This chapter discusses the results of this study. After an overview of the study,
effects of different intervention patterns on vocabulary learning and reading
comprehension will be discussed, and then possible explanations for the learners’
questionnaire responses will be provided.
Study Overview
This study expanded previous gloss research by exploring the combined effects
of gloss plus inferencing or retrieval on EFL learners’ vocabulary learning and reading
comprehension. The investigation was based on the assumption that gloss plus
inferencing or retrieval could induce higher involvement load but reduce the chance
of memorizing incorrect word meaning. To test the assumption, three intervention
conditions for words repeated three times were set up: inference plus two glosses,
retrieval between two glosses, and full glossing. In the quantitative aspect, data of
the learners’ vocabulary learning and reading comprehension performance as well as
their preference and usefulness responses were collected and compared among
treatment conditions via MANOVAs, ANOVA, independent-samples t tests,
Chi-square tests and descriptive statistics. In the qualitative aspect, the participants’
think-aloud data were analyzed based on their reading strategies and sense-making
process.
Effects of Interventions on Vocabulary Learning
Results of the participants’ vocabulary learning performance showed that all of
the intervention conditions led to word learning. Since all the treatments shared
features of repeated word exposures and glossed reading, the findings reconfirm the
value of exposure frequency and glossing for vocabulary learning (e.g. Hulstijn et al.,
1996). Among all the three treatments, the condition that benefited vocabulary
learning the most was the gloss-retrieval-gloss condition, followed by the
inference-gloss-gloss condition and the least in the full glossing condition. The
difference between full glossing and gloss-retrieval-gloss reached a significant level.
Detailed discussions on each treatment condition are presented in the following
sections.
The Full Glossing Condition
The word gain from the full glossing condition supports that glosses increase
noticing (Bowles, 2004; Hulstijn, 2001; Miyasako, 2002; Rott, 2005; Yanguas, 2009),
stimulate correct responses (Holley & King, 1971) and trigger rehearsal (Holley &
King, 1971; Watanabe, 1997). It also agreed with Rott’s (2007) contention that
reappearing glosses could frequently draw learners’ attention to the target words and
offer them repeated opportunities of isolating the glossed word from the surrounding
context to rehearse word meaning in working memory. However, among the three
treatments, full glossing only generated the least word gain. Although Rott found
that words glossed more times yielded better word gain, this study discovered that it
was not always the case. The full glossing condition, containing the highest glossing
frequency in this study, did not outperform the other two conditions with lower
frequency and was even significantly inferior to the gloss-retrieval-gloss condition on
the immediate and delayed meaning recall tests and the immediate meaning
recognition test. When the target words are glossed all the time, the participants may
use glosses more as a means of understanding the text than learning vocabulary
(Bowles, 2004). In addition, from the GGG learners’ self-report of the gloss use, it
was found that as the times of gloss appearance increased, the number of the learners
checking the gloss directly decreased and more learners ignored the words together
with the glosses or thought about the meaning themselves without checking the gloss.
In other words, the continuous reappearance of glosses may not automatically lead to
learners’ repeated processing of glosses in the margin, let alone increase the quantity
of attentional resources spent on a word as mentioned in Rott (2007). Therefore,
although the involvement load hypothesis predicted that a gloss occurrence resulted in
an involvement load of one because it induced learners’ need for word meaning, the
load of reappearing glosses is not necessarily multiplied by the times glosses occur as
Rott assumed.
The Gloss-retrieval-gloss Condition
From the results, the gloss-retrieval-gloss condition was found to be the best
glossing arrangement for a word occurring three times in the input passage. It not
only produced the greatest word gain in form and meaning but also suffered the least
meaning attrition, which supports the contentions that retrieval can strengthen the link
between form and meaning (Nation, 2001), foster retention (Rott, 2007) and attenuate
the forgetting rate (Wheeler, Ewers, & Buonanno, 2003). As Rott (2007) suggested,
the involvement load of a retrieval was assumed to be three because it stimulated need
of understanding a word, search for its meaning in the mental lexicon and evaluation
of whether the searched meaning was correct in context.
In the GRG condition, the initial glossing supports the idea that the connection
between form and meaning had better be established initially (Nation, 2001), or
learners’ meaning mapping tends to remain shallow and vague (Rott & Williams,
2003). The subsequent retrieval might be the key that makes the
gloss-retrieval-gloss condition superior to the full glossing condition because what
contributes to incidental vocabulary learning is not repetition alone but the repeated
opportunity of retrieval (Karpicke & Roediger, 2008; Nation, 2001). Although Rott
(2007) argued that the subsequent retrieval of a word after the initial glossing did not
further strengthen the form-meaning connection because learners might just focus on
comprehension of a continued text rather than attend to any word aspect, it is possible
that learners in the present study were able to pay more attention to the bolded but
unglossed word during retrieval because they were required to retrieve word meaning
without referring back to the previous glossed word encounter. If there is no
retrieval between glossed encounters, learners will not try recalling or inferring
meaning from context, which demands deeper mental processing (Hulstijn, 1992;
Hermann, 2003; Knight, 1994; Mondria & Boer, 1991) or a higher involvement load
(Laufer & Hulstijn, 2001) and thus yields better retention. The third word
appearance with glossing could serve as a confirmation of the retrieval. Without the
final conformational glossing, problems such as forgetting or uncertainty during the
retrieval could only be left unsolved. If Rott had glossed the third word occurrence
in the glossed-retrieval-bolded-bolded treatment, the treatment involving retrieval
probably would have significantly outperformed the four-gloss treatment.
The Inference-gloss-gloss Condition
Inference-gloss-gloss was the second best intervention condition found in this
study. The condition was not significantly better than the full glossing condition, but
it generated greater word learning than the full glossing condition by more than 10%
in immediate form and meaning recall and more than 8% in meaning recognition.
The better immediate word gain in the IGG condition may be linked to the search and
evaluation processes during inferencing that created higher involvement load (Laufer
& Hulstijn, 2001). Although the retention rate in the IGG condition was higher than
that in the full glossing condition as well, the forgetting rates of the former were not
lower than the latter probably because the greater amount of immediate word learning
might be subject to attrition more easily.
The effect of inference-gloss-gloss was also comparable to that of the best
condition, gloss-retrieval-gloss. In fact, the think-aloud analyses showed that
retrieval and inferencing exhibited in the current study had one feature in common.
Either during retrieval or during inferencing, the learners might rely on the context to
evaluate possible meanings for a word. The major difference between the two
conditions was that the former was primed with a glossed word encounter and more
likely to lead to a specific correct response while the latter required a learner to draw
on wider knowledge sources to derive meanings that tended to be vague and
error-prone even after repetitive hypothesis testing. That was why the rate of correct
retrieval was higher than that of correct inference. If the context for the first word
encounters had not been designed to provide enough clues, the rate of correct
inference could have been lower. The low rate of correct inference might limit word
learning in the IGG condition because it was found that correct inference significantly
yielded better form and meaning recall than incorrect inference in this condition.
When wrong inferences are made, as Hulstijn (1992) mentioned, the problem is
that learners need to unlearn incorrect inferences. In fact, not all incorrect inferences
can be easily corrected. Chances are that they can be incompletely corrected,
confused with correct glosses, fossilized in memory, etc. Analyses of think-aloud
data also showed that the IGG learners tended to express uncertainty about their
inferences by using more hedges before their guesses such as kě néng shì18 (maybe)
and providing several possible meanings for a target word. Such uncertainty might
prevent them from linking form and meaning in the memory. Therefore, although
inferencing can induce need, search and evaluation and its involvement load may not
be lower than retrieving, its error-prone and uncertain nature may reduce its effect on
initial form-meaning establishment.
In addition, inferencing has the following disadvantages. First, learners may
pay more attention to the association of context and meaning than the connection
between form and meaning (Mondria, 2003). Second, when the inferences are
correct, learners may think that they already know the words and thus spend less time
committing the words to memory or even will not bother remembering word meaning
as they have fulfilled the need of comprehension (Hulstijn et al., 1996; Mondria, 2003;
Sokman, 1997). Third, learners’ form-meaning mapping may not be stable because
they tend to only have a vague understanding of the target words during inferencing
and are more likely to skip the confirmation or disconfirmation of their guesses (Rott
& Williams, 2003). When an inferred meaning seems plausible in readers’ discourse
model, they may be satisfied with the meaning assignment and transfer the inferred
meaning to the next word encounter (Rott, 2005). Therefore, although inferencing
takes up high involvement load, it does not always result in better retention.
To improve the inferencing method, a follow-up verification can be added.
Fraser (1999) found that consulting a dictionary following inferencing increased word
meaning recall from either inferencing or dictionary consultation alone by 20% and
explained that verification following inferencing not only offset the problem of
making wrong inferences but also led to a greater depth of processing. A similar
reinforcing effect of verification was also observed in direct vocabulary teaching after
reading in Sonbul and Schmitt (2010) and the meaning-inferred method designed in
Mondria (2003). Although Mondria worried that verification such as word list
checking was quite time-consuming, the current design that simply provided
participants with glosses after inferencing could shorten the time needed for
dictionary consultation, instruction or tedious word list checking.
In short, although glossing facilitates vocabulary learning, repeating it every time
does not increase learning as much as leaving one word encounter unglossed for
inferencing or retrieval. Making the first word encounter unglossed for inferencing
and glossing the subsequent encounters may enhance learning but its effect is not as
obvious and stable as inserting a retrieval practice between glossed encounters.
Inference making tends to be error-prone, create a vague link between form and
meaning initially and make learners believe that they have acquired the words, while
retrieving is more likely to strengthen memory that has been established in the
previous glossed encounter. To maximize word learning, a sequence of initial
form-meaning connection, subsequent retrieval and final confirmation can be
followed.
Effects of Interventions on Reading Comprehension
Aside from vocabulary learning, the current study aimed to investigate whether
the treatment conditions fostered or hindered text comprehension. Following the
line of previous findings that glossing enhanced the comprehension of the text (e.g.
Yanguas, 2009), the present investigation further compared the effects of repeated
glossing and the combination of glossing with inferencing or retrieval on text
comprehension. The result showed that the difference in the treatments did not
constitute a significant factor in the learners’ comprehension performance. The
nonsignificant result could be attributed to the test method. This study designed a
choice test to measure the participants’ reading comprehension. Ko (2005) indicated
that there were pros and cons in every test format. Recall protocols are not suitable
for students who cannot express their thoughts accurately but are able to understand
the text perfectly. Results of cloze tests may be confounded with the influence of
contextual clues, the blank number and other decisions on the test format. Choice
tests may fail to detect readers’ comprehension accurately and consistently. Since
there is no perfect test method, the measurement choice should be taken into
consideration in the interpretation of results.
Although no significant treatment effect was found on reading comprehension,
there was a tendency that full glossing facilitated reading comprehension better. The
tendency supported Gettys et al.’s (2001) view that glosses could aid second language
reading by lessening learners’ burden of word-by-word decoding and Rott’s (2007)
argument that repeated glossing facilitated comprehension of the target words and the
entire proposition. In addition, the presence of glosses increased the likelihood of
correctly comprehending the propositions. The reason why the error-prone treatment
conditions involving inferencing or retrieval in one word encounter did not interfere
with comprehension significantly is probably that the readers were assisted with
glosses in the other two word encounters. To maximize learners’ attentional quality
on word learning but minimize misunderstanding in comprehension, provision of
glosses after obtrusive interventions such as inferencing or retrieval is advisable.
Strategy analyses also supported the use of glossing in facilitating
comprehension. During the unglossed word encounter, the learners in the IGG and
GRG conditions seemed to be so occupied with solving local word problems that they
exhibited little global reading. The observation is consistent with Ko (2005) that
glossing fosters the use of high-level strategy use. Although the use of glossing only
induces low involvement with the target vocabulary items (Laufer & Hulstijn, 2001),
it may free up learners’ mental resources for high involvement with the text.
The finding that there was no significant correlation between word gain and
reading comprehension seems to suggest that vocabulary learning and text
comprehension may operate on dissimilar mechanisms. As Rott (2007) reviewed,
word learning involves noticing the gap, isolating word from context, and
elaboratively associating form and meaning to make form-meaning connections in the
mental lexicon, but in comprehension, what matters is whether readers can rapidly
integrate the information existing in the text and possessed by themselves. Although
Rott also speculated that the retrieval task might take up attentional resources of text
comprehension, her speculation was not consistent with the nonsignificant correlation
found in the GRG condition. The inconsistent findings between Rott’s study and
this study might lie in the difference in the adopted test methods. Rott employed an
L1 recall task but this study used a choice test to measure comprehension. The recall
task has been criticized for involving a test of memory retrieval as well (Chang, 2006).
That could partially explain why Rott found that the learners recalled the least main
ideas under the retrieval task that was favorable to word learning. In her case, there
might have been competition for readers’ mental resources between text memory and
word memory rather than between comprehension and word storage.
Although no apparent causal relationship was observed between reading
comprehension and word learning in the current investigation, the word gain in the
IGG condition positively correlated with comprehension to a noticeable degree. The
positive correlation that almost reached a significant level echoes Rott’s (2000)
observation that the learners’ strategic approaches to word inferencing and reading
seemed related. In other words, the IGG condition could have the potential to bring
twofold benefits to word learning and reading comprehension. Nevertheless, the
IGG condition produced the least reading comprehension score. One possible
explanation for the findings is that although the word inferencing strategy may foster
reading comprehension, the error-prone inferences may offset the effect of inferencing
on comprehension. If more assistance such as meaning options or other instances of
a target word in context selected from the concordance data can be provided for the
IGG learners during their inferencing, the IGG condition may be an ideal condition
that can benefit vocabulary learning and reading comprehension greatly at the same
time.
As for the trade-off issue between vocabulary learning and reading
comprehension, no definite conclusion could be made from this study. Although
correlation analyses showed no significant correlation between vocabulary learning
and reading comprehension in each treatment, the concern for the trade-off issue
cannot be ruled out completely due to the following reasons. To begin with, the
comprehension score was found to negatively correlate with form recall and meaning
recognition to a mild degree in the GRG condition. Another possible trade-off could
exist in the full glossing condition as well. Although correlation analyses showed
the relationship between word gain and reading comprehension in the condition
almost reached a neutral state, the learners in this condition exhibited a contrast in
their average vocabulary and reading performance because they obtained the least
word gain but their comprehension score was the highest among all the groups.
Since no further solid evidence can be found in the present investigation, whether a
trade-off exists between vocabulary learning and reading comprehension is still
inconclusive.
In summary, in contrast to vocabulary learning, the treatment effect on reading
comprehension is not significant. Interpretations of the nonsignificant effect cannot
exclude the consideration of the test format. Although no definite conclusion can be
made from the statistical analyses, some tendencies can be observed from the results.
First of all, comprehension seems to be better facilitated in the full glossing condition
because glossing is more likely to lead to correct interpretation and enable readers to
have more mental resources saved from less word-solving for high-level
comprehension. Second, the nonsignificant correlations between word gain and
reading comprehension may imply that vocabulary learning operates on different
mechanisms from comprehension and that no obvious trade-off effect exists between
them. However, some evidence still suggests that there is a relationship between
vocabulary learning and reading comprehension. A positive correlation between
learning and comprehension that almost reached a significant level was noticed in the
IGG condition, so inferencing may play a role in both learning and comprehension.
Nevertheless, the condition also resulted in the poorest comprehension, so the effect
of inferencing may be offset by wrong inferences. In addition, a possible existence
of a trade-off between learning and comprehension cannot be completely eliminated
because the word gain in the GRG condition correlated with comprehension
somewhat negatively and the GGG condition, which yielded the least word learning,
turned out to result in the best comprehension.
Preference and Perceived Usefulness
The present study found that the learners had a positive attitude toward glossed
reading, which is consistent with previous findings (Bell & LeBlanc, 2000; Cheng &
Good, 2009; Jacobs, 1994; Jacobs et al., 1994; Ko, 2005). Unexpectedly, among the
three conditions involving glossing, full glossing was less preferred than glossing
three conditions involving glossing, full glossing was less preferred than glossing