Академический Документы
Профессиональный Документы
Культура Документы
YasushiMitarai
Oitalhiversitly
KazumiAizawa
7bbyoDenkiUhiversily
Glessing unknown words in a passage is useful for helping second language (L2)readers
acqtiire new vocabulary as well as overcome lexicalobstacles. However, there isno conclusive
answer te the question as to what type of glossing bestsenres these purposes. One hundredand
eighty-five Japaneselearnersof Englishread a passage with four differenttypes of glosses(a
Japanesesingle gloss, a Japanese
multiple choice g]oss,an English single gloss and an English
multiple-cheice gloss) and took a reading comprehension and two vocabu]ary tests (immediate
and delayed). Results indicated that an L2
gloss w・as the rnost effective in reading
single
comprehension, It was also discovered that firstlanguage (Ll)glosses worked better for
vocabulary Iearning
as a whole and that there was no differenceof the effect between the Ll single
lntroduction
recognize the meaning of unknown words w・hen they come across them. There are several ways
to accomplish thisiinferring the meaning of unknown words from context, using a dictionary and
referring to a gloss.
Lexicalinference isencouraged as an effect[ve strategy to overcome the lack of vocabulary in
L2 reading. Apart from beinga reading strategy, lexicalgLiessingis claimed to be instrumental
in vocabulary learning (Hulstijn,1992). Accordingto Hulstijn's effort hypothesis",
`Lmental
73
acquisition than guessing the meaning of the target words while reading in English.
Using a dictionaryrnay be expected to clear the hurdle of recognizing the correct meaning of
unknown words. In fact, itwas found that stuclents learned more words when they were allowed
to use a dictionary
while reading, than when they were not (Luppescuand Day, 1993). XVhile
consulting a dictionarymay result in better retention of new words, it is not without serieus
drawbacks. Learners'use of the dictionaryresults in lower reading speed, and may confuse the
learner,
especially ifthereare a largenumber of entries under the head word from which to choose
{Luppescuand Day, 1993). Another concern is thatfrequent lookingup of words interferes with
short-term memory and thus disruptsthe comprehension process. In addition to these problems
Hulstijnet al. (1996)
and Aizawa (1998)
offered evidence that glossing had a greater positive effect
on the retention of vocabulary than dictionaryuse.
From the above discussionwe may determinethe superiority of glossing. Yet the
reasonably
conditions. We must ask ourselves whether a multiple-choice gloss (MCG)ismore effective than
a singlegloss(SG)because the former requires more mental efforts in the semantic decision-making
process. In answer to this question, Watanabe (1997a,b)
reported seemingly contradictory
Method
Subjects
The subjects foranalysis were 185 college-Ievel students who had been selected from 221
74
initial
participants. All of them were Japanesenative speakers and were leaniing English as a
Design
There were two independentvariabl es: language of the glosses {English
the or Japanese) and
-choice). Thus cach group didthe reading task with the
the type of the glosses (single
or multiple-
)tVlteDC・t6
glaze:glaze
igo }J6
1 . )EttF(
2. ag<t6
to make
glaze:glaze something shiny
posttestscores and the scores ef a delayed vocabulary posttest that was administered two weeks
later.
Materials
The 432-word reading passage used forthisstudy was
"Compa$sion
isin the eyes" taken from
Chicken sozip forthe sottg Ler(HealthCommunication Inc.,1996). From the passage we selected as
questions. The former had Japanesequestions to be answered in Japanese and the latter
Japanesequestionsand English choices. Each question was given two points, the maximum
possible score being 16,
75
versa. "re used multip]e-choice tests rather than those that required the subjects to write the
Japaneseequivalent in order to avoid the flooreffect that might be incurredby any unexpectedly
low scores. Our decisionto employ the former type was basedon Knight(1994>, who foundthat
"select-definition"
the (i.e.
multiple-choice) test led to the learningof a greater nurnber of words
than the
`'supply-definition"
test,which required Ll equivalents fortarget words.
Procedure
The study was conducted according to a feur-week scheme in 1998. In the firstweek, we
(Aizawa)
comprising 18 sets of three Japanesewerds as question items. As seen from the
example below, there were six Englishword choices per set, threebeingkeysand the other three
being distractors:
1 kE'fa[ ]2
ge.$.-・x, eets,de [ ] 3 [ ] ltffLbl
{1) balloen (2)issue <3) schedule (4) target (5) truth (6>victory
in the VKS test,we dividedthe remaining participants into feur groups on the basis of the
vocabulary pretestscores. The fourgroups had comparable yocabulary knowledge.
In the second week, the reading comprehension testand the immediate vocabulary posttest
were administered, The participants were instructedto read the passage and answer the reading
eomprehension questionson a separate sheet of paper. During the task each group referred to
one of the fourdifferent
tyPesof glosses printed te the right of the reading passage:a JapaneseSG,
a Japanese
MCG, an EnglishSG, and an EnglishMCG. The participants were specifically asked
to read through the passage before answering the questions so that they would not rush to answer
the questions without referring to thegloss. They were allowed to refer back to the passage and
76
the glosses te answer the questions. All the participants finished the reading test within 20
minutes. During a five-minutes' recess after the test,the participants listenedto an Englishpop
song. The purpose of thisrecess was to divert
theirattention from the glossed target words they
had encountered during the rcading. The first vocabulary posttest, which had not been
announced to the participants beforehand,was conducted immediatelyafter thesong,
Two weeks laterthe second vecabulary posttest was given to investigatethe effect of delay
on vocabulary retention. Again this test had not been announcedbeforehand. It wasconfirmed
thatno participants had studied the target words between the twe posttests.
DataAnelysis
For the analysis we furthernarrowed down the subjects in the two MCG groups. "'e
eliminatecl those who scored lower than 10 out of the maximum possible score of 14 in the choice
of the correct meaning in the MCGs. The reason forthis screening was that we feared the failure
to choose the correct a]ternative in the MCGs would adversely affect the performance in the
vocabulary posttests(Watanabe, 1997a), AIthough the number of subjects in MCG groups was
considerably reduced, no significant differencewas confirmecl between the inean scores ef the
yocabulary pretest of the remaining subjects of fourgroups, F C3,l81)O,69,n.s. =
We used a 2 (language)
x 2 (SGIMCG) de$ign to calculate the analysis of varlance (ANOVA)
on the reading comprehension test. For the vocabulary posttests a three-way ANOVA was
measures on the third factor. Because we had counterbalanced the language of choices forhalfof
the question items intwo posttests,the raw data were converted into T-scoresforana}ysis. The
advantage of this conversion is that we can compare the performances in tests of potentially
different
units of measurement.
Results
VocabularyLearning
Table 1 gives the means and standard deviationsof vocabulary posttest scores. Looking at
very slight decrease of the other three groups. These figuresshould not be interpreted as
indicatingactual rise in the raw score. Because the figuresrepresent T scores, they indicatethe
relative performances of each group within respective tests. Hence this interestingbetter
performance by the EnglishMCG group in the delayed test must be understood not as an absolute
gain but as a gain produceclthrough comparison with the scores of the other groups.
Table 2 shows the results of a three-way ANOVA calculated on the vocabulary posttests.
As seen from Table 2, the main effects of language and SG/MCG were significant in the
vocabulary posttests. Whereas the conversion of the raw scores resulted in no significant
77
Immediate Delayed
group n M so M sa
source ss of MS F p
betweensubjects
Language(A) 3351.57 111T81 3351.571344.l4 25.81IC}.35 O.Oo-qctsO.oe-tsO.05+
SG 1 MCG (B) 1344.14
AxB 493.5Z23500.71 493.52 3.80
error 129.83
withinsubjects
difference
between the two posttests,this isnot a serious drawback in our study. This isbecause
we were concerned with the comparison of the variables within each test and we took it for
granted that there would be overall attrition of newly learned vocabulary retention. VLre,
therefore, did not includeraw scores in our analysis.
A furtheranalysis on interactienrevealed the following:The SGIMCG effect was significant
glossing, F(1,181 )
in English =i 13.35,P< .Oel,but not in Japaneseglossing,F{1,181) =- O.80,n.s.
The language effect was significant both in the SG and MCG, F(1,181) =
4.90,P < .05,F(1,181) =
F(1,181) 2.61,n.s,
=
78
Reading Comprehension
Table 3 presents the means and standard deviationsof the reading comprehension test
scores,
group n M so
Japanese
SG 5349533e 9.59 3.493.693.694.27
Japanese
MCG 8,1611,OO
English SG
EnglishMCG 8,67
The results of an ANOVA performed on the reading comprehension test are presented in
Table 4. As Table 4 shows, the SGs were significantly more effective thanMCGs, and, somewhat
sources ss of MS F p
error 14,25
Discussion
Glossesand VocabularyLearning
Multipte-Choice
Both of the main effects of the Japanese
glosses and the SGs were confirmed in vocabulary
learning. The factthat itwas the Japaneseglesses which were more effective forthe retention of
the rrieaning of targetwords will not be surprising formany L2 teachers. They are empirically
79
the fact,however, that the positivernain effect of SGs can be attributable to thepoor results of the
EnglishMCG the EnglishSG group; there
group cempared with was by contrast no significant
difference
betweenthe two Japaneseglos$ groups, A probable cause of this was the subjects'
English proficiency.As Watanabe (1997b)
found with his subjects, lower proficiency Iearners
benefitedmore from an Ll SG than an L2 MCG. Certainlyitisimpossibleto directlycornparethe
proficiency of subjects in our study and Watanabe's. Yet, compared with Watanabe'ssubjects,
who were students (non-native
pre-university, undergraduate or graduate)in the U.S.,our subjects
are unlikely to have beenlessweak in theirEnglishabilities. Itseems, therefore,reasonable to
conclude that even though we selected those who had scored more than 9 points in the choice of
alternatives of MCGs, theyhave not reached the L2 thresholdlevelto endorse the superiority of
MCGs. If we had furtherrestricted the subjects for analy$is to those who scored, forexample,
mere than 1O,the results might havebeenreversed. However, had we cloneso, we would nothave
been able to validate the equality of the vocabulary knowledge oi the four groups on the basisof
thevocabulary pretest. In addition, a serious imbalancewould have risen betweenthegroups by
eliminating a largenumber of subjects, especially from the English MCG group, in which only 16
subjects scored more than 10. 0ne conceivable way to make the MCG absolutely fail-safe
rnaybe
to provide MCGs with almost synonymous alternatives, whichever learnerschoose and incidentally
learnwill work in the posttest.
In spite of the above discussion,
we cannot totall},rule out the significance of MCGs forthe
following reasons, First of all, as far as the Japanese
gloss groups were concerned, the MCG
group did as well as, ifnot betterthan,the SG group, Furthermore, as the factthat interaction
between SG!MCG and period conditions tended to be significant indicates,MCG groups, in
particularthe English MCG group, performed unexpectedly well inthe delayedposttest. This
leads us to the question whether an L2 MCG has a more positive long-term effect than other
glosses on the retention of newly learnedvocabulary. This point may be interesting
enough to
merit furtherresearch.
L2 Single-Gloss
and Reading Comprehension
To turn to reading comprehension, itwas found that the English SG was the most effective.
relatively easy language. Although this may have been favorable,we must be reminded that
80
comprehension isgenerally easier to attain in Ll than in L2. Therefore we cannot determine that
the simple English glossing excessively contributed to the higher scores. The more like!},
explanation isthat Englishgloss users were encouraged to be rnore deeply involvedin the reading
process because the L2 gloss ma}r have requirecl more striving in reading. The inferiorresults of
MCG groups will be ascribable to the fact thatreaders in thesegroups devoted too much mental
energy to the decision-making process of meaning selection for each glossed word. In other
words, the MCGs had similarly disruptive
effects on the readers, as use <Aizawa,
diddictionary
1998, Luppescu and Day, 1993), We must point out that the study would have been more
complete, if the four groups had been formed with equal ability in L2 reading as well as L2
vocabulary, though itwould have demanded a fairlycomplicated process to realize this.
Conc[usion
favorableeffect in vocabulary learning, Until the questionhas been crarified, teachers have no
single best t},pe of gloss to serve bothpurposes and are leftto choose whichever bestserves fer
theirs.
Note
References
Aizawa, K. (1998).
Incidentalvocabulary learningthrough
Guessing exercises, glossing reading: or
81
Readi7zgin a FoxeignLangucrge,7(2),541-551.
Furuya, T. & Mitarai,Y. (1995). A comparative study of the effect of monofingual and bilingual
London:MacmMan.
J. Hollander,M. & Greidanus,
Hulstijn, , T. (1996).
Incidentalvocabulary learningby advanced
82