Printer Friendly

THE TRILL ISN'T GONE: RHOTIC VARIATION IN SOUTHERN NEW MEXICAN SPANISH.

1. INTRODUCTION. The Spanish language has been spoken in what is modern-day New Mexico since the initial forays into the territory by Spanish colonizers coming north from Mexico at the turn of the sixteenth century. The first permanent settlement of significant importance to the Spanish conquest was Santa Fe, which was founded in 1610. As the final destination from Mexico to New Spain along the Camino Real, Santa Fe proved to be crucial to the influx of a great number of Spanish speakers into what is now the southwestern United States. All aspects of the Spanish of northern New Mexico have been researched extensively, from its phonological characteristics to its lexical characteristics (Bills & Vigil 2008, Clegg 2010, Clegg & Waltermire 2009, Cobos 1983, de la Puente Schubeck 1991, Espinosa 1909, Fernandez 1990, Hills 1906, Lance 1975, Ornstein 1975, Rael 1937, Torres Cacoullos & Ferreira 2000, Torres Cacoullos & Travis 2010). Pioneering studies of northern New Mexican Spanish date back to the beginning of the twentieth century (Espinosa 1909, Hills 1906), before New Mexico even achieved statehood (in 1912). Aurelio M. Espinosa (1909) provides a thorough examination of northern New Mexican Spanish in his series Studies in New Mexican Spanish, which consists of three parts (Part 1: Phonology, Part 2: Morphology, and Part 3: The English Elements). To date, this remains one of the most comprehensive descriptions of northern New Mexican Spanish. More recently, Garland Bills and Neddy Vigil (2008) have provided an exhaustive account of lexical variation in the Spanish spoken in the state, including the varieties that are spoken in the southern part of the state, in The Spanish Language of New Mexico and Southern Colorado: A Linguistic Atlas. This work was based on linguistic interviews carried out with 357 Spanish speakers from the region. As such, it represents the most rigorous study of New Mexican Spanish.

1.1. DIALECTS OF NEW MEXICAN SPANISH. The Spanish spoken in northern New Mexico is often referred to as "traditional" Spanish by many linguists. According to Bills & Vigil (2008:5), speakers of this dialect "represent early settlement prior to the twentieth century and today reside primarily in the upper Rio Grande drainage area of central and northern New Mexico." The more traditional (and in some cases archaic) characteristics of northern New Mexican Spanish are likely the result of historical isolation from other areas of New Spain and the varieties of Spanish spoken in northern Mexico. This isolation was due in no small part to the fact that it was extremely dangerous to cross the terrain from the Mesilla Valley in southern New Mexico to areas of northern New Mexico in earlier times (Bills & Vigil 2008). Subsequent settlements of the southern part of the state, most notably in Las Cruces, which occurred much later, have been more heavily influenced by Mexican varieties of Spanish. Due to the independent linguistic developments resulting from these historical circumstances, linguists have posited two major dialects of New Mexican Spanish - the traditional Spanish of northern New Mexico and the more typically Mexican variety of Spanish spoken in the southern part of the state.

Despite the abundant research that has been conducted on northern New Mexican Spanish, the Spanish of southern New Mexico has been studied only sporadically, with the majority of research focusing mainly on language use and maintenance. Aside from the lexical variation detailed in Bills & Vigil (2008), there have been no thorough descriptions of southern New Mexican Spanish. To date, the research on the phonology of southern New Mexican Spanish is practically nonexistent (with the lone exception being Willis 2005). This relative lack of research may be due to the fact that southern New Mexican Spanish is often assumed to be identical to the varieties of Spanish in northern Mexico, especially Chihuahua, despite the fact that it is a largely bilingual dialect as most New Mexicans also speak English.

1.2. RHOTIC VARIATION IN NEW MEXICAN SPANISH. With the present study, we will attempt to show that the variety of Spanish spoken in southern New Mexico is distinct from that of northern Mexico with respect to its phonological characteristics. More specifically, this research will focus on the variable use of rhotics in southern New Mexican Spanish and in particular the seemingly low use of the multiple trill. In standard or general Spanish, there are two rhotic phonemes - the voiced alveolar tap (also known as a "flap") /r/ and the voiced alveolar multiple trill /r/. Both Spanish rhotic phonemes share the same point of articulation and voicing. The difference between the two is that the tap is produced with a quick tapping gesture of the pre-dorsum of the tongue against the alveolar ridge whereas the trill involves the production of two or three brief closures of the vocal tract as a result of vibrations of the tongue tip caused by considerable back pressure (Blecua 2001, Olsen 2012:71, Schwegler et al. 2010:278, Sole 2002). Following Sole (2002:656), "once trilling is initiated, tonguetip vibration is maintained as a self-sustaining vibratory system." In determining what constitutes a trill, the most important factor to consider is whether there is more than one closure of the vocal tract. When there is not, a tap is easily discerned from a trill. Determining the exact number of closures for trills is of little importance to variationist sociolinguistics, which focuses on the use and conditioning of sound variants rather than their acoustic properties. Though forays into the intersection of acoustic phonetics and sociolinguistics have been pursued in more recent sociophonetic approaches to variation, they will not be pursued in the current research since the goal is to determine the sociolinguistic conditioning of trills among Spanish-English bilinguals not their production.

The phoneme /r/ has two standard allophones, an alveolar tap [r] and an alveolar trill [r]. The phoneme /r/ has only one allophone, the alveolar trill [r], which is used for orthographic <rr>, thereby establishing minimal pairs in word-medial intervocalic contexts (e.g. cero 'zero' ['se.ro] vs. cerro 'hill' ['se.ro]). According to prescriptive accounts of the language, the use of the trill for /r/ is obligatory, though this is clearly not the case in actual speech. One of the most commonly used variants of /r/ in regional varieties of Latin American Spanish is a groove fricative (Lipski 1994). Though this variant is non-existent in the speech of the participants for the current study, neutralization is quite common. In the current data, the Id of words like arriba 'up/above' [a.'ri.[beta]a] and tierra 'land' ['tje.ra] are often realized as taps rather than trills.

Unlike IT/, the conditioning of Id allophones (represented orthographically as <r>) is relatively complex. There is substantial dialectal variation in the use of syllable and word-final Id throughout Latin America, ranging from deletion and lateralization throughout the Caribbean to devoicing and assibilation in many parts of Central and South America (Lipski 1994). Though use of the trill is variable in syllable and word-final contexts, these contexts will not be pursued here for two reasons - (1) the use of a trill in these contexts is virtually non-existent in the current data and (2) the tap, not the trill, is the prescribed standard variant for this context; trilled final Id is seen as emphatic (Schwegler et al. 2010:284). By rule in standard Spanish, the [r] allophone can only occur and is obligatory in word-initial position as in Rodriguez [ro. 'ori.yes] and in syllable-initial position following a consonant as in alrededor 'around' [al.re.oe.'oor]. (1) Examples of variation in the use of taps and trills in these variable contexts for the current study appear in Figure 1.
Phonological Context          Example

Absolute initial position     revolvio 'stirred 3sg.' [re.(3ol.'Pjo]
                              (16/0:05:50)
                              Ramona 'Ramona' [ra.'mo.na]
                              (13/1:08:55)
Syllable-initial after vowel  hile rojo 'red chile' [r/i.le.'ro.ho]
                              (17/0:45:05)
                              esa razon 'that reason' [e.sa.ra.'son]
                              (06/0:07:10)
Syllable-initial after
consonant                     alrededor 'around' [al.re.6e.'6or]
                              (05/0:42:34)
                              tan rapido 'so quickly' [tan.Va.pi.6o]
                              (16/0:27:35)

Figure 1. Variable contexts for tapped and trilled /r/.


According to some researchers (Rael 1937:79-80, Chacon Vigil 1993:58), the use of rhotics in New Mexican Spanish follows patterns of "standard" variant usage, though Figure 1 shows that this is clearly not a reality. A greater number of studies, fortunately, acknowledge that the use of rhotics presents a much wider range of variation (Bills 1997, Bills & Vigil 1999, 2008, Espinosa 1909, Lipski 2008, Vigil 2008). Even going back to the original studies of Espinosa, one finds mention of several different rhotic variants in the Spanish of New Mexico, including a lateralized variant in intervocalic contexts and a devoiced aspirated variant found when Id follows a voiceless consonant. Lipski (2008:204) mentions the use of a groove fricative variant instead of the alveolar trill in the Spanish of New Mexico but does not specify its geographical distribution or its frequency of use compared to that of other rhotic allophones.

Since New Mexican Spanish has been in consistent contact with English since the mid-1800s, one would also expect to find some level of English influence in the articulation of rhotic phonemes. Several researchers have reported the use of an English-like alveolar approximant [J] in the Spanish of northern New Mexico (Bills 1997, Bills & Vigil 1999, Espinosa 1909, Lipski 2008, Vigil 2008). This was first reported by Espinosa, who claimed that the production of rhotics sometimes "approaches the western American English final "r", but the tongue is not arched upward and backward as is often the case with the American sound" (1909:102). The current authors have observed the use of this sound in the Spanish of southern New Mexico too, though its use does not seem to be very common. Though the use of [[??]] may not be very frequent in this variety, a simplification of the multiple trill articulations of both /r/ and /r/ in words such as Enrique 'Henry' [en.'ri.ke], el radio 'the radio' [el.'ra.ojo], and carro 'car' ['ka.ro] seems to be very common. Since the multiple trill does not characterize varieties of American English, it seems to be losing ground in this contact variety in favor of the alveolar tap (which is consequently very common in American English for unstressed medial /t/, as in 'city' ['si.ri] (Ladefoged & Johnson 2011:64)). Foster (1976) observed this among a group of Mexican-American students at the University of Texas at El Paso. He claims that there is "confusion of /r/ with /r/" among these bilingual speakers (Foster 1976:26). To date, no study has focused on the loss of the alveolar trill in southern New Mexican Spanish. The objective of the current study is to determine the linguistic and social conditioning of syllable-initial rhotics in the Spanish of southern New Mexico. It is anticipated that the multiple trill that characterizes monolingual varieties of northern Mexican Spanish will be simplified and replaced by a single tap among bilinguals of the region due to an indirect influence from English, in which a trill is not used.

2. FIELD WORK AND METHODOLOGY. This research is based on 15 interviews conducted by Mayra Valtierrez with 15 Spanish-English bilinguals who currently live in Dona Ana County, New Mexico, one of the southernmost counties of the state, which is located just northwest of El Paso, Texas along the U.S./Mexico border. The first criterion used in selecting potential participants was that they self-identified as bilinguals. Then, their places of birth were considered. All participants were born and raised in the Mesilla Valley that cuts through most of Dona Ana County and includes Las Cruces. It was important to include participants who were both born and raised in this region in an effort to collect data that accurately portray the use of Spanish among bilinguals of southern New Mexico. Participants indicated that they have spent very little time outside of this community, although it should be noted that many of them have spent enough time outside of the area to understand that their use of Spanish differs from that of other bilinguals in the United States, Mexico or other communities within New Mexico. Once these factors had been identified, interviewees were chosen according to the demographic characteristics of the community as described in the 2000 census data for Dona Ana County (U.S. Census Bureau 2008), which ensured a relatively representative sample of the community. The two demographic factors used in the selection of interviewees were sex and age. Of the 15 participants, there are eight females and seven males ranging from ages 24 to 74.

The level of formality during data collection is an important factor to consider given that as the level of formality decreases the likelihood that the speaker will produce colloquial language increases (Holmes 2008, Labov 1984). Social setting influences the level of formality during data collection; and because of this, all of the interviews were conducted either in a participant's home or place of business. An attempt was made to avoid settings that might lead a speaker to use a more standard form of Spanish, such as the researcher's office or any laboratory-like setting. The interviews were recorded with a digital voice recorder that was placed out of view during interviews so as to put interviewers at greater ease. To ensure the highest levels of informality possible, participants were allowed to talk about any topic of their choice throughout the interviews, though all interviews did focus on the participants' experiences in the area.

Another important factor that contributed to lowering the level of formality was the interviewer's familiarity with the community and its particular linguistic reality. The interviewer was raised in the Mesilla Valley and used a local variety of Spanish during the interviews with which participants would closely identity. This is particularly important in research conducted with bilinguals within the United States since many bilingual speakers tend to identify their Spanish as "not very good or incorrect" (Beaudrie & Ducar 2005). Participants were asked to speak Spanish during the interviews, although at times they would switch to English. Even so, the interviewer spoke primarily Spanish in an effort to collect as much Spanish data as possible. To further maintain a low level of formality during interviews, the interviewer took time before interviews to engage participants in polite conversation that centered on positive feelings and attitudes about the community and the need to represent the linguistic reality of bilinguals of the area. This helped to narrow the social distance between the interviewer and the participants and create a sense of solidarity between them.

The use of code-switching by participants was quite common during these informal conversations that preceded interviews. This alone indicates a low level of formality since code-switching by bilinguals often carries covert prestige among ingroup community members, who generally express unfavorable views of language mixing when asked directly about its use (Holmes 2008). An attempt to move the interviewer and the participants from a distant connection to a more intimate one was important since the relationship between speakers in a conversation has a noticeable effect on the language variety that speakers use (Holmes 2008). Unfortunately, when a researcher is regarded as an expert and there is a distant or low-solidarity relationship, difficulties can arise when attempting to collect representative speech samples since participants are more likely to use a formal variety. Overall, considerations of social setting and social distance (solidarity) in lowering the level of formality are necessary given that the use of spontaneous, colloquial speech provides the most accurate representation of a speaker's phonological repertoire.

3. USE OF RHOTICS IN THE COMMUNITY. A total of 388 tokens were extracted from the colloquial speech of the 15 participants for the current study. Both authors, who have extensive experience with the identification of phonological variants, extracted and identified an approximately equal number of tokens. A goal of thirty tokens was desired for each participant, though in several cases this number of tokens could not be reached due to limitations presented by the variable context, which is extremely narrow. All tokens were extracted according to the variable context in which the [r] allophone would be found in standard or general Spanish (i.e. in syllable-initial contexts; see Section 1.2). In addition to [r] and [r], two other rhotic allophones were encountered in the database - a voiced alveolar approximant [[??]] and an assibilated variant [r]. Neither of these allophones occurs with great frequency in the database, but their use by participants will be analyzed alongside the tap and the trill in the current study. It should be noted here that no distinction was made between voiced and voiceless variants as the aim of the current research is not to explore more finely graded acoustic properties of rhotics, but to analyze their sociolinguistic conditioning among bilinguals.

To ensure reliability in the identification of rhotic variants, given that this was largely done by ear, an Inter-Rater Reliability check was conducted in which both researchers independently listened to and identified rhotic variants from a random selection of 100 tokens. These were then compared for both researchers in order to ensure sufficient levels of reliability in the correct identification of tokens. A rate of accuracy of 92% was achieved between both researchers. Since this is just outside the margin of error of 95%, both researchers refined the criteria used in the correct identification of tokens. When determining if a token was a tap or a trill, if a single articulatory gesture was not detected, it was coded as a trill. (2) During the process of token extraction, all contentious tokens (N=160) were subjected to acoustic analysis using Praat (Boersma & Weenink 2011). The articulator/ characteristics of a typical trill used for the accurate identification of these phones follow those described by Sole (2002) and Olsen (2012) in which "at least two successive closures of the vocal tract [are] evident" (Olsen 2012:71). The general realization of all tokens for the current study is displayed in Table 1.

As shown in Table 1, the multiple trill is not the most commonly used variant (at only 42.8%), though its use greatly exceeds those reported for other varieties (see Hammond 1999), even that of northern New Mexico, for which Vigil (2008) found a rate of use of merely 26%. These results provide further evidence that neutralization of the trill is occurring in all varieties of Spanish and that this should by no means be stigmatized as some sort of aberrant phonological feature due to partial acquisition, imperfect learning or the like. Somewhat ironically, the use of trills is more frequent in southern New Mexican Spanish, a bilingual dialect, than it is in many monolingual dialects (see Hammond 1999). The use of an assibilated variant ([r]) or an alveolar approximant ([[??]]) by participants was extremely rare (at 1% each). Furthermore, the use of these variants occurs in very limited contexts. In all four cases in which an assibilated variant was used, it occurred after /s/ and is therefore the result of progressive assimilation. The use of an English-like alveolar approximant occurred for names (Ruben and Rudy) or English-origin loans (root beer) in three of the four cases (with the lone exception being sierra). Given the extremely low frequency of use of these variants and the rather obvious nature of their conditioning, their use will not be further explored in the current study. That said, the goal of the current study will be to fully explain the sociolinguistic conditioning of multiple trills in southern New Mexican Spanish (as opposed to the use of taps). To accomplish this, all extracted tokens were coded manually for multiple linguistic and social factors, which will be described presently.

4. LINGUISTIC FACTORS. Each of the 388 tokens used in the current analysis was coded for relevant social and linguistic factors. First we will discuss the relevant linguistic factors that may condition the use of the multiple trill [r] in southern New Mexican Spanish. The independent linguistic variables that will be analyzed in this study are word position, word stress, orthographic representation, preceding phonological segment, word frequency, and cognate status.

4.1. WORD POSITION. Word position has been shown to be a crucial factor in consonant reduction and deletion in the world's languages (Escure 1977). The most immediately apparent example of consonant reduction and deletion in Spanish is for the voiced obstruents /b/, /d/, and /g/. These phonemes, which have resulted from voicing of intervocalic /p/, /t/, and /k/ from Latin, have two primary allophones in all dialects of Spanish - a fricative articulation as well as a stop articulation. The use of these allophones occurs in complementary distribution, with stop variants occurring in absolute initial position and after nasal consonants and fricative variants occurring in all other contexts. More important to the current study, however, is the fact that deletion of syllable-initial fricative variants occurs almost exclusively in word-medial position whereas deletion is almost non-existent in word-initial position. Consider, for example, the utterance De los dos lados 'On/from both sides'. Whereas it is fairly common in many dialects of Spanish to delete word-medial Idl of the word lados (as ['laus]), the deletion of both absolute-initial /d/ (in this case de) and word-initial /d/ (in this case dos) is practically non-existent (*[e.lo.soz.'la.5os]). This is due to the relative lack of articulatory strength needed to produce word-medial consonants compared to word-initial (especially absolute-initial) consonants. For this reason, it is expected in the current analysis that the multiple trill will be more frequent in word-initial contexts whereas reduced variants are expected in word-medial contexts.

4.2. WORD STRESS. Related to word position is word stress. Though many words in Spanish consist of three syllables or more, the majority of words in Spanish consist of only two syllables. Also, the majority of words, regardless of syllable quantity, receive penultimate stress. Of the 27, 211 vowel-final non-verbs examined in Barkanyi (2002), 23,940 receive penultimate stress (88.0%) while the remainder receive antepenultimate stress (10.8%) or final stress (1.2%). (3) Therefore, the most common word type in Spanish is disyllabic with stress placed on the penultimate syllable. This pattern of stress is so common that when native speakers are tasked with assigning stress to nonce words, the overwhelming majority of assignments occur on the penultimate syllable (Waltermire 2004). Following Penny (1991:35), "energy on the tonic (and the greater audibility this brings) allows tonic vowels to be quite well differentiated and preserved, while lesser degrees of energy devoted, in decreasing order, to initial, final and intertonic vowels imply greater degrees of merger and loss." Consequently, the articulatory force necessary to stress tonic syllables leads to a strengthening in articulation of the onset. In other words, stress "inhibits lenition" (Colantoni & Marinescu 2010:108). Since unstressed syllables require less articulatory effort, the production of

4.3. ORTHOGRAPHIC REPRESENTATION. According to prescriptive accounts of Spanish, the orthographic representation <rr> must be realized obligatorily as a multiple trill. It is this articulation, for this orthographic representation, that has led the vast majority of linguists to claim that Spanish has two rhotic phonemes. This status is easily confirmed with a minimal pairs test (e.g. morro 'hill' ['mo.ro] vs. moro 'moor' ['mo.ro]). Regardless of whether these are two distinct phonemes or not (see Nunez-Cedeno & Morales-Front 1999 for a counterargument), the use of the multiple trill for "double r" is readily apparent in the vast majority of monolingual dialects of Spanish, though there is some evidence that the use of this articulation may not be as common as prescriptive accounts lead us to believe (Hammond 1999). (4) This may be somewhat surprising for monolingual dialects, despite the simplification and reduction of other word-medial consonants (see previous discussions on word position and stress), but it should come as no surprise that Spanish-English bilinguals simplify and reduce these articulations since dialects of American English do not include a multiple trill. With regards to the current study, none of the participants uses a trill while speaking English as this articulation does not exist in local varieties of this language. For this reason, "double r" will likely be produced less often as a multiple trill among these speakers than as a tap.

4.4. PRECEDING PHONOLOGICAL SEGMENT. All rhotics for the current study occurred syllable initially, which greatly diminishes the possibility of deletion of phonological variants, which tends to occur syllable finally. That said, however, levels of strengthening and weakening may depend on the phonological context of Id within discourse. That is, when occurring post-consonantally (as in Enrique [en.'ri.ke]), Id may receive a different pronunciation than in postvocalic contexts (such as la rosa 'the rose' [la. ro.sa]). Given that the coordination of two consonants requires greater muscle control and coordination between articulators, it is suspected that post consonantal Id will be reduced from a trill to a tap, despite the fact that the multiple trill is the standard articulation in this context. When occurring in pre-vocalic, word-initial contexts, it is likely that the multiple trill will be maintained. This is also expected to be the case in absolute initial position. This is expected due to the articulatory effort required to produce consonants in these contexts.

4.5. WORD FREQUENCY. Many recent approaches to phonological phenomena have utilized a usage-based approach that accounts for word frequency as a primary determinant of articulatory reduction (Brown 2004, Bybee 1985, 2000, 2001, Bybee et al. 1994, DTntrono & Sosa 1986, Hooper 1976a, 1976b, Jurafsky et al. 2001, Krug 1998). Each of these studies has shown that phonetic articulations and syllables of frequently occurring words and morphemes tend to reduce. This being the case, it is expected that rhotics will be produced as taps more often for words of high frequency while they will tend to be produced as multiple trills in words of low frequency. Frequency levels were determined using Mark Davies' A Frequency Dictionary for Spanish (2006). This dictionary, which contains the 5,000 most frequently occurring words in the Spanish language, was based on a 20-million word corpus evenly divided among spoken language samples and written (both fiction and non-fiction) texts from Spain and Latin America. For the purposes of the current study, it was determined that there should be a distinction between high, mid, and low frequency words. Since the vast majority of words used in the database do occur as entries in the Davies dictionary, it was decided that not all of these words should be considered highly frequent. Of course, it could be debated whether or not some of these words are truly frequent, every day words. Admittedly, some of them (e.g. redaccion 'editing', reanudar 'to renew', retroceso 'backwards', etc.) are not. For this reason, it was decided that the list of the 5,000 most frequently occurring words should be divided in two in order to account for gradations of frequency, with the 2,500 most frequent words being considered words of truly high frequency and everyday usage while the next 2,500 most frequently occurring words would be considered words of mid frequency. Low frequency words are any words that do not appear in the Davies dictionary. Once these cut-off points had been determined, coding for frequency was fairly straightforward. The coding process was aided greatly given that all 5,000 words appearing in the dictionary are arranged alphabetically in the dictionary index (Davies 2006:183-234).

4.6. COGNATE STATUS. Since bilinguals have access to two phonetically divergent lexical forms (Torres Cacoullos & Ferreira 2000), it is important to consider whether words of a similar shape in both English and Spanish will have a conditioning role in the use of trills among participants. There are many cognates shared by English and Spanish due to the shared historical contribution of Latin to the formation of both languages. Though English did not derive directly from Latin (as in the case of Spanish), there are numerous loanwords of Latin origin that form an important part of the English language. These words present competing phonological influence in the speech of bilinguals, who have access to two distinct phonological realizations of essentially the same word. Speakers' access to typically English realizations of these words could potentially, therefore, result in more English-like articulations. In other words, the Id in a word such as Ramon, which exists in both languages (Ramon in Spanish) could be realized as a tap or an alveolar approximant since multiple trills are not used in English. Non-cognates, on the other hand, do not present competing phonological realizations among speakers. A word like correr 'to run' [ko.'rer], since it has a completely different form in English, would not be susceptible to influence from English. For these reasons, it is expected that there will be a greater incidence of trill use for non-cognates since this articulation does not exist in English. Though the use of a tap, which is expected for cognates, cannot be attributed directly to English influence, since this articulation would not be typical among native English speakers for syllable-initial /r/, the avoidance of the trill in this context may be an indirect result of speaking English.

4.7. RESULTS: LINGUISTIC FACTORS. We will now turn to a quantitative analysis of general patterns of use of multiple trills with respect to the linguistic factors just described. The realization of trills according to these factors appears in Table 2.

As previously stated, the orthographic representation <rr> only occurs word medially in Spanish. Out of the 177 word-medial tokens, the vast majority (175, or 98.9%) receive this written representation. The only example of word-medial <r> that would prescriptively be pronounced as a trill (see Figure 1) was the word alrededor 'around', which was produced twice. This extremely low rate of occurrence provides strong evidence that words of this type are highly uncommon in Spanish, but it does not mean that they should be discarded from the current analysis given the importance of word position as a determining factor in the strengthening and reduction of consonants in this language. Though the raw numbers for the use of the trill for word position and orthographic representation are nearly identical, their relative strength (if any) in the conditioning of use of trills among bilinguals will be determined by way of multivariate analysis. It is likely that the greater use of trills in word-medial contexts is due to their almost exclusive orthographic representation as <rr>. If not for this competing factor, we would expect articulations for rhotics to be reduced in word-medial contexts rather than strengthened. Since word stress is not contingent upon orthographic representation, the greater use of trills in stressed syllables (at 50.5% compared to 35.4% use in unstressed syllables) corresponds with our predictions. The less frequent use of trills after consonants (at 29.5% compared to over 40% in absolute initial and postvocalic contexts) also follows our predictions for the reasons laid out in Section 4.4. With respect to word frequency, rates of trill realization are very similar by level of frequency, thereby making this a statistically non-significant factor group. Finally, rhotics of non-cognates received a trilled pronunciation notably more than those for cognates (49.7% compared to 35.8%). These words are more highly associated with Spanish articulations than are cognates, which have competing phonological representations among bilinguals since they exist in both languages. Though most of our predictions related to these linguistic factors are borne out in the data, their relative strength on the conditioning of variation remains to be seen. It is possible that social factors play a greater role in the conditioning of trills, for example, or that certain factors exhibit less influence than we had thought. Of course, these questions will be answered once we have conducted multivariate analyses of both linguistic and social factors.

Rates of trill use have given us some indication as to the importance of each of the linguistic factors just presented on the use of multiple trills in the Spanish of southern New Mexico. However, in order to determine the statistical significance and relative contributions that each factor presents with respect to the realization of multiple trills, we must conduct a multivariate analysis. This was accomplished using GoldVarb X (Sankoff et al. 2005), a statistical analysis program that generates probability weights corresponding to observed frequencies in a corpus. This program determines the relative contribution of factors on the occurrence of a linguistic variant when all factors are considered together. GoldVarb then assigns a probability to each factor in terms of its relative contribution to the occurrence of the selected variant and determines whether the effect of each factor group is statistically significant. The results of this analysis are shown in Table 3.

Of the six linguistic factors examined in the current study, only orthographic representation and word stress were selected as significant. It was expected that orthographic <rr> would tend to favor multiple trills since this is the context most highly associated with the trill. Of all contexts, it is easily identifiable and often receives the majority of attention among speakers concerning "correct pronunciation". The use of trills for <r> would be less obvious and, given the narrow contexts for its use, less easily distinguished in these contexts. The use of a trill for a single written <r> is statistically unlikely among bilingual speakers (with a factor weight of .39) for these very reasons. The effect of word stress, though not as strong (with a range of only 12), was somewhat more surprising, but one that falls in line with findings from previous research showing the use of strengthened articulations in stressed syllables. The use of multiple trills is statistically likely for stressed syllables (with a factor weight of .56) while somewhat unlikely for unstressed syllables (with a factor weight of .44). This is due to the articulatory strength required for the production of stressed syllables, and in particular syllable onsets (as in the case of the rhotics presented in this study).

5. SOCIAL FACTORS. Since the current study focuses on the variable use of rhotic allophones in the speech of bilinguals living in southern New Mexico, the analysis of social factors is necessary. There are several factors of seeming importance to the study of this kind of variation--namely, how often participants use both languages and the nature of their language backgrounds while growing up. Related to the use of both languages, age may also be a significant factor in the conditioning of rhotic variation since multiple studies have shown that younger generations of New Mexican speakers use Spanish less often than their parents or grandparents do (Bills 1997, Bills & Vigil 1999, Hernandez-Chavez et al. 1996, Hudson et al. 1995). The variable use of rhotics may also be related to education level and the sex of participants, but the strength of these factors is not yet known.

5.1. SEX. A total of eight females and seven males participated in the current study. This slightly greater representation of females is due to the fact that the Dona Ana County population, according to 2008 census data, is 50.6% female and 49.4% male. Sex was used as opposed to gender in selecting participants for interviews. It is important to note that sex is based on biological characteristics, while gender is a social phenomenon that is associated with female and male individuals and is rooted in the social and cultural beliefs of a given community. As stated by Bonvillain, "it is through concepts of gender that society transforms female and male human beings into social women and men, assigning them roles and giving them cultural value" (1997:166). In the current study gender identity was not discussed during data collection with participants since gender is a complex "social construct" (Bonvillain 1997:166) and since this study is concerned with phonological characteristics of a general community sample rather than the use of language by different genders.

Analyses of sociolinguistic data consistently support the notion that women generally use and produce standard language more commonly than men do (Chambers 2003, Holmes 2008, Horvath 1985, Labov 1972, 2001, Milroy & Milroy 1997, Poplack et al. 1988), and with regards to pronunciation that "females use standard and prestige pronunciations at higher rates than do males of similar age and social or racial groups" (Bonvillain 1997:171). It is then expected for women who participated in this research to use trills more often than men. This expectation should be tempered, however, by the fact that the greater use of prestige variants by females is only a general tendency. Several studies have shown that women sometimes use supra-locally stigmatized variants, as long as they carry some sort of covert prestige within the community (Fontanella de Weinberg 1979, Rissel 1989, Trudgill 1972).

5.2. AGE. Participants for the current study were grouped into three generations. The first generation includes participants between the ages of 20 and 40, the second generation includes participants from ages 41 to 60, and the third generation includes participants from ages 61 to 80. These generations were determined based on shared life experiences of participants within the given community, as well as demographic information from the 2000 census on the age of the population of Dona Ana County (U.S. Census Bureau 2008). With regard to shared life experiences, some of the participants indicated that they had children or great-grandchildren that belonged to one of the other generations. For example, one of the participants between the ages of 61 to 80 explained that she had a daughter in her 40s, a grandson in his 20s and great-grandchildren that are toddlers. The shared life experience of starting a family was therefore chosen as the means to determine generation, since in this particular community average age at which people become parents and start a family is 20 to 25.

It is also important to consider that the use of non-standard variants decreases when individuals enter the work force (Holmes 2008, Sankoff & Laberge 1978) due to external pressures to conform to the norms associated with more formal environments. In the case of bilinguals, the two languages are used much more during the time spent in the work force when the use of the two languages is beneficial (Holmes 2008). Given that language use changes during the time in which an individual is in the workforce, we would expect that those participants between the ages of 41 to 60 will produce trills more often than individuals from the other two age groups.

5.3. EDUCATION LEVEL. Participants' level of education was also considered and discussed during data collection, along with information on whether participants had received any formal education in Spanish. According to 2000 census data for Dona Ana County, the level of education among the local population rarely exceeds a high school degree and the median value of owned homes is S90,000. Data from the 2000 census reveal that only 22.3% of individuals in the community hold a bachelor's degree or higher while 70% of individuals have a high school education (U.S. Census Bureau 2008). Thus, the two levels of education chosen to classify participants in the community were high school, which includes less than high school as well as any vocational training and/or an associate's degree, and college, which includes a bachelor's and/or graduate degree. Vocational training and associate's degrees were included with the high school classification given that this type of training rarely includes a requirement for a second language and thus engaging in a formal education in Spanish (Dona Ana Community College 2011). The majority of participants (10 out of 15) expressed that they had received little to no formal education in Spanish, aside from elementary ESL programs in which Spanish was not the focus but was instead used in order to transition them into English. However, five of the participants who received a bachelor's degree or higher (which includes seven of the 15 participants) had received some formal education in Spanish through required course work in college. It is then expected that those individuals who have had a formal education in a university setting will produce trills at higher rates due to having received conventional instruction in the use of standard Spanish (Erickson 1984, Mesthrie et al. 2009). After all, it is through formal education that most individuals learn written and oral prescriptive language conventions.

5.4. FREQUENCY OF SPANISH USE. A questionnaire was used to determine the frequency of Spanish use among participants (see Appendix). Participants were only given the questionnaire after completing interviews. The questionnaire prompted them to indicate their percentage of use of Spanish in different domains of the community with different interlocutors. The domains included were home, work or school, church, and restaurants or bars. Interlocutors included were spouse or significant other, children, parents, relatives, friends, and bosses or teachers. These particular domains and interlocutors were chosen because they represent the most common places where participants use Spanish along with the speakers who most often occupy them. The respondents could choose a percentage from 10 to 100 in increments of 10%. A 0% value was omitted as an option in the questionnaire since Spanish is used in a large variety of contexts and to varying degrees by bilinguals in the community due to its proximity to the U.S./Mexico border. However, if a participant indicated that their use was indeed 0% this number was honored.

Frequency of Spanish use was calculated by taking an average of the participants' ten responses. Participants' frequency of use was then divided into three categories, the first being greater than 70%, the second between 30% and 70%, and the last less than 30%. It is expected that speakers demonstrating a higher frequency of Spanish use will produce multiple trills more often than participants who use this language less frequently. This is expected since Spanish-dominant individuals frequently access typically Spanish articulations on a daily basis while those who speak this language less frequently lack ready access to these articulations and tend to use articulations that are more characteristic of English. The former two categories (> 70% and 30%-70%) were collapsed after running initial analyses in order to better assess the potential impact of a relatively frequent use of Spanish (> 30%) as opposed to a less frequent use of this language (< 30%).

5.5. PARENTS' BIRTHPLACE. The birthplace of participants' parents is another factor that must be considered in determining the likelihood of use of multiple trills within the community. The place of birth and nationality of participants' parents were discussed by participants during interviews in describing where they learned to speak Spanish. However, parents' birthplace was chosen as an indicator for the use of multiple trills instead of nationality since birthplace is much more likely to provide clues to the use of Spanish in the home during formative years by participants given that many of the participants' parents have become naturalized U.S. citizens. The birthplace of participants' parents was coded by the use of three categories - Mexico, United States, or both, seeing as some participants have one Mexican-born parent and one U.S.-born parent. After an initial analysis, we decided to compare participants with two Mexican-born parents as opposed to having one or more U.S.-born parents. This allows for a clearer assessment of the effect that having Mexican-born parents might have on the use of the multiple trill.

The birthplace of participants' parents is an important consideration in the current study since those individuals that were born in Mexico are most likely to produce a multiple trill for syllable-initial Id, given that this is the typical pronunciation in monolingual Mexican Spanish (Lipski 1994:279). Furthermore, and in view of the fact that children learn phonological systems by reproducing the sounds that are used around them during infancy (Salzmann et al. 2012), one would expect that those participants who grew up with Mexican-born parents had more exposure to the trill than those whose parents were born in the U.S. Given that parents contribute greatly to their children's phonological acquisition, it is expected that those participants whose parents were both born in Mexico will use trills to a greater extent than those with either both U.S.-born parents or one parent born in this country.

5.6. RESULTS: SOCIAL FACTORS. The general realization of multiple trills according to the social factors just described will indicate whether or not the use of this typically Spanish articulation is related to specific social categories and, by extension, certain social groups. The rates of use of multiple trills by speakers of these social categories are displayed in Table 4.

As can be observed from the results shown in Table 4, men produced multiple trills at a slightly higher rate (44.7%) than women did (40.9%). It was expected that women would produce trills more often than men given the number of studies that show that women tend to produce standard variants and articulations more often than men. The results with respect to speaker sex, however, are not statistically significant. In considering the age of participants, the expected results correlate with those outlined in Section 5.2. Trill use is very uncommon among younger speakers (at a rate of merely 20.2%) given this group's preference for English. Participants of the second generation (between the ages of 41 and 60) produced trills at a rate of 52.6%, presumably because they are still in the workforce and have a need to use standard speech for professional reasons. The third generation, on the other hand, consists of speakers who have recently retired or who will retire soon and, for this reason, trails slightly behind the second generation at 43.5%.

Factors related to language use and experience yielded mixed results. The social factor of level of education yielded results that were contrary to the expected outcome. Instead of those participants with a college education producing trills at a higher rate, as was expected, it was actually those with less than a college education that produced [r] at a greater rate (48.2%, as compared to 35.7% for those with a college education). It is difficult to provide a definitive reason for this result, although one can speculate that it may be due to other social circumstances. One possibility is that college-educated speakers have greater fluency in English, but this is doubtful given that all speakers for the current study are highly fluent in both languages. Another possibility is that this is due to the overall belief held by many in the area that there is a need to produce language-specific sounds such as a trill to identify as a Spanish speaker. During interviews, some participants expressed their current or past troubles with producing certain sounds that pertain to Spanish such as the trill and their efforts to "correct" this. Perhaps the use of this articulation by speakers with less formal education is the result of pressures to use this articulation, which may be felt less by college-educated speakers.

Unlike level of education, the results for frequency of Spanish use and its correlation with trill production do coincide with the expected results. Those individuals who use Spanish 30% or more of the time produced trills at a rate of 48.8%. This is over twice as often as the rate for those speakers who use Spanish less than 30% of the time (only 23.1%). Since these speakers access phones from English more often than from Spanish, their use of trills is inhibited, showing a clear preference among these speakers for the tap.

The last social factor considered was participants' parents' place of birth. It was anticipated that participants with parents who were both born in Mexico would produce trills at a higher rate than those individuals with both parents born in the U.S. or a combination of one born in the U.S. and another in Mexico. The difference in the rate of trill production according to this factor differed markedly, as was expected. Those with Mexican-born parents produced trills at a rate of 51.9%, compared to those with U.S.-born parents or a combination, who produced trills at a rate of 31.6%. This result is similar to that for the use of Spanish since amount of exposure to a certain language inevitably influences the way in which a bilingual will speak either language; but rather than being based on production/output, it is related to perception/input.

Though these numbers point to several social effects on the use of the multiple trill, in particular generation, Spanish use, and parents' birthplace, the relative impact of each of these factors remains to be determined. We do not expect that education level will have much of an effect on the use of the multiple trill. In order to assess the relative contribution of these factor groups and the conditioning effect of individual factors on the use of multiple trills, we will now turn to a second multivariate analysis of social factors. This type of analysis will uncover hierarchies of conditioning among each of the independent social variables that may have a conditioning effect on the dependent variable ([r]). The results of this multivariate analysis appear in Table 5.

As clearly shown by these results, the use of multiple trills in this variety of Spanish is largely conditioned by social factors (Spanish use, generation, and parents' birthplace), which have higher ranges overall (22 to 28) than those for linguistic factors (12 and 24; see Table 3). The greater number of statistically significant social factor groups, combined with these greater ranges, show that the conditioning of multiple trills is not motivated primarily by linguistic factors. This is an important finding. Speakers who use Spanish often (more than 30% of the time) will likely produce a multiple trill (shown by the probability weight of .57) while the use of this variant is statistically unlikely (with a factor weight of only .29) among those who speak this language infrequently. The potential loss of the multiple trill in this variety is also being impeded by older speakers (who are predicted to use this articulation with factor weights of .58 and .56 for second and third generation speakers, respectively). This is no surprise given that older speakers in New Mexico use Spanish more often than younger speakers. The slightly greater use of the trill among participants of the middle generation is most likely due to workplace pressures to use standard articulations. It is highly unlikely that young speakers will use this articulation (with a factor weight of only .31), which almost certainly has to do with the fact that younger speakers speak Spanish less often, on average, than do older speakers. It is logical to assume that speakers with Mexican-born parents continue to use (as well as receive input from) this language frequently, thereby making parents' birthplace also a significant factor group in the use of multiple trills in this variety. The exposure to English produced by at least one U.S.-born parent from an early age would inhibit production and accurate perception of the multiple trill if in fact articulations from English were more common among speakers while growing up. Multivariate analysis shows that it is statistically unlikely that these speakers will produce a trill (with a probability weight of .38).

6. CONCLUSION. Results from this study indicate that there may be a gradual neutralization of rhotic variants in bilingual Spanish toward single taps. This is only a tendency, however, which is more common among English-dominant younger speakers who grew up with at least one U.S.-born parent. This being the case, this decreased likelihood of using multiple trills may be due in large part to an indirect influence from English, in which multiple trills are not used, since there is a competing phonological influence from this language in the speech of these individuals. Managing a phonological repertoire that includes competing phones from both languages presents a higher cognitive load for bilingual speakers who have access to each of the articulations from the languages that they speak. Furthermore, the physical production of multiple trills requires considerably more effort than that required in the production of a single tap. Further favoring the use of taps among English-dominant younger speakers is the fact that this is an extremely common allophone for unstressed medial /t/ in American English, meaning that tap production among these speakers is not problematic, even though the contexts in which taps are used in Spanish may be. For each of these reasons, the neutralization of rhotics toward a single tap may well be complete within a couple of generations. For now, this is being impeded by older speakers who speak Spanish often and who grew up in households with Mexican-born parents. These individuals are keeping the use of multiple trills alive and well for the time being in the Spanish of southern New Mexico. Their rates of use are even higher than those reported for the Spanish of northern New Mexico in Vigil (2008). This may well be due to the fact that the southern dialect has assimilated more characteristics of Mexican Spanish than has the northern dialect. In this sense, the contribution of Mexican Spanish to the Spanish spoken by bilinguals in the southern part of New Mexico should not be underestimated. This study provides further evidence of the strong influence of Mexican Spanish along the border. Though these influences are often associated with the lexicon, we have shown that they also influence pronunciation and contribute to the maintenance of typically Spanish features in an otherwise bilingual dialect.

APPENDIX: QUESTIONNAIRE
Primer nombre:
Nacionalidad de padre:
Profesion:
Nacionalidad de madre:
Frecuencia con que usa el espafiol:

                                %
En casa         10  20  30  40  50  60  70  80  90  100
En el trabajo/  10  20  30  40  50  60  70  80  90  100
la escuela
En la iglesia   10  20  30  40  50  60  70  80  90  100
En bares/       10  20  30  40  50  60  70  80  90  100
restaurantes
Con marido/     10  20  30  40  50  60  70  80  90  100
mujer/novia/o
Con hijos       10  20  30  40  50  60  70  80  90  100
Con padres      10  20  30  40  50  60  70  80  90  100
Con parientes   10  20  30  40  50  60  70  80  90  100
Con amigos      10  20  30  40  50  60  70  80  90  100
Con el jefe/    10  20  30  40  50  60  70  80  90  100
los maestros


MUCHISIMAS GRACIAS.

REFERENCES

BARKANYI, ZSUZSANNA. 2002. A fresh look at quantity sensitivity in Spanish. Linguistics 40.2.1-20.

BEAUDRIE, SARA, and CYNTHIA DUCAR. 2005. Beginning level university heritage language programs: Creating a space for all heritage language learners. Heritage Language Journal 3.1-26.

BILLS, GARLAND. 1997. New Mexican Spanish: Demise of the earliest European variety in the United States. American Speech 72.2.154-72.

BILLS, GARLAND, and NEDDY VIGIL. 1999. Ashes to ashes: The historical basis for dialect variation in New Mexican Spanish. Romance Philology 53.43-60.

BILLS, GARLAND, and NEDDY VIGIL. 2008. The Spanish language of New Mexico and southern Colorado: A linguistic atlas. Albuquerque: University of New Mexico Press.

BLECUA, BEATRIZ. 2001. Las vibrantes del espanol: Manifestaciones acusticas y procesos foneticos. Barcelona: Universidad Autonoma de Barcelona dissertation.

BOERSMA, PAUL, and DAVID WEENINK. 2011. Praat: Doing phonetics by computer [computer software]. Amsterdam: Department of Language and Literature, University of Amsterdam.

BONVILLAIN, NANCY. 1997. Language, culture, and communication: The meaning of messages. 2nd edn. Upper Saddle River, NJ: Prentice Hall.

BROWN, ESTHER. 2004. The reduction of syllable-initial /s/ in the Spanish of New Mexico and southern Colorado: A usage-based approach. Albuquerque: University of New Mexico dissertation.

BYBEE, JOAN. 1985. Morphology: A study of the relation between meaning and form. Philadelphia: John Benjamins.

BYBEE, JOAN. 2000. The phonology of the lexicon: Evidence from lexical diffusion. Usage-based models of language, ed. by Michael Barlow and Suzanne Kemmer, 65-85. Palo Alto, CA: CSL1 Publications, Stanford University.

BYBEE, JOAN. 2001. Phonology and language use. Cambridge: Cambridge University Press.

BYBEE, JOAN; REVERE PERKINS; and WILLIAM PAGLIUCA. 1994. The evolution of grammar: Tense, aspect, and modality in the languages of the world. Chicago: University of Chicago Press.

CHACON VIGIL, LUCY. 1993. The traditional Spanish of Mora, New Mexico. Albuquerque: University of New Mexico dissertation.

CHAMBERS, JACK K. 2003. Sociolinguistic theory. Oxford: Blackwell.

CLEGG, JENS. 2010. An analysis of the motivations for borrowing in the Spanish of New Mexico. Spanish of the U.S. Southwest: A language in transition, ed. by Susana Rivera-Mills and Daniel Villa, 223-37. Norwalk, CT: Iberoamericana Vervuert.

CLEGG, JENS, and MARK WALTERMIRE. 2009. Gender assignment to English-origin nouns in the Spanish of the southwestern United States. Southwest Journal of Linguistics 28.1.1-17.

COBOS, RUBEN. 1983. A dictionary of New Mexico and southern Colorado Spanish. Santa Fe: Museum of New Mexico Press.

COLANTONI, LAURA, and IRINA MARINESCU. 2010. The scope of stop weakening in Argentine Spanish. Selected proceedings of the 4th conference on laboratory approaches to Spanish phonology, ed. by Marta Ortega-Llebaria, 100-14. Somerville, MA: Cascadilla Proceedings Project.

DAVIES, MARK. 2006. A frequency dictionary of Spanish: Core vocabulary for learners. New York: Routledge.

DE LA PUENTE-SCHUBECK, ELSA. 1991. La perdida del modo subjuntivo en el espanol chicano de Nuevo Mexico. Albuquerque: University of New Mexico dissertation.

D'INTRONO, FRANCISCO, and JUAN MANUEL SOSA. 1986. Elision de la /d/ en el espanol de Caracas: Aspectos sociolinguisticos e implicaciones teoricas. Estudios sobre lafonologia del espanol del Caribe, ed. by Rafael Nunez-Cedeno, Irazet Paez Urdaneta, and Jorge Guitart, 135-63. Caracas: Ediciones La Casa de Bello.

DONA ANA COMMUNITY COLLEGE. 2011. Academic catalog with student handbook. Online: http://dabcc.nmsu.edu/publications/Our_Students/2011-12-Catalog.pdf.

ERICKSON, FREDERICK. 1984. School literacy, reasoning, and civility: An anthropologist's perspective. Review of Educational Research 54.4.525-46.

ESCURE, GENEVIEVE. 1977. Hierarchies and phonological weakening. Lingua 43.1.55-64.

ESPINOSA, AURELIO M. 1909. Studies in New Mexican Spanish, Part I: Phonology. University of New Mexico Bulletin/Language Series 1.2.47-161.

FERNANDEZ, ROSA M. 1990. Actitudes hacia los cambios de codigos en Nuevo Mexico: Reacciones de un sujeto a ejemplos de su habla. Spanish in the United States: Sociolinguistic issues, ed. by John J. Bergen, 49-58. Washington, DC: Georgetown University Press.

FONTANELLA DE WEINBERG, MARIA BEATRIZ. 1979. Dinamica social de un cambio linguistico. Mexico City: Universidad Nacional Autonoma de Mexico.

FOSTER, DAVID WILLIAM. 1976. The phonology of Southwest Spanish. Studies in southwest Spanish, ed. by J. Donald Bowen and Jacob Ornstein, 17-28. Rowley, MA: Newbury House.

HAMMOND, ROBERT. 1999. On the non-occurrence of the phone [r] in the Spanish sound system. Advances in Hispanic linguistics, ed. by Javier Gutierrez-Rexach and Fernando Martinez-Gil, 135-51. Somerville, MA: Cascadilla.

HERNANDEZ-CHAVEZ, EDUARDO; GARLAND BILLS; and ALAN HUDSON. 1996. El desplazamiento del espanol en el suroeste de EEUU segun el censo de 1990. Actas del X Congreso lnternacional de la Asociacion de Linguistica y Filologia de la America Latina, ed. by Marina Arjona Iglesias, Juan Lopez-Chavez, Araceli Enriquez Ovando, Gilda Lopez Lara, and Miguel Angel Novella Gomez, 664-72. Mexico City: Universidad Nacional Autonoma de Mexico.

HILLS, ELIJAH CLARENCE. 1906. New Mexican Spanish. Publications of the Modern Language Association of America 21.706-53.

HOLMES, JANET. 2008. An introduction to Sociolinguistics. 3rd edn. London: Pearson Longman.

HOOPER, JOAN BYBEE. 1976a. Introduction to natural generative phonology. New York: Academic Press.

HOOPER, JOAN BYBEE. 1976b. Word frequency in lexical diffusion and the source of morphophonological change. Current progress in historical linguistics, ed. by William Christie, 96-105. Amsterdam: North Holland.

HORVATH, BARBARA. 1985. Variation in Australian English: The sociolects of Sydney. Cambridge: Cambridge University Press.

HUDSON, ALAN; EDUARDO HERNANDEZ-CHAVEZ; and GARLAND BILLS. 1995. The many faces of language maintenance: Spanish language claiming in five southwestern states. Spanish in four continents: Studies in language contact and bilingualism, ed. by Carmen Silva-Corvalan, 165-83. Washington, DC: Georgetown University Press.

JURAFSKY, DANIEL; ALAN BELL; MICHELLE GREGORY; and WILLIAM D. RAYMOND. 2001. Probabilistic relations between words: Evidence from reduction in lexical production. Frequency and the emergence of linguistic structure, ed. by Joan Bybee and Paul Hopper, 229-53. Amsterdam: John Benjamins.

KRUG, MANFRED. 1998. String frequency: A cognitive motivating factor in coalescence, language processing and linguistic change. Journal of English Linguistics 26.286-320.

LABOV, WILLIAM. 1972. Sociolinguistic patterns. Philadelphia: University of Philadelphia Press.

LABOV, WILLIAM. 1984. Field methods of the Project on Linguistic Change and Variation. Language in use: Readings in sociolinguistics, ed. by John Baugh and Joel Sherzer, 8-53. Englewood Cliffs, NJ: Prentice Hall.

LABOV, WILLIAM. 2001. Principles of linguistic change, volume II: Social factors. Oxford: Blackwell.

LADEFOGED, PETER, and KEITH JOHNSON. 2011. A course in phonetics. 6th edn. Boston: Wadsworth Cengage.

LANCE, DONALD. 1975. Spanish-English code switching. El lenguaje de los chicanos, ed. by Eduardo Hernandez-Chavez, Andrew Cohen, and Anthony Beltramo, 138-53. Arlington, VA: Center for Applied Linguistics.

LIPSKI, JOHN. 1994. Latin American Spanish. New York: Longman.

LIPSKI, JOHN. 2008. Varieties of Spanish in the United States. Washington, DC: Georgetown University Press.

MESTHRIE, RAJEND; JOAN SWANN; ANA DEUMERT; and WILLIAM L. LEAP. 2009. Introducing sociolinguistics. Philadelphia: John Benjamins.

MILROY, JAMES, and LESLEY MILROY. 1997. Varieties and variation. Handbook of sociolinguistics, ed. by Florian Coulmas, 47-64. Oxford: Blackwell.

NUNEZ CEDENO, RAFAEL, and ALFONSO MORALES-FRONT. 1999. Fonologia generativa contemporanea de la lengua espanola. Washington, DC: Georgetown University Press.

OLSEN, MICHAEL K. 2012. The L2 acquisition of Spanish rhotics by LI English speakers: The effect of LI articulatory routines and phonetic context for allophonic variation. Hispania 95.1.65-82.

ORNSTEIN, JACOB. 1975. The archaic and the modern in the Spanish of New Mexico. El lengnaje de los chicanos, ed. by Eduardo Hernandez-Chavez, Andrew Cohen, and Anthony Beltramo, 6-12. Arlington, VA: Center for Applied Linguistics.

PENNY, RALPH. 1991. A history of the Spanish language. Cambridge: Cambridge University Press.

POPLACK, SHANA; DAVID SANKOFF; and CHRISTOPHER MILLER. 1988. The social correlates and linguistic processes of lexical borrowing and assimilation. Linguistics 26.47-104.

RAEL, JUAN BAUTISTA. 1937. A study of the phonology and morphology of New Mexican Spanish based on a collection of 410 folk-tales. Palo Alto, CA: Stanford University dissertation.

RISSEL, DOROTHY. 1989. Sex, attitudes, and the assibilation of /r/ among young people in San Luis Potosi, Mexico. Language Variation and Change 1.269-83.

SALZMANN, ZDENEK; JAMES STANLAW; and NOBUKO ADACHI. 2012. Language, culture, and society. 5th edn. Boulder, CO: Westview Press.

SANKOFF, DAVID, and SUZANNE LABERGE. 1978. The linguistic market and the statistical explanation of variability. Linguistic variation: Models and methods, ed. by David Sankoff, 45-55. New York: Academic Press.

SANKOFF, DAVID; SALI TAGLIAMONTE; and ERIC SMITH. 2005. Goldvarb X: A variable rule application for Macintosh and Windows [computer software]. Toronto: Department of Linguistics, University of Toronto.

SCHWEGLER, ARMIN; JUERGEN KEMPFF; and ANA AMEAL-GUERRA. 2010. Fonetica y fonologia espanolas. 4th edn. New York: John Wiley and Sons.

SOLE, MARIA-JOSEP. 2002. Aerodynamic characteristics of trills and phonological patterning. Journal of Phonetics 30.4.655-88.

TORRES CACOULLOS, RENA, and FERNANDA FERREIRA. 2000. Lexical frequency and voiced labiodental-bilabial variation in New Mexican Spanish. Southwest Journal of Linguistics 19.1-17.

TORRES CACOULLOS, RENA, and CATHERINE TRAVIS. 2010. Variable yo expression in New Mexico: English influence? Spanish of the U.S. Southwest: A language in transition, ed. by Susana Rivera-Mills and Daniel Villa, 185-206. Norwalk, CT: Iberoamericana Vervuert.

TRUDGILL, PETER. 1972. Sex, covert prestige and change in the urban British English of Norwich. Language in Society 1.179-95.

UNITED STATES CENSUS BUREAU. 2008. Dona Ana facts and population. Online: http://quickfacts.census.gov/qfd/states/35/350131k.html.

VIGIL, DONNY. 2008. The traditional Spanish of Taos, New Mexico: Acoustic, phonetic, and phonological analyses. Lafayette, IN: Purdue University dissertation.

WALTERMIRE, MARK. 2004. The effect of syllable weight on the determination of spoken stress in Spanish. Laboratory approaches to Spanish phonology, ed. by Timothy L. Face, 171-91. New York: Mouton de Gruyter.

WILLIS, ERIK W. 2005. An initial examination of southwest Spanish vowels. Southwest Journal of Linguistics 24.185-98.

MARK WALTERMIRE AND MAYRA VALTIERREZ

New Mexico State University

Waltermire Department of Languages and Linguistics

New Mexico State University

Las Cruces, NM 88003

[markw@nmsu.edu]

Valtierrez

Department of Languages and Linguistics

New Mexico State University

Las Cruces, NM 88003

[mayrav@nmsu.edu]

(1) The obligatory use of phones is obviously a myth given the considerable amount of research within the past half-decade which shows that the use of phones is highly variable, even for the same speaker. Hammond (1999) has shown that the use of the trill in these contexts for a wide variety of Spanish dialects is highly variable and not very frequent.

(2) This process was not necessary for distinguishing the two other rhotic variants ([ ] and [r]) found in the database. These were quite easily differentiated due to their divergent characteristics.

(3) If verb conjugations were included in this calculation, the percentage of word-final stress would increase slightly due to the numerous preterit conjugations that receive final stress in Spanish. the onset of these syllables also requires less effort, thereby resulting in their reduction and deletion. Given these findings, it is expected that multiple trills will be preserved in stressed syllables and reduced to taps in unstressed syllables.

(4) Hammond (1999) purports that the diverse consultants for his study (N=229) realized /r/ at a rate of merely 1%, which seems artificially low, especially given more recent findings, such as those of Vigil (2008). This unbelievably low number most certainly has to do with the fact that tokens were identified almost entirely by ear and were never analyzed acoustically. According to Hammond (1999:142), "sound spectrographs were prepared for analysis of randomly selected occurrences of the phoneme /r/ for four subjects." Given that each consultant produced only seven rhotic tokens in the reading passage that he administered, this means that of the total 1,603 tokens used in his analysis, fewer than 28 were analyzed acoustically. If 28 tokens were taken at random from the current investigation, it is possible that none of them would be realized as a trill, which necessitates analyzing a much greater proportion of tokens acoustically, certainly more than 1.7% of them.
Table 1. General realization of rhotic variants for all participants.

         Linguistic variant
[r]      [r]      [r]     [[??]]

166/388  214/388  4/388   4/388
(42.8%)  (55.2%)  (1.0%)  (1.0%)

Table 2. Realization of multiple trills according to linguistic
factors.

Factor group       Factor         N       %

Word position
                   Initial      67/211   31.8
                   Medial       99/177   55.9
Word stress
                   Stressed     96/190   50.5
                   Unstressed   70/198   35.4
Orthographic rep.
                   Double <rr>  99/175   56.6
                   Single <r>   67/213   31.5
Preceding segment
                   None         3/7      42.9
                   Vowel        135/286  47.2
                   Consonant    28/95    29.5

Word frequency
                   Low          34/82    41.5
                   Mid          30/68    44.1
                   High         102/238  42.9
Cognate status
                   Cognate      69/193   35.8
                   Non-cognate  97/195   49.7

Word position: [chi square] = 23.0; df=1; p = 0.000; Word stress: [chi
square] = 9.12; df= 1; p = 0.003; Orthographic rep.: [chi square] =
24.8; df= 1; p = 0.000; Preceding segment: [chi square] = 9.16; df= 2;
p = 0.010; Word frequency: [chi square] = 0.108; df=2; p = 0.947;
Cognate status: [chi square] = 7.76; df= 1; p = 0.005

Table 3. Multivariate analysis of the probabilities of use of multiple
trills and linguistic factors in the Spanish of southern New Mexico.

Factor group        Factor      N   %[r]  Factor
                                          Weight

Orthographic rep.
                   Double <rr>  99  56.6  .63
                   Single <r>   67  31.5  .39
                                          Range 24
Word stress
                   Stressed     96  50.5  .56
                   Unstressed   70  35.4  .44
                                          Range 12

p<.05, N = 388, Input = 0.423, Log likelihood = -250.151 Other factor
groups included in analysis: 1) word position, 2) preceding segment,
3) word frequency, and 4) cognate status.

Table 4. Realization of multiple trills according to social factors.

Factor group         Factor            N        %

Sex
                     Female          81/198   40.9
                     Male            85/190   44.7
Generation
                     First (20-40)   19/94    20.2
                     Second (41-60)  110/209  52.6
                     Third (61 >)    37/85    43.5
Education level
                     High school     106/220  48.2
                     College         60/168   35.7
Spanish use
                     > 30%           145/297  48.8
                     < 30%           21/91    23.1
Parents' birthplace
                     U.S./mixed      55/174   31.6
                     Mexico          111/214  51.9

Sex: [chi square] = 0.580; df=1; p = 0.446; Generation: [chi square] =
27.9; df=2;p = 0.000; Education level: [chi square] = 6.05; df= 1; p =
0.014; Spanish use: [chi square] = 18.9; df= 1; p = 0.000; Parents'
birthplace: [chi square]= I6.1; df= 1; p = 0.000

Table 5. Multivariate analysis of the probabilities of use of multiple
trills and social factors in the Spanish of southern New Mexico.

Factor group         Factor           N   % [r]  Factor
                                                 Weight

Spanish use          > 30%           145  48.8   .57
                     < 30%            21  23.1   .29
                                                 Range 28
Generation
                     Second (41-60)  110  52.6   .58
                     Third (61 >)     37  43.5   .56
                     First (20-40)    19  20.2   .31
                                                 Range 27
Parents' birthplace
                     Mexico          111  51.9   .60
                     U.S./mixed       55  31.6   .38
                                                 Range 22

p<.05, N = 388, Input = 0.405, Log likelihood = -238.136
Other factor groups included in this analysis: 1) sex and 2) education
level.
COPYRIGHT 2013 Linguistic Association of the Southwest
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2013 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Waltermire, Mark; Valtierrez, Mayra
Publication:International journal of the Linguistic Association of the Southwest
Geographic Code:1U8NM
Date:Dec 1, 2013
Words:11534
Previous Article:US IMMIGRATION POLICIES AND THEIR EFFECTS ON SPANISH-SPEAKING COMMUNITIES.
Next Article:Formulaic Language and Second Language Speech Fluency: Background, Evidence and Classroom Applications..
Topics:

Terms of use | Privacy policy | Copyright © 2022 Farlex, Inc. | Feedback | For webmasters |