Second Language Vocabulary Acquisition and
Learning Strategies in ICALL Environments
Thomas M. Segler1, Helen Pain1 and Antonella Sorace2
1 Division of Informatics, University of Edinburgh
mailto: email@example.com, firstname.lastname@example.org
2 Department of Theoretical & Applied Linguistics, University of Edinburgh
The research described in this paper investigates the role of Vocabulary Learning Strategies (VLS)
in ICALL environments. Although VLS taxonomies do exist, they have been developed for traditional
classroom-type learning, and tend to be incomplete in terms of strategies or factors arguably important
for vocabulary learning. Based on existing taxonomies, a VLS taxonomy for an ICALL environment will
be developed that takes these issues into account. The ?nal taxonomy, validated by factor analysis, will
also be evaluated by experienced language teachers in terms of being able to serve as a basis for providing
teaching ICALL materials. The usefulness of a broad range of VLS in this learning environment will
be investigated, both in terms of perceived helpfulness and assessed growth of lexical knowledge. The
taxonomy building is expected to be interleaved with an evaluation of vocabulary acquisition, which also
requires that the question of how to measure and assess lexical knowledge be addressed. The design of the
learning environment will not be considered in this paper.
Keywords: ICALL, Vocabulary Acquisition, Language Learning Strategies
Topics: Teaching and learning strategies, Pedagogical issues for intelligent language tutoring systems
Lexical acquisition has traditionally been neglected in Second Language Acquisition (SLA) research. In par-
ticular, little work has been carried out on Vocabulary Learning Strategies (VLS) in an ICALL environment.
This is unfortunate, as lexical acquisition is clearly central for SLA. Vocabulary is basic to communication,
and often seen as the greatest source of problems by second language learners (“When students travel, they
don’t carry grammar books, they carry dictionaries.” (Krashen, as cited in Lewis, 1993)).
The research described in this paper investigates the role of VLS in ICALL environments. Although tax-
onomies of a broad range of VLS do exist, they have been developed for traditional classroom-type learning
and thus need to be reassessed for an ICALL environment. They also tend to be incomplete in terms of
strategies or factors arguably important for vocabulary learning.
Compared to a traditional classroom-type learning environment, ICALL offers the unique advantage of
logging, analysing and tracking both the development of lexical knowledge, and the usage patterns of VLS.
The main goals are: (a) to investigate the usefulness of a broad range of VLS in an ICALL environment,
both in terms of perceived helpfulness and assessed growth of lexical knowledge, (b) to develop a VLS
taxonomy for this type of learning environment, and (c) to evaluate its pedagogical usefulness in terms of
providing a basis for the development of a sequence of teaching materials.
An investigation of the above raises the underlying question of the assessment of lexical knowledge in
an ICALL environment, i.e. the problem of when a word can be said to be acquired. This issue will also
be addressed below. An ICALL system will be developed in order to provide a comprehensive vocabulary
learning environment, with as broad a range of VLS as possible. The design of this environment will not be
considered in this paper.
Vocabulary Learning Strategies for a Classroom-type Environment
VLS constitute a subclass of language learning strategies, which are applicable to a wide variety of language
learning tasks, ranging from the more isolated (vocabulary, pronunciation, grammar) to integrative tasks like
oral communication and reading comprehension. Studies such as Schmitt and McCarthy (1997) have shown
that language learning strategies are not inherently ‘good’, but depend on the context in which they are used,
their combination with other strategies, frequency of use, and the learners’ pro?ciency level. One of the
?rst attempts at providing a comprehensive overview of language learning strategies can be found in Oxford
(1990). She identi?ed two distinct approaches to language learning, direct (memory/cognitive/compensation)
and indirect (metacognitive/social/affective) strategies1.
The importance of VLS in the group of language learning strategies is re?ected by the fact that the vast
majority of strategies in taxonomies such as Oxford’s are either VLS (all strategies in the memory category),
or can be used for vocabulary learning tasks. Although the use of a wide variety of strategies has been
found to be characteristic of successful learners, the great majority of learners seem to favour some form
of mechanical strategy such as repetition over deeper, more complex ones, such as contextual guessing or
metacognitive strategies (Lawson and Hogben, 1996, Gu and Johnson, 1996). This ?nding is disappointing in
the light of the Depth-of-Processing (DOP) hypothesis (Craik and Lockhart, 1972), which states that ‘deeper’
analysis of a stimulus (with ‘depth’ referring to a greater degree of semantic involvement) leads to better
long-term memory retention.2
Analyse af?xes, roots
Check for L1 cognate
Guess from context
Use word lists
Teacher checks word lists
Interact with L1 speakers
Image of word meaning
Connect to related words
Group words together
Study word sound/spelling
Use physical action
Paraphrase word meaning
Underline initial letter
Put L2 labels on objects
Use L2 media
Continue study over time
Skip/pass new word
Table 1: Excerpts from Schmitt’s Taxonomy
1As regards the distinction between memory and cognitive categories - the goal of both being to help recall words through some form
of language manipulation - Schmitt and McCarthy (1997) characterise the former as more obviously linked to mental manipulation.
2Although the DOP hypothesis in its original version has been subjected to various criticisms and modi?cations (see e.g. Craik and
Tulving (1975), Craik (1979), Lockhart and Craik (1990), and some ideas turned out to be untenable or least questionable, as a whole
their proposal has proved to be widely in?uential. In particular, the idea that the nature of mental processing is crucial for long-term
memory retention and memory performance is generally agreed upon today.
In the area of VLS taxonomy development, the most notable efforts in terms of range of strategies con-
sidered have been Stoffer (1995) and Schmitt (1997).
Stoffer developed a Vocabulary Learning Strategy Inventory (VOLSI) containing slightly fewer items
than Schmitt’s taxonomy (53 compared to 58). Using factor analysis, Stoffer found that the 53 VOLSI items
clustered into nine categories: strategies involving authentic language use, strategies used for self-motivation,
strategies used to organise words, strategies used to create mental linkages, memory strategies, strategies
involving creative activities, strategies involving physical action, strategies used to overcome anxiety, and
Schmitt’s goal was to develop a comprehensive inventory of individual VLS, and classify them along two
dimensions. The ?rst classi?cation dimension was adopted from Oxford (1990), who grouped learning strate-
gies into four3 categories (social (SOC), memory (MEM), cognitive (COG), and metacognitive (MET)). In
order to account for the case where meanings of new words are discovered without recourse to other people’s
expertise, Schmitt introduced a ?fth category, determination (DET) strategies. These seem to be roughly
equivalent to the “guessing intelligently in listening and reading” part of Oxford’s compensation strategies.
The second classi?cation dimension was proposed by Nation (1990) and re?ects the distinction of initial
discovery of word meanings (discovery strategies - DISCOV) and remembering (consolidation strategies -
Schmitt’s taxonomy groups VLS into 6 main categories (according to combinations of possible values of
the two dimensions) with 58 individual strategies in total (see Table 1). The reader will notice that not all 10
possible category combinations are actually realised in existing VLS - this is because discovery is the primary
purpose of only a small range of strategies (determination and social strategies). A caveat to bear in mind
is that some extent of arbitrariness in the classi?cation scheme could not be completely avoided, as some
strategies could be classi?ed under several headings. Although validation by factor analysis of Schmitt’s
taxonomy as a whole is still pending, Kudo’s (1999) study of VLS used by Japanese learners largely supports
Oxford’s (1990) classi?cation scheme (which served as a basis for Schmitt’s taxonomy).
Other notable classi?cation schemes based on description have been proposed by Nation (2001, p. 218)
and Gu and Johnson (1996). Nation distinguishes strategies relating to the planning of vocabulary learning
(e.g. choosing words, planning repetition) from strategies involving access to sources of vocabulary knowl-
edge (e.g. analysing word parts, using parallels with other languages), and learning processes. The latter
involve ways of establishing vocabulary knowledge (noticing, retrieving, and generating). Gu and Johnson
(1996) have developed a vocabulary learning questionnaire containing a considerable number of strategies,
divided into the following major categories: beliefs about vocabulary learning, metacognitive regulation,
guessing strategies, dictionary strategies, note-taking strategies, memory strategies (rehearsal and encoding),
and activation strategies.
Compared to other classi?cation schemes, Schmitt’s taxonomy is probably the most extensive, and has
the advantage of being organised around an established scheme of language learning strategies.
Vocabulary Learning Strategies in an ICALL Environment
A research topic still awaiting investigation is the establishment of a comprehensive taxonomy of VLS in an
ICALL/multimedia-type environment (Duquette et al., 1998). A central question that needs to be addressed
in this context is: What is the range of strategies that should be provided by an ICALL system for vocabulary
To answer this question, Schmitt’s basic distinction of discovery (DISCOV) and consolidation (CONS)
strategies seems to be a useful starting point. DISCOV strategies lend themselves to reading (or listening)
comprehension tasks, while CONS strategies are meant to be used in isolation from a ‘meaningful’ language
learning activity. In an ICALL system, CONS strategies could be implemented by tools to review and consol-
idate partially acquired lexical knowledge, e.g. tools for note-keeping, network-building etc. DISCOV strate-
gies, on the other hand, can be investigated in a reading/discovery-type environment with online glosses4/help
3Actually, Oxford’s scheme consists of six top-level strategy groups (see above) - Schmitt does not include affective and compensation
4A gloss is a translation or brief explanation of dif?cult or technical text (e.g. unusual words).
functions. In this scenario, learners reading texts have the opportunity to click on dif?cult/unknown words and
choose among several explanation options, each of which exempli?es a different type of DISCOV strategy.
Development of the Taxonomy
While Schmitt’s taxonomy will be used as a basis for the development of the taxonomy, it will have to be
revised and updated in several respects. Although it seems reasonable to conjecture that each of Schmitt’s 6
major strategy types keep their relevance in an ICALL environment, the extent to which this is true for social
strategies will most likely depend on the degree of collaborativeness of the system. The other strategy groups
could in some cases be amended by strategies more appropriate for an ICALL environment, or strategies
deserving inclusion that have not been considered in Schmitt’s taxonomy. An example of the former would
be the CONS-MEM strategy ‘use physical action’ which is not applicable to an ICALL system.
Whereas Schmitt’s taxonomy was developed for L2 English (with L1 Japanese), in this research L2 will
be German (with L1 English). Strategies involving internal morphological decomposition (such as analysing
af?xes and roots) can be hypothesised to play a more signi?cant role in L2 German (which is morphologically
richer than English). In particular, structural analysis of compound words (an explicit strategy of this kind is
missing in Schmitt’s taxonomy) can be expected to play a much more important role in L2 German compared
to L2 English.
ICALL environments are also conducive to investigations of a vocabulary learning aspect neglected in
major learning typologies up to now, the ‘unlearning’ of wrongly perceived meanings. This is because mis-
conceptions about a word’s meaning can not only be discovered and logged via some form of online vocab-
ulary assessment, but also linked to a learner’s prior strategy use (or usage patterns). The issue of detecting
misconceptions, as well as successful ‘unlearning’ of them, is of course closely linked to the choice of vocab-
ulary assessment format (see below). Since misconceptions in the area of lexical knowledge can occur along
several dimensions and in different degrees, a decision about what constitutes a misconception will have to
be made within the framework of the chosen vocabulary assessment apparatus. In particular, misconceptions
can be expected to occur between words of similar lexical form, but different meaning, both between L1 and
L2 (false cognates) and within L2 (“synforms”)5.
Speci?c questions that may be investigated include: (a) Are particular strategies/types of glossing/follow-
up exercises, or combinations thereof, particularly effective at the task of unlearning wrongly perceived mean-
ings, and if so, (b) can they be related to particular types of misconceptions? If (a) is the case, factor analysis
might shed some light on the question of whether this subset of strategies is to be elevated to the status of a
primary strategy dimension.
Another dimension not explicitly mentioned in typologies of language/vocabulary learning strategies so
far is depth-of-processing. As there is a consensus in the literature that depth-of-processing (DOP) is a crucial
variable for vocabulary retention, it seems worthwhile to investigate whether it constitutes a valid dimension
in the resulting typology. The main problem here seems to be ?nding an independent index for DOP, i.e.
operationalising the concept in a precise way. The following could be a ?rst approximation to this end: ?rst,
the learning strategies are classi?ed into broad DOP categories, where this can be done in a reasonably self-
evident, straightforward way. For instance, it is obvious that rote repetition would be assigned to a lower DOP
category than a mnemonic technique such as the keyword method. Since the dif?culty of operationalising
DOP arises not across, but within, these broad categories, a second step would be an empirical evaluation to
differentiate the DOP index within categories. This could take the form of having teachers and/or learners
rate the subjective degree of DOP for each learning strategy on a scale.
In sum, it is envisaged that the resulting taxonomy will be based on Schmitt’s primary strategy classi?ca-
tion scheme, but undergo some revision in terms of adding and amending strategies within the major category
5see e.g. Laufer (1985, 1988).
Evaluation of the Taxonomy
The process of taxonomy-building will be interleaved with an evaluation process: initial modi?cations to
Schmitt’s taxonomy based on introspection will be subjected to evaluations of both language teachers and
language learners. Learners will evaluate the strategies in terms of perceived helpfulness, which can be com-
pared to their actual strategy usage. Since Schmitt has investigated these parameters - subjective helpfulness
and actual usage - for his classroom-type taxonomy using Japanese learners of L2 English, a comparison
to Schmitt’s results can shed light on the usefulness of a broad range of VLS for a different L2 and type
of learning environment. The usefulness of the VLS on offer will also be measured in terms of growth of
lexical knowledge, which raises the question of how to assess lexical acquisition in the context of an ICALL
vocabulary learning environment (see below). Finally, an empirical study to evaluate the pedagogical use-
fulness of the resulting taxonomy will be carried out. This could be done by asking experienced language
teachers to judge the taxonomy in terms of its usefulness for providing a basis for developing a sequence of
(ICALL/multimedia) teaching (reading) materials.
Another improvement on Schmitt’s taxonomy will be to run a factor analysis on the classi?ed strategies to
validate Schmitt’s basic classi?cation scheme for the ICALL environment - as Kudo (1999) observes, failure
to do so in Schmitt’s taxonomy calls into question whether the primary categories really share the common
Evaluation of Vocabulary Acquisition
The research issues described so far in this paper raise the question of how to measure lexical knowledge,
or acquisition. This problem points to the underlying question of when a word6 can be said to be acquired,
which in turn raises the fundamental issue of how to de?ne or characterise the concepts of lexical knowledge
and word meanings. While an in-depth look at these issues7 is outside the scope of this paper (see Segler
(2001) for a more detailed discussion), it should be noted that the traditional view of words as having ?xed
meanings has been challenged by recent statistical-based approaches to word meaning. In keeping with the
Wittgensteinian view that words do not have simple meanings in terms of concepts, these approaches are
based on the assumption that “a word is de?ned by its use in a wide range of contexts” (Burgess and Lund,
Latent Semantic Analysis
A relatively recent and very promising example of a statistical-based approach is Latent Semantic Analysis
(LSA)8. In LSA, semantically similar words are believed to exhibit similar contextual distributions. As
Landauer et al. put it, “the aggregate of all the word contexts in which a given word does and does not
appear provides a set of mutual constraints that largely determines the similarity of meaning of words and
sets of words to each other” (Landauer et al., 1998). Using raw text as the only input, LSA produces vectors
representing the frequency distribution of lexical items across a wide range of contexts. Word meaning is thus
represented as an average of the meaning of the passages in which the word is contained. Semantic similarity
can then be measured by vector proximity in the high-dimensional semantic space.
Given that meaning is ?uid and speakers’ understanding of words changes over time, LSA seems promis-
ing as a potential technique to measure this drift as a function of an individual’s language exposure - clearly
an attractive prospect in regard to assessing lexical understanding of (second) language learners.
However, it is an open question whether (and to what extent) statistically-based approaches like LSA
could be used to model lexical knowledge for (second) language learners. Apart from the obstacle of requiring
rather large amounts of training data, i.e. huge corpora, the treatment of homonyms such as ‘bank’ appears
6The general term word is used in this paper as a convenient shorthand for the concept of lexical unit, as de?ned by Cruse (1986,
p. 24). Most ‘words’ are polysemous and therefore correspond to several lexical units; a lexical unit can also be made up of several
‘words’ (e.g. hot dog). A discussion of why the notion of linguistic unit is the more appropriate concept in the context of L2 vocabulary
acquistion can be found in (Bogaards, 2001).
7In particular, the question of what constitutes lexical knowledge, and to what extent it is different from syntactic knowledge
8Landauer and Dumais (1997) have shown that it correlates well with several cognitive phenomena; for an overview of LSA, see
(Landauer et al., 1998).
less than ideal. Since LSA represents the meaning of a word as a sort of average of the meaning of the
paragraphs/passages in which it is contained, any differences in meaning are glossed over, which might be
acceptable for most polysemous words with comparatively slight differences in meaning, but is much more
problematic for homonyms whose meanings are usually unrelated. Finally, while LSA seems to capture
judgements of similarities of meaning quite well (cf. McDonald (1997)), it has little to say about the way
in which words are related - e.g. do they stand in a syntagmatic/paradigmatic or hyponymic/antonymic
relationship9. This sort of information (besides usage and reference) constitutes an important chunk of lexical
knowledge, i.e. what a learner has to be cognizant of in order to ’know’ a word. As Landauer and Dumais
(1997) note, “the similarity relations between words that are extracted by LSA are based solely on usage.”
For these reasons, LSA (or a similar statistical approach) will not be used to assess lexical knowledge of
L2 learners10. Instead, a suitable test of depth of lexical knowledge seems to provide a more viable alternative
for the purposes of the research described.
Vocabulary Knowledge Scale
Tests of depth of lexical knowledge can broadly be classi?ed into two categories: (a) tests attempting to anal-
yse the different aspects of lexical knowledge, and (b) ‘developmental’ tests “identifying levels of knowledge
that may be interpreted as stages in the acquisition of the word.”(Read, 1997, p. 315). Due to the dif?culty of
designing tests that accurately assess the complex multidimensional construct of lexical knowledge, existing
tests mainly fall in category (b), typically using some sort of rating scale, such as the Vocabulary Knowledge
Scale (VKS) mentioned above.
The VKS has the following ?ve steps, or categories:
I I don’t remember having seen this word before.
II I have seen this word before, but I don’t know what it means.
III I have seen this word before, and I think it means
(synonym or translation)
IV I know this word. It means
. (synonym or translation)
V I can use this word in a sentence:
. (Write a sentence)
(If you do this section, please also do Section IV.)
Despite being “a workable instrument, allowing coverage of a reasonable number of words” (Read, 1997,
p. 317), and “sensitive to increases in vocabulary knowledge that result from reading activities” (Read, 2000,
p. 135), various aspects of its validity have been questioned (cf. Read (1997, 2000)). First, it is not clear that
the levels in the scale correspond to acquisition stages. Second, the VKS in its present form does not account
for multiple meanings of a word. While the introduction of additional scales might be a remedy, it would run
counter to the VKS’s original intention of providing a practical single scale. Of course, the general problem
seems to be the attempt to reduce the complex, multidimensional construct of vocabulary knowledge to a
?xed linear scale, which does not seem to re?ect the nature of lexical knowledge very well. This makes it
next to impossible to relate a given increase in score to a speci?c dimension, or level, of lexical knowledge.
Despite these caveats, VKS-type scales are the most attractive candidates among commonly-used vocab-
ulary test formats, avoiding both the pitfalls of simple recall, multiple-choice or cloze tests (which neglect the
multidimensional nature, i.e. depth, of lexical knowledge), and tests attempting to measure all dimensions of
lexical knowledge (which are impracticable to due to the huge battery of tests required). Finally, the VKS’s
proclaimed purpose to “track the early development of speci?c words in an instructional or experimental
situation” (Wesche and Paribakht, 1996, as cited by Read, 2000, p. 33) seems to ?t quite nicely with the
assessment purposes of this research.
9The semantic dimensions in the high-dimensional semantic space are of no help here, as they are abstractions and do not correspond
to ’real’ semantic features or dimensions.
10However, LSA may well be used in the context of providing example sentences explaining an unknown word’s usage - see Segler
(2001) for details.
Multi-state models (see Figure 1) are a relatively recent alternative to traditional vocabulary tests such as
probing breadth or depth of lexical knowledge, and rating scales such as the VKS (cf. Waring (1999)).
Figure 1: An example Multi-state model with 5 states (adopted from Waring (1999))
Figure 2: An example for several words (adopted from Waring (1999))
Instead of arranging levels of knowledge on a linear scale, multi-state models represent them in inter-
connected states “with no necessary assumption that one State is higher or lower in knowledge than another
along a continuum” (Waring, 1999, chapter 4). While there may be empty states, all words should be in a
particular state (see Figure 2).
Since lexical development in multi-state models is not represented along a continuum, or cline, but rather
as the movement between (or addition to) states, vocabulary acquisition is conceptualised as change, rather
than growth as the traditional scale models imply (thus vocabulary attrition can also be accounted for). Fur-
thermore, as all states are interconnected, there is no assumption that a given word has to pass through a state
B on its way from A to C (one of the potential problems of a rating scale). Meara (1996) describes a method
of tracking the global development of a learner’s vocabulary based on a matrix of transitional probabilities.
In order to perform the analysis and provide a basis for probability calculations, data is gathered twice in the
form of learners reporting on the state of their knowledge of a large set of words for a small number (four) of
Thus, in contrast to the VKS which aims to assess the development of individual words, multi-state
models seem more geared to a macro-level, global view of the learners’ lexicon. They rely on learners’
self-assessment, which may sometimes be inaccurate, and generally provide a less straightforward rating
scheme than a more traditional rating scale approach. For these reasons, multi-state models appear to be
less suitable for the main purposes of lexical assessment in the current research. However, they may still
be utilised as a complementary assessment tool for an investigation of effects of the VLS employed on the
overall development of the learners’ vocabulary knowledge.
The research described in this paper will provide a better understanding of the use of VLS in an ICALL
environment. Through a process of interleaved evaluation of vocabulary acquisition and taxonomy building,
the investigation will provide a taxonomy especially geared to VLS used in this environment. The ?nal
taxonomy, validated by factor analysis, will also be judged by experienced language teachers in terms of
being able to serve as a basis for providing teaching ICALL materials. It is hoped that, in addition to revising
and updating existing VLS taxonomies for the new environment, the resulting taxonomy will re?ect important
issues and dimensions which have not received explicit attention in the building of VLS taxonomies so far.
These include the depth-of-processing hypothesis, and the ‘unlearning’ of wrongly perceived meanings.
The work is also expected to shed light on the usefulness of a broad range of VLS in an ICALL environ-
ment for L2 German, both in terms of perceived helpfulness and assessed growth of lexical knowledge. As
has been argued in this paper, the latter also requires that the issue of how to evaluate and measure lexical
acquisition be addressed.
Bogaards, P. (2001). Lexical units and the learning of foreign language vocabulary. Studies in Second
Language Acquisition, 23:321–343.
Burgess, C. and Lund, K. (1997). Modelling parsing constraints with high-dimensional context space. Lan-
guage and Cognitive Processes, 12:177–210.
Craik, F. (1979). Levels of processing: Overview and closing comments. In Cermak, L. and Craik, F., editors,
Levels of Processing in Human Memory, pages 447–461. Erlbaum, Hillsdale, N.J.
Craik, F. and Lockhart, R. (1972). Levels of processing: A framework for memory research. Journal of
Verbal learning and Verbal Behavior, 11:671–684.
Craik, F. and Tulving, E. (1975). Depth of processing and the retention of words. Journal of Experimental
Cruse, D. (1986). Lexical Semantics. Cambridge University Press.
Duquette, L., Reni´e, D., and Laurier, M. (1998). The evaluation of vocabulary acquisition when learning
French as a second language in a multimedia environment. Computer Assisted Language Learning, 11:1:3–
Gu, Y. and Johnson, R. (1996). Vocabulary learning strategies and language learning outcomes. Language
Kudo, Y. (1999). L2 vocabulary learning strategies (n?rc network 14) [html document]. Honolulu: University
of Hawai’i, Second Language Teaching & Curriculum Center. Retrieved [2000, Oct 20] from the World
Wide Web: http://www.lll.hawaii.edu/n?rc/NetWorks/NW14/.
Landauer, T. and Dumais, S. (1997). A solution to Plato’s problem: The latent semantic analysis theory of
the acquisition, induction, and representation of knowledge. Psychological Review, 104:211–240.
Landauer, T., Foltz, P., and Laham, D. (1998). An introduction to latent semantic analysis. Discourse
Laufer, B. (1985). Vocabulary Acquisition in a Second Language: the Hypothesis of ‘Synforms’ (similar
lexical forms). PhD thesis, University of Edinburgh.
Laufer, B. (1988). The concept of ‘synforms’ (similar lexical forms) in L2 learning. Language and Education,
Lawson, M. and Hogben, D. (1996). The vocabulary-learning strategies of foreign-language students. Lan-
guage Learning, 46:1:101–135.
Lewis, M. (1993). The Lexical Approach. Language Teaching Publications.
Lockhart, R. and Craik, F. (1990). Levels of processing: A retrospective commentary on a framework for
memory research. Canadian Journal of Psychology, 44(1):87–112.
McDonald, S. (1997). Exploring the validity of corpus-derived measures of semantic similarity. Paper
presented at the 9th Annual CCS/HCRC Postgraduate Conference, University of Edinburgh, June 18-19.
Meara, P. (1996). The vocabulary knowledge framework. Vocabulary Acquisition Research Group Virtual
Library [Online]. Available: http://www.swansea.ac.uk/cals/calsres/vlibrary/pm96d.htm [2001, Jan 15].
Nation, I. (1990). Tecahing and Learning Vocabulary. Heinle & Heinle, Boston, MA.
Nation, I. (2001). Learning Vocabulary in Another Language. Cambridge University Press, Cambridge.
Oxford, R. (1990). Language Learning Strategies: What Every Teacher Should Know. Newbury House,
Read, J. (1997). Vocabulary and testing. In Schmitt, N. and McCarthy, M., editors, Vocabulary: Description,
Acquisition and Pedagogy, pages 303–320. Cambridge University Press, Cambridge.
Read, J. (2000). Assessing Vocabulary. Cambridge University Press, Cambridge.
Schmitt, N. (1997). Vocabulary learning strategies. In Schmitt, N. and McCarthy, M., editors, Vocabulary:
Description, Acquisition and Pedagogy, pages 199–227. Cambridge University Press.
Schmitt, N. and McCarthy, M., editors (1997). Vocabulary: Description, Acquisition and Pedagogy. Cam-
bridge University Press.
http://www.dai.ed.ac.uk/homes/thomasse/newprop.pdf [2001, Apr 23].
Stoffer, I. (1995). University Foreign Language Students’ Choice of Vocabulary Learning Strategies as Re-
lated to Individual Difference Variables. PhD thesis, University of Alabama. Unpublished doctoral disser-
Waring, R. (1999). Tasks for assessing second language receptive and productive vocabulary. PhD Thesis
[Online]. Available: http://www1.harenet.ne.jp/?waring/papers/phd/title.html [2000, Nov 27].
Wesche, M. and Paribakht, T. (1996). Assessing second language vocabulary knowledge: Depth vs. breadth.
Canadian Modern Language Review, 53:13–39.