See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/264781280

Bottom-Up or Top-Down: English as a Foreign Language Vocabulary

Instruction for Chinese University Students

Article · June 2014

DOI: 10.1002/tesq.170

CITATIONS READS

3 1,620

4 authors, including:

Christo Moskovsky Guowu Jiang

University of Newcastle 1 PUBLICATION   3 CITATIONS   
26 PUBLICATIONS   97 CITATIONS   
SEE PROFILE
SEE PROFILE

Alan Reed Libert

University of Newcastle
37 PUBLICATIONS   25 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

1.EAP (English for Academic Purposes) instruction: The differences between Australia and China in relation to the provision of EAP courses online View project

All content following this page was uploaded by Guowu Jiang on 04 May 2018.

The user has requested enhancement of the downloaded file.

Bottom-Up or Top-Down: English as a
Foreign Language Vocabulary
Instruction for Chinese University
Students
CHRISTO MOSKOVSKY, GUOWU JIANG, ALAN LIBERT, AND
SEAMUS FAGAN
University of Newcastle
Callaghan, New South Wales, Australia

Whereas there has been some research on the role of bottom-up and
top-down processing in the learning of a second or foreign language,
very little attention has been given to bottom-up and top-down
instructional approaches to language teaching. The research
reported here used a quasi-experimental design to assess the relative
effectiveness of two modes of academic English vocabulary instruc-
tion, bottom-up and top-down, to Chinese university students
(N = 120). The participants, divided into two groups—bottom-up
and top-down—were exposed to 48 hours of explicit vocabulary
instruction. Their achievement was measured with two vocabulary
tests, Academic Vocabulary Size and Controlled Productive Knowl-
edge, administered at the start (T1) and at the end (T2) of the treat-
ment. Analyses of the test scores reveal that at T2 the bottom-up
group slightly outperformed the top-down one on both vocabulary
size and controlled productive knowledge. With respect to the for-
mer, the bottom-up group’s superiority was found to be statistically
significant, although with a relatively small effect size (g2 = .05).
doi: 10.1002/tesq.170

B ottom-up and top-down processing are well-established concepts in

a wide range of fields, including psychology, cognitive science, peda-
gogy, and institutional management. They refer to two essentially differ-
ent ways of processing and/or organising information. Broadly
speaking, bottom-up is a form of inductive (or data-driven) processing
starting with smaller and/or lower-ranked units and moving upwards
through larger and/or higher-ranked units. Top-down is a form of
deductive (or schemata-driven) processing working in the opposite

TESOL QUARTERLY Vol. 0, No. 0, xxxx 2014 1

© 2014 TESOL International Association
direction: from higher-to lower-ranked units (Jay, 2003). It is important
to bear in mind that mental processing is almost never only bottom-up
or top-down; most mental operations would involve both (Sun, 2002).
Bottom-up and top-down have also been used with reference to lan-
guage processing. Bottom-up language processing works from pho-
nemes and morphemes through lexemes and phrases to clauses and
larger chunks of text. Top-down language processing works in the
opposite direction, from the overall message and text structure to
lower-ranked units (Celce-Murcia & Olshtain, 2001; Chaudron &
Richards, 1986; Matthews, 2007; Shohamy & Inbar, 1991).
Figure 1 shows how Celce-Murcia and Olshtain (2001) conceptualise
graphically the operation of bottom-up and top-down language pro-
cessing. The two types of language processing seem closely aligned
with the two principal forms of language use, perception and produc-
tion. Language perception by its nature seems to be essentially bot-
tom-up: Phonological, morphological, and lexical aspects of the
sentence/utterance must be processed before it is linked to concep-
tual-semantic content. Production seems to generally operate in the
opposite way: Starting with conceptual-semantic content and only then
assigning grammatical structure to it. De Bot, Paribakht, and Wesche’s
(1997) model of lexical perception and production shown in Figure 2
illustrates this. As widely recognised in relevant literature, language
perception is not exclusively bottom-up and neither is language pro-
duction only top-down, a fact which is clearly reflected by theories of
speech perception and production. Take, for instance, Levelt, Roelofs,
and Meyer’s (1999) speech production model.
As Figure 3 shows, in speech production the processing mode is
clearly from top to bottom, but note the self-monitoring dimension
of production, which is essentially bottom-up. In language perception,
on the other hand, the decoding of language input necessarily starts
in a bottom-up mode—with the processing of the phonemic and
morphemic features of the message (Carrell, 1988; Carrell & Eister-
hold, 1983; Rost, 1990), but when it comes to assigning a meaning to
it the individual’s general knowledge of the world is necessarily
invoked—in a top-down processing mode (Goodman, 1967; Smith,
2004). In addition, the language stimulus is often incomplete and/or
ambiguous, in which case the individual’s conceptual system would
also have to be accessed (in a top-down mode) to help disambiguate
the stimulus.1

1
Take the sentence He arrived in his Beetle. Bottom-up processing provides at least two
meanings for the lexeme beetle: an insect and a car model. The conceptual system rules
out (via top-down processing) the insect interpretation.

2 TESOL QUARTERLY
 FIGURE 1. Vocabulary and language knowledge (Celce-Murcia & Olshtain, 2001, p. 75).

FIGURE 2. Lexical comprehension and poduction model for oral and written modalities
(De Bot, Paribakht, & Wesche, 1997, p. 315).

BOTTOM-UP OR TOP-DOWN: EFL VOCABULARY INSTRUCTION 3

 conceptual preparation in
terms of lexical concepts

lexical concept

lexical selection

lemma
lemmas
MENTAL
morphological encoding LEXICON
word
forms
self- morpheme
monitoring

phonological encoding
syllabification

phonological word

phonetic encoding
SYLLABARY

phonetic gestural score

articulation

sound wave

FIGURE 3. Speech production theory outline (Levelt et al., 1999, p. 3).

The concepts of bottom-up and top-down processing are obviously

also relevant to the field of second/foreign language (L2) learning.
How language learners process different types of L2 input would be
expected to be of key importance for both second language learning
theory and practice. There has been some research on the role of bot-
tom-up and top-down processing in L2 reading and listening compre-
hension—notably, only the two perceptive skills. Researchers seem to
be divided in relation to which of the two processing modes, bottom-
up or top-down, contributes more to reading and listening compre-
hension (Field, 2004; Tsui & Fullilove, 1998).
Earlier studies seem to favour top-down processing. Anderson,
Reynolds, Schallert, and Goetz (1977), for instance, assert that “the
background knowledge structures [or ‘schemata’] that the reader
brings to a text are much more important than linguistic structures in
the text” (cited in Tsui & Fullilove, 1998, p. 434), therefore skilful
readers would presumably favour top-down processing. Hildyard and
Olson’s (1982) study also found that good listeners engaged in top-
down processing while poor listeners mainly attended to lower-ranked
elements. Similarly, in Shohamy and Inbar (1991, p. 29) the less profi-
cient listeners performed less successfully on global questions, that is,

4 TESOL QUARTERLY
questions which required broad understanding of the whole listening
passage, but they were quite good at answering local questions, that is,
questions which required attention to specific detail.
Others (e.g., Eskey, 1988; Perfetti, 1985) have argued that successful
language processing may be impossible without solid bottom-up skills,
suggesting that poor readers may in fact be over-relying on top-down
processing, because their bottom-up skills (e.g., word recognition) are
poor. In accord with that view, Tyler and Warren (1987) point out that
thorough comprehension can take place only when listeners success-
fully decode (in a bottom-up mode) the L2 input before integrating it
into their existing knowledge.
Yet others (Carrell, 1988; Mendelsohn, 2001; Stanovich, 1980) take
a more balanced view, arguing that reading and listening are very com-
plex mental operations necessarily involving both bottom-up and top-
down processing and that over-reliance on either processing mode can
lead to poor outcomes. Vandergrift (2004, p. 4) reinforces this view,
stating that “learners need to learn how to use both processes to their
advantage.”
Bottom-up and top-down also have a pedagogical dimension.
Instructional techniques can be designed as either bottom-up or top-
down (or mixed; Celce-Murcia & Olshtain, 2001). It is reasonable to
assume that L2 instruction which is aligned with the learners’ pre-
ferred mode of processing would be more effective. However, some-
what surprisingly, little attention seems to have been given to bottom-
up and top-down instructional approaches in L2 teaching, nor has
there been an attempt to establish empirically which one produces
better learning outcomes. The present research has specifically been
conducted to fill that gap. As part of this study, we used a quasi-experi-
mental between-group design to assess the relative effectiveness of bot-
tom-up and top-down instructional approaches to English academic
vocabulary teaching for Chinese university students.
Our focus on L2 vocabulary has been driven by two principal con-
siderations. Vocabulary instruction seems particularly well suited for an
L2 delivery which strictly involves either a bottom-up or a top-down
instructional design. In addition, in recent years there has been
increased recognition that explicit vocabulary teaching is linked to
improved learning outcomes, in relation not only to lexical knowledge,
but also to L2 competence generally (Read, 2004; Sonbul & Schmitt,
2010).2
2
S€
okmen (1997), among others, has argued convincingly in favour of explicit vocabulary
teaching, suggesting that implicit (spontaneous) vocabulary learning is by its nature a
very slow process and that it may be suitable only for advanced learners. S€
okmen also
refers to research evidence showing that inferring word meaning from context does not
necessarily result in long-term retention.

BOTTOM-UP OR TOP-DOWN: EFL VOCABULARY INSTRUCTION 5

 Besides, the issue of vocabulary competence in English as a foreign
language (EFL) seems to be of particular relevance to the Chinese
educational context. In China, teaching/learning English is an essen-
tial part of all levels of the country’s educational system. Language
education in China typically involves quite sizable classes of up to 60
students (or sometimes even more!). Despite recent efforts to intro-
duce the communicative approach more broadly, most language
instruction (including vocabulary) is rooted in traditional teaching
methods (grammar translation and/or audiolingual) and strongly
relies on rote learning. Class activities are typically dominated by the
teacher and commonly involve reading aloud, reciting, and repetition;
little or no time is devoted to interactive communicative tasks. Both
teachers and learners frequently resort to their first language (L1) for
clarifying aspects of L2 grammar and meaning (Ma, 2009). While in
secondary school there is a significant focus on developing learners’
grammatical knowledge, at university the students’ primary task is to
develop their vocabulary proficiency in English (Hu, 2002b; Ma, 2009;
Wang, 2010). There seems to be a widespread belief among university
language instructors that extensive reading is the best way to develop
vocabulary knowledge, while explicit teaching/learning is regarded as
relatively unimportant and deployed relatively infrequently. Many
among them, in fact, hold the view that learning vocabulary is the stu-
dents’ own responsibility, but students often lack a sound systematic
approach to learning vocabulary (Ma, 2009), and rote learning is com-
monly their most valued learning strategy (Gan, Humphreys, & Hamp-
Lyons, 2004; Ma, 2009).
The Chinese College English Syllabus (National College English
Testing Committee, 2013) specifies 4,200 English words as the mini-
mum operational vocabulary knowledge university students must have,
although some experts (e.g., Huang, Chen, Xu, Li, & Fu, 2004;
Nation, 2006) have suggested that this number may be too low. In
reality, very few Chinese students achieve even this relatively modest
level of competence in EFL vocabulary (Frederic, 1997).

RESEARCH DESIGN AND OBJECTIVES

The principal objective of our study was to examine the relative

effectiveness of two forms of L2 vocabulary instruction, bottom-up and
top-down. The study addressed the following research question: Which
of two instructional approaches to vocabulary teaching, bottom-up or
top-down, can produce better learning outcomes in relation to EFL
academic vocabulary competence for Chinese university students?

6 TESOL QUARTERLY
 The study involved a quasi-experimental pre- and posttest between-
group design. For the purposes of the study we recruited 120 Chinese
EFL learners who were first-year university students at Hebei Normal
University, in China. The study’s participants were allocated to two
research groups. Treatment in both groups involved 48 hours of expo-
sure to EFL academic vocabulary instruction over 8 weeks (6 hours
per week), but one of the groups was exposed to only bottom-up
vocabulary teaching, whereas the other was exposed to only top-down
vocabulary teaching. Two different vocabulary tests, Academic Vocabu-
lary Size Test and Controlled Productive Knowledge Test, were admin-
istered to participants at the beginning and at the end of the
treatment. The results from the pre- and posttests were then subjected
to descriptive and inferential statistical analyses. The study’s design is
graphically presented in Figure 4.
In summary, the study was designed in a way to ensure that all con-
ditions in the two groups (including allocation of participants, instruc-
tor, syllabus, teaching materials, venues, and technology used) were as
similar as possible, with the exception of the specific teaching
approach (bottom-up or top-down).

METHOD

Participants
The study’s participants were pooled from among first-year students
at Hebei Normal University, in Hebei province, China. We recruited
the first 120 students who volunteered to participate in the research.
The participants had a similar level of English proficiency, similar age,
and similar educational background, based on the selection criteria of
the university entrance exam of this university.
Hebei Normal University is a medium-ranked Chinese university
located in Shijiazhuang, the capital city of Hebei province. As the
name suggests, it is a teacher-training institution. In terms of its

FIGURE 4. A pre-post test between-group design.

BOTTOM-UP OR TOP-DOWN: EFL VOCABULARY INSTRUCTION 7

structure and operation, Hebei Normal University is a typical represen-
tative of tertiary institutions in China.
Participants were given an opportunity to self-allocate to one of
two learning groups, Group A and Group B. This was done to enable
them to be together in the same learning group with their friends or
classmates. Importantly, participants did not know in advance which
instructional design (bottom-up or top-down) would be used in
which group; in other words, their choice of group could not have
been made on that basis. Each of the two learning groups was fur-
ther subdivided into two classes to achieve a manageable class size
(see Table 1).
Subsequent examination of the composition of the two groups
showed that they were quite evenly matched along a range of demo-
graphic variables (including age, gender, level of education, parents’
level of education, ethnicity, primary language, urban/rural, etc.; see
Appendix S1 in the Supporting Information). Also, as established via
the vocabulary pretests, there were only relatively minor preexisting
differences in vocabulary competence between the two groups,
and in any case these differences were factored in in subsequent
analyses.
The vocabulary pretest scores (see Table 2, in the Results section)
also indicate that the participants’ proficiency level in English—at least
as far as their lexical competence was concerned—was relatively low.

Vocabulary Course

An English academic vocabulary course was constructed specifically

for the purposes of the study. The design of the course syllabus was
largely inspired by Coxhead’s (2006) and Nation’s (2001) ideas about
vocabulary teaching. The vocabulary course involved 48 hours of EFL
academic vocabulary instruction over 8 weeks (6 hours per week).
Instruction was delivered per the lesson plan included in Appendix S2
(in the Supporting Information online).

TABLE 1
Allocation of Participants Into Groups

Group A Group B
Group Classes (bottom-up approach) (top-down approach)
Class 1 30 30
Class 2 30 30
Total 60 60

8 TESOL QUARTERLY
TABLE 2
Descriptive Statistics for the AVST (Pre and Post) and CPKT (Pre and Post) Scores by
Groups and the Whole Sample

Group

Bottom-Up Top-Down Whole Sample

M Skewness Kurtosis M Skewness Kurtosis M Skewness Kurtosis

Test (SD) (SEskew) (SEkurt) (SD) (SEskew) (SEkurt) (SD) (SEskew) (SEkurt)
Pre- 29.65 .07 .25 27.67 .29 .49 28.66 .08 .3 8
AVST (4.88) (.31) (.61) (6.38) (.31) (.61) (5.74) (.22) (.44)
Post- 44.87 .68 34 42.70 .54 1.02 43.78 .68 .88
AVST (3.40) (.31) (.61) (4.20) (.31) (.61) (3.96) (.22) (.44)
Pre- 8.52 .11 .54 9.58 .43 .72 9.05 .12 .53
CPKT (2.30) (.31) (.61) (3.63) (.31) (.61) (3.07) (.22) (.44)
Post- 14.67 .55 .65 13.85 .42 .19 14.26 .40 .54
CPKT (2.62) (.31) (.61) (2.28) (.31) (.61) (2.48) (.22) (.44)

Note. SEskew = standard error of skewness; SEkurt = standard error of kurtosis; the mean values
are highlighted.

A member of the research team was involved in teaching both

groups. By the time of the start of the treatment, our colleague had
had 6 years of EFL teaching experience in China, and a substantial
part of that had in fact been devoted to vocabulary teaching. The
instructor had no a priori preference for either of the two instruc-
tional approaches.
The course syllabus involved the following paper and online
resources:
• Academic Words List (AWL) and related sublists (Coxhead,
2000)
• IELTS (academic module) preparing books 5 and 7: listening
and reading content (University of Cambridge Local Examina-
tions Syndicate, 2006, 2009)
• http://wordinfo.info (English Word Information)
• http://www.iciba.com (online bilingual dictionary)
• http://wordnetweb.princeton.edu/perl/webwn (WordNet, a lex-
ical database of English)
• http://www.visualthesaurus.com/trialover (Visual Thesaurus)
• http://www.ozdic.com (Oxford Collocation Dictionary Online)
• http://www.ldoceonline.com (Longman Dictionary of Contem-
porary English)
• http://www.englishvocabularyexercises.com/AWL (online aca-
demic vocabulary exercises for the Academic Word List)

BOTTOM-UP OR TOP-DOWN: EFL VOCABULARY INSTRUCTION 9

 The online resources were deployed for looking up English-based
definitions of lexical meaning as well as Chinese translation equiva-
lents, for checking phonological and morphological aspects of the tar-
get lexeme, for collecting information about relevant collocations, and
for engaging in lexical exercises.
Each lesson was structured in six distinct steps (or events):
• Event 1: Introducing the spoken and written form, as well as
the morphemic structure, of the target word
• Event 2: Introducing the definition and L1 translation of the
target word
• Event 3: Introducing related forms from the same lexical family
• Event 4: Introducing different phrases and collocations of the
target word
• Event 5: Presenting the target word at the sentence level
• Event 6: Presenting the target word at the level of the whole
context
Group A (the bottom-up group) was instructed following the order
of events from 1 to 6 as above. Group B (the top-down group) was
instructed in the reverse order, starting with Event 6 and finishing with
Event 1.
Appendix S2 in the Supporting Information online contains a com-
plete lesson plan for the teaching of the lexeme transport.

Vocabulary Tests

Two dimensions of the participants’ academic English vocabulary

competence (reception and controlled production) were measured
quantitatively with two vocabulary tests, the Academic Vocabulary Size
Test (AVST) and the Controlled Productive Knowledge Test (CPKT);
the two tests were administered at the beginning (T1) and at the end
(T2) of the 8-week treatment. The pretests and posttests involved dif-
ferent items randomly selected from Coxhead’s (2000) Academic
Word List (AWL).
The Academic Vocabulary Size Test was broadly based on Nation’s
(2001) Vocabulary Level Test (academic vocabulary section). The par-
ticipants were presented with a list of 50 lexemes randomly selected
from the AWL. They were instructed to mark each lexical item with a
(√) if they thought they recognised it; with a (?) if they thought the
word was familiar, but were unsure about the meaning; or with a (9)
if they didn’t recognise the word at all. When participants chose the

10 TESOL QUARTERLY
(√) option, they were also required to provide at least one Chinese
translation equivalent.
The AVST results were scored using the following procedure (based
on the Vocabulary Knowledge Scale developed by Wesche and Pari-
bakht, 1996). A lexical item marked with a (√) and involving a correct
Chinese translation equivalent was given a score of 1. A lexical item
marked with a (√) but involving an incorrect Chinese translation
equivalent was given a score of 0.5. Items marked with a (?) were
scored as 0.5, and items marked with a (9) were scored as 0. Partici-
pants could achieve a maximum score of 50.
The CPKT was Nation’s (2001, p. 427) Productive Level Test (uni-
versity word list level). It comprised 18 sentences involving one target
lexical item each (randomly selected from the AWL). Participants were
instructed to recover the target word from the cue provided, for exam-
ple:
The Far East is one of the most populated reg______ of the world.

The afflu_____ of the Western world contrasts with the poverty in other parts.

The book covers a series of isolated epis_____ from history.

Participants could achieve a maximum score of 18.
The full versions of the pre- and posttests can be viewed in Appen-
dix S3 in the Supporting Information online.
These two tests have been widely employed and recognised in the
field (see Laufer & Nation, 1999; Nation, 2001; Schmitt, Schmitt, &
Clapham, 2001), and their validity has been empirically established.
However, as a further measure of reliability, the scores were also sub-
jected to correlational analyses (in essence, test-retest correlations),
and the bivariate correlation coefficients between T1 and T2 test
scores for the whole sample and for each treatment group were found
to be significant at .01, which suggests that the tests were reliable mea-
sures of the participants’ vocabulary acquisition.

Analyses

To address the study’s main objective, that is, the relative efficiency
of the two vocabulary teaching approaches, eight sets of data were col-
lected: four sets of pretest scores and four sets of posttest scores. The
test scores were subjected to a range of descriptive and inferential sta-
tistical analyses. These are presented schematically in Figure 5.

BOTTOM-UP OR TOP-DOWN: EFL VOCABULARY INSTRUCTION 11

 FIGURE 5. A pre-post test between-group design.

The study’s design allowed us to explore the following relationships:

(1). Group A AVST T1 vs. Group B AVST T1. This analysis
enabled us to establish preexisting differences between the
two groups in academic vocabulary size.
(2). Group A CPKT T1 vs. Group B CPKT T1. This analysis
enabled us to establish preexisting differences between the
two groups in controlled productive vocabulary knowledge.
(3). Group A AVST T1 vs. Group A AVST T2. This analysis
enabled us to establish the relative growth in academic vocab-
ulary size among Group A learners.
(4). Group A CPKT T1 vs. Group A CPKT T2. This analysis
enabled us to establish the relative growth in controlled pro-
ductive vocabulary knowledge among Group A learners.
(5). Group B AVST T1 vs. Group B AVST T2. This analysis
enabled us to establish the relative growth in academic vocab-
ulary size among Group B learners.
(6). Group B CPKT T1 vs. Group B CPKT T2. This analysis
enabled us to establish the relative growth in controlled pro-
ductive vocabulary knowledge among Group B learners.
(7). Group A AVST T2 vs. Group B AVST T2. This analysis
enabled us to establish posttreatment differences between the
two groups in academic vocabulary size.
(8). Group A CPKT T2 vs. Group B CPKT T2. This analysis
enabled us to establish posttreatment differences between the
two groups in controlled productive vocabulary knowledge.

12 TESOL QUARTERLY
 The latter two were directly related to the study’s main objective:
establishing the effects of the two different instructional approaches,
bottom-up and top-down, to the acquisition of EFL academic vocabu-
lary.
More specifically, in addition to the descriptive statistics for the pre-
and posttest scores, we conducted a mixed model (between-within sub-
jects) 2 (group: bottom-up vs. top-down) 9 2 (time: T1 vs. T2) ANO-
VA separately on the AVST and the CPKT scores to identify whether
length of treatment had a significant effect on the scores and whether
there was a significant time vs. group interaction. The latter was of par-
ticular interest because it would reveal whether gains in vocabulary
acquisition were unequally distributed between the two groups, with
implications for the relative effectiveness of the two instructional
approaches. In order to gain insight into the specifics of the differ-
ences, the general significance effects and interactions in the ANOVAs
were followed up with within-group (paired samples) and between-
group (independent samples) t-tests. Finally, the test scores were sub-
jected to analysis of covariance (ANCOVA) which was used to establish
whether the differences at T2 between the top-down and bottom-up
groups remained when controlling for preexisting variation in the
scores at T1.

RESULTS
Descriptive Statistics

A summary of the descriptive statistics for the pre- and posttest

scores is presented in Table 2. Visual inspection of the test scores’ his-
tograms suggested a normal bell-curved distribution. Normality of dis-
tribution is further confirmed by the fact that all values for skewness
displayed in Table 2 are well below 1.0, and the ratio between skew-
ness and its standard error, and that between kurtosis and its standard
error, is less than 3.00.3
Inspection of the mean values in Table 2 reveals that participants in
both groups experienced a solid growth in EFL academic vocabulary
knowledge as a result of the treatment, as demonstrated by the higher
mean values for the posttest scores. Notably, the mean posttest values
of the bottom-up group were slightly higher than those of top-down
group. The significance of these differences for each of the tests is
explored in the following sections.

3
Post-AVST skewness–standard error of skewness ratio for the whole sample is the only
exception at a tolerable 3.6.

BOTTOM-UP OR TOP-DOWN: EFL VOCABULARY INSTRUCTION 13

Results for Academic Vocabulary Size

The mixed model 2 (group: bottom-up vs. top-down) 9 2 (time: T1

vs. T2) ANOVA conducted on the AVST scores reveals that there was a
significant effect of time, F (1,118) = 1030.94, p = .000, partial g2 =
.90. In other words, the pre- and post-AVST scores differed signifi-
cantly between T1 and T2. It is noteworthy that the effect of time (i.e.,
length of exposure to instruction) is rather large, as revealed by the
size of the partial eta squared statistic (.90). There was no significant
time*group interaction (p = .85), suggesting that both groups gained
equally in academic vocabulary size.
The significant main effect of time was followed up with within- and
between-group comparisons using t-tests. The within-group compari-
sons show that the group exposed to bottom-up instruction scored sig-
nificantly higher on the post-AVST (M = 44.82) than on the pre-AVST
(M = 29.65, t(59) = .20.13, p = .000). Similarly, the group exposed to
top-down instruction also scored significantly higher on the post-AVST
(M = 42.70) than on the pre-AVST (M = 27.67, t(59) = .20.13,
p = .000). These results suggest that both groups made significant
gains in the acquisition of academic vocabulary size over the instruc-
tion period, regardless of the type of instruction they were exposed to.
Between-group comparisons on the pre-AVST scores reveal that the
existing differences between the two groups at T1 were marginal
rather than significant, unequal variance t(110.39) = 1.9, p = .058, with
the bottom-up group scoring marginally higher (M = 29.65) than the
top-down group (M = 27.67). At T2, however, the differences between
the two groups on the AVST were statistically significant, t(118) = 3.10,
p = .002, with the bottom-up group (M = 44.87) outperforming the
top-down group (M = 42.70). In order to test whether the marginal
difference between the groups at T1 was responsible for the statisti-
cally significant one at T2, ANCOVA was conducted with the post-
AVST scores as a dependent variable and the pre-AVST scores as a co-
variate. The results reveal that even after controlling for the marginal
preexisting between-group difference at T1, the difference in scores
on the post-AVST remained statistically significant, F(2,117) = 6.01,
p = .02, g2 = .05, although the effect size is relatively weak (.05 for
instructional approach compared to .90 for time).

Results for Controlled Productive Knowledge

A similar procedure was used to analyse the CPKT scores. First, the
mixed model 2 (group: bottom-up vs. top-down) 9 2 (time: T1 vs. T2)

14 TESOL QUARTERLY
ANOVA was run, and the results reveal that, again, there was a signifi-
cant effect of time (i.e., length of exposure to instruction), F(1,118) =
577.28, p = .000, partial g2 = .83. The effect of time was nearly as large
as that for AVST, as reflected in the size of the partial eta squared sta-
tistic (.83).
In contrast with the AVST scores, however, this time there was a sig-
nificant time*group interaction, F(1,118) = 18.87, p = .000, partial g2 =
.14, indicating that there was a statistically significant difference
between the two groups in terms of the growth in controlled produc-
tive vocabulary knowledge: the bottom-up group improved significantly
more (mean difference = 6.15) than the top-down group (mean differ-
ence = 4.27).
Within- and between-group comparisons conducted with a series of
t-tests reveal that both the bottom-up and the top-down groups scored
significantly higher on the post-CPKT than on the pre-CPKT. The rele-
vant values for the bottom-up group were post-CPKT (M = 14.67) ver-
sus pre-CPKT (M = 8.52), t(59) = .34.94, p = .000. The values for the
top-down group were post-CPKT (M = 13.85) versus pre-CPKT
(M = 9.58), t(59) = .10.77, p = .000. These results indicate that as a
result of the treatment both groups made significant progress in the
acquisition of controlled productive vocabulary knowledge.
Between-group comparisons (using independent samples t-tests) on
the pre-CPKT scores show that the differences between the two groups
at T1 were marginal, unequal variance t(99.72) = 1.9, p = .057,
although the top-down group scored slightly higher (M = 9.58) than
the bottom-up one (M = 8.52). The analyses of the post-CPKT scores
reveal that, again, the bottom-up group (M = 14.67) slightly outper-
formed the top-down one (M = 13.85). However, in contrast to the
findings in relation to the post-AVST scores, the differences between
the two groups on the post-CPKT scores did not reach the conven-
tional level of statistical significance (p = .071).

DISCUSSION

The most important findings of the study can be summarised as fol-

lows. Both groups made significant gains in the attainment of English
academic vocabulary, in terms of both vocabulary size and controlled
productive vocabulary knowledge. It was not specifically among the
study’s goals to examine the durability of the lexical knowledge
attained as a result of the treatment. However, in view of the fact that
the posttests were administered 2 weeks after the end of the course
(and 10 weeks after its start), we believe that the posttest scores could
be regarded as evidence of durable learning. It would seem that the

BOTTOM-UP OR TOP-DOWN: EFL VOCABULARY INSTRUCTION 15

EFL academic vocabulary course that was specifically constructed for
the purposes of the current study was quite effective in achieving very
good learning outcomes, using either of the two instructional
approaches.
The two groups’ high achievement notwithstanding, the bottom-up
group slightly outperformed the top-down group at T2 on both vocab-
ulary size and controlled productive vocabulary knowledge. The bot-
tom-up group’s superiority on vocabulary size was found to be
statistically significant after controlling for marginal preexisting differ-
ences between the two groups at T1, although the effect size was rela-
tively small. The differences in the two groups’ performance on the
controlled productive vocabulary knowledge test at T2 did not reach a
conventional level of statistical significance. However, the bottom-up
group was found to have undergone a larger growth in controlled pro-
ductive vocabulary knowledge as a result of the treatment; this differ-
ence in growth between the two groups was found to be statistically
significant.
The study’s findings clearly indicate that between the two instruc-
tional approaches the bottom-up approach worked marginally better
for this population of EFL learners (i.e., first-year university students
in China). Put simply, learners seemed to favour the bottom-up over
the top-down instructional design.
This result may be impossible to explain by reference to a single fac-
tor. It is much more likely that the operation of several largely inde-
pendent factors converged to yield such outcomes. In what follows we
briefly consider what we believe to be the most important among
them.
In the first place, it may be the case that by its nature L2 lexical
development is better aligned with bottom-up instruction. According
to Jiang’s (2000) theory, L2 lexical acquisition involves three distinct
stages. An early initial stage involves only phonological and ortho-
graphic features for the L2 lexical entry. With the increase of learners’
exposure to the L2, the L2 lexical entry is linked to an L1 translation
equivalent; notably, at this stage the L2 entry involves L2 phonological
and orthographic features, but L1 syntactic and semantic features.
After more exposure to, and experience with, the L2, the L1 syntactic
and semantic features are gradually replaced by L2 syntactic and
semantic features; this constitutes the third and final stage of L2 lexi-
cal development. Jiang’s model of L2 lexical development is schemati-
cally presented in Figure 6.
The actual sequencing of vocabulary instruction used in the bottom-
up approach seems reasonably well aligned with these three stages. In
the bottom-up instruction, the first step involved presenting the pho-
nological, morphological, and graphemic features of the target word.

16 TESOL QUARTERLY
FIGURE 6. The three stages of L2 lexical development (adapted from Jiang, 2000, and Ma,
2009).

Then the word’s definition and its Chinese translation were intro-
duced, followed by instruction targeting related forms. These first
three steps seem to correspond quite nicely to the first two stages of Ji-
ang’s (2000) model, and Steps 4–6 of the bottom-up approach, which
were concerned with the word’s use in higher ranked linguistic struc-
tures (e.g., collocations, clauses, discourse), seem to correspond to Ji-
ang’s Stage 3. As pointed out earlier, the top-down approach involved
the reverse sequence of instructional steps.
Another factor which is quite likely to have played a role in relation
to this study’s results is the learners’ L2 proficiency level. There is gen-
erally agreement in relevant literature that top-down processing is
more common among high-proficiency learners, whereas lower profi-
ciency learners tend to favour bottom-up processing (e.g., Eskey, 1988;
Shohamy & Inbar, 1991; Tsui & Fullilove, 1998). This pattern is

BOTTOM-UP OR TOP-DOWN: EFL VOCABULARY INSTRUCTION 17

probably at least in part due to the different modes of processing asso-
ciated with language perception and language production. As we sug-
gested in the introductory part of this article, language perception is
essentially (though not exclusively) bottom-up, and the opposite seems
to be the case with language production; Levelt et al.’s (1999) model
of speech production presented earlier is a good illustration of that.
In most natural circumstances the initial phases of (second) language
acquisition are predominantly receptive (and therefore bottom-up). As
L2 teaching professionals know quite well, the productive skills often
lag considerably behind the receptive ones, and in some situations
(especially in foreign language contexts) may not evolve to any sub-
stantial degree. As we noted earlier, the participants’ test scores at T1
(see Table 2, in the Results section) reveal that their English profi-
ciency level was relatively low, indicating a natural tendency for bot-
tom-up processing.
These two factors—proficiency level and nature of L2 lexical devel-
opment—together go a long way towards explaining the advantage of
the bottom-up instructional approach found in our study, although
the participants’ preference for bottom-up processing may have been
additionally reinforced by cultural and traditional factors specific to
the Chinese context. It would appear that the notion of a bottom-up
mode of information processing and knowledge formation is deeply
rooted in the sociocultural and educational tradition of China (Hu,
2002a, 2005; Ma, 2009; Watkins & Biggs, 2001). The concept of grad-
ual evolution from smaller and more basic elements to complex and
elaborate structures seems to have been a continuous theme in Chi-
nese thought and philosophy over a long time and remains a distinct
characteristic feature, it has been argued, of contemporary Chinese
mentality (Li, 2006). It is also a matter of fact that most education in
the Chinese context (including first language literacy) involves bot-
tom-up instruction (Jin, 2010; Peng, 1997). Jin (2010), for instance,
advocates a distinctly bottom-up sequence as the most efficient way of
teaching and learning Chinese literacy: 字[zi] (words), 词[ci] (phrases),
句[ju] (sentences), 段[duan] (paragraph), 篇[pian] (text). And as a mat-
ter of fact, most textbooks and teaching materials for Chinese literacy
seem to follow that order (Ministry of Education of the People’s
Republic of China, 2011). In light of that, it should not be surprising
that Chinese learners favour bottom-up processing.
Our discussion of the study’s findings would be incomplete without
some commentary on the EFL academic vocabulary course that was
specifically created for the purposes of our research. A comparison of
T1 and T2 test scores reveals that the course enabled participants to
achieve substantial gains in both vocabulary size and controlled pro-
ductive vocabulary knowledge. It seems quite clear that the innovative

18 TESOL QUARTERLY
design we used in constructing the syllabus as well as the use of a
range of online vocabulary resources as an essential part of the instruc-
tion were instrumental in producing these notable learning outcomes.
They also reinforce the value of explicit and direct vocabulary teaching
for the development of both receptive and productive vocabulary
knowledge.

CONCLUSION

Although we cannot claim that our results are definitive, they are
indicative with respect to several points. For one thing, it appears that
particular attention to vocabulary teaching, for example, in a dedi-
cated course (whatever approach is used), will lead to successful learn-
ing of this part of a second language. Further, the particular method
of course development that we used, constructing a vocabulary course
which takes considerable advantage of resources available online,
seems to have worked well.
Of course, the main goal of our study was to determine the relative
effectiveness of the bottom-up and top-down approaches to vocabulary
teaching, and our results show a clear, albeit weak, advantage for the
former approach, although this should be confirmed by further stud-
ies. We have linked the mode of processing to two factors, learners’
proficiency level and the nature of L2 lexical development, suggesting
that these may have been in part responsible for the outcomes of this
study. We further suggest that the participants’ Chinese cultural and
educational background may have also played a role in relation to our
results. In this light it would be interesting to carry out similar studies
in other cultural contexts where the bottom-up approach is not so
deeply embedded. Should new studies produce similar results, then it
might mean that the effectiveness of this approach is not (at all) to be
attributed to cultural factors, but rather to features of human process-
ing.
As noted earlier, the current study was specifically designed to
address issues of L2 vocabulary instruction. An obvious extension of
our research would involve examining the effectiveness of bottom-up
and top-down instruction on the other major language skills (reading,
listening, writing, and speaking), especially the productive ones.

ACKNOWLEDGEMENTS

We owe our gratitude to Dr. Silvia Ratcheva for her invaluable assistance with this
study’s statistical analyses.

BOTTOM-UP OR TOP-DOWN: EFL VOCABULARY INSTRUCTION 19

THE AUTHORS

Christo Moskovsky is a senior lecturer at the University of Newcastle, Australia, where

he has been involved in postgraduate education and research in applied linguistics.
He has authored and coauthored a number of publications in this field of study.

Guowu Jiang recently completed a PhD program in applied linguistics at the Uni-
versity of Newcastle, Australia. He has been involved in EFL education and
research for a number of years, with a special interest in EFL vocabulary teaching
and learning.

Alan Libert is a senior lecturer at the University of Newcastle, Australia. He com-

pleted his BA in Greek and Latin at New York University and his PhD in linguis-
tics at McGill University. His most recent book is Adpositions and Other Parts of
Speech (2013, Peter Lang).

Seamus Fagan, director of the English Language and Foundation Studies Centre
at the University of Newcastle, has been involved in EFL/ESL for more than
30 years. He served for 17 years on the Board of English Australia, the last 6 years
as chair. His area of research is World Englishes.

REFERENCES

Anderson, R., Reynolds, R., Schallert, D., & Goetz, T. (1977). Frameworks for com-
prehending discourse. American Educational Research Journal, 14, 367–381.
doi:10.3102/00028312014004367
Carrell, P. (1988). Some causes of text-boundedness and schema interference in
ESL reading. In P. L. Carrell, J. Devine, & D. E. Eskey (Eds.), Interactive
approaches to second language reading (pp. 101–113). New York, NY: Cambridge
University Press.
Carrell, P., & Eisterhold, J. (1983). Schema theory and ESL reading pedagogy. TE-
SOL Quarterly, 17, 553–573. doi:10.2307/3586613
Celce-Murcia, M., & Olshtain, E. (2001). Discourse and context in language teaching:
A guide for language teachers. Cambridge, England: Cambridge University Press.
Chaudron, C., & Richards, J. (1986). The effect of discourse markers on the compre-
hension of lectures. Applied Linguistics, 7, 113–127. doi:10.1093/applin/7.2.113
Coxhead, A. (2000). A new academic word list. TESOL Quarterly, 34, 213–238.
doi:10.2307/3587951
Coxhead, A. (2006). Essentials of teaching academic vocabulary. Boston, MA: Hough-
ton Mifflin.
De Bot, K., Paribakht, T. S., & Wesche, M. B. (1997). Towards a lexical processing
model for the study of second language vocabulary acquisition. Studies in Second
Language Acquisition, 19, 309–329. doi:10.1017/S0272263197003021
Eskey, D. (1988). Holding in the bottom: An interactive approach to the language
problems of second language readers. In P. L. Carrell, J. Devine, & D. E. Eskey
(Eds.), Interactive approaches to second language reading (pp. 93–100). New York,
NY: Cambridge University Press.
Field, J. (2004). An insight into listeners’ problems: Too much bottom-up or too
much top-down? System, 32, 363–377. doi:10.1016/j.system.2004.05.002
Frederic, C. B. (1997). Remembering: A study in experimental and social psychology.
Cambridge, England: Cambridge University Press.

20 TESOL QUARTERLY
Gan, Z., Humphreys, G., & Hamp-Lyons, L. (2004). Understanding successful and
unsuccessful EFL students in Chinese universities. Modern Language Journal, 88,
229–244. doi:10.1111/j.0026-7902.2004.00227.x
Goodman, K. (1967). Reading: A psycholinguistic guess game. Journal of the Read-
ing Specialist, 6(1), 126–135. doi:10.1080/19388076709556976
Hildyard, A., & Olson, D. (1982). On the comprehension and memory of oral ver-
sus written discourse. In D. Tannen (Ed.), Spoken and written language: Exploring
orality and literacy (pp. 19–32). Norwood, NJ: Ablex.
Hu, G. (2002a). The People’s Republic of China country report: English language
teaching in the People’s Republic of China. In R. E. Silver, G. Hu, & M. Iino
(Eds.), English language education in China, Japan, and Singapore (pp. 1–77). Sin-
gapore: National Institute of Education, Nanyang Technological University.
Hu, G. (2002b). Recent important developments in secondary English-language
teaching in the People’s Republic of China. Language, Culture and Curriculum,
15, 30–49. doi:10.1080/07908310208666631
Hu, G. (2005). Contextual influences on instructional practices: A Chinese case
for an ecological approach to ELT. TESOL Quarterly, 39, 635–660. doi:10.2307/
3588525
Huang, J., Chen, Y., Xu, Y., Li, J., & Fu, Y. (2004). On the vocabulary size of the
College English Syllabus. Foreign Language World, 1, 1–9.
Jay, T. B. (2003). The psychology of language. Upper Saddle River, NJ: Prentice Hall.
Jiang, N. (2000). Lexical representation and development in a second language.
Applied Linguistics, 21, 47–77. doi:10.1093/applin/21.1.47
Jin, H. (2010). 随堂导读(字、词、句、段、篇) [Classroom guidance: Words, phrases,
sentences, paragraphs, and texts]. Changchun, China: Jilin Education Press.
Laufer, B., & Nation, P. (1999). A vocabulary-size test of controlled productive
ability. Language Testing, 16, 33–51. doi:10.1177/026553229901600103
Levelt, W. J. M., Roelofs, A., & Meyer, A. S. (1999). A theory of lexical access in
speech production. Behavioral and Brain Sciences, 22, 1–75. doi:10.1017/
S0140525X99001776
Li, G. (2006). What do parents think? Middle-class Chinese immigrant parents’
perspectives on literacy learning, homework, and school-home communication.
School Community Journal, 16, 27–46.
Ma, Q. (2009). Second language vocabulary acquisition. Bern, Switzerland: Peter Lang.
Matthews, P. H. (Ed.). (2007). Oxford concise dictionary of linguistics. Oxford, Eng-
land: Oxford University Press.
Mendelsohn, D. J. (2001). Listening comprehension: We’ve come a long way,
but . . . Contact, 27, 33–41.
Ministry of Education of the People’s Republic of China. (2011). The criteria of Chinese
literacy pedagogy for compulsory education. Beijing, China: People’s Education Press.
Nation, P. (2001). Learning vocabulary in another language. Cambridge, England:
Cambridge University Press.
Nation, P. (2006). How large a vocabulary is needed for reading and listening?
Canadian Modern Language Review, 63, 59–82. doi:10.1353/cml.2006.0049
National College English Testing Committee. (2013). The college English syllabus.
Shanghai, China: Shanghai Foreign Language Education Press.
Peng, D. (1997). 汉语认知研 [Chinese literacy study]. Shandong, China: Shandong Edu-
cation Press.
Perfetti, C. (1985). Reading ability. New York, NY: Oxford University Press.
Read, J. (2004). Research in teaching vocabulary. Annual Review of Applied Linguis-
tics, 24, 146–161. doi:10.1017/S0267190504000078
Rost, M. (1990). Listening in language learning. London, England: Longman.

BOTTOM-UP OR TOP-DOWN: EFL VOCABULARY INSTRUCTION 21

Schmitt, N., Schmitt, D., & Clapham, C. (2001). Developing and exploring the
behaviour of two new versions of the Vocabulary Levels Test. Language Testing,
18, 55–89. doi:10.1191/026553201668475857
Shohamy, E., & Inbar, O. (1991). Validation of listening comprehension tests: The
effect of text and question type. Language Testing, 8, 23–40. doi:10.1177/
026553229100800103
Smith, F. (2004). Understanding reading: A psycholinguistic analysis of reading and
learning to read (6th ed.). Mahwah, NJ: Lawrence Erlbaum.
S€
okmen, A. J. (1997). Current trends in teaching second language vocabulary. In
N. Schmitt & M. McCarthy (Eds.), Vocabulary: Description, acquisition and pedagogy
(pp. 237–257). Cambridge, England: Cambridge University Press.
Sonbul, S., & Schmitt, N. (2010). Direct teaching of vocabulary after reading: Is it
worth the effort? ELT Journal, 64, 253–260. doi:10.1093/elt/ccp059
Stanovich, K. E. (1980). Toward an interactive-compensatory model of individual
differences in the development of reading fluency. Reading Research Quarterly,
16, 32–71. doi:10.2307/747348
Sun, R. (2002). Duality of the mind: A bottom-up approach toward cognition. Mahwah,
NJ: Lawrence Erlbaum.
Tsui, A. B. M., & Fullilove, J. (1998). Bottom-up or top-down processing as a dis-
criminator of L2 listening performance. Applied Linguistics, 19, 432–451. doi:10.
1093/applin/19.4.432
Tyler, L. K., & Warren, P. (1987). Local and global structure in spoken language
comprehension. Journal of Memory and Language, 26, 638–657. doi:10.1016/
0749-596X(87)90107-0
University of Cambridge Local Examinations Syndicate. (2006). Cambridge IELTS 5:
Examination papers from University of Cambridge ESOL examinations—English for
speakers of other languages. Cambridge, England: Cambridge University Press.
University of Cambridge Local Examinations Syndicate. (2009). Cambridge IELTS 7:
Examination papers from University of Cambridge ESOL examinations—English for
speakers of other languages. Cambridge, England: Cambridge University Press.
Vandergrift, L. (2004). Listening to learn or learning to listen? Annual Review of
Applied Linguistics, 24, 3–25. doi:10.1017/S0267190504000017
Wang, Z. (2010). A study on vocabulary learning through English reading. Shanghai,
China: Shanghai Jiaotong University Press.
Watkins, D. A., & Biggs, J. B. (Eds.). (2001). Teaching the Chinese learner: Psychologi-
cal and pedagogical perspectives. Melbourne, Australia: Comparative Education
Research Centre and Australian Council of Educational Research.
Wesche, M., & Paribakht, T. (1996). Assessing second language vocabulary knowl-
edge: Depth versus breadth. Canadian Modern Language Review, 53, 13–39.

Supporting Information

Additional Supporting Information may be found in the online ver-

sion of this article:
Appendix S1. Demographic information on the participants.
Appendix S2. An example of how the word “TRANSPORT” was
taught.
Appendix S3. Vocabulary pre- and post tests.

22 TESOL QUARTERLY

View publication stats