The L2 Acquisition of Spanish Rhotics by L1 English Speakers: The Effect of L1 Articulatory
Routines and Phonetic Context for Allophonic Variation
Author(s): Michael K. Olsen Source: Hispania, Vol. 95, No. 1 (March 2012), pp. 65-82 Published by: American Association of Teachers of Spanish and Portuguese Stable URL: http://www.jstor.org/stable/41440363 . Accessed: 01/09/2014 12:10 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. . American Association of Teachers of Spanish and Portuguese is collaborating with JSTOR to digitize, preserve and extend access to Hispania. http://www.jstor.org This content downloaded from 168.176.5.118 on Mon, 1 Sep 2014 12:10:20 PM All use subject to JSTOR Terms and Conditions The L2 Acquisition of Spanish Rhotics by LI English Speakers: The Effect of LI Articulatory Routines and Phonetic Context for Allophonic Variation Michael K. Olsen University of Pittsburgh, USA Abstract: This article offers a fine-grained investigation of how first-language (LI) phonetics involving English rhotics affect Spanish rhotic production by second-language (L2) learners. Specifically, this study investigates how different LI English rhotic articulatory routines (retroflex-like and bunched-like) and the phonetic context that produces allophonic taps in English affect the accuracy of Spanish rhotic pronunciation by L2 learners. Tap and trill accuracy rates as well as English rhotic articulation were calculated from recordings of forty-eight beginning-level university Spanish students reading texts in Spanish and English. Results from multiple linear regressions show that English rhotic articulation is a significant predictor of trill accuracy and is a predictor of tap accuracy when controlling for amount of Spanish exposure. These results suggest that LI articulatory routines affect the accuracy of tap and trill production. Results from a paired samples /-test show that a significantly high percentage of accurately produced taps in Spanish were found in words that follow the same articulatory context that produces taps in English. These results suggest that an overarching theory of the second language acquisition of phonology should consider how subtle differences in LI articulatory routines and transfer of LI phonetic context of allophones to L2 phonemes influences L2 speech production. Keywords: acquisition expectations/expectaciones de adquisicin, acquisition of rhotics/adquisicin de rticos, LI influence/influencia de LI, L2 phonetics/fontica de L2, L2 phonology acquisition/adquisicin fonolgica de L2, second-language acquisition/adquisicin de segunda lengua, Spanish phonology/ fonologia de espanol, Spanish rhotics/rticos en espafiol 1. Introduction The representations described task that in terms learners and of how how have distinctive the when learner's acquiring acoustic first second material language language is (LI) mapped phonetic (L2) onto phonology different repertoire abstract can and be described in terms of how distinctive acoustic material is mapped onto different abstract representations and how the learner's first language (LI) phonetic repertoire and phonology interact with the L2 phonological system. This study seeks to further examine how LI phonetics influence L2 phonology acquisition. Recently, a number of studies have come to light that have looked at rhotic perception and production in L2 Spanish learners. Face (2006) investigated intervocalic rhotics among intermediate and advanced L2 Spanish learners whose LI was English noting the developmental trajectory of phones used in L2 Spanish rhotics. Rose (2010a) tested the predictions of the Perceptual Assimilation Model (Best 1995) in L2 learners' perceptual discrimination of rhotics and found that perception generally increased with proficiency level. Rose (2010b) described the range of phones that LI English speakers utilize for L2 Spanish rhotics in intervocalic position. The results of this study suggest that L2 learners employ [j] in all Spanish rhotic contexts at first, moving to more [r] -dominant articulations in all rhotic contexts and finally differentiating between [r] and [r] at the more advanced levels.1 While these studies have added to our understanding of what to expect in the development of AATSP Copyright 2012 Hispania 95.1 (2012): 65-82 This content downloaded from 168.176.5.118 on Mon, 1 Sep 2014 12:10:20 PM All use subject to JSTOR Terms and Conditions 66 Hispania 95 March 20 1 2 L2 Spanish rhotics, starting with the LI phones leading to the L2 phones, and have utilized acoustic analysis to categorize rhotics, they did not investigate the influence of LI phonetics on L2 accuracy directly. Another recent study that investigated L2 Spanish rhotics is Hurtado and Estrada (2010). This study investigated linguistic factors (phonological environment and tap/trill pronunciation) as well as various sociolinguistic factors that contributed to the pronunciation of Spanish rhotics by LI English speakers and found that linguistic factors and input received through study abroad and explicit instruction improved pronunciation. This investigation is important because it shows, as did the previously mentioned studies, that L2 Spanish rhotic pronunciation can improve over time. Waltmunson (2005) investigated the relative difficulty of the acquisition of rhotics with respect to Iii and /d/. This study places the acquisition of L2 Spanish rhotics relative to the acquisition of other L2 Spanish phones; which furthers our knowledge of where rhotics are situated in terms of the larger picture of the acquisition of L2 Spanish phonology. Although these studies have contributed to our understanding of the factors that affect the production of Spanish rhotics by LI English speakers, which necessitates the incorporation of LI and L2 phonetics, how LI English articulatory routines and phonetic context for LI allophonic variations influence the acquisition of Spanish rhotics has largely been left out. One LI articulatory phenomenon that may influence the L2 acquisition of Spanish rhotics is the range of articulations that English speakers employ for the English rhotic III. English rhotic articulations are usually thought of as two maximally distinct articulations - retroflex and bunched; although intermediate variation does exist across speakers (Alwan, Narayanan, and Haker 1997; Espy-Wilson et al. 2000). The distinct articulations that LI English speakers employ can be described as within-category subphonemic (as opposed to allophonic) variants since speakers do not perceive differences between articulations, nor are they governed by phonetic context across speakers. In other words, LI English speakers assign a variety of different articulations to one category; in this case /j/. These English within-category differ- ences are relevant to the study of the acquisition of L2 Spanish rhotics because, although these differences are not perceived in the LI, differences in articulatory routine may still affect the accurate production of L2 Spanish rhotics.2 Spanish speakers produce rhotics by raising the apex of the tongue towards the alveolar ridge, making a very brief closure for taps and multiple closures for trills. While dialectal and individual variations do exist, usually occurring as fricatives or approximants (Blecua 2001; Hammond 1999; Lindau 1985), apical taps and trills are generally what English-speaking learners of Spanish are taught in a classroom setting (Face 2006). The important point is that in both rhotics of standard Spanish, speakers raise the apex of the tongue toward the alveolar ridge (Blecua 2001). Because LI English retroflex rhotics require the speaker to raise the tongue apex towards the alveolar ridge, the production of both Spanish rhotics should be facilitated when they are represented in the interlanguage phonology as allophones of 111 since both Span- ish rhotics necessitate a raised tongue apex. On the other hand, production of Spanish rhotics would be impeded when English speakers employ a bunched-like [j] because they are used to the opposite direction of movement of the tongue apex. Another LI English phenomenon that may influence L2 acquisition of Spanish rhotics is the phonetic context that converts the alveolar stops Iii and Idl to taps in English, which are physiologically similar to Spanish taps. While this is a completely different phenomenon than the influence of articulatory routines, because it involves the transfer of an LI phonological rule and not muscle movements, this is another largely unexplored phenomenon in the acquisi- tion of Spanish rhotics. Ladefoged (2006) formulates this production as a phonological rule, stating "alveolar stops become voiced taps when they occur between two vowels the second of which is unstressed" (74). Alternations in atomic and atom exemplify this rule. The word atomic [9'thamik] shows that when Iii is the onset of a stressed syllable, it is realized as [th]. The word atom ['aeram] shows that [r] alternates with [th] when stress falls on a syllable that does not contain Iii in its onset. This content downloaded from 168.176.5.118 on Mon, 1 Sep 2014 12:10:20 PM All use subject to JSTOR Terms and Conditions Olsen / L2 Acquisition of Spanish Rhotics 67 A phonological rule is needed to explain the distribution of taps because they are allophonic in English. Spanish taps are phonemic in the sense that a rule is not needed to explain their distribution. Taps can be realized in both stressed and unstressed environments as exempli- fied in the words mejoro [me'xoro] 'improve (first person singular)' and mejor [mexo'ro] 'improved (third person singular)'. Stress does not influence tap distribution in Spanish as it does in English. While intervocalic taps may be represented phonemically in Spanish, they are neutralized with trills in other contexts and, in that sense, taps are not completely phonemic in the same manner as other Spanish phonemes (e.g., /d/ which has allophonic variants [d] and [] found in complimentary distribution). If the English phonetic context that produces taps influ- ences the production of Spanish taps, learners of Spanish should be able to produce taps more accurately when in environments similar to the predictable environments in which they exist in English (following a stressed syllable and in the onset of an unstressed syllable). Because trills do not exist in Standard American English, no LI phonetic context is predicted to affect their acquisition. Differences in LI articulatory routines, such as retroflex and bunched [j] and the LI phonetic context that produces allophonic taps, may both affect the acquisition of Spanish rhotics. As beginning learners assimilate L2 segments to their existing LI segments, LI phonetics should dominate the production of L2 sounds. Results of this study expand our knowledge of how the LI influences the acquisition of L2 phonology. If both types of LI phonetic phenomena influence L2 production, L2 acquisition of phonology is influenced not only by LI abstract phonological categories, but also by articulatory routines found in the LI as well as phonetic contexts that create allophonic variants, regardless of the similarity of phonemes that the context affects in each language. Colantoni and Steele (2008) discussed the need to incorporate phonetic constraints into hypotheses regarding L2 phonology acquisition. They showed that specific LI articulatory conventions that are not necessarily predisposed by aerodynamics (i.e., tongue shape in English rhotic articulation) also influence L2 speech production. Because of this, it is important to carry out studies on LI phonetic phenomena that may influence the acquisition of L2 phonological systems. Therefore, the hypothesis tested in this study is that LI articulatory routines and LI phonetic contexts influence the production of L2 phonology. The acquisition of Spanish rhotics by English speakers is revelatory regarding the cur- rent hypothesis because distinct predictions can be made regarding the influence of both LI articulatory routines and phonetic context due to the differences in rhotics between these two languages.3 The present study investigates beginning L2 Spanish learners' rhotic production in intervocalic position. Specifically, it examines the possible effect LI articulatory routine differences (i.e., the differing routines involved in retroflex-like articulations vs. bunched-like articulations as described above) have on Spanish rhotic production, and the effect that an English LI phonetic context that produces allophonic variation (i.e., the context that produces taps as allophones of alveolar stops as described above) has on accurate L2 pronunciation to test the hypothesis that both types of phonetic factors contribute to L2 phonology acquisition. This study also provides a better understanding of English-Spanish interlanguage phonology regarding rhotic acquisition of beginners. The specific research questions addressed in this study are: 1 . Does the L 1 within-category difference manifested in different manners of American English rhotic articulation (i.e., retroflex-like or bunched-like) affect the facilitation of Spanish rhotic production? 2. Does the phonetic context that produces allophonic taps in English affect accuracy in Spanish rhotic production? This content downloaded from 168.176.5.118 on Mon, 1 Sep 2014 12:10:20 PM All use subject to JSTOR Terms and Conditions 68 Hispania 95 March 20 1 2 2. Interlanguage Phonology System English has one phonemic rhotic 111 which can be produced utilizing two maximally distinctive articulations with intermediate variations. One is known as the retroflex [j] and the other as the bunched [j]. Speakers who employ more retroflex-like [j] articulations lift the apex of their tongue up and point it toward the alveolar ridge, keeping the tongue dorsum relatively low in the oral cavity. Speakers who employ more bunched-like [j] articulations contract their tongue tip back and raise the tongue dorsum toward the palate. Other constrictions in the vocal tract (lips and pharynx) do contribute to the production of English rhotics; however, the variation across speakers lies in the differences in tongue constriction in the palatal region (Espy- Wilson et al. 2000). Whether or not the different ways of producing English rhotics have an acoustic effect is not as important for the acquisition of Spanish rhotics (although acoustic effects may be important for learners of L2 English) as the physiological aspect of the differing articulations and the way to measure such articulations - F5-F4 distance with a greater F5-F4 difference indicating a more retroflex-like articulation and a smaller difference indicating a more bunched-like articulation (Zhou et al. 2008). The reason that acoustic effects of English rhotics are not important is that the acoustics of English and Spanish rhotics are completely different and there is no logical reason to assume any type of transfer based on acoustic similarities or differences. The physiologically different articulations, however, are comparable (in the case of retroflex articulations) and contrastable (in the case of bunched articulations) between English and Spanish. Although taps are not usually associated with rhotics in English, because they occur as allophones of alveolar stop consonants /t/ and I A! as in the word matter [maeq], they are important to this study because of their rhotic status in Spanish. Speakers produce taps by raising the apex of the tongue towards the alveolar ridge and making a very brief closure. Spanish, unlike English, has two distinct phonemic rhotics - a tap /r/ and a trill /r/, illustrated by the minimal pairs foro [foro] 'forum'/ forro [foro] 'lining'.4 Spanish speakers produce taps in the same way as the allophonic tap in English and are described in a manner similar to the description already mentioned by Tomas (1921). Trills differ in that the number of closures is greater. Sol (2002) noted that trills usually entailed four, and sometimes five or six quick successive closures. It has also been noted that subjects often hyperarticulate in laboratory conditions, and that less taps are more common in speech that is more casual (Blecua 1999). Although variation in articulatory gestures assigned to Irl exist for native speakers, rhotics that L2 learners are taught in classroom settings in the United States can be described simply as alveolar taps and alveolar trills (Face 2006). 2.1 Interlanguage Rhotic Phonological Structure In the case of LI English speakers learning Spanish as an L2, learners must acquire a new speech sound, the trill /r/, and must reassign an English allophone of alveolar stops [r] to become its own phoneme Id. Because English speakers' phonetic inventories do not contain trills, they are more easily acquired than taps (Flege 1995). This is because speakers do not need to reassign an existing phone in their LI, and are able to construct a new phoneme. However, this does not mean that learners accurately produce trills before taps. Due to the articulatory difficulty of trills, they actually show up later in rhotic development than taps (Face 2006). Figure 1 shows the native structure of English taps and Spanish rhotics.5 The allophonic tap is found in complementary distribution of alveolar stops /t/ and Id/. The connecting lines between the phonemes and the [r] indicate this relationship. Because both Spanish rhotics hold phonemic status, the lines show the Spanish phonemes connecting directly to their corresponding phones. Figure 2 shows the process that native English speakers must undergo to acquire a target-like phonological system regarding rhotics. The broken lines indicate the process of disassociation from alveolar stops in the LI and the arrow indicates the phonemicization process that [r] must This content downloaded from 168.176.5.118 on Mon, 1 Sep 2014 12:10:20 PM All use subject to JSTOR Terms and Conditions Olsen / L2 Acquisition of Spanish Rhotics 69 Figure 1. Phonological Structure of English Tap and Spanish Rhotics Figure 2. Phonological Restructuring of English Tap and Spanish Rhotics undergo. Learners must also acquire the trill, which does not exist in English. This process occurs as learners notice (subconsciously) that the distribution of taps is no longer predictable and a separate phoneme must be posited. While the acquisition process alone is quite complicated, there are also other confounding factors. Orthography is one important factor that can have a negative influence on the acquisition of Spanish rhotics (Koda 1989; Munro and Derwing 1994; Zampini 1994). Because taps and non-intervocalic trills are represented orthographically as <r>, classroom learners' (with no prior experience) immediate response is to refer to the English rhotic 111 which shares the same grapheme. Instead of activating the tap already existent in their phonological representation, the alveolar approximant 111 is activated. While this orthographic influence is probably stronger for taps, it is possible that <rr>, the intervocalic grapheme for trills, also activates the alveolar approximant. An interlanguage phonological system regarding rhotics (after a contrast between all rhotics involved is perceived) consists of the two phonemes Id and Irl. The former is realized as the allophones [r] and [i] and the latter is realized as [r] and [j]. Figure 3 shows the typical interlanguage structure of beginning L2 Spanish learners as seen in the data of the present study. This content downloaded from 168.176.5.118 on Mon, 1 Sep 2014 12:10:20 PM All use subject to JSTOR Terms and Conditions 70 Hispania 95 March 20 1 2 Figure 3. Interlanguage Phonological Structure of Rhotics 3. L2 Phonological Acquisition Although learners go through a process of learning phonology starting with their LI and moving towards an L2, the language phenomena evident in this process sometimes do not resemble the LI or the L2 as shown in Figure 3 (Ellis 2008). Differences in interlanguages may occur because the acquisition of phonology relies on phonetically encoded phonemic categories as well as individual differences in learners' abilities to form phonetic-phonemic mappings. One L2 phonological acquisition model that describes acquisition in terms of differences between LI and L2 articulations is Flege's (1995) Speech Learning Model (SLM). According to the SLM, L2 learners perceive all L2 sounds in terms of established LI segments in their phonological system from the onset of L2 exposure. Through experience, L2 learners notice phonetic differences between target segments and LI segments and are able to create a different phonetic category for the target language segment. This means that native English speakers would, at first, interpret the voiced interdental fricative [] found in the word cada [kaa] 'each' as a voiced alveolar stop [d]. Over time, speakers would recognize that the L2 phone is not the same as the LI form and formulate a separate segment in their phonological system. The SLM also predicts that separating L2 segments from LI segments that are phonetically closer to those LI segments (e.g., [t] and [th]) will be more difficult than separating L2 segments from LI segments that have more phonetic distance between them (e.g., [j] and [r]). Because L2 segments that are more distant from LI segments are more easily separated, the SLM also predicts that L2 learners more easily produce them. In terms of the acquisition of Spanish rhotics within the SLM, rhotics would at first be perceived as the English [j] because of a sufficiently small perceived phonetic distance between the L2 and LI phones and because of the orthographic influence mentioned above. Spanish rhotics will also be produced by L2 learners as English [j] because they are influenced by the phonological structure of those LI categories to which they are most similar, at least at the early stages of L2 development. This means that taps should be more accurately produced in contexts where they are produced in the LI (i.e., the onset of an unstressed syllable following a stressed syllable). Because articulatory routines for taps are more similar to routines used by speakers who employ retroflex articulations for LI English rhotics, these speakers are predicted to more accurately produce Spanish rhotics. Therefore, LI within-category differences may affect the accurate pronunciation of Spanish rhotics even when they are not perceived by LI speakers, because of their degree of similarity to target phones. This content downloaded from 168.176.5.118 on Mon, 1 Sep 2014 12:10:20 PM All use subject to JSTOR Terms and Conditions Olsen / L2 Acquisition of Spanish Rhotics 7 1 4. Current Study The current study investigated one LI articulatory routine difference and one LI phonetic context that influence the acquisition of Spanish rhotics by English speakers. Specifically, this study addressed the effects of the different ways to produce [j] in English. It also addressed the influence that the phonetic context involving [r] in English has on the acquisition of the same phone in Spanish. The remainder of this paper consists of a description of the participants and the procedures used in the study, analytical results, and a discussion on what the results indicate. 4.1 Participants and Procedures Fifty-one native English-speaking adults from university beginning Spanish classes par- ticipated in this study. Five native Spanish-speakers (three females and two males) representing dialects from a variety of South American countries and Spain also participated to provide a base for rhotic accuracy rates. The data from three of the native English-speaking participants are not included in this study because either an instrument malfunction or user error occurred rendering their recordings inaudible. Therefore, the data elicited from a total number of forty-eight participants were included in the analyses. Participants filled out a questionnaire that asked them to indicate if English was their native language and to rate their exposure to Spanish (i.e., classroom-based exposure and exposure through media) before taking the Spanish class in which they were currently enrolled on a Likert scale from 1-7 where 1 = 'no exposure' and 7= 'extensive exposure'. All L2 Spanish learners indicated that they were native English speakers. Participants recorded themselves reading a text in Spanish adapted from a reading found in Mosaicos (4th ed.) (Castells, Guzman, Lapuerta, and Garcia 2006) designed to elicit the same number of possible tap and trill articulations from each participant in order to calculate accuracy rates. They accomplished this task on Macintosh computers equipped with headsets using Audio Recorder 3.2. The Spanish text contained a total of thirty-two intervocalic taps. Nineteen taps occurred in the onsets of unstressed syllables where the preceding syllable was stressed (the same environment in which alveolar stops become taps in English), as in the word pero ['pero] 'but'. Thirteen taps occurred in other intervocalic environments (i.e., occurring after an unstressed syllable and comprising the onset of either a stressed or unstressed syllable), as in the words diferente [dife'rente] 'different' or numro ['numro] 'number'. The text also contained four intervocalic trills as in the words cierran ['sjeran] 'close' (third person plural) or correos [ko'reos] 'mail' (plural). In order to determine the type of English rhotic articulation employed, participants also recorded themselves pronouncing four English words containing [j] - arrow , car, proud, and heart , along with a prolonged [j] pronunciation. They were asked to pronounce each word twice and to hold out the [j] for a few seconds. All recordings were analyzed using Praat (Boersma and Weenink 2009). Successful taps were counted when there was a clear closure of the vocal tract indicated in the spectrogram by a brief break in the formant structures. Successful trills were counted following Sol (2002) and Blecua (1999), when at least two successive closures of the vocal tract were evident. Figure 4 shows an example of an accurate tap articulation and Figure 5 is an example of an accurate trill articulation. Manner of articulation of English [j] in each participant was calculated by averaging the distances between F4 and F5 taken from each of the pronounced English words and the pro- longed [j]. The distance between F4 and F5 of the [j] articulations in each word was produced by averaging all of the F4 and F5 frequencies measured at 6.25 millisecond intervals throughout each pronunciation and subtracting the F4 average from the F5 average. Care was taken to exclude any surrounding sounds from the [j] measurements. Figure 6 shows an example of a This content downloaded from 168.176.5.118 on Mon, 1 Sep 2014 12:10:20 PM All use subject to JSTOR Terms and Conditions 72 Hispania 95 March 2012 Figure 4. Example of Accurate Tap Articulation Figure 5. Example of Accurate Trill Articulation window from which F4 and F5 measurements were obtained (measured every 6.25 milliseconds within the window). As mentioned previously, a greater distance between F4 and F5 indicates a more retroflex- like articulation whereas a lower distance between F4 and F5 indicates a more bunched-like articulation (Zhou et al. 2008). Figure 7 is a spectrogram of a bunched [j] articulation and Figure 8 is a spectrogram of a rhotic [j] articulation. As can be seen, the distance between F4 and F5 in Figure 7 is relatively small compared to the distance between F4 and F5 in Figure 8. Zhou et al. (2008) found that differences between F4 and F5 for participants (males with similar vocal tract length) employing retroflex articulations were around 1400 Hz, while differences for speakers employing bunched articulations were around 700 Hz. The average distance between F4 and F5 varied among speakers in the present study, ranging from 525 Hz and 1603 Hz with a mean of 1057 Hz and a standard deviation of 219 Hz. This broad range of differences is not surprising because of the possibility of intermediate articulations across speakers. Because English rhotic articulations in this study fell along a continuum with retroflex and bunched poles, they were analyzed as such. Tap accuracy rates were calculated for each student by dividing the total number of accurate taps by the total number of possible taps. Tap accuracy rates were also calculated for taps in phonetic environments that produce taps in English as well as taps in other environments to test This content downloaded from 168.176.5.118 on Mon, 1 Sep 2014 12:10:20 PM All use subject to JSTOR Terms and Conditions Olsen / L2 Acquisition of Spanish Rhotics 73 Figure 6. Measurement of [j] Articulations Figure 7. Spectrogram of Bunched [j] Articulation Figure 8. Spectrogram of Retroflex [j] Articulation This content downloaded from 168.176.5.118 on Mon, 1 Sep 2014 12:10:20 PM All use subject to JSTOR Terms and Conditions 74 Hispania 95 March 20 1 2 for the possible influence that the English tap rule has on accuracy of the production of Spanish taps. Trill accuracy rates were calculated in the same manner. 4.2 Results The participant responses to the Likert scale ranking task, which asked them to rank themselves according to the amount of exposure to Spanish prior to enrolling in their Spanish class, created a normal distribution with a peak around the second ranking (little exposure). The percentage of participants that indicated having little exposure was 37.5% (N= 18). Two participants indicated that they previously had a fair amount of exposure to Spanish (Likert ranking of 5), and no participants indicated that they had undergone considerable (Likert ranking of 6) or extensive (Likert ranking of 7) exposure to Spanish prior to enrolling in this course. Figure 9 shows the distribution of participant responses. Figure 9. Distribution of Exposure to Spanish Prior to Enrollment Thirty-one English-speaking participants (64.6%) were able to produce at least one tap accurately (as judged by the investigator using spectrograms). While participants that were not able to produce at least one tap where found in all of the experience levels, the majority were found in the lower experience levels, indicating that tap production increases with experience. Successful taps consisted of a clear closure of the vocal tract and successful trills consisted of at least two successive closures. Of the accurately produced taps, accuracy rates ranged from 3.1% (1/32) to 1 00% (32/32) with a mean of 56.3%. Tap accuracy rates for the native Spanish- This content downloaded from 168.176.5.118 on Mon, 1 Sep 2014 12:10:20 PM All use subject to JSTOR Terms and Conditions Olsen / L2 Acquisition of Spanish Rhotics 75 speaking participants ranged from 96.9% (31/32) to 100% (32/32) with a mean of 97.5%. An independent-samples /-test was performed to test the significance of the difference between the means of Spanish tap accuracy rates between the participant groups. The difference between English-speaking participants' tap accuracy rates (M=36.3, SD=36.1) and Spanish-speaking participants' tap accuracy rates (M=97.5, SD=1.4) was significant, f(51)=11.661,/?<.001. A linear regression was performed (a= .05) to test whether English rhotic articulation (more retrofiex-like articulations predicting better accuracy) predicted tap accuracy with participants' tap accuracy rates as the dependent variable and English rhotic articulation (distance between F4 and F5 for each participant) as the predictor variable. For tap accuracy, rhotic articulation (fi2=.057) alone was not a significant predictor (=.239,p=.102). Figure 10 is a scatter plot of the relationship between tap accuracy rates and English rhotic articulation. A multiple linear regression was then performed to test whether English rhotic articulation was a predictor of tap accuracy, this time controlling for exposure to Spanish (a second predictor variable). This test showed a significant effect for English rhotic articulation (=.320, p=.010) as well as Figure 10. Correlation between Tap Accuracy and English Rhotic Articulation amount of exposure to Spanish (=.564,/?<.001), which combined accounted for 37% of the variance (R2=. 369). Because English rhotic articulation was only a significant predictor when controlling for amount of exposure to Spanish, participants were split into groups according to the amount of exposure to Spanish they had indicated and separate regressions were performed on each This content downloaded from 168.176.5.118 on Mon, 1 Sep 2014 12:10:20 PM All use subject to JSTOR Terms and Conditions 76 Hispania 95 March 20 1 2 group. English rhotic articulation (ft2=.235) alone proved to be a significant predictor (=.484, /?=.042) for the participants who ranked themselves as having little prior exposure to Span- ish (2 on the Likert scale). For participants who reported having some prior exposure to Spanish (3 on the Likert scale), English rhotic articulation (R2=. 321) approached significance as a predictor of tap accuracy (=. 567,/?=. 088). For all other participants, English rhotic articula- tion was not a significant predictor of tap accuracy (R2=. 215, =. 464,/? =.209 for group 1; R2 = . 1 6 1 , = .40 1 , p = .285 for group 4).6 F igure 1 1 shows the relationship between tap accuracy rates and English rhotic articulation separated by Spanish exposure group. The data points in the second group better fit the regression line, showing a significant correlation. The data points in the third group also indicate a weak relationship to the regression line. The other groups do not show strong correlations. Figure 11. Correlation between Tap Accuracy and English Rhotic Articulation by Exposure Group To see whether the English phonetic context that produces taps affected Spanish tap accu- racy, a paired-samples t- test was performed. This test compared the means of the accuracy rates of taps (including only the participants that produced accurate taps, N=31) found in phonetic environments that pattern like the English tap rule (M=61 .6, SD=3 1 .4) and taps found in other environments (M=45.4, SD=30.9). A significant difference was found between these means; /(30)=4.845,/?<.001. Figure 12 shows the means of the accurate taps in the English tap rule environment compared to the accurate taps in other environments. As the bars indicate, the This content downloaded from 168.176.5.118 on Mon, 1 Sep 2014 12:10:20 PM All use subject to JSTOR Terms and Conditions Olsen / L2 Acquisition of Spanish Rhotics 77 Figure 12. Means of Accuracy Rates of Taps in Different Environments mean accuracy of taps in English tap rule environments is significantly higher than the mean accuracy of taps in other environments. Seven English-speaking participants out of the forty-eight (14.6%) were able to produce at least one accurate trill. Of the accurately produced trills, accuracy rates ranged from 25% (1/4) to 50% (2/4) with a mean of 35.7%. Trill accuracy rates among the native Spanish-speaking participants ranged from 50% (2/4) to 100% (4/4) with a mean of 85%. An independent-samples t-tQSt was performed to compare Spanish trill accuracy rates between the participant groups. The test showed a significant difference, /(5 1 ) = 1 1 .724, p< .00 1 , between the English-speaking participants' trill accuracy rates (N=48, M=5.2, SD= 13.6) and Spanish-speaking participants' trill accuracy rates (N=5, M=85.0, SD=22.4). A linear regression was performed (a=.05) to test whether English rhotic articulation predicted trill accuracy with participants' trill accuracy as the dependant variable and English rhotic articulation (distance between F4 and F5 for each participant) as the predictor variable. For trill accuracy, rhotic articulation alone (/?2=.110) was a significant predictor (=.332, p=. 021). A multiple linear regression was also performed to account for exposure to Spanish in trill accuracy rates. This test showed that English rhotic articulation (=. 364, p =.01 2) was also a significant factor controlling for exposure to Spanish (= .2 1 S,p= . 122), which combined accounted for 16% of the variance (/?2=.157). Figure 13 shows the correlation between trill accuracy rates and English rhotic articulation. This content downloaded from 168.176.5.118 on Mon, 1 Sep 2014 12:10:20 PM All use subject to JSTOR Terms and Conditions 7 8 Hispania 95 March 2012 Figure 13. Correlation between Trill Accuracy and English Rhotic Articulation 5. Discussion and Implications Results from the regressions performed provide answers to the first research question repeated here: 1 . Does the LI within-category difference manifested in different manners of American English rhotic articulation (i.e., retroflex-like or bunched-like) affect the facilitation of Spanish rhotic production? The regressions showed that English rhotic articulation is a predictor of accurate trill production at the beginning levels of L2 Spanish phonology acquisition. That is to say, those learners who employ more retroflex articulations in English were initially able to produce trills with a higher accuracy rate than those who employ more bunched articulations. However, the small number of trills produced by participants should be taken into consideration and is a limitation of this study. Although results were positive for trill production both alone and accounting for prior exposure to Spanish, results of this study showed that there was no significant relationship between tap accuracy rates and English rhotic articulation alone. When taking prior exposure to Spanish into consideration, however, English rhotic articulation did correlate with tap accuracy rate. This result is in line with Face's (2006) findings that with more experience, rhotic accuracy rates improve. These results also extend the idea that experience is one of the main factors in the acquisition of Spanish rhotics. Even relatively small differences in the amount of exposure This content downloaded from 168.176.5.118 on Mon, 1 Sep 2014 12:10:20 PM All use subject to JSTOR Terms and Conditions Olsen / L2 Acquisition of Spanish Rhotics 79 time at the beginner level make a difference in L2 Spanish rhotic accuracy. When controlling for amount of exposure to Spanish, English rhotic articulation did have an effect on tap accuracy rates. Therefore, although experience is the main factor in accurate tap production in Spanish, English rhotic articulation does influence the accuracy of tap production, at least at certain stages of acquisition, and should be taken into consideration by learners and instructors. Because the English rhotic articulation became a predictor of tap accuracy only when controlling for exposure, self-ranked exposure to Spanish groups were split to further examine the effect of English rhotic influence on each individual group. Results showed an effect for the second and third self-ranked groups (only the second group was statistically significant) and not for other self-ranked groups. These results, along with the results from the multiple linear regressions showing that exposure to Spanish is a significant predictor of tap accuracy, suggest a crucial difference measure much like Wode's (1976) "crucial similarity measure," where transfer only begins when learners notice similarities between the LI and the L2. Influence involving LI phonetic factors (e.g., English rhotic articulation) and production only become evident at the point where L2 learners notice differences between the LI and the L2 articulations. At first, the amount of exposure to Spanish and speaking practice is minimal and may not be sufficient for learners to distinguish between English and Spanish rhotic articulations. At the point where a crucial difference is noticed (shown by the results in groups two and three), learners begin to produce Spanish rhotic articulations because they recognize when they are producing English rhotic articulations and when they are producing Spanish rhotic articulations, and that they are not the same (onset of segmental acquisition). At this point, English rhotic articulation becomes a factor in tap accuracy rates. Retroflex-like articulations facilitate accurate tap production relative to bunched-like articulations because, once target language articulation is distinguished from LI articulations, speakers that employ retroflex-like articulations are used to raising the apex, as is necessary for Spanish rhotics. After more experience with Spanish, LI phonetic factors become less important. Learners who employ bunched-like articulations also learn to produce accurate taps. Further research employing cross-sectional or longitudinal design and a higher number of elicited trills would be able to test the validity of such a crucial difference measure and indicate whether trills also follow the described measure. Concerning the effect of LI phonetic context on the production of Spanish rhotics, results show that out of the accurately produced taps, a significantly high percentage of them were in words in which primary stress was contained in the syllable before the tap, equal to those found in English words that create tapped realizations of Iii and Id/. These results provide an answer to the second research question repeated here: Does the phonetic context that produces allophonic taps in English affect accuracy in Spanish rhotic production? The results to this question are interesting for two reasons. First, the correlation of similar patterns of stress in relation to taps across languages and effective production of taps indicates that the subconscious phonological English tap rule interacts with L2 Spanish utterances in a way that facilitates similar phonetic effects across the two languages. That participants more accurately produced taps in environ- ments that also render taps in their LI shows an LI influence of phonetic context. Second, that phonetic contexts of an allophone associated with Iii and Idl (both of which are also phonemes in Spanish) facilitates the production of taps only in environments that are similar to English provides evidence that these participants have not yet sufficiently augmented the possible phonetic contexts in which Spanish [r] can occur past LI phonetic contexts. Although this process is underway, indicated by instances of accurate tap productions in environments unlike English, beginners have not yet reallocated taps in their phonemic inventories from an English allophone of Iii and /d/ to a separate Spanish phoneme. In other words, these learners have begun to associate [r] with Spanish rhotics, but hold onto the phonetic contexts that produces taps in English. The phonological structure of learners in this study is in an interlingual state, as seen in Figure 3 above. This content downloaded from 168.176.5.118 on Mon, 1 Sep 2014 12:10:20 PM All use subject to JSTOR Terms and Conditions 80 Hispania 95 March 20 1 2 Interestingly, results showed that the phonetic context that produces allophonic taps in English have a longer lasting effect than within-category subphonemic variants of [j] on the production of L2 Spanish rhotics. Because LI English taps are represented in the phonological system as an allophone of alveolar stops It / and /d/, it is more salient to the learner's LI system. Because the differences in LI articulatory routine are represented at an even less conscious subphonemic level, the effects shown in this study are easily overcome because of the lack of salience to the LI system. These results are interesting because they indicate that, although differences in LI articulatory routines do affect L2 phonology acquisition, influence from LI phonetic context is more difficult to overcome. Including more advanced learners in future studies that focus on LI phonetic influence would reveal the extent of these phonetic factors in L2 phonology acquisition. These results can be partially explained by current models that have been purported to explain L2 phonology acquisition. The Speech Learning Model (SLM) (Flege 1995), the Native Language Magnet Model (NLM) (Iverson and Kuhl 1996), and the Perceptual Assimilation Model (Best 1995) all predict that L2 input will at first be categorized with respect to LI cat- egories. Learners are able to form L2 categories as they receive more exposure. Results herein showed that learners with less experience produce LI phones that they perceive (largely due to orthographic influence) to be closest to the actual L2 phone. As learners perceive phonetic differences between L2 and LI rhotics, they begin to stop using phonetic articulations mapped to /i/ and start using phonetic articulations mapped to [r], which are closer to the L2 articulation. They begin to formulate a separate category for taps. However, at this point, the phonetic context that produces allophonic taps in English continues to influence the accuracy of tap production in L2 Spanish. In this sense, the LI category that is associated with L2 taps changes with more experience. That the LI influence shifts from one LI phone ([j], as evidenced in the production of learners with very little exposure to Spanish) to a different LI phone ([r], as evidenced in the accuracy rates of learners with more exposure to Spanish) is not predicted in the current models mentioned. A more fine-grained phonetic analysis is necessary to account for the data analyzed in this study. The current study has implications for interlanguage phonological theory. As proposed by Colantoni and Steele (2008), theories that explain and predict interlanguage phonology should address both phonological and phonetic phenomena. The results obtained by this study provide evidence for this proposal. That phonetic factors (LI articulatory routine variation and LI phonetic context for allophonic variation) have an effect on the accuracy rates of Spanish rhotics by L2 learners lends support for the hypothesis tested in this study. Both LI articulatory routine variation, exemplified by different English rhotic articulations, and LI phonetic context, exemplified by the phonetic context that produces taps in English, do affect the acquisition of an L2 phonology in terms of native-like pronunciation, albeit at different intensities. Because results showed that both types of factors did influence target-like learners' pronunciation, both should be considered in L2 acquisition theories. The current study has also shown that the LI phones that influence accurate pronunciation in Spanish rhotics change over time. Current L2 phonology acquisition models do not account for such shifts. This may be accomplished by attributing this shift to different levels of representation. The first LI phone that influences L2 Spanish rhotic production, /j/, is at the phonemic level and is the closest phoneme that exists in the LI to the target phoneme. As differences in articulation are noticed, and a separate tap category is created, the phonetic context that produces allophonic taps in the LI continue to influence L2 Spanish taps because of the allophonic status that this phone has in the LI . As for the effect of articulatory routines, because they are neuromotor routines that LI English speak- ers use to articulate /j/, once learners have made the transition to articulating taps, differences in English rhotic articulatory routines have less of an influence on accurate tap production. In other words, once learners with bunched-like articulations have a sufficient amount of practice articulating taps in Spanish, the advantage for learners with retroflex-like articulations disap- This content downloaded from 168.176.5.118 on Mon, 1 Sep 2014 12:10:20 PM All use subject to JSTOR Terms and Conditions Olsen / L2 Acquisition of Spanish Rhotics 8 1 pears. Future studies focusing on LI phonetic influence at different levels of proficiency as well as studies investigating the influence of different LI phones are needed to further understand the effect of LI phonetics on L2 phonology development. ACKNOWLEDGMENTS I would like to thank Alan Juffs, Erin O'Rourke, Carlos Ramirez, and Jeff Lingwall for their helpful comments and indispensable insights. NOTES !Rose (2010b) suggests a more fine-grained sequence of phone production than presented here. However, I have presented the general sequence suggested by her results. 2 The term "articulatory routine" is used herein to refer to neuromotor routines employed by speakers which involve the physical movement of the articulators to consistently produce the acoustic effects that listeners interpret as particular phones. In this case, the articulatory routines investigated are the physical differences in tongue position between retroflex-like and bunched-like rhotic articulations. 3 All instances of the word English refer to Standard American English spoken in the mid- Atlantic United States. 4 Some researchers only posit one phonemic rhotic in Spanish (Hualde 2005). 5 A complete phonological structure would include other allophones of Iii and /d/; only the phonologi- cal structure relating to taps is relevant for this study. 6 Because only two participants indicated that they had a fair amount (Likert ranking of 5) of exposure to Spanish prior to enrolling in the course, a linear regression could not be performed. WORKS CITED Alwan, Abeer, Shrikanth Narayanan, and Katherine Haker. (1997). "Toward Articulatory-acoustic Models for Liquid Approximants Based on MRI and EPG data. Part II. The Rhotics." Journal of the Acoustical Society of America 101: 1078-89. Print. Best, Catherine T. (1995). "A Direct Realist Perspective on Cross-language Speech Perception." Speech Perception and Linguistic Experience: Issues in Cross-language Speech Research. Ed. Winifred Strange. Timonium, MD: York. 167-200. Print. Blecua, Beatriz. (1999). "Caracteristicas acusticas de la vibrante multiple del espanol en habla espontnea." Proceedings of the 1st Congress of "Experimental Phonetics". Tarragona, Spain. 119-26. Print.
. (2001). Las vibrantes del espanol: Manifestaciones acusticas y procesos fonticos. Diss. U Autnoma de Barcelona. Print. Boersma, Paul, and David Weenink. (2009). Praat: Doing Phonetics by Computer. Version 5.1.04. Software. Carlisle, Robert S. (1988). "The Effect of Markedness on Epenthesis in Spanish/English Interlanguage Phonology." Issues and Developments in English and Applied Linguistics 3: 15-23. Print. Castells, Matilde O., Elizabeth Guzman, Paloma Lapuerta, and Carmen Carcia. (2006). Mosaicos: Spanish as a World Language. 4th ed. Upper Saddle River, NJ: Pearson. Print. Colantoni, Laura, and Jeffrey Steele. (2008). "Integrating Articulatory Constraints into Models of Second Language Phonological Acquisition." Applied Psycholinguistics 29.3: 489-534. Print. Ellis, Rod. (2008). The Study of Second Language Acquisition. 2nd ed. Oxford: Oxford UP. Print. Espy- Wilson, Carol Y., Suzanne E. Boyce, Michel Jackson, Shrikanth Narayanan, and Abeer Alwan. (2000). "Acoustic Modeling of American English /r/." Journal of the Acoustic Society of America 108.1: 343-56. Print. Face, Timothy L. (2006). "Intervocalic Rhotic Pronunciation by Adult Learners of Spanish as a Second Language." 7th Conference on the Acquisition of Spanish and Portuguese as First and Second Langauges. Ed. Carol. A. Klee and Timothy L. Face. Somerville, MA: Cascadilla. 47-58. Print. Flege, James E. (1995). "Second Language Speech Learning: Theory, Findings and Problems." Speech Perception and Linguistic Experience: Issues in Cross-language Speech Research. Ed. Winifred Strange. Timonium, MD: York. 233-72. Print. This content downloaded from 168.176.5.118 on Mon, 1 Sep 2014 12:10:20 PM All use subject to JSTOR Terms and Conditions 82 Hispania 95 March 20 1 2 Hammond, Robert. (1999). "On the Non-occurrence of the Phone [r] in the Spanish Sound System." Advances in Hispanic Linguistics. Ed. Javier Gutirrez-Rexach and Fernando Martinez-Gil. Somer- ville, MA: Cascadilla. 290-304. Print. Hualde, Jos I. (2005). "Quasi-Phonemic Contrasts in Spanish." WCCFL 23 Proceedings. Ed. Benjamin Schmeiser, Vineeta Chand, Ann Kelleher, and Angelo Rodriguez. Somerville, MA: Cascadilla. 374-98. Print. Hurtado, Luz. M., and Chelsea Estrada. (2010). "Factors Influencing the Second Language Acquisition of Spanish Vibrants." Modem Language Journal 94.1 : 74-86. Print. Iverson, Paul, and Patricia K. Kuhl. (1996). "Influences of Phonetic Identification and Category Good- ness on American Listeners' Perception of Irl and /l/." Journal of Acoustic Society of America 99.2 : 1130-40. Print. Koda, Keiko. (1989). "Effects of LI Orthographic Representation on L2 Phonological Coding Strategies." Journal of Psycholinguistic Research 18.2: 201-22. Print. Ladefoged, Peter. (2006). A Course in Phonetics. Boston: Thomson Wadsworth. Print. Lindau, Mona. (1985). "The Story of /r/." Phonetic Linguistics : Essays in Honor of P Ladefoged. Ed. Victoria Fromkin. New York: Academic. 157-67. Print. Munro, Murray J., and Tracey M. Derwing. (1994). "Evaluations of Foreign Accent in Extemporaneous and Read Material." Language Testing 1 1.3: 253-66. Print. Rose, Marda. (2010a). "Differences in Discriminating L2 Consonants: A Comparison of Spanish Taps and Trills." Selected Proceedings of the 2008 Second Language Research Forum. Ed. Matthew T. Prior, Yokiko Watanabe, and Sang-Ki Lee. Somerville, MA: Cascadilla. 181-96. Print.
. (2010b). "Intervocalic Tap and Trill Production in the Acquisition of Spanish as a Second Language." Studies in Hispanic andLusophone Linguistics 3.2: 1-41. Print. Sol, Maria-Josep. (2002). "Aerodynamic Characteristics of Trills and Phonological Patterning." Journal of Phonetics 30.4: 655-88. Print. Tomas, Tomas N. (1921). Manual de pronunciacin espanola. Madrid: Filologia espafiola. Print. Waltmunson, Jeremy C. (2005). "The Relative Degree of Difficulty of Spanish It, d /, Trill, and Tap by LI English Speakers: Auditory and Acoustic Methods of Defining Pronunciation Accuracy." Diss. U of Washington. Print. Wode, H. (1976). "Developmental Sequences in Naturalistic L2 Acquisition." Working Papers on Bilingualism 11:1-13. Print. Zampini, Mary L. (1994). "The Role of Native Language Transfer and Task Formality in the Acquisition of Spanish Spirantization." Hispania 77.3: 470-81. Print. Zhou, Xinhui, Carol Y. Espy- Wilson, Suzanne Boyce, Mark Tiede, Christy Holland, and Ann Choe. (2008). "A Magnetic Resonance Imaging-Based Articulatory and Acoustic Study of 'Retroflex' and 'Bunched' American English Irl." Journal of the Acoustic Society of America 123.6: 4466-81. Print. This content downloaded from 168.176.5.118 on Mon, 1 Sep 2014 12:10:20 PM All use subject to JSTOR Terms and Conditions