You are on page 1of 14

Available online at www.sciencedirect.

com

ScienceDirect
System 41 (2013) 1056e1069

www.elsevier.com/locate/system

Effects of semantic and phonological clustering on L2 vocabulary


acquisition among novice learners
Amanda Wilcox a,1, Almitra Medina b,*
a
b

6030 Haley Center, Department of Foreign Languages and Literatures, Auburn University, AL 36849-5204, USA
6030 Haley Center, Department of Foreign Languages and Literatures, Auburn University, AL 36849-5204, USA
Received 2 October 2012; revised 4 October 2013; accepted 4 October 2013
Available online 26 October 2013

Abstract
Vocabulary is critical in second language acquisition, yet there is no consensus as to how to present vocabulary in a way that
facilitates learning. Scant empirical evidence supports the claim that presenting words in semantic clusters is helpful, while
increasing evidence asserts that grouping words semantically impedes L2 vocabulary acquisition. How best to cluster L2 vocabulary, therefore, is an unresolved question. Research in cognitive psychology shows that including repetitive aspects (such as certain
consonant-vowel combinations) in verbal information facilitates information retrieval. Presenting words with a repetitive aspect
makes these vocabulary items phonologically similar. Therefore, the present study investigated semantic and phonological clustering effects among 32 novice learners of L2 Spanish. Vocabulary was presented in four categories ([semantic, phonological]),
and within-subjects comparisons were made as to which clusters were learned better and retained longer over time. Results show
that the semantically clustered (and not phonologically similar) words were more difficult to learn. Performance improved by
presenting vocabulary in random and/or phonological groups both in the short term and over time. Even though the mind may tend
to store accumulated vocabulary in semantic fields, it may initially be prepared to receive words that need to be organized in
semantic fields, not to receive words that have already been semantically pre-organized.
2013 Elsevier Ltd. All rights reserved.
Keywords: Semantic clustering; Phonological clustering; Novice learners; L2 vocabulary acquisition; Interference theory; Distinctiveness hypothesis; Semantic field theory; Componential analysis; Hebb repetition effect

1. Introduction
Vocabulary learning is a crucial factor in second language acquisition (SLA). Learners must acquire hundreds of
words at a minimum in order to be able to communicate in the target language at the most basic level. Increasing
vocabulary knowledge in novice second language (L2) learners should not only help them communicate more
effectively, but also help learners comprehend a broader range of input from the target language. The question then
arises as to how to package vocabulary in the most beneficial way to facilitate learning and improve comprehension.
With this query in mind, the present study explored the effects of grouping words in semantic and phonologically
* Corresponding author. Tel.: 1 334 844 6363; fax: 1 334 844 6378.
E-mail addresses: aew0018@tigermail.auburn.edu (A. Wilcox), medina@auburn.edu (A. Medina).
1
Tel.: 1 334 844 6364; fax: 1 334 844 6378.
0346-251X/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.system.2013.10.012

A. Wilcox, A. Medina / System 41 (2013) 1056e1069

1057

similar/disparate clusters. In order to contextualize the study, Section 2 reviews research that both support and conflict
with the commonly held assumption that presenting vocabulary to novice learners in linguistically linked groups
facilitates its acquisition. A review also follows (in Section 3) of the Hebb repetition effect (Hebb, 1949) and how
the associated first language (L1) cognitive research in lexical learning could be pertinent in an L2 context as a method
of vocabulary presentation.
2. Semantic clustering
A number of L2 vocabulary acquisition studies have been designed to test the underlying assumption that vocabulary should be presented in groups of words that are related semantically, or groups of words whose meaning
would fall under one superordinate concept (Finkbeiner and Nicol, 2003; Tinkham, 1993, 1997; Waring, 1997). Most
of these studies point out that the vast majority of available textbooks organize words in this way, under headings such
as body parts or kitchen utensils, word groupings that are linguistically linked by meaning (but not necessarily
synonyms of each other). Following Tinkham (1997), a semantic cluster is defined in the current study as a group of
words with semantic and syntactic similarities, such as apricot, peach, plum, nectarine, pear, apple, which fall under
the superordinate concept fruit and come from a single syntactic word class, in this case, nouns. Note that semantic
clusters differ conceptually from thematic clusters. Thematic clusters, such as frog, pond, swim, hop, green, slippery,
are argued to be based on psychological associations between words and a shared thematic concept (Tinkham, 1997,
p. 141e142) and contain lexical items of variable word class, such as a mixture of nouns, verbs, and adjectives.
Therefore, while semantic categories are linguistically based and of a single word class (e.g., all nouns, all verbs, or all
adjectives), thematic clusters are cognitively based and of multiple word classes. Though these two types of clusters
are not necessarily mutually exclusive (some concepts could be related both semantically and thematically), the
nuance of difference between the above groupings is still apparent.
There are theories (e.g., semantic field theory, Lehrer, 1974) that seem to support clustering vocabulary presentation in a semantic way, as well as authors who assert the effectiveness of semantic set presentation (e.g., Grandy,
1992; Hashemi and Gowdasiaei, 2005; Haycraft, 1993; Seal, 1991; Stoller and Grabe, 1995; Wharton and Race,
1999; Hoshino, 2010). However, there is also theoretical support (e.g., interference theory, Baddeley, 1990; Higa,
1963) and a growing body of empirical evidence that suggest semantic clustering deters L2 vocabulary acquisition
at the novice level (e.g., Erten and Tekin, 2008; Finkbeiner and Nicol, 2003; Tinkham, 1993, 1997; Waring, 1997).
2.1. Theoretical framework
The idea of presenting vocabulary in semantic clusters to improve acquisition and retention is not a new one and has
its roots in L1 research. In a summary of studies on L1 vocabulary acquisition and their implications for L2, Stoller
and Grabe (1995) noted that L1 research supports the claim that new learning occurs when new material can be
attached to current knowledge (schema). Stoller and Grabe argue that vocabulary, therefore, should be taught in
semantically related sets in order to activate schema and thereby connect target material with material that was
previously learned. Since many of the strategies for learning vocabulary in an L1 or L2 are purportedly the same (e.g.,
vocabulary learning in both languages often occurs by means of definition, example, context, and morphological
analysis), it is argued that the benefits for semantic set presentation can be extended to SLA (Stoller and Grabe, 1995).
The utility of connecting new words with preexisting schema has also been put forth in educational psychology
research. Ausubel (1968) explained that advance organizers, which function as a superordinate word or concept under
which target words or concepts can be organized, should be introduced to learners in advance of receiving new information in order to activate existing schemas, allowing learners to therefore connect newly presented material with
already known material. Although arguments in favor of semantic clustering do not make claims as to whether the
superordinate word or concept should be presented to learners before they are exposed to the new information or not,
the notion of connecting vocabulary with existing schemas is nonetheless evident.
Two other concepts, namely, semantic field theory (Lehrer, 1974) and componential analysis (Nida, 1975), have
also served to provide support for semantic clustering (Channell, 1981). Semantic field theory is based on the
assumption that rather than being organized in lists of random words, vocabulary is cognitively organized by interrelationships and networks between words. The mind classifies vocabulary by making connections in meaning, and
these connections are considered semantic fields (Channell, 1981; see Grandy, 1992; for further discussion of

1058

A. Wilcox, A. Medina / System 41 (2013) 1056e1069

semantic fields). In Channells (1981) interpretation, words that are close in meaning are literally located closer to
each other in the mental lexicon. The conclusion drawn suggests that since vocabulary is organized in the mind into
groups of words that are linked in meaning (Miller and Johnson-Laird, 1976), vocabulary should thus be presented to
students in semantically-related groups, mimicking the configuration of the lexicon, to facilitate learning. Using
componential analysis provides a systematic way of describing vocabulary words. Componential analysis requires
that the meanings of words be broken down into what is known as semantic components. Channel differentiates which
meanings are shared by words and which connotations distinguish them from one another. Once the semantic
components are derived, distinguishing one words meaning from another becomes more apparent (Channell, 1981).
Two concepts that argue against semantic clustering are that of interference theory and the distinctiveness hypothesis. Interference theory (Baddeley, 1990; Higa, 1963) states that when words are being learned at the same time,
but are too similar or share too many common elements, these words will interfere with each other thus impairing
retention of them (Waring, 1997; pp. 261e262) because traces in memory often compete with each other. Waring
(1997) therefore concludes that vocabulary should not be presented in sets that are grouped under one superordinate concept given their similarity. Nation (2000), concurring with the argument against semantic clustering, also
elaborates by stating that interference occurs when items presented together are both unfamiliar, or when one is
unfamiliar and the other poorly established (p. 9). This would seem to be particularly applicable to novice L2 learners
and could help explain results from Tinkham (1993, 1997) and Waring (1997), who used nonsense words to mimic
novice L2 learning.
The distinctiveness hypothesis focuses on differences rather than similarities and, in essence, comes to the same
conclusion. Since similarity seems to confound the mind, distinctiveness should help organize it. Research (Hunt and
Elliot, 1980; Hunt and Mitchell, 1982) behind the distinctiveness hypothesis proposes that increasing the nonsimilarity of information increases its ease of learning, and as such, vocabulary should be presented in a nonrelated
fashion so that the mind is presented with information organized in a way that is conducive for learning. Ausubel
(1968) also acknowledges that the differences between new and existing ideas and can be confusingly similar,
but explains that a comparative organizer is employed to help discriminate between such similar ideas (p. 149).
2.2. Support in favor of and against semantic clustering
Researchers (e.g., Finkbeiner and Nicol, 2003) acknowledge that there are L1 findings, albeit limited, supporting
the idea that learners benefit from semantic categorization of vocabulary presentation. There is little empirical evidence, however, that supports this claim in SLA, and only a handful of studies exist in this strand of research. Tinkham
(1993), in his study of advanced English as a Second Language (ESL) speakers learning artificial words, found evidence against semantic clustering. In a trials-to-criterion test to determine how easily participants could learn English/artificial words in sets of related and nonrelated pairs, Tinkhams participants were able to learn unrelated words
in statistically fewer trials than related words, a finding also corroborated by Waring (1997) with L1 Japanese/L2
artificial word pairs. Waring furthermore noted that 25% of the errors participants made when pairing related words
were from the same superordinate category, whereas this occurred only 5% of the time for nonrelated words. These
studies thus suggest an interfering effect of grouping words semantically.
In a later study, Tinkham (1997) examined both semantic and thematic clusters, but similar results surfaced with
respect to the semantically grouped words. That is, with L1 English/L2 artificial word pairs, it took participants
significantly longer to reach criterion when tested on semantically related words than when assessed on unrelated
pairs, and this was the case regardless of translation modality (i.e., written versus oral). Responding to the researchers posttest questionnaire, participants commented that the semantic sets were too similar, or all the
same (p. 379), which made it more difficult for them to remember the corresponding artificial word. Furthermore,
most participants indicated that the thematically based sets (rather than the semantic groupings) were easier to
learn.
Studies on semantic clustering have also examined the impact of translation direction (i.e., L1 to L2 vs. L2 to
L1) on the speed of responses (Finkbeiner and Nicol, 2003), but the findings have been consistent (consistent for
response speed; see, however, the immediate posttest results of Schneider et al., 2002; who examined response
accuracy). In a study of L1 English/L2 artificial pairs, Finkbeiner and Nicol (2003) found that translation in the
forward direction (L1eL2) took statistically longer in the semantically clustered condition than those exposed to
the random grouping of the same words. Translation in the opposite direction was also statistically slower for the

A. Wilcox, A. Medina / System 41 (2013) 1056e1069

1059

semantically related words. The semantic category effect, therefore, was negative in this experiment. A plausible
explanation offered was that the repeated conjuring of the same lemma (lexical hub of form, meaning, and
syntax) during the semantically clustered training condition caused interference and delayed translation in both
directions.
Most L2 studies on semantic clustering have incorporated only immediate (and not delayed) posttests, which
therefore limits what they say about L2 development. Hence, although it can be argued that presenting words in
semantic clusters is not facilitative for novice learners on the short term, little can be said about its effect on the long
term. Crow and Quigley (1985) attempted to investigate the impact of categorical and unrelated sets immediately after
treatment and over time (four and eight weeks later). The problem, however, is that type of presentation was conflated
with learning approach, calling into question whether they, in fact, were indeed testing for semantic set presentation
(see Hoshino, 2010; for a discussion). Papathanasiou (2009), therefore, arguably gave us the first indication that
semantic clustering has a negative long-term impact as well. Assessing her participants immediately after their final
vocabulary lesson and again two weeks later on translations from L2 English to L1 Greek, Papathanasiou found that
beginning-level adults in the semantic cluster group remembered significantly less words than those in the unrelated
group, regardless of time of assessment (immediate or delayed).
Immediate and delayed advantages for semantically unrelated sets are also corroborated in a study by Erten and
Tekin (2008), who found similar results with an image-vocabulary matching test among younger participants
(fourth grade Turkish students of English) at the novice level.
Clearly, irrespective of translation modality, i.e., written vs. oral, (Tinkham, 1997), translation direction, i.e., L1 to
L2 vs. L2 to L1 (Finkbeiner and Nicol, 2003; see Schneider et al., 2002; for an exception), and time of testing, i.e.,
immediate vs. delayed (Papathanasiou, 2009), most studies investigating the impact of semantic clustering among
novice L2 learners have come to the same general conclusion: Participants perform (as determined by speed of recall,
number of correct items, or number of trials to reach a predetermined criterion) statistically more poorly with
semantically related groups than with clusters that are unrelated. A few studies, however, provide (limited) evidence to
the contrary (Hashemi and Gowdasiaei, 2005; Hoshino, 2010; Schneider et al., 2002).
Schneider et al. (2002) found that only in a very specific circumstance did their L1 English/L2 French learners
show a statistical advantage with semantically related vocabulary. In their study, Schneider et al. examined the effects
of semantic (blocked category) versus nonsemantic (mixed) presentation methods and translation direction on vocabulary retention and transfer in two sessions. Session 1 was the learning phase, and Session 2, offered a week later,
was termed the relearning phase, whereby learners were exposed to the word pairs again. The effects found pertaining
to the blocked and mixed presentation of vocabulary hinged on both translation direction and session, as indicated by a
three-way interaction. Specifically, there was a sizeable advantage of the semantic clusters over the mixed groupings
during the learning phase only (not after relearning) and only for the participants asked to translate from L2 French to
L1 English (not those asked to translate from L1 to L2). There was essentially no difference between the two types of
presentations (regardless of translation direction) in Session 2, which means that although participants had less difficulty producing L2eL1 semantic sets initially, there was a greater loss of these words by Session 2, despite having
been re-exposed to the pairs. This studys results, therefore, do not offer convincing support in favor of semantic
clustering.
Another study with only some support for semantic set presentation is that of Hoshino (2010). Hoshino tested
Japanese university EFL students in a classroom setting on L2eL1 translations in five different types of word
groupings: synonyms, antonyms, category, thematic, and unrelated. Hoshino found that, independent of learning style,
categorically clustered words were the easiest to learn (participants performed statistically better upon examination),
and that there was no significant difference among the remaining groupings, findings that run contrary to other studies
(e.g., Tinkham, 1997; Higa, 1963). Although these results appear to offer some support for semantic clustering, the
findings are limited for two reasons. First, the EFL level of the participants was not specified (participants were
identified as freshman, sophomores, or juniors, but their L2 level was not indicated), which therefore lowers the
studys external validity. Second, the number of target words per set was limited to two (e.g., Set 1: moth, wasp; Set 2:
asthma, diabetes, Set 3: calf, chick), lower than in most other studies. Tinkham (1993) and Waring (1997), for
example, targeted three (Experiment 1) or six (Experiment 2) semantically related words per set, while Finkbeiner and
Nicol (2003) targeted eight. Targeting fewer words per set was necessary in order to create parallels with the antonym
clusters, but doing so decreased the similarities among semantically clustered items, which thus can reduce possible
interfering effects.

1060

A. Wilcox, A. Medina / System 41 (2013) 1056e1069

Interesting is the study of Hashemi and Gowdasiaei (2005), who found semantic sets to produce a significant
advantage over unrelated sets on both breadth and depth of vocabulary knowledge (i.e., how many and how well target
words are known). Sixty intermediate level EFL students from Iran were divided into two groups and presented with
new vocabulary either in lexical sets (LS) or semantically unrelated sets (SU). Results show that post-instruction, the
LS group fared better than the SU group in both measures of vocabulary knowledge. These participants, however,
were intermediate learners who had studied EFL for at least six years; therefore, the results are not necessarily
comparable to those of novice level learners nor generalizable to the novice population. In fact, Hashemi and
Gowdasiaei also found that upper level intermediate students had significantly greater gains than their lower level
intermediate peers. These findings, therefore, suggest that learners L2 level is indeed important in the ongoing debate
of semantic presentation.
Compiling the data on novice adult participants, it is clear that further exploration into vocabulary clustering is
warranted for several reasons. First, learners of Schneider et al. (2002) experienced an initial advantage with semantic
sets when given L2 French cues, but these findings were not corroborated in any other L2 study (e.g., Papathanasiou,
2009). Second, Tinkhams (1993, 1997) research strongly supports organizing words into nonrelated or thematically
related groups, but the implications of his investigations along with those of Warings (1997) and Finkbeiner and
Nicols (2003) are limited given that the words being tested were not from natural languages, but rather artificial,
and only had short-term components, which restricts the conclusions that can be drawn for L2 development. Moreover, Papathanasiou (2009) included both immediate and delayed L2 to L1 translation assessments; there is a need,
however, to expand this research to examine the opposite translation order (i.e., L1 to L2) in immediate and delayed
conditions (without a relearning phase). Therefore, in order to address the paucity of experimental research in this
strand of investigation and address limitations in previous studies, the present study sought to compare semantic and
nonsemantic vocabulary presentation of L1 English-L2 Spanish word pairs immediately after presentation and again
two weeks later. Furthermore, there are clearly a number of ways in which vocabulary words can be grouped for
presentation during L2 learning, yet other than Tinkhams (1997) thematic contribution, current research seems to lack
examination of the effects of other types of clustering. We thus present another type of vocabulary set presentation in
Section 3 below.
3. Phonological clustering: another method of vocabulary presentation
It has been argued that entries in the lexicon are mainly of a phonological nature in both L1 and L2 (Frost, 1998;
Duyck, 2005), even when participants are presented the words only visually (Szmalec et al., 2009). Furthermore, L1
research on novel or nonword learning shows that memory for lexical items is strengthened when a repetitive
phonological aspect is present among words (Szmalec et al., 2009). Little is known, however, about the effect of
phonological repetition across words in L2 vocabulary acquisition.
L1 research on the Hebb repetition effect shows that repetitive aspects embedded in information presented for serial
recall facilitates information retrieval. In his seminal investigation, Hebb (1949) found that by repeating a nine-digit
list every third trial among other numerical lists, participants were progressively more able to recall the repeated list
even though they were not informed, and for some participants, aware, of the repetition. The Hebb effect refers to the
findings that recall performance significantly increases for sequences that are repeated than for those that are not. This
Hebb repetition effect is argued to be applicable to novel word learning (Szmalec et al., 2009). Specifically, L1 researchers have provided evidence for a connection between the Hebb repetition effect and the learning of phonological
word forms by repeated presentation (Page and Norris, 2009).
Hebbs (1949) model shows that the repeated sequences encoded in short-term memory (STM) (or the phonological loop of working memory for verbal information, e.g., Baddeley, 2003) generate traces in long-term memory
(LTM) that outlive short-term storage. The recall of the repetition gradually generates a stable representation in LTM.
The Hebb effects implications in L1 novel and nonword learning is applicable to the current study given that they may
also be pertinent to L2 vocabulary acquisition. The following brief literature review is provided in order to justify the
testing of another method of vocabulary presentation, namely, phonological clustering.
Short-term memory is claimed to be a precursor in moving information to LTM during acquisition, and repeated
presentation of information increases retention levels (Baddeley, 1990; Melton, 1963). With this in mind, Szmalec
et al. (2009) presented a study that connects STM and LTM theory to the Hebb effect and produced what they call a
laboratory analogue of naturalistic novel word learning. In Experiment 1 of their study, Dutch participants were

A. Wilcox, A. Medina / System 41 (2013) 1056e1069

1061

visually presented with nonsense sequences of consonant-vowel (CV) syllables for immediate serial recall. As
expected, participants were more accurate in recalling repeated sequences than filler sequences (i.e., sequences
viewed only a single time). In Experiment 2, given an auditory lexical decision task, participants were instructed to
determine whether the words they heard were words or nonwords. Participants took longer to reject nonwords with
these repeated sequences than other sequences, since they seemed to have formed some type of subvocal phonological representation of the visually recurring segments in the first experiment and applied them in the second
experiment to the auditory task. Szmalec et al. (2009) thus demonstrated that Hebb repetition sequences enter the
mental lexicon in the same way that novel word forms do. They also found that phonological repetition aids lexical
retrieval, which supports the argument that greater repetitions allow for increased subvocal rehearsal in the
phonological loop of working memory, thereby making retention crucially dependent on phonological characteristics (Baddeley, 2003).
The results above generally support the findings of Mosse and Jarrold (2008). Using a sample of English-speaking
children, Mosse and Jarrold (2008) correlated the magnitude of Hebb learning with word and nonword pairedassociate learning. In the paired-associate word learning task, images were linked with familiar names (e.g.,
Simon, Michael); in the nonword learning task, the phonemes of the aforementioned familiar names were then
rearranged to create nonwords. In up to ten trials, learners named each image, thereby repeating Hebb sequences.
Mosse and Jarrold found that nonword, but not word, learning was significantly related to the magnitude of Hebb
learning, thereby providing evidence suggesting that the Hebb repetition effect involves the same processes as those
involved in novel word learning. These findings are similar to those of Szmalec et al. (2009) and further support a
rationale for connecting this branch of research with L2 vocabulary presentation. In fact, we argue that these
domain-general principles can be taken from the Hebb strand of cognitive research and applied to new word learning
in L2.
Learning a novel word requires retention of its phonemes in the correct serial order (Mosse and Jarrold, 2008). One
of the principal objectives of the current study, therefore, is to use the implications of this line of research to test
phonological category effects on vocabulary presentation in an L2. Hebb research shows that when a series has a
repetitive aspect and participants are required to recall these series, learning is positively affected. If recalling a series
with a repetitive aspect facilitates non/novel word learning in L1, then it is reasonable to investigate if L2 words can be
acquired in a similar way. The question is whether repeating a combination of phonemes in a group of words will be
enough to trigger and then create associations that carry novel recalled words from STM into LTM. Since there are no
studies that investigate the above question in L2, this study not only provides further evidence to support or refute
current research on the role of semantic clustering in vocabulary learning, but also provides preliminary results in a
new line of investigation on the effects of phonological categorization.

4. Research question and hypotheses


Given the previously mentioned gaps in the literature, the following research question and hypotheses guided our
study:
Research Question: What are the effects of clustering (semantic, phonological) on L2 vocabulary learning
among novice-level learners of Spanish immediately after practice and over time?
Hypothesis 1 Based on the findings of Tinkham (1993, 1997), Waring (1997), Finkbeiner and Nicol (2003), and
Papathanasiou (2009), we predict that vocabulary presented in the semantically clustered conditions
will be more difficult to learn in the short term than vocabulary presented in a semantically nonrelated fashion. According to Schneider et al. (2002), this difficulty could facilitate better long-term
retention.
Hypothesis 2 We also predict that the phonologically similar groups will be easier to learn than the nonphonologically clustered sets since the participants will be exposed to the repetition of serial information that is conducive to novel word learning (Mosse and Jarrold, 2008; Szmalec et al., 2009).
Athough there is no L2 precedent from which to speculate, it seems reasonable that since one of the
basic features of the Hebb effect is its participation in the process of moving information from STM
to LTM, participants will better retain those word pairs in the long term.

1062

A. Wilcox, A. Medina / System 41 (2013) 1056e1069

5. Methods
5.1. Participants
Fifty-four native-English speaking students at a university in the southeastern United States taking the first level of
Elementary Spanish originally attended the first session; only 38 of those participants remained in the final participant
pool. Selection for inclusion in the study was based on meeting three criteria: attending all three sessions, claiming
English as ones native language, and scoring 0 on the pretest. Of the 38, six participants took part in a think-aloud
protocol where their verbal thoughts were recorded as they completed the second session (for more details, see Section
5.4 below). The recordings were collected to obtain a verbal account of participants vocabulary learning strategies in
each clustered condition. In an effort to avoid any type of interaction, these participants data were collected separately; their scores were not included in the following analyses. Data from the remaining 32 participants were used in
the statistical analyses.
5.2. Experiment design and materials
In a pre-post-delayed design, with time and category as independent variables, each participant was tested on four
categories that consisted of sets of five words that were semantically linked or not, and phonologically similar or not:
[semantic, phonological]. Participants were given an L1 English prompt and were asked to produce the written L2
Spanish equivalent, which, according to Schneider et al. (2002), is a more difficult task than producing the L1
equivalent and should aid in long-term retention.
During the first session, participants completed a background questionnaire and a pretest on the computer using EPrime (2.0). The pretest consisted of an L1eL2 translation task of 30 nouns typically found in an elementary textbook.
Twenty of those words were target lexical items and 10 were distractors. The English prompts were presented one at a
time. Participants were asked to type in the Spanish form of the word if they knew it or thought they knew it.
Otherwise, they were asked to type a question mark (?) in the response field. Participants were not given any feedback.
During the second session, one week later, learners were trained only on the target vocabulary, which was controlled
for syllable length. The target words were two or three syllables long and were grouped into four categories, based on
whether the words were semantically linked or not and whether or not each word shared the same beginning sounds (see
Appendix). Words in the [semantic, phonological] ([S  P]) category all fell under the superordinate concept
tools including Spanish words for drill (taladro), saw (serrucho), hammer (martillo), etc., but had no common wordinitial phonemic cluster. This type of semantic cluster (i.e., categorical rather than, for example, a set of synonyms) was
chosen to best imitate the types typically seen in elementary Spanish classes, though careful attention was paid to word
choice to avoid presenting participants with vocabulary already introduced to them during the course of their study.
Words in the [semantic, phonological] ([S  P]) group, were chosen based on their lack of similarity both in
meaning and sound, such as knuckle (nudillo) and thunder (trueno). The [semantic, phonological] ([S P])
category was a grouping of words that did not share meaning, but did share common beginning phonemes like wood pile
(tinada) and cheap red wine (tintorro). And finally, the words in the [semantic, phonological] ([S P]) group
shared not only the common beginning of the word, but also shared meaning in that each one pertained to the superordinate concept torment, a thump or blow (mazazo) and dungeon (mazmorra) are two examples. The mean syllable
length in each group was as follows: [S  P] 3, [S  P] 3, [S P] 3, and [S P] 2.6.
On a computer monitor, word pairs along with the sentence ________ means_______, as seen in Tinkham (1993,
1997), appeared on the screen for approximately two seconds following Schneider et al. (2002). Tinkhams sentence
was modified slightly for the [S  P] group to A__________ is a __________. This modification was made to
ensure participants did not confuse nouns in that category with verbs. Once participants viewed a conditioned 5-pair
set of vocabulary words, they were assessed on that set of words. During this partial assessment, participants saw an
English prompt and were asked to produce the Spanish equivalent. Once the test on those five words was complete,
participants continued on to the next condition for a new set of five word-pairs. Participants were tested again, in the
manner stated above, and the process was repeated until all 20 target words were presented and tested. That was
considered one trial. Participants did three trials, with the order of words within the 5-pair sets and the order of sets for
each trial randomized, following Schneider et al. (2002). Although participants were not given any feedback, the
intermediate recall of vocabulary followed models provided by Finkbeiner and Nicol (2003) and Schneider et al.

A. Wilcox, A. Medina / System 41 (2013) 1056e1069

1063

(2002) and was also supported by Nation (2001). Nation affirms that having to recall a meaning when seeing or
hearing a particular cue strengthens the form-meaning connection. After the third trial, participants took the immediate posttest, which consisted of the 20 target words in random order presented in English one-by-one on the
computer screen. Participants typed in the Spanish equivalent. Words were not presented in the same order as in the
pretest.
Approximately two weeks later (following Papathanasiou, 2009), participants returned to take the delayed posttest,
which assessed participants on the 20 target vocabulary words. Testing learners two weeks after completing the
training in an effort to see what was retained over time helped expand on previous research that did not include a longterm component. Each of the three sessions lasted approximately 20 min, and the final session concluded with a postexperiment questionnaire.
5.3. Scoring procedure
Two separate scoring methodsdstrict and lenientdwere used and analyzed. In the strict scoring procedure,
participants were awarded 1 point for exact reproduction of the target word and 0 points otherwise. No partial credit
was given for spelling inaccuracies. Comparisons were made in the number of correct answers for each condition, and
the change in accuracy over the course of the sessions. In the lenient scoring procedure, participants received 1 point
for accurately producing 2 syllables in three-syllable words and 1 syllable in two-syllable words. This means that
participants had to accurately spell 2 of the 3 (or 1 of the 2) syllables correctly to receive a point. Zero points were
awarded otherwise. The lenient scoring procedure was included in an attempt to account for a broader definition of
vocabulary learning.
5.4. Think-aloud protocol
For the think-aloud protocol (TA), six participants were asked to verbalize their thoughts while practicing the target
vocabulary during the second session. Each participant received special instructions and practiced thinking aloud prior
to beginning the target vocabulary task. During the TA training session, participants were given a written example of a
TA (i.e., they read a passage of what a TA might sound like, using a simple math problem to exemplify). As practice
with a language task, they were then given three Spanish-English word-pairs and were asked to say out loud whatever
passed through their minds as they tried to remember the new words. Once training was complete, participants began
the target vocabulary task while verbalizing their thoughts. Oral data were collected using headsets and Audacity
software. Examples of the recordings can be found at the end of Section 6 below.
6. Results
Since each participant scored 0 on the pretest, data from that assessment were not used in the analysis. The strict
scoring procedure data2 from the 32 non-TA participants were submitted to a 4  2 repeated measures analysis of
variance (ANOVA), with category ([S P], [S P], [S  P], [S  P]) and time (immediate and delayed
posttests) as the within-subjects variables and .05 as the alpha value. A significant main effect was found for time, F(1,
31) 101.21, p < .001, h2p .77, and category F(3, 31) 19.84, p < .001, h2p .39, and a significant
Time  Category interaction also surfaced, F(3, 31) 7.71, p < .001, h2p .20. The effect size (partial eta-squared)
was large in each case (.77, .39, and .20, respectively). See Fig. 1 for a plot graph of the overall means. The main effect
for time shows that participants scored significantly lower two weeks after vocabulary training than immediately
following it, as would be expected.
In order to gain a clearer understanding of the interaction between categories, separate one-way ANOVAs were
carried out for the immediate and delayed posttests with category as the independent variable. To control for Type I
errors that can result from the use of multiple tests, a corrected Bonferroni alpha value of .025 was used to determine
significance of the individual one-way ANOVAs. As illustrated in Table 1, the analysis of the immediate posttest
scores showed a significant main effect for category, F(3, 31) 17.41, p < .001. Descriptive statistics of the
2
The data from the lenient scoring procedure were also submitted to a repeated measures ANOVA. However, the same statistical patterns
emerged from this analysis as from the strict scoring procedure. Therefore, tables and figures display results from the strict scoring procedure only.

A. Wilcox, A. Medina / System 41 (2013) 1056e1069

1064

Fig. 1. Plot graph of mean correct over time. Note: IMPOS Immediate posttest; DLPOS Delayed posttest.

immediate posttest (Table 2) show [S  P] words to be by far the most difficult to learn (M 1.38); that difference
was significant when compared to the other three groupings. Though the [S  P] group fared the best (M 3.03),
analysis also showed that both phonological groups of words, [S P] (M 2.16) and [S P] (M 2.38), were
significantly improved when compared to the [S  P] cluster, but that neither phonological cluster was significantly
different when compared to each other. Though the phonological clusters fared better than the [S  P] category,
they still did not fare statically as well as the [S  P] category.
Finally, the effects of category in the delayed posttest were examined (see Table 2 for the descriptive statistics). It is
important to note that retention was generally poor for all four categories. This is understandable since participants
practiced the words for only 20 min two weeks prior to the delayed posttest assessment. All students also stated on the
post-experiment questionnaire that they did not practice these words outside the lab. Even with the poor retention,
however, there was also a significant main effect for category in the delayed posttest, F(3,31) 3.56, p .017. The
results of this assessment follow the tendencies presented earlier, but vary slightly. The [S  P] category still had the
lowest average (M .06) of all the categories, but this was not significantly different (p .161) from the retention
level of the [S P] category (M .25). And though the [-S-P] cluster again had the highest average (M .47)
overall in the delayed posttest, this average was not significantly different from either of the phonologically similar
categories, [S P] p .109, [S P] p .52. It seems that in the long term, participants remembered little in
general, but significantly more in the [S  P] and [S P] categories than in the [S  P] cluster.
In the real world, partial recollection is usually more valuable than no recollection at all. In this case it seems
pertinent to look at the lenient scores for the delayed posttest. Using the lenient scoring method, partial learning is
accounted for, which, after such a long break with no new practice or priming before the final assessment, is warranted. The means (and standard deviations) of the lenient scoring data are as follows: immediate posttest [S  P]

Table 1
One-way ANOVA results for category comparisons on the immediate and delayed posttests.
Source

df

Sum of squares

Mean squares

Partial eta2

Immediate
Delayed

3
3

44.78
2.96

14.93
.99

17.41
3.56

.000*
.017*

.36
.10

*p < .025 (corrected Bonferroni value).

A. Wilcox, A. Medina / System 41 (2013) 1056e1069

1065

Table 2
Descriptive statistics for mean correct for each category in the immediate and delayed posttests.
Category

[S
[S
[S
[S




P]
P]
P]
P]

Immediate posttest

Delayed posttest

Mean

SD

Mean

SD

1.38
3.03
2.16
2.38

.23
.25
.28
.25

.06
.47
.25
.38

.25
.67
.67
.55

N 32; min 0; max 5.

M 1.88 (1.28), [S  P] M 3.78 (1.24), [S P] M 3.13 (1.60), [S P] M 3.22 (1.07); delayed posttest
[S  P] M .28 (.52), [S  P] M .91 (1.00), [S P] M .72 (1.20), [S P] M 1.13 (.79). The results of
the delayed posttest analysis (using one-way ANOVAs) actually mimic the results seen in the strict scoring of the
immediate posttest: both phonological categories had significantly higher means when compared to the [S  P]
word group. Neither phonological category had a significantly different mean from the [S  P], so the phonological
categories were arguably as easy to remember as the [S  P] category in the long term. These results provide an even
stronger case for the proposal that grouping vocabulary either randomly or phonologically could better facilitate longterm retention than presenting words exclusively clustered semantically ([S  P]).
Several interesting observations were made after listening to the recordings of the six TA participants. First, there
was an extremely marked difference in recording content between participants. For example, either the participants
consistently did nothing other than repeat the Spanish and English words and say ____ means ______, for the
duration of the practice session, or the participants attended to the phonological or orthographical aspects of the word
and tried to make some type of sense out of it. For example, if the target word was trueno (thunder), then a
participant might say something like, the letters are almost the same in both words, thunder.theres a storm, true,
no? Notice that when the space is removed from true, no? trueno is formed. Another example was taladro is
drill.that sounds like la drill o.taladro. More than one participant used the double m in hammer to help them
remember martillo, which has a double l. The point here is that many learners largely looked for similar sounds and
spellings to help them remember the different sets of words. But one participant also said, mazo, maza, mazorca.thats a lot of ms.Im confused. This statement might help explain the difference in means between the
[S  P] and the [eS P] and [S P] categories. No participant mentioned noticing the superordinate grouping of
words during the think aloud.
7. Discussion and conclusion
The research question investigated in this study sought to observe effects of lexical clustering on L2 vocabulary
learning. Clustering seems to have a strong impact on L2 vocabulary learning among novice-level Spanish participants
in the short term and some effect in the long term. These results provide partial support for the hypothesis presented.
First, the results speak to the notion that semantically clustered word groupings are more difficult to learn than
nonrelated groups of words among novice learners (Tinkham, 1993, 1997; Waring, 1997; Finkbeiner and Nicol, 2003;
Papathanasiou, 2009; Erten and Tekin, 2008), thus suggesting support for interference theory (Baddeley, 1990; Higa,
1963). This difficulty in the short term did not, however, translate into better long-term retention as had been posited
by Schneider et al. (2002). The two phonologically linked categories consistently had significantly higher means than
the [S  P] category, and significantly lower means than the [S  P] category on the immediate posttest but not on
the delayed, and were not significantly different from one another. The following discussion seeks to explain these
findings.
For proponents of semantic clustering, the semantic set in question e tools e should have arguably been a fairly
easy one to learn, but it proved more difficult for participants. Many learners stated in their post-experiment questionnaire that this particular group of words was harder than the others. About half stated it was because Spanish
forms were so different from their English forms; the other half did not know why they seemed more difficult.
Though Channell (1981) provides arguments concerning how words are organized and then stored in the brain, her
conclusion does not seem to apply to the novice learner. Waring (1997) explains that novice learners must first create

1066

A. Wilcox, A. Medina / System 41 (2013) 1056e1069

the L2 semantic and vocabulary knowledge network (i.e., semantic fields) into which new vocabulary must be put
(p. 269). It seems that at the novice level, even though the mind may tend to store accumulated vocabulary in semantic
fields, initially it is prepared to receive words that need to be organized in that way, not to receive words that have
already been organized semantically.
This may be because beginning L2 learners, by definition, have little background in the target language. A common
theme in the argument for semantic categorization is the reliance on background knowledge. However, it is difficult to
anchor new linguistic information to old (previously learned) information when most of the target linguistic information is overwhelmingly new.
The need for the mind to organize words on its own rather than receive words pre-arranged in semantic sets at the
novice level explains the findings of most empirical studies on the subject (e.g., Tinkham, 1993, 1997; Finkbeiner and
Nicol, 2003; Papathanasiou, 2009; Waring, 1997). Specifically, no advantage is generally found for semantic sets over
mixed sets at the beginning level because L2 learners at this stage do not have a sufficient lexical base on which to
anchor incoming lexical stimuli to be able to compare and contrast the similarities efficiently. This explanation, in fact,
helps to explicate the contrasting results of Hashemi and Gowdasiaei (2005), who found that intermediate learners
benefited more from semantic set presentation than unrelated set presentation. It is argued that while novice L2
learners must first create the semantic and lexical knowledge network (Waring, 1997), intermediate L2 learners or
above already have an existing L2 network, and thus are able to anchor novel lexical information to previously learned
information to facilitate processing and storage of the words. A similar proposal was put forth by Papathanasiou
(2009), who posited that intermediate or more advanced learners would likely have pre-existing knowledge of
many words from the semantic groups in question, and thus need only to add new words to an existing store rather than
create one completely anew (p. 319).
In comparison to [S  P], both phonologically similar groups demonstrated significantly better performance on
the immediate and delayed posttests, whether or not the words were semantically linked. It seems fair to say that the
phonological category effect was positive in the short term when comparing [S P] to the [S  P] group, and in
the long term, if scoring from the lenient procedure is considered.
It appears that some of the domain-general mechanisms of L1 word learning are applicable to L2 vocabulary
learning, as the results provide some support for the Hebb effect in SLA. This is an interesting new finding because it is
a reminder of the role phonology can play in SLA, especially in novel word learning. The phonological similarities
between the select words seem to be enough to have aided in their learning. It should be obvious that phonology would
be an important factor in SLA, but there is a general lack of examination of its effects on L2 vocabulary acquisition.
Grouping words together based on phonological similarities makes that repetition more salient to the learner, and
repeating phonological information seems to increase the chances of learning.
It is not uncommon for L1 vocabulary to be presented to early readers in phonemic groups (e.g., play, stay, way),
probably for the above reason. It is curious, though, that there was no significant difference between the
two phonological categories in the immediate or delayed posttests. One might have expected [S P] to perform
significantly worse than [S P] given that semantic clustering negatively affected performance when examining
nonphonologically similar groups ([S  P] and [S  P]). This lack of significance, however, might be explained by
word choice.
A closer look at the [S P] category reveals that it is the only group that contained two words whose meaning
was changed by a single phoneme (in maza and mazo). Not only are they only a phoneme apart, but several students
responded on their post-experiment questionnaire that these words were particularly easy to remember because they
were so similar in spelling and in terms of their English equivalents, mace and mallet. If two of the five words of that
group were particularly easy to remember,3 it makes sense that their learning would be increased as well, perhaps
unfairly so when compared to its nonsemantically related counterpart, [S P], thus potentially creating no significant difference between the two groups when maybe there really is. Even so, the increased means of the [S P]
category when compared to [S  P] on both the immediate and delayed tests indicate that phonological clustering
can be beneficial to L2 Spanish learners at the novice level.

3
A post-hoc examination of the immediate posttest responses in the [S P] group supports this notion. Over twenty participants correctly
responded with maza (N 29) and mazo (N 23). In comparison, only 5, 7, and 12 participants responded correctly to mazmorra, mazorca, and
mazazo, respectively.

A. Wilcox, A. Medina / System 41 (2013) 1056e1069

1067

8. Implications, limitations, and future research


L2 research, including the study presented herein, has not found convincing support for clustering words
semantically. Most studies, in fact, note that semantic clusters serve as a detriment to novel word learning. After
reviewing 10 current elementary Spanish textbooks from various publishers, we found that all of them group vocabulary into semantic clusters. Although the manner in which this vocabulary is presented to students may vary from
teacher to teacher, many students are nonetheless asked to study lexical items in semantic fields already created for
them. For the benefit of novice L2 learners, therefore, L2 curriculum coordinators or language educators who base
their vocabulary presentation on semantic clusters may need to reconsider presentation methods and instead explore
the possibility of facilitating vocabulary learning with semantically and syntactically unassociated words (as evident
by our [S  P] and [S  P] comparisons) or other presentation methods, such as with thematic clusters, as argued
in Tinkham (1997), or phonological clusters, as supported with by our [S P] and [S  P] comparisons. Language
textbook writers, furthermore, should consider the building evidence in support of these presentation methods (thematic or phonological) for the beginning stages of adult L2 acquisition. Nation (2000) recommends that instructors
take frequency of usage into account. Preparing lexical groups based on usage frequency could provide a bank of
vocabulary words that are not semantically clustered. Song lyrics and simple poetry could also be used as resources for
random and/or phonologically similar vocabulary.
Grouping words based on phonology is a common practice in teaching early readers (e.g., Kindergarten
through Grade 2) in the L1. Those students are learning how to spell and read a new label for a word/concept with
which they are familiar. L2 students have to do that as well, but they are presumably also focused on producing
the new label and understanding it when it is presented to them. It seems that replicating the presentation practice
of L1 novice word learning could be beneficial to L2 students by making the sound of the target words a more
salient feature.
In terms of the phonological categories, this study provides preliminary results that connect theory pertaining to
brain mechanisms used in general L1 novel word learning to L2 learning of vocabulary. Considering the fact that the
phonologically grouped words showed a marked improvement compared to the semantically clustered nonphonologically similar words, this presentation method justifiably calls for further research. Also serving as motivation for future investigation is the limited generalizability of the present study to lengthier processing times (greater
than two seconds per trial), an increased word base (greater than five lexical items per set or greater than one set per
category), other pedagogical approaches (contextualized instruction rather than rote memorization), and other
assessment measures (receptive rather than productive). Given that the repetitive aspect among [P] words seems to
have alleviated learners cognitive load during written production relative to the [S  P] words, the query as to
whether the same can be said in a receptive task (e.g., L2 to L1 translations) remains an open question.
In future research, sets of semantically linked words that are near synonyms can be explored (as in the nonquantitative classroom study of Jullian, 2000). After all, it maybe the case that semantic clustering based on categories such as tools differ from groupings based on near synonyms, such as words whose meanings approximate that
of to hit (e.g., strike, punch, smack), especially if learners are exposed to a componential analysis of those lexical
items. Additionally, semantic clusters could be chosen by learners in a pilot session, given that semantic fields can
differ between languages, meaning that what constitutes a semantic cluster in L1 may not constitute a semantic cluster
in L2. For example, participants might make different lists of common fruits depending of the fruits typically available
to them in their region (e.g., dragon fruit, tamarillo, mangosteen, and genip maybe prevalent in Colombia, but perhaps
not in the Midwest of the United States). These considerations would enable researchers to create semantic groupings
that are clear and sensible to their particular population. New sets of words with phonologically similar endings,
instead of beginnings, could also be examined in order to see if word-final phonological similarity can likewise impact
vocabulary acquisition. In the think-aloud protocols, a number of references were made to spelling. Research into
whether it is orthographical or phonological similarity, or both (graphophonemic), that primarily contributes to the
results should also be carried out by examining, for example, aural presentation rather than visual, or graphemes that
are different from each other but are associated with the same phoneme (such as the graphemes v and b, which are
both associated with /b/ in Standard Spanish). It would benefit future research to include a larger sample of TA
participants, which would give a more comprehensive view of learner strategies and processing while learning vocabulary. Understanding how vocabulary is best presented and learned should help pedagogues and learners alike in
achieving their shared goal of second language acquisition.

1068

A. Wilcox, A. Medina / System 41 (2013) 1056e1069

Appendix
Word clusters
[S  P]
Drill e taladro
Hammer e martillo
Chisel e escoplo
Rachet e carraca
Saw e serrucho
[S  P]
Knuckle e nudillo
Broom e escoba
Thunder e trueno
Deer e venado
Sword e espada
[S P]
Inkwell e tintero
Cheap red wine e tintorro
Water tank e tinaco
Large earthenware jar e tinaja
Wood pile e tinada
[S P]
Mace e maza
Mallet e mazo
Dungeon e mazmorra
Torture e mazorca*
Thump/blow e mazazo

A taladro is a drill.
A martillo is a hammer.
An escoplo is a chisel.
A carraca is a ratchet.
A serrucho is a saw.
Nudillo means knuckle.
Escoba means broom.
Trueno means thunder.
Venado means deer.
Espada means sword.
Tintero means inkwell.
Tintorro means cheap red wine.
Tinaco means water tank.
Tinaja is a large earthenware jar.
Tinada means wood pile.
Maza means mace.
Mazo means mallet
Mazmorra means dungeon
Mazorca means torture
Mazazo is a thump or blow

*Mazorca generally means cob (e.g., mazorca de maz corncob). However, mazorca is also used by some in Argentina to mean torture, a
meaning that is derived from a Rosista military group named Mazorca. Participants were informed of this dual meaning in the post-experiment
questionnaire.

References
Ausubel, D., 1968. Educational Psychology: a Cognitive View. Holt, Rinehart & Winston, Inc, New York.
Baddeley, A.D., 1990. Human Memory Theory and Practice. Erlbaum, Hillsdale, NJ.
Baddeley, A.D., 2003. Working memory and language: an overview. J. Commun. Dis. 36 (3), 189e208.
Channell, J., 1981. Applying semantic theory to vocabulary teaching. ELT J. 35 (2), 115e122.
Crow, J.T., Quigley, J.R., 1985. A semantic field approach to passive vocabulary acquisition for reading comprehension. TESOL Q. 19 (3),
497e513.
Duyck, W., 2005. Translation and associative priming with cross-lingual pseudohomophones: evidence for nonselective phonological activation in
bilinguals. J. Exp. Psychol. Learn. Mem. Cogn. 31, 1340e1359.
Erten, I.H., Tekin, M., 2008. Effects on vocabulary acquisition of presenting new words in semantic sets versus semantically unrelated sets.
System 36 (3), 407e422.
Finkbeiner, M., Nicol, J., 2003. Semantic category effects in second language word learning. Appl. Psycholinguistics 24 (3), 283e369.
Frost, R., 1998. Toward a strong phonological theory of visual word recognition: true issues and false trails. Psychol. Bull. 123, 71e99.
Grandy, R., 1992. Semantic fields, prototypes, and the lexicon. In: Lehrer, A., Kittay, E.F. (Eds.), Frames, Fields, and Contrasts: New Essays in
Semantic and Lexical Organization. Lawrence Erlbaum Associates, Hillsdale, NJ, pp. 103e122.
Hashemi, M., Gowdasiaei, F., 2005. An attribute-treatment interaction study: lexical-set versus semantically-unrelated vocabulary instruction.
RELC J. 36, 341e361.
Haycraft, J., 1993. An Introduction to English Language Teaching. Longman, Malaysia.
Hebb, D.O., 1949. The Organization of Behavior. Wiley, New York.
Higa, M., 1963. Interference effects of intralist and word relationships in verbal learning. J. Verbal Learn. Verbal Behav. 2, 170e175.
Hoshino, Y., 2010. The categorical facilitation effects on L2 vocabulary learning in a classroom setting. RELC J. 41, 301e312.
Hunt, R.R., Elliot, J.M., 1980. The role of nonsemantic information in memory: orthographic distinctiveness effects on retention. J. Exp. Psychol.
Gen. 109, 49e74.
Hunt, R.R., Mitchell, D.B., 1982. Independent effects of semantic and nonsemantic distinctiveness. J. Exp. Psychol. Learn. Mem. Cogn. 8, 81e87.
Jullian, P., 2000. Creating word-meaning awareness. ELT J. 54 (1), 37e46.
Lehrer, A., 1974. Semantic Fields and Lexical Structure. North-Holland, Amsterdam.

A. Wilcox, A. Medina / System 41 (2013) 1056e1069

1069

Melton, A.W., 1963. Implications of short-term memory for a general theory of memory. J. Verbal Learn. Verbal Behav. 2, 1e21.
Miller, G.., Johnson-Laird, P.N., 1976. Language and Perception. Belknap Press of Harvard University Press, Cambridge, MA.
Mosse, E.K., Jarrold, C., 2008. Hebb learning, verbal short-term memory, and the acquisition of phonological forms in children. Q. J. Exp.
Psychol. 61 (4), 505e514.
Nation, P., 2000. Learning vocabulary in lexical sets: dangers and guidelines. TESOL J. 9 (2), 6e10.
Nation, P., 2001. Learning Vocabulary in Another Language. Cambridge University Press, Cambridge, UK.
Nida, E.A., 1975. Componential analysis of meaning: an introduction to semantic structures. Mouton, The Hague.
Page, M.P.A., Norris, D., 2009. A model linking immediate serial recall, the Hebb repetition effect and the learning of phonological word forms.
Philosophical Trans. R. Soc. 364, 3737e3753.
Papathanasiou, E., 2009. An investigation of two ways of presenting vocabulary. ELT J. 63 (4), 313e322.
Schneider, V., Healy, A., Bourne, L., 2002. What is learned under difficult conditions is hard to forget: contextual interference effects in foreign
vocabulary acquisition, retention, and transfer. J. Mem. Lang. 46, 419e440.
Seal, B.D., 1991. Vocabulary learning and teaching. In: Celce-Murcia, M. (Ed.), Teaching English as a Second or Foreign Language, second ed.
Heinle and Heinle, Boston, pp. 296e311.
Stoller, F., Grabe, W., 1995. Implications for L2 vocabulary acquisition and instruction from L1 vocabulary research. In: Huckin, T., Haynes, M.,
Coady, J. (Eds.), Second Language Reading and Vocabulary Learning. Ablex Publishing Corporation, Norwood, NJ, pp. 24e45.
Szmalec, A., Duyck, W., Vandierendonck, A., Mata, A., Page, M., 2009. The Hebb repetition effect as a laboratory analogue of novel word
learning. Q. J. Exp. Psychol. 62 (3), 435e443.
Tinkham, T., 1993. The effect of semantic clustering on the learning of second language vocabulary. System 21 (3), 371e380.
Tinkham, T., 1997. The effects of semantic and thematic clustering on the learning of a second language vocabulary. Second Lang. Res. 13 (2),
138e163.
Waring, R., 1997. The negative effects of learning words in semantic sets: a replication. System 25 (2), 174e261.
Wharton, S., Race, P., 1999. 500 Tips for TESOL: Teaching English to Speakers of Other Languages. Kogan Page Limited, London.

You might also like