Toward a Taxonomy of Written Errors: Investigation Into the Written Errors of Hong Kong Cantonese ESL Learners

Toward a Taxonomy of Written Errors: Investigation Into the Written Errors of Hong Kong Cantonese ESL LearnersALICE Y.. CHAN City University of Hong Kong Hong Kong SAR, China This articl

Toward a Taxonomy of Written Errors: Investigation Into the Written Errors of Hong Kong Cantonese ESL Learners

ALICE Y W CHAN

City University of Hong Kong

Hong Kong SAR, China

This article examines common lexicogrammatical problems found inCantonese English as a second language (ESL) learners’ writtenEnglish output A study was conducted with 387 student participants,who were asked to do two untutored and unaided free-writing tasks ofabout 200–300 words each A range of lexicogrammatical error typescommonly found among Hong Kong Cantonese ESL learners wasidentified Errors from the lexical level included vocabulary compensa-tion and inaccurate directionality; errors from the syntactic levelincluded calquing, existential structures, incorrect ordering of adver-bials, and independent clauses as subjects; and those from thediscourse level included periphrastic-topic constructions Mother-tongue influence was inevitably an important source of the problems,but inadequate mastery of correct usage of the target language anduniversal processes were also important factors The results of the studyhave potential for enhancing our understanding of the interlanguagegrammar of learners and the nature, sources, and prevalence of learnerproblems The results also have promising pedagogical implications, asthey inform teachers of the levels, nature, sources, prevalence, andgravity of learner errors and equip them with the key ingredientsneeded for the design of appropriate remedial instructional materials

A discussion of how the taxonomical classification would be useful forlanguage teachers is also given

doi: 10.5054/tq.2010.219941

E nglish is a ‘‘value-added’’ language in Hong Kong, indispensable forboth upward and outward mobility, rather than a typical second orforeign language (Li, 1999, p 97) It is compulsorily taught at allsecondary and primary schools and is the medium of instruction ofabout one third of the total number of secondary schools and themajority of tertiary institutions Despite its official status and addedvalue, it is used in Hong Kong only in the formal domains ofgovernment, business, education, and law, typically in the presence of

native English speakers or non-Chinese speakers (Li, 1999) Forintraethnic communications, Cantonese or Cantonese-English mixedcode is preferred, and many Hong Kong Cantonese are under greatsocial pressure not to switch entirely to English when communicatingamong themselves orally (Li, 2000) Many informal written exchangesare conducted in Chinese-Cantonese,1 standard written Chinese, orEnglish-Chinese mixed code Cantonese English as a second language(ESL) learners’ exposure to English is very limited.

Many Hong Kong Cantonese ESL learners encounter problems

in learning the morphology, lexis, syntax, and semantics of English(e.g., Budge, 1989; Chan, 2003, 2004b, 2004c, 2004d; Chan, 1991;Chan, Kwan, & Li, 2002, 2003; Chan, Li, & Kwan, 2003; Gisborne, 2002;Green, 1991; Li, 2000; Webster & Lam, 1991; Webster, Ward, & Craig,1987; Yip & Matthews, 1991; Yu, 1988a, 1988b) Despite claims thatHong Kong English should be viewed as a legitimate new varietybecause of the existence of unique, systematic features (Bolton, 2002;Bolton & Lim, 2002; Hung, 2002), it has been argued that Hong KongEnglish is not appropriately characterized as a new variety of English,because of its limited social role and the predominance of StandardEnglish as the norms of reference (Li, 2000) Li contends that features inHong Kong English can be more appropriately viewed as interlanguagefeatures

ESL teachers need to have a good understanding of the cognitive andpsycholinguistic mechanisms at work in learners’ learning process inorder to help them overcome their second language (L2) problems.Because errors are indicative of a learner’s interlanguage2and the errorsmade along a learner’s interlanguage continuum are often due to acomplex interplay between both first language (L1)- and L2-relatedfactors (Li & Chan, 1999), there is a need to investigate the writtenoutput of Cantonese ESL learners in order to uncover the extent ofnegative transfer and the interaction between transfer and other non–L1-related factors Results of such research should have considerablepotential for alleviating English language teachers’ workload and forquickening students’ learning process Despite various attempts todiagnose Hong Kong ESL learners’ writing problems, there is a lack ofsystematic, large-scale studies which scrutinize a full range of writtenlexicogrammatical errors, analyze the possible causes, and establish a

1 The word Cantonese used in this article refers to the variety per se, which may be spoken or written The word Chinese refers to standard written Chinese The written Chinese used in Hong Kong is a mixture of spoken Cantonese and standard written Chinese (Snow, 2004), and the term Chinese-Cantonese is used to refer to this special medium.

2 There have been some criticisms of the term ‘‘interlanguage.’’ Cook (1993) points out that the term is often used to refer ‘‘both to the learner’s knowledge of the second language and to the actual speech of L2 learners’’ (p 19) No such distinction is made in this article.

taxonomical classification The present study aimed at bridging such aresearch gap.

ERROR ANALYSIS AND TRANSFER ANALYSIS

The approach used in the study could be seen as arising from theparadigms of error analysis (EA) and transfer analysis (TA) in L2acquisition (SLA) research EA (Corder, 1967) compares learners’interlanguage with the target language to locate mismatches Errors areseen as evidence of learning and could be described without the need torefer to a learner’s native language Earlier interest in error analysiswaned, because it was thought that, even if learner errors could bepredicted and understood, such errors could not be ameliorated EA wasalso attacked as ‘‘a pseudoprocedure in applied linguistics’’ (Bell, 1974,

p 35) and was insufficient because of its biased practice of ‘‘analyzingout the errors and neglecting the careful description of the non-errors’’(Hammarberg, 1974, p 185) Despite these criticisms during the 1970sand early 1980s, this paradigm has been revitalized followingsignificant research in the past decades, such as James (1998),Kellerman (1995), Kellerman and Sharwood Smith (1986), and Odlin(1989), leading to the redefinition of the concept of TA Though EA isnot a theory of acquisition, it is argued as a methodology for dealing withdata (Cook, 1993), and teachers are attracted to this paradigm ‘‘by itspromise of relevance to their everyday professional concerns’’ (James,

1998, p x)

TA (James, 1990) compares learners’ interlanguage strings with theirmother tongues It is a subprocedure in the diagnostic phase of EA anddeals with interlanguage and target language mismatches assumed to bethe results of mother-tongue interference Crosslinguistic influence isacknowledged, and learner errors are seen as a register of learners’current perspective on the target language (James, 1998) Althoughstructural comparisons of two languages are often uncertain correlates

of learner behavior (Kellerman, 1995), with the data-handling ology in EA and the acknowledgement of crosslinguistic influence in theidentification of target language–interlanguage mismatches in TA, thesetwo paradigms still remain useful means to understanding the cognitiveand linguistic complexities involved in SLA

method-PREVIOUS RESEARCH INTO WRITTEN ERRORS MADE BY CHINESE SPEAKERS

Numerous small-scale studies have been carried out to investigate thewritten errors made by Hong Kong Cantonese ESL learners Among the

problems documented include those with relative clauses, pluralmarking, and topicalization Mother-tongue interference has often beenargued as the major cause Webster et al (1987) present various localESL errors and conclude that all the errors may be attributed to mother-tongue influence Green (1991) examined the overuse of topic-comment structure in Hong Kong English and argues that the structure

is evidence of typological transfer Budge (1989) attributes Hong Kongstudents’ failure to mark plural nouns with -s in writing to the influence

of Cantonese phonology Outside of Hong Kong, Deterding (2000), Tan(2005), and Zhu (2007) examined the influence of Chinese on writtenSingaporean English Poedjosodarmo (2000) investigated the influences

of Malay on the written English of university students in Singapore.Although Tan (2005) argues that Singlish arose due to the influences ofthe students’ mother tongues on all the lexical, syntactic, and discourseaspects of English, Zhu (2007) claims that not all errors can beattributed to Chinese influences To the author’s knowledge, none ofthese studies has attempted to establish a systematic taxonomy oflexicogrammatical learner errors to arrive at a generalizable conclusionabout the cognitive and psycholinguistic mechanisms underlying thelearning process

OBJECTIVES

The present study aimed to identify a range of lexicogrammaticalerrors commonly found in Hong Kong Cantonese ESL learners’ writtenoutput and to establish an error taxonomy The underlying belief wasthat the nature and the causes of the errors could be more systematicallyand reliably generalized, if similar errors are classified into the sametype

PROCEDURE

Data Collection: Phase I

A study was conducted with 387 Hong Kong Cantonese ESL learners,including 65 students from three local universities and 322 studentsfrom five local secondary schools (124 students from Form 6 and 198students from Form 33) The Form 3 students (about 50% of the totalnumber of participants) and Form 6 students (about 30%) could becategorized as lower intermediate (L-I) and upper intermediate (U-I)learners, respectively, whereas the university students (about 20%) could

3 Form 3 and Form 6 students in Hong Kong are comparable to Grade 9 and Grade 12 students in the United States, respectively.

be categorized as advanced (A) learners Elementary learners were notincluded, because they are not accustomed to doing free writing inEnglish.

The participants did two free-writing tasks administered at twodifferent time slots of 40 minutes at an interval of about 2 weeks Aword limit was set at 200, 250, and 300 words for students at the threeproficiency levels, respectively The participants produced 696 pieces offree writing (totaling about 158,000 words) Of these, 187 pieces, 116pieces, and 47 pieces of narrative writing (a total of 350 pieces) werefrom the Form 3, Form 6, and university students, respectively; and 157pieces, 124 pieces, and 65 pieces of descriptive writing (a total of 346pieces) were from the students at the respective levels

Data Analysis

A research assistant, very proficient in English and holding a master’sdegree in English, was engaged in identifying anomalous structures fromthe corpus The errors were then assigned to a working error taxonomyunder the supervision of the researchers (the author and hercollaborator) To ensure that the research assistant could extend thepatterned anomalies to the entire corpus, the researchers coached theassistant in a series of error identification and categorization sessions.Accuracy and consistency were maximized by having a second researchassistant with similar linguistic background and training double-checkthe error taxonomy Where the two assistants’ judgments diverged,either one or both of the researchers reviewed the categorization tomake a third judgment (see appendix)

A comparison between the interlanguage strings and equivalentstrings in the learners’ mother tongue was then carried out to determinewhether crosslinguistic influences (Kellerman, 1995; Kellerman &Sharwood Smith, 1986) may have been at work Attempts were alsomade to ascribe the errors to possible sources where mother-tongueinterference could not be observed

Data Collection: Phase II

The error taxonomy established (see Results: Phase I) gave primafacie evidence of syntactic transfer from Chinese to English In thesecond phase of the study, five error types thought to be the results of L1interference, namely, omission of copulas, incorrect order of adverbials oradverbs, existential structures, misuse of relative clauses, and transitivity patternconfusion, were isolated, and the extent of syntactic transfer was furtherinvestigated with the use of individual interviews (including translation

tasks [from L1 to L2], explanations of translations, and self-reports)and group tests (including grammaticality judgment questionnairesand translation tasks [from L1 to L2]) The individual interviewswere administered to a focus group of 42 students, and the group tests

to a large sample of 710 students Differences between the performance

of different age groups were also examined (for details, see Chan,2004c)

1 *their academic results are still dissatisfactory

Overuse of affixes (11 tokens: 18% from L-I, 27% from U-I, 55% from A).Overuses of affixes were exemplars of overgeneralization, where theneed for an affix in word formation had been overgeneralized.Inadequate knowledge of the word class of a stem word was probably areason for such overgeneralizations, because the learners did not seem

to be aware that the original stems without the unwanted affixes sufficefor the meanings conveyed

2 *The happiness we have now cannot enlast

Lexical Level Altogether, 617 error tokens belonged to the lexical level.Inaccurate directionality (9 tokens: 22.2% from L-I, 66.7% from U-I, 11.1%from A) Such confusion was probably the result of mother-tongueinterference, as the substitution words and the target words often havesubstitutable L1 Chinese-Cantonese equivalents with no directionalitydifferences

3 *I borrowed money from my friends and borrowed the money to him [cf lend;Chinese-Cantonese employs the same word ze3 for both borrow and lend.4]

Synonym confusion (73 tokens: 58.9% from L-I, 35.6% from U-I, 5.5% fromA) These errors showed learners’ difficulties in differentiating theappropriate uses of near synonyms and the contexts in which they should

be used L1 influence may have been at work, as the confusable Englishsynonyms often share the same or similar Chinese-Cantonese equivalents

4 *My mother is nice, she didn’t fight me [cf beat; the Chinese-Cantoneseequivalents of fight and beat are daa2 gaau1, and daa2, respectively, which aresimilar.]

Vocabulary compensation (199 tokens: 37.7% from L-I, 47.2% from U-I,15.1% from A) For this error type, the synonymy relation between thesubstitution words and the target words holds only in the learner’smother tongue The substitution words (groups) and the target words(groups) have very different meanings and usage in English Mother-tongue influence was one probable cause

5 *Open TV and open the playstation [cf turn on; the Chinese-Cantoneseequivalent of turn on is hoi1, the same as the Chinese-Cantonese equivalent ofopen.]

Synforms (336 tokens: 54.5% from L-I, 33% from U-I, 12.5% from A).Synforms are lexical mis-hits selected because of formal resemblance toother L2 forms (Hall, 2002, p 71; Laufer, 1997) The learners’ insecureknowledge of both the target forms and their corresponding mis-hits wasprobably the major cause of the problem No mother-tongue inter-ference was observed

6 *I sleep on the bed, my mother also sleep nearly [cf nearby]

Syntactic Level Altogether, 4,295 error tokens were found at thesyntactic level

Pseudotough movement (11 tokens: 9.1% from L-I, 81.8% from U-I, 9.1%from A) Characterized by the use of a tough adjective (e.g., easy) in anerroneous structure (Yip, 1995), pseudotough movement has been regarded

a high-frequency erroneous structure for Cantonese ESL learners (Li &Chan, 1999)

4 All Chinese characters in this article are transliterated using the Jyutping system (Tang et al., 2002) The number at the end of each romanized Cantonese syllable is a tone mark, indicating one of the six distinctive tones in Hong Kong Cantonese.

7 *Up to now, we are not easy to work together [cf ngo5 mun4 bat1 jung4 ji6 jat1hei2 gung1 zok35(we not easy together work)].

The Chinese-Cantonese equivalents of the English erroneous structuresare both acceptable and common Mother-tongue interference wasprobably a major cause Acceptable sentences in the target language,such as John is not easy to convince, may also have led the learners intobelieving that these sentences with tough movement were positiveevidence in support of their interlanguage hypothesis for the pseudo-tough movement structure (Chan & Li, 2002)

misuse of until (33 tokens: 6.1% from L-I, 75.8% from U-I, 18.2% fromA) Rather than using the preposition to show that something happensduring a period before a particular time and stops at that time (Quirk,Greenbaum, Leech, & Svartvik, 1985), the learners used it with timeadverbials like forever or now to emphasize the truth of the precedingstatements at the time of speaking, including the time periods specified

by the time adverbials, traceable to Chinese-Cantonese expressions zik6zi3 or zik6 dou3 (Chan, 2003)

8 *Until now, I enjoy the school life

Misuse of conjunctions (42 tokens: 31% from L-I, 59.5% from U-I, 9.5% fromA) Many of these errors had correlative pairs attached to both clauses of acomplex sentence The influence of Chinese was evident: Chinesecomplex sentences are symmetrical and allow double conjunctions(correlative pairs) The Chinese equivalents of because and so, althoughand but are good examples of such correlative pairs (Chan, 2004a)

9 *Although we can’t have our own life there, but now we are happy

Duplicated comparatives or superlatives (47 tokens: 44.7% from L-I, 51.1%from U-I, 4.3% from A) Comparable constructions in Chinese-Cantonese may have affected the learners’ use of a redundant more ormost, because the corresponding comparisons in Chinese-Cantonese areformed by the addition of the words bei2 (than) or gang3 (more) and zeoi3(most) preceding the comparative adjectives and superlative adjectives,respectively (Li & Thompson, 1981; Matthews & Yip, 1994).Overgeneralization of the use of English more or most may also havebeen a probable cause, because the two words are used for makingcomparatives and superlatives for all polysyllabic English words andmany bisyllabic English words

5 All the Cantonese sentences used for comparison are grammatical Cantonese sentences acceptable to native speakers of Cantonese.

Chinese-10 *That is the most happiest time in [cf naa5 si6 ngo5 zoi6 haai4 tung4 si4zeoi3 faai3 lok6 dik1 si4 hau6 (that is I in childhood most happyNOMINALIZER

time)]

Misordering of constituents in indirect questions (47 tokens: 23.4% from L-I,70.2% from U-I, 6.4% from A) The incorrect placement of subjectfollowing operator mirrored the order of the two constituents in a directwh- question:

11 *I don’t know where is it

L1 interference was not evident, because in Chinese, wh- words occur inthe same position in a sentence as do non-question words having thesame grammatical function (Li & Thompson, 1981) No reordering ofsubject and operator is required in an indirect wh- question Inadequatemastery of the correct ordering of constituents in English indirectquestions was probably the main cause

In-prepositional phrases (49 tokens: 73.5% from L-I, 24.5% from U-I, 2%from A) These consisted of the preposition in, either used redundantly

or chosen inappropriately, in an in-prepositional phrase Such structurescould be traceable to equivalent Chinese prepositional phrases, whichoften require the presence of the word zoi6 (in)

12 *In many years ago, my father [cf zoi6 han2 do1 nin4 cin4 (in many years ago)]

Independent clauses as objects or subjects (54 tokens: 53.7% from L-I, 31.5%from U-I, 14.8% from A) L1 effects should have been at work, because it

is acceptable to have two or more verb phrases or clauses in the samesentence (i.e., serial verb constructions) and for the first verb phrase orclause to be the subject of the whole sentence in Chinese (Li &Thompson, 1981; Matthews & Yip, 1994) A lack of awareness that anindependent clause cannot be the object or subject of an Englishsentence may also have been a reason (Chan, Kwan, & Li, 2003)

13 *You don’t need to worry about the problem will struck at you [cf nei5 bat1seoi1 jiu3 daam1 sam1 man6 tai4 wui5 jing2 hoeng2 nei5 (you not needworry problem will affect you)]

Be + -ed (99 tokens: 29.3% from L-I, 50.5% from U-I, 20.2% from A) Inthese sentences, the verb to be coexists with the past participle (or pastform) of the main verb This error can be traceable to Chinese-Cantonese structures with si6 (is) serving as a marker of specialaffirmation (Li & Thompson, 1981, p 151), linking the two major

constituents of the sentences to mean ‘‘It is true that.’’ However, it isunclear whether the forms of the verbs were past forms or pastparticiples, because most of the words ended with -ed It is also unclearwhy the -ed form instead of the base form was used (see be + base form).

14 *She is always cried [cf taa1 si6 si4 seong4 huk1 dik1 (he IS always cry

PARTICLE)]

Pseudopassives and undergeneration of passives (110 tokens: 29.1% from

L-I, 53.6% from U-L-I, 17.3% from A) Many of these could be regarded aspseudopassives (cf Yip, 1995), which are ‘‘one reflection in theinterlanguage of the Chinese typological characteristic of topic-prominence’’ (p 97) Because the learners failed to generate the fullrange of English passive constructions, these errors could also be seen ascases of undergeneration of the target passive (Yip, 1995) Mother-tongue interference was apparent

15 *The floor can automatic clean [cf ze5 dei6 baan2 ho2 ji5 zi6 dung6 cing1git3 (the floor can automatic clean)]

Omission of subjects (114 tokens: 64% from L-I, 22.8% from U-I, 13.2%from A) This was often associated with compound or complex sentenceswhere both clauses shared the same subject, the subject was present inone of the clauses, the missing subject could be identified with thesubject present in the other clause, or the missing subject wasunderstood in the immediate context

16 *First, ‘ talk about the traffic [cf sau2 sin1 taam4 taam4 gaau1 tung1man6 tai4 (first talk talk traffic problems)]

Mother-tongue interference was observed, because a coreferential nounphrase in the second clause of a sentence or in subsequent sentences of

a discourse is not normally mentioned in Chinese A coreferentialpronoun may be used, but it is not obligatory (Li & Thompson, 1981).Subjects (and objects) may be omitted in Chinese when the constituentshave been the topics of previous utterances, or when the reference isclear from the context (Matthews & Yip, 1994)

Existential structures (118 tokens: 61% from L-I, 26.3% from U-I, 12.7%from A) Mother-tongue interference may have been at work, becausethe corresponding existential meaning in Chinese-Cantonese isexpressed using jau5 (have) Students’ inadequate mastery of thedifferent forms of the verb to be in English may also have been aprobable cause Given that the perfect forms have been and has been of the

verb to be are orthographically similar to the verb have, probableconfusion because of such acceptable structures as There may have beensome problems may also have led to the anomaly (Chan et al., 2002).

17 *There had many people at there [cf daan6 naa5 leoi5 jau5 han2 do1 jan4(but there have many people)]

Misuse of prepositions (126 tokens; 45.2% from L-I, 36.5% from U-I,18.3% from A) In these sentences, an inappropriate preposition waschosen in place of an appropriate one, a superfluous preposition wasadded, or a required preposition was omitted

18 *We played card games on the bus although it was crowded of people

19 *I could meet more new friends and play with them besides from mybrothers

No L1 interference could be identified Direct translations of sentencessuch as Examples 18 and 19 do not suggest the need for such inaccurateconstituents Because the uses of English prepositions are not easy togeneralize and choices of prepositions are often lexically determinedand idiosyncratic, inadequate mastery of the choice and use of Englishprepositions was probably the major cause

However, mother-tongue interference may have been at work for theproduction of sentences with missing prepositions:

20 *Which kind of examination system is appropriate‘ the situation now?[cf naa5 jat1 zung2 haau2 si5 zai3 dou6 si6 sik1 hap6 jin6 zoi6 dik1 cing4 fong3(which one kind exam system appropriate nowNOMINALIZERsituation)].Sentences such as Example 20, which had an adjectival complement(e.g., appropriate) used with a postmodifier lacking an appropriatepreposition (e.g., the situation), were actually direct Chinese translations,because the corresponding Chinese constituents for the adjectives areoften used as transitive verbs in Chinese (e.g., sik1 hap6 [appropriate]).Verb form selection (144 tokens: 36.8% from L-I, 47.2% from U-I, 16% fromA) In these sentences, an -ing participle was used in place of a present-tense verb, a past form in place of a base-form verb, and the like Verbs inChinese do not exhibit different verb forms, so insufficient mastery of verb-formation processes in English was probably a major cause

21 *Every day he driving his car

Misuse of relative clauses (158 tokens: 18.4% from L-I, 53.2% from U-I,28.5% from A) No direct L1 interference could be traceable for some

relative clause errors, because there are no relative pronouns in Chinese:Relative clauses in Chinese are formed by the nominalizer dik1 (e.g., oi3[love] ngo5 [me] dik1 [NOMINALIZER]jan4 [person] [the person who lovesme]) The erroneous English sentence errors (e.g., Example 22) and theChinese translations do not resemble each other.

22 *She will cook the food what I like to eat

Other errors identified were erroneous reduced relative clauses with amissing finite verb (e.g., Example 23) Inadequate mastery of therestrictions that a finite relative clause requires the co-occurrence of arelative pronoun and that a nonfinite relative clause precludes anexplicit relative pronoun was probably the major reason

23 *I have a large family which including grandmother,

The omission of relative pronouns, especially subject relative pronouns(e.g., Example 24), however, could be seen as resulting from L1 Directtranslations of the English sentences showed great resemblance toChinese relative clauses without the nominalizer dik1

24 * You are the first‘ come to Hong Kong [cf nei5 si6 dai6 jat1 go3 loi4heong1 gong2 (You are the firstCLASSIFIERcome Hong Kong)]

Incorrect order of adverbials or adverbs (172 tokens: 91.3% from L-I, 7%from U-I, 1.7% from A) Most of these errors were associated with theincorrect placement of the adverb very, though other adverbs, such asnever, were also sometimes misplaced

25 *I was very work hard to read [cf ngo5 han2 nou5 lik6 duk6 syu1 (I veryhard read book)]

In Chinese, han2 (very) is typically placed before verbs (e.g., han2 hei2fun1 [very like]) and predicative adjectives or adjectival verbs (e.g., han2jau5 jung6 [very useful]; Chan, Li, & Kwan, 2003; Matthews & Yip, 1994).Such resemblance between the syntactic behavior of Chinese verbs andadjectives, together with the acceptability of a similar very + ADJECTIVE

structure in English, such as very good, may have led the students to thinkthat the structure very +VERBwas acceptable in English

The acceptability of expressions such as I very much want to go may alsoexplain the error Overgeneralization resulting from their inadequateunderstanding of the differences in forms and functions between thedegree adverb very and adverbials such as very much and of the contextwhich allows fronted adverbials may also have been the cause