Sociolinguistic variation in U.S. Spanish (Post-print, pre-typesetting)

Abstract

In this chapter, we survey research on correlations of linguistic behavior with speakers’ sociodemographic characteristics. The effects of social factors suggest stable variation in some cases, and changes in progress in others. In cases of stable variation, the patterns reported correspond to those found throughout the Spanish-speaking world. For example, women tend to use standard variants, such as sibilant rather than aspirated syllable-final /s/, more than men do among Puerto Ricans in Philadelphia. Change in progress in immigrant communities may be detected synchronically in apparent time and inferred from comparison with variation patterns in communities of origin. One scenario may be the receding of a phonological variant in both its overall rate and associations with social factors, as with the decline of hiatus-breaking [ʝ] (cayer from caer ‘to fall’) in Salvadoran communities in Houston. Another may be the spread of a variant and the development of social conditioning, for example, the reduction of intervocalic ⟨ll⟩ (eØa from ella ‘she/her’) among Mexicans and Salvadorans living in the same neighborhoods in Houston. Contributing to the paucity of studies on social factors in U.S. Spanish has been the familiar problem that social characteristics of speakers are generally less well-defined than linguistic categories, particularly in minority language situations, and that social factors are often highly interdependent. A solution to this conundrum can be found in principal component analysis, as a heuristic for grouping speakers strictly on the basis of their linguistic behavior, demonstrated here in a community-based corpus of New Mexican Spanish.

Full text

To appear in: Potowski, K. (ed.) Handbook of Spanish as a Heritage/Minority Language, Routledge.
(Final version, pre-typesetting)
!
1
Sociolinguistic variation in U.S. Spanish1
Rena Torres Cacoullos and Grant M. Berry
Abstract
In this chapter, we survey research on correlations of linguistic behavior with speakers’
sociodemographic characteristics. The effects of social factors suggest stable variation in some
cases, and changes in progress in others. In cases of stable variation, the patterns reported
correspond to those found throughout the Spanish-speaking world. For example, women tend to
use standard variants, such as sibilant rather than aspirated syllable-final /s/, more than men do
among Puerto Ricans in Philadelphia. Change in progress in immigrant communities may be
detected synchronically in apparent time and inferred from comparison with variation patterns in
communities of origin. One scenario may be the receding of a phonological variant in both its
overall rate and associations with social factors, as with the decline of hiatus-breaking [ʝ]!(cayer
from caer ‘to fall’) in Salvadoran communities in Houston. Another may be the spread of a variant
and the development of social conditioning, for example, the reduction of intervocalic ⟨ll⟩ (eØa
from ella ‘she/her’) among Mexicans and Salvadorans living in the same neighborhoods in
Houston. Contributing to the paucity of studies on social factors in U.S. Spanish has been the
familiar problem that social characteristics of speakers are generally less well-defined than
linguistic categories, particularly in minority language situations, and that social factors are often
highly interdependent. A solution to this conundrum can be found in principal component analysis,
as a heuristic for grouping speakers strictly on the basis of their linguistic behavior, demonstrated
here in a community-based corpus of New Mexican Spanish.!

To appear in: Potowski, K. (ed.) Handbook of Spanish as a Heritage/Minority Language, Routledge.
(Final version, pre-typesetting)
!
2
1. THE SOCIAL CONTEXT OF SPANISH IN THE U.S.: CONCEPTUAL AND
METHODOLOGICAL CHALLENGES
The neglect of sociolinguistic patterns of language variation in U.S. Spanish is due at least in part
to the abiding preoccupation with contact-induced change and a methodological predilection for
acceptability judgments, experimental tasks, or cherry-picked examples. This “Hispanic tradition”
of language study as Bills (1975:vi-vii) characterized it nearly half a century ago is hampered by
an “interest in the accumulation of speech fragments with little concern for linguistic or
sociological context” and “almost exclusive interest in deviations from standard Spanish”.
Adherence to the analyst’s idealizations as the benchmark for evaluations leaves working-class
varieties of U.S. Spanish in a no-win situation, as pointed out by Ana Celia Zentella; for example,
New Mexican Spanish is branded ‘archaic’ “porque se describe en referencia a la norma de otra
comunidad [because it is described in reference to the norm of another community],” yet “tampoco
se vale ser innovador…al notar la reacción…en contra de…lonche…y otros préstamos [it isn’t
worthwhile to be innovative either...when one notes the reaction...against...lonche... and other
borrowings]” (Zentella 1990:157).
Methodological issues begin with data collection. As Peñalosa (1981:7) asserted, appropriate data
come from “Labovian-type studies of Chicano speech in a natural setting.”. This is because the
vernacular—the unreflecting use of language in the absence of the observer, when minimum
attention is paid to monitoring speech—is the style that is most regular in structure (Labov
1972:112). In contrast, when speakers of subordinate varieties are asked direct questions about
their language, as is the case with acceptability judgments, their answers shift toward (or away
from) the prestige variety in irregular and unforeseeable ways (Labov 1972:111). Whether data
are gathered by an in- or out-group member has also been demonstrated to make a difference, for
example, in rates of word-final nasal velarization by Salvadorans interviewed by a Mexican in
Houston (Hernández 2011:67; see Section 3 below).
When we turn our attention to continuity with Spanish varieties spoken across the Americas and
to community-based samples of vernacular speech (e.g., Otheguy and Zentella 2012), social
factors in tandem with linguistic constraints become important for diagnosing stability vs. change
in U.S. Spanish varieties; these factors also serve to detect parallels vs. divergences vis-à-vis
Spanish varieties outside the U.S. Moreover, even for assessing contact-induced change, it has
become clear that neglect of the social context of bilingualism is risky, because once speakers are
adequately characterized with respect to social factors, phenomena attributed either to majority
language influence or to minority language loss may turn out to be conditioned by social class
instead.
An instructive example concerns use of the subjunctive mood, which undergoes attrition among at
least some second and third generation speakers of Spanish in the U.S. (e.g., Ocampo 1990; Silva-
Corvalán 1994: 86-90). A parallel presumed loss of the French subjunctive in Canada is imputed
to contact with English. To test this, Poplack (1997) considered external measures of contact at
the individual and community level. If contact with English is playing a role, speakers with higher
English proficiency and those living in neighborhoods with a higher proportion of English speakers
should show a lower rate of the subjunctive than speakers with lower indices of contact with
English. Neither measure of contact correlated with subjunctive rate. Instead, after accounting for

To appear in: Potowski, K. (ed.) Handbook of Spanish as a Heritage/Minority Language, Routledge.
(Final version, pre-typesetting)
!
3
a strong lexical effect of the governing verb, systematic quantitative analysis of both internal and
external constraints exposed the unsuspected effect of social class, with professionals displaying a
proclivity for the subjunctive (Poplack and Levey 2010:402-4). Moreover, subjunctive use is
characterized not by change, but by long-term stable variability, despite centuries of normative
injunctions (Poplack, Lealess and Dion 2013).
In this chapter, we survey the scant number of reports on social factors in U.S. Spanish, first for
stable linguistic variables—those with distribution patterns that persist across time and
communities—and then for possible changes in progress—where age distributions are gradient.
In the final section, we apply statistical procedures (principal component analysis and regression
analysis) to a community-based corpus of New Mexican Spanish to infer and test social factors
relevant to conditioning language variation.
2. SOCIAL CLASS AND GENDER IN STABLE VARIATION
The “central dogma” of sociolinguistics stated by Labov is that “the community is prior to the
individual” (Labov 2006:5). Individual speaker behavior can be understood only once the
community pattern is known, since individual linguistic behavior results from social histories and
memberships. For Spanish-speaking communities in the U.S., the social factor most considered,
though often in isolation, has been speaker gender. Usually, gender is implicated in claims of
changes in progress, with women being seen in some cases as conservative and in others as leaders
of linguistic change. Hasty pronouncements of change can stem from equating language change
with perceived departures from an idealized norm or even with ordinary variation; that is, failing
to differentiate between situations of stable variation and situations of change in progress.
Crucially, while all change implies the existence of variability in language, the converse is not
true: “not all variability and heterogeneity in language structure involves change” (Weinreich,
Labov and Herzog 1968:188).
The “gender paradox” is the pattern of gender differentiation whereby “women conform more
closely than men to sociolinguistic norms that are overtly prescribed, but conform less than men
when they are not” (Labov 2001:292-93). This follows from two generalizations for the distinct
scenarios of variation: women use stigmatized variants at lower rates than men for stable
sociolinguistic variables, but adopt innovative forms earlier, both for prestige variants and for
linguistic changes from below, i.e, changes from within the system which occur below the level
of conscious awareness and consequently lack style shifting in initial stages (Labov 2001:261-93).
A stable sociolinguistic variable in Spanish is variation in the realization of the forms of the copula
estar. This dates back approximately half a millennium, judging from the recommendation of Juan
de Valdés (1535) that the verb be written without e- to distinguish it from the demonstrative
pronoun esta:
“me ha parecido, por no hacer tropezar al letor, poner la e cuando son pronombres, porque
el acento está en ella, y quitarla cuando son verbos, porque, estando el acento en la última,
si miráis en ello, la primera e casi no se pronuncia, aunque se escriba [It has seemed a good
idea to me, in order not to trip up the reader, to put the letter e before the words that are
pronouns, because it is accented there, and not to include it in verbs, because, being that

To appear in: Potowski, K. (ed.) Handbook of Spanish as a Heritage/Minority Language, Routledge.
(Final version, pre-typesetting)
!
4
the accent is on the last syllable, if you think about it, the first e is almost not pronounced,
even though it is written] ”.
Drawing on 32 interviews recorded in San Antonio with Mexican-Americans raised in South
Texas, Garcia and Tallon (2000) examined three variants of estar: está, ‘stá (with apheresis of the
vowel) and ‘tá (N = 1,025) in multivariate analysis. They find phonological conditioning by
preceding segment. In addition, and in accordance with the generalization for stable variation,
female speakers favored “canonical” está, while males favored the ‘tá variant, leading the authors
to suggest that the latter may be “the least formal variant” and “a marker of male speech” (García
and Tallon 2000: 356-57).
Another variable showing social stratification is the alternation between para and pa’. This is
conditioned by morphosyntactic and phonological, but also social, factors. A study based on data
extracted from recordings with 171 speakers in San Antonio (token numbers not reported)
indicates that males use the abbreviated form of the preposition at nearly double the rate of females
(42% vs. 23%) (Lantolf 1982). What is more, gender interacts with education and occupation:
although males display a higher rate of pa’ across all occupation and education levels, the gender
difference is as high as 48 percentage points for blue-collar workers and as low as 5 for
professionals, with a 20-point difference within white-collar workers (the largest group sampled)
(Lantolf 1982:172). Considering internal factors, the reduced form was favored in directional
(locative) uses (Lantolf 1982:167). The social and linguistic conditioning of para ~ pa’ in San
Antonio parallels that found outside the United States. Based on two analyses in Venezuelan
corpora (48 speakers and 1599 tokens in one, 72 speakers and 2144 tokens in the other),
Bentivoglio and Sedano (2011:169-71) report that expressions of directionality are favorable
contexts for pa’ (me fui pa Nueva York ‘I went to New York’) while para is preferred for
purposives (para terminar ‘in order to finish’) and furthermore that following consonants promote
the abbreviated form (as in pa’ comprar ‘in order to buy’), but that the strongest effect is that of
socioeconomic level: the “low level” showed the highest rates of the reduced variant pa’.
A class-based account is also proposed for a higher Spanish subject pronoun rate among
Colombians and possibly Cubans who have lived in New York City for more than five years as
compared with newcomers from those same countries. Shin and Otheguy (2013:442-43) point to
the high affluence rankings of these Latino national-origin groups in Census data, offering the
conjecture that affluent Latinos are susceptible to influence from English due to looser social
networks and more interaction with speakers of English. However, among the Colombians and
Cubans sampled (N = 45), no effect is found for social class or education. Nevertheless, a “woman
effect” is reported, which is most pronounced among those who were Latin-American born but
had lived in NYC more than five years (Shin and Otheguy 2013:439). While the gender effect
among Colombians and Cubans may be because, as suggested by Shin and Otheguy (2013:446),
women have more contact with US-born children or friends than men do, it also appears to be the
case that women have higher rates of pronominal subject expression than men in Colombia to
begin with (Orozco 2015:30; see also Martín Butragueño and Lastra 2015:50).
Understanding variation in Spanish among Latino New Yorkers necessitates knowledge of their
social context (Otheguy and Zentella 2012:149-50), in the same way that it is imperative to
distinguish language contact settings due to immigration, as a result of conquest, or across national
boundaries (Guadalupe Valdés 1982; cf. Poplack and Levey 2010: 396-97). The hypothesis of

To appear in: Potowski, K. (ed.) Handbook of Spanish as a Heritage/Minority Language, Routledge.
(Final version, pre-typesetting)
!
5
susceptibility to English or other-dialect influence would necessarily be tested by measures of
degree of contact. Such measures have relied on self-reports, for example, speakers have been
divided into in-group vs. out-group orientation groups based on reported frequency of interactions
with speakers from other dialect regions (Otheguy and Zentella 2012:109-12). Metrics for degree
of contact with English may be derived from demographic data; in particular, the proportion of
Spanish vs. English speakers in neighborhoods of residence (see Poplack and Levey 2010:399,
402). A parallel measure could be applied to gauge participants’ level of interaction with speakers
of other dialects, as has been done for contact between Salvadorans and Mexicans in Houston
(Hernández 2009:598-600; see below). Direct measures of contact are best developed from
sociolinguistic profiles culled from content analysis of recorded conversations constituting a
corpus; for example, concerning time and location of acquisition of English, preferred or “most
comfortable” language, language choice according to interlocutor, and general affect toward the
bilingual situation (Poplack, Walker and Malcolmson 2006:196-207; Torres Cacoullos and Travis
2018: Chapter 4).
Poplack’s (1979) dissertation with Puerto Ricans living in Philadelphia remains a model study that
has yet to be repeated in a U.S. Spanish community. Based on 24 sociolinguistic interviews
collected from a neighborhood block in Philadelphia over a period of one year (Poplack 1979:28-
37), it was one of the first to apply rigorous statistical analysis (logistic regression and principal
component analysis) to data on linguistic variation and show the effects of social predictors. Of
the 24 participants, 15 were female, and most were working class or unemployed and had limited
formal education and social mobility (Poplack 1979:38-43). Poplack tested gender, age, education,
language proficiency, and geographic origin (1979:48-50) as they conditioned lenition of coda /s/
(N = 19,284), /n/ (N = 8,648), and /r/ (N = 7,142) (1979:64, 108,143)2. While there was very
limited social conditioning of /n/ lenition (cf. Poplack 1979:123,127) for monomorphemic /r/ she
found an increased lenition rate for males (Poplack 1979:165), while for infinitival /r/ she found a
slight effect of education (Poplack 1979:172). With respect to coda /s/ lenition, social factors were
particularly important. For plural /s/, only geographic origin and language proficiency were
selected by the model (Poplack 1979:86). For monomorphemic and verbal coda /s/, however,
Poplack found that each of the five social factors tested (age, speech style, education, geographic
origin, and language proficiency) conditioned lenition in word-final position (1979:75,96). It is
important to note that the speakers studied here were from an immigrant population with a
relatively short history (less than fifty years) in Philadelphia, which raises the question of the role
of social factors given varying degrees of community stability and geographic permanence.
3. CHANGES IN PROGRESS
Linguistic change in Spanish in the U.S. is often proclaimed, though not as often demonstrated.
Making a reasonable case for language change requires, above all, a robust quantitative pattern,
which is verified in the speech of a community-based sample of speakers selected in a principled
manner (Poplack et al. 2012). Changes in progress can be detected synchronically in apparent
time—the distribution of variant forms across age cohorts (Labov 1994:43-72).
For example, in her pioneering sociolinguistic study of Panama City in 1969-71, Henrietta
Cedergren (1973) observed a process of deaffrication from [tʃ] to the fricative [ʃ] in apparent time,
with!an!inverse!relationship!between!lenition! and! age. The lenited variant increased as age

To appear in: Potowski, K. (ed.) Handbook of Spanish as a Heritage/Minority Language, Routledge.
(Final version, pre-typesetting)
!
6
decreased, peaking in the second youngest age group of 27-32 year olds and slightly declining in
the 15-26 group. Cedergren obtained data using the same sampling procedure in 1983 to see
whether real-time evidence would suggest a genuine change in progress or age-grading—that is,
change with age that repeats in each generation and results in stable community behavior in
aggregate. In fact, her comparison revealed age-grading—the same pattern was followed across
apparent time at each point in real time—but with lenition incrementally higher for all but the two
youngest groups; this is interpreted to mean that [tʃ] lenition in this community had peaked (Labov
1994: 94-97).
Deaffrication of [tʃ] showed a correlation with age in Tomé, in the Río Abajo region of New
Mexico, just south of Albuquerque. Excluding postnasal and postlateral cases (planchar ‘to iron’,
el chile ‘the chili pepper’), which are categorically realized as affricates, Jaramillo and Bills (1982)
give an apparent time interpretation to the distribution of the variants across age groups in a sample
of 36 speakers (N = 1,029). They find a shift from the fricative variant to the affricate, as the rate
of [ʃ] is nearly halved in the youngest (17-30) age group compared to approximately 80% in the
older groups. The interpretation of a shift toward the more standard pronunciation is supported by
considering the effect of education, operationalized as years of formal instruction. Since age and
educational attainment partially overlap (a greater proportion of younger than older people had
college education), Jaramillo and Bills (1982:161) cross-tabulated age and education, to show an
independent effect for education. In fact, within the young group, 8 speakers with a college
education had a lenition rate approximately four times lower than that of the other 4 young
speakers. Speakers with more than two years of formal study of Spanish also tended to lenite less
often. The researchers conclude that the “perceived change” away from the “long-established”
fricative variant “appears to simply reflect a sociological change related to education” whereby
some residents are “expanding their command of different varieties of Spanish” (Jaramillo and
Bills 1982:163-164).
Change in progress may be inferred from comparison of variation patterns in communities of
origin. In Salvadoran communities, sequences of front vowels in hiatus with other vowels (as in
vea ‘he/she/you(formal) sees (Subj)’) alternate with a hiatus-breaking [ʝ] variant (veya) (Lipski
1994:258). Hernández (2015) compared rates and conditioning of the hiatus-breaking [ʝ] variant
for the immigrants in Houston to comparable data from San Sebastián, El Salvador, the
municipality of origin for most families. He reports that the rate of hiatus-breaking [ʝ] in Houston
(6%, N = 737) is less than a third of that in San Sebastián (20%, N = 811), receding to 2% (N =
288) in the second immigrant generation. While in San Sebastián the hiatus-breaking variant was
favored by older speakers and disfavored by women and those with a secondary school education,
in Houston, with the now overwhelming preference for the hiatus variant, none of the social factors
investigated—education level, gender and age of the speaker—make a statistically significant
contribution.
Contrariwise, speaker gender does appear to contribute to linguistic variation in Kennett Square,
Pennsylvania, though in a diminished way. Matus-Mendoza (2004) analyzed variable assibilation
of word-final /r/ to a voiceless retroflex sibilant (/ɾ/→ [ř] \ __#; e.g., deci[ř] ‘to say, tell’) in a
corpus of 83 sociolinguistic interviews with speakers in Moroleón, Guanajuato and Kennett
Square, where many mushroom industry workers are from Moroleón. The linguistic conditioning
of assibilated /r/ was the same in Moroleón (N = 2,796; Matus Mendoza 2004:21) and Kennett

To appear in: Potowski, K. (ed.) Handbook of Spanish as a Heritage/Minority Language, Routledge.
(Final version, pre-typesetting)
!
7
Square (unknown N). As for extralinguistic factors, rates in Moroleón differed across locales, with
more frequent assibilation in urban than in rural areas, and across genders, with women assibilating
more than men (Matus-Mendoza 2004:20-22). Differences according to occupation and education
level also indicate that assibilation is a prestige variant in Mexico (Matus-Mendoza 2004:26-27).
In Kennett Square, the rates of assibilation increase with more schooling and among women, but
the percentages are “extremely low…compared to…Moroleón” (6% among women in Kennett
Square vs. 24% in Moroleón), suggesting an “equalizing situation” in the shared working
environment (Matus-Mendoza 2004:27).
Contraction of a phonetic variant has also occurred in Houston, where Salvadorans live and work
alongside Mexicans. Composition of neighborhoods of residence provides one measure to
approximate degree of dialect contact. In Houston’s Segundo Barrio, Hispanics make up 90% of
the population and the ratio of Salvadorans to Mexicans is on the order of one-to-ten, while in
Holly Spring, where Hispanics constitute just 12% of the population, it is closer to one-to-two
(Hernández 2011:55). Hernández (2011) capitalizes on this difference to compare variable word-
final nasal velarization (/n/→[ŋ] \ __#, as in los pueblos fueron [ˈfwe.ɾoŋ] los que sufrieron
[su.ˈfɾje.ɾoŋ] más ‘the towns were the ones that suffered the most’) in three Salvadoran
communities. The rate of nasal velarization declines in Houston compared with San Sebastián, El
Salvador, the community of origin (23%, N = 430), but more so in Segundo Barrio (3%, N = 476)
than in Holly Spring (14%, N = 981) (Hernández 2011:66). On this basis, Hernández proposes that
differences between the two Houston communities are explained by amount of exposure to
speakers of Mexican Spanish (cf. Trudgill 1986:39). One scenario of possible change in progress
in a U.S. Spanish immigrant community, then, is dropping an alternation that constitutes a
linguistic variable in the community of origin (see Weinreich [1968:18-19], and Erker [this
volume] on Spanish dialectal contact in the U.S.).
A contrary development may be the spread of a new linguistic variable. This appears to be the case
with intervocalic ⟨ll⟩ ([ʝ]) deletion (e.g., iba a ir el bos por eØa a Brownsville ‘the bus was going
to go to Brownsville for her’ vs. estudiar a Matamoros con ella ‘to study in Matamoros with her’)
in the Segundo Barrio and Holly Spring neighborhoods in Houston. Hernández (2015) compared
speech data from sociolinguistic interviews conducted among Salvadoran and Mexican
immigrants. The participants were first generation, second generation or (in the case of the
Mexican speakers) third generation, and most were from families from San Sebastián, El Salvador,
or Matamoros, Mexico. The rate of intervocalic ⟨ll⟩ deletion in Houston is twice as high in the
second and third than in the first generation among Mexicans (N = 383) and three times as high in
the second generation than in the first among Salvadorans (N = 622) (Hernández 2015). This
means that second generation Mexican and Salvadorans show a closer elision rate (31% and 23%,
respectively) than do their first-generation counterparts (17% and 5%, respectively). Though not
significant in multivariate analysis, there appears to be a tendency for higher elision rates among
men than women, in both national origin groups, for this expanding phonological variable.
As indicated by the studies surveyed in this chapter, linguistic and social categories are linked, yet
social factors remain understudied — particularly socio-economic status. Common belief holds
that linguistic patterns in U.S. Spanish are unaffected by speakers’ socioeconomic status (Bills and
Vigil 2008:250). Some researchers even assert that speakers’ occupation or education should not
be expected to correlate with the language of minority communities, since Spanish is not

To appear in: Potowski, K. (ed.) Handbook of Spanish as a Heritage/Minority Language, Routledge.
(Final version, pre-typesetting)
!
8
instrumental for success in the employment market (e.g., Garcia and Tallon 2000:358, n.1).
Contributing to the lack of studies of social factors is the problem of grouping speakers according
to sociological characteristics. This is at least no less exacting in minority-language situations than
elsewhere, as the appropriateness of the criteria must be independently established for each
community. For example, a solution for immigrant communities is offered by Orozco (2007:105),
who classified New York City Colombians into three groups by taking into consideration their
occupations both in NYC and in Colombia: those who retained white-collar jobs, blue-collar
workers before and after immigration, and blue-collar workers in NYC who held white-collar
positions in Colombia.
But a remaining problem in general is that social categories, unlike linguistic categories, have no
standard or agreed-upon methods of demarcation. An additional obstacle is that social groupings
often correlate with one another, and as such it is disadvantageous to include them in omnibus in
a statistical model. We now illustrate an alternative approach which can circumvent this problem
by grouping speakers based on of their linguistic behavior to infer social grouping.
4. PREDICTING SOCIAL VARIATION WITH LINGUISTIC BEHAVIOR:
CLUSTERING AND STRATIFICATION IN NEW MEXICAN SPANISH
Hints of the social conditioning of variable usage can already be discerned in the earliest linguistic
study of Spanish in the U.S. Over a century ago, Espinosa (1911:10) in The Spanish language in
New Mexico and Southern Colorado suggests the social evaluation of the aspirated variant of /s/,
qualifying it as “widespread among the rural uneducated classes”. Dating back to 16th-17th
century settlement from New Spain (today, Mexico), Northern New Mexico is home to
(Traditional) New Mexican Spanish. As Lipski (2000:2-4) has noted, New Mexican Spanish was
deemed by Espinosa and contemporary linguists in Latin America and Spain to be nothing less
than another national variety of the language.
In New Mexico, it is the speakers of English, not Spanish, who are (descendants of) immigrants.
In 1850, the area became a U.S. territory, and in 1878, the railroad arrived along with accelerating
Anglo-American immigration.!In!1912, New Mexico was admitted to the Union as the 47th state
and English increasingly displaced Spanish in schools—even in northern, longstanding Spanish-
speaking communities—by the 1940s. Today Spanish is taught as a foreign language and, while
Hispanics represent as much as 80% of the population in some northern counties, there is a
continued shift toward English (Bills and Vigil 2008 inter alia). The remaining speakers of
Traditional New Mexican Spanish provide an invaluable window into Spanish language use in a
native community.
The New Mexico Spanish-English Bilingual (NMSEB) corpus consists of spontaneous speech
collected by in-group community members and thoroughly transcribed in prosodic units.
Participants were selected to cover a range of demographic backgrounds to permit the assessment
of extra-linguistic constraints on linguistic variation (Travis and Torres Cacoullos 2013; Torres
Cacoullos and Travis 2018 Chapter 2).
To identify those social factors that may contribute a consistent effect on linguistic variation in the
data, we cast a wide net by looking at the problem in reverse (Horvath and Sankoff 1987; Poplack

To appear in: Potowski, K. (ed.) Handbook of Spanish as a Heritage/Minority Language, Routledge.
(Final version, pre-typesetting)
!
9
1979:190-223). We will use the linguistic behavior of speakers in NMSEB to cluster them via a
principal component analysis (PCA), and then interpret the resultant configurations in terms of our
extralinguistic knowledge of the speakers to identify the social characteristics that individuals
within those clusters have in common. PCA is a data optimization method used to partition a
multidimensional space into several orthogonal components that reduce the dimensionality of that
space; the dimensions that contribute toward partitioning the variance of that space are called the
principal components. A PCA works best when there is a high amount of variance in the space;
typically, this is found when each row vector (in this case, speaker) has more than 10 numeric
variables, or dimensions (in this case, linguistic features) (cf. Horvath and Sankoff 1987:186).
We focus on four phonetic variables, which yield 14 such dimensions: onset (syllable-initial) /s/
lenition, coda (syllable-final) /s/ lenition, intervocalic /d/ elision, and intervocalic ⟨ll⟩ lenition.
Each of these has been studied and implicated as either characteristic of New Mexican Spanish or
as a stable, socially-stratified variable in other dialects of Spanish (e.g., Espinosa 1909:72,75;
Gutiérrez 1981; Lapesa 1968:354, 356; Lipski 2011:75-83; Samper Padilla 2011:105-14). The
dimensions for the PCA were based on the linguistic constraints for each of these four phonetic
variables:
Onset /s/ lenition: Favored and other contexts. Since onset /s/ lenition was most strongly favored
by preceding non-high vowels—as in ese ‘that one’ or la señora ‘the woman’ (cf. Brown 2005a)—
counts of onset /s/ in favorable phonetic contexts (preceding non-high vowels) were separated
from counts of onset /s/ in other phonetic environments. Additionally, since complete elision of
onset /s/ was rare, tokens were divided into full ([s]) and lenited variants and counts were included
separately. Doing so produced counts of four variants: full onset /s/ preceded by a non-high vowel,
lenited onset /s/ preceded by a non-high vowel, full onset /s/ in other contexts, and lenited onset
/s/ in other contexts.
Coda /s/: Favored and other contexts. Lenition was most strongly favored when the following
phone was a voiced consonant, as in desde ‘since/from’ or los viejitos ‘the old people’ (cf. Brown
2005b). Unlike onset /s/, coda /s/ showed a mix of full ([s]), aspirated ([h]), and elided (Ø) variants,
so we considered counts of each separately. This produced six additional variants per speaker (each
of the three variants followed by a voiced consonant and in other environments).
Intervocalic /d/ elision: We take counts of intervocalic approximants against the number of elided
intervocalic tokens (in which there was no perceptible frication as well as audible vowel
coarticulation, e.g., casado [kasau]; casada [kasa=ː] ‘married’).
⟨
ll
⟩
lenition: In words like ellos ‘they/them’, reduced and completely elided forms were grouped
together.
In all, there were 14 variants across the four phonetic variables. Since onset and coda /s/ were
further subdivided by phonetic environment, this produced six categories: onset /s/ in a favorable
environment for lenition, onset /s/ in other environments, coda /s/ in a favorable environment for
lenition, coda /s/ in other environments, intervocalic /d/, and intervocalic ⟨ll⟩. If a participant had
fewer than 20 tokens in total for any of these categories (summing up all variants within those
categories), their counts for all variants in that category were zeroed out to keep low token counts

To appear in: Potowski, K. (ed.) Handbook of Spanish as a Heritage/Minority Language, Routledge.
(Final version, pre-typesetting)
!
10
from warping the PCA output. The principal components resulting from the PCA3 were then
plotted based on the amount of variance each principal component accounted for. Three principal
components accounted for 78% of the total variance in the dataset. We then examined the
associations of each of the 14 variants with each of these three principal components. Many of the
variants showed moderate associations, or loadings (with magnitude greater than 0.3; |PCx| >0.3)
(cf. Horvath and Sankoff 1987:194), indicated by bolded text and cell shading in Table 1; variants
with weaker associations (0.25 ≤ |PCx| < 0.3) are listed in bold without shading.
We interpret the loadings as follows. Principal Component 1 (PC1), which accounts for 46% of
the variance, appears to represent lenition in general. That is, speakers who have (negative)
associations with this component are more likely than other speakers to aspirate onset /s/ and will
also tend to lenite (aspirate) coda /s/ and to use lenited intervocalic /ʝ/ as well. Principal Component
2 (PC2), accounting for another 21% of the variance, is largely the complement of PC1. Here
retention of full variants is (positively) associated with PC24. Principal Component 3, which
accounts for 11% of the total variance, is more complicated. Both onset /s/ retention in other than
preceding non-high-vowel environments (i.e., in disfavorable contexts for /s/ aspiration) and coda
/s/ lenition, especially ø, pattern in the same direction (negatively), and these are contrasted with
intervocalic ⟨ll⟩#lenition and, though its association is marginal, onset /s/ aspiration in preceding
non-high-vowel contexts. This component, then, groups more standard and general Spanish
linguistic patterns, namely onset /s/ retention and coda /s/ lenition, in opposition to traditional New
Mexican variants, that is, intervocalic ⟨ll⟩#lenition and onset /s/ aspiration.
Effectively, the Principal Component Analysis has taken a 14-dimensional space representing each
variant of our four variables and reduced it to a three-dimensional space where highly correlated
items pattern together. This permits a spatial representation of the data, which can elucidate
similarities in speaker behavior, but it also, crucially, illustrates the associations of the phonetic
variants to one another. Through such an analysis, we apprehend that participant groupings are
strongly determined by patterns of lenition.

To appear in: Potowski, K. (ed.) Handbook of Spanish as a Heritage/Minority Language, Routledge.
(Final version, pre-typesetting)
!
11
Table 1: Loadings of 14 Consonantal Variants
in New Mexico (NMSEB) on Principal Components
!
Variable
Variant (Dimension)
PC1
PC2
PC3
Onset /s/
[s].Preceding NonHighV
0.24
0.40
-0.11
[h]/ø.Preceding NonHighV
-0.33
-0.11
-0.01
[s].Other environments
0.13
0.38
-0.45
[h]/ø.Other environments
-0.31
-0.03
0.21
Coda /s/
[s].Following Voiced C
0.20
0.34
-0.15
[h].Following Voiced C
-0.30
0.13
-0.33
ø. Following Voiced C
-0.28
0.09
-0.40
[s].Other environments
0.28
0.30
-0.04
[h].Other environments
-0.34
0.05
-0.26
ø .Other environments
-0.28
-0.09
-0.36
Intervocalic
/d/
Approximant intervocalic /d/
-0.15
0.43
0.26
Elided intervocalic /d/ (ø)
-0.25
0.31
0.21
Intervocalic
⟨ll⟩
Full intervocalic /ʝ/
-0.23
0.34
0.22
Lenited intervocalic /ʝ/
-0.30
0.21
0.32
Variance accounted for:
46%
21%
11%
Interpretation:
Lenition
(general)
Retention
(general),
except /d/
NM Spanish
vs. other
dialects
With a general linguistic interpretation of the principal components in mind, we then ask how
individual speakers associate with each of the principal components. In Figure 1, by plotting each
participant according to their loading on the first two Principal Components, and letting the
shading indicate the third Principal Component, we capture the results of the PCA visually and use
those results to cluster participants. In doing so, we observe that the participants naturally fall into
three main groups, primarily delineated by PC1 (indicated by shape in Figure 1).
Using these speaker clusters based on linguistic behavior, we compared sociodemographic
characteristics of the speakers to assess what was shared among most members. Group 1 mostly
consists of miners, factory workers, or ranchers who are men with a middle or high school
education. Group 2 is mainly constituted by middle or high school educated men and women, some
in production (e.g., factory workers), and some in service (e.g., in dry cleaning) occupations.
Group 3 is a more urban, predominately female group, in which we find most of the participants
with (some) college education and/or professional occupations (e.g., teachers).

To appear in: Potowski, K. (ed.) Handbook of Spanish as a Heritage/Minority Language, Routledge.
(Final version, pre-typesetting)
!
12
Figure 1: Grouping of NMSEB Speakers by Linguistic Behavior (from PCA)
Based on these clusters, then, it appears that socioeconomic status (occupation and education),
gender, and rural vs. urban locale should be considered as candidates for conditioning linguistic
variation in NM Spanish. For a composite socioeconomic index based on occupation and
education, we grouped speakers into ‘production workers’ (N = 14), ‘service employees’ (N = 15)
and ‘professionals’ (N = 9). There were 22 women and 16 men. As to locale, ‘urban’ were those
participants from cities with 10,000 or more residents (Albuquerque, Española, Las Vegas, Los
Lunas, Santa Fe) (N = 11).
To obtain additional evidence that these social factors may be predictors of variation, we
determined whether they were distributed unevenly among the three speaker clusters, via Fisher’s
exact tests. According to these, gender (p < 0.05) was disproportionately distributed among groups,
with males being more common in Group 1 and females in Group 3. Subsequent Fisher’s exact
tests conducted pairwise indicated that occupation-education was also differentially distributed
across groups 1 and 3 (p < 0.05), suggesting that both social class and gender may be useful
categories for conditioning linguistic behavior. Additionally, there seems to be a slight bias toward
rural speakers in Group 1 (7 of 8 are rural), though this does not reach statistical significance.
Thus, we include Rural vs. Urban locale as a social predictor, understanding that this characteristic
may not be as robust as gender or socioeconomic status in distinguishing participants’ linguistic
behavior.
These three social predictors were considered together with linguistic factors in generalized linear
mixed models (conducted using the lme4 package [Bates et al. 2014] in R [R Core Team 2015]).5
A separate model was fit for each of the four phonetic variables, and this model was compared to
models with only linguistic and only social predictors (via likelihood ratio tests). While both onset
and coda /s/ lenition were primarily determined by linguistic factors (as also reported by Brown
2005a, 2005b), model fits for intervocalic /d/ elision and ⟨ll⟩ lenition improved with the inclusion
of a combination of linguistic and social predictors.

To appear in: Potowski, K. (ed.) Handbook of Spanish as a Heritage/Minority Language, Routledge.
(Final version, pre-typesetting)
!
13
Tables 2 and 3 show the results of generalized linear mixed models for intervocalic /d/ elision and
⟨ll⟩#lenition, respectively. The Intercept refers to the log-likelihood of a dependent variable at a
given reference level (reference levels are listed below each table). Levels of each predictor are
assigned a weighting (β-coefficient), or Estimate, with positive values indicating an increased
likelihood and negative values indicating a decreased likelihood for a given level (factor). For
example, in the case of intervocalic /d/ elision, the positive Estimate for a preceding non-high
vowel suggests that this phonetic context increases the likelihood of elision.6 Also indicated in the
model outputs is significance (determined by estimated p-values computed via a Wald test).
Table 2: Social and Linguistic Factors Conditioning
Intervocalic /d/ Elision in NMSEB (N = 3,447)*
Factor
Estimate
Std. Error
Sig.
N
% Elision
Intercept
-3.74
0.54
***
Preceding Non-high V
3.08
0.57
***
2788
16%
(Preceding High V)
659
3%
Participle
1.65
0.59
**
471
34%
(Not a participle)
2976
10%
Production vs. Other Occupations
-1.28
0.42
**
Service vs. Professional
-0.12
0.42
(Production Occupation)
1753
18%
(Service Occupation)
1132
10%
(Professional Occupation)
444
9%
Rural Locale
-0.09
0.29
2376
14%
(Urban Locale)
1065
12%
Male Gender
0.60
0.29
*
1633
17%
(Female)
1808
11%
Preceding Non-high V:Participle
1.52
0.78
*
320
48%
(Prec. Non-high V, Non-participle)
753
14%
(Prec. High V, Participle)
150
3%
(Prec. High V, Non-participle)
288
1%
*Generalized linear mixed model, lme4 package (Bates et al. 2014) in R (R Core Team 2015)
Random Effects (SD): Speaker (0.55); Word (2.01)
Reference Level: /d/ Present, Prec High V, Non-participle, Urban, Female
| *** p<0.001 | ** p<0.01 | * p≤0.05 |
Intervocalic /d/ deletion is strongly affected by social class. In agreement with reports on Latin
American varieties of Spanish including Panamanian (Cedergren 1973) and Venezuelan
(D’Introno and Sosa 1986), intervocalic /d/ elision is favored in working class speech. Also
replicating reported patterns, men elide intervocalic /d/ more often than women. The primary
factors conditioning intervocalic /d/ elision, however, are still linguistic. Phonetic context and
participle status work together to vastly increase lenition rates with participles from the first
conjugation (-ado) relative to non-participles when the preceding phone is a non-high vowel.
Intervocalic ⟨ll⟩ lenition is also conditioned by a combination of social and linguistic factors. We

To appear in: Potowski, K. (ed.) Handbook of Spanish as a Heritage/Minority Language, Routledge.
(Final version, pre-typesetting)
!
14
find that while the strongest predictors of ⟨ll⟩ lenition—phonetic context—are linguistic, there is
also an effect of occupation: speakers from production occupations lenite ⟨ll⟩ most often, followed
by speakers from service and professional occupations.
Table 3: Social and Linguistic Factors Conditioning
Intervocalic
⟨
ll
⟩
Lenition in NMSEB (N = 1,335)*
!
Factor
Estimate
Std. Error
Sig.
N
% Lenition
Intercept
-0.15
0.53
Preceding Front V
-0.04
0.86
600
74%
(Preceding Non-front V)
735
49%
Following Front V
0.26
0.56
26
46%
(Following Non-front V)
1309
61%
Asymmetry in Height
0.89
0.50
1222
59%
(Symmetry in Height)
113
74%
Production vs. Other Occupations
-0.8
0.35
*
Service vs. Professional
0.13
0.43
(Production Occupation)
784
67%
(Service Occupation)
365
54%
(Professional Occupation)
186
43%
Rural Locale
0.55
0.36
1000
63%
(Urban Locale)
335
51%
Male Gender
0.02
0.36
753
63%
(Female)
582
56%
Preceding Front V: Following Front V
-3.05
1.04
**
15
40%
(Prec. Front V; Foll. Non-front V)
585
74%
(Prec. Non-front V; Foll. Front V)
11
55%
(Prec Non-front V; Foll. Non-front V)
724
49%
*Generalized linear mixed model, lme4 package (Bates et al. 2014) in R (R Core Team 2015)
Random Effects (SD): Word (1.2), Speaker (0.76)
Reference Level: Full token, Preceding Non-Front V, Following Non-Front V,
Height Symmetry, Urban Locale, Female Gender
| *** p<0.001 | ** p<0.01 | * p≤0.05 |
5. CONCLUSION
Although social factors in U.S. Spanish have received inadequate attention to date, the few
available reports confirm the need to account for sociolinguistic variation. As we have seen, where
social factors have been tested, sociolinguistic patterns generally replicate those found across the
Spanish-speaking world, revealing the systematic character of varieties of Spanish in the US.
A contributing factor to the paucity of studies has been the familiar problem that social
characteristics of speakers are generally less well defined than linguistic categories, particularly in
minority language situations, and that social factors are often highly interdependent. A solution to
this conundrum can be found by using a data optimization method such as principal component

To appear in: Potowski, K. (ed.) Handbook of Spanish as a Heritage/Minority Language, Routledge.
(Final version, pre-typesetting)
!
15
analysis (PCA) as a heuristic for grouping speakers strictly based on their linguistic behavior, with
the groups thus defined then interpreted according to social characteristics. We have illustrated
one such analysis in a corpus of New Mexican Spanish. By applying PCA to counts of known
phonetic variables, we determined that occupation-education, gender, and demographic locale
were likely social factors of variation. This was confirmed via regression analysis for two phonetic
variables. For intervocalic /d/ and ⟨ll⟩, the highest lenition rates are found in speakers with
production occupations and, for /d/, among men. In each case, we observe social stratification
common not only to many other dialects of Spanish, but to many language varieties in general.
The study of Spanish in the US can advance with data from community-based speech corpora that
are constituted by participants of known sociodemographic characteristics sampled in an informed,
principled way. These are, effectively, the principal components of accountable sociolinguistic
research.
1 This work was made possible by funding from the National Science Foundation (Grant 1019112/1019122) to Rena
Torres Cacoullos and Catherine Travis.
2 Poplack grouped coda /n/ and /r/ by morphemic status, separating verbal /n/ and infinitival /r/ from monomorphemic
/n/ and /r/.
3 The PCA was conducted in R (R Core Team 2015) using the prcomp() function. The counts for each column were
scaled to account for different overall token counts for the distinct variables.
4 We note that both approximant and elided intervocalic /d/ correlate positively with this component with similar
magnitudes, which indicates that the component makes no distinction between the two. In fact, the variants of
intervocalic /d/ pattern similarly for each principal component, indicating that /d/ elision is not a phonetic variable
which contributes much meaningful variance for grouping our speakers.
5 Weighted effect coding was used to account for inevitable token imbalances in discourse data.
6 Because Occupation has three levels, we compared production workers to other workers as one contrast, and service
to professional occupations for the second contrast. Thus, weightings are reflective of how the second group behaves
with respect to the first. In /d/ elision, for example, the negative estimate for Production vs. Other indicates that
speakers from non-production occupations elide less than speakers in production occupations.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

To appear in: Potowski, K. (ed.) Handbook of Spanish as a Heritage/Minority Language, Routledge.
(Final version, pre-typesetting)
!
16
6. References
Bates, D., Maechler, M., Bolker, B., and Walker, S. (2014). lme4: Linear mixed-effects
models using Eigen and S4. R package version 1.1-5. http://CRAN.R-
project.org/package=lme4
Bentivoglio, P., and Sedano, M. (2011). Morphosyntactic variation in Spanish speaking
Latin America. In Díaz-Campos, M. (ed.), The Handbook of Hispanic
Sociolinguistics. Oxford: Blackwell. 168-186.
Bills, G.D. (1975). Linguistic research on United States Hispanics: State of the art. In
Teschner, R.V., Bills, G.D., and Craddock, J.R. (eds.), Spanish and English of
United States Hispanos: A critical, annotated, linguistic bibliography. Arlington,
VA: Center for Applied Linguistics.
Bills, G.D. and Vigil, N.A. (2008). The Spanish language of New Mexico and Southern
Colorado: A linguistic atlas. Albuquerque: University of New Mexico Press.
Brown, E.L. (2005a). Syllable-initial /s/ in Traditional New Mexican Spanish: Linguistic
factors favoring reduction ahina. Southwest Journal of Linguistics, 24(1-2), 13-
30.
Brown, E.L. (2005b). New Mexican Spanish: Insight into the variable reduction of “La
ehe inihial” (/s-/). Hispania 88(4), 813-824.
Cedergren, H. (1973). The interplay of social and linguistic factors in Panama. Doctoral
dissertation. Cornell University.
D’Introno, F., and Sosa, J.M. (1986). La elisión de la /d/ en el español de Caracas:
Aspectos sociolingüísticos e implicaciones teóricas. In Núñez Cedeño, R., Páez,
I., and Guitart, J. (eds.), Estudios sobre la fonología del español del Caribe. 135-
163. Caracas: La Casa de Bello.
Espinosa, A.M. (1909). Studies in New Mexican Spanish. Part I: Phonology. Doctoral
dissertation. University of Chicago.
Espinosa, A.M. (1911). The Spanish language in New Mexico and Southern Colorado.
Santa Fe, NM: Historical Society of New Mexico (No. 16).
Garcia, M.E., and Tallon, M. (2000). Estar in Mexican-American Spanish: Phonological
or morphological variability? In Roca, A. (ed.), Research on Spanish in the
United States: Linguistic issues and challenges. Somerville, MA: Cascadilla
Press. 348-359.
Gutiérrez, J. (1981). An analysis of the phoneme /s/ in New Mexico Spanish. In Elerick,
C. (ed.), Proceedings of the Ninth Annual Southwestern Areal Language and
Linguistics Workshop, El Paso: University of Texas at El Paso. 234-239.
Hernández, J.E. (2009). Measuring rates of word-final nasal velarization: The effect of
dialect contact on in-group and out-group exchanges, Journal of Sociolinguistics,
13(5), 583-612.
Hernández, J.E. (2011). Measuring rates and constraints of word-final nasal velarization
in dialect contact. In Ortiz-López, L.A. (ed.), Selected Proceedings of the 13th
Hispanic Linguistics Symposium.Somerville, MA: Cascadilla Proceedings Project.
54-69.
Hernández, J.E. (2015). Yod travels north: Measuring rates and constraints of two
phonological processes in dialect contact. Invited talk. Department of Spanish,
Italian and Portuguese, Penn State University. 26 January.

To appear in: Potowski, K. (ed.) Handbook of Spanish as a Heritage/Minority Language, Routledge.
(Final version, pre-typesetting)
!
17
Horvath, B., and Sankoff, D. (1987). Delimiting the Sydney speech community.
Language in Society, 16(2), 179-204.
Jaramillo, J.A., and Bills, G.D. (1982). The phoneme /ch/ in the Spanish of Tomé, New
Mexico. In Barkin, F., Brandt, E.A., and Ornstein-Galicia, J. (eds.), Bilingualism
and Language Contact: Spanish, English, and Native American Languages. New
York: Teachers College, Columbia University. 154-165.
Labov, W. (1972). Some principles of linguistic methodology. Language in Society 1(1),
97-121.
Labov, W. (1994). Principles of linguistic change, vol. 1: Internal factors. Oxford:
Blackwell.
Labov, W. (2001). Principles of linguistic change, vol. 2: Social factors. Malden, MA:
Blackwell.
Labov, W. (2006). The social stratification of English in New York City. 2nd Ed.
Cambridge: Cambridge University Press.
Lantolf, J.P. (1982). Speaker sex and para reduction in Chicano Spanish. In Barkin, F.,
Brand, E.A., and Ornstein-Galicia, J. (eds.), Bilingualism and Language Contact:
Spanish, English, and Native American Languages. New York: Teachers College,
Columbia University. 166-176.
Lapesa, R. (1942/1968). Historia de la lengua española. 7a edición. Madrid: Escelicer,
S.A.
Lipski, J.M. (1994). Latin American Spanish. London: Longman.
Lipski, J.M. (2000). Back to zero or ahead to 2001? Issues and challenges in U.S.
Spanish research. In: Roca, A. (ed.), Research on Spanish in the United States:
Linguistic issues and challenges. Somerville, MA: Cascadilla Press. 1-41.
Lipski, J.M. (2011). Socio-phonological variation in Latin American Spanish. In Díaz-
Campos, M. (ed.), The handbook of Hispanic sociolinguistics. Wiley-Blackwell.
72-97.
Martín Butragueño, P., and Lastra, Y. (2015). Subject pronoun expression in oral Mexical
Spanish. In Carvalho, A.M., Orozco, R., and Shin, N. L. (eds.), Subject pronoun
expression in Spanish: A cross-dialectal perspective. Washington, D.C.:
Georgetown University Press. 39-57.
Matus-Mendoza, M. (2004). Assibilation of /-r/ and migration among Mexicans.
Language Variation and Change, 16, 17-30.
Ocampo, F. (1990). El subjuntivo en tres generaciones de hablantes bilingües. In Bergen,
J. (ed.), Spanish in the United States: Sociolinguistic issues. Washington:
Georgetown University Press. 39-45.
Orozco, R. (2007). Social constraints on the expression of futurity in Spanish-Speaking
urban communities. In Holmquist, J., Lorenzino, A., and Sayahi, L. (eds.),
Selected proceedings of the Third Workshop on Spanish Sociolinguistics.
Somerville, MA: Cascadilla Proceedings Project. 103-112.
Orozco, R. (2015). Pronominal variation in Colombian Costeño Spanish. In Carvalho,
A.M., Orozco, R., and Shin, N.L. (eds.), Subject pronoun expression in Spanish:
A cross-dialectal perspective. Washington, D.C.: Georgetown University Press.
17-37.
Otheguy, R., and Zentella, A.C. (2012). Spanish in New York: Language contact,
dialectal leveling, and structural continuity. New York: Oxford University Press.

To appear in: Potowski, K. (ed.) Handbook of Spanish as a Heritage/Minority Language, Routledge.
(Final version, pre-typesetting)
!
18
Peñalosa, F. (1981). Some issues in Chicano sociolinguistics. In Duran, Richard P. (ed.),
Latino language and communicative behavior. Norwood, N.J.: ABLEX, 3-18.
Poplack, S. (1979). Function and process in a variable phonology. Doctoral dissertation.
University of Pennsylvania.
Poplack, S. (1997). The sociolinguistic dynamics of apparent convergence. In Guy, G.R.,
Feagin, C., Schiffrin, D., and Baugh, J. (eds.) Towards a social science of
language. Papers in honor of William Labov: Social interaction and discourse
structures, Volume 2. Amsterdam: John Benjamins. 285-309.
Poplack, S., and Levey, S. (2010). Contact-induced grammatical change. In Auer, P., and
Schmidt, J.E. (eds.), Language and space—An international dandbook of
linguistic variation: Volume 1—Theories and methods. Berlin: Mouton de
Gruyter. 391-419.
Poplack, S., Lealess, A., and Dion, N. (2013). The evolving grammar of the French
subjunctive. Probus, 25, 139-195.
Poplack, S., Walker, J.A., and Malcolmson, R. (2006). An English "like no other"?:
Language contact and change in Quebec. Canadian Journal of Linguistics / Revue
canadienne de Linguistique, 51(2),185-213.
Poplack, S., Zentz, L., and Dion, N. (2012). What counts as (contact-induced) change.
Bilingualism: Language and Cognition, 15(2), 247-254.
R Core Team. (2015). R: A language and environment for statistical computing. R
Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-
project.org/.
Samper-Padilla, J.A. (2011). Sociophonological variation and change in Spain. In Díaz-
Campos, M. (ed.), The handbook of Hispanic sociolinguistics. Wiley-Blackwell.
98-120.
Shin, N.L., and Otheguy, R. (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, C. (1994). Language contact and change: Spanish in Los Angeles.
Oxford: Clarendon.
Torres Cacoullos, R., and Travis, C.E. (2018). Bilingualism in the community: Code-
switching and grammars in contact. Cambridge: Cambridge University Press.
Travis, C.E., and Torres Cacoullos, R. (2013). Making voices count: Corpus
compilation in bilingual communities. Australian Journal of Linguistics, 33(2),
170-194.
Trudgill, P. (1986). Dialects in contact. Oxford: Blackwell.
Valdés, G. (1982). Bilingualism in a Mexican border city: A research agenda. In Barkin,
F., Brandt, E.A., and Ornstein-Galicia, J. (eds.), Bilingualism and Language
Contact: Spanish, English, and Native American Languages. New York: Teachers
College, Columbia University. 3-17.
de Valdés, J. (1535/2010). Diálogo de la lengua. Biblioteca Virtual Miguel de Cervantes,
online edition based on José F. Montesinos (Madrid, Espasa-Calpe, 1976).
Weinreich, U. (1968). Languages in contact: Findings and problems. The Hague:
Mouton.

To appear in: Potowski, K. (ed.) Handbook of Spanish as a Heritage/Minority Language, Routledge.
(Final version, pre-typesetting)
!
19
Weinreich, U., Labov, W., and Herzog, M. (1968). Empirical foundations for a theory of
language change. In Lehmann, W. P., and Malkiel, Y. (eds.), Directions for
historical linguistics. Austin: University of Texas Press. 95–188.
Zentella, A. C. (1990). El impacto de la realidad socio-económica en las comunidades
hispanoparlantes de los Estados Unidos: Reto a la teoría y metodología
lingüística. In Bergen, J. (ed.), Spanish in the United States: Sociolinguistic
issues. Washington: Georgetown University Press. 152-166.