between these two last aspects; see also Travis (2007). Reported speech favors overt subjects, due to the discursive requirement of making clear who the

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!8 Some of the real questions are embedded in fragments of reported speech, and such contexts seem to disfavor null SPPs.

! 17

speaker is; (28) shows a dynamic narrative where the informant talks to two different persons and reports the dialogue between them and herself, putting the resources provided by presence and absence of pronouns into practice, and making evident the construction of text by the informant. By contrast, echo-interrogatives in dialogue (as in 29) does not especially favor explicit subjects, if we consider the nature of interviews we have analyzed, where usually it is not necessary to establish the referents of tú and yo. (27) mucho muy distinto a l- <~l::>/ los tiempos de ahora// yo pienso que mi vida [Ø] he vivido

demasiado/ demasiado yo <~yo:>/ doy gracias a la vida <~vida::>/ por darme esa oportunidad de vivir/ fracasos <~fraca:sos> [sufrimientos <~sufrimie:ntos>]/ desventuras <~desventu:ras>/ todo (ME-283-23M-06, turn 116).

(28) entonces <~tonces>/ [Ø] le dije a la muchacha “ay/ ¿qué crees [Ø]? que/ pues yo me quiero ir de aquí/ pero [Ø] no tengo a nadie” entonces <~entóns> [Ø] dice “yo conozco a una señora que es partera/ si [Ø] gustas/ [Ø] vete con ella” [Ø] le digo “sí”/ yo le dije a la señora/ “[Ø] me voy a ir” [Ø] dice “pero ¿por qué?”/ [Ø] le digo “porque pues <~pues> es que yo/ yo ya no quiero estar”/ “pero si tú no tienes familia”/ “no pero <~pero:> este/ [Ø] voy a ver si/ [Ø] me voy a mi fa-/ con mi familia” y [Ø] dice “pues <~pus> no/ [Ø] no te vayas” [Ø] dice “aquí nosotros te pagamos tu/ tu cama/ y el bebé que [Ø] tengas/ pues <~pus> si [Ø] quieres [Ø] nos lo regalas”/ y [Ø] le digo “no/ es que/ este/ mi compañerita este/ conoce a mi familia y yo me voy a ir” (ME-313-13M-07, turn 19).

(29) ¿qué hacía [Ø]?/ limpieza de oficinas (ME-307-11M-07, turn 14). f) Verbal tense. Verbal tense is also an important independent variable. The most favoring factor is the Co-preterite (p= .60, 31.4%, 153/488); Present is indifferent (p= .50, 21.1%, 187/888); and Preterite does not favor over SPPs (p= .44, 15.5%, 86/556). The rest of verbal tenses have been grouped as “other tenses” and are disfavoring (.40, 15.7%, 17/108). Such a distribution is in the line of Silva-Corvalán’s observations about distribution of tenses among Mexican-Americans at Los Angeles (1997, 126; 2001, 161) for types A, B, and C verbs. As it is well known, the distinction proceeds from Hochberg (1986), but it is readapted by Silva-Corvalán, observing the relation between verbal tenses and their functional values: A corresponds to Preterite (factual, assertive, dynamic, focus in the event, foregrounded); B is essentially the Present (factual and assertive, but not always dynamic and focal; Future is scarce in American Spanish, as Lastra & Martín Butragueño 2010 confirms for MC); C includes Co-preterite, Conditional, and subjunctive tenses (backgrounded, irrealis, non-dynamic, non-assertive, polite). Future, Conditional, and subjunctive tenses remain as “other tenses” in our data (N= 108); though subjunctive mood is perfectly alive in MC (Lastra y Martín Butragueño 2012), analysis of verbal mood (see above) has already revealed the disfavoring character of non-indicative forms for overt SPPs. Silva-Corvalán offers 36% for C, 33% for B, and 27% for A. Our data show 31% for Co-preterite, 21% for Present, and 16% for Preterite; i.e., the same order of factors (in a grosso modo comparison), but at a lower absolute rate.

On the other hand, more than half of cases of Co-preterite are morphologically ambiguous (58.0%, 283/488), but if only ambiguous Co-preterite with overt SPPs is considered, the proportion rises to three fourths of the sample (72.5%, 111/153). This difference is remarkable, so that it could suggest a partial functional effect based on the

! 18

ambiguity constraint. Additionally, it has to be said that some examples show an accumulation of Co-preterite with overt SPP (as in 30).9 (30) porque yo agarraba y/ yo veía a mis compañeros que en ocasiones ellos agarraban y [Ø] se

subían a los trenes/ y yo también lo <~lo:> trataba de hacer/ al principio [Ø] me dio mucho trabajo/ pero y después [Ø] lo dominaba a la perfección (ME-283-23M-06, turn 76).

g) Ambiguity. The next factor in the regression model is morphological ambiguity (between 1st and 3rd sg persons in Co-preterite, Post-preterite and subjunctive tenses). Even though the level of textual ambiguity is extremely low, the factor group is significant, and morphological ambiguity favors overt SPPs (p= .61, 36.1%, 127/352). MC rates are very similar to “less distinctive” (36%) and “more distinctive” (17%) verb tenses/moods in overall YS, and 31% and 14% in monolingual YS (Michnowicz, this volume). It is likely that low discursive ambiguity is caused precisely by the presence of overt SPPs with ambiguous morphological verbal forms. See (31), in which podía could have a co-referent as a subject yo (the speaker) or ella (the speaker’s mother). Probably, the unmarked interpretation would be ella due to the short distance and the absence of a contrastive focus (yo), but certainly the ambiguity is not resolved.10 (31) mi este/ mi mamá trabajaba en el correo// y ella empezó a <~a:>/// a ver si me [Ø] podía meter/

¿no?/ pues <~pus> [Ø] ya empecé a ir al sindicato/ a pasar listas/ a que/ me conocieran/ casi un año/ de estar yendo// y mientras pues <~pus> a tu casa a hacer el aseo y todo/ ¿no?// hasta que ya/ [Ø] entré a trabajar/// [Ø] voy a cumplir/ en este/ primero de julio/ treinta años de servicio/ trabajando (ME-274-22M-06, turn 43).

In WLA data (Silva-Corvalán 1982, 109), ambiguous verbs were linked to 69% of expressed subjects (vs. 39% with unambiguous verbs in context, and 38% with unambiguous verbs; p= .28 for ambiguous V forms, .68 for unambiguous V, .55 for context unambiguous in the case of absence, p. 113). In Silva-Corvalán (1994, 156), Group 1 offers 59% of subject expression and 45% of overt SPP with ambiguous verbs (vs. 34% and 21% with non-ambiguous forms); Group 2 shows 64% of subject expression and 33% of overt SPP with ambiguity (vs. 37% and 19% with non-ambiguity); Group 3 presents 37% of subject expression and 21% of overt SPP in ambiguous cases (vs. 40% and 22%) —the differences are significant for Groups 1 and 2, but not for Group 3. Our results, however, show probabilistic weights very similar to those offered by Blanco (1999) in her study of 1st person SPPs at Alcalá de Henares (close to Madrid): .64 for ambiguous and .49 for non-ambiguous (vs. .61 and .48 in MC). Cameron argues in favor of seeing the effects of morphological ambiguity "not as independent constraints [...] but as effects that interact with switch reference, the central constraint" (1993, 307). Our data reveal this same tendency. In a cross-tabulation of overall data, ambiguous Vs with overt SPPs are more frequent with change in reference: 54% vs. 43% (co-reference with an object) and 29% (with a subject). Nevertheless, both effects are at least partially independent.11 The same logic can be applied to the relation between

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!9 By the way, this is one of the few cases where a relative subject and a pronominal subject coexist. 10 Some Mexican speakers point to the fact that the culturally expected co-reference is ella. 11 Our arguments are four: a) in a log-model with only these two factor groups, both are significant (log likelihood= -1008.326, sign= .000, input= .294); b) regarding the fit, χ2 (3 d.f.)= .199, accepted, p = .977; c)

! 19

ambiguity and verbal tenses. In overall data, overt SPPs linked to morphological ambiguity are concentrated in the Co-preterite (39%, 111/283); percentages of overt/ambiguous in other verbal tenses rest in few data (8/37 for Present, 2/12 for Preterite, 6/20 for other tenses). A link among Co-preterite, ambiguity and switch/same reference with overt SPPs is not evident. Truly Co-preterites that show change of reference have 45% of overt SPPs > Present (31%) > Preterite (26%) > Other tenses (19%), but the same first more frequent place always appears: Co-Preterite (45%) > Preterite (19%) > Present (18%) > Other (1%) for co-reference with objects, and Co-Preterite (23%) > Other (17%) > Present (16%) > Preterite (11%) for co-reference with subjects. h) Frequency of verbs. There are 53 verbs that occur ten or more times in the initial collection (of 3600 utterances); the specific number of occurrences is showed in the second column of Appendix (e.g., sentir and mirar appear 10 times, vs. 605 cases of ser). The third column presents the number of occurrences that make part of the sample (i.e., of the variable [Ø]/SPP; so that, we have 0 cases —0.0%— of haber vs. 134 —72.8%— of decir and 163 —76.9%— of tener). Finally, fifth and sixth columns show number and percentage of overt SPPs in the variable contexts, as 26 cases of creer (54.2%) or 10 of hacer (14.7%). The list is divided in two halves: the verbs above and below the rate of overt SPPs in the overall sample (21.7%). So that, 22 verbs stay above this threshold, and 31 below it. The 22 upper verbs occur with 211 overt SPPs (47.6%), which confirms their frequent character. Does every verb have its own history, or is it possible to establish a regular variation? While there were 115 examples with haber in the first collection, none of them was eligible for the sample; the immense majority was impersonal, and a very reduced number showed lexical or relative subject. The verb decir shows a slightly low rate; many of the examples introduced reported speech, while some of them were evidential-like (digo yo). It is likely that the relatively low rate is linked to the serial occurrence of decir in some sequences of reported speech, with no problem in establishing the referent. On the other hand, tener shows a frequency (36/163, 22.1%) close to the threshold. Interestingly, it refers to the possession of objects only a few times; it rather refers to age (tenía, tendría X años) or other meanings, like ‘to give birth’ in Ella cada año tenía un bebé. Regarding gustar, only 2 cases in 60 were eligible (A como tú gustes, Si gustas vete con ella), and the rest were data with more canonical uses, that had lexical or sentential subjects (its 50% of overt SPPs is misleading). In the case of llegar, we have a rate of 28.6% (10/35); near all the cases are locative, but there are some examples with metonymical meaning, as in Yo he llegado a la conclusión. The verb estar rises to 32.9% (26/79), but usually without a locative sense; rather, we find subjective situations in descriptions in co-preterite, such as De repente yo estaba solito. In a higher position, the rate of vivir is 34.0% (18/53); sometimes, its meaning refers to a locative, concrete sense, but many other times to an

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!this dual log likelihood is higher than in independent models (log likelihood= -1044.021, sign= .000, input= .265 for ambiguity; log likelihood= -1038.371, sign= .000, input= .233 for co-reference); d) R2 is very low= .087. In fact, in a model of our data with only switch/same reference relationship between both groups only in overt SPPs, the comparison is significant (χ2= 8.917, p= .003; correction of continuity= 8.289, p= .004; Fisher, p= .003), with 29.9% of switch reference and 70.1% of same reference for ambiguous cases, and 45.4% and 54.6%, respectively, for non-ambiguous cases, for overt SPPs. However, this is not all the story. A cross-tabulation with the three factors in our co-reference group is also significant (χ2= 9.958, p= .007), but the order of frequencies in ambiguous cases is subject (54.3%) > change of reference (29.9%) > object (15.7%), and in non-ambiguous cases is change (45.4%) > subject (39.3%) > object (15.3%), with some non-expected distributions. In other words, an independent behavior between both factor groups seems to exist.

! 20

abstract state, as in Yo vivía con mucho miedo. We have also the remarkable case of nacer, with a rate of 46.2% (6/13); likely, this high pronominal presence has to do with the position in the beginning of the text and the testimonial character of the verb: Yo nací, Nací yo. The verb with the highest rate of overt SPPs is creer (54.2%), what seems plausible if we consider the importance of subject in the acceptance of an argument. Maybe the most important partial conclusion is to observe the degree of differences among individual verbs, from 0% with haber (and others) to 54.2% in the case of creer. Once again, a syntactic base for the expression and type of subjects seems to be constrained by contextual meanings and discursive necessities. Nevertheless, the 22 more frequent verbs add 1603 examples to the initial collection (i.e., 44.5%), and only 815 to the final sample (40.0%), but their input to overt SPPs rate is slightly higher (47.6%, 211/443), which suggests that frequency supports a certain degree of variation, maybe enhancing the existing trends (Erker & Guy 2012). i) Style. The variable “style” was significant (in eighth position in the global ranking), and the second-part style favors overt SPPS (p= .54, 23.2%, 236/1016) more than the first-part style (p= .46, 20.2%, 207/1024). Only a few studies have considered style, but in very different senses. Silva-Corvalán (1982, 118) compared written Spanish and Spoken Mexican American Spanish and the percentages of expressed subjects were very similar (45% and 43%, respectively). Blanco (1999) encountered .56 in formal register vs. .44 in informal register; due to her study being limited to the 1st person, it is possible that formal register has included a higher proportion of argumentative-like style. Shin (2012, 136) shows significant differences of overt SPPs expression by speech context: 3.3% in picture book, 3.9% in story from memory, and 11.5% in sociolinguistic interviews. In our case, first-part style includes a number of introductory turns, established in a more ritual form, so that they can have propitiated a lesser proportion of overt SPPs, vs. second-part style, with a higher quantity of long turns by the interviewee, as well as a higher proportion of argumentation and reported speech. j) Pronoun position. Overt SPPs occur, for the most part, in pre-verbal position 85.6% (379/443). In general, it is accepted that post-verbal position for subject is more marked than pre-V in Spanish, and quantities suggest that our data are not the exception. Only 64 examples appear after the verb (14.4%, 64/443). Silva-Corvalán finds 43% of expressed post-V subjects and 57% of expressed pre-V subjects (1982, 118). However, the percentages seem to make reference to non-pronominal subjects also, and basically in narratives, so that a detailed comparison is not possible. In contrast, data about Yucatan Spanish (88% of pre-V, and 12% of post-V, Michnowicz, this volume) are very similar to that of MC. Crucially, Cameron's prediction (1993, 321, see above) about the inverse relation between subject pronoun expression and postposed subjects is in contradiction with our results. He offers as a proof the differences between Puerto Rico data (81% of pre-V subjects and 19% of post-V subjects, following Morales 1982), and WLA data (65% of pre-V subjects and 35% of post-V subjects, following Silva-Corvalán 1977). We emphasize words subject pronoun and subject because the prediction referred to overt SPPs (at least in its first part), and not to overall subjects. That is very important, because our data show 85.6% of pre-V overt SPPs, and 14.4% of post-V overt SPPs, closer to Puerto Rico data than WLA data on subjects. We can draw two consequences: a) while MC variety has a lower rate of overt SPPs, the level of overt post-V is also low; b) the discussion reinforces the arguments expressed above in order to consider

! 21

the overall variable problem in reference to all kind of subjects and not only reduced to SPPs in the near future. Even though the numbers are small, it has been possible to realize a log-p analysis of the material (table 3).

Table 3. Distribution of post-V overt SPPs in MC, ordered by range Factors F/N % P

a) Grammatical person and number Uno12 3rd pl 1st pl 1st sg 3rd sg 2nd sg

Range= 63

21/27 3/20 3/19

29/232 6/96 2/49

77.8 15.0 15.8 12.5 6.2 4.1

.90

.75

.73

.52

.31

.27

b) Educational level Low

Medium High

Range= 47

47/161 13/158 4/124

29.2 8.2 3.2

.73

.45

.26

c) Ambiguity Ambiguous

Non-ambiguous Range= 36

41/126 23/317

32.5 7.3

.75

.39

d) Gender Male

Female Range= 29

47/201 17/242

23.4 7.0

.66

.37

Log likelihood= -120.994, significance= .000, input= .071 Only four factor groups were selected: grammatical person and number of the pronoun, educational level, morphological ambiguity, and gender. Two of these variables are also present in the general model (table 2), but other two are new, educational level and gender. Regarding grammatical person and number, two of the factors are really remarkable: uno and 1st sg. Undoubtedly, it is not casual that both actually refer to first discursive person. Almost all cases of uno are post-V (77.8%, 21/27, as in 32); this object-like position could be linked with the generic meaning of the pronoun: ‘me, but only as a member of a group’ (cf. Serrano 2012). The situation is different with yo; only 12.5% (29/232) are post-V. After reviewing all the examples one by one, many of them appear with Co-preterite or even with Post-preterite as a descriptive fragment in a wider narrative (frequently in the aperture): tenía yo, estaba yo, tendría yo, etc. Other group of examples indicate an important point in the narrative, and then they appear in Preterite: nací yo, me casé yo, etc.

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!12 Given its interesting behavior, uno has now been considered separately.

! 22

(32) porque/ pues <~pus> el Distrito se/ se compuso de/ pura gente// pues <~pus> pobre casi la mayoría/ no hombre que nomás ganaba uno/ si ganaba uno diez o veinte pesos/ ¡uh! era mucho/ era mucho ganar/ diez pesos (ME-314-13H-07, turn 72).

The selection of educational level as significant in post-position is noticeable. A cross-tabulation with person and number reveals a clear pattern. All cases of post-posed uno were produced by people of low educational level (in fact, 24 in 27 of every example of uno comes from low level). Besides, the 29 post-V yo shows also a very stratified distribution: 19 in low, 7 in medium and only 3 in high-level. Educational level is, then, an active social force in post-position. In (33), the speaker choses post-V probably to mitigate her presence and her sister’s in the narrative, in a recurrent way in people with low educational level, at least in our data. (33) pues de/ mamá ahorita <~orita> bueno tuviera yo otra hermana/ que tuviera <~tuviera:>/ un

año más que yo/ pero falleció/ ella chiquita (ME-308-12M-07, turn 105). The third factor group is ambiguity, with a clearly higher weight over ambiguous verbal forms. Two aspects can be converging in this result. First, the presence in post-V of a number of ambiguous Co-preterite tenses, so that the presence of SPPs serves strictly to avoid ambiguity (vivía, comía, ganaba), besides forms with Post-preterite, forms with usted, etc. Second, the abundance of uno and yo in post-V implies, of course, the occurrence of first and third morphological forms, precisely those that are more linked to morphological ambiguity. The last significant independent variable is gender. Men favor post-V, while they have less data with overt SPPs (in general, overt SPPs are more favored by women); the profile is completed with a cross-tabulation with age: 40 in 47 male examples are from men with 55 and more years. In conclusion, people who produce post-V overt SPPs are especially low-education older male speakers. Of course, pre-V or post-V position is not a deterministic question, and there is a large degree of freedom to make decisions. There are also a number of data where it is not obvious if we have pre-V or post-V. Both questions emerge in (34) to (37). (34) ¿qué pienso [Ø]?/ pues realmente yo como les he dicho yo “es que eso/ eso no e-/ no es justo

que lo hagan” (ME-307-11M-07, turn 176). (35) pero <~pero:>/ ahí ya tenía yo creo como dieciocho años yo/ mh (ME-308-12M-07, turn

173). (36) ni idea/ o sea [Ø] nunca me enteré por qué// pero/ pero [Ø] era un chavo así medio este/ pues

<~pus> ya/ para su edad que tenía entre// se-/ ¡muy infantil!// para su edad/ realmente porque/ pues <~pus> ya/ yo sí me acuerdo que estaba <~taba> chico yo/ y este// y ya [Ø] se veía con una apariencia ya mayor/ con bigote/ y [Ø] seguía conviviendo con un (ME-197-31H-01, turn 14).

(37) porque yo ahorita ya no nada nada/ yo vivía en la calle/ en la calle vivía yo/ y [la mejor (ME-314-13H-07, turn 84).

In (34), it is not easy to establish which yo is the subject and which an extra-predicative theme. In (35), the first yo can be a post-V subject of tenía, or a pre-V of creo, and the second yo could be a post-V of creo, or most probably a post-posed theme (or appendix). On the other hand, (36) and (37) seem to be a sort of chiasmus, with inversion

! 23

of terms (SV and VS) in the completive of (36)13 and in the conjoined sentences of (37). Especially in (37), it is difficult not to perceive a rhetorical proposal in the male speaker. Social factors As can be seen in table 4, the only significant social factor group is age. Educational level and gender were discarded. Regarding education, people of low educational level showed the highest rate of overt SPPs (24.2%, 161/666), vs. medium-level (20.9%, 157/752) and high-level (20.1%, 125/622). The relative correlation between overt SPPs and lower education suggested a possible significant social structure; however, even a model with only two factors (low vs. non-low level of education) was not statistically significant. Neither in NYC Mainlanders social class and education are significant, but in frequencies working class employs more overt SPPs than middle class (30 vs. 25%); and Elem.-H.S. speakers employ more than College-Grad. speakers (30 vs. 20%). However, lower SES (socioeconomic status) people use more overt forms than higher SES in a significant way (31% vs. 25%) (Otheguy & Zentella 2012, 121-122). It is remarkable that in MC and in NYC Mainlanders low-level groups always use more overt SPPs. For Otheguy and Zentella, their figures suggest a change from below in the form of a converging force among Mainlanders (but not among Caribbeans) (2012, 122). From the point of view of frequencies, women favor overt solutions (23.6%, 243/1028) slightly vs. men (19.8%, 200/1012). The point is interesting, because the same pattern appears in every age group: 31 vs. 28% for 55 an more years; 23 vs. 18% for 35-54 years; and 17 vs. 16% for 20-34 years. The predominance of females also appears in a significant way among overall Mainlanders in NYC (30 vs. 25%), and LAR Mainlanders (29 vs. 22%), but the difference is non-significant among NYR Mainlanders and among Caribbeans) (Otheguy & Zentella 2012, 117-121).14 NYC male and female Mexicans show almost the same percentages: 18.8% (men) and 18.2% (women) (apud Shin 2012, 136). One of our intuitions was that some older women produced a higher number of overt SPPs, especially in the case of yo. In fact, women use in general more pronouns than men (54.7% vs. 45.3%), but the proportion rises with yo (60.7% vs. 39.3%). Gender was non-significant in YS, while men produced higher rates (22% vs. 18%) (Michnowicz, this volume).

Table 4. Social distribution of overt SPPs (age) Factors F/N =% P

Age 55 and more years

35-54 years 20-34 years Range= 19

190/641 130/650 123/749

29.6 20.0 16.4

.61

.48

.42

Log likelihood= -1049.361, significance= .000, input= .212

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!13 Instead of interpreting in (36) a null subject ([Ø] estaba chico yo). In that case, the subject could refer to un chico or to the first yo (the speaker); the last yo would be an appendix. 14 Almost all demographic groups differ significantly among Mainlanders (Colombians, Ecuadorians and Mexicans) in Otheguy & Zentella, when gender, age, education, class, and socioeconomic status are considered (2012, 75).

! 24

What age reflects in table 4 is a clear pattern of decreasing rates in younger speakers, both in frequency (29.6 > 20.0 > 16.4%) and probability (.61 > .48 > .42). There is no doubt about the significance of the fact (= .000); the point, of course, is to decide between an age effect and a change in progress (Labov 1994, chapters 3-4). Maybe it is possible to talk about a floor effect with respect to the 16.4% of younger people: a minimum in the necessities of overt SPPs in the adult language (cf. the child language in Shin 2012), similar to the rate reported by Michnowicz in monolingual Yucatan Spanish (16%) —but neither discrete nor continuous age was significant in YS variety. This floor effect would be the counterpart of the ceiling effect (Shin & Otheguy 2013, 431).

Data from Barranquilla show a very similar pattern, with different percentages, but almost identical probabilistic weights (groups by age are not exactly equivalent): 50 or more years (41%, p= 0.60) > 20-50 years (33%, p= 0.49) > adolescents (32%, p= 0.40) (Orozco & Guy 2008, 78). If MC data were evidence of a change in progress, the data would be doubly defiant, given the Romance tendency towards higher rates of occurrence (in American Spanish, Brazilian Portuguese and of course French); i.e., “from this perspective lower SPP rates are archaic” (Erker & Guy 2012, 531; see also Mayol 2012). The significant role of age, and the tendencies of educational level and gender (clearly more overt SPPs in older people, and slightly lower educational speakers and women) suggest a more detailed discussion, outlined below. 4. Discussion of hypotheses The CRH question MC data analysis suggests that CRH is verified in a weak sense, more than in a strong sense: a) It is true that the overall catalogue of significant internal factors is similar to that of many other studies, especially in the more salient ranges (as grammatical person and number or co-reference); in some cases, the effect of some factor groups has been non-significant (as syntactic domain), but other related groups have emerged as significant (as verbal mood). Particularly, the probabilistic behavior of the factors included in some groups is virtually identical to the weight reported in other works (it is the case of person and number, co-reference, or ambiguity). These facts suggest the validity of CRH, taking dialects as witnesses of apparent time (or real geography), and underlining a similar behavior of linguistic constraints about them; b) however, there is a number of particular features that differentiate MC from other varieties: a very lower rate of overt SPPs with specific tú; morphological ambiguity and switch reference are at least partially independent, and the same can be said about ambiguity and verbal tenses; a low rate of overt SPPs, but also a low rate of post-posed overt SPPs (in contrast with the prediction by Cameron 1993). These aspects seem to weaken a strong acceptation of CRH; c) in any case, MC would not be engaged in a change in the address “more free overt SPPs position” → “pre-V overt SPPs”, as it has been proposed for Caribbean data, as a fragment of a more general solidification of SVO order. The main problem with MC data is to explain why the rate of overt SPPs is decreasing in apparent time; taking into account the relatively stable behavior of linguistic factors (in the probabilistic comparison with other dialects), it is necessary to look at social factors in search of a clearer portrait. Social variation The results have shown that hypothesis (1b) is false, at least in our sample. The role of age is significant in MC data (see above), so that the question is if our data suggest a change in

! 25

progress, an age-specific use of language, or a combined model (Cheshire 2005, Cameron 2011), in terms of a decreasing pronoun rate (vs. cases of an increasing pronoun rate, as in NYC, see Shin & Otheguy 2013, but similar to decreasing rates observed by Orozco & Guy 2008 in Barranquilla).15 In that sense, we examine the effect of individual speakers and the effects of subgroups by gender and educational level.

a) To examine the role of individual speakers is essential for considering a change in progress scenario and establishing who the leaders in the change are (cf. Labov 2001, part C), here in the form of a decreasing pattern. Otheguy & Zentella also consider rates in every individual speaker (2012, 76-81), with the Caribbean speakers in top positions. Their figures vary from 0.60 in ranking 1 to 0.09 in ranking 140.16 In the case of MC, table 5 shows the individual speaker17, every speaker’s age, the absolute frequency (F) of overt SPPs, and the rate (%) of overt SPPs per speaker, from 11.6% to 50%.

Table 5. Overt SPPs per speaker in MC Speaker Age F Rate (%)

32H 45 13 11.6 31H 29 14 11.8 23H 59 16 12.5 21M 27 19 14.8 11H 27 19 16.0 11M 21 19 16.5 22H 40 24 18.0 21H 20 29 19.7 12M 37 24 20.2 31M 21 24 20.2 22M 51 23 23.5 13M 78 33 24.1 32M 40 23 24.5 12H 52 24 25.3 33H 72 21 27.6 33M 58 31 31.0 23M 56 46 38.3 13H 73 41 50.0

As can be seen, there are important differences among informants: so, 23M, the speaker with a higher absolute number of overt SPPs, shows a figure more than three times higher than 32H (first in the column), with only 13 cases; the same can be said about rate. Even though there are important differences among speakers, these differences do not rise in an abrupt, but in a gradual and stable form, without extreme outliers, maybe with the

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!15 It is interesting that Mexican children (from 6;4 to 7;8) studied by Shin (2012) show low rates of overt SPPs (6.3%), with slightly higher rates in girls (6.7% vs. 5.7% in boys). We are not thinking of a simplistic connection between these and our data (for age/gender effects along life on SPPs, see Shin 2012, 138), but it is worthwhile to observe that both observations are compatible. Dominican monolingual children (from 4 to 12 years) studied by Ortiz (2011, 424) show 39% of overt SPPs (vs. 49% in monolingual adult speakers). 16 Bayley et al. (2012, 58) offer the individual results as well. 17 The meaning of speaker's label is: first number, educational level (1, low; 2, medium; 3, high); second number, age (1, 20-34; 2, 35-54; 3, 55-); letter "H", male, and letter "M", female.

! 26

exception of 13H (ten speakers stay under the overall rate, 21.7%, and eight are over the overall rate). If we observe the extremes of the series (Martín Butragueño 2006, 2014), we can divide the list of informants in three thirds, and to assign a low, a medium and a high value to each segment. Among six lower rates, there are four men (in fact, three men have the lowest rates), four young speakers, and two people of each educational level. In the high extreme, there are the same number of male and female, four old people and three high-educational level speakers. A crosstab between age and gender reveals that older women have the highest percentage of overt SPPs, 31% (vs. 28% in their male peers), almost ten points over the mean (21.7%).

Regarding a possible ongoing change, this would be a retraction movement (i.e., the younger people have a lesser rate of overt SPPs), so that older people would be losing its traditional weight. It is true that the same process is taking place with other linguistic variables in Mexican Spanish (as vowels and rhotics, Martín Butragueño 2006, 2014), but the leaders of change in those processes are basically self-made women with a personal history of social rise, in the line of description of leaders by Labov (2001). In NYC, Shin & Otheguy observe also that "in addition to revealing an inverse relationship between affluence and contact-induced change in bilingual New York, we find that the ongoing change [increasing overt SPPs] is led by first-generation women [...] the position of linguistic leadership occupied by women is the result of their extensive contact with high-pronoun using second-generation Latinos" (2013, 431-432). In short, if we would have a change in progress in the expression of SPPs, it would necessary to explain why the men are the leaders (if they are) and how they could have a social effect.

b) In all previous observations, age has been taken as a discrete variable, divided in three generations. It is possible to refine the perspective on social pattern when age is considered in a continuous way (see second column in table 5). A series of complementary linear analyses (see 38, realized with SPSS 21, 2012) confirms statistical tendencies observing the role of age in subgroups by gender and educational level (in a similar way to Labov 1994, ch. 3). R2 indicates the amount of observed variation in the dependent variable that can be explained by variation in independent variable; FANOVA and Standardized β test the signification of regression equation (Hernández Campoy & Almeida 2005, 240-248). (38) a. Age. Overall data: rate (%)= .311*age + 8.595, R2= .355. [FANOVA= 8.803, p= .009; Standardized β= .596, p= .009]. b. Age and gender.

Men: rate (%)= .397*age + 2.972, R2= .397. [FANOVA= 4.610, p= .069; Standardized β= .630, p= .069].

Women: rate (%)= .240*age + 13.308, R2= .402. [FANOVA= 4.702, p= .067; Standardized β= .634, p= .067]. c. Age and educational level.

Low: rate (%)= .378*age + 7.207, R2= .505. [FANOVA= 4.079, p= .114; Standardized β= .711, p= .114]. Medium: rate (%)= .194*age + 12.967, R2= .112. [FANOVA= .506, p= .516; Standardized β= .335, p= .516]. High: rate (%)= .261*age + 9.571, R2= .363. [FANOVA= 2.282, p= .205; Standardized β= .603, p= .205].

! 27

The linear trend for the overall data by age shows an interesting R2 (.355, i.e., it accounts for the 35.5% of the variance) and its ANOVA and Standardized β are significant. That is, age as a continuous variable remains significant. In the subgroups of age by gender, men and women exhibit a similar enough R2 (.397 and .407), and ANOVA and Standardized β are non-significant. However, p (.067 and .069) is not far from being significant; the slope is slightly steeper for men, but women are moving through a higher level. Regarding subgroups of continuous age and educational level, all of them show non-significant ANOVA and Standardized β. Only low-level people present (besides a high R2= .505), not so far of significance the ANOVA and the Standardized β (p= .114).

What is the meaning of these results? First, we have here a notorious difference linked to continuous age, with more null SPPs in youngest people. Second, what linear data say besides is that men and women are moving basically in parallel. Third, younger low-level people offer a non-significantly lesser degree of overt SPPs than older low-level people (vs. medium and high-level people).

With only this evidence in hand, we cannot say if we have here an ongoing change or not. It is likely that a future model has to include both components: stratification by age (and partially by gender and educational level), based in the development of a different conversational style through the individual's life;18 and change in progress, based in the floor effect (a variant of the economy principle, in the end). And maybe the background to the SPPs expression has to do with the fast change that is in progress in many Mexican cities, and particularly in MC, in relation to forms of address (see Hummel, Kluge & Vázquez Laslop 2009), especially regarding the actual reduction in the use of yo, usted and uno. But this is a hypothesis that will have to be explored later on. 6. Conclusions a) The overall rate of overt SPPs in MC (21.7%) suggests that the variety makes part of conservative dialects. b) For overt SPPs, linguistic significant factor groups in MC are grammatical person and number; co-reference; verbal mood; enunciative type; textual genre; verbal tense; ambiguity; style. There are remarkable coincidences with the groups and the probability of factors mentioned in the literature about SPPs.

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!18 It is interesting to consider qualitatively some older people: 23M is a medium level, 56 year-old woman; she only studied secondary school, plus a series of formative courses at a later moment: she produces the impression of a self-made woman. In the fragment we have examined, she is speaking about her childhood, and the high level of yo (32 in 47 SPPs) is worth noticing (yo me peleaba, yo pienso, yo doy gracias a la vida). Her profile fits quite well with the fact that some older women in Mexico City reveal remarkable coefficients of overt SPPs. 13H would seem to be a partial exception in the picture, as a male speaker; he is a man with no studies, 73 years old, who worked as a loader and had had an especially hard life. His interview is halfway between a description of life in popular neighborhoods, and a more personal narrative of himself as a member of that landscape, but from the background, what converges with a medium level of yo (15/42) and especially of uno (18/42). 13M is a 78 years old woman, also without studies, housewife and maid (especially in the past); the fragment selected in our sample is a particularly shocking narrative of her childhood and youth, and it shows a high number of yo (24/34, as in yo sufría, yo me quiero ir de aquí, yo ya no quiero estar), in consonance with the older female behavior with SPPs, but above all building a vivid account of her experiences. Eventually, 33M is another old woman (58 years old), but in this case she has a high educational level (she works as a teacher in a university), and the topic of the fragment is the life of a female writer; this circumstance particularly explains the high number of ella (26/31). Apparently, the high general number of pronouns in this interview is associated with the necessity of textual disambiguation.

! 28

c) Age is the most robust social factor group; it is significant as a discrete variable in logistic regression analysis. Besides, it is significant as a continuous variable in linear regression analysis; gender and educational level behave as non-significant variables when they are subordinated to continuous age in linear regression analysis. It could be established a floor effect around a rate of 10-15% of SPPs, reached by some speakers in MC. d) MC data seems to respect a weak version of CRH, and to offer some evidence of age variation, in order to decrease the number of overt SPPs in younger speakers. The meaning of this second aspect needs additional research, considering more speakers and exploring the causes of reduction of overt SPPs. 7. References AMASTAE, JON, BETTY DAJLALA T., XÓCHITL DÍAZ, ALBERTO ESQUINCA M., SAIDAH YAZMÍN OCHOA, PHILIP

PORTER, & MARÍA JOSÉ TORRES 2000. "Ø Hablo: el uso del sujeto en el español fronterizo" [El Paso, Juárez, Parral], in Memorias del V Encuentro Internacional de Lingüística en el Noroeste. Ed. María del Carmen Morúa Leyva & Gerardo López Cruz. Hermosillo: Universidad de Sonora, t. 2, 219-232.

BARRENECHEA, ANA MARÍA, & ALICIA ALONSO 1977. “Los pronombres personales sujetos en el español hablado en Buenos Aires”, in Estudios sobre el español hablado en las principales ciudades de América. Ed. Juan M. Lope Blanch. México: Universidad Nacional Autónoma de México, 333-349.

BAYLEY, ROBERT, & LUCINDA PEASE-ÁLVAREZ 2006. "Null and expressed pronoun variation in Mexican-descent children's Spanish", in Sociolinguistic Variation: Data, Theory, and Analysis. Ed. J. Arnold, R. Blake, B. Davidson, S. Schwenter & J. Solomon. Stanford: CSLI, 85-99.

-------- NORMA L. CÁRDENAS, BELINDA TREVIÑO SCHOUTEN, CARLOS MARTÍN VÉLEZ SALAS 2012. "Spanish dialect contact in San Antonio, Texas: An exploratory study", in Selected Proceedings of the 14th Hispanic Linguistics Symposium. E. K. Geeslin & M. Díaz-Campos. Somerville: Cascadilla Proceedings Project, 48-60.

BENTIVOGLIO, PAOLA 1980. Why canto and not yo canto? The problem of first-person subject pronoun in spoken Venezuelan Spanish. M. A. thesis. Los Angeles: University of California.

-------- 1987. Los sujetos pronominales de primera persona en el habla de Caracas. Caracas: Universidad Central de Venezuela.

--------, LUIS A. ORTIZ, & CARMEN SILVA-CORVALÁN 2011. "La variable expresión del sujeto pronominal. Guía de codificación", in <http://preseea.linguas.net/Metodolog%C3%ADa.aspx>.

BLACKWELL, SARAH & MARGARET LUBBERS QUESADA 2012. “Third-Person Subjects in Native Speakers' and L2 Learners' Narratives: Testing (and Revising) the Givenness Hierarchy for Spanish”, in Selected Proceedings of the 14th Hispanic Linguistics Symposium. E. K. Geeslin & M. Díaz-Campos. Somerville: Cascadilla Proceedings Project, 142-164.

BLANCO CANALES, ANA 1999. “Presencia/ausencia de sujeto pronominal de primera persona en español”, Español Actual, 72, 31-39.

CAMERON, RICHARD 1993. “Ambiguous agreement, functional compensation, and non-specific tú in the Spanish of San Juan, Puerto Rico, and Madrid, Spain”, Language Variation and Change, 5, 305-334.

-------- 1995. "The scope and limits of switch reference as a constraint on pronominal subject expression", Hispanic Linguistics, 6/7, 1-28.

-------- 1996. "A community-based test of a linguistic hypothesis", Language in Society, 25, 61-111. -------- 1997. "Accessibility theory in a variable syntax of Spanish", Journal of Pragmatics, 28, 29–67. -------- 2011. “Aging, age, and sociolinguistics”, in The Handbook of Hispanic Sociolinguistics. Ed. M. Díaz-

Campos. Oxford: Wiley-Blackwell, 207-229. --------, & N. FLORES-FERRÁN 2003. "Perseveration of subject expression across regional dialects of Spanish",

Spanish in Context, 1, 41-65. CANTERO SANDOVAL, GUSTAVO 1978. "Observaciones sobre la expresión innecesaria de los pronombres

personales sujeto en el español de México", Anuario de Letras, 16, 261-264. -------- 1986. "Tipos de expresión obligatoria de los pronombres personales sujeto en español", in Actas del V

Congreso Internacional de la Asociación de Lingüística y Filología de América Latina. Caracas: Instituto de Filología Andrés Bello, 243-248.

! 29

CHESHIRE, J. 2005 "Age and generation-specific use of language", in Sociolinguistics: An Introductory Handbook of the Science of Language and Society. Ed. U. Ammon, N. Dittmar, K. Mattheier & P. Trudgill Berlin: Mouton de Gruyter, 1552-1563.

CLAES, JEROEN 2011. “¿Constituyen las Antillas y el Caribe continental una sola zona dialectal? Datos de la variable expresión del sujeto pronominal en San Juan de Puerto Rico y Barranquilla, Colombia”, Spanish in Context, 8, 191-212.

COMPANY COMPANY, CONCEPCIÓN 2012. “Historical morphosyntax and grammaticalization”, in Handbook of Spanish Linguistics. Ed. J. I. Hualde, A. Olarrea & E. O’Rourke. Oxford: Wiley-Blackwell, 673-693.

ERKER, DANIEL, & GREGORY R. GUY 2012. “The role of lexical frequency in syntactic variability: variable subject personal pronoun expression in Spanish”, Language, 88, 526-557.

FLORES-FERRÁN, NYDIA 2007a. "A bend in the road: Subject personal pronoun expression in Spanish after thirty years of sociolinguistic research", Language and Linguistics Compass, 1, 624-652.

-------- 2007b. “Los Mexicanos in New Jersey: Pronominal expression and ethnolinguistic aspects”, in Selected Proceedings of the Third Workshop on Spanish Sociolinguistics. Dir. Jonathan Holmquist, Augusto Lorenzino & Lotfi Sayahi. Somerville, MA: Cascadilla Proceedings Project, 85-91.

HERNÁNDEZ CAMPOY, JUAN MANUEL, & MANUEL ALMEIDA 2005. Metodología de la investigación sociolingüística. Málaga: Comares.

HOCHBERG, JUDITH G. 1986. “Functional compensation for /s/ deletion in Puerto Rican Spanish”, Language, 62, 609-621.

HOLMQUIST, JONATHAN 2012. "Frequency rates and constraints on subject personal pronoun expression: Findings from the Puerto Rican highlands", Language Variation and Change, 24, 203-220.

HUMMEL, MARTIN, BETTINA KLUGE, & MARÍA EUGENIA VÁZQUEZ LASLOP (eds.) 2009. Formas y formulas de tratamiento en el mundo hispánico. México – Graz: El Colegio de México – Karl-Franzens-Universität.

HURTADO, LUZ 2005. “El uso de tú, usted y uno en el español de los colombianos y colombo-americanos”, in Contactos y contextos lingüísticos: el español en los Estados Unidos y en contacto con otras lenguas. Ed. Luis A. Ortiz López & Manuel Lacorte. Frankfurt - Madrid: Vervuert - Iberoamericana, 185–200.

KROCH, ANTHONY 1989a. “Function and Grammar in the History of English Periphrastic do”, in Language Change and Variation. Ed. R. Fasold & D. Schiffrin. Amsterdam: John Benjamins, 133-172.

-------- 1989b “Reflexes of grammar in patterns of language change”, Language Variation and Change, 1, 199-244.

LABOV, WILLIAM 1994. Principles of Linguistic Change. Vol. 1: Internal Factors. Oxford: Blackwell. -------- 2001. Principles of Linguistic Change. Vol. 2: Social Factors. Oxford: Blackwell. -------- 2010. Principles of Linguistic Change. Vol. 3: Cognitive and Cultural Factors. Oxford: Wiley –

Blackwell. LASTRA, YOLANDA, & PEDRO MARTÍN BUTRAGUEÑO 2000. “El modo de vida como variable

sociolingüística”, in Estudios de variación lingüística. Ed. P. Martín. México: El Colegio de México, 13-43.

--------, & PEDRO MARTÍN BUTRAGUEÑO 2010. “Futuro perifrástico y futuro morfológico en el Corpus sociolingüístico de la ciudad de México”, Oralia, 13, 145-171.

--------, & PEDRO MARTÍN BUTRAGUEÑO 2012. “Aproximación al uso del modo subjuntivo en la ciudad de México”, Boletín de Filología, 47, 101-131.!

MARTÍN BUTRAGUEÑO, PEDRO 1999. "¿Es funcional la variación sintáctica, in Estudios de variación sintáctica. Ed. Ma. J. Serrano. Madrid - Frankfurt: Iberoamericana - Vervuert, 221-235.

-------- 2006. “Líderes lingüísticos en la ciudad de México”, en Líderes lingüísticos. Estudios de variación y cambio. Ed. P. Martín. México: El Colegio de México, 185-208.

-------- 2010. "La posición extrapredicativa de tema en la lengua hablada", en Estudios de gramática descriptiva del español. Sintaxis, semántica y entonación del orden de palabras. Ed. S. Bogard. México: El Colegio de México, 117-183.

-------- 2014. Fonología variable del español de México. Vol. I: Procesos segmentales. México: El Colegio de México.

--------, & YOLANDA LASTRA 2011. Corpus sociolingüístico de la ciudad de México. Vol. 1: Hablantes de instrucción superior. México: El Colegio de México.

--------, & YOLANDA LASTRA 2012. Corpus sociolingüístico de la ciudad de México. Vol. 2: Hablantes de instrucción media. México: El Colegio de México.

! 30

--------, & YOLANDA LASTRA forthcoming. Corpus sociolingüístico de la ciudad de México. Vol. 3: Hablantes de instrucción baja. México: El Colegio de México.

--------, & MARÍA EUGENIA VÁZQUEZ LASLOP 2002. “Variación y dinamimsmo lingüístico: problemas de método”, Lexis, 26, 305-344.

MAYOL, LAIA 2012. “An account of the variation in the rates of over subject pronouns in Romance”, Spanish in Context, 9, 420-442.

MICHNOWICZ, JIM. “Subject pronoun expression in Yucatan Spanish”, this volume. MONTES MIRÓ, ROSA GRACIELA 1986. "Factores discursivos en el análisis de los pronombres personales

sujeto en español", Morphé, 2, 45-71. MORALES, AMPARO 1982. "La perspectiva dinámica oracional en el español de Puerto Rico", in El español

del Caribe: Ponencias del VI Simposio de Dialectología. Ed. Orlando Alba. Santiago de los Caballeros: Universidad Católica Madre y Maestra, 203-219.

-------- 1989. "Hacia un universal sintáctico del español del Caribe: el orden SVO", Anuario de Lingüística Hispánica, 5, 139-152.

OROZCO, RAFAEL, & GREGORY GUY 2008. "El uso variable de los pronombres sujetos: ¿qué pasa en la costa Caribe colombiana", in Selected Proceedings of the 4th Workshop on Spanish Sociolinguistics. Ed. Maurice Westmoreland & Juan Antonio Thomas. Somerville, MA: Cascadilla Proceedings Project, 70-80.

ORTIZ LÓPEZ, LUIS A. 2009. “Pronombres de sujeto en el español (L2 vs L1) del Caribe”, in El Español en Estados Unidos y otros contextos de contacto. Ed. Jennifer Lehman, & Manel Lacorte. Frankfurt - Madrid: Vervuert – Iberoamericana, 85-110.

-------- 2011. "Spanish in contact with Haitian Creole", in The Handbook of Hispanic Sociolinguistics. Ed. M. Díaz-Campos. Oxford: Wiley-Blackwell, 418-445.

OTHEGUY, RICARDO, & ANA C. ZENTELLA 2012. Spanish in New York. Language Contact, Dialectal Leveling, and Structural Continuity. Oxford: Oxford University Press.

--------, ANA C. ZENTELLA, & INGRID HEIDRICK 2012. "Coding manual", in OTHEGUY & ZENTELLA 2012, 225-273.

--------, ANA C. ZENTELLA, & D. LIVERT 2007. “Language and dialect contact in Spanish in New York: Towards the formation of a speech community”, Language, 83, 770-802.

PRESEEA 2011. Proyecto para el estudio sociolingüístico del español de España y de América. “Guía PRESEEA para la investigación lingüística”, in <http://preseea.linguas.net/Metodolog%C3%ADa.aspx>.

QUESADA, MARGARET LUBBERS, & SARAH BLACKWELL 2009. “The L2 acquisition of null and overt Spanish subject pronouns: A pragmatic approach”, in Selected Proceedings of the 11th Hispanic Linguistics Symposium. Ed. J. Collentine et al. Somerville: Cascadilla Proceedings Project, 117-130.

RAE-ASALE 2009. Real Academia Española - Asociación de Academias de la Lengua Española. Nueva Gramática de la lengua española. Madrid: Espasa.

SANKOFF, D., S. A. TAGLIAMONTE, & E. SMITH 2012. Goldvarb Lion. A Multivariate Analysis Application. Toronto: University of Toronto, in <http://individual.utoronto.ca/tagliamonte/goldvarb.htm>.

SERRANO, MARÍA JOSÉ 2012. "El sujeto pronominal usted/ustedes y su posición. Variación y creación de estilos comunicativos", Spanish in Context, 9, 109-131.

SHIN, NAOMI LAPIDUS 2012. "Variable use of Spanish subject pronouns by monolingual children in Mexico", in 14th Hispanic Linguistics Symposium. Ed. Kimberly Geeslin & Manuel Díaz-Campos. Somerville, MA: Cascadilla Proceedings Project, 130-141.

--------, & RICARDO OTHEGUY 2013. "Social class and gender impacting change in bilingual settings: Spanish subject pronoun use in New York", Language in Society, 42, 429-452.

SILVA-CORVALÁN, CARMEN 1977. A Discourse Study of Word Order in the Spanish Spoken by Mexican-Americans in West Los Angeles. M.A. thesis. Los Angeles: University of California.

-------- 1982. “Subject variation in spoken Mexican-American Spanish”, in Spanish in the United States: Sociolinguistic aspects. Ed. J. Amastae & L. Elías-Olivares. New York: Cambridge University Press, 93-120.

-------- 1994. Language Contact and Change. Oxford: Oxford University Press. -------- 1997. "Variación sintáctica en el discurso oral: problemas metodológicos", en Trabajos de

sociolingüística hispánica. Ed. Francisco Moreno Fernández. Alcalá de Henares: Universidad de Alcalá, 115-135.

-------- 2001. Sociolingüística y pragmática del español. Washington: Georgetown University Press. TAGLIAMONTE, SALI A. 2006. Analysing Sociolinguistic Variation. Cambridge: Cambridge University Press.

! 31

-------- 2012. Variationist Sociolinguistics. Change. Observation, Interpretation. Oxford: Wiley-Blackwell. SPSS 21 2012. Statistical Package for the Social Sciencies, version 21. International Business Machines Corp. TRAVIS, CATHERINE 2005. “The yo-yo effect: Priming in subject expression in Colombian Spanish”, in

Theoretical and Experimental Approaches to Romance Linguistics: Selected Papers from the 34th Linguistic Symposium on Romance Linguistics (LSRL). Ed. Randall Gess & Edward J. Rubin. Amsterdam: John Benjamins, 329-349.

-------- 2007. “Genre effects on subject expression in Spanish: Priming in narrative and conversation”, Language Variation and Change, 19, 101-135.

--------, RENA TORRES CACOULLOS 2012. "What do subjects pronouns do in discourse? Cognitive, mechanical and constructional factors in variation", Cognitive Linguistics, 23, 711-748.

APPENDIX Frequent verbs in MC by overt SPPs rate

Verb F

(collection) F

(sample) %

(samp./collec.) F

(overt SPPs) %

(overt SPPs) creer 50 48 96.0 26 54.2 gustar 60 2 3.3 1 50.0 nacer 18 13 72.2 6 46.2 comer 16 14 87.5 6 42.9 ganar 19 16 84.2 6 37.5 vivir 58 53 91.4 18 34.0 crecer 10 6 60.0 2 33.3 mandar 13 3 23.1 1 33.3 poner 15 12 80.0 4 33.3 querer 40 36 90.0 12 33.3 estar 175 79 45.1 26 32.9 bailar 18 13 72.2 4 30.8 pensar 11 10 90.9 3 30.0 trabajar 73 51 69.9 15 29.4 llegar 56 35 62.5 10 28.6 quedarse 26 22 84.6 6 27.3 acordarse 20 20 100.0 5 25.0 buscar 14 8 57.1 2 25.0 ser 605 131 21.7 31 23.7 estudiar 21 17 81.0 4 23.5 ver 73 63 86.3 14 22.2 tener 212 163 76.9 36 22.1 conocer 34 28 82.4 6 21.4 entrar 22 14 63.6 3 21.4 venir 20 14 70.0 3 21.4 andar 20 15 75.0 3 20.0 llevarse 12 10 83.3 2 20.0 pagar 18 10 55.6 2 20.0 sentir 10 10 100.0 2 20.0 terminar 18 10 55.6 2 20.0 dar 78 26 33.3 5 19.2 decir 184 134 72.8 24 17.9 irse 21 17 81.0 3 17.6 saber 91 74 81.3 12 16.2

! 32

ir 73 57 78.1 9 15.8 vender 27 19 70.4 3 15.8 hacer 102 68 66.7 10 14.7 darse 12 7 58.3 1 14.3 mirar 10 7 70.0 1 14.3 durar 11 10 90.9 1 10.0 encontrar 11 11 100.0 1 9.1 empezar 34 24 70.6 2 8.3 aprender 15 15 100.0 1 6.7 llevar 33 27 81.8 1 3.7 dedicarse 19 11 57.9 0 0.0 dejar 20 14 70.0 0 0.0 haber 115 0 0.0 0 0.0 hacerse 13 4 30.8 0 0.0 llamarse 32 2 6.3 0 0.0 meterse 13 12 92.3 0 0.0 pasar 23 4 17.4 0 0.0 perder 11 9 81.8 0 0.0 traer 21 15 71.4 0 0.0