DOCUMENT RESUME ED 427 521 FL 025 693 AUTHOR Reeder, Jeffrey T. English Speakers' Acquisition of Voiceless Stops and Trills TITLE in L2 Spanish. ISSN-0898-8471 ISSN PUB DATE 1998-00-00 20p.; For the complete volume of working papers, see FL 025 NOTE 687. PUB TYPE Journal Articles (080) Reports Research (143) JOURNAL CIT Texas Papers in Foreign Language Education; v3 n3 p101-18 Fall 1998 EDRS PRICE MF01/PC01 Plus Postage. DESCRIPTORS Computer Oriented Programs; Higher Education; Language Proficiency; Language Research; *Learning Processes; Linguistic Theory; *Phonology; *Pronunciation; Second Language Learning; *Spanish; Time Factors (Learning) *English Speaking IDENTIFIERS ABSTRACT A study attempted to (1) define acoustically the learner's progress in acquisition of a set of phonological features, specifically interlingual differences in the voice onset time of voiceless stop consonants and the number of taps produced when attempting to produce the Spanish (2) determine if and at what stage of acquisition the phoneme /r/ (trill); given L2 targets are realized; and (3) use data from adult English-speaking learners of Spanish to test Flege's framework of second language speech acquisition known as the Speech Learning Model. Subjects were 40 native English-speaking college students of Spanish. The learners, representing four different proficiency levels, provided data that were analyzed acoustically using computer-based speech analysis software. In addition to tracing the acquisition of a set of sounds through the four levels, the study provides evidence that the Spanish trill is acquired differently than the voiceless stops. Furthermore, Flege's model is shown to be a relevant model of second language speech acquisition with respect to the sounds and language combination examined here. Contains 17 references. (MSE) ******************************************************************************** Reproductions supplied by EDRS are the best that can be made from the original document. ******************************************************************************** English Speakers' Acquisition of Voiceless Stops and Trills in L2 Spanish JEFFREY T. REEDER, Sonoma State University U.S. DEPARTMENT OF EDUCATION Office of Educational Research and Improvement EDUCATIONAL RESOURCES INFORMATION PERMISSION TO REPRODUCE AND CENTER (ERIC) DISSEMINATE THIS MATERIAL HAS ffi/This document has been reproduced as BEEN GRANTED BY received from the person or organization originating it. e&rptrit 0 Minor changes have been made to improve reproduction quality. Points of view or opinions stated in this TO THE EDUCATIONAL RESOURCES document do not necessarily represent INFORMATION CENTER (ERIC) official OERI position or policy. BEST COPY AVAILABLE English Speakers' Acquisition of Voiceless Stops and Trills in L2 Spanish JEFFREY T. REEDER, Sonoma State University The purpose of this paper (1) to define is threefold: acoustically the learner progress in the acquisition of a set of phonological features, spe- cifically interlingual differences in the voice onset time of voiceless stop consonants and the number of taps produced when attempting to pro- duce the Spanish phoneme irk (2) to determine if and at what stage of the given L2 targets are realized; and (3) to use data from acquisition adult English-speaking learners of Spanish to test Flege's framework of second language speech Learning acquisition known the as Speech Model. The basis for addressing the above questions in this paper is a cross-sectional study of 40 native English-speaking learners of Spanish in a U.S. university. These learners, representing four different levels, pro- tided data that were analyzed acoustically using computer-based speech analysis software. In addition to tracing the acquisition of a set of sounds through the four levels, this study provides evidence that the Spanish trill is acquired differently than the voiceless stops. Furthermore, Flege's SLM is shown to be a relevant model of second language speech acquisi- tion with respect to the sounds and language combination examined in this study. INTRODUCTION This paper is divided into four sections. The first section provides the background for this study by summarizing current views of some of the fac- tors constraining second language (L2) speech acquisition and by describing Flege's (1995) Speech Learning Model (SLM). Also in the first section is a de- scription of the Spanish sounds that are studied in this work, accompanied by an overview of previous research done on the L2 acquisition of those Spanish sounds by English speakers. The second section of this paper describes the method used in this cross-sectional study, including a description of the par- ticipants and their experience with the study of Spanish, an explanation of the data collection procedure, and an outline of the scoring procedure. The next section presents the results of the perception test and the data from the par- ticipants' production attempts of the trill and the voiceless stops, and the final section summarizes the results and their implications for what is known about the acquisition of L2 Spanish speech. BACKGROUND In this section, current views of some of the factors governing second language (L2) speech acquisition are summarized. After this background in- formation is given, a description of relevant portions of Flege's (1995) SLM is provided, given that it is a model of L2 speech acquisition that claims to be 102 Texas Papers in Foreign Language Education of all value according to the categoriza- applicable to learners ages. tions in their Ll. Next is a description of the Spanish sounds /r/, /p/, /t/, and /k/ that are Also seeking evidence of how categories are established, Williams studied here, as well as an encapsu- lation of the research that has al- (1979) engaged bilingual adults in perceptory discrimination tasks and ready been done on the L2 acquisi- found that the subjects established tion of those sounds by native Eng- compromise VOT values falling be- lish speakers. tween the categories present in the Ll and L2, suggesting that learners L2 Speech Perception Constraints may alter their perceptive categories Early in their linguistic experi- in response to stimuli. This finding ence, humans learn to organize the was later confirmed by Flege (1987) enormous set of perceived sounds by his classifying them as speech or non- the and forms basis for hy- "equivalence the classification" and, former, speech as for pothesis, which proposes that learn- sounds with phonemic significance ers may group similar Ll and L2 in the Ll phonological inventory, in phones into one category based on much the same way as they organize variety of color into the infinite such compromise values. According categories such as red, pink, or or- to this hypothesis, a learner projects Ll phonetic categories onto the L2 ange. With respect to L2 acquisition, whenever the sounds are judged by the research suggests that learners in the learner to be equivalent; new early stages routinely categorize L2 phonetic categories are formed only speech sounds that they perceive in when the the terms of their L1 phonological in- perceives learner sounds as different. One effect that ventory. To test the hypothesis that such proposed rules L1 phonemic for categorization Flege (1995) that influence perception of other speech classification equivalence is cases of continued perceptual link- sounds, Scholes (1967) designed an experiment in which listeners from age of Ll and L2 sounds limit the ac- a variety of language backgrounds curacy with which L2 sounds may be were presented aurally with a set of produced (Flege, 1995). synthetic vowel stimuli. The result- L2 Speech Production Constraints ing data showed that subjects tended Just as perceptual categories are to organize the stimuli in accordance established for Ll processing, Bor- with the vowel systems of their na- and Yoshioka den, Harris, tive languages, suggesting an Ll in- Fitch, fluence in vowel perception. In ex- (1981) claim that speakers have men- periments in which subjects were tally pre-established representations of muscular gestures that are neces- asked to categorize initial occlusives, sary to produce the articulatory tar- for which voice onset time (VOT) is production an acoustic cue, Lisker & Abramson Accordingly, get. L2 (1964) reported that listeners regu- would presumably be limited, either larly separated a continuum of stop by the degree of similarity between consonant stimuli varying in VOT the L1 and L2 targets or by the degree 4 Acquisition of Stops and Trills in L2 Spanish 103 to which the learner is able to suc- The Speech Learning Model cessfully establish new gestural rep- In its current form, the Speech resentations. Borden (1980) suggests Learning Model (Flege, 1995) pres- ents four postulates and seven hy- that self-perception plays an impor- tant role in establishing a link be- potheses concerned with the ulti- tween the perception and produc- mate attainment of L2 pronuncia- tion. The SLM claims that learners tion of novel phonetic targets in that the learner progressively modifies of an L2 must create accurate percep- gestural representations until audi- tual "targets" to guide them in the production of L2 sounds; failure to tory feedback indicates to the learner that the L2 target has been met satis- do so will result in inaccurately pro- factorily. Direct realist accounts of duced targets. The first postulate of speech learning, such as that pro- the SLM proposes that the same de- vices that are used by learners to posed by Best (1995), point out that learners have proprioceptive access learn their native language (L1) can to the gestures used to create speech be accessed at any age and applied to sounds and are able to learn effi- L2 learning. Since the present study ciently the important elements of treats the adult [1] acquisition of L2 Spanish, this postulate is of consid- the gestures used to create L1 speech sounds. Best proposed that this ges- erable importance since it provides tural proprioceptivity leads to the the assumption that speech learning formation of relational "lower-order processes remain accessible to all L2 which may gradually learners, regardless of age. Of the invariants," model hypotheses give seven language-specific, the way to of "higher-order invariants," causing a given by Flege (1995, p. 239), the sec- ond, third, and seventh are particu- amount lower-order reduced of phonetic detail to be detected and larly relevant to the present study thus potentially interfering with the and are listed here: mechanisms used in the learning of Hypothesis 2: A new phonetic new sounds. A question that has generated category can be established for an L2 controversy literature sound that differs phonetically from in the is whether L2 perception precedes pro- the closest Ll sound if bilinguals dis- cern at least some of the phonetic duction or whether accurate produc- tion can come before (or without) differences between the Ll and L2 perception. Some studies, such as sounds. (1979), and (1963), Neufeld Lane The greater the Flege (1987), have suggested that ac- Hypothesis 3: curate perception must come before perceived phonetic dissimilarity be- production, but other research, such tween an L2 sound and the closest Ll as Gass's (1984) study on English sound, the more likely that it is learners' VOT production, suggests phonetic the between differences that accurate production may pre- sounds will be discerned. cede perception. 5 104 Texas Papers in Foreign Language Education The production acoustic correlate used to measure Hypothesis 7: stop consonant production is VOT of a sound eventually corresponds to the VOT continuum, properties the represented [2]. Within its in Lisker and Abramson found that phonetic category representation. most languages tend to cluster VOT around three or fewer categories of Hypotheses 2 and 3 predict that learners will be able to create a new values: long voicing lead, zero onset or short lag, and long voicing lag. phonetic category once they perceive that a sound differs from a corre- Keating (1984) proposed describing these as phonetic categories, such sponding sound in the Ll; the like- that Lisker and Abramson's catego- lihood of this occurring increases as the differences between the Ll and ries would be phonetically realized as the following: voiced (also re- L2 sounds magnify. According to Hypothesis 7, once learners have es- ferred to as prevoiced), voiceless un- tablished such a phonetic category aspirated, and voiceless aspirated, representation for a novel sound, respectively. Figure 1 illustrates the their production of that sound will phonological differences for stop between English and eventually correspond to that of na- consonants tive speakers of the L2, provided Spanish along the VOT continuum. The figure shows that English uses their phonetic categories were accu- all three phonetic categories, with rately represented. the voiced and voiceless unaspirated categories used for voiced phonemes Description of Sounds and the voiceless aspirated category The Phonemes /pl, Itl, /k/ used for voiceless phonemes. Span- The phonemes /p/, /t/, and /k/ ish, however, uses only two of the are present in both English and phonetic categories. In Spanish, the voiced category represents voiced apparent their Spanish. Despite similarity, however, in English the phonemes and the voiceless unaspi- rated category is used for voiceless allophones [ph] (aspirated voiceless phonemes. The voiceless aspirated bilabial stop), [th] (aspirated voiceless phonetic category is not used in stop), and alveolar (aspirated [kh] voiceless velar stop) may occur in Spanish. word-initial position or at the be- The Phoneme Irl ginning stressed of syllable, a /r/ has no whereas in Spanish the three pho- The Spanish trill nemes in question each have only counterpart in any dialect of Ameri- can English, nor in most other dia- syllable-initial one allo- possible lects of English. Although this pho- realization phonic voiceless (the neme's allophonic distribution may bilabial stop [p] for /p/, the voiceless include allophones that are voiced dental stop [t] for /t/, and the voice- or voiceless fricatives, uvular trills, less velar stop [k] for /k/). and voiced proposed originally alveolar voiceless or As by trills, the most common allophone Lisker and Abramson (1964), and as by many since expanded further in most dialects and the one that is other standard most frequently taught to learners of the researchers, 6 Acquisition of Stops and Trills in L2 Spanish 105 9E C 0 9'4' R 8C R E 8 C 8( T - L2 Spanish is the voiced alveolar of Spanish stop consonants, tion while a control group received no trill represented by the International Phonetic Alphabet symbol /r/. The special instruction. Over the course of one semester, she found most readily that identifiable acoustic learners in the experimental group correlate of the /r/ is a regular inter- were able to significantly shorten ruption in the waveform and spec- VOT in their production of /p/ and trograph caused by the brief, periodic cessations of phonation that corre- /g/, thus rendering their production spond to each contact between the more Spanish-like. the Although /t/, /k/, /13/, and /d/ also improved, tongue and the alveolar area. the degree of improvement was not statistically significant, which Gon- Existing Research On L2 Spanish Phonemes /p/, /V, /k/, and Id zdlez-Bueno attributes to the interac- Although interest in the acqui- tive- operation of developmental and sition of L2 Spanish stop consonants transfer processes. Flege and Eefting (1988) con- by English-speaking populations has developed relatively recently, there ducted an experiment in which Eng- have been a number of important lish monolinguals, Spanish mono- linguals, and native Spanish bilin- advances. González-Bueno (1997) tested two groups of intermediate guals imitated a consonant-vowel learners of Spanish in a foreign lan- continuum in which the VOT of the guage setting. The first, an experi- As ex- initial consonant varied. mental group, received explicit in- pected, the Spanish monolinguals struction and practice in the produc- showed a tendency to produce stops 7 106 Texas Papers in Foreign Language Education with Spanish-like VOT values (short subjects were recruited from among and those who had no significant child- English long the lead) lag, monolinguals tended produce hood background with Spanish, nei- to English-like stops (short lag, long through ther formal significant lag). With the bilingual group, how- study prior to age 12, through resi- ever, the researchers reported stop dence or extensive travel in Span- production in all three VOT ranges, nor ish-speaking communities, suggesting that those subjects had through family contact. The native processed the stops in terms of the English-speaking subjects are catego- three phonetic categories present in rized as below and are described in both languages. Table 1. Contrasting with the amount of attention that has been given to the Learners): (Beginning Level 1 The 10 participants in this group acquisition of L2 Spanish stop con- sonants as evidenced by the afore- were students enrolled for credit in a mentioned studies, the trill has not university Spanish first-semester been the subject of recent published language course. Consequently, study studies. this hopes to fill a gap in the research by Level 2 (Intermediate Learners): The 10 participants in this group directing inquiry to the L2 acquisi- tion of the /r/ among adult English were students enrolled for credit in a Spanish speakers. third-semester university language course. METHOD This section of this paper de- 3 (Advanced Learners): Level scribes the research method em- The 10 participants in this group ployed in this cross-sectional study. were enrolled for credit in an upper- division or graduate-level univer- A description of basic characteristics of the participants as well as infor- sity course in Spanish language, lit- mation about their background with erature, or culture. L2 Spanish is given. Following the description of the subjects, an expla- Level 4 (Very Advanced Learn- ers): The 10 participants nation of the data collection proce- this in dure and an outline of the scoring group were native speakers of Eng- lish who were full-time procedure are provided. instruc- faculty teaching university tional courses in Spanish language, litera- Subjects The present study examines ture, or culture and who otherwise met the criteria for inclusion in the data gathered in 1997 from 45 volun- teers affiliated with the Spanish pro- study. gram at a medium-sized, private native Spanish-speaking The university in the United States. Of participants from were recruited these participants, 40 were native among teachers of Spanish at the speakers of English and the remain- native university. same These ing 5 were native speakers of Span- speakers, from Colombia, Chile, native English-speaking The ish. a Proficiency of Bilingual Education Teachers 107 Table 1 Participant Characteristics (Group Means) Level 2 Level 1 Level 3 Level 4 Age (years) 19.7 41.8 21.6 18.5 H.S. Span. (years) 2.2 2.3 2.5 2.2 Univ Span. (semesters) 2.6 21.6 7.9 1.1 Hrs Span used/day 0.9 1.8 4.1 1.2 Days in Span country 4.2 5.0 18.3 1615.5 Mexico, and Spain, were selected to ticipate in the study, each participant provide speech data for purposes of completed the first part of the study, a questionnaire that identified the represent and comparison, they many of the most common varieties relevant individual learner charac- teristics and established a profile of of Spanish. each level of subjects, as seen in Ta- Data Elicitation ble 1. The second part of data collec- tion was a two-alternative, forced- The data collection process con- choice discrimination test designed sisted of three separate types of data- to measure perception of discrete gathering methods. A questionnaire target language items among collected relevant biographical data 12 and learner characteristics, a forced- minimal pairs. The stimulus for this portion of the experiment was a choice minimal pair test prompted by a Spanish audio recording pro- previously recorded audio recording vided a measure of auditory dis- of a native Spanish speaker reading crimination, and one of the minimal pair items; sub- different three types of audio recordings of each jects circled the item they believed subject's Spanish captured L2 pro- they heard on a score sheet. The re- duction. To ensure consistency and cording was presented via a Sony cross-level comparability of data, the TCM-919 cassette player. data collection procedure was identi- The third part of this study re- corded the subjects' production of cal for each participant, regardless of level. All instructions were given in Spanish. This part consisted of three English. different segments, each involving a After signing a form indicating different elicitation protocol. In the their informed willingness to par- first these segments, the of re- 9 108 Texas Papers in Foreign Language Education searcher instructed subjects to read RESULTS from a list of ten Spanish words or The results section presents the phrases embedded in a carrier phrase findings of the perception and speak- common to all items (e.g. "digo - ing elicitation tests. These findings, tabular and graphic presented in esta vez"). These items were selected formats, are also analyzed statisti- to provide a wide variety of phone- cally to evaluate the significance of mic targets and to measure sensitiv- ity to (and influence from) ortho- the findings. graphic cues in an elicitation proto- col where all or most of the partici- Perception Test pant's attention could be directed to the results two- The of pronunciation. In the second seg- alternative forced-choice perception test for each level are presented in ment, also designed to measure dis- Figure 1. The perception test in this crete lexical items, participants were study provides a general assessment sequentially shown 11 picture cards, of the participants' perception of each with a drawing of a relatively Spanish minimal pairs, including 5 common item. The task was for the vocalic and 6 consonantal features. subjects to say the names of the ob- jects in Spanish. Finally, in the third These results show that in this study at more advanced levels speech elicitation protocol, partici- learners pants provided a 30-second guided discriminate better to were able among Spanish minimal pairs than narration in Spanish in response to a written cue. This segment pro- those at beginning levels, although learners at all levels showed a high vided data from a context in which the participant's mental resources percentage of correct responses, sug- gesting that most learners correctly were presumably engaged in many perceive phonemic features in Span- elements of language production be- sides pronunciation. ish. Speech data for all participants Spanish Production Data: Stop Con- were recorded as follows: Each sub- sonants ject was recorded individually in the The data collected from the researcher's office (which, while not an anechoic chamber, speech elicitation tests appear in Ta- includes a bles 2 and 3. Table 2 indicates the number of sound muffling features). Having been prompted, shortest, longest, and mean VOT subjects spoke into a Sony SV-9 microphone values recorded for each stop conso- (the data are also mounted on a Nissin tripod stand. nant arranged graphically in Figures 2 through 7). The resulting input fed directly into The results of the ANOVA shown the audio input of a Macintosh desk- in Table 3 suggest that stop conso- top computer, which then processed nants are produced with a signifi- and recorded the signals digitally us- ing Signalyze speech analysis soft- cantly more Spanish-like VOT when ware. Jo