CHAPTER FOUR
RESULTS AND DISCUSSIONS
This chapter contains two sections. The first section presents the results of
different test measures concerning the four research questions and the second section
discusses the performances of the Explicit Group and the Inferential Group under
different exposure frequency.
4.1 Results
In this section, analyses of data are presented as follows. First, the participants’
baseline knowledge of the target words is described. Second, the participants’ word
gains between the pretest and the immediate posttest and the retention of the word
gains between the pretest and the delayed posttest are examined through
paired-samples t test. Then, effects of exposure frequency on vocabulary growth are
analyzed through one-way repeated measures ANOVA. Next, effects of contextual
situations on vocabulary growth are also analyzed through independent-samples t test.
Finally, the interaction between exposure frequency and contextual situations is tested
through two-way repeated measures ANOVA.
4.1.1 Participants’ Baseline Knowledge of the Target Words
Since the target words were chosen from a level of the CEEC word list that was
the target participants was double-checked by two experienced teachers, the two
chosen classes were sure to have little knowledge of the target words and thus had an
equal start. To confirm the two groups had fair baseline knowledge of the target
words, the independent-samples t test was administered to compare their mean scores
on the pretest. As seen in Table 3, the two groups’ mean scores for the target items
on the pretest were different by .68 and the difference was not significant (p > .05).
Therefore, the two groups were not statistically different in the baseline knowledge of
the target words.
Table 3
The Comparison of Two Groups’ Correct Target Items on the Pretest
Group N Mean SD Sig.
Explicit 37 3.03 1.32
Inferential 41 3.71 1.83
.07
4.1.2 Word Gains from Incidental Learning
Research Question 1: Can EFL senior high school students acquire vocabulary
incidentally from reading?
To know if there was any immediate vocabulary growth after reading, both
groups’ performances on the immediate posttest were compared with those on the
pretest through paired-samples t test. As shown in Table 4, the mean scores of both
groups’ immediate posttest were higher than those on the pretest. Among items
testing target words between the pretest and the immediate posttest, the average score
of the Explicit Group increased by 2.13 and the Inferential Group gained 2.53, with
the score differences reaching a significant level (p < .01). Thus, as expected, it was
evident that senior high school students could learn vocabulary incidentally through
reading. Furthermore, if the participants did not know any of the target words and
their scores on the pretest were simply obtained from wild guesses,
3it was estimated
at most that the average students in the Explicit Group were likely to remember 43%
(i.e. 5.16 out of 12) of the target word meanings on the immediate posttest and those
in the Inferential Group could still retain as high as 52% (i.e. 6.24 out of 12) in their
memory.
3
Such speculation is not groundless. The two groups’ mean scores for the target
items were 3.03 and 3.71 out of the total score 12, so their percentages of correct
answers were 25.25% and 30.92% respectively. Since both percentages were close
to 25%, the chance of guessing on a multiple-choice test with four options, it was
likely that most of these participants did not know the meanings of the target words
Table 4
Scores for All Target Items
Pretest (P)
Immediate
Posttest (IP)
Delayed
Posttest (DP) Sig.
Group N Mean SD Mean SD Mean SD P – IP P - DP
Explicit 37 3.03 1.32 5.16 1.88 4.16 2.22 .00** .01**
Inferential 41 3.71 1.83 6.24 2.47 4.98 2.14 .00** .00**
**p < .01, two tailed.
Not only did the beneficial effect of incidental learning take place in students’
immediate word gains, but it contributed to significant vocabulary retention as well.
As shown in Table 4, both groups had a better performance on the delayed posttest
than on the pretest by 1.13 and 1.27 respectively, with the differences of their mean
scores between the pretest and the delayed posttest reaching a significant level (p
< .01).
4.1.3 Students’ Performance Under Different Exposure Frequency
Research Question 2: Does exposure frequency influence EFL learners’ vocabulary
acquisition?
As illustrated in Tables 5, 6, and 7, there was a general tendency that the more
frequently a word was exposed, the better mean scores both groups got on the two
posttests. To be more specific, words under six exposures were scored the highest
among all kinds of exposures, words under five exposures were scored higher than
those under fewer exposures on the delayed posttest and on the Inferential Group’s
immediate posttest, words under four exposures were scored higher than those under
fewer exposures, and words under three exposures were scored higher than those
without exposures on the immediate posttest and on the Explicit Group’s delayed
posttest.
Table 5
All Students’ Scores Under Different Exposure Frequency on the Two Posttests
Immediate Posttest Delayed Posttest
Exposure
Frequency
N Mean SD Mean SD
0 78 .91 .74 .79 .75
3 78 1.15 .94 .79 .76
4 78 1.40 1.01 1.13 .90
5 78 1.32 .76 1.21 .90
6 78 1.86 .89 1.46 .94
Total 390 1.33 .93 1.08 .89
**p < .01
Note. Data of zero exposures were obtained from students’ performance on the three items testing no-context control words in the instrument.
Table 6
The Explicit Group’s Scores Under Different Exposure Frequency on the Two Posttests
Immediate Posttest Delayed Posttest
Exposure
Frequency
N Mean SD Mean SD
0 37 .81 .70 .76 .75
3 37 1.08 .89 .84 .76
4 37 1.32 .91 1.03 .90
5 37 1.11 .70 1.05 .90
6 37 1.65 .86 1.24 .94
Total 185 1.19 .86 .98 .89
*p < .05
Table 7
The Inferential Group’s Scores Under Different Exposure Frequency on the Two Posttests
Immediate Posttest Delayed Posttest
Exposure
Frequency
N Mean SD Mean SD
0 41 1.00 .77 .83 .74
3 41 1.22 .99 .76 .80
4 41 1.46 1.10 1.22 .91
5 41 1.51 .78 1.34 .91
6 41 2.05 .89 1.66 .86
Total 205 1.45 .97 1.16 .90
**p < .01
Unexpectedly, there were three pieces of deviance found in these three tables.
The first one was observed on the immediate posttest, where the mean score of four
exposures was higher than that of five exposures by .08 among all participants (see
Table 5). When analyzed within each group, the irregular result was only found in
the Explicit Group on the immediate posttest (see Table 6) but not in the Inferential
Group (see Table 7). Another unusual drop of the Inferential Group’s mean scores
from zero exposures to three exposures by .07 on the delayed posttest was noticed in
Table 7. Because of the drop, the other unexpected result was found in the same
mean score for words with no exposures and those exposed three times among all
participants on the delayed posttest (see Table 5).
Apart from the above three minor unexpected findings, the majority of the
results supported the general tendency that the increase of exposure frequency
facilitated vocabulary acquisition.
4.1.3.1 The Significant Exposure Frequency
Since exposure frequency was a within-subject factor in the study, one-way
repeated measures ANOVA was conducted to test the significance of the factor.
Table 8 shows on the immediate posttest, the differences between different word
occurrences in all participants (F = 15.15, p < .01) and in each group (the Explicit
significant level.
Table 8
One-way Repeated Measures ANOVA Comparison of Students’ Scores Under
Different Exposure Frequency on the Immediate Posttest
SS df MS F Sig.
All Participants 38.35 4 9.59 15.15 .00**
Explicit Group 14.45 4 3.61 6.17 .00**
Inferential Group 25.35 4 6.34 9.30 .00**
**p < .01
However, further pairwise comparisons shown in Table 9 revealed that such
significant differences were mainly found between words under six exposures and
those under other exposure frequency. To be more specific, scores of six exposures
were significantly higher than those of any other exposure frequency except in the
Explicit Group, in which the score of words under six exposures was only higher than
that without exposures.
There were other minor significant differences found between the no-exposure
words and words exposed more than three times among all participants and between
the no-exposure words and word exposed four times in the Explicit Group.
Nevertheless, no more significant difference was found among scores of exposures
fewer than six times.
Table 9
Pairwise Comparisons of Students’ Scores Under Different Exposure Frequency on
the Immediate Posttest
All Participants Explicit Group Inferential Group
fre-
quency
fre-
quency
Mean
Difference Sig.
Mean
Difference Sig.
Mean
Difference Sig.
0 3 -.24 .36 -.27 1.00 -.22 1.00
4 -.49 .01** -.51 .03* -.46 .35
5 -.41 .00** -.30 .62 -.51 .12
6 -.95 .00** -.84 .00** -1.05 .00**
3 4 -.24 .97 -.24 1.00 -.24 1.00
5 -.17 1.00 -.03 1.00 -.29 1.00
6 -.71 .00** -.57 .06 -.83 .00**
4 5 .08 1.00 .22 1.00 -.05 1.00
6 -.46 .01** -.32 .97 -.59 .02*
5 6 -.54 .00** -.54 .06 -.54 .03*
*p < .05, **p < .01
In addition, on the delayed posttest, one-way repeated measures ANOVA showed
there was no significant difference between each exposure frequency in the Explicit
Group (F = 2.25, p > .05),
4but the differences between different exposure frequency
in all participants (F = 10.17, p < .01) and in the Inferential Group (F = 8.90, p < .01)
reached a significant level (see Table 10).
Table 10
One-way Repeated Measures ANOVA Comparison of Students’ Scores Under
Different Exposure Frequency on the Delayed Posttest
SS df MS F Sig.
All Participants 25.44 4 6.36 10.17 .00**
Explicit Group 5.44 4 1.36 2.25 .07**
Inferential Group 22.86 4 5.72 8.90 .00**
**p < .01
Nevertheless, further pairwise comparisons revealed that such significant
differences were mainly found between words exposed less than four times and those
under five and six exposures (see Table 11).
4
The differences in the Explicit Group might be neutralized by the guessing effect.
If only items answered correctly on both posttests could be counted as true retention
on the delayed posttest, the group’s scores of words under six exposures were still
significantly different from those under three and zero exposures and scores of words
with exposures were significantly different from those without exposures.
Table 11
Pairwise Comparisons of Students’ Scores Under Different Exposure Frequency on
the Delayed Posttest
All participants Inferential Group fre-
quency
fre-
quency Mean Difference Sig. Mean Difference Sig.
0 3 .00 1.00 .07 1.00
4 -.33 .16 -.39 .58
5 -.41 .02* -.51 .06
6 -.67 .00** -.83 .00**
3 4 -.33 .08 -.46 .10
5 -.41 .01** -.59 .03*
6 -.67 .00** -.90 .00**
4 5 -.08 1.00 -.12 1.00
6 -.33 .09 -.44 .11
5 6 -.26 .47 -.32 .51
*p < .05, **p < .01
The prevailing significant difference only found between words exposed six
times and those exposed fewer times suggested that the minimum number of
exposures to the target words was six for the senior high students in the study to
acquire words from context significantly.
4.1.4 Effects of Contextual Situations
Research Question 3: Do types of contextual situations of words in a text influence
EFL learners’ vocabulary acquisition?
Although there is no definite answer to this question in the study, the following
results shed some light on inferential contexts as the preferable type of contextual
situations for senior high school students in acquiring vocabulary from reading.
4.1.4.1 Better Vocabulary Acquisition Under Inferential Contextual Situations
Since the two classes under different contextual situations were statistically equal
on the pretest, whether one contextual situation was superior to the other in
facilitating vocabulary acquisition was decided by which class scored higher on both
posttests. As displayed in Table 12, the Inferential Group outperformed the Explicit
Group on both posttests except on words under three exposures on the delayed
posttest, and the differences between the two groups reached a significant level on the
sum of all target words and words under five exposures on the immediate posttest.
Although there were no more significant differences found in Table 12, the
differences between these two groups were getting closer to a significant level as the
number of word exposures increased, with the difference under six exposures almost
reaching a significant level on both posttests.
Table 12
Target Word Gains on Two Posttests in Each Contextual Situation
Immediate Posttest Delayed Posttest
Frequency Group N Mean SD Sig. Mean SD Sig.
All TWs Explicit 37 5.16 1.88 4.16 2.22
Inferential 41 6.24 2.47
.03*
4.98 2.14
.10
6 Explicit 37 1.65 .86 1.24 .98
Inferential 41 2.05 .89
.05
1.66 .86
.05
5 Explicit 37 1.11 .70 1.05 .88
Inferential 41 1.51 .78
.02*
1.34 .91
.16
4 Explicit 37 1.32 .92 1.03 .90
Inferential 41 1.46 1.10
.55
1.22 .91
.35
3 Explicit 37 1.08 .89 .84 .73
Inferential 41 1.22 .99
.52
.76 .80
.64
*p < .05
The Inferential Group outperformed the Explicit Group not only within the same
exposure frequency but also across different times of exposures in some cases (see
Table 13). On the immediate posttest, the score of words occurring fewer than five
times in the Inferential Group were higher than that of words occurring five times in
the Explicit Group; on the delayed posttest, the score of words occurring five times in
the Inferential Group were higher than that of words occurring six times in the
Explicit Group and the same situation appeared in the four exposures of the
Inferential Group versus the five exposures of the Explicit Group.
Table 13
Rank of Mean Scores on Both Posttests
Immediate Posttest Delayed Posttest
Rank Mean Group Frequency Mean Group Frequency
2.05 Inferential 6 1.66 Inferential 6
1.65 Explicit 6 1.34 Inferential 5
1.51 Inferential 5 1.24 Explicit 6
1.46 Inferential 4 1.22 Inferential 4
1.32 Explicit 4 1.05 Explicit 5
1.22 Inferential 3 1.03 Explicit 4
1.11 Explicit 5 .84 Explicit 3
1
2
3
4
5
6
7
8 1.08 Explicit 3 .76 Inferential 3
4.1.4.2 Smaller Forgetting Rates in Inferential Contextual Situations
When target words were embedded in inferential contextual situations,
participants not only obtained more word gains but also demonstrated less forgetting.
As shown in Table 14, the Inferential Group forgot 16.67% of the words learned under
four exposures and 19.05% under six exposures at the rate 5.78% and 5.54% less than
their counterparts in the Explicit Group respectively.
Table 14
The Forgetting Rates Between the Two Contextual Situations
Explicit Group Inferential Group
Mean Mean
Expo- sure Frequ-
ency
Immediate Posttest
Delayed Posttest
Mean Difference
FR (%)
Immediate Posttest
Delayed Posttest
Mean Difference
FR (%)
3 1.08 .84 .24 22.50 1.22 .76 .46 38.00
4 1.32 1.03 .30 22.45 1.46 1.22 .24 16.67
5 1.11 1.05 .05 4.87 1.51 1.34 .17 11.29
6 1.65 1.24 .41 24.59 2.05 1.66 .39 19.05
Note. The forgetting rate (FR) was calculated as suggested by Groot (2000): the difference between the
mean score on the immediate posttest and that on the delayed posttest, divided by the mean score on
the immediate posttest, and then converted into percentage
.
Although words learned under three and five exposures in the Explicit Group
were less forgotten than those in the Inferential Group, the contradictory results could
be explained. First, the low forgetting rate of words under five exposures in the
Explicit Group actually resulted from their poor gains on the immediate posttest, not
from a better retention rate. Second, since deriving word meanings from inferential
context requires more mental effort as suggested by Hulstijn (1992) and thus demands
more exposures before correct meanings are inferred, three exposures may be too few
for the Inferential Group to secure their word knowledge. Therefore, although the
Inferential Group scored higher on words under three exposures than the Explicit
Group on the immediate posttest, it is suspected that the higher gains were built up
largely by guesswork, which explains why the score of the Inferential Group under
three exposures decreased so dramatically on the delayed posttest that the score was
even lower than that of words with no exposure (see Table 7). The inability to infer
word meanings from three exposures is also consistent with Hulstijn, Hollander, and
Greidanus’s (1996) nonsignificant difference of participants’ word meaning retention
performance between words appearing three times and those appearing only once
when no external meaning assistance was provided. Aside from these two
explainable exceptions, words embedded in inferential contextual clues that required
more mental inferring ability generally were not forgotten as easily as those
embedded in explicit clues.
4.1.5 The Relationship Between Exposure Frequency and Contextual Situations
Research questions 4: What is the relationship between exposure frequency and types
of contextual situations in EFL incidental learning of
vocabulary?
Although two-way repeated measures ANOVA with exposure frequency as a
within-subject factor and contextual situations as a between-subject factor revealed
nonsignificant interactions between these two factors in Table 15, there is an
interrelationship observed between exposure frequency and contextual situations,
which will be discussed in Section 4.2.4.
Table 15
Tests of Within-subjects Effects in Two-way Repeated Measures ANOVA
Source SS df MS F Sig.
Frequency 20.94 3 6.98 10.36 .00**
Frequency * Context 1.35 3 .45 .67 .57 Immediate
Posttest
Error (Frequency) 153.58 228 .67
Frequency 16.95 3 5.65 9.09 .00**
Frequency * Context 2.59 3 .87 1.39 .25 Delayed
Posttest
Error (Frequency) 141.73 228 .62
**p < .01
4.2 Discussions
The results concerning the four research questions of the study are discussed as
follows.
4.2.1 Benefits of Incidental Learning of Vocabulary
The results in Section 4.1.2 correspond with previous studies that have attested to
word gains from incidental learning (Day, Omura, & Hiramatsu, 1991; Eeds &
Cockrum, 1985; Horst, Cobb, & Meara, 1998; Hulstijn, Hollander, & Greidanus, 1996;
Knight, 1994; Nagy, Herman, & Anderson, 1985; Nagy, Anderson, & Herman, 1987;
Nation, 2001; Nation & Waring, 1997; Oxford & Scarcella, 1994; Paribakht &
Wesche, 1997; Paribakht & Wesche, 1999; Pitts, White, & Krashen, 1989; Swanborn
& de Glopper, 1999; Tekman & Daloglu, 2006; Zahar, Cobb, & Spada, 2001). In
addition, the present study finds significant delayed retention from incidental
vocabulary learning. To further explore the value of such retention, it is worthwhile
to compare forgetting rate percentages between incidental learning and rote learning.
According to Groot’s (2000) definition, forgetting rate percentages are “the
percentage of the words learned that was forgotten during the period of time between
the immediate and the delayed tests” (p. 76). In the case of the present study, the
average number of words learned on the immediate posttest for the Explicit Group
was 5.16 and the Inferential Group 6.24, with the average score decreasing by 1.00 for
the former group and 1.27 for the latter between the two posttests. After these
figures are converted into forgetting rate percentages, it is found that 19.38% of the
words learned are forgotten in the Explicit Group and 20.32% in the Inferential Group
(see Table 16).
Table 16
The Forgetting Rates of All Target Items Between Two Posttests
Immediate
Posttest
Delayed
Posttest
Group N Mean SD Mean SD
Mean
Difference
Forgetting
rate (%) Sig.
Explicit 37 5.16 1.88 4.16 2.22 1.00 19.38 .02*
Inferential 41 6.24 2.47 4.98 2.14 1.27 20.32 .00**
*p < .05, **p < .01
Compared with the bilingual list method in the four experiments analyzed in
Groot (2000), in which 33%, 40%, 60%, and 26% of words learned were forgotten
respectively by students aged 16-20 in a two-to-three-week interval, the forgetting
rate percentages over a four-week span in the present study are much lower.
Probably it is meaningfulness and context that prevent forgetting from happening too
quickly, as reviewed in Beheydt (1987), because the method of learning in the present
study is more meaningful and context-dependent than rote learning of bilingual lists.
1994; Mondria & Boer, 1991), derived meaning in incidental learning may demand
deeper mental processing and thus yield higher retention than memorizing given
meaning.
Though there are significant drops in both groups’ average scores between the
immediate posttest and the delayed posttest (see Table 16), the significant learning
outcome and better-retained progress than rote learning is observed in incidental
vocabulary learning.
4.2.2 Effects of Exposure Frequency on Vocabulary Growth
In Section 4.1.3, the present study finds a general tendency that supports positive
effects of repeated exposures on vocabulary learning in many studies (Horst, Cobb, &
Meara, 1998; Hulstijn, Hollander, & Greidanus, 1996; Jenkins, Stein, & Wysocki,
1984; Kyongho & Nation, 1989; Nagy, Herman, & Anderson, 1985; Paribakht &
Wesche, 1997; Rott, 1999; Saragi, Nation, & Meister, 1978; Schmitt & Carter, 2000;
Tekman & Daloglu, 2006; Zahar, Cobb, & Spada, 2001); however, the effects are not
statistically significant if words occur less than six times, so a minimum requirement
for exposure frequency in yielding significant vocabulary acquisition is suggested in
Section 4.2.2.1. Apart from the suggestion for the minimum frequency requirement,
the consistency of learning percentage found between Rott (1999) and the present
study implies four ranges of learning probabilities for words occurring four and six
times in sufficient contexts for incidental vocabulary learning in Section 4.2.2.2.
Finally, the reasons why word gains are found less predictable in words occurring less
than six times are explained in Section 4.2.2.3.
4.2.2.1 The Minimum Exposures for Significant Word Acquisition
The findings in Section 4.1.3.1 suggest that learners need at least six encounters
of a word in reading to have significant vocabulary acquisition. Such findings that
repetitions around or more than six times are crucial for incidental vocabulary
learning confirm those of the previous investigations. For example, Nation’s (2001)
literature review states at least five or six repetitions are needed for vocabulary
learning to occur and Ghadirian (2002) notes “a significant increase in learning when
a new word was encountered five times and an even greater increase for six
encounters (p. 162)” from Saragi, Nation, and Meister (1978). In addition, Rott
(1999) reports six-exposure groups are significantly better than two-exposure and
four-exposure groups in the enhancement of vocabulary knowledge and Liu (2002)
finds learners acquire words repeated in the range from six to twelve times
significantly better than those from one to five times. However, the university
students in Rott’s study could demonstrate significant productive as well as receptive
word gains after only two exposures to an unknown word while the senior high school
freshmen in the present study need at least four exposures to start a significant
immediate growth of receptive word knowledge (see Table 9).
When words occur less than six times, learners in the present study have shown
difficulty acquiring them only from context, so there are some nonsignificant results
found between no-exposure words and five-or-fewer-exposure ones (see Table 9).
Such results are consistent with Hulstijn, Hollander, & Greidanus’ (1996) comparison
between one versus three word encounters in a single text. Hulstijn et al hypothesize
that the effect of exposure frequency will be limited if readers without any external
assistance in meaning ignore unknown words or infer their meanings unsuccessfully.
The hypothesis is supported in the present study when the number of exposures is
three in all participants and even when the numbers of exposures are five within each
group and four in the Inferential Group (see Table 9).
4.2.2.2 The Consistent Learning Percentage of Words Under Four and Six
Exposures Found Between the Present Study and Rott (1999)
When the mean score under each exposure frequency is converted into Knight’s
(1994) learning percentage (the difference between the exposure mean score and the
no-exposure
5mean score, divided by the number of target words), there is a
remarkable resemblance between participants’ test performance in the present study
5
The no-exposure mean score was assumed to be zero in each exposure frequency in
the present study.
and that in Rott (1999). Although Rott tested six target words in the format of
definition-selection, unlike the present research which only tested three target words
and employed a test format that required selections of target words to fit the sentence
context, the two studies achieve similar learning percentage in receptive knowledge.
As illustrated in Table 17, the average learning percentage of all participants on
immediate acquisition is about 46.67% under four exposures and 62% under six
exposures, which is close to the counterparts’ 44% and 61 % in Rott’s respectively
(see Table 18). When each group in the present study is discussed separately, the
Explicit Group is similar to Rott’s at 44% under four exposures but its performance
under six exposures is 6% lower than Rott’s, while the Inferential Group is 4.67% and
7.33% higher than Rott’s under the two kinds of exposures respectively.
In terms of the four-week delayed retention, the delayed posttest in Table 17
reveals all participants in the present study retain 37.67% of the target words under
four exposures and 48.67% under six exposures, which is a little higher than the
corresponding 32% and 48 % in Rott’s in Table 18. Except for the Explicit Group
under six exposures at 6.67% lower, both groups in the present study tend to
outperform Rott’s on the four-week delayed posttest, with the Explicit Group 2.33%
higher under four exposures and the Inferential Group 8.67% and 7.33% higher under
four and six exposures respectively.
Table 17
Percentage of Words Learned Incidentally in the Present Study
All Participants Explicit Group Inferential Group
Exposure Frequency