SSLA, 21, 225–241. Printed in the United States of America.

LEXICAL PROCESSING
STRATEGY USE AND
VOCABULARY LEARNING
THROUGH READING

Carol A. Fraser
York University

This article reports on a strategy training study that investigated the

lexical processing strategies (LPSs; ignore, consult, infer) used by
L2 learners when they encounter unfamiliar vocabulary while reading
and the impact of these strategies on vocabulary learning. A time-
series with repeated-measures design was used. Introspective data
were gathered from eight participants (Francophone university stu-
dents, intermediate ESL proficiency) on eight texts over 5 months to
elicit LPS use on self-identified unfamiliar words. Then, 1 week after
each reading, participants completed a cued recall task to measure
their learning of these words. An analysis of overall LPS use (changes
in patterns and effectiveness of strategy use with LPS-focused instruc-
tion) and word retention rates demonstrates the potential for vocabu-
lary learning through reading and indicates that some LPSs lead
to higher retention rates than others. This research increases our
understanding of the role of LPS use in vocabulary learning and
suggests some reevaluation of current pedagogic practice.

Although it is generally acknowledged in theory and research that some in-

cidental L2 vocabulary learning occurs in the course of reading for compre-
hension, there is less consensus as to whether reading is a rich event for

I gratefully acknowledge support for this research from Educational Testing Services in the form of
copies of the TOEFL Institutional Test and Glendon College, York University in the form of a Minor
Research Grant. I also thank Esther Geva and Birgit Harley, who provided constructive feedback on
an early draft of this paper. An earlier version of this paper was presented at the 11th World Con-
gress of Applied Linguistics (AILA), Jyväskylä, Finland, August 1996.
Address correspondence to Carol Fraser, Glendon College, York University, English Department,
2275 Bayview Avenue, Toronto, Ontario, Canada M4N 3M6; e-mail: [email protected].

 1999 Cambridge University Press 0272-2631/99 $9.50 225

226 Carol A. Fraser

vocabulary learning. A variety of studies have found evidence of vocabulary

growth (e.g., Huckin & Bloch, 1993; Hulstijn, 1992; Paribakht & Wesche, 1993,
1997; Pitts, White, & Krashen, 1989), but views differ as to whether the demon-
strated retention rates represent low (Hulstijn, 1992) or high (Krashen, 1989)
learning outcomes.
Little is understood regarding how this incidental learning occurs, par-
ticularly with respect to the cognitive and metacognitive processes learners
engage in when encountering unfamiliar words while reading and how differ-
ences in these processes affect vocabulary learning. The underlying assump-
tion has been that incidental vocabulary learning primarily occurs through the
process of inferring word meaning. But some research suggests that L2 learn-
ers who are left on their own generally ignore unfamiliar words, infer only
when there is a specific need, and consult sparingly and on a very selective
basis (Bensoussan & Laufer, 1984; Paribakht & Wesche, 1997). Because atten-
tion to an unknown word seems to be a prerequisite for any learning to occur
(Ellis, 1994; Gass, 1988; Schmidt, 1994), high rates of ignoring would thus se-
verely limit the learning potential.
Furthermore, there is a lack of consensus regarding the benefit for L2 learn-
ers of the use of lexical inference while reading. On the one hand, some
researchers have argued that inferring word meanings is potentially a produc-
tive strategy for vocabulary learning. First, it is a communicative event in
which learners engage in considerable hypothesis formation and testing about
word meaning (Ellis, 1994; Haastrup, 1989, 1991) and, second, the rich psycho-
logical and linguistic context that text provides can act as a cognitive hook for
the memory of new words (Schouten-van Parreren, 1989). However, research-
ers (Bensoussan & Laufer, 1984; Haastrup, 1989, 1991; Haynes, 1984; Hulstijn,
1992) have also found that inferring is not always an easy or efficient strategy
for L2 learners to use either because of text complexity (i.e., the text may not
provide sufficient clues to infer) or because of reader limitations (i.e., L2
learners often do not make use of available cues or misuse them). Evidently,
lexical inference is neither an efficient nor an effective vocabulary-learning
strategy when L2 readers fail in their attempts to infer meaning for new words
or when the meaning determined is inaccurate.
This article reports on a study of the lexical processing strategies that L2
learners use while reading for comprehension. Lexical processing strategies
(LPSs) refer to the three strategic options an L2 reader has when confronting
an unfamiliar word: ignore and continue reading, consult a dictionary or an-
other individual, or infer word meaning on the basis of linguistic and contex-
tual cues. Within the theoretical framework of classroom-based research, the
goal of this study was, first, to describe the LPSs that L2 learners use while
reading and, second, to trace the impact of LPS-focused instruction on LPS
use and other relevant variables including reading rate, reading comprehen-
sion, and vocabulary learning. This paper presents results that address the
issue of vocabulary learning and the role of lexical processing strategies. Spe-
Lexical Processing Strategy Use 227

cifically, (a) Are the LPSs adult L2 learners use potentially productive for vo-
cabulary learning? (b) Does LPS-focused instruction affect word learning? and
(c) Are different word retention rates associated with reading proficiency, LPS
use, or previous familiarity with the word?

DESIGN OF THE STUDY

Using a quasi-experimental, interrupted, time series with repeated-measures
design, introspective and performance data were gathered individually from
eight participants. All were Francophone university students enrolled in an in-
termediate level ESL course, “Academic Foundations,” taught by the principal
researcher.
There were two phases to the instructional treatment: metacognitive strat-
egy training, followed by focused language instruction. Both phases were inte-
grated into the regular content of the English for Academic Purposes (EAP)
course involving 19 students. Each phase consisted of approximately 8 hours
of directed instruction (4 in and 4 out of class) given over 1 month. Phase 1,
metacognitive strategy training, focused on developing student awareness of
the use and viability of the three LPS options. Overall, a direct instructional
approach to strategy training was followed. This approach consisted of an ex-
plicit presentation of the LPSs (what the strategy is; why it is useful; and
when, where, and how to use it), guided practice applying the strategy
(teacher-student modeling and activities to apply outside class), and discus-
sion of the results of the trial application (the effectiveness and efficacy of
strategy use and problems encountered). In phase 2, instruction focused on
building up the kinds of language knowledge thought to underlie the ability to
use the LPSs—effective inferencing procedures, in particular. It included in-
struction on cognates, word structure (morphological word stems, prefixes,
and suffixes), grammatical function (noun, verb, adjective, adverb, and con-
junction), lexical cohesion (reiteration and collocation), and structural redun-
dancy (appositive constructions, intersentential relations such as defining
relative and adverbial clauses, and summary and rephrasing discourse mark-
ers). The presentation and practice of language items was contextualized and
integrated into course reading tasks and focused on how learners could use
this language information to determine the meaning of unfamiliar words.
The eight volunteer participants1 for the individual data collection sessions
were selected to represent higher and lower levels of English reading profi-
ciency on the basis of their results on the Vocabulary Levels Test (Nation,
1990) and a subsection of the Institutional TOEFL, Section 3, Vocabulary and
Reading Comprehension (Educational Testing Service, 1992). These two tests
were given to all students in the research class (N = 19). On the Vocabulary
Levels Test, the class mean was 61.4 (maximum score 90; range 43–82; SD 9.9).
An analysis of class means for words at different frequency levels, according
to test guidelines, indicated that vocabulary knowledge at the 2,000- and
228 Carol A. Fraser

3,000-word frequency levels was acceptable but that this ESL group needed to
develop more knowledge of words at the 5,000- and 10,000-word frequency lev-
els. On the Vocabulary and Reading Comprehension Section of the TOEFL, the
class mean was 46.4 (maximum score 67; range 39–53; SD 4.4). According to
published performance references (Educational Testing Service, 1993), a mean
of 46.4 for an undergraduate student is equivalent to a total scaled TOEFL
score of about 490. Class results on the two tests were highly correlated (r =
.79). A scatter plot was used to differentiate the performance of participants
on the two measures. Whereas the higher performance group scored above
the mean on the Vocabulary Levels Test and above the score of 50 on the
TOEFL (Section 3), those in the lower group scored below the class mean on
the Vocabulary Levels Test and below 50 on the TOEFL (Section 3).
Participants met individually with the researcher nine times over 5 months
for one training and eight data-collection sessions. These meetings repre-
sented four measurement periods: baseline, after metacognitive strategy train-
ing, after language-focused instruction, and a delayed measure given 1 month
after the end of the instructional treatment to assess maintenance. In each
data-collection session, participants first studied comprehension questions,
read an article, answered comprehension questions, and skimmed the article
to identify unknown words. They then engaged in an oral interview that fo-
cused on eliciting a retrospective think-aloud protocol of the LPSs they had
used to deal with unfamiliar vocabulary while reading. A structured interview
format (employing the same questions, in the same order, to all participants,
for all texts) was used to increase the stability of the data in the repeated-
measures design. All probes began with the general question “What did you
do and think about when you first saw [the word] ‘x’?” followed by one to
three more nondirective questions depending on the indicated LPS option.
The texts were selected to be challenging; all came from the Science & Tech-
nology section of The Economist and were 1,000–1,200 words long with a read-
ability range of 9.4–12.3.2 Both a bilingual and an English dictionary were
available for consultation.
Finally, 1 week after each reading, participants completed a cued recall task
to monitor the effect of instruction on word learning. It consisted of 10 words
selected from those each participant had indicated to be unknown and that
had been focused upon in the retrospective think-aloud interview. Words were
selected both to reflect a participant’s frequency of use of the various LPS
options and to ensure a balance of two levels of unknown words (i.e., either
previously seen but not recalled or never seen before). Participants were
asked to indicate their level of knowledge of the words using the 5-point Vo-
cabulary Knowledge Scale (VKS, Paribakht & Wesche, 1993; Wesche & Pari-
bakht, 1996). In this cued recall task, participants are asked to indicate their
knowledge of words from (a) no familiarity to (b) recognition but no recall of
meaning to (c) some recall (synonym or translation) to (d) use of the word in
a sentence.
Lexical Processing Strategy Use 229

Table 1. Frequency distribution of use of

lexical processing strategies
LPS option Frequency Percent

Consult 330 29
Ignore 269 24
Infer 491 44
No attention 37 3

Total 1127 100

Note. “No attention” refers to the think-aloud protocols; partici-
pants occasionally indicated that they did not notice the unfamiliar
word item while reading. The total is based on 878 unfamiliar word
encounters.

LEXICAL PROCESSING STRATEGY USE

Data Analysis

Overall, 878 unknown words were coded and analyzed from the think-aloud
protocol data (a maximum of 15 words per participant per text). Initially, fre-
quency distributions based on total word encounters were examined to deter-
mine overall patterns of LPS use and the knowledge resources associated with
those patterns. Subsequently, after confirmation that the distributions were
normal using the K-S (Lilliefors) statistic, a multivariate repeated-measures
analysis of variance was carried out on each of the variables for LPS use and
success to investigate reading proficiency and instructional treatment effects.
This analysis was based on each participant’s rate of use of the different op-
tions and percentage of success at each time period. Additionally, orthogonal
contrasts were used to break down the time effect into linear, quadratic, and
cubic effects.3

Results

LPS Use. Table 1 presents the overall use of the various lexical processing
strategies. Participants used the three LPS options both alone and in combina-
tion with each other (e.g., infer and consult), which accounts for the 1,127 re-
sponses elicited on 878 word encounters. Furthermore, in 37 cases (3%),
participants reported that they had not noticed the unknown word while read-
ing; these items were not analyzed further. Of the 841 remaining cases of
actual strategy use, 71% represent single-LPS use and 29% represent multiple-
LPS use. As Table 1 indicates, overall, these participants inferred more fre-
quently than they ignored or consulted.
Further examination of the frequency distributions associated with LPS use
230 Carol A. Fraser

Figure 1. Mean rate of use of the three LPS options (consult, ignore, infer)
by L2 readers (N = 8) over the four measurement periods. Rate of use is per
100 words. Rates do not add up to 100 because participants sometimes used
LPS options in combination.

revealed that inferencing was the preferred as well as the primary strategy
used by these participants. When the frequency of use presented in Table 1 is
transformed into a rate of use based on the total number of words (N = 841)
rather than on instance of LPS use (N = 1127), the outcome is that these learn-
ers inferred in 58% of the cases in which they encountered an unfamiliar word.
Moreover, over half of these inferences occurred alone and, in contexts where
inferencing was combined with other strategies, it was the first LPS option al-
most all of the time (96%). In contrast, they consulted a dictionary 39% of the
time—about half the time alone (55%) and half the time following an inference
(45%)—and they ignored unfamiliar words 32% of the time, generally alone
(62%) but sometimes following an attempted inference (35%).
Furthermore, this overall hierarchy of LPS use was maintained over time at
all measures (Figure 1). That is, these L2 readers inferred more than they con-
sulted and consulted more than they ignored. Although there were fluctua-
tions and an overall slight increase in the rates of inferencing and consulting,
the rate of ignoring decreased steadily. The repeated-measures ANOVA indi-
cated that this decreasing linear trend for ignoring was significant, F(3, 4) =
12.97, p < .05. There were no significant time effects for either inferencing or
consulting.
Effectiveness of LPS Use. Each instance of LPS use in the protocol data
was also rated in terms of how successful it was in determining a meaning for
the word that was appropriate for text comprehension. A 3-point scale was
used: no comprehension (i.e., no or an inappropriate meaning was deter-
mined), partial comprehension (i.e., the meaning determined worked generally
for the text context although there was some distortion or loss of the text
Lexical Processing Strategy Use 231

Figure 2. Mean success rates associated with the use of the LPS options,
consult and infer, to determine appropriate word meaning for text comprehen-
sion over the four measurement periods.

representation, e.g., corpses as “bodies” instead of “dead bodies”), and com-

prehension (i.e., the meaning determined was appropriate for the text context
with little or no meaning distortion). Each participant’s rate of success when
consulting and inferencing was computed by awarding a score of 0 for no com-
prehension, .5 for partial comprehension, and 1 for full comprehension.
An examination of the overall frequency distributions indicated that, when
participants did consult or infer, they were generally successful; 78% of con-
sults and 52% of inferences resulted in full comprehension and another 20%
of inferences resulted in partial comprehension (i.e., the meaning determined
worked adequately for text comprehension). Figure 2 further demonstrates
that over all four time periods these participants were more successful in de-
termining unfamiliar word meaning when consulting than when inferencing.
More importantly, Figure 2 also illustrates a sharp increase in success when
inferencing after the onset of metacognitive strategy training. The repeated-
measures ANOVA indicated that this was a significant time effect, F(3, 4) =
7.80, p < .05, which was both linear and quadratic.
Type and Effectiveness of Inference Processes Used. Additionally, all in-
ferences were characterized according to the type of process involved. A dis-
tinction between word identification and sense creation processes previously
made in word recognition research (Clark & Gerrig, 1983; Kintsch & Mross,
1985; Swinney 1979) was adopted. Inferencing through word identification is
characterized as a fast, automatic, data-driven process in which the form of
the unfamiliar word activates an L1 or L2 association in the learner’s mental
lexicon (e.g., stalking–talking; inherit–hériter and inhérant). The thematic con-
text of the text plays a minimal role in word identification inferences. In the
232 Carol A. Fraser

protocol data, such associations based on the phonological or orthographic

form of the word seemed to “jump off the page” with the reader exerting little
control over the association, as illustrated in these examples:

(1) T: What did you do and think about when you first saw “snapshots”?
S: Term of hockey.
T: What does it mean in hockey?
S: Just the snapshot, when someone shot in the goal. But here I didn’t know the
meaning so. . . .
(2) T: What did you do and think about when you first saw “failure”?
S: I thought this is a kind of mistake or . . . It sounds like the same in French “faillir
une émission” but I can’t be sure.

In contrast, inferencing through sense creation is a context-centered, more

deliberate and effortful process whereby meaning is created on the basis of
language and situational cues from the text. As examples (3) and (4) illustrate,
learners typically utilized both grammatical and situational cues from the im-
mediate sentence context.

(3) T: What did you do and think about when you first saw “rotten”?
S: I knew that was a symptom of morning sickness, I know there was one was vomit-
ing, and this one was probably, nausea or dizzy, so . . . I didn’t . . . really find the
real meaning because I knew it was a symptom.
(4) T: What did you do and think about when you first saw “soothe”?
S: It’s the first time I see this word but I think it’s the action to feed a baby.
T: How do you know?
S: Breast milk contains enough to feed a baby for a few hours so . . .

An examination of the frequency distributions of LPS success by inference

process type (Table 2) indicates that sense creation processes were used
most frequently—that is, 65% of the time—and quite effectively, in that com-
prehension was achieved 78% of the time. Furthermore, even though L1 and
L2 word identification processes were used equally (17% and 18%, respec-
tively), these Francophone participants had more success using their L1 (77%
success) compared to their L2 (44% success).
Figure 3 illustrates the use of inference process types over time and indi-
cates that there was a steady increase in the proportion of inferences based
on sense creation processes. In other words, these readers increasingly gener-
ated word meaning on the basis of linguistic and situational elements in the
text rather than associating the unknown word with a phonologically or ortho-
graphically similar word in their L1 or L2 lexicons. Whereas the repeated-mea-
sures ANOVA indicated that this effect was approaching significance, F(3, 4) =
4.14, p < .10, the univariate trend analysis indicated a significant linear trend,
F(1, 6) = 15.79, p < .01. Over the course of the study, these L2 readers gradu-
ally increased the proportion of inferences based on sense creation processes
but not to a point that reached statistical significance.
Lexical Processing Strategy Use 233

Table 2. Frequency distribution of LPS success by inference process type

Inference process type

L1 word L2 word
LPS success identification identification Sense creation All

No comprehension 23% 56% 22% 28%

Partial 17% 9% 23% 20%
Complete 60% 35% 55% 52%

Total 100% (n = 83)a 100% (n = 87) 100% (n = 316) 100% (n = 486)

% use of process type 17%b 18% 65% 100%
Note. The values represent percent of success in determining an appropriate meaning when inferring. No comprehen-
sion = no or inappropriate meaning determined; Partial = meaning determined worked generally for the text context
with some loss or distortion; Complete = meaning determined was appropriate with little or no distortion. L1 word
identification = inferring on the basis of L1 word association; L2 word identification = inferring on the basis of L2
word association; Sense creation = using linguistic and situational context to infer.
a
Counts are in parentheses. bThese values represent percentage of use of each of the inference process types.

Figure 3. Mean rate of use of inference process types over the four measure-
ment periods. L1 Word Id = use of L1 word association to infer; L2 Word Id =
use of L2 word association to infer; Sense Creation = use of linguistic and situ-
ational context to infer.

Furthermore, if we examine the rate of success these learners had when

using the various inference processes over time (Figure 4), we see that they
became increasingly successful when inferencing with sense creation pro-
cesses. The repeated-measures ANOVA indicated that this time effect was ap-
proaching significance, F(3, 4) = 4.42, p < .10. Also, this upward trend was both
234 Carol A. Fraser

Figure 4. Mean success rates associated with the use of the various infer-
ence process types in determining appropriate word meaning for text compre-
hension over the four measurement periods.

linear, F(1, 6) = 12.54, p < .05, and quadratic, F(1, 6) = 11.99, p < .05, and was
characterized by two distinct increases, one after the onset of metacognitive
strategy training and one after the onset of language-focused instruction. Fig-
ure 4 also demonstrates that these Francophone learners had more success
using their L1 than their L2 to determine word meaning. However, although
the success rates using L2-based word-identification processes were quite sta-
ble, the success rates using L1-based word-identification processes fluctuated
dramatically.

LEXICAL PROCESSING STRATEGY USE

AND VOCABULARY LEARNING

Data Analysis
The VKS, which was the format used in the cued recall task to monitor the
effect of LPS instruction on vocabulary learning, uses a 5-point scale to distin-
guish different levels of a learner’s knowledge of a word: 1 (not familiar), 2
(familiar but meaning is not known), 3 (correct synonym or translation is
given), 4 (word is used with semantic appropriateness in a sentence), and 5
(word is used with grammatical and semantic appropriateness in a sentence).
Initially, all items were scored according to the VKS. However, because of the
low frequency of use of the categories distinguishing intermediate levels of
knowledge for analysis (i.e., categories 1, 4, and 5), the 5-point VKS score was
collapsed to a 2-point score (meaning recalled vs. not recalled). Additionally,
because the goal of this word learning measure was to track retention associ-
ated with LPS use, the rating of participants’ learning of a word was based on
Lexical Processing Strategy Use 235

Figure 5. Percentage of words recalled according to LPS use by L2 readers

(N = 8).

the meaning determined by participants in their initial LPS use and not a stan-
dard meaning of the word. For example, through LPS use, one participant de-
termined that sights meant taille (“size”) and, subsequently, in the cued recall
task supplied the translation taille to the prompt sights; when scored, she was
awarded a score of 3 (correct synonym or translation is given) even though
the meaning determined was incorrect.
A total of 622 cued recall items were analyzed.4 The word-learning score for
each time period represents the percentage of demonstrated retention
of words over two texts (i.e., 20 words). A multivariate repeated-measures
ANOVA was carried out to examine the effects of instruction and reading profi-
ciency level on word learning scores. Additionally, frequency distributions
were analyzed to examine patterns of word retention associated with LPS use,
inference process type, and level of familiarity with the word.

Results

Word-learning scores for the four time periods had a range of 5–50% and an
overall mean of 28% (SD = 12); individual participant means ranged from 13%
to 40%. Results of the repeated-measures ANOVA indicated no significant main
effects for group (i.e., L2 reading proficiency) or time (i.e., instruction), and
no group-by-time interaction effect.
The analysis of the frequency distributions of word-learning scores associ-
ated with LPS option, inference process type, and level of word familiarity in-
dicated the following patterns. First (see Figure 5), when participants
consulted or inferred alone, they recalled the word meaning they had deter-
236 Carol A. Fraser

Figure 6. Percentage of words recalled according to type of inferencing pro-

cess by L2 readers (N = 8).

mined about 30% of the time (30% and 31%, respectively). However, when
they inferred and then consulted, their recall increased to 50%.
Second (see Figure 6), when participants used L1-based word-identification
processes to infer, they had a mean retention rate of 50%, which decreased
to 39% for L2-based word-identification inferences and 28% for sense creation
inferences.
Third (Figure 7), word retention rates were higher for words that partici-
pants initially indicated they had previously seen but could not recall (42%)
compared to words they had not seen before (25%).

DISCUSSION AND CONCLUSION

This article has reported on results from a study of the lexical processing
strategies L2 readers use while reading. The results support the view that
reading for comprehension in an L2 can be productive for incidental vocabu-
lary learning. When dealing with unfamiliar vocabulary, these L2 learners used
LPSs that were productive for word learning (i.e., consulting and inferencing)
more frequently than unproductive ones (i.e., ignoring or not paying atten-
tion). They were also generally successful in determining a meaning that was
adequate for text comprehension when both consulting and inferencing.
Moreover, when inferencing, they accessed sense creation processes that
made use of cues from the text context more frequently than word-identifica-
tion processes that relied on L1 or L2 form-based associations. Previous re-
search in L2 inferencing (Bensoussan & Laufer, 1984; Haynes, 1984; Huckin &
Lexical Processing Strategy Use 237

Figure 7. Percentage of words recalled according to level of previous famil-

iarity by L2 readers (N = 8). The level of familiarity represents participants’
perception of their knowledge: either never seen before or previously seen
but not recalled.

Bloch, 1993) has indicated that inferences based on word form associations
rather than the use of cues from the text context often lead to inappropriate
word-meaning determination. The potential for L1 cognate associations to be
especially misleading was highlighted in this study by the dramatically fluctu-
ating success rates of the learners when inferencing using L1-based word-iden-
tification processes.
Although instruction focusing on enhancing lexical processing strategy use
did not have a direct effect on vocabulary learning, there was evidence of an
indirect effect. First, the rate of ignoring words decreased with instruction,
and success when making inferences increased. This increase in the effective-
ness of the inferencing strategy was related to steady increases in both the
rate of use and percentage of success these L2 readers had when inferencing
using sense creation processes. In other words, not only did these learners
increase the number of inferences that were based on the linguistic and situa-
tional context of the text, but these inferences became increasingly appro-
priate. The LPS-focused instruction thus seemed to improve the underlying
conditions needed for effective vocabulary learning while reading: More unfa-
miliar words were attended to, more extensive and elaborative processing
was undertaken, and more appropriate meanings were determined.
With respect to vocabulary learning, substantial individual differences were
indicated by the wide range of word retention scores across participants
238 Carol A. Fraser

(5–50%) and the high standard deviation (12). Differences in L2 reading pro-
ficiency level did not affect the retention of the meaning of new words encoun-
tered while reading. Sanaoui (1992) reported a similar finding in a study that
investigated the vocabulary learning of adult L2 (French) learners at different
language-proficiency levels.
Word-retention scores on the cued recall task indicate that these learners
did acquire vocabulary while reading for meaning.5 Furthermore, the mean re-
tention rate of 28% is considerable in the light of previous research. For exam-
ple, in the context of a single encounter with a new word while reading, with
no prompt to infer, both Krashen (1989) and Hulstijn (1992) reported reten-
tion rates of about 7%. In the context of multiple encounters with target words
with no prompt to infer, Paribakht and Wesche (1997) reported a retention
rate of 11% for verbs and 24% for nouns, and, in a single reading in which the
inferring process was prompted and guided, Hulstijn (1992) reported retention
rates of 13% and 17%.
The higher retention rates in this study could relate to several factors.
First, the words tested for retention were self-selected by the learners in the
sense that they initially identified these words as unfamiliar. These words
thus represented a corpus of unknown words that were both noticed and po-
tentially highly salient for these learners (e.g., partially known words, L1 cog-
nates, and words encountered multiple times). Additionally, the retrospective
think-aloud interview may in itself have constituted an added word-learning
opportunity. Minimally, participants’ attention was focused on the ortho-
graphic and phonological form of each word, but participants also retrieved,
rehearsed, and elaborated earlier thought processes on the semantic and syn-
tactic features of the words. This factor could have allowed for more explicit
learning involving a higher level of awareness than might normally occur in
reading and thus may have enhanced retention (Ellis, 1994; Pressley & Affler-
bach, 1995).
Finally, this study found evidence that some LPSs were associated with
higher retention rates than others. Predictably, these learners had higher re-
tention rates when an inference was based on an already existing L1 or L2
word association. There was already a cognitive hook to hang the new item
on. This facilitating effect of existing memory structures was also seen in the
higher retention rates for words with which learners had some previous famil-
iarity compared to those reported as never before seen. This supports the
view that vocabulary is learned incrementally over multiple encounters (Nagy,
Herman, & Anderson, 1985).
Furthermore, when these learners inferred and then consulted, they had a
higher retention rate than if they inferred or consulted alone. Schouten-van
Parreren (1989) reported a similar finding in her case study of adult L2 learn-
ers. Inferring unknown word meaning and then consulting a dictionary for ver-
ification may thus be a routine that involves a greater depth of processing,
that is, more rehearsal and more elaboration, than either inferring or consult-
ing alone and, accordingly, results in enhanced retention (Craik & Lockhart,
Lexical Processing Strategy Use 239

1972). Furthermore, as Kelly (1990) noted, this is also a routine that can, to a
significant extent, offset the risk of inferring an incorrect meaning.
In conclusion, this research adds to our understanding of incidental L2 vo-
cabulary growth through reading and also provides some insight into how the
lexical processing strategies L2 learners use when encountering unfamiliar vo-
cabulary affect vocabulary learning. It suggests that, although instruction that
enhances lexical processing strategy use does not directly affect vocabulary
learning, it can have an indirect effect by decreasing the number of unfamiliar
words ignored and increasing the quality of inferences.

Instructional Implications
This research supports the efficacy of instruction that aims to improve L2
learners’ ability to infer new word meanings and provides specific direction
regarding the kind of instruction on inferencing that is appropriate. Overall, it
supports the instructional focus outlined by Nation (1990) whereby students
are provided with a strategic procedure that guides them to use elements in
the immediate sentence context, first as the basis for their inference and, sub-
sequently, to monitor its appropriateness.
However, the results of this study also bring into question Nation’s (1990)
recommendation to direct students to avoid using word form cues until after
they have inferred on the basis of sentence context. Nation’s rationale for se-
quencing the use of word form cues to the end as a monitoring procedure was
that research had consistently shown that a common cause of error in lexical
inferencing was use of the word form rather than the context to infer. Data
from this study also support this observation. Additionally, however, results
suggest that the use of word form cues are typically accessed in a fast, auto-
matic manner as part of the process of inferencing through word-identification
processes. Other researchers have also noted the primacy of word form cues
in the inferencing process (Bensoussan & Laufer, 1984; Haynes, 1984; Huckin &
Bloch, 1993). Accordingly, it seems unlikely that we can train students to hold
off their potentially automatic use of word form cues. Rather, instruction
needs to guide students to monitor and elaborate inferences based on word-
identification processes through the use of the immediate sentence and the
wider text context.
Results from this study highlight the importance of consulting as an ex-
plicit vocabulary-learning strategy, particularly when it is used to verify an in-
ference. Thus, this study suggests a reevaluation of the minimal role often
accorded to dictionary use in the L2 reading class. Results accentuate the im-
portance of consulting both for effective determination of word meaning and
for vocabulary learning. This finding is consistent with research that has
found that consulting a dictionary enhances the reading comprehension
(Knight, 1994) and vocabulary learning (Knight, 1994; Luppescu & Day, 1993)
of L2 learners. In the present study, the use of consulting was an area of great
individual difference in both the frequency and effectiveness of its use. Several
240 Carol A. Fraser

participants rarely consulted and one participant had continuing difficulty ac-
cessing an appropriate meaning when consulting the dictionary. Therefore,
there may be a need to focus in the classroom on the efficient and effective
use of a dictionary while reading.
A final pedagogical implication is that L2 learners will likely benefit from
instruction that focuses on developing strategies for acquiring vocabulary
through reading. Results indicated that the learners in this study did acquire
a considerable amount of vocabulary in the context of a single reading of a
text for meaning. Moreover, some learners demonstrated substantially higher
rates of retention than others, and some LPSs were associated with higher re-
tention rates than others. Direct strategy training—the kind that makes stu-
dents aware of the differential impact of various LPS routines on vocabulary
learning and that introduces them to heuristics that facilitate the development
of rich and elaborated memory traces for words—should enable L2 learners
to more fully exploit the rich potential reading provides for vocabulary
learning.

NOTES

1. The eight student participants received an honorarium.

2. Readability was assessed using the Readability Estimator (Berta-Max Educational Software,
1988). A readability average was determined on the basis of five readability formulas (The Dale-Chall
Index, The Fog Index, The Flesch Index, The Flesch-Kincaid Index, and The Fry Index).
3. Because of the small number of participants, the power of analysis was low. Accordingly, al-
though an alpha level of .05 was used to assess statistical significance, effects that were approaching
statistical significance, that is, an alpha level of .10, were examined as well. This was done to offset
the risk that important, though statistically nonsignificant, effects would be overlooked (see Borg,
1987; Pressley & Afflerbach, 1995).
4. From the total of 640 cued recall items collected, 18 were not analyzed. Occasionally, partici-
pants did not indicate 10 unfamiliar words in a text. To ensure that all vocabulary learning measures
consisted of 10 recall items, the researcher selected supplementary items (18 in total), which were
subsequently deleted from the data base.
5. As one reviewer noted, these remembered word meanings could be approximate and even in-
correct. However, if vocabulary learning is viewed as a continuous and incremental process, then
the vocabulary knowledge that any individual has at a given time is never fully complete and accu-
rate (De Bot, Paribakht, & Wesche, 1997; Nagy & Herman, 1987). Within this framework, in the pres-
ent study vocabulary learning occurred to the extent that participants remembered the meanings
that they had determined for unfamiliar words encountered in the course of reading. Even though
these meanings could be partial and inaccurate, participants’ memory of them demonstrates learn-
ing. The quality or accuracy of this vocabulary learning was examined through the analysis of the
success or effectiveness of participants’ use of the various lexical processing strategies for determin-
ing appropriate word meaning for text comprehension.

REFERENCES

Bensoussan, M., & Laufer, B. (1984). Lexical guessing in context in EFL reading comprehension. Jour-
nal of Reading Research, 7, 15–32.
Berta-Max Educational Software. (1988). Readability estimator. Anacortes, WA: Berta-Max, Inc.
Borg, W. R. (1987). Applying educational research. London: Longman.
Clark, H. H., & Gerrig, R. J. (1983). Understanding old words with new meanings. Journal of Verbal
Learning and Verbal Behavior, 29, 591–608.
Craik, F. I. M., & Lockhart, R. R. (1972). Levels of processing: A framework for memory research.
Journal of Verbal Learning and Verbal Behaviour, 11, 671–684.
De Bot, K., Paribakht, T. S., & Wesche, M. B. (1997). Toward a lexical processing model for the study
Lexical Processing Strategy Use 241

of second language vocabulary acquisition: Evidence from ESL reading. Studies in Second Lan-
guage Acquisition, 19, 309–329.
Educational Testing Service. (1992). The Test of English as a Foreign Language. Princeton, NJ: Educa-
tional Testing Service.
Educational Testing Service. (1993). TOEFL: Institutional Testing Program. Princeton, NJ: Educational
Testing Service.
Ellis, N. C. (1994). Consciousness in second language learning: Psychological perspectives on the role
of conscious processes in vocabulary acquisition. AILA Review, 11, 37–56.
Gass, S. (1988). Integrating research areas: A framework for second language studies. Applied Linguis-
tics, 9, 198–217.
Haastrup, K. (1989). The learner as word processor. AILA Review, 6, 34–46.
Haastrup, K. (1991). Lexical Inferencing Procedures or Talking About Words. Tübingen: Gunter Narr.
Haynes, M. (1984). Patterns and perils of guessing in second language reading. In J. Handscombe,
R. A. Orem, & B. P. Taylor (Eds.), On TESOL ’83 (pp. 163–176). Washington, DC: TESOL Publica-
tion.
Huckin, T., & Bloch, J. (1993). Strategies for inferring word meaning in context: A cognitive model. In
T. Huckin, M. Haynes, & J. Coady (Eds.), Second language reading and vocabulary learning (pp.
153–176). Norwood, NJ: Ablex.
Hulstijn, J. H. (1992). Retention of inferred and given word meanings: Experiments in incidental vo-
cabulary learning. In P. J. L. Arnaud & H. Béjoint (Eds.), Vocabulary and applied linguistics (pp.
113–125). London: Macmillan.
Kelly, P. (1990). Guessing: No substitute for systematic learning of lexis. System, 18, 199–207.
Kintsch, W., & Mross, E. F. (1985). Context effects in word identification. Journal of Memory and Lan-
guage, 24, 336–349.
Knight, S. (1994). Dictionary use while reading: The effects on comprehension and vocabulary acqui-
sition for students of different abilities. The Modern Language Journal, 78, 285–299.
Krashen, S. (1989). We acquire vocabulary and spelling by reading: Additional evidence for the input
hypothesis. The Modern Language Journal, 73, 443–463.
Luppescu, S., & Day, R. R. (1993). Reading dictionaries and vocabulary learning. Language Learning,
43, 263–287.
Nagy, W. E., & Herman, P. A. (1987). Breadth and depth of vocabulary knowledge: Implications for
acquisition and instruction. In M. G. McKeown & M. E. Curtis (Eds.), The nature of vocabulary
acquisition (pp. 19–35). Hillsdale, NJ: Erlbaum.
Nagy, W. E., Herman, P. A., & Anderson, R. C. (1985). Learning words from context. Reading Research
Quarterly, 20, 233–253.
Nation, I. S. P. (1990). Teaching and Learning Vocabulary. Rowley, MA: Newbury House.
Paribakht, T. S., & Wesche, M. (1993). Reading comprehension and second language development in
a comprehension-based ESL program. TESL Canada Journal, 11, 9–29.
Paribakht, T. S., & Wesche, M. (1997). Vocabulary enhancement activities and reading for meaning in
second language vocabulary acquisition. In J. Coady & T. Huckin (Eds.), Second language vocabu-
lary acquisition: A rationale for pedagogy (pp. 174–200). New York: Cambridge University Press.
Pitts, M., White, H., & Krashen, S. (1989). Acquiring second language vocabulary through reading: A
replication of the Clockwork Orange study using second language acquirers. Reading in a Foreign
Language, 5, 271–275.
Pressley, M., & Afflerbach, P. (1995). Verbal protocols of reading: The nature of constructively respon-
sive reading. Hillsdale, NJ: Erlbaum.
Sanaoui, R. (1992). Vocabulary learning and teaching in the French as a second language classroom.
Unpublished doctoral dissertation, University of Toronto (OISE).
Schmidt, R. (1994). Deconstructing consciousness in search of useful definitions for applied linguis-
tics. AILA Review, 11, 11–26.
Schouten-van Parreren, C. (1989). Vocabulary learning through reading: Which conditions should be
met when presenting words in texts? AILA Review, 6, 75–85.
Swinney, D. A. (1979). Lexical access during sentence comprehension: (Re)consideration of context
effects. Journal of Verbal Learning and Verbal Behaviour, 18, 645–659.
Wesche, M., & Paribakht, T. S. (1996). Assessing second language vocabulary: Depth versus breadth.
The Canadian Modern Language Review, 53, 13–40.