Chapter 3 Methodology 3.1 Method and Subjects

Chapter 3
Methodology
The following section will introduce the method adopted in this study, including
ways of collection, subjects and classification of speech errors. The relative examples
will also be provided.
3.1 Method and Subjects
The method adopted in this study was to collect speech errors of normal speakers
in natural settings for analysis. The data contained 421 single-syllable paradigmatic
lexical substitutions in Mandarin Chinese. 244 were collected by the author from
2005 to 2008 and 177 were adopted from Prof. Wan’s speech-error corpus between
2005 and 2008 (who granted me access to the corpus of normal slips of the tongue).
These errors were all collected under natural conversational settings, which were not
prepared speeches read from manuscripts. The excerpts were mainly taken from two
sources. First, the researcher recorded radio programs and live lectures, typed the
transcription word-by-word, picked up speech errors by multiple-checking and finally
transcribed them in IPA symbols. The same way of collection was contributed to Prof.
Wan’s speech error corpus. The second major source was by ‘pen and paper.’ The
researcher often obtained speech errors from daily conversations. As soon as detecting
any slips of the tongue in chatting or free talks, the researcher will write it down
quickly in a notebook. In addition to the error itself, its related information will also
be recorded, such as the complete utterance, self-correction and other relevant context.
The researcher will then ask for the subjects’ permission to use their data.
As for the subjects, more than 30 speakers contributed to the data. Their ages
ranged from 20 to 50 years old, including both males and females. They may have
37
38
Mandarin or Taiwanese as their first language, but there were no major differences
between them because any errors doubted as bilingual errors will be excluded from
the corpus.
3.2 Classification System
The speech error type collected in this study was limited in single-syllable
non-contextual lexical errors, i.e. paradigmatic errors. Such limitation has two reasons.
First, those pure phonological errors were not included because the study aims to
investigate the phonological similarities in lexical errors. Secondly, errors involving
more than a single syllable, such as compounds, were also excluded. Because the
processing of compounds turns to be another issue, this study only focuses on the
activation of single words.
Paradigmatic errors happened when the speaker unintentionally retrieves another
word, which was not present in the neighboring context, to replace his intended
utterance. One important clue to identify the speaker’s intended utterance was through
self-correction. Speakers usually correct themselves after uttering the wrong lexicon
and later produce the intended one.
(10) ti51-i55 tÛj'n51 ‡Ó51 tÛjow51-‡Ó51 w‹n51 ni21...tÛj#w51 ni21 tÛ*y51
first CL
thing just
enquire you ask
you go
‘The first thing is to enquire you…..ask you to go there.’
Take (10) for example, the speaker uttered [w‹n51] ‘enquire’ first, but corrected
himself as [tÛj#w51] ‘ask’ immediately. It implied that what the speaker intended to
say was [tÛj#w51] ‘ask’ rather than [w‹n51] ‘enquire.’ Through self-correction,
speakers’ intended utterance could be recognized.
Next, how paradigmatic errors were classified in previous studies will be reviewed
in 3.2.1. Based on them, the classification system formulated in this study will be
39
provided through 3.2.2 to 3.2.4. Note that in the following examples, ‘I’ means the
speaker’s intended utterance and the second line is its gloss in English. ‘ĺ
means
the intended expression is replaced by ‘E’—the erroneous utterance. Both the
erroneous and intended utterance are emphasized in bold and underlined. The final
sentence is the smooth translation in English.
3.2.1 Classification in Previous Studies
The previous studies about speech errors set some standards to define the
semantic and phonological relationship between lexical errors. For the semantic
connection, they may be antonyms (boy-girl, left-right), coordinate relationship
(summer-winter, tongue-thumb) or super-ordinate (cup-dishes) (Dell & Reich, 1981).
For the phonological similarities, it could be evaluated from three aspects and their
relative examples were given in (11)-(13).
The first criterion was the initial segment because initial consonants were often
preserved in lexical substitutions (e.g., Boomer & Laver, 1968; Nooteboom, 1969;
Mackay, 1970; Fromkin, 1971; Harley, 1984).
(11) prepare ĺ prepay; combination ĺ contamination
(from Nooteboom, 1969)
In (11) the initial consonant [p] and [k] were remained in the target respectively.
Moreover, Dell and Reich (1981) investigated numbers of identical phonemes
between 289 lexical substitutions and found nearly 40% errors involving the identical
segment in the first position.
The second criterion was to see the similarity of phonetic features between a pair
of errors (Fromkin, 1971). Fay and Cutler (1977) concluded that in 156 malapropisms,
nearly 25% of them differed in only one feature.
(12) pedestrian ĺ tebestrian ([p] ĺ [t] and [d] ĺ [b] )
(from Fromkin, 1971)
40
For example, in (12) the value of the feature [coronal]—the place of
articulation—was changed between interacting units [p, t] and [d, b].
The third criterion was the number of syllables with identical stress pattern.
Harley (1984) claimed in 30 paradigmatic errors, 21 of them contained the same
syllable pattern.
(13) I: You eat the kippers?
ĺ E: You eat the catters?
(from Harley, 1984)
Just as example (13), the interacting units kippers and catters contained the same
syllable structure.
In addition to initials, phonetic features and syllable patterns, the current study
added more criteria to evaluate phonological similarities in paradigmatic errors, which
will be introduced in following sections.
3.2.2 Classification in Semantically-Relatedness
The interacting units of paradigmatic errors may be similar in semantics,
phonology or both. This section intends to set criteria to evaluate the relatedness in
semantics.
First of all, semantically-related paradigmatic errors could be classified as
antonyms, hyponyms, near-synonyms, coordinates and associates, based on the
semantic relationship between the target and intended utterance. In detail, antonyms
are words with opposite meanings, mostly adjectives and verbs. Hyponyms are those
of the same semantic group, such as classifiers, animals and physical actions.
Near-synonyms, including nouns, adjectives and verbs, have similar meanings but are
considered lexical substitutions due to speakers’ self-correction. Coordinates are
words of a limited set of vocabulary, such as numbers, colors, etc. Associates are
errors in which the target and intended utterance both related to a certain concept in
the sentence. Examples of each lexical relation are provided in (14)-(23).
41
(14) Paradigmatic lexical errors: (semantically-related: antonyms/adjectives)
I: tsaj51 to055-tÛi055 t‡¢51-t‡o021 ta51 t‡*‹035-‡Ó51ĺ
in Tokyo
this
kind big city
E: tsaj51 to055-tÛi055 t‡¢51-t‡o021 Ûj#w21 t‡*‹035-‡Ó51
in Tokyo
this
kind small
city
‘in a big city like Tokyo’ ĺ‘in a small city like Tokyo’
(15) Paradigmatic lexical errors: (semantically-related: antonyms/verbs)
I: ni21 tÛjow51 pi51 k*aj55 na51-k( ‡Ó35-tÛj'n55 ĺ
you just
avoid open
that-CL time
E: ni21 tÛjow51 t*j#w55 k*aj55...pi51 k*aj55 na51-k( ‡Ó35-tÛj'n55
you just
pick
open avoid open that-CL time
‘You could avoid that time.’ ĺ ‘You could pick up that time.’
Interacting utterances in examples (14)-(15) are both antonyms. In (14) two adjectives
[ta51] ‘big’ and [Ûj#w21] ‘small’ are converse in describing shapes. Likewise, in (15)
two verbs [pi51] ‘to avoid’ and [t*j#w55] ‘to pick up’ are contrastive actions.
(16) Paradigmatic lexical errors: (semantically-related: hyponyms/classifiers)
I: na51-‡Ó51 san55-‡Ó35 nj'n35 tÛ*j'n35 t( w#021-‡Ó51ĺ
that is thirty
year before GEN past thing
E: na51-‡Ó51 san55-‡Ó35 swej51…san55-‡Ó35 nj'n35 tÛ*j'n35 t( w#021-‡Ó51
that is thirty
years-old thirty
year
before GEN past
thing
‘Those things happened before thirty years.’ĺ ‘Those things happened when you
were thirty years old.’
(17) Paradigmatic lexical errors: (semantically-related: hyponyms/ nouns)
I: t*a t(
m#w55 Æ E: t*a t(
kow21
she GEN cat
she GEN dog
‘her cat’ Æ ‘her dog’
(18) Paradigmatic lexical errors: (semantically-related: hyponyms/ verbs)
I: swn35-i35 ni21 j#w51 Ûj'21 t‡(51 k( Ûin55-t( n?ĺ
so
you want write this-CL heart-get
E: swn35-i35 ni21 j#w51 k*an51…Ûj'21 t‡(51 k( Ûin55-t( n?
so
you want read
write this-CL heart-get
‘So do you have to write down what you’ve learned?’ ĺ‘So do you have to read what
you’ve learned?’
42
Examples (16)-(18) illustrate hyponyms in which the interacting units are members of
the same semantic group. In (16) [nj'n35] ‘year’ and [swej51] ‘years old’ are both
classifiers while in (17) [m#w55] ‘cat’ and [kow21] ‘dog’ are both animals. Besides,
in (18) both verbs [Ûj'21] ‘write’ and [k*an51] ‘read’ are physical actions.
(19) Paradigmatic lexical errors: (semantically-related: near-synonyms/nouns)
I: t*a55 t( Ûj'n51 tÛ*i35-‡Ó35 pu51 ›o035-i51 twan51 ĺ
it GEN thread actually
NEG easy
cut
E: t*a55 t( m#w35... t*a55 t( Ûj'n51 tÛ*i35-‡Ó35 pu51 ›o035-i51 twan51
it GEN hair
it
GEN thread actually
NEG easy
cut
‘Actually the thread is not easy to be cut off.’ ĺ ‘Actually the hair is not easy to be
cut off.’
(20) Paradigmatic lexical errors: (semantically-related: near-synonyms/adjectives)
I: p‹n21-laj35 tÛjow51 x‹n21 kwej51 ĺ
originally just
very expensive
E: p‹n21-laj35 tÛjow51 x‹n35 x#w21
originally just
very good
‘(It) is indeed always expensive.’ ĺ‘(It) is indeed always good.’
(21) Paradigmatic lexical errors: (semantically-related: near-synonyms/verbs)
I: mej21 k( ›‹n35 tow55-jow21 tÛj#w55 p*‹035-jow21 t(
tÛ*Ä'n35-li51 ĺ
every CL person both have
make
friend
GEN right
E: mej21 k( ›‹n35 tow55-jow21 tswn51 p*‹035-jow21 t(
tÛ*Ä'n35-li51
every CL person both have be
friend
GEN right
‘Everyone has the right to make friends.’ ĺ‘Everyone has the right to be others’
friends.’
Examples (19)-(21) describe three types of near-synonyms—nouns, adjectives and
verbs. They will not be detected unless the speaker corrected himself. In example (19)
two nouns [Ûj'n51] ‘thread’ and [m#w35] ‘hair’ are similar in meanings while in (20)
two adjectives [kwej51] ‘expensive’ and [x#w21] ‘good’ and two verbs [tÛj#w55]
‘make’ and [tswn51] ‘to be’ in (21) both are.
43
(22) Paradigmatic lexical errors: (semantically-related: coordinate/numbers)
I: t*a55 jow21 tÛ*i55 k( ‡Ó51-tÛj'51 i35-t‡*an21 ĺ
it has seven CL world
heritage
E: t*a55 jow21 sÓ51…jow21 tÛ*i55 k( ‡Ó51-tÛj'51 i35-t‡*an21
it has
four has seven CL world
heritage
‘It has seven world heritages.’ ĺ ‘It has four world heritages.’
In (22) [tÛ*i] ‘seven’ and [sÓ51] ‘four’ are coordinates of numbers. Besides,
coordinates have some other sub-categories such as colors ([xo035] ‘red’ replacing
[lan35] ‘blue’), demonstratives ([t‡(51] ‘this’ replacing [na51] ‘that’) and personal
pronouns ([t*a55] ‘he’ replacing [wn21] ‘I’).
(23) Paradigmatic lexical errors: (semantically-related: associate/ verbs)
I: t*(51-t‡‹n55-tsu21
j#w51 tÛj'51-›u51 t‡wa55 ts*aj51- t‡*o035ĺ
Special-investigative-unit want interfere arrest vegetable worm
E: t*(51-t‡‹n55-tsu21
j#w51 tÛj'51-›u51 ta21...t‡wa55 ts*aj51- t‡*o035
Special-investigative-unit want interfere
hit arrest vegetable worm
‘The Special Investigative Unit tried to arrest those illegal vegetable dealers’ ĺ
‘The Special Investigative Unit tried to hit…arrest illegal vegetable dealers’
In (23) the interacting words [t‡wa55] ‘to arrest’ and [ta21] ‘to hit’ are not directly
related in meanings. However, both actions regard the other concept ‘illegal vegetable
dealers’ because people of Special Investigative Unit are able to hit or arrest those
illegal men. It shows how lexical errors exhibit the associative relation.
According to these five criteria (antonyms, hyponyms, near-synonyms,
coordinates and associates), the composition of 421 paradigmatic errors was shown in
(24). Notice there are 62 errors are not related in any semantic sense.
(24) Classification of Semantically-related Paradigmatic Errors: (N=421)
1. Semantically-related: 359
a. Antonyms:
41 (Adjectives: 24; Verbs: 17)
b. Hyponyms:
116 (Classifiers: 35; Nouns: 44; Verbs: 37)
c. Near-synonyms: 88 (Nouns: 15; Adjectives: 23; Verbs: 50)
d. Coordinates: 38 (Numbers: 20; Colors: 8; Demonstratives: 7; Pronouns: 3)
e. Associates: 76 (Nouns: 18; Adjectives: 8; Verbs: 39; Others: 11)
2. Unrelated:
62
44
3.2.3 Classification in Phonologically-Relatedness
This section intends to define the phonological similarities in paradigmatic errors.
In addition to initial consonants, phonetic features and syllable structures mentioned
in 3.2.1, words containing the same rhyme and tone were included for evaluation. The
relevant examples were exhibited in (25)-(30).
(25) Paradigmatic lexical errors: (phonologically-related: initial)
I: Ûj'21 l(
x‹n21 twn55 p‹n21 kwan55-y35 t‡*a35 t( ‡u55 ĺ
write PERF very many CL about
tea
GEN book
E: ÛÄ'35 l( x‹n21 twn55 p‹n21 kwan55-y35 t‡*a35 t( ‡u55
learn PERF very many CL about
tea
GEN book
‘(He) wrote many books about tea.’ ĺ ‘(He) learned many books about tea.’
(26) Paradigmatic lexical errors: (phonologically-related: feature/consonant)
I: jow21 t( nan35-‡‹055 tÛjow51-xwej51 ts*o035 ‡#051 t#w51 Ûja51 ĺ
have LINK male born just
will
from up
to
down
E: jow21 t(
nan35-‡‹055 tÛjow51-xwej51 ts*o035 lj#051... ‡#051 t#w51 Ûja51
have LINK male born just
will
from calculate up
to
down
‘Some boys will judge you from top to toe.’ ĺ ‘Some boys will judge you from
calculating….’
(27) Paradigmatic lexical errors: (phonologically-related: feature/ vowel)
I: k*(21-‡Ó51 ni21 k*an51 t#w51 t( tÛjow51-‡Ó51 ‡wn55 ĺ
but
you see
get LINK just
say
E: k*(21-‡Ó51 ni21 f‹n55…k*an51 t#w51 t( tÛjow51 ‡Ó51 ‡wn55
but
you separate see
get LINK just
say
‘But what you saw was…’ĺ ‘But what you got was…’
(28) Paradigmatic lexical errors: (phonologically-related: rhyme)
I: j#w51 tÛja55 jow35 ĺ
want fill oil
E: j#w51 tÛja55 tÛjow21
want fill wine
‘(It) needs to fuel up.’ ĺ ‘(It) needs to be filled with wine.’
45
(29) Paradigmatic lexical errors: (phonologically-related: syllable structure)
I: wn21 j#w51 p#055 wn21 t( t*o035-‡Ó51 ta21-tÛ*i51 ĺ
I
want help I
GEN together thing hit air
t*o035-‡Ó51 ta21-tÛ*i51
E: wn21 j#w51 k‹n55…p#055 wn21 t(
I
want with help I
GEN together thing hit air
‘I would like to cheer for my colleges.’ ĺ‘I would like to cheer with my colleges.’
(30) Paradigmatic lexical errors: (phonologically-related: tone)
I: pu35-j#w51 t‡u21 t*aj51 tÛjow21 tÛjow51 pu35 xwej51 t*aj51 la51 ĺ
NEG-want boil too
long just
NEG-will
too spicy
E: pu35-j#w51 t‡u21 t*aj51 tÛjow21 tÛjow51 pu35 jo051 t*aj51 la51
NEG-want boil too long just
NEG-use
too spicy
‘Don’t cook it too long so it won’t be too spicy.’ ĺ‘..it doesn’t have to be spicy.’
Examples (25)-(30) introduce how lexical substitutions related in phonology. In (25)
two verbs [Ûj'21] ‘write’ and [ÛÄ'35] ‘learn’ share the same initial consonant [Û].
Examples (26)-(27) show the similarity of phonetic features in consonants and vowels
respectively. In (26) the different initial consonants [‡] and [l] share two features—
[!nasal] and ["continuant] while in (27) two vowels [‹] and [a] only differ in
[±low]. Besides, the interacting words [jow35] ‘oil’ and [tÛjow21] ‘wine’ in (28)
signify the identical rhyme—[jow]. And in (29), the intended unit [p#055] ‘help’ and
the target [k‹n55] ‘with’ share the same syllable structure CVC. Finally, in (30) the
pair of errors [xwej51] ‘will’ and [jo051] ‘use’ both contain the falling tone (51).
Initials, similarity of phonetic features, rhymes, syllable structures and tones are
taken as criteria to evaluate the phonological similarities in paradigmatic errors. As
(31) displayed, in 421 paradigmatic lexical errors, 318 were phonologically-related in
at least one of above aspects and 103 errors were not related in any of it.
46
(31) Classification of Phonologically-related Paradigmatic Errors: (N=421)
1. Phonologically-related: 318
a. Initial:
(same: 81; different: 237)
b. Phonetic Features/consonants: (share 5 features: 82; 4 features: 59;
3 features: 69; 2 features: 62; 1 feature: 28; 0 feature: 18)
c. Phonetic Features/vowels:
(share 5 features: 101; 4 features: 66;
3 features: 68; 2 features: 72; 1 feature: 11)
d. Rhymes:
(same: 54; different: 264)
e. Syllable Structures: (same: 94; different: 224)
f. Tones:
2. Unrelated:
(same: 240; different: 78)
103
3.2.4 Examples of Semantically- and Phonologically-Related Errors
Based on the classification mentioned above, errors which are related in both
semantics and phonology are called mixed errors. More examples are given below.
(32) Paradigmatic lexical errors: (mixed error: associative/initial)
twej51-pu35-twej51 ĺ
I: wn21 t‡wan51 l(
I
earn
PERF true- NEG- true
E: wn35 t‡#021 l(…. wn21 t‡wan51 l(
twej51-pu35-twej51
I
rise PERF I
earn
PERF true- NEG- true
‘(My stocks) have earned money, hasn’t it? ĺ‘…have risen…earned money…’
(33) Paradigmatic lexical errors: (mixed error: hyponym/rhyme)
I: ts‹n21-mn pa35 wn21 tswn51-t‡*‹035 kow21 t( wan35-tÛy51 ĺ
how
BA I
make as
dog GEN toy
E: ts‹n21-mn pa35 wn21 tswn51-t‡*‹035 xow35 t(
wan35-tÛy51
how
BA I
make as
monkey GEN toy
‘Why do you make me as the toy of dog?’ ĺ#‘..make me as the toy of monkey’
(34) Paradigmatic lexical errors: (mixed error: near-synonym/tone)
I: pa21 pj'n51t#055 so051 laj35 ĺ
BA convenient when send come
E: pa21 pj'n51- t#055 tÛi51…so051 laj35
BA convenient when mail send come
‘Send me the lunch box.’ ĺ‘Mail me the lunch box.’
47
Examples (32)-(34) are mixed errors. In (32), the interacting units [t‡wan51] ‘make’
and [t‡#021] ‘rise’ both associate to the concept—stocks—in the sentence and both
begin with the identical initial consonant [t‡]. Example (33) presents the rhyme effect
in that [kow21] ‘dog’ and [xow35] ‘monkey’ are hyponyms and have the same rhyme
[ow]. Finally, tone effect is exhibited in example (34) since the intended utterance
[so051] ‘send’ and the target [tÛi51] ‘mail’ contain the falling tone (51) and both are
actions about delivering.
3.3 Principles for Ambiguous Cases
As previous sections suggested, the paradigmatic errors in the corpus are related
in semantics, phonology or both. As for some vague cases, two principles are applied
for classification. First, lexical substitutions may be confused with phonological errors
which tend to be real words, i.e. lexical bias (Baars, Motley & Mackay, 1975). The
principle to distinguish them is that the interacting units of lexical errors often have
the same syntactic category and are semantically related while phonological errors are
not. The examples are given in (35)-(36).
(35) Unclear case: (different syntactic category)
I:kow35 k*(21-n‹035 ÛÄ'35 t#w51 t‡u21-›‹n35 ‡wn55-xwa51 t( t*(51-s(51ĺ
dog
probably
learn get master person speak
GEN feature
E: kow35 k*(21-n‹035 ÛÄ'35 tj#w51 t‡u21-›‹n35 ‡wn55-xwa51 t( t*(51-s(51
dog
probably
learn lost master person speak
GEN feature
‘The dog may learn the feature of its master’s accent.’ ĺ unacceptable
(36) Unclear case: (not semantically related)
I: k#055-ts*aj35 swan51,
just
count
wn21 ts*wn51-l¢
x#w21-twn55 ĺ
I
wrong PERF good many
wn21 ts*wn51-l¢ x#w21-twn55
E: k#055-ts*aj35 t‡wan51,
just
make (money) I
wrong PERF good many
‘I just counted and found that I made so many mistakes.’ ĺ unacceptable
48
Example (35) is a phonological error rather than a lexical substitution. Although the
error [tj#w51] ‘lost’ is a real word, the intended utterance [t#w51] ‘get’ and the target
[tj#w51] ‘lost’ do not have the same lexical category (adverb and verb respectively).
Thus it should be analyzed as a non-contextual phonological error in which the
consonant [j] is added. Similarly, example (36) is not a paradigmatic lexical error,
either. Although the interacting units [swan51] ‘count’ and [t‡wan51] ‘make (money)’
possess the identical lexical category (both are verbs), neither they are semantically
related nor the utterance is acceptable to native speakers. Thus it should be analyzed
as a non-contextual phonological error that the initial consonant [t‡] replaces [s]. In
conclusion, examples (35)-(36) tell apart the lexical bias in phonological error and
true paradigmatic lexical errors.
Secondly, the error is categorized as paradigmatic only when it has no
identifiable source in the current context. The boundary of context is not limited to the
phrase or sentence that the target occurred in, but is extended to the preceding
utterance and even the topic in the discourse. The restriction is due to the avoidance of
the ‘environment contamination’ in Levelt et al. (1999), who asserted lexical errors
with phonological similarity are intrusions of environment rather than the feedback
from the phonological level. If the environmental intrusion is removed from the
corpus, it will enhance the conclusion that the target is the unselected node at the
semantic level and then the information leaking between levels caused phonological
similarities. Some vague cases involving contextual influence are exhibited as
(37)-(39), which are not categorized as paradigmatic errors due to the boundary of
context.
49
(37) Unclear case: (with identifiable source in the utterance)
I: ›Ó51-p‹n35 Ûj#w21-xaj35 t‡*#051 t‡*u55 i35-twan51 ĺ
Japan
small kid
sing
out one-CL
E: ›Ó51-p‹n35 Ûj#w21-xaj35 t‡*w#051 t‡*u55 i35-twan51
Japan
small kid create
out
one-CL
‘Those Japan kids sang a song…’ĺ ‘Those Japan kids created a song…’
In (37), two possible analyses could be adopted: One is to clarify it as the
phonological anticipation of [t‡*u55] ‘out’. The vowel [u] is anticipated and resulted
in an addition in the target as [t‡*w#051] ‘to create.’ The other way is to report it as a
lexical error, in which the intended utterance [t‡*#051] ‘to sing’ is replaced by the
target [t‡*w#051] ‘to create.’ In this case, although [t‡*#051] ‘to sing’ and [t‡*w#051]
‘to create’ contain the same lexical category and are semantically-related, the error is
categorized as a phonological one with lexical bias due to the identifiable source
[t‡*u55] ‘out.’
(38) Unclear case: (with identifiable source in the preceding utterance)
A: k*(35-i21 tsaj51 k*#w21-ly51 tÛjow21 i51-tj'n21
can
again think
long
one dot
B: pu51 tÛjow21…pu51 tÛi35
NEG long
NEG hurry
‘A: You may think about that for a while.’ B: I’m not so long….I’m not in a hurry.’
Moreover, in B’s utterance of example (38), the target [tÛjow21] ‘long’ replaced the
intended utterance [tÛi35] ‘hurry.’ It is not counted as a paradigmatic error because the
target [tÛjow21] ‘long’ has appeared in the current context, i.e. A’s preceding utterance.
Here the context extends to the previous turn in the conversation. In the next example,
the boundary of context will be further expanded.
(39) Unclear case: (with identifiable source in the discourse)
I: mej35-jow21 k*o051 t*y51 k*#w21 y51-mi21 l( ĺ
NEG have
space go bake corn
PERF
E: mej35-jow21 k*o051 tÛÄ'n35...t*y51 k*#w21 y51-mi21 l(
NEG have
space roll
go bake
corn
PERF
‘ He had no more spare time to bake the corns.’ ĺ ‘… to roll the corns.’
50
At the first sight (39) seems to be a perfect paradigmatic error in which [tÛċn21]
‘roll’ replaced [k*#w21] ‘bake’ are both verbs and refer to a way of cooking. However,
this error came from the conversation in a TV program regarding a successful
enterpriser who sold baked corns before egg rolls. The target [tÛċn21] ‘roll’ has
appeared repeatedly in the previous discourse and thus the error should be excluded
from the corpus. Based on all the principles for classification, the data will be
examined in the next chapter.