Frequency Eff Ects on the Morphological Conditioning of Syllable-Fi Nal / S / Reduction in Border Uruguayan Spanish

Journal of Language Contact 4 (2011) 26–55 brill.nl/jlc

Frequency Eﬀ ects on the Morphological Conditioning of Syllable-ﬁ nal / s / Reduction in Border Uruguayan Spanish

Mark Waltermire New Mexico State University [email protected]

Abstract Th is research provides a usage-based account of the phonological reduction of syllable and word- fi nal /s / in the bilingual Spanish of the Uruguayan-Brazilian border. As with monolingual dialects of Uruguayan Spanish, patterns of aspiration and deletion seem to be conditioned by lexical frequency, with words of high frequency demonstrating higher rates of reduction than words of low frequency. Unlike these dialects, however, the deletion of /s / in border varieties may be conditioned by Portuguese morphological patterns in which plural / s / is deleted for head nouns and adjectives but retained for determiners. Th is possibility is explored by examining the linear and relative positions of plural NP constituents as well as the plural marking patterns of preceding constituents. Multivariate analysis shows that / s / deletion is statistically probable in highly frequent words whereas aspiration is favored among words of low frequency. Th e latter fi nding may be the result of a social distribution of aspirated variants among more upwardly mobile residents. Th e use of both variants is conditioned by the morphological patterns of Portuguese, in which plural marking is carried out through the use of overt articulations for plural determiners and fi rst position NP constituents while plural markers are deleted for nouns and adjectives, especially for second and third position constituents. Th ese results provide further evidence of the uniqueness of bilingual dialects and support the claim that bilinguals have more expansive linguistic repertoires than monolinguals.

Keywords Uruguay ; Spanish ; Portuguese ; usage-based ; frequency ; phonology ; morphology

1 Spanish and Portuguese Sibilant Development

Th e Portuguese sibilant system has remained relatively stable since the medi- eval period, having retained four of the original six sibilant phonemes (/s/, / z/, / ʃ/, and / ʒ/). Of the original seven sibilant phonemes in Old Spanish (/s/, /z/, / ts/, / dz/, / ʃ/, /ʒ /, and / tʃ/), only / s / and / tʃ / have been retained in modern dialects of this language. Th e phonemes / ts / and / dz / weakened to dental

Downloaded from Brill.com09/23/2021 10:14:22AM via free access M. Waltermire / Journal of Language Contact 4 (2011) 26–55 27

fricatives and merged into a single voiceless phoneme, which was later fronted to the interdental fricative /θ /, which is characteristic of northern and central Spain. Th e distinction between /θ / and /s / never emerged in Andalucian Spanish, nor did it carry over into dialects of Latin American Spanish (Lipski, 1994 ). A similar change occurred to /ʃ / and /ʒ /, which neutralized to /ʃ / and backed to the velar fricative /x / in the mid-seventeenth century (Penny, 1991 : 98-101).

1.1 Aspiration of / s / in Uruguayan Spanish

Th ough the sibilant phoneme system of Latin American varieties of Spanish has been greatly reduced, the variable phonetic realizations of the sibilant phoneme /s / (in both syllable-ﬁ nal and syllable-initial positions, depend- ing on the dialect) as [h] or [ø] have greatly enriched the phonetic inventories of these varieties, thereby distinguishing them from dialects of Castilian Spanish and highland varieties of Latin American Spanish, in which the use of such reduced articulations is much less common. As is the case for many dialects of South American Spanish, the reduction of syllable-ﬁ nal / s / to an aspirant or a phonetic zero is common in monolingual varieties of Uruguayan Spanish. Of these variants, aspiration is by far the most common. As illustrated in Table 1 , the use of aspirated variants is most frequent in pre-consonantal position.

Table 1 Realizations of /s / (in percentages) according to ﬁ ve linguistic contexts in monolingual Uruguayan Spanish1

Linguistic context [s] [h] [ø] sC 20 79 1 s#C 4 88 8 s## 85 13 2 s#V 98 2 0 s#v 93 7 0 C = consonant, V = tonic vowel, v = atonic vowel, # = word boundary, ## = absolute ﬁ nal

1 Th ese data were taken from Lipski (1988 : 315), who compares the reduction of syllable- ﬁ nal / s / in 13 dialects of modern Spanish. Unfortunately, raw numbers are not provided.

Downloaded from Brill.com09/23/2021 10:14:22AM via free access 28 M. Waltermire / Journal of Language Contact 4 (2011) 26–55

Pre-consonantal / s/, regardless of word position, is aspirated frequently in monolingual Uruguayan Spanish (at 79% in word internal position and at 88% in word fi nal position). Many studies have reported similar results for this distribution of syllable-fi nal / s / in Uruguay (Lipski, 1987 : 98-99, 1994: 340-341; Pedretti de Bolón, 1983 ; Ricci, 1963 ). Rates of aspiration of /s / are extremely high in this context, but low in pre-vocalic and absolute fi nal positions. Somewhat surprisingly, aspiration has not been reported for the Spanish of border communities in northeastern Uruguay, with the exception of Carvalho ( 2006a ), who examines the aspiration of syllable-fi nal / s / as a prestige marker in the Spanish of Rivera, the most populous city on the Uruguayan side of the border with Brazil. Th is may be due to the fact that aspiration has only recently entered into the Spanish of this contact variety and has tradi- tionally, perhaps even stereotypically, been regarded as a dialect in which /s / was maintained (Carvalho, 2006a ). It also may be due to the fact that past studies of language use on the border have focused primarily on Portuguese (alternately referred to as Dialectos portugueses del Uruguay or fronterizo by linguists and portuñol by non-linguists) (Carvalho, 1998 , 2003a , 2003b , 2004 ; Elizaincín 1976 , 1992a , 1992b ; Elizaincín et al., 1987 ; Hensey, 1972 ; Rona, 1965 ).2

1.2 Deletion of / s / in Uruguayan Spanish

Th e deletion of word-fi nal, pre-vocalic /s / is nonexistent in this dialect since it becomes the onset of the following syllable during the process of resyllabifi cation. In colloquial Spanish, there is a tendency to form new syllables based on the prototypical CV syllable pattern for this language, as in los animales ‘the animals’ /los.a.ni.má.les / > [lo.sa.ni.má.les] or [lo.ha.ni.má.leh] (in aspirating dialects) (Barrutia & Schwegler, 1994 ; López Morales, 1993). Th e deletion of word-fi nal / s/, though perhaps constrained somewhat in pre-vocalic contexts by the process of resyllabifi cation, is frequent only in a limited number of dialects of Spanish, including Cuban, Panamanian, and Puerto Rican varieties (Lipski, 1988 : 315). Th e deletion of word-fi nal /s/, therefore, cannot be considered a major phonological phenomenon in the majority of Spanish dialects, including Uruguay, as shown in Table 1 .

2 Th e term portuñol , which most Riverenses use to refer to border Portuguese, is used very loosely within the community as a mixed dialect without speciﬁ c reference to either a Spanish or Portuguese base.

Downloaded from Brill.com09/23/2021 10:14:22AM via free access M. Waltermire / Journal of Language Contact 4 (2011) 26–55 29

Numerous studies have shown that the deletion of / s / in most dialects of Spanish is not contingent upon morphological status, either as a plural marker or the infl ectional marker of second person singular (-s ) and fi rst person plural (-mos ) conjugations (Alba, 1990 : 86; Cedergren, 1973 : 44, 1978: 40; López Morales, 1983 : 47). Evidence from the current investigation also supports the idea that / s / deletion is the result of phonological, rather than morphological, simplifi cation in this language, as will be demonstrated shortly. Th e deletion of / s / in the current database, which includes approximately 50 hours of colloquial speech data from the border Spanish of Rivera, Uruguay, occurred in all potential syllable-fi nal positions, regardless of other factors such as morphological status, word position, or word stress. Examples of the deletion of syllable-fi nal / s / according to each of these factors from the Rivera database are displayed in Table 2 . Th e examples in Table 2 show that the deletion of plural and monomorphemic / s/, whether word-internally or word-fi nally, is possible in this linguistic community. Rates of deletion, however, may be accelerated for plural markers, due to the high level of frequency with which these morphemes occur in discourse. Th is possibility will be explored in detail in Section 6.

1.3 Variants of / s / in Southern Brazilian Portuguese

In the Portuguese of Rio Grande do Sul, monomorphemic /s / is maintained as a sibilant (either voiced or voiceless), regardless of syllable or word positions (Koch et al., 2002 : 32-43, 189-192).3 Voicing assimilation of / s / occurs both before voiced consonants (i.e. mesmo ‘same’ [mέz.mu]) and intervocalically (os animais ‘the animals’ [u.za.ni.májs]). Th e use of voiceless [s] is standard in all other contexts (i.e. sério ‘serious’ [sé.ri.u], falso ‘false/untrue’ [fáw.su], isto ‘this (dem.)’ [ís.tu], isso ‘that (dem.)’ [í.su], etc.). In Peninsular Portuguese and that of Rio de Janeiro, syllable-fi nal / s / is frequently realized as an alveo-palatal fricative (Azevedo, 1981 ; Stavrou, 1947 ). Voicing assimilation also aff ects its production so that mesmo ‘same’ is realized as [mέʒ.mu] whereas / s / remains voiceless when followed by a voiceless consonant (such as [íʃ.tu] for isto ‘this (dem.)’). Given the ample diachronic and synchronic evidence provided thus far, it can now be proposed that the phonological weakening of syllable-fi nal /s / is a characteristically Spanish phenomenon whereas the maintenance of this phoneme is typical of Portuguese.

3 Th ough the aspiration of / s / apparently does not occur in the Portuguese of southern Brazil (Koch et al., 2002 ), the aspiration of this phoneme has been reported in northeastern states of this country, most notably in Recife (Ferreira, 2001; Palácio, 1989).

Downloaded from Brill.com09/23/2021 10:14:22AM via free access 30 M. Waltermire / Journal of Language Contact 4 (2011) 26–55

Table 2 Examples of the deletion of syllable-ﬁ nal / s / in the Spanish of Rivera, Uruguay

Linguistic context Example Pre-consonantal, word-internalmismo ‘same’ (Monomorphemic) [míø.mo] (39/A:191/61) Pre-consonantal, word-fi nal muchas cosas ‘many things’ (Plural) [mú.tʃaø.kó.sas] (26/A:364/41) Pre-vocalic (atonic), word-fi nal grupitos así ‘groups like that’ (Plural) [gru.pí.toø.a.sí] (09/A:167/09) Pre-vocalic (tonic), word-fi nal entonces esa ‘so that (dem.)’ (Monomorphemic) [en.tón.seø.é.sa] (04/A:165/03) Phrase-fi naluno de ellos ‘one of them’ (Plural) [ú.no.ðé.yoø] (17/A:117/24)

While a full sibilant (either voiced or voiceless) is by far the most common variant of / s / in Southern Brazilian Portuguese, the deletion of this consonant is common in plural noun phrases for plural markers of nouns and adjectives (Azevedo, 1989 ; Guy, 1981 ; Scherre, 2001 ). In vernacular varieties of Brazilian Portuguese, even among the highly educated, Azevedo ( 1989 : 867) claims that “the plural marker is borne out not by the head noun(s) but rather by the leftmost determiner,” as shown in examples (1) and (2):

(1)se o senhor quer nós traz as mala [standard = malas] agora e os pacote grande [standard = pacotes grandes] depois ‘if you wish we’ll bring the suitcases now and the large packages later’ (2)o meu ﬁ lho pegou umas cinco carpa [standard = carpas] ‘my son caught ﬁ ve carp’ Azevedo ( 1989 : 867)

Th e deletion of / s / in Portuguese, since it is limited to plural contexts only, represents a process of morphological, rather than phonological, simpliﬁ cation. Since the deletion of / s / does not occur in other linguistic contexts in Portuguese, such as word-internally, this reduction cannot be explained as phonological weakening. Previous studies of border phenomena have also reported the deletion of plural / s / in Uruguayan Portuguese (Elizaincín et al., 1987 : 61) and Spanish (Carvalho, 2006b ). Th e high token frequency with which plural /s / occurs in discourse may also be related to the deletion of plural markers of head nouns and adjectives in the Spanish of Rivera. Th at is, deletion may not be solely the result of contact with

Downloaded from Brill.com09/23/2021 10:14:22AM via free access M. Waltermire / Journal of Language Contact 4 (2011) 26–55 31

Portuguese, but also as the natural result of phonological erosion in highly frequent infl ectional suffi xes such as the English progressive (– ing ® – in ) or Spanish past participles (–ado ® – ao, –ido ® – io). Furthermore, the possibility that token frequency serves as a conditioning factor in /s / deletion in this variety must be explored given the plethora of evidence from usage-based studies on phonological change showing that articulations for words of high token frequency tend to reduce at a more accelerated rate than those for words of low token frequency (Brown, 2004 ; Brown & Torres Cacoullos, 2002; Bybee, 1985 , 2000 , 2001 ; Bybee et al., 1994 ; D’Introno & Sosa, 1986 ; Haiman, 1994 ; Hooper, 1976a , 1976b ; Jurafsky et al., 2001 ; Krug, 1998 ; Langacker, 2000 ). Brown and Torres Cacoullos (2002: 25), for example, found that the reduction of syllable-initial / s / in the Spanish of four men from Ascención, Chihuahua is statistically favored for words of high token frequency but disfavored for words of low token frequency. Also supporting usage-based theories of consonant reduction are the results reported in Waltermire (2006 ). Multivariate analysis revealed that the deletion of intervocalic /d / in the Spanish of Rivera is favored among words and morphemes of high token frequency. Th e deletion of / s / in the Spanish of Rivera, therefore, may be conditioned by token frequency, whether as part of a morpheme (which tend to occur frequently in discourse due to their high functional load) or not. Th e objective of the current research is to examine the role of morphological conditioning in the use of reduced variants of syllable and word-fi nal /s / in border Uruguayan Spanish. Since high token frequencies condition consonant reduction in other varieties of Spanish, it is proposed that this factor will also infl uence the reduction of syllable and word-fi nal / s / in the Spanish of Rivera, though perhaps at a less accelerated rate than in other varieties due to contact with Portuguese.

2 Data Collection and Methodology

Field work (sociolinguistic interviews, participant observation, and collection of questionnaire data) was conducted in Rivera in the spring of 2003. So as to achieve the most accurate representation possible of the linguistic community of Rivera, I decided to choose consultants based on age, sex, and occupation proportionately to their actual populations within the community as indicated by 2003 census data. Regarding age, three generational groupings (16-25; 26-50; 51-78) were designated based on the emic approach described in Eckert (1997 ). Although a roughly equal number of consultants was chosen to ﬁ ll each of these generational categories, there is a slightly higher

Downloaded from Brill.com09/23/2021 10:14:22AM via free access 32 M. Waltermire / Journal of Language Contact 4 (2011) 26–55 representation of community members belonging to the second generation given that there are more speakers belonging to this generation than to the other two generations in Rivera. A total of 27 consultants between the ages 26 and 50 were interviewed while only 18 consultants were interviewed from the oldest and youngest generations. Th e selection of consultants by sex was also based on 2003 Uruguayan census data. According to these data, the female population of Rivera (N=53,430) slightly outnumbers that of males (N=51,491). Th is demographic trend is also consistent for the entire Uruguayan population. Th e same statistics also show that women live longer than men on average. With this in mind, a greater number of women were interviewed than men. Out of the 63 total consultants interviewed for the corpus, 35 are women and 28 are men. Unfortunately, no census data were readily available for occupations within the community. Th is being the case, three occupational designations for the selection of consultants were chosen - professional, non-professional, and stu- dent. Th ese designations are general enough to encompass all possible professions within Rivera. Professional refers to any member of the community that practices an occupation requiring advanced training and/or education. Professionals (N=28) include teachers, lawyers, doctors, health inspec- tors, and engineers. Non-professionals of the community (N=23) work in professions that do not require advanced education. Th ey are hotel owners and their employees, shopkeepers and their employees, secretaries, police offi - cers, currency exchangers, waiters, street vendors, and taxi drivers. Th ough the majority of members of the fi rst generation are students (N=12) due to their age, six of the younger consultants had already entered the local workforce at the time of interviews. In order to achieve an accurate representation of professions for this generation, fewer professionals (N=2) and non-professionals (N=4) were selected than students. A total of 3,091 tokens of syllable and word-fi nal /s / were extracted from the Rivera database, with approximately 50 tokens for each speaker.4 All tokens of /s / were extracted using Sound Walker. Each token was coded as either a sibilant [s], an aspirant [h], or a phonetic zero [ø]. Only syllable- fi nal and word-fi nal occurrences of / s / were used for the current data set. Word-internal tokens, whenever they occurred in syllable-fi nal position, were also included. Th e only word that was not coded for the current study was the tag question ‘ta ‘alright’. Th is simplifi ed lexical chunk, which is

4 Speakers 21, 24, and 55 did not speak for a long enough period of time to reach a total of 50 tokens.

Downloaded from Brill.com09/23/2021 10:14:22AM via free access M. Waltermire / Journal of Language Contact 4 (2011) 26–55 33 most likely a reduced form of está ‘is’ from the expression tá bom (Portuguese) or tá bien (Spanish) ‘it’s alright’, is highly frequent in the Spanish of Rivera. However, given that the entire fi rst syllable of the verb está has been deleted, it does not seem that this form is simply a result of synchronic reduction. Th e tendency to delete /s / in syllable and word-fi nal positions is pre- sumed to be extremely infrequent in the monolingual Spanish of Uruguay, as shown in Table 1. As mentioned in Section 1.2, word-fi nal /s/, when followed by a vowel, becomes the onset of that syllable, thereby resisting deletion in many dialects of Spanish. Th e use of a phonetic zero in this linguistic context does occur in the Spanish of Rivera, however, and must be included as a variant of /s / for this very reason. Th e very fact that deletion of /s / occurs in this variable context in and of itself is signifi cant to the research of phonological change in this bilingual dialect. Rates of fi nal /s / deletion in Rivera may be more accelerated than in monolingual varieties of Uruguayan Spanish, however, due to the frequency of deletion of word-fi nal plural / s / in Brazilian Portuguese. Tokens of deletion of syllable and word-fi nal /s / correspond to the complete absence of articulation of this phoneme. Perceiving the lack of articulation of syllable and word-fi nal /s / was accomplished without much diffi culty, especially given the benefi ts of variable speed control mechanisms which are standard functions of the playback apparatus used for extraction. Th e ability to distinguish between [s] and [h] was relatively easy given the highly divergent points of articulation of these sounds. A series of self-reliability checks were conducted in order to ensure the internal accuracy of token identifi cation. Th is was achieved by revisiting and recoding tokens for three speakers after all tokens had been extracted. Out of a total of 150 tokens, a level of 97.3% reliability was achieved during this procedure (146/150). Th is level of agreement might have potentially decreased if another researcher had also identifi ed the same tokens. I sincerely doubt, however, that this level of reliability would change drastically due to the overt perceptual diff erences between these sounds.

3 Community-wide Realizations of / s/

Th ough the reduction of syllable and word-ﬁ nal /s / in the Spanish of Uruguay may be conditioned by morphological status and lexical frequency, which will be analyzed shortly, it is crucial to survey the general distribution of variants of this phoneme in the community of Rivera. Since this

Downloaded from Brill.com09/23/2021 10:14:22AM via free access 34 M. Waltermire / Journal of Language Contact 4 (2011) 26–55 is a bilingual speech community, in contact with a dialect of Portuguese in which /s / is not aspirated, the overall rates of aspiration of syllable and word-fi nal /s / in Rivera may be lower than those of other dialects of Spanish in the River Plate region. Th e relative frequency of use of each of the three variants of syllable and word-fi nal /s / in the Spanish of Rivera is shown in Table 3 . Th ough rates of aspiration and deletion are most likely dependent on speaker diff erences, the retention of syllable-fi nal / s / at 63.5% is far more frequent than its reduction for the community as a whole. Th is may be a result of the inherent formality of the interview situations from which these data were derived, in which the speaker feels social pressure to speak a standard, grammatically correct Spanish. However, if this were true, it is unlikely that the community sample as a whole would delete /s / at such a high rate (12.0%), especially since this is the most socially stigmatized of all three variants (Lipski, 1994 : 340-341). Th ough the formality of the interview sessions most likely elicited a somewhat more formal register of Spanish for some consultants, the observer’s eff ect was apparently not substantial enough to completely formalize the speech data. Th us, an alternative explanation for the high retention of this sibilant is necessary. Th e use of [s] in Rivera is relatively high when compared to its use in monolingual dialects of Uruguayan and Argentine Spanish. Several studies confi rm the frequent aspiration of syllable and word-fi nal / s / in the monolingual Spanish of Montevideo (Barrios, 1991 ; Calvis de Bon, 1987 ; Ricci, 1963 ; Pedretti de Bolón, 1983 : 109-110; Vásquez, 1953 ). Barrios (1991 ), for example, reports that [s] is maintained at a rate of only 28% in the Spanish of Montevideo, while aspiration dominates at 54%. Bybee ( 2001 ) reports similar frequencies of use of variants of / s / using a corpus of colloquial Spanish from Buenos Aires (Terrell 1978 ). Results from this study show that the aspiration of /s / is highly frequent when followed by a consonant, especially in word- internal position (80.0%) (Bybee, 2001 : 140). Unfortunately, the overall distribution for these variants was not given in this study (since the research

Table 3 Distribution of variants of syllable and word-ﬁ nal / s / in Rivera

Linguistic variant [s] [h] [ø] 1962/3091 758/3091 371/3091 (63.5%) (24.5%) (12.0%)

Downloaded from Brill.com09/23/2021 10:14:22AM via free access M. Waltermire / Journal of Language Contact 4 (2011) 26–55 35 objective was the examination of the contextual conditioning of / s / variants only). In order to determine the overall distribution of each variant, the total number of all occurrences of syllable-ﬁ nal / s / for this corpus was calculated (N=14,681). Next, the percentages given by Bybee ( 2001 : 140) were con- verted back into raw totals for each linguistic context and totaled in order to show the overall distribution of each variant in this dialect, irrespective of context. Of all three variants, an aspirant is the most frequently used (N=7,548, 51.4%), followed by sibilant [s] (N=5,309, 36.2%) and phonetic zero (N=1,824, 12.4%). Note that the use of an aspirant for syllable-ﬁ nal /s / in the Spanish of both capital cities is more than double that of the fronterizo Spanish of Rivera (at 24.5%, Table 3). Interestingly, the deletion of /s / occurs at approximately the same rate (12%, Table 3), which suggests that this phenomenon in Rivera may not be the result of language contact. Th is possibility, however, will not be ruled out, especially since the deletion of /s / in Rivera may be conditioned primarily by morphological constraints rather than phonological factors.

4 Word Frequency

Token counts for all words in the current data set were compiled with token counts for the same words from the Corpus del Español (Davies, 2002 ), hereafter CdE. Th is corpus is an extremely useful tool due to its immense scope. It is the only on-line frequency dictionary of Spanish that spans eight centuries (from the 1200s to the 1900s), and, having been derived from multiple sources from these time periods, encompasses a wide variety of speech styles. Th is is especially true for the twentieth century data, which can be searched by oral and written modes. Word frequencies were obtained by using only the oral corpus from the 1900s, which alone consists of 5,000,000 tokens. Th is procedure ensured the closest approximation possible to an accurate lexical frequency representation of modern word usage. Frequency counts for tokens occurring ten times or more in the current data set and their corresponding counts from the CdE can be found in the appendix. Th e most frequent lexical types encountered in the Rivera data set appear fi rst, along with the token frequencies with which they occurred in this set, followed by the token frequency count for the same lexical type in the CdE. Th ree frequency levels (low, mid, and high) were chosen for the examination of frequency eff ects on the reduction of syllable and word-fi nal / s / in

Downloaded from Brill.com09/23/2021 10:14:22AM via free access 36 M. Waltermire / Journal of Language Contact 4 (2011) 26–55

Rivera. Th e token frequencies of all lexical types from both the Rivera data set and the CdE were compared in order to judge acceptable cut-off points for levels of frequency. Alternative methods pursued in the classifi cation of frequency levels, such as dividing the Rivera data set in thirds, either by type or by token, did not achieve an accurate representation of word frequency for this community. Words occurring 15 times or more in the speech of more than one consultant in the data set were classifi ed as highly frequent (the most frequent being es ‘it/he/she is’, somos ‘we are’, más ‘more’ and the defi nite articles los (masc.) and las (fem.)) whereas words occurring two times or less were coded as low frequency items (such as rótulos ‘signs’ and anillos ‘rings’). Words of mid token frequency comprise all words occurring more than three times but less than 15 (i.e. desde ‘since’, lados ‘sides’, and maestro(s) ‘teacher(s)’). Out of 508 total lexical types, there are 38 types of high frequency, 102 types of mid frequency, and 368 types of low frequency. Same speaker repetitions of several words occurring three times were classifi ed as low frequency items. Th e distribution of variants of /s / according to word frequency is shown in Table 4 . Th ough the distribution of variants by word frequency is statistically signifi cant, rates of use of these variants for words of all frequency levels are approximately the same. Given the breadth of evidence supporting the hypothesis that frequently occurring lexical items tend to reduce phonologically, it is expected that the reduction of / s / will be conditioned by high token frequency. Only by conducting a multivariate analysis of the distribution of / s/, however, will the contribution of this factor become more transparent.

Table 4 Distribution of variant use according to word frequency Linguistic variant Word frequency [s] [h] [ø] Total High 1219/1955 502/1955 234/1955 1955/3091 (62.3%) (25.7%) (12.0%) (63.3%) Mid 420/631 126/631 85/631 631/3091 (66.5%) (20.0%) (13.5%) (20.4%) Low 323/505 130/505 52/505 505/3091 (64.0%) (25.7%) (10.3%) (16.3%) χ² = 10.4194; df = 4; p < .05

Downloaded from Brill.com09/23/2021 10:14:22AM via free access M. Waltermire / Journal of Language Contact 4 (2011) 26–55 37

5 Lexical Eﬀ ects on the Deletion of Word-ﬁ nal / s / in Rivera

Th e entrenchment of lexical eff ects from Portuguese could also condition the deletion of /s / in certain frequently occurring verb forms in Portuguese in which / s / is deleted, such as: 1) vamo’ ‘let’s go’ as in vamo’la ‘let’s go there’ [và. mu.lá] rather than [và.muz.lá] (Spanish = vamos, vámonos) and 2) digamo’ ‘let’s say/say (subj.)’ [dʒi.gá.mu] rather than [dʒi.gá.mus] (Spanish = digamos ). Th e use of these reduced forms is frequent in the colloquial Portuguese of all Brazilian dialects. Th e use of such forms by such a socially diverse population has become so frequent that these words have routinized into lexical chunks. Th e rates of use of all three / s / variants for these frequently occurring cognates are shown in Table 5 . Th e deletion of / s / is frequent for fi rst person plural conjugations of both ir ‘to go’ and decir ‘to say/tell’ (despite relatively low token counts for each). Th is may be due to the occurrence of / s / in word-fi nal position, which seems to favor deletion more than word-medial position, and also due to its morphological status. Th is would at least partially explain the fact that the deletion of / s / for these items is much higher than overall rates of deletion, as shown in Table 3 . Overall rates of syllable and word-fi nal / s / deletion for the community do not exceed 12%. Rates of deletion for vamos and digamos are three to four times higher than for all other words in the current data set. When one also considers the extremely low rates of / s / deletion presented by Lipski (1988 ) for monolingual dialects of Uruguayan Spanish, which were shown in Table 1, it is plausible to posit that Portuguese is having a direct eff ect on the deletion of fi nal / s / for these verb forms.

Table 5 Distribution of / s / variants in the Spanish of Rivera for two frequently occurring cognates in which / s / is typically deleted in Brazilian Portuguese Linguistic variant Lexical type [s] [h] [ø] Vamos ‘let’s go/we go’ 3/22 11/22 8/22 (13.6%) (50.0%) (36.4%) Digamos ‘let’s say/we say (subj.)’ 4/15 3/15 8/15 (26.7%) (20.0%) (53.3%) χ² = 67.7899; df = 4; p < .05

Downloaded from Brill.com09/23/2021 10:14:22AM via free access 38 M. Waltermire / Journal of Language Contact 4 (2011) 26–55

6 Morphological Status

Th e decision to include morphological status as a factor in the conditioning of /s / reduction in the Spanish of Rivera was not motivated by functional concerns, which have often served as the reason for examining the deletion of plural /s / (Alba, 1990 : 112; Poplack, 1979 : 87-89, 1986). Th at is, these studies have often addressed the question of whether or not the deletion of this plural morpheme occurs less due to the potential loss of grammatical informa- tion which occurs in the process. Th e motivation for examining the eﬀ ect of morphology on the realization of /s / in Rivera was motivated more by the fact that the deletion of /s / in Portuguese is only frequent for plural markers of head nouns and adjectives, as discussed in Section 1.3. For this reason, plural tokens were coded as either determiners (i.e. los (masc.), las (fem.), etc.) or plural nouns and adjectives.5 If the deletion of /s / in Rivera is conditioned by morphological status, a case could be made for the inﬂ uence of Portuguese on the Spanish spoken in Rivera. It should be noted, however, that morphemes, such as plural / s/, occur frequently in discourse and, due to this high type frequency, tend to reduce phonologically. Table 6 shows the distribution of the use of variants of / s / in Rivera according to their morphological status.

Table 6 Distribution of variant use according to morphological status Linguistic variant Morphological status [s] [h] [ø] Total Plural (Nouns/ 517/841 175/841 149/841 841/3091 Adjectives) (61.5%) (20.8%) (17.7%) (27.2%) Non-plural 1305/2013 495/2013 213/2013 2013/3091 (64.8%) (24.6%) (10.6%) (65.1%) Plural 140/237 88/237 9/237 237/3091 Determiner (59.1%) (37.1%) (3.8%) (7.7%)

χ² = 61.4953; df = 4; p < .05

5 All tokens of plural deletion for the current study correspond simply to the absence of the plural marker / s/. While the possibility of leaving nouns and adjectives uninﬂ ected exists (as in los animal rather than los animales , ‘the animals’), no examples of these types of constructions were encountered in the current data set.

Downloaded from Brill.com09/23/2021 10:14:22AM via free access M. Waltermire / Journal of Language Contact 4 (2011) 26–55 39

Since rates of retention of plural / s / approximate those for monomorphemic words, the functional hypothesis does not seem to hold in this community. Furthermore, the deletion of / s / is actually higher for plural markers (at 17.7%), which contradicts the idea that the loss of grammatical infor mation impedes the process of deletion. Th e low rate of deletion for plural determiners (at 3.8%), however, shows that plurality is manifested morphologically by this element of the noun phrase, which partially supports the functional hypothesis. Th e high rate of use of some sort of overt plural marker for determiners (either a sibilant [s] or an aspirant [h]) at 96.2% suggests that the deletion of this consonant patterns like that of colloquial Portuguese, in which only the / s / of non-determiners suﬀ ers deletion. Th e relative con tribution of each of these factors will now be analyzed in order to fully understand the linguistic conditioning of variable realizations of / s / in the Spanish of Rivera.

7 Aspiration of / s / in Border Uruguayan Spanish

As determined previously, the aspiration of syllable and word-fi nal /s / is only characteristic of Uruguayan Spanish, not Portuguese, as such rea lizations of / s / have never formed part of the phonological repertoire of Southern Brazilian Portuguese. Th e aspiration of this phoneme has attained a certain level of prestige in Rivera due to its association with non-border, monolingual varieties of Uruguayan Spanish, with Montevideo at its center (Carvalho, 2006a ). Among the consultants for the current study, the rate of aspiration is twice as high as that of deletion, as shown in Table 3 . Th e use of an aspirant for syllable and word-fi nal /s / is quite robust in Rivera, despite claims to the contrary among residents from the capital (and even some residents of Rivera itself, ironically). We know the social value that this articulation has within the community, but how do word frequency and morphological status condition its use? As a reduced variant, it is possible that aspiration will occur more for plural nouns and adjectives while sibilant articulations will be maintained for plural determiners, in accordance with the norms of local Portuguese. Furthermore, aspiration is expected to be more likely among frequently occurring words, as phono logical erosion aff ects these words fi rst. Using GoldVarb X, the probabilities of co-occurrence between the applied variant, in this case aspiration, and the two independent factors groups (word frequency and morphological status) were determined. Th e results of this analysis are shown in Table 7 . Interestingly, the aspiration of / s / is statistically disfavored for plural nouns and adjectives with a probability weight of .45. As in Portuguese, number

Downloaded from Brill.com09/23/2021 10:14:22AM via free access 40 M. Waltermire / Journal of Language Contact 4 (2011) 26–55

Table 7 Multivariate analysis of the probabilities of /s / aspiration in border Uruguayan Spanish according to morphological status and word frequency (p<.05, N = 3091, Input = 0.243, Log likelihood = −1705.410)

Factor group Factor N % [h] Factor Weight Morphological status Plural 237 37.1 .65 determiner Non-plural 2013 24.6 .50 Plural (noun/ 841 20.8 .45 adjective) Range 20 Word frequency Low 50525.7 .55 High 1955 25.7 .51 Mid 63120.0 .45 Range 10

marking is carried out by the infl ectional suffi x of plural determiners. Unlike in Portuguese, however, the aspiration of infl ectional /s / is also possible in the same context. Th is is a classic case of the richness of bilingual phonological repertoires, which are more expansive than monolingual repertoires due simply to the existence of divergent articulations in diff erent linguistic systems, even those that are structurally quite similar, such as Spanish and Portuguese. Very surprisingly, the aspiration of /s / is statistically likely for low frequency words (with a probability weight of .55), yet not for words of high or mid frequency. We would expect rates of aspiration to be much higher among frequently occurring words, especially in a dialect in which this realization of / s / is somewhat recent. Th e prestige associated with this variant may provide a partial explanation for why this might be the case. Th e use of aspirated variants of /s / is most frequent among young members of the middle class (Carvalho, 2006a ), young professionals, and students (Waltermire, 2006 ). Th ese individuals likely incorporate a greater number of low frequency words in everyday speech since their access to them has been greater than for members of the community that have less formal education. Further research will need to be conducted to determine whether this is actually true. Th e likelihood that this is the case, however, provides some evidence that aspiration may be a primarily social phenomenon in Rivera. Th e natural phonological erosion that we see in other varieties of Spanish seems to have been deterred to a certain extent in Rivera due to such long periods of sustained

Downloaded from Brill.com09/23/2021 10:14:22AM via free access M. Waltermire / Journal of Language Contact 4 (2011) 26–55 41 contact with Portuguese. Th e aspiration of /s/, rather, is entering the community as a marker of prestige.

8 Deletion of / s / in Border Uruguayan Spanish

Unlike the aspiration of / s/, which does not occur in southern varieties of Brazilian Portuguese, the deletion of this consonant is characteristic of the Portuguese of Rio Grande do Sul. As explained in Section 1.3, the deletion of this consonant in colloquial Brazilian Portuguese (including southern dialects) is the result of morphological simplifi cation. It is important to note that the deletion of plural / s / in Portuguese for both masculine and feminine defi nite articles results in the singular form (os ® o , as ® a ) whereas the simplifi cation of these plural determiners in Spanish results in the same form as the singular only for feminine articles (los ® lo , las ® la ). Th ese changes and their potential ambiguity with regards to number, however, have not impeded the deletion of this plural marker in either language. Unlike Portuguese, though, in which the deletion of /s / occurs almost exclusively for plural nouns and adjectives of complex noun phrases and certain frequently occurring fi rst person plural verb infl ections, the deletion of / s / in Spanish is not confi ned to these contexts alone, but rather occurs in a wide array of linguistic contexts (including in word medial position and for determiners). Th e deletion of plural /s / for determiners in Spanish is a very frequent phenomenon in dialects in which /s / is deleted, which may or may not have to do with the lesser need in this language to mark masculine plural determiners in order to disambiguate number. Since this investigation seeks to examine competing linguistic infl uences on the conditioning of / s / deletion in the bilingual community of Rivera, phonetic zero was chosen as the application value for a separate multivariate analysis. Th e results of this analysis are shown in Table 8 . Th e factor that most signifi cantly conditions the deletion of / s / in Rivera is morphological status (with a range of 43). Th e deletion of this consonant is favored for plural nouns and adjectives (with a factor weight of .65) while it is highly disfavored for plural determiners (with a factor weight of .22). Th e deletion of / s / is slightly disfavored for non-plural / s / in monomorphemic words, providing further evidence that the deletion of / s / in the Spanish of Rivera is dictated primarily by morphological patterns. Th e tendency among consultants to retain some sort of overt articulation of /s / solely for plural determiners indicates that the patterning of variant use for plurals in the Spanish of Rivera is very similar to that of Brazilian Portuguese, only that in the Spanish of Rivera, the aspiration of this plural marker is also possible. Th e patterns of deletion with regards to word frequency are exactly as

Downloaded from Brill.com09/23/2021 10:14:22AM via free access 42 M. Waltermire / Journal of Language Contact 4 (2011) 26–55

Table 8 Multivariate analysis of the probabilities of /s / deletion in border Uruguayan Spanish according to morphological status and word frequency (p<.05, N = 3091, Input = 0.112, Log likelihood = −1105.396)

Factor group Factor N % [ø] Factor Weight Morphological status Plural (noun/ 841 17.7 .65 adjective) Non-plural 2013 10.6 .47 Plural determiner 237 3.8 .22 Range 43 Word frequency High 1955 12.0 .53 Mid 63113.5 .50 Low 50510.3 .39 Range 14

suspected. Words of high token frequency suﬀ er deletion whereas this process is unlikely for words of low token frequency. It would be very easy to simply say that the normal processes of phonological reduction are in place in this variety, with the deletion of /s / representing the most advanced stage of a change in progress. Aspiration in Rivera, however, is quite recent. It is not likely that it represents an intermediate stage of reduction since it has histori- cally not formed part of the phonological repertoire of the community.

9 Portuguese Inﬂ uence on the Morphological Conditioning of / s / Deletion in Rivera Spanish

Recent research on plural / s / deletion in Brazilian Portuguese (especially Scherre (2001 ) and Scherre and Naro (1998 )) has shown that patterns of deletion are not conditioned solely by lexical category, as has been proposed by Azevedo ( 1989 ) and which has been pursued thus far in the current research with regards to the Spanish of Rivera. Patterns of /s / deletion also seem to be conditioned by the linear and relative positions of plural NP constituents as well as by the plural marking patterns of preceding constituents. With respect to linear and relative position (i.e. where the NP constituent occurs relative to other constituents), Scherre and Naro (1998 : 10) have demonstrated that “position(s) to the left … favor explicit plural markers more than position(s) to the right” (translation mine). Th is ﬁ nding is of great signiﬁ cance to the current research in that constituents appearing in initial (left) positions, regardless of lexical category, may favor the retention of plural / s / as proposed

Downloaded from Brill.com09/23/2021 10:14:22AM via free access M. Waltermire / Journal of Language Contact 4 (2011) 26–55 43 by Scherre and Naro ( 1998 ). For this reason, this factor will be explored in detail in the following section. Also of great importance to the current research are the fi ndings presented in Scherre ( 2001 ) with respect to the plural marking patterns of preceding constituents within the NP. Th is research shows that immediately preceding zero plurals highly favor the deletion of plural / s / for subsequent NP constituents. Scherre ( 2001 ) eff ectively demonstrates that this constraint is the result of a phrase-level parallel processing eff ect. We shall now examine these two factors in the conditioning of plural /s / deletion in the Spanish of Rivera.

9.1 Linear and Relative Positions of NP Constituents

Scherre and Naro (1998 ), using an extensive corpus of colloquial Portuguese from Rio de Janeiro, examined the role of linear and relative positions on plural marking patterns in this dialect. Linear position refers to linear order within the NP, in this case position 1, position 2, or position 3. Relative position refers to the constituent’s position with respect to the head noun of the NP. As such, there are pre-head constituents (to the left of the head), post- head constituents (to the right of the head), and head constituents them- selves. Using all possible permutations of both linear and relative positions, they propose the following combinations, each of which is demonstrated by an original example from the Rivera corpus.

Table 9 Examples of linear and relative positions of NP constituents in the Spanish of Rivera Positions Example Pre-head, Position 1tantas cosas ‘so many things’ [tán.taø.kó.saø] (26/A:210/42) Pre-head, Position 2sus propios familiares ‘their own family members’ [sus.pró.pjoø.fa.mi.ljá.res] (10/A:198/10) Head, Position 1 cosas graciosas ‘funny things’ [kó.saø.gra.sjó.sas ] (34/B:117/63) Head, Position 2 cinco años ‘ﬁ ve years’ [síŋ.kwá.ñoø] (34/A:167/53) Head, Position 3 todos los días ‘every day’ [tó.doh.loz.ðí.aø] (15/A:176/19) Post-head, Position 2 compañeritos brasileros ‘Brazilian friends’ [kom.pa.ñe.rí.toz.ßra.si.lé.roø] (08/A:185/07) Post-head, Position 3 las hijas mayores ‘the oldest daughters’ [la.sí.haø.ma.yó.reø] (26/A:107/43)

Downloaded from Brill.com09/23/2021 10:14:22AM via free access 44 M. Waltermire / Journal of Language Contact 4 (2011) 26–55

In addition to the positions that appear in Table 9, single nouns should also be included in the analysis of plural / s / deletion. Th ese nouns, though obviously not occurring in strings, are nonetheless signifi cant to the current research in that, aside from also constituting noun phrases, they may reveal further details concerning patterns of deletion for nouns that are not preceded by any addi- tional constituents. It is predicted that plural /s / for these nouns will resist deletion since they bear plural marking without recourse to other elements. Since deletion and aspiration of syllable and word-fi nal /s / are the focus of the current study, patterns of maintenance will not be analyzed explic- itly. Th at said, patterns of maintenance, which would also included aspiration, can be inferred through the thorough examination of patterns of deletion. In order to determine the contributions of each of the linear and relative positions to the deletion of plural /s / for NP constituents in the Spanish of Rivera, a third multivariate analysis was conducted using GoldVarb X. All tokens of plural / s / in the current data set (N=1,078) were extracted and coded for linear and relative positions within the NP, where applicable. Th e application value for this analysis was zero given the need to contrast patterns of maintenance (either the use of a full sibilant or an aspirant) with patterns of deletion. Th e results of this multivariate analysis appear in Table 10 . As the results from this analysis clearly show, the deletion of plural / s / is favored for head and post-head constituents occurring in second and third positions. Th e opposite is true for pre-head constituents, heads in fi rst position, and single nouns. Th ese patterns correspond directly with those attested to in Scherre and Naro (1998 ) for Brazilian Portuguese. Deletion of plural /s / is statistically probable in right-hand positions whereas it is statistically improbable for constituents in left-hand positions. Th is is an extremely

Table 10 Multivariate analysis of the probabilities of /s / deletion for plural NP constituents in border Uruguayan Spanish according to linear and relative positions within the NP (p<.05, N = 1078, Input = 0.117, Log likelihood = –402.373) Factor N % [ø] Factor Weight Head, Position 2 317 28.9 .75 Head, Position 3 38 28.7 .75 Post-head, Position 3 34 26.5 .73 Post-head, Position 2 52 13.5 .54 Pre-head, Position 2 24 8.3 .41 Pre-head, Position 1 304 6.6 .35 Head, Position 1 33 6.1 .33 Single Noun 276 5.8 .32

Downloaded from Brill.com09/23/2021 10:14:22AM via free access M. Waltermire / Journal of Language Contact 4 (2011) 26–55 45 important fi nding in that it confi rms that lexical category alone is not the strongest predictor of deletion. Th e deletion of plural /s / is not contingent upon morphological status alone, but rather a more complex interplay of factors related to a constituent’s position within the NP. As Table 10 clearly shows, heads favor deletion as long as they occupy non-initial positions. Th e same is true for post-heads (adjectives), which by their very nature do not occur in initial positions. As predicted, deletion of plural /s / for single nouns is highly disfavored with a probability weight of only .32. Th e same patterns of retention of plural /s / for NP constituents found in Scherre and Naro ( 1998 ) for Brazilian Portuguese were also found for the Spanish of Rivera by Carvalho ( 2006b ). Using a corpus of colloquial border Spanish that predates that used for the current study, she found that “pre-head elements in both positions 1 and 2 retain plural marking” (Carvalho, 2006b : 158). Results from the multivariate analyses of all three studies are compared in Table 11 . Probability weights for the deletion of plural /s / for the current study show an inverse ranking to those found for the retention of plural /s / in Brazilian Portuguese by Scherre and Naro (1998 ) and Rivera Spanish by Carvalho (2006b ). Given that there is no overlap between factor weights for deletion and retention, it can now be confi rmed that the plural marking of NP constituents is conditioned in exactly the same way in the Spanish of Rivera as it is in Brazilian Portuguese. Th e morphological conditioning of /s / deletion in the Spanish of Rivera seems to be a direct result of contact with Portuguese. Further corroborating these claims are the fi ndings of Waltermire (2006 ). When morphological status was considered simultaneously with

Table 11 Comparison of probability weights for / s / deletion for plural NP constituents for the current study with probability weights for / s / retention in Brazilian Portuguese (Scherre and Naro, 1998 ) and Rivera Spanish (Carvalho, 2006b ) according to relative and linear positions within the NP

Factor Braz. Portuguese Rivera Spanish

Head, Position 2 .75 .20 .29 Head, Position 3 .75 .27 .21 Post-head, Position 3 .73 .15 .23 Post-head, Position 2 .54 .28 .47 Pre-head, Position 2 .41 .84 .69 Pre-head, Position 1 .35 .88 .71 Head, Position 1 .33 .67 .64

Downloaded from Brill.com09/23/2021 10:14:22AM via free access 46 M. Waltermire / Journal of Language Contact 4 (2011) 26–55

phonological factors (following phoneme, word stress, word position), morphological status was determined to be the most signiﬁ cant factor in the deletion of this consonant with a range of 39. Th ough there are dialects of Spanish that demonstrate similar patterns of plural marking for complex NPs, especially Caribbean dialects, these patterns have not been attested to for monolingual Uruguayan Spanish, in which the deletion of /s / is the result of synchronic phonological erosion.

9.2 Plural Marking of Preceding Noun Phrase Constituents

Another signifi cant factor in the maintenance and deletion of plural /s / for NP constituents in Brazilian Portuguese is the marking of preceding elements of the NP. Scherre (2001 : 96) found that “an immediately preceding zero plural leads almost categorically to a following zero plural” in Brazilian Portuguese. Given the importance of this constraint, it was decided that a fourth multivariate analysis should be conducted using the same factors explored by Scherre ( 2001 ). Each of the plural tokens for the current data set was coded according to the plural marking patterns of preceding elements in the NP as described by Scherre ( 2001 ). Th e only plural marking pattern that was not included from Scherre’s original analysis was for the string “N ø _” (preceding number + zero) since no examples of this string were encountered in the data set. Although Scherre (2001 ) does not include the string “ø _” in her analysis, it has been included here in order to further test the claim that immediately preceding zeroes condition the deletion of subsequent plural markers. Examples of the plural marking patterns of preceding NP constituents in the current data set are displayed in Table 12. In addition to plural tokens occurring in two or three-element strings, single nouns and other elements with no preceding constituents were also included in the multivariate analysis. It is expected that plural / s / tokens for these elements will be retained given the signifi cance of the position constraint analyzed in the previous section. Th e results of the fi nal multivariate analysis including the factors just described appear in Table 13 . Th e probability weights generated during multivariate analysis reveal that plural /s / deletion is statistically probable for all factors except for single nouns (.32) and other elements with no preceding constituents (.35). Constituents preceded by cardinal numbers, regardless of the presence or absence of a fi nal / s / or whether they occur as part of a two or three-element string, also favor deletion in the Spanish of Rivera. Th is provides even stronger evidence in support of the claim that initial (left-hand) elements of NPs will likely retain overt plural marking while plural markers for non-initial elements,

Downloaded from Brill.com09/23/2021 10:14:22AM via free access M. Waltermire / Journal of Language Contact 4 (2011) 26–55 47

Table 12 Examples of plural marking patterns of preceding NP constituents in the Spanish of Rivera (S = [s] or [h]; N = cardinal number) Plural Marking Example S _ los casamientos ‘the weddings’ [loh.ka.sa.mjén.toø] (28/A:347/14) ø _muchas cosas ‘many things’ [mú.tʃaø.kó.saø] (22/A:380/32) S S _las causas importantes ‘the important causes’ [lah.káw.sa.him.por.tán.teø] (04/A:166/03) S ø _muchos compañeros brasileros ‘many Brazilian friends’ [mú.tʃoh.kom.pa.ñé.roø.ßra.si.lé.roø] (08/A:185/07) ø S _ todos los cuadros ‘all of the pictures’ [tó.ðoø.los.kwá.ðroø] (24/A:178/34) S N _los dos lados ‘both sides’ [loz.doz.lá.ðoø] (37/A:340/59) N S _dos maneras parecidas ‘two similar ways’ [doh.ma.né.rah.pa.re.sí.ðaø] (06/A:180/05) Number ending in / s/dos olimpiadas ‘two Olympics’ [do.so.lim.pjá.ðaø] (24/A:146/34) Number not ending in / s/siete pesos ‘seven pesos’ [sjé.te.pé.soø] (37/A:283/58)

Table 13 Multivariate analysis of the probabilities of / s / deletion for plural NP constituents in border Uruguayan Spanish according to the plural marking patterns of preceding constituents (p<.05, N = 1078, Input = 0.116, Log likelihood = –387.750) Factor N % [ø] Factor Weight S ø _ 7 71.4 .95 ø _ 28 57.1 .91 Number not ending in / s/ 34 50 .88 N S _ 4 50 .88 ø S _ 3 33.3 .79 S N _ 15 33.3 .79 Number ending in / s/ 42 31 .77 S _ 288 18.8 .64 S S _ 44 15.9 .59 No preceding constituent 337 6.5 .35 Single noun 276 5.8 .32 (S = [s] or [h]; N = cardinal number)

Downloaded from Brill.com09/23/2021 10:14:22AM via free access 48 M. Waltermire / Journal of Language Contact 4 (2011) 26–55 regardless of the plural marking patterns of preceding constituents, will likely be deleted. Th is is true for two and three-element strings that include preceding plural markers (S _ and S S _, respectively), which favor deletion with probability weights of .64 and .59. Th is result is expected for the two-element strings as this pattern is consistent with patterns of plural marking for Brazilian Portuguese. With respect to the three-element construction with two preceding explicit plural markers, this is somewhat unexpected and does not refl ect a process that is typical of Brazilian Portuguese. Scherre (2001 ) found that the deletion of plural / s / is disfavored for these types of strings (with a factor weight of .35). In Brazilian Portuguese there is only a preference for the deletion of plural / s / in three-element strings of the type “S ø _”. As with the Spanish of Rivera, this is the string type that most highly favors plural / s / deletion. Th ese results also provide strong evidence in support of Scherre’s claim that immediately preceding zeroes condition plural / s / deletion for subsequent constituents. Both strings with immediately preceding zeroes highly favor plural /s / deletion as demonstrated by the extremely high probability weights for these factors (.95 for “S ø _”; .91 for “ø _”). Th ough these results corroborate Scherre’s claims, it should be noted that she did not include the string “ø _” for analysis. She did, however, examine the deletion of plural / s / for embedded NPs of this type and found that deletion is also statistically signifi cant for these elements in Brazilian Portuguese. In order to more eff ectively compare the results of these two analyses, probability weights generated for both analyses are provided in Table 14 .

Table 14 Comparison of probability weights for / s / deletion for plural NP constituents for the current study with those for Brazilian Portuguese (Scherre, 2001 ) according to the plural marking of preceding constituents Factor Braz. Portuguese S ø _ .95 .93 ø _ .91 ---- Number not ending in / s/ .88 .39 N S _ .88 .65 ø S _ .79 .60 S N _ .79 .58 Number ending in / s/ .77 .43 S _ .64 .50 S S _ .59 .35 No preceding constituent .35 ---- Single noun .32 ---- (S = [s] or [h]; N = cardinal number)

Downloaded from Brill.com09/23/2021 10:14:22AM via free access M. Waltermire / Journal of Language Contact 4 (2011) 26–55 49

A comparison of probability weights for these two analyses reveals several similarities between plural marking patterns in Brazilian Portuguese and Rivera Spanish. For three-element strings in both data sets (with the exception of “S S _” strings), the deletion of / s / for fi nal constituents is statistically favored. Th ough non-initial constituents favor plural /s / deletion in the Spanish of Rivera categorically to lesser and greater degrees, deletion is disfavored in Brazilian Portuguese for NP constituents immediately following cardinal numbers and for those that are preceded by two overtly marked plurals (S S _). With regards to the presence of cardinal numbers in two-element strings, the probabilities generated for Brazilian Portuguese are counterintuitive and were left unexplained in Scherre ( 2001 ). According to her, “With regard to linguistic parallelism, I consider that this fact is not adequately interpretable. If numbers as a whole do not exhibit overt plural makers, we would expect them to favor more zeroes” (Scherre, 2001 : 97). Th is is exactly what is shown for the Spanish of Rivera, with probability weights for deletion of constituents following numbers at .88 for numbers not ending in / s / and .77 for numbers that do end in / s/. With respect to the deletion of fi nal constituents of “S S _” strings, the two varieties in question display opposite tendencies, with deletion favored in the Spanish of Rivera (with a probability weight of .59) while disfavored in Brazilian Portuguese (with a probability weight of .35). Th is discrepancy may be due to the relatively small number of tokens for this factor for the current study (N=44) when compared to those used in Scherre (2001 ) (N=439). A greater number of tokens would provide more reliable results with regards to these types of strings. On the other hand, this diff er- ence could simply be attributed to diff ering processes for both language varieties. Th is makes sense given that the Spanish of Rivera is impacted by competing infl uences from monolingual varieties of Spanish, in particular that of Montevideo, and multiple varieties of Brazilian Portuguese. Regardless of this discrepancy, the comparison of the results from these two studies shows unequivocally that similar patterns of retention and deletion exist for the Spanish of Rivera and Brazilian Portuguese, which is expected given the long- standing contact that has existed between these two languages in northeastern Uruguay.

10 Conclusion

Th e results from this study provide evidence that the deletion of / s / in the Spanish of Rivera is conditioned by Portuguese plural marking patterns

Downloaded from Brill.com09/23/2021 10:14:22AM via free access 50 M. Waltermire / Journal of Language Contact 4 (2011) 26–55 and high levels of token frequency in discourse. As with Portuguese, plural marking is carried out by leftmost NP constituents in the Spanish of Rivera while plural / s / is deleted for subsequent constituents. Th e deletion of plural /s / is likely for all non-initial positions irrespective of the plural marking patterns of preceding constituents. Probabilities for plural /s / deletion, however, are much greater for tokens that are immediately preceded by a zero, which further corroborates the parallel processing theory laid out in Scherre (2001 ). Interestingly, in the Spanish of Rivera, the use of an aspirant for overtly marked plural determiners is common. In this way, the use of a typically Spanish variant is employed in morphological constructions that are characteristic of Brazilian Portuguese, which seems to be unique to this dialect and a direct result of language contact. Th is fi nding supports the idea that bilinguals possess rich, elaborate phonological repertoires that result from the use of two distinct linguistic systems. Th e deletion of syllable and word-fi nal / s / in Rivera is also favored for words of high token frequency while it is disfavored for words of low token frequency. Somewhat surprisingly, this frequency eff ect was not found for the aspiration of this phoneme, which is favored for words of low token frequency. Th is may be the result of a social eff ect. From a social perspective, the aspiration and deletion of syllable and word-fi nal / s / represent opposite sides of the spec- trum. Unlike the aspiration and full realization of syllable and word-fi nal / s/, which are characteristic of prestige varieties of Uruguayan Spanish (Carvalho, 2006a ; Elizaincín, 1992; Pedretti de Bolón, 1983 ; Ricci, 1963 ; Vásquez, 1953 ), the deletion of /s / is highly stigmatized (Lipski, 1994 ; Fontanella de Weinberg, 1992 ). Deletion may be even more stigmatized in border varieties of Uruguayan Spanish due to its association with border Portuguese. As such, young middle class speakers, in particular those with professional occupations, and students, most of whom have had some formal education, use aspirated variants of / s / more frequently than other groups within the community (Carvalho, 2006a ; Waltermire, 2006 ). Since higher levels of education equate with more expansive lexical repertoires, which would likely include a greater number of low frequency words, it is possible that these speakers use words of low frequency to a greater extent than members of the community who have had less formal education. Conversely, speakers with less formal education would tend to use words of higher frequency, which favor deletion, more often. Th ough the social conditioning of deletion has yet to be determined for this community, it does seem likely that deletion is more common among speakers who prefer Portuguese and use this language often, despite its lack of prestige and its associations with mixed, border varieties of Uruguayan Spanish.

Downloaded from Brill.com09/23/2021 10:14:22AM via free access M. Waltermire / Journal of Language Contact 4 (2011) 26–55 51

References

Alba , Orlando . 1990 . Variación fonética y diversidad social en el español dominicano de Santiago . Santiago , República Dominicana : Pontifi cia Universidad Católica Madre y Maestra. Azevedo , Milton M . 1981 . A Contrastive Phonology of Portuguese and English . Washington, D.C. : Georgetown University Press. Azevedo , Milton M . 1989 . Vernacular features in educated speech in Brazilian Portuguese . Hispania 72 : 862 - 872 . Barrios , Graciela . 1991 . Dos variables sociolingüísticas: (s) y (z) en Montevideo . Unpublished ms. , Universidad de la República. Barrutia , Richard and Armin Schwegler . 1994 . Fonética y fonología españolas . New York : John Wiley and Sons. Brown , Esther . 2004 . Th e Reduction of Syllable-initial /s / in the Spanish of New Mexico and Southern Colorado: A Usage-based Approach . PhD dissertation , University of New Mexico . Brown , Esther and Rena Torres Cacoullos . 2002 . ¿Qué le vamoh aher? : Taking the syllable out of Spanish / s / reduction . In Daniel Ezra Johnson and Tara Sanchez (eds.), Working Papers in Linguistics: Papers from NWAV 30 , 17 - 32 . Philadelphia : University of Pennsylvania Press. Bybee , Joan . 1985 . Morphology: A Study of the Relation between Meaning and Form . Philadelphia : John Benjamins. Bybee , Joan . 2000 . Th e phonology of the lexicon: Evidence from lexical diff usion . In Michael Barlow and Suzanne Kemmer (eds.), Usage-based Models of Language , 65 - 85 . Palo Alto : CSLI Publications, Stanford University . Bybee , Joan . 2001 . Phonology and Language Use . Cambridge : Cambridge University Press. Bybee , Joan , Revere Perkins , and William Pagliuca . 1994 . Th e Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World . Chicago : University of Chicago Press. Calvis de Bon , Elena . 1987 . Aproximaciones al habla de Montevideo . In Humberto López Morales and María Vaquero (eds.), Actas del Primer Congreso Internacional sobre el Español de América , 619 - 632 . San Juan : Academia Puertorriqueña de la Lengua Española . Carvalho , Ana Maria . 1998 . Th e Social Distribution of Uruguayan Portuguese in a Bilingual Border Town . PhD dissertation , University of California , Berkeley . Carvalho , Ana Maria . 2003a . Th e sociolinguistic distribution of (lh) in Uruguayan Portuguese: A case of dialect diff usion . In Silvina Montrul and Francisco Ordóñez (eds.), Selected Papers of the 5th Hispanic Linguistics Symposium: Advances in Hispanic Linguistics , 30 - 44 . Somerville : Cascadilla Press . Carvalho , Ana Maria . 2003b . Rumo a uma defi nição do português uruguaio. Revista Internacional de Lingüística Iberoamericana 2 : 135 - 159 . Carvalho , Ana Maria . 2004 . ‘I speak like the guys on TV’: Palatalization and the urbanization of Uruguayan Portuguese . Language Variation and Change 16-2 : 127 - 151 . Carvalho , Ana Maria . 2006a . Spanish (s) aspiration as a prestige marker on the Uruguayan- Brazilian border . Spanish in Context 3-1 : 85 - 114 . Carvalho , Ana Maria . 2006b . Nominal number marking in a variety of Spanish in contact with Portuguese . In Timothy Face and Carol Klee (eds.), Selected Proceedings of the 8th Hispanic Linguistics Symposium , 154 - 166 . Somerville, MA : Cascadilla Proceedings Project . Cedergren , Henrietta . 1973 . Th e Interplay of Social and Linguistic Factors in Panama . PhD dissertation , Cornell University. Cedergren , Henrietta . 1978 . En torno a la variación de la S fi nal de sílaba en Panamá: Análisis cuantitativo . In Humberto López Morales (ed.), Corrientes actuales en la dialectología del Caribe hispánico: Actas de un simposio , 35 - 50 . Hato Rey, Puerto Rico : Editorial Universitaria, Universidad de Puerto Rico .

Downloaded from Brill.com09/23/2021 10:14:22AM via free access 52 M. Waltermire / Journal of Language Contact 4 (2011) 26–55

Davies , Mark. 2002 . Corpus del Español (100 million words, 1200s-1900s) . Available online at : http://www.corpusdelespanol.org . D’Introno , Francisco and Juan Manuel Sosa . 1986 . Elisión de la / d / en el español de Caracas: Aspectos sociolingüísticos e implicaciones teóricas . In Rafael Núñez-Cedeño , Irazet Páez Urdaneta , and Jorge Guitart (eds.), Estudios sobre la fonología del español del Caribe , 135 - 163 . Caracas : Ediciones La Casa de Bello . Donni de Mirande , Nélida . 1996 . Argentina-Uruguay . In Manuel Alvar (ed.), Manual de dialectología hispánica , 209 - 221 . Barcelona : Editorial Ariel . Eckert , Penelope . 1997 . Age as a sociolinguistic variable . In Florian Coulmas (ed.), Handbook of Sociolinguistics , 151 - 167 . Oxford : Blackwell . Elizaincín , Adolfo . 1976 . Th e emergence of bilingual dialects on the Brazilian-Uruguayan border . International Journal of the Sociology of Language 9 : 123 - 134 . Elizaincín , Adolfo . 1992a . Dialectos en contacto: Español y portugués en España y América . Montevideo : Arca . Elizaincín , Adolfo . 1992b . El español actual en el Uruguay . In César Hernández Alonso (ed.), Historia y presente del español de América , 159 - 174 . Valladolid : Pabecal . Elizaincín , Adolfo , Luis Behares , and Graciela Barrios . 1987. Nos falemo brasilero . Montevideo : Editorial Amesur . Fontanella de Weinberg , María Beatriz . 1992 . El español de América . Madrid : MAPFRE . Guy , Gregory . 1981 . Parallel variability in American dialects of Spanish and Portuguese . In David Sankoff and Henrietta Cedergren (eds.), Variation Omnibus , 85 - 95 . Edmonton : Linguistic Research . Haiman , John . 1994 . Ritualization and the development of language . In William Pagliuca (ed.), Perspectives on Grammaticalization , 3 - 28 . Amsterdam : Benjamins . Hensey , Frederick . 1972 . Th e Sociolinguistics of the Brazilian-Uruguayan Border . Th e Hague: Mouton . Hooper , Joan Bybee . 1976a . Introduction to Natural Generative Phonology . New York : Academic Press . Hooper , Joan Bybee . 1976b . Word frequency in lexical diff usion and the source of morphophonological change . In William Christie (ed.), Current Progress in Historical Linguistics , 96 - 105 . Amsterdam : North Holland . Jurafsky , Daniel , Alan Bell , Michelle Gregory , and William D. Raymond . 2001 . Probabilistic relations between words: Evidence from reduction in lexical production. In Joan Bybee and Paul Hopper (eds.), Frequency and the Emergence of Linguistic Structure , 229 - 253 . Amsterdam : Benjamins . Koch , Walter , Mário Silfredo Klassmann , and Cléo Vilson Altenhofen . 2002 . Atlas Lingüístico- Etnográfi co da Região Sul do Brasil. Porto Alegre: Universidade Federal do Rio Grande do Sul. Krug , Manfred . 1998 . String frequency: A cognitive motivating factor in coalescence, language processing and linguistic change . Journal of English Linguistics 26 : 286 - 320 . Langacker , Ronald W . 2000 . A dynamic usage-based model . In Michael Barlow and Suzanne Kemmer (eds.), Usage-based Models of Language , 1 - 3 . Palo Alto : CSLI Publications, Stanford University . Lipski , John . 1987 . Fonética y fonología del español de Honduras . Tegucigalpa : Editorial Guaymuras . Lipski , John . 1988 . La discontinuidad fonética como criterio dialectológico . Th esaurus 43 : 310 - 326 . Lipski , John . 1994 . Latin American Spanish . New York : Longman. López Morales , Humberto . 1983 . Estratifi cación social del español de San Juan de Puerto Rico . Ciudad de México, D.F. : Universidad Nacional Autónoma de México.

Downloaded from Brill.com09/23/2021 10:14:22AM via free access M. Waltermire / Journal of Language Contact 4 (2011) 26–55 53

Pedretti de Bolón , Alma . 1983 . El idioma de los uruguayos: Unidad y diversidad . Montevideo : Ediciones de la Banda Oriental. Penny , Ralph . 1991 . A History of the Spanish Language . Cambridge : Cambridge University Press . Poplack , Shana . 1979 . Function and Process in a Variable Phonology . PhD dissertation , University of Pennsylvania. Poplack , Shana . 1986 . Acondicionamiento gramatical de la variación fonológica en un dialecto puertorriqueño . In Rafael Núñez-Cedeño , Irazet Páez Urdaneta , and Jorge Guitart (eds.), Estudios sobre la fonología del español del Caribe , 95 - 107 . Caracas : Ediciones La Casa de Bello . Ricci , Julio . 1963 . Un problema de interpretación fonológica en el español del Uruguay . Montevideo : Instituto de Estudios Superiores de Montevideo . Rona , José Pedro . 1965 . El dialecto fronterizo del norte del Uruguay . Montevideo : Universidad de la República. Scherre , Maria Marta Perreira . 2001 . Phrase-level parallelism eff ects on noun phrase number agreement . Language Variation and Change 13 : 91 - 107 . Scherre , Maria Marta Perreira and Anthony Julius Naro . 1998 . Sobre a concordância de número no portugués falado do Brasil . In Giovanni Ruffi no (ed.), Dialettologia, geolinguistica, sociolinguistica. Atti del XXI Congresso Internazionale di Linguistica e Filologia Romanza , 509 - 523 . Tübingen : Max Niemeyer Verlag . Stavrou , Christopher . 1947 . Brazilian Portuguese Pronunciation . Philadelphia : David McKay Company. Terrell , Tracy . 1978 . La aspiración y elisión de / s / en el español porteño . Anuario de Letras 16 : 41 - 66 . Vásquez , Washington . 1953 . El fonema / s / en el español uruguayo . Montevideo : Facultad de Humanidades y Ciencias. Waltermire , Mark . 2006 . Social and Linguistic Correlates of Spanish-Portuguese Bilingualism on the Uruguayan-Brazilian Border . PhD dissertation , University of New Mexico . Waltermire , Mark . 2008 . Social stratifi cation and the use of language-specifi c variants of intervocalic / d / along the Uruguayan-Brazilian border . Sociolinguistic Studies 2-1 : 31 - 60 .

Downloaded from Brill.com09/23/2021 10:14:22AM via free access 54 M. Waltermire / Journal of Language Contact 4 (2011) 26–55

Appendix

Word Frequency Counts for Coded Rivera Data (variable / s/) Source: Codes s (Microsoft Excel document)

Ser (sos, es, somos) (313) (93,413) Más (163) (31,504) Los (158) (80,665) Estar (estoy, estás, está, estamos, están, estuvo, estuvieron, estaba, estábamos, estuviera) (149) (43,915) Las (86) (52,093) Entonces (70) (14,470) Tener (tenés, tenemos, tendrías, tendríamos) (53) (6,452) Este (49) (16,470) Gustar (gusta, gustan, gustó, gustaba, gustaría) (47) (4,598) Ellos (45) (5,840) Ir (vas, vamos, fuiste, fuimos) (45) (16,441) Español (42) (1,194) Cosas (41) (7,496) Portugués (41) (35) Nosotros (40) (7,181) Vez/veces (38) (11,019) Mismo(s)/a(s) (30) (9,076) Dos (29) (9,608) Muchos/as (29) (8,506) Todos/as (28) (10,898) Después (26) (6,555) Inglés (25) (892) Estudiar (estudiar, estudio, estudia, estudiamos, estudian, estudié, estudiaron, estudiaba, estudiando) (24) (3,142) Años (23) (8,183) Esto/a(s) (28) (13,178) Nos (22) (13,739) Vos (22) (2,265) Tres (20) (4,994) Brasileños, brasileros/as (19) (33) Escuela(s) (19) (1,799) Saber (sabés, sabemos, sabíamos) (19) (1,471) Estado(s) (18) (7,125) Nuestro(s)/a(s) (18) (10,103) Menos (17) (6,014) Uruguayos/as (17) (9) Otros/as (16) (7,287) Digamos (15) (2,463) Usted(es) (15) (14,944)

Downloaded from Brill.com09/23/2021 10:14:22AM via free access M. Waltermire / Journal of Language Contact 4 (2011) 26–55 55

Bastante (13) (2,615) Unidos (13) (1,634) Viste(s) (13) (335) Desde (12) (7,632) Hasta (12) (7,129) País(es) (12) (6,602) Chiquinines (11) (0) Escuchar (escucho, escucha, escuchamos, escuchaba, escuchara, escuchando) (11) (690) Hablar (hablás, hablamos, hablaste, hablábamos, hables) (11) (632) Seis (11) (1,680) Hacer (hacés, hacemos, hicimos, hacías) (10) (1,262) Hijos/as (10) (1,643) Poder (podés, puedes, podemos) (10) (3,124) Problemas (10) (2,627)

Downloaded from Brill.com09/23/2021 10:14:22AM via free access