The Role of Native Language Transfer and Task Formality in the Acquisition of Spanish Spirantization
Mary L. Zampini
University of Arizona
Abstract: This study investigates the role that native language transfer and task formality play in the second language acquisition of the Spanish voiced stop phonemes /b d g/ and their spirantized variants, , in order to identify specific problems that beset learners. The results of a data-based experiment involving two groups of native English speakers studying Spanish reveal that native language transfer plays a prominent role in hampering the acquisition of the voiced spirants . Students largely fail to spirantize the voiced stops in L2 speech and incorrectly transfer the phonemic status of English // to Spanish, leading to a slower rate of acquisition of this phone. The presence of orthographic v also interferes with the acquisition of Spanish [b] and  and leads to a decrease in accurate pronunciation during formal reading tasks.
Key Words: second language acquisition (Spanish), phonology, spirantization, pronunciation, transfer, interlanguage variation, applied linguistics.
The goal of the present study is to investigate the role that native language transfer plays in the acquisition of the Spanish voiced stops by native speakers of English studying Spanish, as well as the effect that task formality (or attention paid to speech) has on the pronunciation of these phones. The acquisition of Spanish /b d g/ poses a problem for native speakers of English, because they behave differently in the two languages. In Spanish, they undergo a process known as spirantization in certain contexts; that is, their manner of articulation varies according to the position within the phonetic phrase in which they occur, resulting in either stop or fricative (i. e., �spirantized�) pronunciations. The distribution of the stop and fricative variants of /b d g/ has been traditionally described as follows: stops occur in initial position after a pause (as in Vamos [bámos], �let's go�; or Dámelo [dámelo], �Give it to me�), after a nasal segment (e. g., lindo[lido], �pretty�; un gato [ún gáto], �a cat�), and for /d/ only, after a lateral segment (e. g, aldea [adéa], �village�, but algo [álo], �something�). Fricatives (represented as  for /b d g/, respectively) appear in all other contexts. Furthermore, spirantization occurs both within words and across word boundaries. (77)
English also contains the voiced stop phonemes /b d g/. However, they exhibit stop pronunciations in all environments and fail to undergo spirantization. Another difference between the two languages involves the relationship between the two phones [d] and . The difference between occlusive [d] and fricative  in Spanish is purely allophonic; in other words, the phonetic realization of /d/ may vary without a corresponding change in meaning (as in tarde, �late�, pronounced [tárde] or [táre]). In English, on the other hand, both /d/ and // have phonemic status, as evidenced by minimal pairs such as den /dn/ vs. then /n/.
A third important difference concerns the pronunciation of English /d/ as an alveolar flap, /D/, in post-tonic intervocalic position: for example, reading, [rí:Di] (post-tonic /t/ exhibits this variation as well, as in little [lIDl]). The English flap [D] is perceptually very similar to the Spanish vibrant /r/ and may thus cause problems in the  acquisition of Spanish /d/ and /r/ by native English speakers. It may also cause problems in comprehension; to illustrate, if an English-speaking student pronounces the Spanish word, todo, as [tóDo], native speakers of Spanish may in fact perceive [tóro].
These differences between the two languages raise several interesting questions concerning acquisition. For example, given that  do not exist as variants of /b d g/ in English, do English-speaking students of Spanish acquire these phones readily? What kind of effect, if any, does the amount of attention paid to speech have on the pronunciation of /b d g/? Finally, does the acquisition of Spanish  and  differ significantly from the acquisition of Spanish , given the phonemic status of // in English? Questions such as these became the impetus for the current study.
The role of the flap [D] may also have important implications for the acquisition of Spanish /d/ and /r/, as mentioned. However, the current study focuses primarily on the acquisition of spirantization, and the influence of /D/ does not figure prominently in the analysis of the results. Since the environment in which [D] may arise in L2 speech is one in which Spanish always requires the spirantized variant  (i. e., intervocalic position), an examination of L2 pronunciation in this context will consider any non-spirantized form as incorrect, whether it be a stop [d] or flap [D]. The influence of English /D/ on L2 speech remains, however, an important issue for future research.
The transfer of phonological knowledge from a speaker's first language (L1) plays an important role in the acquisition of a second or foreign language phonology -so much so that early studies in second language (L2) acquisition often ignored the acquisition of phonology. Many researchers believed that errors in pronunciation resulted primarily, if not exclusively, from negative L1 transfer and therefore contained little of interest to them. The past decade, however, has witnessed a renewed interest in second language phonological acquisition, as well as a recognition that a variety of factors, in addition to L1 transfer, interact to shape learners' developing L2, or interlanguage, phonology.
Interlanguage (see Selinker, 1972; Corder, 1981) refers to the language learner's knowledge of L2 at any given point in time in the language learning process. It constitutes an incomplete model of the structure of L2 that contains only those forms and rules mastered by the learner. In addition, it may consist of generalizations that do not form part of the structure of L2 at all, such as information and knowledge that the learner transfers from L1. Interlanguage phonology, therefore, concerns itself with the models of pronunciation and phonemic inventories developed by students as they learn to speak a foreign language.
The proliferation of research on the many factors shaping the acquisition process does not belie the importance of L1 transfer in L2 phonology. As Gass and Selinker (1983) assert, �...language transfer is indeed a real and central phenomenon that must be considered in any full account of the second language acquisition process� (Gass and Selinker, 1983: 7). The behavior of the voiced stops in Spanish, for example, suggests that L1 transfer may play a prominent role in their L2 acquisition by native speakers of English. A common error committed by native English speakers studying Spanish involves the failure to produce spirantized variants of /b d g/. Such an error results from the direct transfer of L1 knowledge concerning the phonetic realization of voiced stops, since L1 (English) requires that /b d g/ be pronounced as stops, regardless of the phonetic environment in which they occur. Thus, the potentially influential role of L1 transfer in the acquisition of Spanish /b d g/ seems evident, and a concrete, data-based investigation will allow a more precise determination of the ways in which transfer shapes L2 speech production.
In addition to transfer, one of the many factors proposed to affect L2 speech production  involves the amount of attention paid to speech. Researchers have often investigated the effects of this in terms of the formality of specific linguistic activities. In fact, Tarone (1979, 1982, 1983) postulated that the amount of attention learners pay to speech directly causes variation in L2 pronunciation during tasks of differing levels of formality, and she proposed the Chameleon Model to account for this variation along a range of different types of L2 performance. According to the model, each type of task fits into a continuum of styles ranging from careless speech to the most formal of situations, with many intermediate styles found in between the two. The least careful (or least formal) styles, Tarone proposed, will contain the lowest number of accurate L2 forms. Accuracy will increase, however, as the student moves toward more careful styles of speech.
Formal tasks used to investigate variation in L2 pronunciation include the reading of word lists or paragraphs, as well as the recitation of learned material, while conversations and spontaneous speech typify more informal tasks. Dickerson & Dickerson (1977), for example, conducted an experiment that investigated the pronunciation of English /r/ by native Japanese speakers. Each participant read aloud a list of words (the most formal task), read a dialogue, and participated in a conversation (least formal task). According to the results, the highest number of errors in the pronunciation of English /r/ occurred during the conversational task, while the lowest number of errors appeared in the reading of the word list.
Major (1986,1987) made a proposal similar to Tarone's with respect to the incidence of L1 transfer errors and the formality of L2 speech. He proposed the Ontogeny Model of phonological acquisition to describe the relationship between both transfer errors and developmental errors as L2 pronunciation changes over time or according to discourse formality. Unlike L1 transfer errors, developmental errors emerge as a result of principles guiding Universal Grammar or general learning processes; they may resemble errors found in first language acquisition and may also exhibit pronunciations not found in either the native or target languages. With respect to the formality of a given situation, the Ontogeny Model predicts that transfer errors will appear frequently in casual speech but will decrease as speech becomes more formal. Developmental errors, on the other hand, will appear infrequently in the least formal styles of speech, will increase for a time as speech becomes more formal, but will decrease again in the most formal of situations.
The evidence offered by Major (1987) in support of his prediction concerning L1 transfer and stylistic shifts in speaking comes primarily from a pilot study that investigated several aspects of L2 English pronunciation by native speakers of Brazilian Portuguese. For example, the incidence of [i] insertion both word finally (as in the pronunciation, [dógi] for dog) and in word final consonant clusters was examined for four tasks of differing formality by two groups of learners at different stages in the acquisition process. Major notes that Portuguese does not contain word-final consonant clusters, nor word-final single consonants other than /s/; furthermore, [i] insertion in Portuguese is an L1 phenomenon that consistently appears in loan words containing such sequences. Therefore, the appearance of this same process in L2 speech would indicate that L1 transfer had taken place.
The results of Major's experiment showed a general decrease in errors involving [i] insertion as speech became more formal, although the differences found did not prove statistically significant. He nevertheless concluded that the results displayed a trend supporting the Ontogeny Model with respect to styles of speech, and he did find significant support for the model based on changes in pronunciation over time. Major recognized the need, however, for additional research in order to more adequately test the claims made by the Ontogeny Model. In addition, he noted that the model may not always predict the correct results in situations that involve the transfer of stylistic  variation rules from L1.
Beebe (1987), for example, carried out an investigation of L2 speech variation in which she obtained results that contradicted both the Chameleon and Ontogeny Models' predictions concerning the relationship between transfer errors and task formality. She tested native Thai speakers' pronunciation of English /r/ in word-initial and word-final position. The two tasks given for this experiment entailed an interview with a native speaker of English and the reading of a word list. Her results differed according to the position of /r/ within the word. She found that in word-final position the Thais' pronunciation of English /r/ corresponded to the predictions made by the Chameleon and Ontogeny Models -fewer errors occurred during the formal task of the experiment. For word-initial /r/, however, Beebe found that the Thai speakers committed more pronunciation errors during the formal task. Many of these errors revealed an L1 pronunciation of /r/. To explain this, Beebe pointed out that Thai contains a prestigious variant of word-initial /r/ -a trilled /r/- used only in careful speech. This trilled /r/ did not appear in the participants' L2 speech at all during the informal task, but accounted for 24. 4% of all pronunciations of /r/ in the formal task. Thus, Beebe concluded that the Thai subjects transferred the L1 notion of prestige variants for /r/ to L2, resulting in more errors during the formal, rather than the informal, task.
A similar situation may prove relevant for the acquisition of Spanish /b d g/ by native speakers of English. Hieke (1987) discusses several cases of absorption in English, a phenomenon which he broadly defines as any of a number of phonetic changes that may occur in informal or rapid speech. Examples of absorption include assimilation, elision, and the reduction of particular sounds. Hieke further states that, �[absorption processes] do not occur in strictly obligatory fashion any more than casual speech as such is obligatory. Usually such processes have been referred to as general tendencies and the matter is left at that; as a result, we really have no clear notion of their actual prevalence in [native] -and still less in [non-native]- speech� (Hieke, 1987: 42). His transcribed examples of casual speech in English include some instances of voiced stop spirantization (e. g., the word, about, pronounced as a  out; see Hieke, 1987: 43).
Other researchers have also documented voiced stop spirantization (primarily for /b/ and /g/) in casual English speech. In an examination of British English, Gimson (1989) found that �in rapid, familiar speech, where easy intelligibility rather than articulatory precision is the aim, the closure of plosives [i. e., stops] is often so weak that the corresponding fricative sound, without a preceding stop, is produced, especially in weakly accented intervocalic positions� (Gimson, 1989: 160). His illustrations include fricative  and  in the pronunciation of dagger and rubber, respectively. Brown (1990) also provides examples of this, showing weakened  in sequences such as we've been and he's been, and weakened  in ago (Brown, 1990: 79). However, she also notes that in the production of these weakened forms, although the articulators fail to make contact, allowing continuous airflow, they do not exhibit the �length and robust friction� (p. 78) of true fricatives.
The spirantization of English stops differs from Spanish, however, in two key respects. First, English spirantization is an occasional, non-obligatory phenomenon of casual speech, whereas Spanish spirantization exemplifies an instance of obligatory allophonic variation in that it occurs consistently in minimally intervocalic position for all dialects. Secondly, English speakers may spirantize not only the voiced stops in casual speech, but the voiceless stops as well. Gimson (1989) and Brown (1990) both mention the appearance of a voiceless bilabial fricative  in casual speech as a variant of occlusive [p]. Bronstein (1960) cites such a pronunciation for the word, cupful, in a discussion of casual American speech: [kfl] (Bronstein, 1960: 72). Spirantization of the voiceless stops in Spanish, however,  does not occur. Despite these differences, native speakers of English learning Spanish may exhibit a stronger tendency to spirantize the Spanish voiced stops in informal situations since this corresponds to possible behavior in L1. If so, this will result in patterns of pronunciation that contradict the claims made by the Chameleon and Ontogeny models of phonological acquisition.
Based upon both the behavior of /b d g/ in Spanish and English and the insights obtained from previous investigations in second language phonological acquisition, the present study tested the following two hypotheses:
1. Errors in the L2 pronunciation of Spanish /b d g/ will prove the result of L1 transfer.
2. English-speaking students of Spanish will exhibit an improved pronunciation with respect to /b d g/ during an informal conversational task rather than a formal reading task.
The study involved a total of 32 native English-speaking university students: seventeen enrolled in a second-semester intensive Spanish course and fifteen enrolled in a fourth-semester intensive Spanish course. A control group of five native speakers of Spanish also participated in the experiment to verify the Spanish spirantization context. A detailed discussion of the results obtained by the control group appears below.
The materials used for the experiment consist of a series of questions answered by each student (the informal task) and an excerpt from a culture text which the student read aloud (the formal task). The question s for the informal task dealt with aspects of student and university life; they all utilized the present tense and contained simple vocabulary, so as to ensure that each participant would understand them without difficulty. In addition, the questions remained very general in order to give each student an opportunity to speak at length and thus allow for a greater possibility that the phones in question would emerge during speech.
The reading passage for the formal task came from an intermediate culture text (Copeland, Kite, & Sandstedt, 1981: 2-3) and was chosen for its use of grammatical structures and lexical items to which all students had received considerable exposure. In addition, the text contained several instances of the target phones, but not blatantly so. Some few alterations were made to the original text, however, for the experiment, These changes entailed the elimination of some paragraphs, so that the passage would not appear too long to students unaccustomed to reading aloud in a foreign language, as well as the substitution of some present tense verbs to the past tense, so as to advance the text to the level of second and fourth-semester students enrolled in an intensive course. A native speaker of Spanish verified the edited version of the reading passage as grammatically correct, as well as natural and flowing in style. The materials used for both tasks appear in the Appendix.
Each student participant met individually with the interviewer in one audio-taped session lasting approximately ten minutes. Half of the students completed the informal task first, followed by the formal task, whereas the other half completed the two tasks in reverse order, so as to avoid the possibility that task order could somehow influence pronunciation and, hence, skew the results. After all students had completed both tasks, each occurrence of /b d g/ was transcribed. In addition, two other researchers (one a native speaker of Spanish with a linguistics background, the other a native speaker of English and doctoral candidate in Spanish linguistics) transcribed a subset of the data independently, with approximately 95% agreement to the original transcription found both times. Unresolved discrepancies were subsequently eliminated from the analysis.
As mentioned, a control group of native  Spanish speakers participated in the experiment in order to verify the spirantization context. Since the rules offered by theoretical linguistics are formulated so as to reflect the competence of an ideal native speaker, they usually do not account for dialectal or individual variation in their description of the phenomenon. Nevertheless, spoken language changes continually, and researchers often encounter deviations from the traditional norm, or rule, among native speakers of a given language. As an example, rules that describe Spanish spirantization usually derive spirantized variants in all contexts except after a pause, nasal consonant, or /l/ (in the case of /d/). However, several researchers have reported that for some dialects, spirantization occurs in a much more restricted context. Canfield (1981), for example, notes that a stop, rather than spirantized, pronunciation of Spanish /b d g/ represents a general tendency in much of Central America, Colombia, and parts of the Caribbean after any consonant or semivowel. Specifically, /b d g/ exhibit stop pronunciations after all consonants and semivowels in Nicaragua, Honduras, El Salvador, and most of Colombia. As for Puerto Rico, �[s] ince /l/ and /r/ are somewhat neutralized, along with /s/ syllable final, /b/, /d/, and /g/ tend to be occlusive more often than would otherwise be the case after these consonants: verdad [beldá], desde [déhde], or [déhe], Margarita [malgaríta]� (Canfield, 1981: 78). Stop pronunciations after /l r s/ occur in other areas of the Caribbean as well, in those dialects that show the same type of syllable final neutralization processes found in Puerto Rican Spanish. In fact, the nasal phonemes comprise the only consonants after which the realization of /b d g/ does not vary; spirantization never occurs in this context.
In another study, Fernández (1982) found the occlusive and fricative variants of /b d g/ in free variation after consonants and glides in Costa Rican Spanish, with a clear preference for occlusives. He states, �[t]he evidence... indicates that the allophones of /b d g/ in Costa Rican Spanish are tending towards a system in which the occlusives are heard in group-initial position and after consonants and glides, while the fricatives are heard elsewhere. This constitutes a re-ordering of the allophones within each phoneme, but does not affect the phonemic system, nor is comprehension in the least impaired� (Fernández, 1982: 135). He also notes that this re-ordering results in a simplification of the traditional system still found in peninsular Spanish. This parallels much of the data found in Canfield (1981).
Although some researchers have considered the pronunciation of stops in traditionally spirantizing contexts an archaism, Fernández proposes instead that it represents a case of regression, involving a change from a fricative pronunciation to an occlusive one. Evidence comes from studies which find spirantization after consonants a more frequent phenomenon among older speakers and comparisons of earlier descriptions of Costa Rican Spanish to more recent ones showing a change from fricative to occlusive pronunciations. Furthermore, Fernández considers the lack of occlusive pronunciations after consonants in even the most conservative peninsular dialects evidence against a theory of archaism. Finally, Fernández suggests that this change may result from hypercorrection, since in cases of emphasis or very careful speech, occlusives may appear in environments that otherwise exhibit spirants, even in standard dialects. He also tentatively suggests the possibility of indigenous influence on the pronunciation of /b d g/ in American Spanish, although he finds a lack of convincing evidence.
Given the evidence for variation in pronunciation of the voiced stops in American Spanish dialects, a group of native Spanish speakers read the same text given to the English-speaking students for the current experiment. The members of this control group were all Spanish instructors; some had even instructed the student participants of the experiment. Five native speakers participated, and the results mirrored the variation found in Canfield (1981). The  lone informant from Spain spirantized in the traditionally expected environments, whereas the other members of the control group -from either Puerto Rico or Colombia- all pronounced /b d g/ as stops to varying degrees after non-nasal consonants.
Due to this variability in pronunciation, all instances of /b d g/ following non-nasal consonants were excluded from the analysis of L2 speech. Students of Spanish will not necessarily acquire spirantization in these contexts if they receive little or no positive input. Furthermore, since either a stop or spirantized pronunciation would constitute an acceptable pronunciation in postconsonantal position, including them would have skewed the results. Therefore, the only spirantizing context retained for analysis in the current study involved /b d g/ in postvocalic position.
Results and Discussion
The tables contain the statistical results of the experiment. Table I shows the percentages of correct pronunciations for each phone ([b d g ]) for the two groups of students, as well as for the two tasks performed. The remaining tables provide information concerning statistical analyses of the data. A Chi-Square analysis that examined the relationship between the second and fourth-semester students for each phone failed to reveal any significant differences between the two groups. Therefore, all subsequent testing of the relationships between particular phones within each task and across tasks considered both groups of students as one. As mentioned, the two hypotheses guiding the study predicted that, first, errors in pronunciation would result from L1 transfer, and secondly, that students would exhibit more accurate L2 pronunciation of /b d g/ during the informal, rather than formal, task. The relevant results for each of these hypotheses are discussed in turn.
Table I Percentages of Correct Pronunciations
Type of Error Found
As seen in Table1, neither group of students experienced any difficulty with Spanish phonetic [d] and [g], an unsurprising result since Spanish and English behave similarly in this respect. The occasional mispronunciations of these phones (two involving [d] and two involving [g]) among the fourth semester students came about as the result of performance errors, such as slips of the tongue or stumbling over the pronunciation of a word. Unlike the results obtained for [d] and [g], however, mispronunciations of the expected stop [b] appeared rather frequently. All errors of this type involved the pronunciation of a voiced labiodental fricative [v] in words containing an orthographic v. As known, /v/ constitutes part of the English phonemic inventory, but not a part of Spanish at either a phonemic or allophonic level. Rather, an orthographic v in Spanish simply represents an alternative spelling for the phoneme /b/. Therefore, the appearance of phonetic [v] in the L2 Spanish speech of the English-speaking students clearly implies  that L1 transfer has taken place. Furthermore, this error occurred more often during the reading task than the conversational task, a fact which supplies additional evidence that orthography and knowledge of L1 influenced pronunciation. A more detailed discussion of the role of orthography in L2 pronunciation appears in the next section.
In the environments where Spanish required the spirantized variants , most errors in pronunciation entailed the realization of these sounds as stops. In addition, some mispronunciations of the expected bilabial fricative  involved the realization of labiodental [v] in words containing orthographic v. In both cases, the errors illustrate again the effects of native language transfer on L2 pronunciation.
Consider next the percentages of correct pronunciations of Spanish  (Table1). Both groups of students experienced greater difficulty with expected  as opposed to  or . These differences in pronunciation proved significant when subjected to a matched T-test that compared the means of various pairs of phones during both tasks. As shown in Table 2, a comparison between  and  in both the informal and formal tasks, as well as a comparison between  and  in the informal task, produced significant results. As suggested earlier, the phonemic status of // in English may account for the disparity found. If  cannot replace [d] in English without the possibility of creating confusion in comprehension, the transfer of this knowledge to L2 will cause students to instinctively reject  as a possible pronunciation of /d/. Thus, learners may acquire Spanish  at a slower rate than either  or . The results obtained here provide support for this hypothesis.
|Table 2 Matched T-test:|
|Comparison of Spirantized Variants|
|Degrees of Freedom = 31|
Paired t Value
|Is value significant?|
| vs. ||21.138||Yes||4.922||yes|
| vs. ||3.385||
| vs. ||17.753||
| vs. ||10.584||
| vs. ||5.257||
| vs. ||5.327||
To summarize, all errors in pronunciation of the Spanish voiced stops show the effects of negative L1 transfer, both in terms of the failure to spirantize /b d g/ and the transfer of English /v/ and /d/ to the developing phonemic inventory of L2 Spanish. The findings thus support and confirm the first hypothesis.
L2 Pronunciation and Task Formality
The second hypothesis predicted that both groups of students would exhibit more accurate pronunciation during the informal, rather than formal, task. The results, shown in Tables 1 and 3, support this hypothesis as well. As seen in Table 1, both groups of students exhibited less accurate pronunciation of Spanish [b ] during the reading task than during the conversational task. A series of matched T-tests compared the means of each phone in the two tasks, and as shown in Table 3, the differences proved significant for [b ].
Table 3 Matched T-test:
Comparison of Informal and Formal Tasks
|Degrees of Freedom= 31|
|Mean||Paired t Value||Is value Significant?|
|[b]i vs. [b]f||11737||2262||Yes|
|i vs. f||11487||3471||Yes|
|[d]i vs. [d]f||163||303||No|
|[g]i vs. [g]f||1.562||1||No|
|i vs. f||
The presence of absorption phenomena in informal English speech provides one possible account of the differences found,  at least for the expected spirant . Absorption may have influenced the pronunciation of expected  as well; however, stronger evidence for an alternate solution appears below. As discussed above, researchers have documented the optional spirantization of /b/ and /g/ in informal native English speech. This possibility for spirantization may have created a greater tendency to allow spirants to emerge in L2 speech during the informal task. In conversational tasks, learners generally concentrate more on communicating effectively through the use of accurate vocabulary and perhaps grammatical structures and less on an accurate enunciation of each word. A lack of attention to pronunciation may thus increase the probability of a weakening of particular sounds in context. In the case at hand, this lack of attention appears to have resulted in a more frequent emergence of the voiced velar spirant  during the informal task.
As for the differences obtained with expected [b] and , absorption could not have contributed to a more accurate pronunciation of [b] during the informal task. Although it may have affected the pronunciation of expected , the evidence suggests that orthography contributed more prominently to these results. Recall that all mispronunciations of expected [b], as well as some of expected , involved the realization of a labiodental fricative in words containing an orthographic v. If knowledge of L1 caused the students to pronounce an orthographic v as [v] in Spanish, as suggested above, it comes as no surprise that this would happen more often in a reading assignment than conversation, for the visual reinforcement of the written word serves to influence pronunciation. In a speaking exercise, on the other hand, the student focuses on orthography less and may even be unaware of the spelling of some words. The presence of orthographic v and the transfer of L1 phonological knowledge, therefore, provide the most probable explanation for the significant decrease in accuracy in pronunciation of Spanish [b] and  during the formal task.
The influence of orthography raises an interesting question concerning the possible negative effects it may have on overall L2 pronunciation in learning environments that emphasize or focus exclusively on the development of reading and writing skills. Although the overall success rates in pronunciation between the second-and fourth-semester students of the current experiment did not prove significantly different, the second-semester students nevertheless exhibited a higher percentage of correct pronunciations of expected [b] during the reading task: 54. 90% vs. only 44. 44% for the fourth-semester group; see Table 1. However, both groups performed almost equally with respect to expected [b] during the conversational task: a 61.18% success rate for the second-semester group vs. a 62. 67% success rate for the fourth-semester students. The difference in success rates for the reading task may have resulted from an increased emphasis on the development  of reading skills in the more advanced class. Many of the assignments given to the fourth-semester students that participated in this experiment involved the reading of short stories, novels and /or plays, and many of the discussions in class centred upon literary topics. The second-semester students, on the other hand, concentrated on developing primarily oral and aural skills, although they had started to develop reading skills as well. Hence, does a change in focus from oral /aural skills to reading /writing skills in the classroom adversely affect students' developing pronunciation abilities? The results of the current experiment suggest that orthography may indeed adversely affect pronunciation, and the possible effects of reading on overall L2 pronunciation require further investigation.
To summarize, the present results confirm the predictions made concerning pronunciation of the Spanish voiced stops and task formality, although the possibility for absorption in English appears to have played a smaller role than originally hypothesized. This seems logical, however, for the degree of informality in L2 speech does not often equal that found in L1 situations where absorption phenomena generally emerge. Nevertheless, the tendency to spirantize all L2 stops more often during the informal task parallels behaviour in L1, and the possible links between native language absorption and L2 speech production warrant further investigation.
Implications for Pedagogy and Language Acquisition Research
The results of the current study are important in several respects. First of all, the findings demonstrate that native language transfer hinders the acquisition of /b d g/ by native speakers of English studying Spanish, as evidenced by both the failure to spirantize /b d g/ in the necessary contexts and the phonetic realization of an orthographic v, as [v] instead of [b] or . Furthermore, of the three phonemes studied, the student participants pronounced spirantized  least often. These results suggest that Spanish language instructors must help students become aware of the allophonic variation exhibited by Spanish /b d g/ through specific oral exercises that highlight these phones in various positions within the word. In addition, instructors must place particular emphasis on the proper pronunciation of Spanish /d/ in order to help students overcome the L1 tendency of treating occlusive [d] and fricative  as two distinct phonemes.
Secondly, the fact that the differences in pronunciation between second- and fourth-semester students proved insignificant points to the difficulty with which students master Spanish spirantization. This further highlights the need for instructors to make learners aware of the allophonic variation inherent to the Spanish voiced stops. In order to achieve accurate L2 pronunciation, students must reconcile not only the phonemic differences between L1 and L2, but the allophonic differences as well. Although a normative speaker's failure to properly produce allophonic variants in L2 may not prevent comprehension by native speakers of the language, it will contribute to the persistence of a foreign accent and delay mastery of the L2 phonological system.
The effects of orthography on pronunciation also underscore the importance of continued pronunciation practice in the classroom, especially in classes that focus on the development of reading and writing skills. Such practice should not limit itself to free conversation; rather, instructors should incorporate exercises that specifically target those areas of orthography that prove most problematic and that may adversely affect pronunciation in L2. This calls attention to the need for additional research in both second language phonological acquisition and classroom pedagogy. First, researchers must investigate more fully the effects of orthography on students' developing L2 phonological competence and identify those areas that may negatively influence L2 speech. This in turn will allow instructors to develop more effective classroom materials and provide students an opportunity to overcome the difficulties  presented by orthography.
In conclusion, the present study has demonstrated that native language transfer plays a prominent role in the L2 acquisition of the Spanish voiced stops and spirantization in three principal ways. First, the absence of an obligatory allophonic rule of voiced stop spirantization in English manifested itself in the failure of students to spirantize Spanish /b d g/ in the necessary contexts. Secondly, the transfer of phonemic status of English // to Spanish produced a much slower rate of acquisition of Spanish  when compared to the acquisition of the other voiced spirants,  and . Thirdly, the transfer of English /v/ reduced accuracy of pronunciation for both Spanish [b] and  and occurred at a significantly greater rate during the formal reading task than the informal conversational task.
The results of the current study raise interesting questions for future acquisition research, including the role of orthography in L2 speech production and the possible effects of L1 absorption phenomena on L2 pronunciation. From a pedagogical point of view, the determination of the types of obstacles that students confront in the acquisition of the Spanish voiced stops provides an opportunity for instructors to develop specific materials aimed at overcoming those obstacles and facilitating acquisition. The acquisition of a second language phonology is a highly complex process, one that most adult learners will not master completely without some form of formal instruction or conscious awareness of the differences between L1 and L2. The development and reinforcement of proper L2 pronunciation skills should therefore remain a central component to any language program that seeks to achieve fluency and communicative competence. (78)
APPENDIX: Materials Used for the Experiment
Questions used for the informal task:
|�De dónde eres? �Dónde vives?|
|�Cómo es tu rutina diaria?|
|�Qué te gusta hacer durante el fin de semana?|
|�Cómo es tu clase de español? �Qué hacen Uds. durante la clase?|
Reading passage used for the formal task:
Los primeros habitantes de la península ibérica, en tiempos históricos, eran las tribus celtíberas, de origen no muy bien conocido. En el tercer siglo antes de Cristo, llegaron los romanos y convirtieron la península en una colonia romana. Establecieron la lengua latina, su sistema de gobierno y su organización social y económica. Más tarde introdujeron la religión cristiana. Se ha dicho que la península llegó a ser la colonia más romanizada de todas.
La lengua que adoptaron los habitantes de la península es la que se llama �el latín vulgar�, o sea la lengua del pueblo y no la lengua clásica. El español de hoy desciende de esa lengua. Todas las lenguas �neolatinas� como el portugués, el francés, el italiano, el rumano y el español se parecen tanto porque todas tienen como base la lengua latina.
Los conceptos del gobierno también tienen sus raíces en la época romana. La idea de formular leyes ideales que se puedan aplicar a todos los casos y la tendencia a refinarlas en los casos especiales sigue como base de la ley hispánica.
Los romanos consideraban a los pueblos conquistados como ciudadanos del imperio y este concepto determinó el sistema usado por los españoles en el Nuevo Mundo. La empresa colonial era una actividad dirigida por el rey, y las tierras descubiertas eran de él. Los productos de las colonias españolas se consideraban iguales a los de la península. (Adapted from Copeland, Kite, & Sandstedt, 1981: 2-3).
Beebe, Leslie M. �Sociolinguistic Variation and Style Shifting in Second Language Acquisition�. Interlanguage Phonology: The Acquisition of a Second Language Sound System. Eds. Georgette Ioup and Steven H. Weinberger. Cambridge, MA: Newbury, 1987: 378-88.
Branstine, Zoann. �Stop/Spirant Alternations in Spanish: On the Representation of Contrast�. Studies in the Linguistic Sciences 21 (1991): 1-22.
Bronstein, Arthur J. The Pronunciation of American English: An Introduction to Phonetics. New York: Appleton -Century-Crofts, 1960.
Brown, Gillian. Listening to Spoken English, 2nd ed. New York: Longman, 1990.
Canfield, D. Lincoln. Spanish Pronunciation in the Americas. Chicago: U Chicago P, 1981.
Copeland, John G., Ralph Kite, and Lynn Sandstedt. Civilización y cultura, 2nd ed. New York: Holt, Rinehart and Winston, 1981.
Corder, S. Pit. Error Analysis and Interlanguage. London: Oxford UP, 1981.
Cressey, William W. Spanish Phonology and Morphology: A Generative View. Washington, D.C.: Georgetown UP, 1978.
Dickerson, Lonna, and Wayne Dickerson. �Interlanguage Phonology: Current Research and Future Directions�. Actes du 5ème Colloque de Linguistique Apliquée de Neuchâtel: The Notions of Simplication, Interlanguages and Pidgins and Their Relation to Second Language Learning. Eds S. Pit Corder and E. Roulet. Geneva: Droz, 1977: 18-29.
Fernández, Joseph A. �The Allophones of /b d g/ in Costa Rican Spanish�. Orbis 31 (1982): 121-46.
Gass, Susan M., and Larry Selinker, eds. Language Transfer in Language Learning. Rowley, MA: Newbury, 1983.
Gimson, A. C. An Introduction to the Pronunciation of English. 4th ed. London: Edward Arnold, 1989.
Harris, James W. �La espirantización en castellano y la representación fonológica autosegmental�. Estudis Gramaticals 1 (1984): 149-67.
Hieke, Al. �Absorption and Fluency in Native and Non-Native Casual English Speech�. Second Languge Acquisition. Eds. Allan James and Jonathan Leather. Dordrecht: Foris, 1987: 41-58.
Lozano, María del Carmen. Stop and Spirant Alternations: Fortition and Spirantization Processes in Spanish Phonology. Bloomington, IN: Indiana U Linguistics Club, 1979.
Major, Roy C. �The Ontogeny Model: Evidence from L2 Acquisition of Spanish r�. Language Learning 36 (1986): 453-503.
_____. �A Model for Interlanguage Phonology�. Interlanguage Phonology: The Acquisition of a Second Language Sound System. Eds. Georgette Ioup and Steven H. Weinberger. Cambridge, MA: Newbury, 1987: 101-24.
Selinker, Larry. �Interlanguage�, International Review of Applied Linguistics 10 (1972): 209-31.
Tarone, Elaine. �Interlanguage as Chameleon�. Language learning 29 (1979): 181-9 1.
_____. �Systematicity and Attention in Interlanguage� Language Learning 32 (1982): 69-84.
_____. �On the Variability of Interlanguage Systems�. Applied Ligüistics 4 (1983): 142-63.