Universality and Typology in Phonotactic Constraints: Evidence from Amharic and English

Lado’s (1957) work Linguistics across Cultures lays the theoretical foundations for the CAH¹. Lado (1957: 2) states, “Those elements that are similar to the learner’s native language will be simple for him, and those areas that are different will be difficult.” Based on the Contrastive Analysis Hypothesis, areas of interference and/or learning difficulties arise, particularly where the two languages diverge. Similarly, Gluth (2003: 1) defined CA as “the methodical examination of pairs of languages to determine the structural distinctions and similarities between them.” This would imply that CAH asserts that the L1’s “interference” is mostly responsible for the majority of mistakes committed when learning the L2.

According to Lado (1957), ESL or EFL learners tend to transfer their native elements of language and cultural forms and meanings to the target language and culture. This transfer can happen both receptively and constructively when attempting to actively speak the language and behave appropriately in the culture, as well as comprehend and absorb the language and culture as performed by native speakers, respectively.

As the 1970s began, CA was losing its credibility in some ways. Abbas (1995) claims that the fundamental flaw of CA is that it overemphasizes a single kind of error, “interference.” The other criticism is the lack of sufficient empirical evidence that supports the CA hypothesis. It was discovered that many of the faults or mistakes predicted or identified by CA were not evident in learners’ language. However, according to Fisiak (1981: 7), CA should be conducted despite some flaws because not all CA hypotheses are incorrect. It is advised or proposed that teachers should use CA in conjunction with error analysis to address its inadequacies or weaknesses. It serves a helpful explanatory role because it is implemented by pinpointing the mistakes that the students have practically made in the classroom. That is to say, it can still be used to justify some faults and blunders. “Research shows that contrastive analysis may be most predictive at the level of phonology and least predictive at the syntactic level” (Schackne 2002: 199).

The study of error, both in type and number, especially in applied linguistics, is known as EA. EA focuses on actual production errors; It assumes that faults signify learning challenges and that the frequency of a specific error indicates how hard it is for learners to acquire that particular form (Brown 1994). According to the EA approach, CA is insufficient as a technique for pinpointing every area where second language learners struggle (Hammarberg 1972). EA proponents say that to identify the source of interference, we should first list the mistakes of EFL learners, followed by an appeal to conflicts between the learners’ first language and EFL structures (Jain 1975). This indicates that EA proponents are content to employ CA, but only to explain observed errors rather than to forecast errors. However, they deny the notion of error prediction, pointing out that errors that Contrastive Analysis cannot forecast can still be empirically identified using error analysis. In this sense, doing an error analysis study is helpful to pedagogic implications by shedding light on specific language areas to focus on teaching and syllabus writing (Ellis 1997). Nonetheless, EA is also criticized. Some scholars, such as Schackne (2002), claim that EA is insufficient as a foundation for identifying all possible learning issues because it detects only the errors that L2 learners actually make.

We can conclude from the above discussion that CA and EA have advantages and disadvantages of their own. The fundamental distinction between CA and EA is that CA attempts to forecast potential L2 faults, whilst the latter empirically detects errors observed in L2 production. Therefore, in this article, CA and EA are used complementarily since the errors made by the students can be explained by CA and, empirically, can determine the actual errors made by them.

The initial position clusters in English can contain no more than three consonants (CCC-). In the case of three-consonant clusters, /s/ must be the first phoneme; the voiceless stops /p/, /t/, or /k/ must be the second element in the cluster, and the set /l, r, w, y/ must be the third sounds. This type of cluster demonstrates the peculiar behavior of (/s/ + consonant) clusters, which defy the rules that other two-consonant clusters follow, such as spew /spj/, stew /stj/, skew /skj/, splash /splæʃ/, and script /skrɪpt/. Uniquely, /s/ can form a three-consonant cluster initially with either /p/, /t/, or /k/, followed by one of the phonemes in the set /l, r, y, w/ as a third element. A list of these kinds of clusters is adapted from Gimson (1962/1989: 245) as follows:

Download Original Figure

Roach (2014: 68) posits that the two-consonant clusters on the onset in English are of two types. One type consists of the voiceless, alveolar fricative phoneme /s/ followed by /p, t, k, w, j, l, m, n/ as in the words speak, steak, skin, swim, sue, slow, smuggle,and snuggle. He refers to the /s/ phoneme in these clusters of pre-initial consonants, whereas /p, t, k, w, j, l, n, m/ are the initial consonants. The second type starts with /b, d, ց, p, t, k, l, m, n, f, θ, ʃ, h, v, or (z)/ followed by one of the sets of the four approximants /l, r, w, j/ as in, for example, blade, dwell, grant, plate, twins, and quite, etc. Roach refers to the first consonant of these clusters as the initial and the second cluster as the post-initial consonant.

In English, if two consonant clusters occur syllable-initially, one from the set /p, t, k, b, d, g, m, f, v, θ, s, h/ must be the first element with the patterns: /sm/, /sn/, /st/, /sk/, /sp/, /sw/, /sl/, /dw/, /dr/, /θw/, /θr/, /pl/, /pr/, /py/, /tr/, /tw/, /bl/, /br/, /by/, /fl/, /fr/, /fy/, /kl/, /kr/, /ky/, /kw/, /vy/, /hy/, /my/, /gr/, /gl/, /gw/ (cf. Gimson 1962/1989: 244; Kreidler 2003: 115).

Since monosyllabic words do not have any medial consonant(s), a syllable-based approach is inappropriate when analyzing medial consonant clusters. To analyze consonant sequences that cross syllable boundaries, or medial consonant clusters, polysyllabic words are suitable. To examine sequences of consonants spanning word margins, an utterance is also deemed an idea. According to O’Connor (1998: 76–77), four consecutive consonants are the maximum number accepted by median sequences across syllable boundaries, and seven successive consonant clusters can be allowed across word boundaries. Three types of medial sequences exist across syllable boundaries as shown in Kreidler (2003: 128).

Download Original Figure

The clusters which occur post-vocalically, i.e., after a vowel at the end of a syllable, are called word-final clusters. In English, there are two, three, or four final consonant clusters as presented below.

2.4.4.3. Final Four-Consonant Clusters (-VCCCC)

Four consonant clusters in the word/syllable-final position are the maximum size allowed in English, but this pattern is rare. It can only be detected when the VCCC form has the inflectional affixes /s/ and /t/ attached to it. According to Roach (2014: 69–73), some good instances are /ntθs/, /mpts/, /ksts/, /mpst/, /lfθs/, /ksθs/, which stand for thousands, exceptions, texts, glimpses, twelfths, and sixths, respectively. These are frequently reduced to triconsonantal clusters by some speakers.

Various phonologists have proposed different maximum syllable sizes for English. For instance, Selkirk (1982) suggests CCVVCC to be the maximum syllable size, where VV is either a long vowel or a diphthong, as in flounce. On the contrary, Blevins (1995) proposed CCVVC as the maximum syllable size, as in dream. I think such proposals will need to find an alternative method or syllable size that can handle terms like strengths /streŋθs/, which exhibits a CCCVCCC syllable pattern.

Generally speaking, there are two main types of syllables: closed and open. A closed syllable is a syllable that contains a consonant in the coda, whereas a syllable that does not contain a consonant in the coda is called an open syllable. Although English has more syllable types, the following are instances of English syllable structures.

Download Original Figure

Two languages might vary greatly in the combinations they allow. When learning to enunciate SL or FL, this could be a major cause of difficulties. There are two or three initial consonant clusters in English, as in place and split. Despite English admitting three consonant clusters word-initially, the English word rsping is not feasible since /s/ is always the first in clusters of three. In addition, word/syllable-finally, English can allow up to four consecutive consonant clusters as in texts /teksts/.

In Amharic, however, there can never be more than two consonants in a cluster. Besides, Amharic does not permit initial consonant clusters, but if it does, initial consonant clusters are limited to Stop + Liquid types only. When more than three consonant clusters appear in the medial position across a morpheme boundary, the epenthetic vowel /ɨ/ will be inserted to break up the impermissible cluster. For example, while pronouncing words such as sport, spring, split, place, drive, blow, flow, etc., students frequently insert [ɨ] between the English consonant clusters that are not allowed in their native language, Amharic. As a result, learning the complex initial and final consonant clusters in English is challenging for Amharic learners of English. This, in turn, can lead to pronunciation discrepancies between English words, such as blow vs. below, flow vs. follow, sport vs. support, spirit vs. sprite, and drive vs. derive, and thus can hinder communication.

An illustration of the English and Amharic syllable structures is presented in the following table.

Download Original Figure

Since Amharic does not permit consonant clusters on the onset of a syllable, and more than two consonant clusters word-finally, Amharic learners of English will be challenged by the initial position CC, CCC and, final CCC, CCCC of English syllables.

Table 1 presents the results of the respondents after hearing ten of the word pairs. The last vowel -i is unstressed in every instance.

Except for “eight,” which was sometimes perceived as “eighty,” most final -C words were correctly perceived. Although they were minor, the problems were with words that ended with -Ci. As Table 1 depicts, in perceiving words like lucky, funny, honesty, city, and difficulty, almost no significant number of students committed any errors. The percentages were greater for the remaining five words. 5%–10% of the respondents misheard the words eighty, party, and sixty as eight, part, and six, respectively. This could be because /ɪ/ is devoiced because of the preceding voiceless consonant /t/. However, the mistake rate was 18% in Germany and 32.8% in dirty. There are specific explanations for the high percentages in these final two instances. It is common to see some Ethiopians use German to refer to the country, the language, and the people; Germany may be used very rarely by the participants. According to my observations, the term dirty has been used by many Ethiopians as a portmanteau word to refer to a noun and an adjective (see Amharic k'oʃaʃa). For native Amharic speakers and most Ethiopians, Germany and dirty are foreign. These students misheard German and dirt instead of the correct words Germany and dirty.

The possibility of adding an epenthetic schwa across a word boundary between final -C and initial C- in Amharic may be the reason for errors. This could cause some people to perceive English words with the final -Ci incorrectly, believing that the unstressed final -i is just an epenthetic vowel and is hence “not there.”

The majority of the issues related to clusters CC are being mistaken with CVC; only rarely did the reverse occur. This is a strategy for modifying English clusters that are not likely to occur in their native language. With a CCC triple cluster—a type that is not present in Amharic—the word disprove contained the greatest percentage of errors (45.9%). Overall, the mistake percentages are fairly low in other words. In the word pair flood/fluid, both directions were difficult (27.9%, 14.5%). This is most likely because university students are more familiar with the word fluid (e.g., correction fluid), which is pronounced as [flud], which is halfway between the canonical English pronunciations of [flᴧd] ‘flood’ and [flu:ɪd] ‘fluid.’ Regarding the word chains, a sizable portion (36.1%) mistook it for Chinese. This looks puzzling because regular Amharic words, like chains, can end in the cluster /nz/ in words such as /wənz/ ‘river.’ The perception error may emanate from the un-Amharic /eɪ/ diphthong or non-existent diphthong found in the English word chain (Table 2).

The results of the production of word pairs with final -C vs final -Ci are presented in Table 3 below.

Table 3 depicts that the majority of participants had no trouble producing words with the syllables C and Ci. There was a single outlier: 38 participants (62.3%) pronounced lucky as luck by dropping the last /ɪ/. My view is that, for the participants, lucky and luck are lexically treated equally; there is typically no linguistic distinction between the two, and hence treated as luck. This could be explained by the vowel becoming devoiced, which makes it harder to hear.

Although the error percentage is minimal, except for luck, Amharic learners of English found it more difficult to pronounce the final -Ci English words than those with the final -C.

Word-initial, medial and ultimate three-consonant sequences proved to be substantially more challenging for the participants than sequences containing only two consonants. Table 4 demonstrates that none of the participants were able to accurately produce the words weaknesses, sixth, thanks, students, strong, disprove and strengths. For instance, the students pronounced the initial consonant cluster in the word strong /strɑːŋ/ as /ɨstɨrong/ or /ɨstɨron-g/; they pronounced the word’s medial consonant cluster in the word disprove /dispru:v/ as /disɨpruv/, /dis-²pruv/, /dispərov/, or /dis-apruv/. The students uttered or pronounced the final consonant clusters of the word students /stuːdənts/ as /ɨstudentɨs/, /studentɨs/, /ɨstudent-s/, or /studenɨts/.

In dealing with two-consonant clusters, 21 (34.4%) students produced blow disyllabically as below, pronouncing it as /bɨloʊ/, /bɨlaʊ/, /bɨlo/, and /bIloʊ/. 38 (62.3%) of the participants pronounced flow disyllabically, pronouncing it as /foloʊ/, /fɨloʊ/, /fɔ:loʊ/, and /fʊlo/; only 2 (3.3 %) students correctly said /floʊ/.

The word health was the other problematic word that the students found troublesome to correctly pronounce. All the 61 (100%) participants pronounced the word health /helθ/ as /helzɪ/ or /helz/ with the phoneme /z/. Predictably, the “dark” velarized /l/ of American English was replaced by the Amharic clear /l/.