A Minimalist approach to gender agreement

in the Afro-Bolivian DP: Variation and the

specification of uninterpretable features
Sandro Sessarego & Javier Gutirrez-Rexach

University of Wisconsin, Madison and The Ohio State University

This article analyzes gender-agreement variation in the Determiner Phrase (DP)
of Afro-Bolivian Spanish (ABS). We claim that this is an instance of crossgenerational change, consisting in the systematic substitution of stigmatized
basilectal Afro-Bolivian features with more prestigious Bolivian Spanish ones. In
light of recent minimalist models, the variability encountered in the corpus can
be accounted for systematically as a by-product of the differential specification
of uninterpretable features in a derivation. This study highlights an on-going
process of post-bozal Spanish approximation to a more prestigious Spanish
variety. Such transition seems to be driven by social factors and is significantly
regulated by syntactic constraints.
Keywords: Afro-Bolivian Spanish, gender agreement, determiner phrase,
variation, feature specification, minimalism

1.Introduction
This article evaluates variation in the gender-agreement system of AfroBolivian Spanish, an Afro-Hispanic vernacular developed from what
was once a bozal language spoken in Los Yungas, Department of La Paz,
Bolivia. We present data showing evidence of cross-generational change,
consisting of the systematic substitution of stigmatized basilectal AfroBolivian features with more prestigious Bolivian Spanish ones. One of the
outcomes of this situation is the transition from one gender-agreement
system to another one. The purpose of this work is to shed light on the
Folia Linguistica 45/2 (2011), 465488.
issn 01654004, e-issn 16147308 Mouton de Gruyter Societas Linguistica Europaea
doi10.1515/flin.2011.017

466 Sandro Sessarego & Javier Gutirrez-Rexach

linguistic constraints regulating gender agreement within a theoretical

framework capable of accounting for the variability encountered in this
specific transition.
African slavery persisted in Bolivia from the very beginning of its colonization, in the first decades of the sixteenth century, until 1826, when in
the new constitution, immediately after independence from Spain, slaves
were declared free. Slavery was formally reestablished in 1830, but it was
abolished again in 1831. Nevertheless, in practice, until the Land Reform
of 1952, it represented a common practice, so that Afro-Bolivians had no
other choice but to work for their owners. The date of the introduction of
forced African labor to Los Yungas is not clear. The first records suggesting
a significant number of slaves in the area date from the eighteenth century.
They are documents concerning the inventory of goods belonging to some
plantations, or sale certificates that attested the passage from one owner
to another (Portugal Ortiz 1977: 7783). Since that period until 1952, when
the Land Reform took place, Afro-Bolivians have been employed in Los
Yungas as slaves in haciendas. After the Land Reform, a majority remained
in the region, becoming the new owners of small parcels from the land
that once belonged to plantations. Even though Afro-Yungueos are nowadays free people and their life conditions have improved during the last
60 years, their situation is far from being optimal, as the daily per-capita
income, mostly derived from growing and selling coffee and coca, does not
even reach $4 (Lipski 2005).
All contemporary Afro-Yungueos speak a variety of Highland Bolivian Spanish (HBS), which may be more or less influenced by the traditional
Afro-Bolivian dialect according to the speakers level of education and
mobility outside of Yungas. Typically, until 1952 black workers were not
allowed to attend school. Several older members of these communities are
therefore illiterate. However, after that date, the hacienda system ended and
basic public education began to develop in Afro-Yungueo communities.
The study of Spanish at schools resulted in a gradual drop of the traditional
dialect. This process has been described as a systematic substitution of stigmatized basilectal ABS features with more prestigious HBS ones (Lipski
2006a). As far as ABS gender marking is concerned, this substitution is not
random. Rather, what can be observed is the transition from one agreement system to another one, according to specific syntactic constrains.
The focus of the present study is on the linguistic implications of this
transition. The remainder of the article is organized as follows: section 2

Gender agreement in the Afro-Bolivian DP 467

is a brief illustration of ABS DP features; section3 elaborates on certain

methodological issues; section4 describes the attested gender-agreement
patterns; sections 5 and 6 provide a formal analysis of the phenomena; section7 presents the statistical results; finally, section8 interprets the structural repercussions of our findings, and section9 contains our conclusions.

2. Afro-Bolivian Spanish DP features

Lipski (2006b) describes five quintessential features of the traditional AfroYungueo DP which distinguish this dialect from other Spanish varieties:
(a) lack of noun-adjective gender agreement; (b) invariant plurals, that is,
no plural suffixes on nouns, adjectives, or determiners; (c) use of a single
invariant definite article; (d) elimination of definite articles in generic constructions; and (e), frequently, the retention of plural /s/ only on the first
element of plural NP. Instances of (a) and (b) are exemplified in (1).
(1) a. Siempre cont-aba algun-o-s cosa.
always tell-PST some-M-PL thing.F.SG
She always told some things.
b. es-o-s
fiesta
this-M-PL party.F.SG
these parties

Lipski notices that inflection is more likely to appear on determiners and

prenominal adjectives, while postnominal adjectives present a lower rate
of concord (2).
(2) a. con la
gente
antig-o
with the.F.SG people.F.SG old-M.SG
with the old people
b. es-a-s
casa
chic-o
this-F-PL house.F.SG small-M.SG
those small houses

Interestingly, no cases of postnominal agreement have been reported,

unless the prenominal elements agree too. With respect to (c), a robust
presence of the invariant plural definite article (lu) is attested in ABS.
(3) a. lu
taza
the.M.PL cup.F.SG
the cup

468 Sandro Sessarego & Javier Gutirrez-Rexach

b. lu
juama
the.M.PL family.F.SG
the families

Definite articles may also be eliminated in argument position (d) leading

to a generic reading, while such a reading requires the use of the article in
standard Spanish (4).
(4) a. Perr-o
ta floj-o.
dog-M.SG is worthless-M.SG
Dogs are worthless.
b. Patrn
huasqui-aba muj.
landowner.M.SG beat-PST
woman.F.SG
The landowners beat the women.

The last quintessential Afro-Bolivian DP feature reported is the frequent

retention of plural /s/ only on the first element of plural NP (e), as illustrated in (5).
(5) a. es-o-s
dirigente
this-M-SG leader.M.SG
these leaders
b. en idioma
antig-o di mi-s abuel-o
in language.M.SG old-M.SG of my-PL grandparent-M.SG
in the old language of my grandparents

For the purpose of this article, only tokens involving gender agreement across
the DP (a, c) are analyzed, leaving the study of number marking (b, e) and
bare nouns (d) for further research (Delicado-Cantero & Sessarego 2010).

3.Methodology
Formal syntactic theories have traditionally been built on partly-idealized
standard languages, on the basis of well-formedness judgements from a
limited set of informants. This methodology produced an impressive
amount of data, generalizations and insights, mostly because it abstracted
away from certain kinds of empirical complexities (Barbiers 2009: 1608).
Yet, for these same reasons, such a model has often been criticized by sociolinguists, who instead base their observations on bigger corpora of naturalistic production data, and have developed several techniques to study the

Gender agreement in the Afro-Bolivian DP 469

real vernacular, namely, the real language spoken by people when paying
no metalinguistic attention to their speech (Labov 1972).
Recent works on microvariation attempt to combine these previously
contrasting approaches to compare a speakers intuitions with real production data, with the goal of developing more fine-grained, empiricallytestable generalizations (Cornips & Poletto 2005). In collecting data for
microparametric analysis, it is therefore crucial to gather both grammaticality judgments as well as naturalistic data. For this reason, the informants who participated in the study were first interviewed and then asked to
answer grammaticality judgments from an oral questionnaire.
Atotal of 13 recorded interviews, lasting an hour or more each, were
conducted during July 2008 with Afro-Bolivian speakers residing in the
communities of Tocaa, Mururata and Chijchipa, North Yungas. The
informants were native speakers of the dialect who did not speak any other
language spoken in Bolivia, such as Quechua or Aymara. The interviews
were conducted by letting the informant talk about any topic of their liking
and asking them follow-up questions, in line with the principle of Tangential Shift (Labov 1984: 37).1 The goal was reducing the Observers Paradox
(Labov 1972) as much as possible. Only later, usually after one or two days
from the time of the interview, the same informant was asked for grammaticality judgments. This was done in order to not disrupt the outcome
of the interview by telling the speaker the nature of the phenomena under
analysis in advance.
Responses on acceptability-judgment tasks rely at least in part on explicit, prescriptive notions held by the speakers (Cornips & Poletto 2005).
One way of diminishing this effect, which proved successful according to
the experimental methods described in Labov (1984), is to ask for grammaticality judgments in an indirect way. Thus, to discover whether or not a
variable was present in the community, not only the direct intuitions were
elicited: Do you judge X a grammatical/better sentence than Y?; Can you
say X?; also indirect questions were asked: Is variant X present in this
community?; Do you know anybody who can say X?
1

According to the Principle of Tangential Shift, interviews may be arranged into a network
of topics which do not need to be followed according to a prescribed sequence. The conversation between the interviewer and the informant should start with the least personal
questions and progress, step by step, toward more intimate topics. The shift between topics
should be as smooth as possible. It should be based on follow-up questions to what has just
been said by the informant.

470 Sandro Sessarego & Javier Gutirrez-Rexach

The comparison of these two different sources of data resulted in the

interesting but not unexpected finding that almost everybody who
claimed not to say X, but to know people who could say it, were found
using X or something structurally similar several times during the naturalistic interview. This would indicate that such a structure was indeed part
of their grammar and confirmed the importance of relying on different
sources. While grammaticality judgments can give us a good insight into
the abstract idealized language of an informant, only a comparison of such
information with empirical data can help us build a robust fine-grained
generalization.

4. The grammar of gender agreement, a variable one?

With respect to what pertains to DP gender-agreement operations, grammaticality judgments, oral questionnaires and indirect questions indicated the presence of several possible patterns. Different informants indicated that in the most traditional ABS variety gender agreement occurs
only on singular definite articles, while the rest of the DP elements show
default-masculine concord, as shown in (6). These findings are in line
with the data presented by Lipski (2009) for the most traditional dialect.
However, the examples in (6) must not be considered an observed pattern but rather a reported one; in fact, none of our informants recognized
(6) as their own grammar; nevertheless, the Afro-Bolivians interviewed
indicated that some people in their community use these more traditional
constructions.
(6) a. tod-o la
comida delicios-o
all-M.SG the.F.SG food.F.SG delicious-M.SG
all the delicious food
b. tod-o lu
comida delicios-o
all-M.SG the.M.PL food.F.SG delicious-M.SG
all the delicious foods
c. est-e/es-e
comida delicios-o
this/that-M.SG food-F.SG delicious-M.SG
this/that delicious food
d. much-o/un comida delicios-o
much/a-M.SG food-F.SG delicious-M.SG
much/a delicious food

Gender agreement in the Afro-Bolivian DP 471

One subject (1/13) presented the grammatical intuitions exemplified in (7).

He showed gender agreement on plural and singular definite articles and
demonstratives, but not on other categories.
(7) a. tod-o la
comida delicios-o
all-M.SG the.F.SG food.F.SG delicious-M.SG
all the delicious food
b. tod-o las
comida delicios-o
all-M.SG the.F.PL food.F.SG delicious-M.SG
all the delicious foods
c. est-a/es-a
comida delicios-o
this/that-F.SG food.F.SG delicious-M.SG
this/that delicious food
d. much-o/un comida delicios-o
much/a.M.SG food.F.SG delicious-M.SG
much/a delicious food

The majority of the informants (7/13) used agreement on plural and singular definite articles, demonstratives, pre-nominal adjectives, and also on
weak quantifiers2 (8):
(8) much-a/un-a comida delicios-o
much/a-F.SG food-F.SG delicious-M.SG
much/a delicious food

The last group (5/13) claimed to use gender agreement for all the elements,
including strong quantifiers and post-nominal adjectives (9):
(9) tod-a la
comida delicios-a
all-F.SG the.F.SG food.F.SG delicious-F.SG
all that delicious food

This would intuitively lead us to argue in favor of four different grammars; however, a closer look at the empirical data from the oral interviews
complicates the picture. In fact, it was common for somebody to claim
to speak a certain grammar but use patterns belonging to another. Sometimes, speakers would freely alternate between forms within the same sen2

Since Milsarks dissertation (1974), natural-language quantifiers have been classified into
two different groups, called weak and strong quantifiers. Weak quantifiers (some, many,
etc.) can occur in existential (there) sentences. On the other hand, strong quantifiers (all,
most, etc.) cannot. Milsark suggests that this phenomenon would be due to the fact that
strong quantifiers refer to subsets of previously established sets, while weak quantifiers
establish such sets and for this reason can occur in existential constructions.

472 Sandro Sessarego & Javier Gutirrez-Rexach

tence. For example, the conversation fragment in (10) is from an informant

whose grammaticality judgments indicated a grammar of type (9):
(10) Tod-o la
comunidad
particip-aba; much-a gente
all-M.SG the.F.SG community.F.SG participate-PST much-F.SG people.F.SG
ven-a, much-o
gente
ven-a
desde lejos.
come-PST much-M.SG people.F.SG come-PST from far
Tod-a-s, todit-a-s la-s
persona-s se reun-an.
all-F-PL all-F-PL the.F-PL people-F-SG REFL meet-PST
Muy bonit-o la
fiesta
era...
very nice-M.SG the.F-SG party.F.SG be-PST
All the community used to participate, many people used to come, many
people used to come from far away. All, all the people gathered. The
party was very nice...

Speakers like this present gender agreement on strong quantifiers in 50

to 60 percent of instances, thus indicating that cases of agreement mismatches are very common and should not be regarded as just E-language
errors.

5. Variation and checking relations

The study of language variation seems at first to be at odds with formal
linguistic analyses that strive to hypothesize principles and generalizations
based on Universal Grammar. Nevertheless, in the Principles and Parameters paradigm, several attempts were made to capture dialectal and
inter-speaker variation, exploring the notion of parametric variation and,
in the last decade, fine-graining it to include so-called micro parameters
(Beninc 1989; Kayne 1996; etc). Language-internal variation, on the other
hand, has traditionally been excluded from formal linguistic analyses.
However, during the last years, several scholars have taken into account
variation beyond the usual parametric inter-language domain (Adger &
Smith 2005; Adger 2006). Individual variation becomes the core of linguistic research, bringing previously disregarded phenomena considered as
belonging to E-language to the fore (Adger & Trousdale 2007).
The advent of the Minimalist Program entailed a derivational approach
that was inconsistent with parameter-based accounts. Thus, an extension
of the parameter/micro-parameter idea to individual variation seems to
not be a straightforward matter. In the most recent formulations of the

Gender agreement in the Afro-Bolivian DP 473

Minimalist Program (Chomsky 2001, 2002, 2006), syntactic derivations

are viewed as strictly dependent on feature valuation and checking. The
distinction between interpretable and non-interpretable features, though
controversial, has proven very useful. Several features have an interpretation at LF, thus they are semantically-interpretable features. Other features, however, lack such semantic import and are present to trigger the
necessary merger or agreement operations during the derivation. Said
uninterpretable features have to be matched via Agree and are finally
deleted before Spell-out.
Given these theoretical assumptions, the Minimalist Program seems to
show enough promise to account for important aspects of variation because
it admits several phonological outputs for a given semantic interpretation.
Adger & Smith (2005) argue for characterizing syntactic variation in terms
of (un)interpretable features. Certain uninterpretable features may be present in one category but absent in another one. Since they are uninterpretable, they would have no semantic repercussion, thus being equally
legitimate for a convergent derivation. Therefore, variation is reduced to
the specification of the uninterpretable features in a derivation (Adger &
Smith 2005: 161). As expected, syntax per se remains invariable or perfect
(Brody 1997), given that variation is located only in the lexical component.
Variation will occur when one item or other enters the numeration and
takes part in a syntactic derivation. Such (variable) underspecification for
gender is what we observe when analyzing the ABS DP. Several (social)
factors may affect the outcome: ease of lexical access (probably linked to
frequency of use), speaker-hearer relationships, age, gender, social identity,
etc. (Adger & Smith 2005: 164). Due to the reduced number of participants,
we are not in the position to provide a detailed sociolinguistic analysis of
the phenomenon; nevertheless, we will comment on the effect of the factor age in patterning gender-agreement variation and cross-generational
agreement evolution.

6. Agreement and valuation

Taking the array of DP features described in section2 into account, ABS has
been classified by Lipski (2006c: 9) as a case of DP impoverished agreement. Lipski suggests that gender and number features in DP percolate
up from the noun to the determiner and eventually to the post-nominal

474 Sandro Sessarego & Javier Gutirrez-Rexach

e lement. He considers data from several speakers of different ages and levels of education presenting variable gender-agreement configurations, and
he notices that no case of post-nominal gender concord is found unless
pre-nominal elements agree, as shown in (11):
(11) a. un-a curva
anch-a
a-F.SG curve.F.SG large-F.SG
b. un-a curva
anch-o
a-F.SG curve.F.SG large-M.SG
c. un
curva
anch-o
a.M.SG curve.F.SG large-M.SG
d. *un
curva
anch-a
a.M.SG curve.F.SG large-F.SG
alarge curve

Our data, when analyzed cross-generationally, are generally in line with

those reported by Lipski. However, an important difference can be noticed
in the speech of several informants. Many speakers present almost a complete lack of gender agreement in strong quantifiers (Milsark 1974), as
shown in (12) and also in examples (6)(9):
(12) a. tod-o la-s
cosa
bonit-o
all-M.SG the.F-PL thing.F.SG nice-M.SG
all the nice houses
b. tod-o la
chic-a de Tocaa
all-M.SG the.F.SG girl-F.SG of Tocaa
all girls from Tocaa
c. tod-o la
comunidad
all-M.SG the-F.SG community.F.SG
all the community

Cases like these seem to violate the pre-nominal to post-nominal percolation order, unless we postulate that strong quantifiers are elements external to the DP, and then we argue independently in favor of a different
mechanism for the checking of the gender feature in languages where they
agree in gender and number with N, like Standard Spanish. Additionally,
the feature-percolation account of gender agreement runs into problems
when compared with data from other Romance varieties in which postnominal adjectives may agree with N and disagree with D (cf. Pomino &
Stark 2009 for Fassano Ladin). Alternatively, one might propose a system
with one or more agreement projections inside DP and with the relevant
displacement operations applied to agreeing elements so that they enter

Gender agreement in the Afro-Bolivian DP 475

into a specifier-head relation (Sportiche 2002, Koopman 1997). This type of

approach would be problematic too in that the number of internal agreement projections required would not be limited, and would most likely be
established on an ad-hoc basis. It also runs against recent minimalist ideas
supporting the elimination of Agreement projections (Chomsky 2002).
For these reasons, an account of gender agreement based on a minimalist model seems more adequate to describe the phenomena found ABS.
Within the minimalist framework, gender agreement is conceived as the
result of valuation processes which do not necessarily require movement
but just a configurational feature-checking mechanism (Picallo 2008).
Gender agreement, in fact, involves the transmission or sharing of features with nominal origin to other lexical items (adjectives) or to functional elements (determiners, quantifiers). Neither the demonstrative nor
the adjective comes from the lexicon with a value for gender. The gender
feature of determiners and adjectives is lexically unvalued (Chomsky 2001)
and becomes valued as a consequence of a syntactic process of agreement
with the gender feature of the noun (cf. Pesetsky & Torrego 2007).
Recent work on agreement operations advocates a version of agreement which departs from the previous view of this operation as a feature
assignment mechanism (Chomsky 2000). Rather, the process is seen as an
instance of feature sharing (Frampton & Gutmann 2000; Pesetsky & Torrego 2007), in line with the view of agreement as feature unification prevalent in the HPSG framework (Pollard & Sag 1994). Within the probe-goal
theory of the syntactic computation, the operation Agree can be formally
defined as in (13).
(13) Agree (Pesetsky & Torrego 2007: 4)
a. An unvalued feature F (a probe) on a head H at syntactic location
(F) scans its c-command domain for another instance of F (a goal )
at location (F) with which to agree.
b. Replace F with F, so that the same feature is present in both
locations.

If a goal is valued for F, replacing the token-value of the probe with the
value of the goal results in an instance of valued F substituting for the specification of the unvalued probe. Avalued F may now serve as the goal for
some ulterior operation of Agree triggered by an unvalued, higher instance
of F serving as a new probe. The result is that a single feature F will be
shared by several positions, and the process could iterate further.

476 Sandro Sessarego & Javier Gutirrez-Rexach

An element drawn from the lexicon with an interpretable valued feature

will be specified as i(nterpretable)F(eature)[val]; an uninterpretable unvalued feature that has not participated in Agree is annotated (where relevant)
by an empty pair of brackets ( uF[ ]) and, after the Agree operation takes
place, it turns into uF[val]. On the other hand, an element coming from
the lexicon without a specification for such a feature, will be annotated as
no-F[ ], and it will not be able to act as a probe for Agree operations of that
particular kind.
(14) uF[val] . . . uF[val] . . . iF[val] . . . uF[val] . . . no-F[ ]

If we postulate that an uninterpretable feature such as gender may be present in certain nominal elements but absent in others, and that variation is
the result of lexical differences in the feature specification of certain items,
it follows that contrasts in overt syntax will be the result of differences in
the computation of varying specifications. We propose an account of the
different gender-agreement configurations across DP strings in ABS that
can be summarized in the following fashion:
(15) a. [DP una [NP curva
uG[fem]....... iG[fem]...........
b. [DP una [NP curva
uG[fem]....... iG[fem]...........
c. [DP un [NP curva
no-G[]......... iG[fem]...........
d. [DP a
[NP curve
alarge curve

ancha]
uG[fem]
ancho]
no-G[]
ancho]
no-G[]
large]

Therefore, this approach can account for all the gender-agreement configurations encountered in the ABS Determiner Phrase by postulating the
presence/absence of unvalued gender features on the different DP components. Having clarified this point, we can now proceed to solve a different empirical issue, namely why no instances of post-nominal gender
concord can be found on adjectives unless pre-nominal articles agree (11),
in both Lipskis (2006c) and our corpus.

7. Changes in feature valuation

Results from statistical analyses (Sessarego 2009) suggest that ABS is
undergoing a cross-generational change, in which stigmatized basilectal

Gender agreement in the Afro-Bolivian DP 477

ABS features are being substituted by more prestigious HBS ones. One
result of this transition is the introduction of a wider range of genderagreement configurations in a language which originally made little use
of it. In minimalist terms, this phenomenon can be seen as the emergence
and development of unvalued features on elements which previously were
not specified for them.
The non-occurrence of strings like (16) might indicate that, cross-generationally, indefinite articles developed an unvalued gender-feature specification before post-nominal adjectives. Therefore, speakers mastering
post-nominal agreement would also present concord on indefinite articles.
(16)

a. *[DP un [NP curva ancha]

no-G[]........ iG[fem]............ uG[fem]
[DP a
[NP curve large]
alarge curve
b. [DP una [NP curva ancho]
uG[fem]...... iG[fem]............ no-G[fem]
[DP a
[NP curve
large]
alarge curve

Ahypothesis that emerges from this data is that unvalued gender features
developed gradually in ABS: first on certain elements (e.g. articles) and
only later on others (e.g. strong quantifiers, adjectives). Although this is the
general tendency, different linguistic and social factors may affect the selection of an item or other, and therefore the overt syntax result. Such variation can probably be modeled in sociolinguistic terms. Nevertheless, the
reduced number of participants in our survey does not allow us to conduct
a complete sociolinguistic analysis here. We will leave this study for further
research. In the present article we will only comment on cross-generational
differences and on the probability of nominal gender agreement associated
with different lexical and functional categories.
Findings from grammaticality judgments, oral questionnaires and
indirect questions led to the identification of four different patterns of
agreement (see 610). However, the comparison of such results with the
data recorded by means of sociolinguistic interviews revealed a considerable amount of variability between the grammatical intuitions ideally
assumed or accepted by the speakers and the patterns attested in the actual
survey, thus indicating that agreement paradigms may be not completely
stable in the grammar of these informants. For these reasons, the model

478 Sandro Sessarego & Javier Gutirrez-Rexach

proposed by Adger & Smith (2005) to account for unvalued uninterpreable features seems ideal to capture the nature of the phenomena found
in ABS. The nature of the element occurring with the nominal head (e.g.
articles, adjectives, strong/weak quantifiers, etc.), has a clear effect on the
output; however, not only internal factors condition the agreement operation, there are also external ones playing a crucial role. Acloser look at the
interview transcripts revealed that gender mismatches on adjectives and
determiners between ABS and Standard HBS are common, with the masculine gender prevailing over the feminine one. We claim that these differences are due to two separate factors: (a) Certain words listed in the HBS
lexicon as feminine are listed in the ABS one as masculine and vice versa;
(b) The valuation process of agreement in ABS departs from the standard
Spanish one in that certain ABS elements lack the unvalued features present in their Spanish counterparts.
While grammaticality judgments were discordant for certain syntactic
categories among informants, every participant agreed on the use of el and
la as respectively the masculine singular definite article and the feminine
singular definite one. Lipski (2009) reports no instances in ABS where el is
used with nouns that are grammatically feminine in HBS, while he comments on some cases in which la is used with nouns that are grammatically
masculine in HBS. Compare examples in (17) with their Spanish counterparts: el pulmn the chest, el patio the doorway.
(17) a. Mi quit-a mi
gorro
pa pon- aqu la
pulmn.
to.me take-PRS my.SG hat.M.SG to put-INF here the.F.SG chest.F.SG
He takes off my hat to put it here over my chest.
b. Yo lleg-aba la
patio.
I arrive-PST the.F.SG doorway.F.SG
Iarrived to the doorway.

On the other hand, our corpus includes several cases indicating that agreement mismatches involving definite articles can be found under both conditions, thus suggesting that a differential lexical specification exists and is
bidirectional. Compare the ABS examples in (18) with their Spanish counterparts: la maxima autoridad the top authority, la serpiente the snake; el
problema the problem, el sistema the system.
(18) a. l dic-e que es el
mxim-o
autoridad.
he say-PRS that is the.M.SG maximum-M.SG authority.F.SG
He says he is the top authority.

Gender agreement in the Afro-Bolivian DP 479

b. El
serpiente no se com-e; se mat-a con escopeta.
the.M.SG snake.F.SG no REFL eat-PRS REFL kill-PRS with rifle.F.SG
The snake cannot be eaten; it must be killed with a rifle.
c. La
problema
de Tocaa era lu
patrn.
the.F.SG problem.F.SG of Tocaa is-PST the.M.PL owner.M.SG
The land owners were the problem of Tocaa.
d. La
sistema
de hacienda no sirv-e
pa nada.
the.F.SG system.F.SG of plantation no work-PRS for nothing
The plantation system is useless.

Setting aside those mismatches that are due to different specifications in

the lexicon, all the others have to be viewed as the by-product of a specific
valuation constraint, which is conditioned by the type of grammatical category entering the derivation with the nominal head. Cross-generational
VARBRUL results for internal factors (Table 1) indicate that the unvalued
gender-feature distribution among grammatical categories is highly variable (Range 72): post-nominal adjectives disagree the most (Factor Weight
.95), while plural and singular definite articles show the highest level of
concord (Factor Weight .23).
Within a variationist minimalist framework, several external factors
may affect the selection of an item: ease of lexical access (probably linked
to frequency of use), speaker-hearer relationships, gender, social identity,
age, etc. (Adger & Smith 2005: 164). Quantitative results not only provide
Table 1. Cross-generational variable rule analysis of the contribution of internal
factors to the probability of lack of gender agreement in Afro-Bolivian DP
Grammatical
category

Factor
weight

%Lack
agreement

%data

Post-Nom. Adj.
Strong Q.
Pre-Nom. Adj.
Indef. Art.
Weak Q
Dem
Def. Art.

.95
.66
.64
.62
.60
.24
.23
Range
72

50
35
14
12
10
3
2

272
275
220
280
102
84
1371

19
11
19
11
4
3
53

Total = 2604; log likelihood = 624.215; total chi-square = 202.0101; chi-square/cell =

21.6291; significance = 0.001; input = 0.041

480 Sandro Sessarego & Javier Gutirrez-Rexach

Table 2. Cross-generational variable rule analysis of the contribution of external
factors to the probability of lack of gender agreement in Afro-Bolivian DP
Generation

Factor weight

%Lack agreement

%data

80+
5180
2150

.67
.56
.35
Range
32

21
11
1

651
927
1026

25
36
39

Total = 2604; log likelihood = 624.215; total chi-square = 202.0101; chi-square/cell =

21.6291; significance = 0.001; input = 0.041

empirical support for a structural analysis, they also provide more evidence
for microvariation between closely related grammatical systems exhibiting
orderly heterogenity which can, in turn, be correlated with external variables (Cornips & Corrigan 2005: 7). While the reduced number of informants does not allow a complete sociolinguistic analysis, it is worth pointing out that the age factor significantly affects variation in this case (see
Table 2).
Generation is, in fact, a significant factor group (Range 32), with the
oldest group (80+) strongly favoring disagreement (Factor Weight .67) and
the 2150 group disfavoring it (Factor Weight .35). These data reflect the
presence of a cross-generational change, pushing ABS in the direction of
HBS. Young generations did not experience the segregation imposed by
the hacienda system and had more chances to have contact with the Spanish variant spoken outside the community. These conditioning elements,
in addition to the stigmatization attached to the Afro-Hispanic vernacular,
are pushing the younger members of the community to quickly replace the
basilectal features with more prestigious HBS ones.3
3

Processes of this kind have been labeled as cases of decreolization in the literature
(DeCamp 1971; Bickerton 1973; Rickford 1987; Winford 1997). The notion of decreolization
consists of a series of linguistic approximations ranging from the creole to the superstrate
language. This term, therefore, seems to efficiently describe the gradual shift in genderagreement patterns found in ABS. Nevertheless, we would like to point out that the term
decreolization could be a bit misleading in that it might lead people to wrongly assume
that ABS was once a radical creole, a language derived from an earlier pidgin stage (Lipski
2009: 186). On the other hand, in line with the sociohistorical evidence provided by Sessarego (2010a, b), we would like to claim that ABS did not evolve from a pidgin, as it was
probably a language which highly resembled the superstrate from its inception. Obviously,
during the last decades especially after 1952 ABS has approximated standard Spanish
even more. This, however, does not imply that before such date ABS was radically different.

Gender agreement in the Afro-Bolivian DP 481

Although there are no wider diachronic data available, looking at the

synchronic results for the three generations under analysis, a clear picture
emerges of how the gender-agreement domain might have expanded in
ABS. The statistical findings presented in Table 1 and 2 give only a partial sketch of how unvalued gender feature might have developed. Only if
we look at the result for each generation individually, gender-agreement
development patterns become clearer.

8. The dynamics of gender feature domains

The above data have a clear structural repercussion. As the syntactic trees
in Figures 1 to 3 show, the gender-feature domain is locally limited in the
syntactic representation. DP constituents are arranged in a hierarchical
order, according to their syntactic position (see Bosque & GutirrezRexach 2010): strong quantifiers, demonstratives/definite articles, weak
DP
No-G[ ]
STRONG Q
39%(29/59)
uG[fem]

DEMDEF ART
96%(306/319)

Gender feature domain

uG[fem]

WEAK Q
81%(95/117)
uG[fem]

PRE-NOM ADJ
74%(42/57)
iG[fem]
no-G[ ]

NOUN
POST-NOM ADJ
21%(21/98)
Figure 1. Gender agreement patterns for the 80+ generation according to
grammatical category (percentages and raw numbers)

482 Sandro Sessarego & Javier Gutirrez-Rexach

DP
No-G[ ]
STRONG Q
57%(61/107)
uG[fem]

DEMDEF ART
97%(496/510)

Gender feature domain

uG[fem]

WEAK Q
89%(118/143)
uG[fem]

PRE-NOM ADJ
85%(70/82)
iG[fem]
no-G[ ]

NOUN
POST-NOM ADJ
39%(37/95)
Figure 2. Gender agreement patterns for the 5180 generation according to
grammatical category (percentages and raw numbers)

quantifiers, pre-nominal adjectives, nouns and post-nominal adjectives.

As can be observed, nouns carry an interpretable valued gender feature
iG[fem], while other elements present uninterpretable valued gender features uG[fem], or no gender features altogether no-G[ ].
As can be noticed, the three generations show three different levels of
gender agreement. While for the 81+ generation agreement is mainly limited to demonstrative, definite articles, weak quantifiers and pre-nominal
adjectives; for generation 5180 also strong quantifiers agree in the majority of instances. On the other hand, post-nominal adjectives agree more
than 50% of the cases only for the 2150 generation informants.
Gender-agreement evolution seems to develop cross-generationally in a
systematic way. In fact, for all the three figures, the following gender-agreement ranking is maintained across the grammatical categories analyzed:
(19) DEM/DEF ART > WEAK Q > PRE-NOM ADJ > STRONG Q > POST-NOM ADJ

This hierarchy, in addition to the fact that all singular definite articles agree

Gender agreement in the Afro-Bolivian DP 483

DP
No-G[ ]
STRONG Q
88%(96/110)
uG[fem]

DEMDEF ART
99%(577/581)

Gender feature domain

uG[fem]

WEAK Q
96%(139/144)
uG[fem]

PRE-NOM ADJ
94%(78/83)
iG[fem]
no-G[ ]

NOUN
POST-NOM ADJ
78%(85/109)
Figure 3. Gender agreement patterns for the 2150 generation according to
grammatical category (percentages and raw numbers)

with the gender of the noun, might indicate that in a previous phase gender agreement was limited to singular definite articles, and it gradually
extended to the rest of the categories.
Interestingly, these findings are in line with SLA research on the acquisition of gender agreement in DP. In fact, Hawkins (1998) showed that
English students speaking French as a second language presented more
agreement on definite articles than on indefinite ones, and also more agreement on determiners than on adjectives; similar findings have also been
reported for English speakers of Spanish by Bruhn de Garavito & White
(2000), and more recently by Franceschina (2005) who tested advanced
speakers of Spanish coming from a variety of backgrounds (Italian, Portuguese, English, Arabic, German and French). All these studies on gender
agreement also share the common view that masculine is the default value,
as it appears significantly more on determiners, and on adjectives in cases
of agreement mismatches. These data indicate that language evolution follows certain hierarchical steps, as proposed by Pienemanns (1998) Pro-

484 Sandro Sessarego & Javier Gutirrez-Rexach

cessability Theory. According to Processability Theory, there is a universal,

implicational hierarchy of processing procedures derived from the general architecture of the language processor. Additionally, there are specific
procedural skills needed for the production of utterances from a language
to be learned, the target language. Based on these assumptions, several
predictions can be made for second language development which can be
tested empirically. The central claim of this theory is that these processing procedures not only reflect their sequence of activation in language
generation but also that the acquisition of these procedures will follow
this implicational hierarchy. More research must definitely be carried out
to achieve a broader generalization and decide on whether the path of
gender-agreement development reported in (19) can really be considered
as a list of activations in language generation; nevertheless, the data and the
framework seem to be highly promising.

9.Conclusions
This study offers a quantitative approach to variable gender agreement
within the DP in Afro-Bolivian Spanish. Our findings and proposal try
to bridge the gap between the study of variation and general theories
about syntactic structure. Variation is a component of human languages,
and our results confirm that it should be taken into account when analyzing structural properties in specific syntactic domains, such as agreement in the DP. Our goal is to characterize the ingredients of variation
in a structurally systematic fashion, as computationally determined by
differences in the specification of lexical items and by restrictions on syntactic operations, more specifically, as a locality condition on agreement.
Accounts of this sort are now possible after recent developments in the
minimalist (and related) frameworks, which are trying to account for
alternation and variation phenomena affecting syntactic elements (Adger
& Smith 2005).
Our proposal also has important sociolinguistic consequences. The
underlying reasons pushing Afro-Bolivian in the direction of a more prestigious Spanish variety are essentially the stigmatization of the Afro-Hispanic vernacular and the increasing contact with a more prestigious Spanish dialect. Contact with Bolivian Spanish increased substantially after

Gender agreement in the Afro-Bolivian DP 485

1952, the year of the Bolivian Land Reform, which freed Afro-Bolivians
from slavery and introduced education in the black communities. These
changes, which have affected the socio-economic scenario of black Bolivia
during the last six decades, are reflected in the speech of the members of its
community. This scenario would explain why generation was proven to be
a significant factor group affecting the studied variation.
Moreover, the different factor weights displayed within the factor group
grammatical category suggest that different grammatical elements have
different probabilities of agreeing with the gender feature of N. This fact,
in addition to the constant ranking showed cross-generationally and to the
100 percent agreement for singular definite articles, might indicate that in
a previous stage of development agreement was limited to these elements,
and then it gradually spread to other categories. The nature of this transition is shown to be gradual, thus containing much alternation between
forms.
From a theoretical perspective, this work sheds some light on the linguistic constraints regulating gender agreement in an Afro-Hispanic vernacular approximating to a more prestigious Spanish dialect. The process
is driven by social factors through a path that is highly constrained by syntactic restrictions and configurations (cf. also Cornips & Corrigan 2005).

Abbreviations
ABS
Afro-Bolivian Spanish
D Determiner
DEF ART Definite article
DEM Demonstrative
DP
Determiner Phrase
F Feminine
G
Gender feature
HBS
Highland BolivianSpanish
HPSG Head-driven Phrase Structure Grammar
i Interpretable
INF Infinitive
M Masculine
N Noun

NP
Noun Phrase
PST Past
PL plural
POST-NOM ADJ Post-nominal adjectives
PRE-NOM ADJ Pre-nominal adjectives
PRS Present
PST Past
REFL Reflexive
SG Singular
SLA
Second langage acquisition
STRONG Q Strong quantifier
u Uninterpretable
Val Value
WEAK Q Weak quantifier

486 Sandro Sessarego & Javier Gutirrez-Rexach

References
Adger, David. 2006. Combinatorial variability. Journal of Linguistics 42: 503530.
Adger, David & Jennifer Smith. 2005. Variation and the minimalist program. In
Leonie Cornips & Karen P. Corrigan, eds. Syntax and variation: Reconciling the
biological and the social. Amsterdam: John Benjamins, 149178.
Adger, David & Graeme Trousdale. 2007. Variation in English syntax: Theoretical
implications. English Language and Linguistics 11: 261278.
Barbiers, Sjef. 2009. Locus and limits of syntactic microvariation. Lingua 119: 1607
1623.
Beninc, Paola. 1989. Dialect variation and the theory of grammar. Dordrecht: Foris.
Bickerton, Derek. 1973. The nature of a creole continuum. Language 49: 640669.
Bosque, Ignacio & Javier Gutirrez-Rexach 2010. Fundamentos de sintaxis formal.
Madrid: Akal.
Brody, Michael. 1997. Perfect chains. In Liliane Haegeman, ed. Elements of grammar. Dordrecht: Kluwer Academic Publishers, 139167.
Bruhn de Garavito, Joyce & Lydia White. 2000. L2 acquisition of Spanish DPs:
The status of grammatical features. In Catherine Howell, Sarah A. Fish & Thea
Keith-Lucas, eds. Proceedings of the 24th Annual Boston University Conference
on Language Development. Somerville, MA: Cascadilla Press, 164175.
Chomsky, Noam. 2000. Minimalist inquiries. In Roger Martin, David Michaels &
Juan Uriagereka, eds. Step by step: Essays on minimalist syntax in honor of Howard Lasnik. Cambridge, MA: MIT Press, 89156.
Chomsky, Noam. 2001. Derivation by phase. In Michael Kenstowicz, ed. Ken Hale:
Alife in language. Cambridge, MA: MIT Press, 152
Chomsky, Noam. 2002. Beyond explanatory adequacy. MIT Occasional Papers in
Linguistics 20. Department of Linguistics and Philosophy, MIT.
Chomsky, Noam. 2006. Language and mind. 3rd edn. Cambridge, MA: MIT Press.
Cornips, Leonie & Karen Corrigan. 2005. Toward an integrated approach to syntactic variation: A retrospective and prospective synopsis. In Leonie Cornips
& Karen Corrigan, eds. Syntax and variation: Reconciling the biological with the
social. Amsterdam: John Benjamins, 127
Cornips, Leonie & Cecilia Poletto 2005. On standardising syntactic elicitation
techniques (Part 1). Lingua 115: 939957.
De Camp, David. 1971. Towards a generative analysis of a post-creole continuum.
In Dell Hymes, ed. Pidginization and creolization of languages. Cambridge:
Cambridge University Press, 349370
Delicado-Cantero, Manuel & Sandro Sessarego. 2010. Variation and syntax in
number expression in Afro-Bolivian Spanish. In Luis Ortiz, ed. Proceedings of
the Hispanic Linguistic Symposium 2009 (HLS2009). San Juan: Universidad de
Puerto Rico.

Gender agreement in the Afro-Bolivian DP 487

Frampton, John & Sam Gutmann. 2000. Agreement is feature sharing. http://www.
math.neu.edu/ling /pdffiles/agrisfs.pdf.
Franceschina, Florencia. 2005. Fossilized second language grammars. Amsterdam:
John Benjamins.
Hawkins, Roger. 1998.The inaccessibility of formal features of functional categories in second language acquisition. Paper presented at the Pacific Second Language Research Forum, Tokyo, March 1998.
Kayne, Richard. 1996. Microparametric syntax: Some introductory remarks. In
James R. Black & Virginia Motapanyane, eds. Microparametric syntax and dialect variation. Amsterdam: John Benjamins, ixxxviii.
Koopman, Hilda. 1997. The SpecHead configuration. In Felicia Lee & Edward
eds. Syntax at sunset, UCLA Working Papers in Syntax and Semantics 1: 3765.
Labov, William. 1972. Language in the inner city. Philadelphia: Universiy of Pennsylvania Press.
Labov, William. 1984. Field methods of the project on linguistic change and variation. In John Baugh & Joel Sherzer, eds. Language in use. Englewood Cliffs:
Prentice Hall, 84112.
Lipski, John M. 2005. Nuevas fronteras de dialectologa afrohispnica: Los Yungas
de Bolivia. Proceedings from Conferencias sobre la lengua y cultura del mundo
de habla hispana. Departamento de estudios Hispnicos. Universidad de Estudios extranjeros de Kyoto.
Lipski, John M. 2006a. Morphosyntactic implications in Afro-Hispanic language:
New data on creole pathways. Paper presented at The 35th New Ways of Analyzing Variation Conference (NWAV35). Columbus: The Ohio State University.
Oct. 2006.
Lipski, John M. 2006b. El dialecto afroyungueo de Bolivia: En busca de las races
del habla afrohispnica. Revista Internacional de Lingstica Iberoamericana 3:
137166.
Lipski, John M. 2006c. Afro-Bolivian Spanish and Helvetia Portuguese: Semi-
creole parallels. Papia 16: 96116.
Lipski, John M. 2009. Afro-Bolivian Spanish. Madrid & Frankfurt: Iberoamericana Vervuert.
Milsark, Gary. 1974. Existential sentences in English. MIT Ph.D. dissertation.
Pesetsky, David & Esther Torrego. 2007. The syntax of valuation and the interpretability of features. In Simin Karimi, Vida Samiian & Wendy K. Wilkins,
eds. Phrasal and clausal architecture: Syntactic derivation and interpretation.
Amsterdam: John Benjamins, 262294
Picallo, Carmen. 2008. Gender and number in Romance. In Lingue e Linguaggio 1: 4766.
Pienemann, Michael. 1998. Language processing and second language development:
Processability theory. Amsterdam: John Benjamins.

488 Sandro Sessarego & Javier Gutirrez-Rexach

Pollard, Carl & Ivan Sag, 1994. Head-Driven Phrase Structure Grammar. Chicago:
University of Chicago Press.
Pomino, Natasha & Elisabeth Stark. 2008. Adnominal adjectives in Romance:
Where morphology seemingly meets semantics. In Mara T. Espinal, Manuel
Leonetti & Louise McNally, eds. Proceedings of the IV Nereus International
Workshop Definiteness and DP Structure in Romance Languages. Constanz:
Arbeitspapiere des Fachbereichs Sprachwissenschaft.
Portugal Ortiz, M. 1977. La esclavitud negra en las pocas colonial y nacional de
Bolivia. La Paz: Instituto Nacional de Cultura.
Rickford, John. 1987. Dimensions of a creole continuum. Stanford: Stanford University Press.
Sessarego, Sandro. 2009. Gender agreement variation in the Afro-Bolivian determiner phrase: The interplay of social and linguistic factors. Revista Iberoame
ricana de Lingstica 4: 135146.
Sportiche, Dominique. 2002. Movement, agreement and case. http://www.linguistics.ucla.edu/people/sportich/papers/mac.pdf.
Winford, Donald. 1997. Re-examining Caribbean English creole continua. World
Englishes 16: 233279.
Authors addresses
Sandro Sessarego
Department of Spanish & Portuguese
1018 Van Hise Hall
1220 Linden Drive
University of Wisconsin at Madison
Madison, WI 537061557, USA
E-mail: sandrosessarego@yahoo.it
Javier Gutirrez Rexach
Department of Spanish and Portuguese
Hagerty Hall 298
The Ohio State University
Columbus, OH 43210, USA
E-mail: gutierrez-rexach.1@osu.edu

Copyright of Folia Linguistica is the property of De Gruyter and its content may not be copied or emailed to
multiple sites or posted to a listserv without the copyright holder's express written permission. However, users
may print, download, or email articles for individual use.