According to Larsen-Freeman & Long (1991: 59), errors are a sign that learners are actively engaged in hypothesis testing which would

CHAPTER TWO LITERATURE REVIEW

Current research in the areas of learner errors, L1 speech errors, L2 speech errors, crosslinguistic variation in phonological processing, and the segmental inventories, phonotactics and syllable structures of English and Mandarin Chinese will be reviewed in this chapter to provide material with which to serve as the theoretical basis for the analyses of this study.

The organization of this chapter is as follows:

2.1 Learner Errors

According to Larsen-Freeman & Long (1991: 59), errors are a sign that learners are actively engaged in hypothesis testing which would

ultimately result in the acquisition of target language rules. Corder (1967), focusing attention on errors from language processing and language acquisition perspectives, introduced the idea that errors are a necessary part of linguistic development and are of significance because they may represent the discrepancy between the grammar of the learner’s

transitional competence and that of the target language (Lennon 1991:

181). In addition, he noted:

A learner’s errors…are significant in [that] they provide to the researcher evidence of how language is learned or acquired, what stages or procedures the learner is employing in the discovery of the language. (Cited in Brown 2000: 216)

Corder (1974) further elaborates on the procedures for error analysis, distinguishing between five stages, which are given below:

(1) selection of a corpus of language (2) identification of errors in corpus (3) classification of the errors identified

(4) explanation of the psycholinguistic causes of the errors (5) evaluation or error gravity ranking of the error

(Cited in Lennon 1991: 181)

As Corder (1967: 167) put it, errors analysis is indispensable to the

teacher in that errors reveal how far the learner has progressed and what

remains for the learner to learn (Corder 1967: 167). Dulay et al. (1982: 140)

point out that the study of learners' errors has been a primary focus of L2

research during the last decade. Studying learners' errors serves two major

purposes: (1) it provides data from which inferences about the nature of

the learning process can be made; and (2) it indicates to teachers and

curriculum developers which part of the target language students have

most difficulty producing correctly and which error types detract most

from a learner's ability to communicate effectively (Dulay et al., 1982: 138).

It is based on this belief that lexical errors are taken as the subject of this study.

2.2 Orthography and Cross-linguistic Differences in Phonological Processing

Fromkin (1993: 371) divides the major writing systems of the world into three categories: word or logographic writing, syllable writing, and alphabetic writing. In a logographic system (e.g. Chinese), the written symbol or character represents both the meaning and pronunciation of a word or morpheme. In syllabic systems (e.g. Japanese), each symbol represents a syllable. In alphabetic systems (e.g. English), each symbol represents one phoneme.

Would differences in orthography lead to differences in phonological storage strategy? If so, given that fundamental differences in the three major orthographic systems exist in the type of phonological information provided by a single graphemic unit, the most salient distinctions in lexicon retrieval should be found in the process of obtaining phonological code, or phonological processing.

According to Frost (1992), two types of abstract codes, orthographic

and phonologic, can theoretically mediate lexical access. (Frost 1992: 256)

Koda (1989: 202) further groups the three orthographic systems into two

contrasting types: phonography (e.g. English and Japanese Kana) and

morphography (e.g. Chinese and Japanese Kanji). The main difference

between the two kinds of writing system is that there is strong

script-speech correspondence in the former system, whereas the latter system maintains a strong relation between form and meaning.

Koda (1997) has established that prior orthographic experience has a strong impact on the development of L2 lexical processing, and that L2 learners from different L1 orthographic backgrounds utilize distinct L2 processing strategies. From this we would expect to find

orthography-motivated differences between the lexical encoding strategies of Chinese EFL students and English native speakers, but need to be

careful in separating what is motivated by orthographic factors, and what is motivated by phonological structure.

2.3 L1 Speech and Recall Errors

In discussing the nature of the human mental lexicon, Aitchison (1994:

19) noted that evidence in the form of speech and recall errors (i.e., “slips of the tongue”, “slips of the ear”) is valuable for several reasons. First, when speakers pick a wrong order, they often think, perhaps only momentarily, that they have grabbed the right one. Second, everyone makes slips of the tongue, no matter how well-educated they are, so speech errors reflect the working of the normal brain. Third, we can build up a data bank of recurring types of errors from which we can investigate the nature of lexical storage and retrieval processes.

According to Aitchison, speech and recall errors fall into two major

categories: assemblage errors and selection errors. An example of an

assemblage error would be the substitution of par cark for ‘car park’, in which the right items have been chosen but assembled in the wrong order, whereas in selection errors, a wrong item appears to have been selected from the mental lexicon. For example, corporal punishment may be replaced with ‘capital punishment.’ The second type of error can shed light on the mental lexicon if it is assumed that anyone who accidentally produces a wrong word is likely to pick one closely related to the intended word or

‘target’ (Aitchison 1994: 19). Aitchison further pointed out that selection

errors may be based on meaning similarity, sound similarity, or both. A

summary of error types gleaned from the work of Aitchison is given in

Table 1.

Table 1. Types of selection errors

F EATURE PRESERVED S INGLE ERROR B ^LEND Initial consonants condom

(condiment) Final consonants monument

(ornament)

lustrious

(lustful & illustrious)

Stressed vowel orgasms (organisms)

Templemead station (temptation)

SOUND

Number of syllables

copulating

(percolating) porcupines (concubines)

Uncle very thing (Huckleberry Finn)

S INGLE ERROR B LEND

MEANING

crosswords (jigsaws)

clarinets

(castanets) tummach

(tummy & stomach) expose

(expect & suppose)

(Aitchison & Straf 1982: 200, 236; Aitchison 1994: 20)

The typology in Table 1 is compiled from some 680 speech errors collected by Aitchison & Straf (1982), of which 472 were uttered by British adults and 208 by British children. Aitchison and Straf concluded that there are some important similarities between adult speech errors and children’s speech errors. Both support the assumption that in lexical storage and retrieval, some phonological characteristics of a word are more salient than others. Furthermore, for both adults and children, the phonological characteristics are independent of one another. Both adults and children look for clusters of salient features as if in retrieval a person first searched for the initial consonant of a word, then sought the final among those with the same initial, and so on. Aitchison and Straf (1982:

211) specified that differences between adults and children show up most clearly in the treatment of initial consonant, number of syllables, and stressed vowel. The majority of speech errors preserved the initial consonant and the number of syllables, but when one of the two

conditions had to be dropped, adults preferred to change the number of syllables, whereas children altered the initial consonant. In other words, adults gave a higher priority to the initial consonant than children, while children valued the syllabic rhythm of a word more than adults.

To determine whether similar selection errors are made by speakers

of languages using a logographic script, e.g., Chinese, Chu-Chang and

Loritz (1977) studied how a group of 22 Cantonese-speaking Chinese, aged

16 to 20, process Chinese ideographs in the short-term memory. A Chinese

Word Recognition Test, which consisted of stimulus and response lists,

was conducted. The subjects were exposed to a total of 90 stimulus

characters and 105 response characters, with phonological, semantic, visual, and random distracters. The subjects’ recall errors were predominantly phonological; 51.9% of the errors were

phonologically-motivated, as opposed to 34.7% for visual errors. The results suggested that the word recognition strategy in the subjects’ native language is predominantly phonological, brining us to the conclusion that lexical storage in a language with a logographic writing system may be identical to that of languages with alphabetic scripts.

2.4 L2 Speech and Recall Errors

How do human beings acquire, organize and process lexical

knowledge when more than one language is involved? Singleton (1999: 80) made clear that

The two major differences between the L1 and the L2 learner are that the latter, on the one hand, is at a more advanced stage of development in both physical and cognitive terms and, on the other hand, has already been through the process of acquiring a language. L2 learners do not coo or babble, and when they utter in their target languages their utterances are from the outset mostly comprised of

meaningful elements.

Chu-Chang and Loritz (1977) had a group of 22 Cantonese-speaking

Chinese, aged 16 to 20, take part in an experiment in which the subjects

were shown a four-word list, after which they were asked to choose from a response wordlist the words they had just seen. The process was repeated 18 times. On the response wordlists, 36 of the words were correct answers while the rest were phonological, semantic, visual, and random distrcters.

A group of Spanish-speaking subjects served as the comparison group.

Both groups made significantly more visual than phonological errors. The Chinese subjects had a visual error rate of 58.1%, against a phonological error rate of 22.5%. The Spanish subjects had a visual error rate of 51.0%

against 30.6% for phonological errors. In second language lexical storage and retrieval, it appears that both groups adopted a predominantly visual strategy.

Chu-Chang and Loritz suggested that the development of a mature phonological strategy in the first language can be discussed in terms of a two-stage model. Stage 1 corresponds to the period when short-term memory representation of written material is predominantly visual, and Stage 2 corresponds to a later period of predominantly phonological representation. Both Chinese and Spanish learners of English as a second language used a visual strategy for remembering words, leading Chu- Chang and Loritz to hypothesize that strategies employed in second language lexical storage are often similar to those used in Stage 1 in first language acquisition.

Hayes (1988) replicated the first part of Chu Chang and Loritz’s study with Mandarin as the target language using seventeen native speakers of Chinese from Taiwan and seventeen Americans who were highly

proficient readers of Mandarin Chinese. Both groups of subjects were

educated to college level. The Taiwanese made significantly more phonological errors than the American subjects. When the errors were broken down into different types, it was found that the Taiwanese also made significantly more phonological than graphic or semantic errors of recall. In contrast, no such clear-cut finding for the non-native readers was found.

Hayes also conducted another experiment -- a character recognition test disguised within a sentence validity test. One distracter, graphic, phonological, or semantic, invalidated each sentence. The data revealed that at the sentence level, the predominant strategy for native readers seemed to be a mix of graphic and semantic strategies. In contrast,

non-native readers made significantly more graphic errors than the natives, and significantly more graphic errors than either phonological or semantic errors. This would indicate that native readers are reading more for

meaning at the sentence level, while non-native readers are sill attending predominantly to graphic features. Hayes’s results support Chu- Chang and Loritz’s conclusion that short-term memory storage for speakers of a language with a logographic writing system is identical to that of speakers whose language has an alphabetic system. The American subjects’

performance suggests that literate adult second language learners are more dependent on visual representation for retention of lexical items in the target language.

Channell (1998, in Carter & McCarthy: 88), citing a well-known

experiment by Brown and McNeil in 1966, states that at some level L1 and

L2 lexicons are similar in that L2 learners experience the tip of the tongue

phenomenon in the same way as L1 speakers. However, evidence that the L2 user's mental lexicon of a given language resembles the L1 user's lexicon is sparse. There are both similarities and differences in lexical behavior. Research in L2 speech errors is urgently needed to enable comparisons and provide a clearer understanding of the nature of the L2 mental lexicon.

2.5 Phonological Encoding of English and Mandarin Chinese

Service and Kohonen (1995) followed 44 Finnish school-aged children from the time they started learning English over a period of three school years. They found that phonological short-term memory, reflected in those children’s ability to repeat English-sounding pseudowords, is a good predictor of learning English as a foreign language. Service and Kohonen propose that it is time to turn our attention to the role of phonological representations in teaching and learning new words.

At the phonological level, words can be divided into syllables, which in turn are made up of melodic strings of sounds. All major approaches to phonology have recognized the syllable as a fundamental unit in

phonological analysis (Blevins 1996: 206). Linguists have observed restrictions in syllable structures of different languages and many

proposals have been made regarding the internal structures of syllables, as

illustrated in the following figure, which shows three possible syllable

structures for the English word cap.

Figure 1. Flat structure

Figure 2. Onset-rime structure

Figure 3. Body-coda structure Syllable

Margin Nucleus Margin k æ p

Syllable

Onset Rime

k Nucleus Coda æ p

Syllable Body Coda Onset Nucleus k æ p

Of the three possible structures, the “Onset-Rime” model is adopted in

our treatment of English and Mandarin Chinese. The decision is based on

Kessler and Treiman’s (1997: 295) analysis of some 2001 English words, in

which they found a significant connection between the main vowel and

the following consonant (coda), but no significant association between the initial consonant (onset) and the main vowel. Their finding supports the idea that English syllables are composed of an onset and a vowel-coda rime.

The Onset-Rime model can also be applied to Chinese: after examining the pros and cons of various proposals for the Mandarin Chinese syllable structure, Li (1999: 82) opts for this model due to evidence from speech errors and the tone-bearing ability of the main vowel.

In the Onset-Rime partition of the syllable, the Onset refers to consonants at the beginning of the syllable; the Rime is the rest of the syllable after the onset. The Rime can be subdivided into a Nucleus and a Coda. In the following sections, we present a classification of constituents allowed in these positional slots within the syllable—Onset, Nucleus, Coda—in both English and Mandarin Chinese. Through this, we wish to clarify fundamental differences in the syllable internal structures of the two languages and determine whether certain English syllable structures may present difficulties for speakers of Mandarin Chinese trying to learn English.

A. Onset:

(A) English:

English syllable onsets may consist of from zero to three

consonants. Any single consonant from the inventory of English

consonants can begin a syllable, except for /
/. In addition, /  / only

comes initially in borrowed words such as “genre”. (Bird, S. 1999)

Table 2. A basic inventory of English consonant phonemes

bilabial labio- dental

dental alveolar palato- alveolar

palatal velar glottal

stop p b t d t  d  k g

nasal (stop) m n

fricative f v   s z   h (central)

approximant

 j w

lateral

(approximant)

l

(Ladefoged 1993: 37)

Below we present a classification of the English syllable onset according to the number of consonants contained therein:

(1) Zero onset:

e.g. in, on, egg

(2) Single consonant onset:

Table 3. Single consonant onset and examples of English

p pill t till
ring

b bill d dill h heal

m mill n nil l leaf

f feel s seal  reef

v veal z zeal j you

 thigh t  chill w which

 thy d  Jill

 shill k kill

 genre g gill

(From Fromkin & Rodman 1993: 185; Giegerich 1992: 34)

(3) Two consonant cluster:

i. /s/ followed by /p, t, k, m, n, f, l, w, y/:

Table 4. Two consonant cluster onset in English (1)

p t k f m n l r w y s + + + + + + + - + +

(Jensen 1993: 66)

e.g. spa, stop, sky, sphere, smooth, snow, slow, sway, suit

ii. A consonant followed by /l, w, r/:

Table 5. Two consonant cluster onset in English (2)

l r w p + + b + + f + + v

t + + d + +

 + +



z

 + +



t  d 

k + + g + +

(Jensen 1993: 66; Giegerich 1992: 155)

e.g. blow, beauty, pray, thwart

(4) Three consonant cluster:

In English, a word may begin with up to three consonants, but no more than three. If a word does begin with three consonants, the first will always be [ s ], the second must be chosen from among the voiceless stops [ p t k ] and the third from among the liquids [ l r ] or glides [ w y ].

Table 6. Three-consonant onset in English

r l w

k screen sclera squeak

p sprain splice --

t strip -- --

(Jensen 1993: 67)

(B) Chinese

Standard Mandarin Chinese has nineteen consonants and three

semivowels. The Mandarin Chinese consonant onset may consist of

zero, a consonant or semivowel, or a consonant and a semivowel. All

consonants except [
] can occur in onset positions.

Table 7. A basic inventory of Mandarin consonants and semivowels

bilabial labial dental retroflex palatal velar

stop p p

t t

k k

nasal m n

fricative f s   x

affricate ts ts

t  t 

t  t 

liquid l

glide j w

(Duanmu 2000:26; Wan 1997: 419)

Next, we classify the Mandarin Chinese onset according to the number of consonants contained therein:

(1) Zero onset:

e.g. /an/ ‘safe’, /aj/ ‘love’

(2) Single consonant onset:

i. CV: e.g. /ta/ ‘he or she’, /pau/ ‘hold’

ii. GV: e.g. /wa/ ‘baby’, /ja/ ‘tooth’, / e/ ‘moon’

Table 8. Mandarin Chinese consonantal and semivocalic onsets p pa ‘dad’ k

k

a ‘card’
i
‘hard’

p

p

‘afraid’ x xu ‘lake’ t  t  a ‘home’

m ma ‘mom’ ts tsai ‘at’ t 

t 

a ‘just’

f fa ‘law’ ts

ts

a ‘wipe’   a ‘blind’

t ta ‘big’ s sa ‘throw’

t

t

a ‘he/she’ t  t  au ‘seek’

n na ‘there’ t 

t 

au ‘noisy’

l la ‘spicy’   au ‘less’

k ku ‘cruel’ r rau ‘circle’

(Duanmu 2000: 26; Li 1999: 53)

(3) CG combinations:

Table 9. CG combinations in Mandarin Chinese

j w

p p

an ‘change’ - -

p

p

an ‘cheat’ - -

m m

an ‘noodles’ - -

f - - -

t t

an ‘store’ t

an ‘broken’ - t

t

an ‘sky’ t

an ‘roll’ - n n

an ‘year’ n

an ‘warm’ n

e ‘cruel’

l l

an ‘link’ l

an ‘messy’ l

e ‘omit’

ts t  an ‘sharp’ t

wan ‘drill’ t 

an ‘donate’

j w

ts

t 

an ‘owe’ ts

an ‘’usurp t 

an ‘persude’

s  an ‘thread’ s

an ‘garlic’ 

an ‘select’

k - k

an ‘close’ -

k

- k

an ‘broad’ -

j w

x - x

an ‘joyful’ -

t  - t 

an ‘brick’ -

t 

- t 

an ‘boat’ -

 - 

an ‘bolt’ -

r - r

an ‘soft’ -

(Duanmu 2000: 29)

B. Nucleus:

(A) English

As syllable nuclei are most often sonorant sounds [-consonant]

and since all vowels are [-consonant], syllable nuclei are vowels in most cases. But some segments that have been treated as consonants are also [-consonant], namely glides /j/ and /w/. Glides are

sonorants because like other sonorant sounds, the air is essentially unobstructed as it moves through the oral cavity. At this point, a clear understanding of what is meant by “vowel” and “consonant” is needed.

Giegerich (1992: 166) makes clear that these terms are defined partly in terms of the phonetic content of a segment and partly in terms of its phonological function, i.e. of the role it plays in a syllable.

Glides are different from vowels in that they do not form the peak of a syllable. [-consonantal] segments are consonants if they occur in the onset ([j] and [w]), and vowels if they occur elsewhere. In addition, in the English syllable, liquids /l, r/ and nasals /m, n/ can be

syllabic—that is, they may constitute separate syllables, as in the

words medal, feather, mutton, and rhythm (Fromkin & Rodman 1993:

199).

Table 10. A basic inventory of vowels of American English

i heed hid

e hayed  head

æ had a  hard

a hod  hawed

hood o hoed

u who’d  Hudd

 herd a hide

a how  hoy

(Ladefoged 1993: 37) (B) Chinese

As syllable nuclei are most often sonorant sounds, syllable nuclei are vowels in most cases. In Mandarin Chinese, the glides [j w ] can be identified with the corresponding high vowels [i u y ] since the two sets do not contrast with each other (Duanmu 2000: 25).

Following Duanmu’s convention (2000: 79), these segments will be

represented as glides in the prenuclear position and as high vowels in

the nuclear or the coda position.

Table 11. Vowels of Standard Chinese

i li ‘force’  kan ‘dry’

y ly ‘filter’  ku  ‘fruit’

u lu ‘road’  sr: ‘color’

œ n

œ ‘torment’ *z tsz ‘word’

a ja ‘duck’ *  r: ‘sun’

 x  u ‘monkey’ *   : ‘two’

 

 ‘thank’

(Duanmu 2000: 38; Li 1999: 38~52)

* These vowels are described as apical vowels in Li and syllabic consonants in Duanmu.

C. Coda:

(A) English

(1) Zero coda:

If the nucleus contains a tense vowel, the coda may be empty.

e.g. too, see, do

(2) Single consonant coda:

All the consonants except /h, r, j, w/ can occur alone at the end of a syllable.

e.g. shake, fill, rim, grass, bed, leave

(3) Two consonant cluster:

(i) One of /m, n,
, l, s/ followed by a consonant.

e.g. jump, thank, round, old, risk,

(ii) A consonant followed by one of /s, z, t, d,  /, which may often represent inflectional suffixes.

e.g. sixths, bags, best, robed, fifth (4) Three consonant cluster:

Three-consonant codas are more limited.

Table 12. Three-consonant coda in English

/dst/ midst /lpt sculpt

/kst/ next /lkt/ mulct

/ks  / sixth /lks/ calx

/mpt/ tempt /lf  / twelfth

/mps/ mumps /rkt/ infarct

/nst/ against /rm  / warmth

/
st/ amongst /rpt/ excerpt

/
kt/ instinct /rps/ corpse

/
ks/ lynx

(Jensen 1993: 69)

(5) Four consonant cluster:

According to Jensen (1993: 70), four-consonant codas appear to be limited to cases where the three-consonant codas are followed by an inflectional suffix.

e.g. sixths, excerpts, instincts

(B) Chinese

(1) Zero coda:

e.g. /kwo/ ‘nation’, (2) Single consonant coda:

In coda position, only [n] and [
] can occur.

e.g. /an/ ‘safe’, /ban/ ‘move’, /ta
/ ‘soup’, /hwa
/ ‘yellow’

(3) Glide coda:

e.g. /k

waj/ ‘fast’, /fej/ ‘fly’, /taw/ ‘knife’

Using the classification presented above, we can compile a list of segments allowed in certain syllable positions in English but not allowed in the same positions in Mandarin Chinese. These structures may be difficult for EFL learners whose native language is Mandarin Chinese.

A. Onset:

(A) Single consonants :

/v,  ,  ,  ,  , z, t  , d  , h/

(B) Two consonant clusters

/sp, st, sk, sf, sm, sn, sl, pl, pr, bl, br, fl, fr, tr, dr, r, w,  r,  w, kr, gr / (C) Three consonant cluster:

/skr, skl, skw, spr, spl, str/

B. Nucleus:

/ , e , æ , a  , , o ,  ,  / C. Coda:

(A) Tense vowels:

/e , o ,  / (B) Single consonant:

/v,  ,  ,  ,  , z, t  , d  / (C) Two consonant cluster:

(i) /m, n,
, l, s/ followed by a consonant.

(ii) a consonant followed by /s, z, t, d,  / (D) Three consonant cluster:

/dst, kst, ks  , mpt, mps, nst,
st,
kt,
ks, lpt, lkt, lf  , rkt, rm  , rpt, rps/

(E) Four consonant structures

In the five sections above, we have provided an overview of previous

research in areas related to the issue at hand, namely L2 word recall in

Chinese native speakers learning English. It is hoped that these theoretical

tools will provide us with a framework for probing into the phonological

processing of Taiwanese EFL learner's lexical storage and retrieval in the

English language. The main points discussed above will be kept in mind

as we embark on test error data collection and immediate word-recall

experiment design, and enter into the discussion of preserved and altered

phonological features in the following chapters.