Final consonants occurring at the end of English words are difficult for English language learners (ELLs) to pronounce accurately; furthermore, consonant clusters that occur word-finally seem particularly challenging. According to universal language theories, most world languages have a preference for a syllable that ends in a vowel. Second, most languages greatly restrict the types and numbers of consonants that can occur at the end of a syllable or word-finally (Blevins, 1995). English is an unusual language in that it functions contrary to these two language universals as any English consonant (except /h/) may occur word-finally, and English word-final consonant clusters may contain three or, more rarely, four consonants. Therefore, because of this particular phonological markedness of English, many groups of English language learners struggle to produce English word-final consonant(s). Empirical studies conducted on native speakers of East Asian languages, Spanish, and Brazilian Portuguese have demonstrated that English language learners tend to favor one or two ways of modifying English word-final consonant clusters based on constraints of their native phonologies (Hancin-Bhatt & Bhatt, 1997; Kim, 2001; Tarone, 1980). In addition, language universals also appear to be a factor in the systematic errors that English language learners make in their interlanguage (Broselow, Chen, & Wang, 1998; Tarone, 1980).

Based on the existing research on East Asians’ pronunciation difficulty of English word-final consonants, the present study sought to investigate how consonant(s) that cannot occur in the native language might be repaired or modified according to L1 constraints and interlanguage. Korean was selected as a medium for examination based on its simple syllable structure (consonant-vowel-consonant). Korean restricts which consonants can occur syllable-finallyand word-finally; further, it does not allow consonant clusters. The research questions guiding the present study were as follows: Does the number of English word-final consonant(s) prove to be a factor in the types of modifications native Korean participants make? What environment conditions the processes or repairs? What characterizes repair processes according to English proficiency level?

Six native Koreans enrolled at an intensive English program (IEP) at an American university participated in the present study. Two participants were selected from each of the following levels (as determined by the IEP placement test, which includes both an oral interview and a listening task): beginning, intermediate, and advanced English proficiency. Participants were individually interviewed by the researcher once in an hour-long session. Participants completed three speaking tasks: reading a list of 150 words, watching and retelling a film clip by Disney Pixar, and answering questions in a personal narrative. All participants were digitally recorded; the speech samples were then uploaded onto a computer and phonetically analyzed by the researcher using Praat waveform and spectrogram software to determine if correct pronunciation was achieved or if errors of a phonetic nature occurred to the target consonant(s). Data was coded for type of error, number of final consonants, environment of the final consonant(s), and level of participant. Chi-squares were run on pair-wise variables to test for significant differences using SPSS.

The first research question addressed the types of modifications and the number of word-final consonant(s) in the target word. Modifications consisted of phonetic changes to the word-final consonant(s). Repairs were coded as correct, deletion, epenthesis, substitute manner, substitute place, substitute voice, or multiple repairs. A segment was correct if it was produced correctly by the participant. For deletion, a consonant was omitted; “grab” became /græ/. For epenthesis a vowel was added to the final consonant(s): “just” became /jus.tə/. For substitute manner the manner of articulation was changed; a stop became a fricative in /greifs/ “grapes.” For substitute place, the place of articulation changed as in /greiv/ “graze.” For substitute voice, obstruent voicing was changed; “cab” was pronounced /cæp/. Finally, in multiple repairs more than one of the aforementioned repairs occurred. Table 1 displays the interaction between number of consonants and type repairs. The results were significant: χ²(12, n = 3,269) = 464.913, p < .001.

Table 1

The Number of Word-Final Consonants and Type of Repair Process Used by All Participants

Table 1 shows that as the number of coda consonants increased, accuracy of production decreased. Overall correct production was 54.5 percent. Correct production of a single word-final consonant was 61.3 percent. The accuracy for two consonants fell to 38.0 percent. Correct production of three consonants fell to 22.6 percent. Among the repairs that were used to modify final consonants, deletion was the most frequent repair, particularly for one and two consonants. However, in the case of three consonants, multiple repairs and epenthesis were frequently used as well. Furthermore, three consonants also appeared to trigger an increase in multiple repairs.

The second research question addressed whether environment could trigger a specific modification used by participants. The seven repair types from Table 1 were again used in these environmental analyses tables. Environment was divided into four categories based on sonority: vowel, voiced obstruent, voiceless obstruent, and sonorant. Vowel described a vowel immediately preceding the word-final consonant as in “hat.” Voiced obstruent referred to the segment before the epenthetic repair /gɪvə/ “give.” Voiceless obstruent was the segment before the repair /fɪʃi/ “fish.” Sonorant was the segment before repair in /drɪlə/ “drill.” Tables 2 to 4 display the interactions of one, two, and three final consonants and the repairs that were most commonly used. The results were significant: χ²(18, n = 2344) = 2017.088, p < .001.

Table 2

The Preceding Segment of One Consonant and Repair Process Used by All Participants

Table 2 shows that in the case of one word-final consonant, overall accuracy was 61.3 percent. In accurately produced tokens, a vowel directly preceded the final consonant. The two most common repairs to a single final consonant were deletion and substitution of voicing (devoicing). In the cases of final obstruents, the numbers of epenthetic repairs were much higher than those in the preceding vowel and sonorant categories.

Table 3 shows production of two consonant clusters; the accuracy fell to 38 percent for these segments. Deletion remained the overall favored repair. The number of multiple repairs increased with the addition of a second final consonant. Again, epenthetic repairs, which were not favored overall, remained the preferred repair for environments of preceding voiced and voiceless obstruents. The results were significant: χ² (18, n = 863) = 879.094, p < .001.

Table 3

The Preceding Segment of Two Consonants and Repair Process Used by All Participants

Table 4 shows production of three consonant clusters; the accuracy fell to 22.6 percent. Deletion was again the favored repair, accounting for an overall total of 32.3 percent of all utterances, which exceeded the number of correct tokens. Multiple repairs were the second most common repair, accounting for 22.6 percent of utterances. Epenthesis accounted for 21.0 percent of all utterances. The results were significant χ² (12, n = 62) = 57.987, p < .001.

Table 4

The Preceding Segment of Three Consonants and Repair Process Used by All Participants

The third research question addressed whether English fluency level had an influence on the type of repair used. The fluency levels of the participants represented beginners, intermediates, or advanced English language learners. Participant fluency levels were determined by the IEP placement listening test and an oral interview. The seven repair categories from the previous tables were used again. The interaction of English fluency level and repair are shown in Table 5. The results were significant χ² (12, n = 3269) = 48.712, p < .001.

Table 5

Participants’ Levels of English Proficiency and Type of Repair Process

The results of Table 5 show that accuracy improved with proficiency level. Furthermore, the more advanced students were better able (or more willing) to communicate in English as evidenced by the total utterances produced in the far right column. Among all participants, deletion was the primary repair used in consonants and clusters. Substitution of voice was a distant second repair among all three proficiency levels. While overall errors did improve with proficiency, the repairs that were used fluctuated between the proficiency levels. Epenthesis in particular was used more frequently by the advanced learners than by beginning and intermediate learners.

The present findings that number of consonants and proficiency level were factors in pronunciation accuracy corroborated with previous studies on Korean (Kim, 2001; Kwon, 2007). However, one interesting finding in this study was the larger number of epenthetic repairs used by advanced learners. There appeared to be a trend that might be attributable to the participants’ experiences in which epenthetic repairs led to fewer misunderstandings than deletion repairs, a finding that corroborates Abrahamsson (2003).

The present study demonstrated that deletion was a preferred repair overall. Epenthesis was a secondary repair in the cases of preceding voiced and voiceless obstruent environments. The preference for epenthesis among obstruents could be a possible result of the numerous English obstruents that either do not exist or are not permitted word-finally in Korean. Korean linguists described both epenthetic vowel insertion and deletion as common repairs to native Korean illicit sequences and to loanwords from other languages (Ha, Johnson, & Kuehn, 2009). These two repairs could be attributed to native language influence and to universal preference for reduction of word-final consonants as well. The final repair to emerge with significance among the proficiency levels was that of devoicing. This final repair was again inconclusive as to whether it was a factor of native phonological processes or interlanguage preference for universals. Korean stops are differentiated by aspiration rather than voicing, so essentially all Korean stops are voiceless. A preference for voiceless final consonant(s) by participants in this study could be attributed to either, or perhaps both, native language and interlanguage influences. Based on the results of the study, it was difficult to differentiate the specific influences of native and interlanguage universal preferences.

Potential areas of future investigation would be a further division among final consonant environments to possibly include those segments native to Korean and those that are nonnative to examine how these two variables interact with data. A second area of potential research might be to attempt to isolate some of the native Korean phonetic factors and those that might be caused by a possible preference for universals in interlanguage.

REFERENCES

Abrahamsson, N. (2003). Development and recoverability of L2 codas: A longitudinal study of Chinese-Swedish interphonology. Studies in Second Language Acquisition, 25(3),313-349.

Blevins, J. (1995). The syllable in phonology theory. In J. Goldsmith (Ed.), A Handbook of Phonological Theory, (pp. 206-240). Oxford, England: Blackwell.

Broselow, E., Chen, S. I., & Wang, C. (1998). The emergence of the unmarked in second language phonology. Studies in Second Language Acquisition, 20(2),261-280.

Ha, S., Johnson, C. J., & Kuehn, D. P., (2009). Characteristics of Korean phonology: Review, tutorial, and case studies of Korean children speaking English. Journal of Communication Disorders, 42, 163-179.

Hancin-Bhatt, B., & Bhatt, R. M. (1997). Optimal L2 syllables. Studies in Second Language Acquisition, 19, 331-378.

Kim, H. K., (2001). The interlanguage phonology of Korean learners of English: A computational implementation of optimality theoretic constraints. Dissertation Abstracts international: Section A. Humanities and Social Sciences, 61(11), 4362A-4363A.

Kwon, B. Y. (2007). Korean speakers’ production of English consonant clusters: Articulatory and perceptual accounts. Dissertation Abstracts international: Section A. Humanities and Social Sciences, 67(10), 3800.

Tarone, E. E. (1980). Some influences on the syllable structure of interlanguage phonology, IRAL, 18(2), 139-152.

Elizabeth Conway is a PhD student at Rutgers University. She is interested in segmental differences between languages and the phonetic/phonological interface that present challenges to English language learners’ pronunciation.

LinkedIn: http://www.linkedin.com/pub/elizabeth-conway/14/250/263