Title

Other Contributor(s)

Author(s)

Speech errors and the language processing in Cantonese

University of Hong Kong

Lee, Vin-yan, Vivian; 李穎欣

Citation

Issued Date

URL

Rights

2001

http://hdl.handle.net/10722/56413

Creative Commons: Attribution 3.0 Hong Kong License

Speech errors and the language processing in Cantonese

Lee Vin Yan, Vivian

A dissertation submitted in partial fulfillment of the requirements for the Bachelor of Science (Speech and Hearing Sciences), The University of Hong Kong, May 4,2001.

1

Speech errors and the language processing in Cantonese Lee Vin Yan, Vivian Abstract There is a sizable literature based on English speech errors (Fromkin,1971; Garrett, 1980; Stemburger,1985). In contrast, no reports about Cantonese speech errors can be found. This study examined Cantonese speech errors in a comprehensive way. References are made from Garrett's and Stemberger's language processing model for the interpretation of Cantonese speech errors. The Hong Kong Cantonese Adult Language Corpus (HKCAC) (Law & Leung, 2000) and an additional 285 minutes of recorded radio programs between December, 2000 and February,2001 were analyzed for finding different types of speech errors. In the Cantonese corpus, there were 168 errors found. They were word substitution, exchange, sound movement, shift, blending, insertion, deletion and tone errors. Among the Cantonese data, bound morphemes involved in different types of errors, which was not reported in English. Further research can be done to investigate the role of bound morpheme in Cantonese language processing.

2

Introduction During speaking, it is not uncommon to find speech errors in normal speech. According to Garrett, the definition of the speech error is as follow: "Speech errors we will consider are those from the spontaneous speech of normal speakers. Normal speech is, of course, not pristinely grammatical and fluent. Nonfluencies, misarticulations and various infelicities of expression are fairly common in normal conversational interchanges." (Garrett, 1990,p. 159) The speech errors indicate that there is a discrepancy between the speaker's knowledge about the form of the utterance and the actual production of the utterance. The reason about the occurrence of the errors and their implications on the language production mechanism are the questions arousing linguists' and psychologists' interests. There is a sizable literature based on English speech errors. Different speech production models derived from the observation and analyses of different types of speech errors are proposed(Fromkin,1971; Garrett, 1980; Stemburger,1985). Garrett (1980) reported different types of speech errors - word substitutions, word exchanges, sound movement errors and shift errors. (1) Word substitution-Td like to have mustache for dinner' (intended: mushroom) (2) Word exchange- 'Add the Est to the roof' (intended: roof to the list) (3) Sound movement error-'It is the golly green giant' (intended: jolly green giant) (4) Shift error-6You have to do learn that' (intended: you do have to learn that) Garrett (1988,1990) tried to divide the above errors into two main classes: one was word

3

substitution errors which involved the lexical selection processes, the other was movement errors which included word and sound exchange related to the phrasal construction and phonological coding respectively. Word substitution errors had two different sub-types: a) meaning related (e.g. 'mother5 for 'wife') and b) form related (e.g. 'mushroom5 for 'mustache'). And there was a dissociation between meaning and form relations. The intrusion words were either semantically related or phonologically related to the target words. There were rare cases to have both form and meaning relation between the target words and intrusion words. Based on the error patterns, the two-step lexical retrieval hypothesis was generated (Garrett, 1988). It meant that the meaning and form retrieval were separate processes (Garrett, 1988). Word retrieval was based on the direct mapping from meaning to words in the initial stage, if failure occurred, it would cause the semantically related word substitutions. If no error occurred in the meaning based retrieval process, the form based retrieval process would take place. The segmental structure is specified in this stage. Word forms were stored in places on the basis of phonological resemblance. If failure occurred in the form based retrieval process, it would yield form-related substitutions (GarretU990, p. 161). Garrett (1980) found that word exchange errors occurred between phrases rather than within phrases and had constraints of grammatical category. The interacting words need not be phonologically similar. In contrast, the sound movement errors involve the words of different syntactic categories but there were phonological similarities between the interacting sound

4

elements (Fromkin, 1971;Gairett, 1988,1990). The sound eirors also occurred within phrases and the distance between the interacting elements was closer together. Based on the differences between word exchange and sound movement errors, Garrett's model (1980,1988) proposed that there were two discrete levels-functional level and positional level for phrasal construction. At functional level, the syntactic relations among words were formed.

The errors occurred at

functional level caused word exchange errors. At positional level, the detailed phonological information of the words was constructed. The errors occurred at positional level caused sound movement errors. The grammatical encoding and phonological encoding took place respectively and independently. Garrett (1988) suggested that there was a relationship between two- step lexical retrieval in lexical selection and the two discrete levels (functional and positional) in phrasal construction. 'The meaning related word substitutions may be associated with selection of lexical content for functional structures, and form-based substitutions with the selection of lexical forms for positional level structures.' (Garrett, 1988, p.77) Shift errors involved closed class elements including words or inflectional affixes (Garrett,1990). Closed class elements rarely involved in the exchange errors. They just appeared in the shift errors. For example: (5)'You have to do learn that' (intended: you do have to learn that) (6) 'He go hacks' (intended: he goes back)

In example (5), the stress went with the moved word 'do' so it distorted the stress pattern of the original sentence. It was the major difference between exchange and shift errors. In example (6), the inflectional affix 's' moved from the word 'go' to the word 'back'.

Based on the

characteristic of the shift errors, it was hypothesized that closed class elements were not on the lexical retrieval but they belonged to the features of the framework of the sentence. The mislocations of the closed class element arose in the process of mapping features of the frames into the final sentence structure and caused shift errors (Garrett, 1990,p. 165). There are also other scholars proposing different hypothese about processing mechanism of language production.

Stemberger (1985) suggested an interactive model of language production

to account for the speech errors just discussed. According to the model, when the speaker produced an utterance, the activation was passed to the target unit and the other units which were similar to the target items. For example, if the target word was 'feather', other phonologicaUy or semantically similar words were also activated such as 'leather, hair'. Then when the word 'feather' got more activation energy than others, it inhibited other related words so that the word 'feather' could be selected and articulated. However, if the activation level of the related item was higher than the target item, the related item would be selected instead of the target and the word-substitution occurred. Stemberger (1985) suggested that the access of syntactic structure was similar to the access of words. The units of the syntactic level were phrase structure. Phrase structure includes NP

6

and VP. The syntax provided a set of syntactic slots to which words are attached. All lexical items in a clause could be accessed in parallel, with all items compete with each other for all slots. The syntactic structure of the phrase structure determined serial order directly. Execution of the sentence begins with passing the activation from the highest node (sentence) to the daughter node (NP and VP). NP got the activation at a faster rate than VP so that NP came first. Take the example (7) 'The little bird ate the seeds'. NP passed the activation faster to the determiner 'the' and then to the adjective 'little5 and than to the noun 'bird'. As the activation spread, all the words gradually became activated until reaching the threshold for execution in the intended order. The error arose when the adjacent units were produced in reverse order because the later unit reached the threshold accidentally before the unit in earlier position. The later unit was executed earlier and it was shift error. Stemberger(1985) argued that shift errors were errors of execution, not errors of access. Most shift errors involved the execution of the word one word early or late. The shift occurred within the phrase rather than the reversal of whole NP and VP as the activation between the NP and VP was great. The difference in activation between lexical items in the same phrase was small. Stemberger (1985) also suggested that there was interaction between levels of the language system: syntax and lexical access, phonological elements and lexical access. The word and syntactic level interact (Stemberger, 1985). For example: (8) 'Your teeth are all red' (intended: Your tongue is all red)

7

In (8), the word 'teeth5 substituted the word 'tongue5 and then a plural verb instead of a singular verb was selected, which agreed with the plural noun 'teeth5. The particular words selected to fill the subject NP caused changes in the surface syntactic structure. It was a piece of evidence that the lexical access affected the syntactic access. Stemberger(1985) also suggested that there was interaction between the access of lexical items and phonological information.

There were word substitutions in which the errors and the

target words were phonologically similar but not semantically similar. Example(9): coupons (target) -> croutons (error) According to Stemberger5s model, when the lexical item was activated, it passed the activation to its phonological units. And then the phonological units passed activation back to the lexical item as feedback, reinforcing the target word and spreading the secondary activation to other phonologically similar words. The phonologically similar word had higher activation energy than the target word and then selected. Word substitution with phonological similarity occurred. Up to now, most of the researches about speech errors are based on European languages. However, Cantonese is different from European languages in many aspects. Only a few studies about Mandarin Chinese speech errors can be found. They are mainly focused on tone and syllable structure (Shen, 1993, Yang,1997, Chen,1999 and Chen,2000X Based on the data found by Shen (1993), the tone of the syllable is represented and processed independently and the syllable branching structure in Mandarin Chinese is composed of initial, medial and final.

8

Example (10): the word 'tian' which means 'sky', the initial is 't\ the medial is T, the final is 'an'. Yang (1997) suggested that Garrett's language processing model could be used to describe the language production processes in Mandarin Chinese, supported by the error data found

He

further suggested that tonal errors indicated that there was a special phonological process for tonal specification, which was absent from non-tonal languages. Chen (1999) found 24 tone errors out of 987 errors in his corpus. Tone errors were much rare than the sound movement errors. He suggested that the lexical tones in Mandarin Chinese processed independently, detached from segmental sounds. Chen (2000) found that there were 10 pure syllable errors among 191 sound errors. Pure syllable errors meant that the target and the source syllable did not share the tone or any constituent in the syllable. Example (11)- 'qing1 zhuo2 du4'(means clarity) was replaced by the word 'qing1 du2 du4\

The source syllable 'du4' moved to the position of the syllable

'zhuo2 'without taking the tone and became 'du2'. The syllable 'zhuo2' and 'du4' had different tones and syllable constituents, so the error was classified as pure syllable error. In Chen's data (2000), the syllable movement errors provided evidence to support the claim that syllables were represented and processed as planning units in the form of a stored chunk. It was different from the role of syllables in English. In contrast, no reports about Cantonese speech errors can be found. This study examined the speech errors of Cantonese in a comprehensive way, focusing on different types of speech errors: word substitution, sound movement errors, shift errors and blending errors. Similarities

9

and differences between Cantonese and English errors were discussed in details.

References

were made from Garrett's and Stemberger's language processing model for the interpretation of the errors. Method Data The Hong Kong Cantonese Adult Language Corpus (HKCAC) (Law & Leung, 2000) and an additional 285 minutes of recorded radio programs between December,2000 and February,2001 were analyzed for finding different types of speech errors. HKCAC consisted of 487 minutes conversations in phone-in programs and forums on the radio. All the recorded conversations were spontaneous speech production. Spontaneous speech production was used because it reflects normal planning processes (Poulisse & Nanda,1999). Data Analysis The speech errors found were classified into 9 different types: word substitution, exchange, sound movement, tone, shift, blending, insertion, deletion and others based on the following definitions. Word substitution errors In this type of errors, the target word is replaced with another word in a planned utterance. Word substitution errors are divided into four sub-types a) form-related word substitution the target and the error are phonologically similar, b) meaning-related word substitution -the

10

target and the error are semantically similar, c) mixed word substitution -both semantically and phonologically similar, d) unrelated word substitution. Exchange errors Exchange errors mean two segments exchange in position. The units involved in the position exchange including syllables, bound-morphemes, words and phrases. The position exchange of the segments makes the target sentence ungrammatical. Sound movement errors This kind of errors involves units including onset, vowel, rime, coda and even the phonetic feature ,which are smaller than a syllable. The sound movement errors are divided into sub-types including anticipation, perseveration, substitution and exchange. Anticipation-one segment at a later position in the planned utterance replaces another segment at an earlier position of the same utterance. Perseveration-one segment in an earlier position replaces another segment in the later position of the same utterance. Exchange-two segments in the same utterance exchange positions. Substitution- the target segment is replaced with an error segment which is not found in the target sentence. Tone errors In Cantonese, there are nine contrastive tones. The assignment of a wrong tone to a

11

syllable causes the loss of the semantic content of a word is classified as tone errors. Shift errors This type of errors means the improper placement of a single item (which can be bound-morphemes or words) in a planned utterance. And this improper placement makes the sentence ungrammatical. Sometimes the speaker offered self-correction after the misplacement of the single item and it appeared that there was an insertion of a single item. However the difference between the shift errors and insertion errors is the added elements in shift errors can be found in the original sentences. Blending errors The errors are the combinations of two competing units, which can be syllables, bound morphemes, words and phrases. Insertion eirors The addition of a segment (which can be a bound-morpheme or a word) in the target sentence that make the original sentence ungrammatical. Deletion eirors The omission of a syllable of multisyllabic words which makes the sentence ungrammatical. Others The errors which cannot be classified into the above groups are grouped into 'others'.

12

Only descriptive statistics were used to describe each type of speech errors in detail. Results In the Cantonese corpus, there were 168 errors found. They were word substitution, exchange, sound movement, shift, blending, insertion, deletion and tone errors. The distribution of different types of errors is shown in Table L Table 1: The distribution of different types of errors in the corpus. Different types Word substitution

Percentage form-related

2/168(1.2%)

Meaning-related

21/168 (12.5%)

Mixed

14/168(8.3%)

others

3/168(1.8%)

Sub-total of word substitution

40/168(23.8%)

Exchange

11/168(6.5%)

Sound movement: anticipation

21/168(12.5%)

perseveration

16/168 (9.5%)

Exchange

1/168(0.6%)

Substitution

2/168(1.2%)

13

Sub-total of sound movement errors

40/168(23.8%)

Tone

4/168(2.4%)

Shift

45/168(26.8%)

[Blending

13/168(7.7%)

Insertion

8/168(4.8%)

Deletion of syllable

3/168(1.8%)

Others

4/168(2.4%)

Total

168

1

The details of each type of errors were reported as follow. Word substitution errors In English, there were only meaning and form related word substitutions. In Cantonese, there were four different types of word substitutions: the word-form substitution (5%), the word meaning substitution (52.5%), the word substitution (mixed) (35%) and others (7.5%). The word meaning substitution was the largest sub-type and the mixed word substitution was the second largest. All the word substitution errors were of the same word classes with the targets. Word-form substitution (12)

Actual m Intend

m

A

B

mou>J tV^ fenl minJ 'no' 'pain' 'delivery' M M ft ^ moii>l tV^ fun! suiil 'no* 'pain' 'separation' 'no pain separation is quite good9

m

u

kerl 'quite' & keil

houl 'good' #? hoiil

14

Word meaning substitution (13)

Actual^ m m

M

t s i ^ fill koH 'government' 'cl' IntendiSc Jft m tsiiH fill o^

i

S

»

J

*

«

^

8

i

h

ts oU tsiijH hail tsil feil soerj>ltsil w m 1 kin-l 'financial' 'expenditure' 'very' 'stable' M & & A # 1 R Z n m h ts oU tsirjH SBU! up-l feil oerjvJ tsil wm1 kin-l 'income' 'Government has stable financial income'

'cl' represents classifier in all the examples. Mixed substitution (14) Actual I I kernl go/1 T Intend € a krail goJ

i IP teml tseM 'think' H SP ICTQI tseM

1 purl 'sell' ^ pinH

1 mai4 K mai-l

m M kal tshan1 'home asset' * M tsil ts h an1 'property'

'I think selling the property' Others (15) Actual

Intend

B

#

m

m

t^il 'watch'

pmrl 'half

tsoijl fcH 'classifier' 'hour'

#

#

m

ft

punH

ko4

tsoijl

'wait' 'wait for half an hour' There were three errors classified as 'others' as they were neither phonologically nor semantically with the target words. However all the three errors were about verbs and the only similarity between the target and the error was they were of the same grammatical categories.

15

Exchange errors In the Cantonese data, the units involved in the exchange errors are only of adjacent positions (91%). Exchange errors occur within phrases (45%) and between phrases (55%). Between phrases (16) Actual

Intend

f!l Lil

m

%

tH

m.

—•

32

ei-J

& A kogl kog-l

gal

kogl fok-l

'cl'

'is'

'public'

'one'

'kind' 'service'

'this' UJ§ Lil

m

%

— a

&

£

tH

ei-l

gal

kogl

kog-l fok-l

'This

'is' is one kind

kogl

'one' 'kind' of pubhc

m

m mou-l

85 mou-J

'public' service'

Within phrases (17) Actual

m £ h kwcH jyH

Hi KoH

kerl MEN

'past'

'cl'

'few5 'year'

Intend M

£

W

¥ m

h

koH

linvi

kwoH jyH

keil

'year' 'few' 'The

past

one year

The exchange errors in Cantonese were of different grammatical categories, unlike those in English. The units involved can be of different levels like syllable-syllable (1/11), bound-morpheme-bound-morpheme (1/11), word-morpheme (2/11), word-word (3/11), word-phrase (4/11). Syllable-syllable (18)

Actual^

m m m

mm

mm

Mm

kap-l tsap-l m\l tamvl M-J ham/I keHkwoij'l tog! wam-l 'mixed with' 'not' 'tasteless' 'not' 'salty' 'Cantonese'

16

Intend ^ i ^ i Pg $. 1^ H I i / kap4 tsap-l m\l ham^ M4 tam/l ke^kworJl torjl wam4 'mixedwith' 'nof'salty' 'nof'tastelss' 'Cantonese' 'Mixed with not pure Cantonese' Bound-morpheme-bound-morpheme (19) Actual

fi *H jam-l tou-l 'important' ii M jam-l tsogJ 5 'important 'The important job'

Intend

Jl tsog-l

7&

m

X

fP

jynJ

keH

korjl

tsobl

'job'

M touJ

jyn/i

m

X

fF

ke^

korjl

tsobl

'job'

Word-word (20)

Actual m h£

«i

m

kwoH jy4 koH 'past' 'classifier' Intend i i £ Iff kw^ *jyH ko4 'The Past

¥

keil lin>l 'few' 'year' ¥ H linvl keil 'year' 'few'

one year

Word-morpheme (21) Actual

Intend

m jml 'because' S jtml

J§ WBi-l Ji WBi-l

fft leiJ 'you' ffr leU

^ "S hBp-l m>l 'legal' 'not' ng -£ m>J hBp-l 'not' 'legal'

ffi fa?4 Sc fa?H

'because you are not legal' Word-phrase (22)

Actual m m & & Ml

tH

Bi-I

koijl

*

-

koijJ g a l

n

m

kogl fok-l

m mou-J

17

Intend

'this' US lil

6

is' fl ei-l

Oft tH

'This is one kind

'public' 'one' ® & gal kogl kogl 'one' of

'kind' 'service' * ffi » kog-l fokJ mouJ

'kind' 'public' public service'

Sound movement errors Cantonese sound movement errors were similar with English data. There were strong phonological similarities among the interacting elements and their environments. Among the sound movement errors in Cantonese, the largest subtype was anticipation (52.5%). The second largest was perseveration(40%). The average distance between the error and the source sound in anticipation, perserveration were 0.9 intervening syllable, 0.75 intervening syllable respectively. The sound movement errors occurred within word (61%) and between words (39%). Anticipation (23) Actual

Intend

„

m

II

*

jtfi 'one'

t^M 'watch'

forjl 1*0jrt 'depressed wind'

—

m

M

m

JBtl

SBM

&0y4

forjl

'depressed wind' 'wash' 'one' 'has been cheered up' Perseveration (24) Actual

Intend

*

M h

6ai-l

tS 3N

'big'

'corporation'

*

ftf h

6ai4

tS 3W

'big' 'a big

'corporation' corporation'

H tshynN|

EEEI

t^yiH

18

Exchange (25) Actual

£5?

'find' 'root' H ts^mvl kBiil 'find' 'root' 'find out the truth'

Intend

keirl 'study'

teil 'bottom'

7h

kmri 'study'

teil 'bottom'

Substitution (26) Actual fg £F f! a Tff h k 0y^ ^ o u 1 heil Mo-l tekl t o ^ si/I 'he' 'very' 'hope' 'citizen' Intend {E £F # ?l h k 0y/l houl heil Mo-J tekl toiH si/I 'he' 'very' 'hope' 'citizen' 'he hope to win citizen's acceptance'

nren^i kol

'heart' men^ ko-l

Within words (27) Actual

M tSBU-l

Intend

wei-l 'make'

hyn>J 'circle'

hynl

WBU 'make' 'make a

yiH 'circle' circle'

hynl

M tSBU-J

Between words (28) Actual H kai^

m Kal

*- m m m $

jig-l touH hoerjl 'make' 'Hong Intend I JD ^ S f kaij4 kal lig-J touH hoegl 'make' 'Hong 'make Hong Kong become various'

kogl Kong' i kogl Kong'

srail

7U jynt>i

tol 'various' ^ 7C tol oeynvl 'various'

it faH ft fa4

svm\ 'heart'

19

However, sound movement eirors also involved units smaller than phoneme such as features. There were two errors involved the feature. Between the errors, the manner of the consonant was anticipated or perseverated but the aspiration feature was kept Feature (29) Actual

Intend

& ± Mi fe h kBml ts uiiNl tsigH tsi-J 'gold' *c 'ftojietf'po sfe H Wc tp h keml K unvl tsiijH tsi-l 'gold' 'power9 'politics' 'politics controlled by financial power'

Tone errors It has been found that tone can be shifted, anticipated and perseverated independently like other phonemic features (e.g. aspiration) during production (Yang, 1997). Tonal errors found in the Cantonese data were similar to previous studies. There are 4 tonal errors among 168 speech errors found in my data. Example is as follow: (30) Actual

Intend

H lun-J

& 0E i! JBm-l mmN lun-J

^ &§ W B M jcwj>l

'whatever' ^ H pBl/l lun-l

'conspiracy' & IK H JBml mmN lun4

'or'

7 pst/1

'whatever' 'conspiracy' 'whatever it is a conspiracy or not'

'not conspiracy'

wek-4 joer^ 'or'

fflt it mmN lun-J

mmi

lunJ

'not conspiracy'

Shift errors Unlike English, not only closed class words but also noun, verb, adjective and adverb were

20

involved in the shift errors in Cantonese. Moreover, the units involved can be monosyllabic words (51.1%), disyllabic words (ll.l%)and even units smaller than word, which is bound morpheme (37.8%). And the average distances between the target position and the error position was 2.8 intervening syllables respectively. Monosyllabic word (31)

Actual ffr m m 9t M % jH

lil

1&H sin! tsH hBi-l oeetl koi tail ts0y tai-l koH

'you' 'this' 'cl'

'is'

Intend I M I i jH

lil

m * m ^c «E

'one' 'cl'

'you' 'this' 'cl' 'is' 'this is the greatest problem'

'one' 'cl'

mim t^eM

'great' 'the greatest' problem'

^ 1 1! - I I

ko^ sin! tsiH hei4 oeetl k ^

mm

^1

H J5

tsoy tai-JkoH

man

'the greatest'

t ^

'problem'

Disyllabic word (32)

Actual m

m

8?

lil joei)4 aH 'this' Intend HJB ft B? lil

joeij-l

m siijvl 'city

w m

m m

m

w

«

sD

tin-l nouJ

SIIJNI

si/I

hrn-J jixrl haitf kfy-l

'computer'

'city'

tseM

5

tseM

aH

tin-l nouJ sig^I 'computer

'this' 'Computer City need to consider this'

5

'city

'need'

m

X

J&

tseM

faiH

mvl siiH 'mature' 'not'

^ SUJN!

Wi sok-l

i

'most' Intend

m

IP

X

m

tseM

farl 'most'

m4

'It is not mature'

'not'

'mature'

J$

f& sok-J

'mature'

#

A

'consider'

&U tei-l jiuH haul Lay-J 5

Bound-morpheme (33) Actual

s

'need'

'consider'

21

Blending errors Similar to the observations in English, blending occurs mostly at the levels of words and phrases (8/13,4/13 respectively). Unlike English, there was one error involved the blending of phonemes from two syllables. The examples are as follow: Word level (34)

Actual m n m m& m WL t? g i turjl soerHlil

til weil lei 9

'normally' 'this Intend

M S m tug! JcerH lil

£ m

tssu-l hou1ji-l jmJi 5

'position'

'easy

gaU kin-)

'have' 'quarrel'

m ft «5B WL if % W til weil lei tsBU-J houl ji-l JBUJ

'normally' 'this'

'position'

'easy'

'argument'

& X & H gall kaul gaiH kiu-i

'have' 'quarrel'

'argument'

'This point is easy to have argument or quarrel' Phrase level (35)

Actual

1£ h

k 0yJ

'he' fg

Intend

h

k 0y/l

nr

.

JBUyl

JBtl

ffl> tOUl 5

m

W

m

lam>l

JBU/I

£

p\n>\

JBIlJ

'has'

'one

'also have'

'boyfriend'

M

—*

m

IP

m

m

m

JBU/I

JBtl

k»H

toul

JBU/I

lam>J

p1^^

£ JBU/I

'boyfriend'

'also have' 'one' 'he' 'has' 'He has one or also has boyfriend' Syllable-Syllable (36) Actual cf5 tsogl

# m

M

%

koH

Jyl

tsenl

hei-l leri kik-l kigl lik-l tsenl hei-l houl to!

'cl'

'vacation'

'is'

'cl'

£

ffi

m

WL M

®

ff

g-l

koH

Jyl

kaH ko1

koH kfajl lik-l tsral hra4 houi tol

. m

5 g/l

'Form 5' Intend ^ tsogl

krl

m ill

M « ft \

'experience' ffi

K

'Form 5' 'cl' 'vacation' 'cl' 'experience' "There were lots of experience in Form 5's vacation'

# 'is'

'is' 'much' ft

#

'much'

22

Insertion errors Most of the units involved in insertion errors were monosyllabic words (5/8). There was one error for bound-morpheme, disyllabic word and nonsense syllables each. Example (monosyllabic word): (37)

Actual mm

ft

&

rjO/l MN mou/l T 'not' 'not' Intend a ft rpJ mou/l T 'not' 'I do not pay attention'

m

u

timl jcegl 'pay attention' 15 W. timl joerjl 'pay attention'

M

weW jtf ft wei