Professional Documents
Culture Documents
The Negation Gap in ESL
The Negation Gap in ESL
Abstract
The present study examines how Korean-speaking learners of
English as a second language understand English ‘bi-negative’
sentences, which contain both not and a negative pronoun (e.g., No
one didn’t cry, I didn’t do nothing). A semantic similarity judgment
task demonstrates that native speakers of English clearly accept
patterns of this type and assign them an affirmative interpretation.
However, the same task reveals that our 31 Korean participants are
largely unaware of this usage and reject the affirmative meaning
of the bi-negative sentences. A translation experiment involving
a different group of 32 Korean-speaking learners reveals a strong
tendency to interpret the bi-negative patterns as if they were the
sort of negative concord pattern typical of Korean. These results
suggest the presence of first-language transfer, even among
advanced speakers.
1. Introduction
A second option, less common but still widely used, makes use of a
negative pronoun (also sometimes called a ‘negative quantifier’) that
denotes a null set. We will refer to this option as ‘argument negation.’
1
An alternative way to express negation involves the ‘long form’ -ci anhta, which
we will set to the side for the purposes of this paper.
528 Miseon Lee, William O’Grady, and Gayoung Lee
‘With nothing/nobody etc., we do not use a negative verb (isn’t, didn’t etc.):
I said nothing. (not I didn’t say nothing)’
(Murphy 2019:172)
The Negation Gap in ESL 529
2. Methods
2.1.1. Participants
A total of 20 native speakers of English took part in a semantic similarity
judgment task to provide a baseline against which to assess the performance
of second-language learners. The participants ranged in age from 19 to 70
years (mean age = 45.1 years, 7 males); all were college graduates or college
students. They had normal or corrected-to-normal vision and no known
530 Miseon Lee, William O’Grady, and Gayoung Lee
2.1.2. Materials
There were a total of 40 critical test items, 10 for each of the four critical
conditions exemplified below in Table 1. Half of these items contained
just one negative – either the negative pronoun no one in subject position
or the negative pronoun nothing in direct object position. The other half
were bi-negative patterns, with the sentential negative not in addition to the
negative pronoun in the subject or direct object position. As exemplified in
Table 1, each sentence was paired with another sentence that had roughly
the same meaning.
Paired semantically
Condition Test sentence
similar sentence
Mono-negative with …
subject no one No one cried all day. Not a single person cried all
day.
direct object nothing She often read nothing. She often didn’t read anything.
Bi-negative with …
subject no one No one didn’t cry Everyone cried during the
during the movie. movie.
direct object nothing Actually they didn’t do They actually did something.
nothing.
2.1.3. Procedure
A semantic similarity judgment task was created and implemented
using the PsyToolkit platform (Stoet, 2010, 2017). Each trial began with a
fixation cross displayed at the center of the monitor for 500 milliseconds
(ms). Then a stimulus consisting of a pair of sentences was presented in a
centered position on the monitor. For each trial, participants were instructed
to indicate whether the two sentences had roughly the same meaning by
pressing the key designated as “Same” or “Different.” Once the participant’s
response was registered or after a maximum duration of 10 seconds, the
task automatically proceeded to the next trial.
The test session began by presenting the procedural instructions on the
computer monitor. Five practice pairs were then given to familiarize the
participants with the task. The whole experiment lasted about 10 minutes
on average. Responses (indicated by pressing a button) were recorded for
subsequent statistical analysis.
532 Miseon Lee, William O’Grady, and Gayoung Lee
2.1.4. Results
The participants judged the two sentences in each pair of critical items as
having the Same meaning in the majority of cases, as documented in Table
2. A repeated measures ANOVA was conducted with two within-subjects
factors: Negation type (mono-negative, bi-negative) and Location of the
negative pronoun (subject, object). The results showed no main effects of
Negation type (F(1, 19) = 2.42, p > .1) or Location (F(1, 19) = 3.24, p > .05).
The interaction between the two factors was also not significant (F(1, 19) =
3.10, p > .05).
Table 2. Number of critical pairs that native speakers judged as Same (Mean
percentage and SE in parentheses)
Mono-negative Bi-negative
With Subject no one 97/100 (97%, SE=1.8) 90/100 (90%, SE=5.7)
With Direct Object nothing 96/100 (96%, SE=2.0) 83/100 (83%, SE=5.9)
193/200 (97%, SE=3.0) 173/200 (87%, SE=3.2)
2.1.5. Discussion
Our findings confirm a very strong tendency for native speakers of
English to assign an affirmative interpretation to bi-negative constructions.
Even in the absence of a supporting context, 90% of the subject patterns
and 83% of the object patterns received this reading.
2.2.1. Participants
A total of 31 native speakers of Korean took part in the same semantic
similarity judgment task. One participant was removed from the analysis
due to scores 2.5 SDs below the group mean. We were left with 30
participants (mean age = 24.3 years, 14 males), all of whom had started
The Negation Gap in ESL 533
2.2.3. Results
The number of critical pairs that Korean participants judged as ‘Same’ is
reported for each proficiency group in Table 3. Of particular relevance are
the mean scores for the bi-negative constructions, which were far below
those of mono-negative constructions – 50% compared to 78% in the case
of subject patterns and 46% versus 91% in case of the object patterns.
534 Miseon Lee, William O’Grady, and Gayoung Lee
Mono-negative Bi-negative
With Subject no one
Advanced (N=8) 37/40 (93%, SE=5.3) 21/40 (53%, SE=15.1)
High-intermediate (N=12) 50/60 (83%, SE=8.2) 26/60 (43%, SE=9.9)
Low- intermediate (N=10) 30/50 (60%, SE=8.5) 28/50 (56%, SE=12.4)
Subtotal 117/150 (78%, SE=4.8) 75/150 (50%, SE=6.8)
With Direct object nothing
Advanced (N=8) 40/40 (100%) 28/40 (70%, SE=9.3)
High-intermediate (N=12) 56/60 (93%, SE=8.0) 21/60 (35%, SE=9.6)
Low- intermediate (N=10) 41/50 (82%, SE=5.5) 20/50 (40%, SE=8.5)
Subtotal 137/150 (91%, SE=3.9) 69/150 (46%, SE=5.9)
254/300 (85%, SE=3.1) 144/300 (48%, SE=4.5)
2.2.4. Discussion
In sum, Korean learners of English, including those who are at an
advanced level, accept the affirmative interpretation of English bi-negative
patterns only about half the time – 50% of the time when there is a subject
negative pronoun and 46% when there is object negative pronoun. These
scores are far lower than those for mono-negatives (78% and 91.3%,
respectively), clearly pointing to a fundamental gap in the L2 learners’
knowledge of English bi-negatives.
These findings raise a new question. Given that Korean learners of
English reject the standard interpretation of bi-negative sentences, what
meaning (if any) do they assign to these patterns? We turn to this question
next.
2.3.1. Participants
A total of 35 native speakers of Korean who had not participated in
Experiment 2 took part in an English-to-Korean translation test. Three
participants were removed from the analysis due to accuracy scores 2.5 SDs
below or above the group mean, leaving 32 participants (mean age = 24.5
years, 12 males). All the participants had been born and raised in Korea,
and were undergraduate students at a university in Seoul who had started
learning English as a second language before puberty in a school context.
Based on the standard score ranges established by the TOEIC committee,
ten participants were identified as advanced learners (TOEFL iBT scores of
95 or better, or TOEIC scores of 945 or better), and 20 as high-intermediate
learners (TOEFL iBT scores ranging from 68 to 94, or TOEIC scores
536 Miseon Lee, William O’Grady, and Gayoung Lee
ranging from 785 to 944).3 All had normal or corrected-to-normal vision and
no known history of neurological, psychiatric, or language disorders. They
gave informed written consent as required by our IRB, but were unaware of
the purpose of the study. They received modest monetary compensation for
their participation.
3
Given that even the advanced learners differed to a high degree from native
speakers in their acceptance of the affirmative meaning of bi-negative patterns in
Experiment 2, we did not include low-intermediate learners in Experiment 3.
The Negation Gap in ESL 537
2.3.3. Results
Our accuracy measure corresponded to the proportion of correct
translations of critical items. The mean accuracy for the critical conditions
was 57.7% (SE = 4.2%), which was much lower than the mean accuracy for
the fillers 96.7% (SE = 2.3%). Table 5 summarizes the mean accuracy for
each of the critical conditions.
4
We also judged translations of bi-negatives as ‘correct’ if they contained two
negative forms: for example, amwukesto an-po-n kes-un anita for ‘They didn’t see
nothing.’
538 Miseon Lee, William O’Grady, and Gayoung Lee
2.3.4. Discussion
The translation data confirm that Korean ESL learners reject the standard
affirmative interpretation of English bi-negative sentences. Whereas native
English speakers take patterns such as No one didn’t cry to mean (roughly)
‘Everyone cried,’ the Korean ESL learners interpret them as a simple
negative (‘No one cried’) in the vast majority of cases. A similar but weaker
tendency occurs for object bi-negatives such as She didn’t do nothing, for
which the preferred translation was the Korean equivalent of ‘She did
nothing’ rather than ‘She did something.5 Clearly, then, exposure to English
bi-negative patterns creates very significant confusion for Korean second
language learners.
5
We set to the side for now the question of why performance on the subject
patterns was somewhat lower than on the object patterns.
The Negation Gap in ESL 541
English negation. (That is, they take No one didn’t cry to mean ‘Everyone
cried’ and I didn’t do nothing to mean ‘I did something.’)
Second, it is also overwhelmingly clear that even advanced Korean-
speaking learners of English as a second language are often unaware of
bi-negatives and their usage. As we have seen, the participants in our second
experiment commonly rejected the match between bi-negative patterns and
the paraphrases with which they were paired. This points to a substantial
gap in the proficiency of college-level learners of English.
These gaps in the proficiency of even advanced learners of English can
almost certainly be traced to the lack of exposure to bi-negative patterns.
The English curriculum that is followed throughout Korea makes no
mention of bi-negative patterns. Our search of 54 English textbooks that
are widely used in middle schools and high schools (published by NE
books, YBM, Chunjae Education Inc. and Visang) found no instances
of bi-negatives. Various popular commercial reference books for high
schoolers (e.g., HB Grammar published by Chunjae Education Inc.) also
do not explain or exemplify bi-negative patterns. Tellingly, the two most
popular machine translators (i.e., Naver Papago, Google Translate) produce
incorrect translations for bi-negatives.
Third, our translation experiment demonstrated that Korean-speaking
learners of English manifest a strong tendency to interpret English
bi-negatives as if they were the sort of negative concord pattern typical
of Korean – a classic example of first-language transfer. The result is the
interpretation normally associated with mono-negative sentences in English:
No one didn’t cry is wrongly taken to mean ‘No one cried’ and I didn’t do
nothing is misinterpreted as ‘I did nothing.’
These findings constitute a promising starting point for further study.
Important experimental work on the acquisition of different negative
patterns has been done by Thornton et al. (2016), who compared the
interpretive preferences of adults and 3- to 5-year old children. The test
542 Miseon Lee, William O’Grady, and Gayoung Lee
Acknowledgements We are grateful to Miho Choo and two anonymous reviewers for
their comments and advice.
Funding This work was supported by the Ministry of Education of the Republic of Korea
and the National Research Foundation of Korea (NRF-2019S1A5A2A03053390).
Declarations
Ethics Approval This study was approved by the Institutional Review Board of the
University of Hawaii at Manoa.
Consent to Participate and Consent for Publication All participants gave informed
written consent in accordance with the requirements of the Institutional Review Board.
References