SCO-KOHNSTAMM INSTITUUT
The training of a word learning strategy: results in process and product
Drs. M. van Daalen-Kapteijns
Dr. C. Schouten-van Parreren
Dr. K. de Glopper
SCO-report 463
KOHNSTAMM
Daalen-Kapteijns, M.M. van
The training of a word learning strategy: results in process and product
Maartje van Daalen-Kapteijns, Carolien Schouten-van Parreren, Kees de Glopper
Amsterdam : SCO-Kohnstamm Instituut, Faculteit der Pedagogische en Onderwijskundige
wetenschappen, Universiteit van Amsterdam
(SCO-rapport : 463)
Met lit. opg.
ISBN 90-6813-517-1
Trefw : vocabulary ; verba! learning
Alle rechten voorbehouden. Niets uit deze uitgave mag worden verveelvuldigd, opgeslagen in een
geautomatiseerd gegevensbestand, of openbaar gemaakt, in enige vorm of op enige wijze, hetzij
electronisch, mechanisch, door fotocopieën, opnamen, of op enige manier, zonder voorafgaande
schriftelijke toestemming van de uitgever.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or
transmitted, in any form or by any means, electronic, mechanical, photocopying, or otherwise,
without the prior written permission of the publisher.
Uitgave en verspreiding :
SCO-Kohnstamm Instituut / Faculteit POW
Mevrouw W. Sargentini
Wibautstraat 4
1091 GM Amsterdam
tel : 020-5251357/1201
Copyright © SCO-Kohnstamm Instituut, 1997.
The training of a word learning strategy:
results in process and product
Authors:
Maartje van Daalen-Kapteijns
SCO-Kohnstamm Institute
Faculty of Educational Sciences
University of Amsterdam
Carolien Schouten-van Parreren
Psychology and Pedagogical Sciences
"Vrije Universiteit"
Kees de Glopper
SCO-Kohnstamm Institute
Faculty of Educational Sciences
University of Amsterdam
SCO-report 463
May 1997
word learning strategy
Preface
This paper reports the results of the training of a strategy to infer word
meanings from context.
The research was supported by the Institute for Educational Research
(SVO: grant numbers 93501 and 95734) and was performed in the SCOKohnstamm Institute of the Faculty of Educational Sciences of the
University of Amsterdam from 1993 to 1995.
We thank the members of the 'HogCog' studygroup for their useful
comments on an earlier draft of this paper.
Table of contents
Abstract
Introduction
Theoretical basis of the training program
Setup of the training program
Description of the Word Learning Strategy (WLS)
The design of the experiment
Analyses of the product results The think-aloud sequel to investigate the inference process Protocol differences between trained and control students Analysis of the process codes
Analysis of the definitions as the products of the think-aloud process
Summary of the differences between trained and control students General conclusions and discussion
References
Appendix A
Figures 1 to 3
Table 1
Tables 2a to 2e
Table 3
Table 4
Tables 5 and 6
Tables 7.1 and 7.2
2
3
5
6
7
11
13
14
18
18
23
26
28
32
35
37
38
39
41
42
43
44
2
word learning strategy
Abstract
This paper reports a study into the effects of a rich instructional
procedure on the process and product of learning to infer word meanings
from written context. The product effects were measured in an experiment
with paper-and-peneil tests, while the process effects were assessed in a
think-aloud sequel.
An experimental design was used in which students were allocated
to an experimental and control condition by blocking. Thirty two gradesix primary school children followed eight training sessions in their
mother tongue (Dutch) and 32 control students followed their regular
lessons in which no special attention was given to learning from context.
The training consisted of text-driven, wordschema-driven and selfregulatory elements, and was given in the form of general cognitive
strategy instruction. On a paper-and-pencil test of skill in inferring word
meaning from context, the difference between the scores of the
experimental and control group was 0.2 standard deviation, which is
statistically not significant. Think aloud-protocols of 16 trained and 16
untrained students were analyzed to obtain insight into how often and
with what quality elements from the training course were used. Trained
students, in comparison with the untrained ones, more often used textdriven and self-regulatory elements. No difference was found for
wordschema-driven elements. The higher frequency of use of the elements
mentioned did not go together with significantly higher quality of the
word meanings inferred. The results are discussed in relation to the
content of the training given and the inherent difficulty of the inference
task.
3
word learning strategy
Introduction
The importance of vocabulary knowledge in a variety of tasks
involving verbal information processing has been widely acknowledged.
Vocabulary knowledge plays a major role in reading comprehension
(Beek, Perfetti & McKeown, 1982; Kameenui, Carnine & Freschi, 1982),
and is therefore strongly related to success in school (Jenkins, Matlock
Slocum, 1989). The question of how children acquire and expand their
vocabulary has been investigated by theorists and researchers (McKeown
& Curtis, 1987). One of the research methods has been to explore
individual differences in vocabulary knowledge and word learning ability
(Shefelbine, 1990). One of the aims of this kind of research is to find
ways to improve vocabulary learning and instruction.
The two most prominent instructional approaches are direct
teaching of individual words and teaching skills that promote independent
learning of words. As it is impossible to teach huge numbers of words
through direct instruction, many vocabulary studies are focused on
teaching vocabulary skills. Using context is one of the most frequently
explored skills (Sternberg & Powell, 1983; Graves, 1987; Buikema &
Graves, 1993). This is potentially a very useful skill because with each
year of schooling students are confronted with a growing amount of texts
that provide opportunities for inferring word meanings.
A previous study (De Glopper, van Daalen-Kapteijns & Schoutenvan Parreren, 1997) shows that skill in inferring word meaning from
context is substantially correlated with vocabulary knowledge. Inferring
word meaning from context is a very complex skill (cf. Van DaalenKapteijns & Elshout-Mohr, 1981). This is probably why a number of
intervention studies bas tried to reduce the complexity of this skill.
In the first place, some intervention studies have focused on only a
part of the skill, e.g. by training students to use a limited set of context
eines, such as synonym and contrast clues (Carnine, Kameenui & Coyle,
1984; Patberg, Graves & Stibbe, 1984). One of the problems with the
training of clues is that some of the more helpful ones occur only rarely
in natural text (Graves, 1987, p. 173). The generalizability of the effect of
such training is therefore limited.
In the second place a reduction of the complexity of the skill has
been attempted by choosing a specific kind of target word for training. In
many interventions the target words (often nonsense words or low
frequency words) can be substituted by a relatively simple synonym or
expression (for instance the word 'oam' meaning 'steam', in the study by
Sternberg & Powell, 1983). In such interventions hardly any target words
occur for which the meaning can only be given by using a more complex
description, for instance in the form of a superordinate and one or more
specifications. An example of such a target word is the word 'receipt' that
we used in the study presented here, meaning 'a written acknowledgement
of something received'. In interventions working with a set of relatively
simple target word meanings, no instructional help is given which is
directed at finding the meaning of more complex ones. This may be one
4
word learrdng strategy
of the reasons that most training studies have resulted in no more than
small gains. Still, as Beek & McKeown conclude (1991, p. 803), "the
potential importance of context as a vocabulary-learning source and the
apparent difficulty in fully utilizing that source warrant a continued search
for more effective instruction".
In the study presented here, we have made an attempt to develop
an enriched and possibly more effective instructional procedure. We did
so by combining different instructional elements in order to prepare
students to deal with the task of inferring word meaning from context in
its full complexity. The training contained not only elements that work
bottom-up from the context, but also elements that work top-down from
knowledge about the structure of word meanings. Besides these "textdriven" and "wordschema-driven" elements, some self-regulatory elements
were part of the training. In this way, we tried to teach students to handle
all kinds of words, that is, also words for which no simple synonym
exists and whose meaning has to be described in a more complex form.
As far as we know, testing such an enriched instructional procedure,
directed at the full complexity of the inference process, has not been
attempted before. In this study we trained students (primary school
children, aged 11-12 years) of the whole verbal ability range; in this
respect too, we tried to avoid a limitation of some previous studies which
trained students of a more restricted verbal ability range.
As there is little experience with such an instructional procedure,
we assessed not only the quantitative results in the form of the products
of the training, but also the qualitative results, i.e. the relevant process
aspects of the students' performance after training. In this way we tried to
gain insight into how the different training elements worked out and with
what effect. Process aspects of the inference skill are also receiving more
attention in other recent studies, such as that by Goerss (1995) and by
Goerss, Beek & McKeown (in preparation).
In the present study, the instructional procedure sketched above has been
tested in order to answer the following research questions. To what
degree is it possible to teach students to infer word meanings from
context in its full complexity? What is the role of the various training
elements on the process and product of the inference activities by students
thus trained? Possible transfer effects of the training from Dutch to
English as a foreign language are also explored.
In the next section, the training program is described in detail in
order to give a clear picture of its constituent elements and the word
learning strategy in which these were embedded.
5
word learning strategy
Theoretical basis of the training program
Let us look first at the situation in which students read a piece of
text and encounter an unknown word. Different actions can be taken,
depending on a variety of task and learner variables. If the students
understand what the text is about and are not hindered by the presence of
the unknown word, they will probably read on because there is no need to
stop and think about what the word might mean. If they understand the
thrust of what the text is about, but full comprehension is hindered by the
unknown word, they have a motive for trying to figure out the word
meaning in order to help comprehension. If they also have come to
appreciate reading as a source for vocabulary learning, they might make
an extra effort to reflect on the inferred word meaning as such.
Figuring out a word meaning from context is like trying to pull
yourself up by your bootstraps. How is it possible to infer the meaning of
a word from a context that is not fully understood because of the verg
presence of the word? We think a solution to this paradox can be found
by considering the beginring and the end of the inference process.
Working from the text
The ground from which one wants to pull oneself up may have
one or more footholds in the form of clues to the meaning of the word;
these clues can be used in a text-driven or bottom-up process. Clues can
be found in the word itself and in the surrounding text, especially when
the text is redundant.
Working from a schema
If one knows what it usually takes to define a word meaning, this
kind of knowledge can be used in a schema-driven or top-down process.
Students can actively search the context for the necessary. definitional
information. If the student has knowledge of the make-Up of definitions,
he can go and search the context to fill the empty slots of the schema.
Decontextualization
Somehow the text-driven and schema-driven processes must be
attuned to one another. In the ideal case, the text-driven search for
meaning is steered from the schema and information found in the context
is decontextualized in order to give it a form that fits the schema. The
person searching goes back and forth between context and schema until
the fit is satisfactory. An important means for the purpose of
decontextualization is 'negotiation of meaning'. The issue to _be negotiated
is to what extent the information found is context-specific and to what
extent it can be generalized. The student, in this negotiation, must weigh
the contextual information against the 'conventional' form that
definitional information usually takes.
Our analysis resembles the process distinction of Sternberg (1987).
Sternberg distinguishes three processes that come into play when a new
6
word learning strategy
word is learned from context. The first process, called 'selective
encoding', entails separating information that is relevant for the meaning
of the unknown word from irrelevant information. This process operatel
(in our view) mainly on a text-driven basis. The second process,
'selective combination', would in our view operate mainly on a schemadriven basis, as in this process a selection from the relevant information
found in the context about the new word is combined to form a
definition. 'Selective comparison', the third process in which new
information is integrated with information already in memory, could have
the function of a go-between, just as decontextualization has.
Setup of the training program
We constructed a training program that would stimulate students
to use botte text-driven and schema-driven processes in order to arrive at a
decontextualized meaning of the target word. The training material
consisted of 'pedagogical contexts' (Beek, McKeown & McCaslin, 1983)
meant to give relevant information about the unknown word without
stating the definition. We tried to make the contexts as natural as possible
by describing situations in which the word typically occurs. We did not
systematically vary the types of clues provided, but saw to it that various
types of clues occurred in different contexts.
As a first help in working on a text-driven basis, we wanted to
teach students to use the sentence in which the unknown word occurs as a
starting point for the inference process by deleting the unknown word to
make a ' cloze sentence'. A second help was a set of questions to be asked
of the context in order to gather information about the word meaning.
Where
This set consisted of six 'Wh-questions': What ..., Who
When ..., Why ... and How ...? The questions are along the same lines as
the '20 questions' used by Sternberg (1987) in a training program for
learning word meanings from context intended for students older than in
our case (high school sophomores and juniors). We taught our students to
use the six questions as examples for ones that they would like to ask of
the context themselves.
We also wanted to make students aware of the information that a
word itself may reveal about its meaning. We wanted them to look for
this kind of information by 'zooming in' on the word. At the same time,
we told them explicitly to use this kind of information to check the
information gained from the surrounding context. This procedure prevents
students from making a wild guess on the basis of the word itself, and
consecutively twisting the interpretation of the context to confirm this
guess (Nation, 1990, p. 163, warns about this kind of twistmg.) By
zooming in on the word as a check on information gathered from the
context, a student combines working on a text-driven basis with the use
of knowledge about word form and word parts.
As a help in working on a schema-driven basis we wanted to
make students familiar with the form that most word meanings take. For
that purpose we introduced a 'Definition Schema' to them (see Figure 1):
7
word learning strategy
[Figure I here]
The Definition Schema resembles the Word Map that Schwartz &
Raphael (1985) use for vocabulary learning. What we call 'general
characteristic' is the 'class' component in the Word Map, and the
'spectfic characteristic' has its equivalent in the Word Map under the
heading of 'What is it likeT The Word Map comprises a third
component: 'What are some examplesT, information about subordinates
that is not represented in the Definition Schema. Schwartz & Raphael
suggest that it is possible to teach primary school students in the middle
grades and above to use the Word Map as a help in vocabulary learning.
An important part of the training program is decontextualization
through negotiation of meaning in a dialogue between the trainer and the
students. The exact content of the dialogue is not prescribed, because it
has to be tailored to the suggestions and reactions of the students. This
issue is closely connected with the instructional approach of the training
program. (See also the separate section on the instructional approach).
Self-regulatory elements form the last ingredient of the training
program. One element consists of making students aware of unknown
words that they may encounter in contexts and of the different ways they
can react when this happens. Such awareness is a condition that has to be
met in order for the student to engage, in some situations at least, in
deliberate attempts to infer the meaning of the word from the context. A
second self-regulatory element in the training program is the stress placed
on the need to check the information derived from the context in this
same context and in new contexts yet to be encountered.
•
Description of the Word Learning Strategy (WLS)
We tried to combine the principIes mentioned above in the form
of a coherent strategy, called 'Word Learning Strategy'. It is made up of
three tactics that can be executed in consecution.
Tactic 1: the 'Brake-tactic' is meant to be used on encountering an
unknown word. This tactic makes students aware of unknown words they
come across while reading text.
Tactic 2: the 'Track-tactie' is meant to track down the meaning of the
word. The 'Track-tactic' is the heart of the strategy. It consists of a
number of steps; the first is to make a 'cloze sentence'. This is a sentence
or part of a sentence from the text in which the unknown word is
replaced by dots. The second step is to ask a few Wh-questions on the
basis of the 'cloze sentence', the answers to which are to be found in the
sentence itself and in the surrounding sentences. The students could use
Wh-questions that are provided or make up one or two by themselves. It
is made clear that a question does not necessarily have to start with a
'Wh-word'. The third step is to find, from the information thus gathered,
one of two things: another word with the same meaning (a synonym); or,
8
word learning strategy
if this is not available, a definition made up of a general characteristic
and one or more specific characteristics. The answers to the Wh-questions
must somehow be transformed into information that fits the definition
format. So it is in this step that the actual decontextualization must take
place. The fourth and last step of the 'Track-tactic' is to check whether
the solution (synonym or definition) fits in the context given.
Tactic 3: the 'Zoom-tactic' embodies zooming in on the word itself in
order to find out whether this yields information that confirms the word
meaning inferred so far.
The tactics were summarized on a Strategy Card that the students could
use during training.
Instructional approach and number and structure of the lessons
The students were trained in groups of eight by 'general cognitive
strategy instruction' (Rosenshine & Meister, 1994). The role of the trainer
in this kind of instruction is a modeling, guiding and checking one. Every
new part of the training was first modeled by the trainer, who demonstrated what she wanted the students to learn while thinking aloud. In the
second step the students as a group tried to do the same with a different
example, in interaction with the trainer who guided and steered them.
After that the students practiced what they had learned for themselves, or
in couples, while the trainer checked what they were doing. The essential
parts of the training program (making a cloze sentence and asking Whquestions, using the Definition Schema, decontextualiation and zooming
. in) were all taught in these three steps.
To teach the Word Learning Strategy, eight lessons of
approximately one hour were constructed with Dutch material. In addition
to these, two lessons were constructed for practicing the strategy with
English words. This was done in order to explore possible transfer effects
from Dutch to English. For Dutch students, English is taught as a foreign
language in grades five and six. Our sixth grade subjects, at the time of
training, had had formal instruction in English for about one and a half
school years.
A short description of all ten lessons is given below. The trainer worked
from a trainer's guide in each lesson, and the students worked in a
booklet of their own.
At the beginning of every lesson, reference is made to what was
done in the previous one. Then the subject matter of the lesson is
modeled by the trainer or practiced by the students under the guidance of
(and checked by) the trainer. During the lessons the students fill out
practice schemas, with room for the cloze sentence and the Wh-questions,
and for a synonym or a definition for the new word. Every lesson ends
with a few reflection questions that are answered by the students for
themselves and are then discussed in the group.
To make the lessons stimulating the practice items for the students
take different forms. Some items have the form of a game. Others have a
9
word learning strategy
form resembling the 'vignettes' that Palincsar & Brown (1989) used in
their instruction for self-regulated reading. In such vignettes the students
read the account of an imaginary pair of children who have used the
Word Learning Strategy to figure out the meaning of an unknown word;
they then discuss the way this was done.
The words used in the lessons
To construct the training material, words were chosen from a
dictionary that defines Dutch words for Dutch students in graden 5
through 8 (Huijgen & Verburg 1987). Words were selected that we, and
grade 6 teachers, presumed would be unknown to students in grade 6.
Among these were difficult words referring to a well-known concept, for
which the students would probably know a synonym (a translated
example is 'accuracy' as the synonym for 'preciseness'). But most of the
words selected referred to a meaning that was at least partly new,
consisting of, for instance, an unknown refinement of a well-known
meaning. An example is the word 'anecdote' as a refinement of the more
general meaning of 'story'.
The Word Learning Strategy was introduced and practiced with
nouns, nouns being the largest category of words in natural language.
However, the students also practiced with verbs and adjectives. In the
case of adjectives, special attention was given to the Definition Schema.
The students were instructed in this case to look only for specific
characteristics that specified the noun to which the adjective refers.
For the English lessons, difficult English words were selected from
appropriate textbooks. Care was taken that the contexts of the target
words were made up off, words that most students would be familiar with.
The contexts used in the lessons
The contexts for the unknown words were constructed as
'pedagogical contexts'. One context per word was offered, describing a
situation in which the word was fairly typically used. It was made clear to
the students that one context is not enough to determine the definite
meaning of a word, but that it is often enough to get a good hunch about
the meaning, that can be checked when the same word is encountered
again. We hoped in this way to make students aware of the fact that
knowledge of a word is not .an all-or-none matter, but that it develops
gradually.
Example of the modeling phase of the instruction
One of the practice items reads as follows:
I don't know much about my grandfather but some anecdotes.
This is the anecdote that my father likes best: how my grandfather,
as a young man, once fell through the ice. Grandfather thought
that he would get less cold without all those wet clothes. So he
took everything off except his pants. That is how he came home,
with icicles hanging off his moustache and eyebrows. It must have
been a funny sight. We have to laugh every time my father tells it.
10
word learning stralegy
Students would typicaliy proceed as follows, in interaction with
the teacher.
They start out using the 'Brake-tactic', by stopping at the word 'anecdote'
and underlining it. Then they go on using the 'Track-tactic' to track down
the meaning of 'anecdote'. In the first step they make a 'cloze sentence'
that my father likes best'. In the second step
such as 'This is the they ask a few Wh-questions on the basis of this cloze sentence. In the
left-hand column of Figure 2 possible Wh-questions are given, and in the
right-hand column possible answers to these questions are indicated.
[Figure 2 here]
In the third step the students try to find, from the information thus
gathered, a synonym or a definition. In the case of 'anecdote' no
synonym is available, so a definition has to be made up, consisting of a
general characteristic and one or more specific characteristics. The
dialogue that guides the decontextualization process would take the
following form (slightly different with different groups of students, as it
had to be adapted to the specific reactions of the students):
Trainer: 'So what we have got is: an anecdote is a story that makes us
laugh and that is about my grandfather going through the ice. What does
that teil us about the meaning of 'anecdote'? Does it tell us something
about a general characteristic? About what sort of thing an anecdote
might be?'
A student: 'An anecdote is a story about my grandfather'.
Trainer: 'Right, here it is a story about my grandfather, correct. Would
that always be so? Is ah anecdote always a story about my grandfather, or
could it also be a story 'bout someone else?'
Most students think that anecdotes might be about other people, so only
the general characteristic 'is a story' is written down in the definition
schema. The dialogue is continued with the trainer reacting to what the
students say as much as possible while guiding them towards a more or
less context-free meaning for anecdote.
Explicit attention is given to the uncertain status of the information
inferred. 1f the answer to the second Wh-question 'it makes us laugh' is
transformed into the specific characteristic 'is funny', the trainer discusses
the need to check this when the word 'anecdote' is encountered in a
different context.
To conclude the dialogue, a definition is formulated on the basis of the
information in the schema and written down on the bottom line.
Figure 3 shows what the schema now contains:
[Figure 3 here]
In the fourth and last step of the 'Track-tactic' the students check whether
this definition fits the context given.
To conclude the inference process the students use the '.Zoom-tactic'.
They look at the word 'anecdote' to find out whether the word itself
11
word leaming strategy
contains information that would confirm the meaning inferred so far.
However, zooming in on 'anecdote' does not give any clues. Then the
practice item is concluded by looking up the definition of the word in the
dictionary. This reads (translated from the Dutch): 'a short story, usually
about a funny- event in somebody's life'.
The trainer and the students discuss the dictionary definition in
comparison with the definition as inferred from the context. The main
point in this discussion is not the correctness or incorrectness of the
definition inferred, but rather the status of the information found. In the
example of 'anecdote', the trainer points at the piece 'it is about my
grandfather falling through the ice when he was a young man'. Looking
back, this could have been decontextualized to 'an event in somebody's
life'. The trainer adds that, strictly speaking, you cannot tell from this
single piece of text what applies only to this text, and what applies to the
meaning of 'anecdote' in general; and that repeated occurrences of the
word must gradually make this more clear.
The design of the experiment
To assess the effects of the training program we used an
experimental design with blocking (Campbell 84 Stanley, 1963). Students
were assigned to matched pairs, and then one member of each pair was
assigned to the experimental group, the other to the control group.
Firstly, effects on the products of the students' inference activities
were measured by the results on pre- and post-tests for inferring word
meaning from context. Vocabulary pre-tests were used as covariates to
control for the effect of verbal ability. Reading comprehension post-tests
were included to measure possible transfer effects. Secondly, we explored
the program's effects on the students' inference processes in a think-aloud
sequel. We first describe the experiment and the assessment of the
product results. The think-aloud sequel is then reported.
Subjects
In order to select experimental and control students we started out
by pre-testing Dutch elementary school children (11 to 12 years old) in
four sixth-grade classes from four different schools. On the basis of the
results of these 93 students on the pre-test for skill in inferring word
meaning from context, 16 students from each class were selected, 4 from
each quartile of scores. From each set of 4 students, 2 were randomly
assigned to the experimental condition and 2 to the control condition. The
32 experimental students were to be trained in groups of eight students
each, and the 32 control students were to follow their regular lessons.
Training and control program
The students in the experimental condition followed the training
program described above, consisting of ten weekly one-hour lessons, eight
with Dutch material, and two with English material. The lessons were
12
word learning strategy
given during school hours. The control students took part in the regular
program in which no special attention was given to inferring word
meaning from context.
Trainer
All four experimental groups were trained by the same instructor,
who had also given the lessons in their try-out version. She was trained to
work along the guidelines that were provided in the trainer's manual.
Before each lesson, several possible ways of going about the word
learning items were discussed with one of the researchers, in order to
prepare the trainer as well as possible for her interactive teaching task.
Measures for the product results
Measures were paper-and-pencil tests, taken at three moments (see
also Table 1). Four tests were administered at moment 1, three weeks
before the training started; three tests at moment 2, about 10 days after
the eight lessons with Dutch material had been given; and two tests at
moment 3, one week after the two English lessons. In the following, the
mother tongue (Dutch) is indicated by Ll, English as a foreign language
by EFL.
Tests of skill in inferring word meaning from context
The tests of skill in inferring word meaning from context are
abbreviated throughout by SIWMC-tests. The L I SIWMC-test has two
parallel versions; version A is administered at moment 1 and version B at
moment 2. Both contain 25 open-ended items. Version A was used to
select experimental and control subjects (see the subsection 'subjects'
above.) Version B is the main criterion task for the training program, as
the test items are constructed in the same way as the training items.
The EFL SIWMC-test has three parallel versions; version A is
administered at moment 1, version B at moment 2 and version C at
moment 3. All three contain 12 open-ended items. Version A is used as a
covariate; Versions B and C are intended as measures of near transfer of
the skill to infer word meanings from context.
Before the administration of each SIWMC-test, knowledge of the
target words was checked in a short pre-test. The target words were
presented underlined in short non-informative sentences. The students
were asked to indicate if they knew anything about the meaning of each
underlined word, and if so to write down what they thought the word
meant. If a student had any knowledge about a word (and this was the
exception), the corresponding item in the S1WMC-test was not taken into
account for this student and the student's score was estimated on the basis
of the remaining items.
,
Vocabulary tests
The L 1 Vocabulary Test (moment 1) contains 60 four-choice
items. This test is used as a covariate in the analysis of covariance to take
into account pre-existing differences in vocabulary between students.
13
word learning strategy
The EFL Vocabulary Test (moment 1) is made up of 64 open-ended
items. Like its L1 counterpart, it is used as a covariate.
Reading comprehension tests
The Ll Reading Comprehension Test contains 24 four-choice
items. This test is intended as a measure of transfer of the skill in
inferring word meanings to text comprehension. This transfer can be
expected if the students, trained in the thorough analysis of a piece of text
in order to find information about an unknown word, use the same kind
of analysis to better understand what the text is about.
The EFL Reading Comprehension Test consists of five stories of
about ten lines each; four comprehension questions are asked about each
story (resulting in 20 items). This test, like its L1 pendant, is meant to
measure any transfer that occurs from the inference of word meanings to
text comprehension.
Hypotheses
The main hypothesis is that the trained students, compared to their
control classmates, have improved their skill in inferring word meanings
from context in Ll, resulting in higher test scores on the L1 SIWMC-test
version B. If confirmation of this hypothesis is found, three other
hypotheses are put forward: improvement in inference in L1 shows
transfer to improvement in inference in EFL for the trained students;
improvement in inference in L1 transfers to improvement in reading
comprehension in L1 for the trained students; both improvements transfer
to improvement in reading comprehension in EFL.
Analyses of the product results
Analyses
Analyses of covariance were used to compare the results of the
experimental and control students at moment 2 and 3. Measures taken at
moment 1 were used as covariates, in order to partial out pre-existing
differences on relevant variables. Dependent measures at moment 2 were:
the Ll S1WMC-test Version B, the EFL SIWMC-test Version B, and the
L1 Reading Comprehension Test. Dependent measures at moment 3 were:
the EFL SIWMC-test Version C, and the EFL Reading Comprehension
Test.
The internal consistency (coefficient alpha) of the dependent
measures mentioned above ranges from .68 to .97 (see Table 1). For all
dependent measures the homogeneity of variance was tested by
computing Cochran's C; the p-values concerned ranged from .48 to .83,
meaning that this condition for analysis of covariance is fulfilled.
Resuits
Means and standard deviations for all measures are given in Table
1. The analyses of covariance are summarized in Table 2a to 2e.
14
word learning strategy
[Table 1 and Table 2a to 2e here]
We tested at alpha =.05. As Tables 1 and 2a show, the difference in mean
scores between the experimental and the control students on the most
relevant dependent measure, L1 SIWMC-test (B), with L1 SIWMC-test
(A) and L1 Vocabulary as covariates, is not significant. The effect size of
the training program on the L1 SIWMC-test (B), estimated by taking the
difference between the mean scores of the experimental and control group
and dividing this by the pooled standard deviation: (8.52 minus
Olkin, 1985). Correction for attenuation
8.24)/3.76, is .074 (Hedges
due to measurement error results in an effect size of .084.
The absence of a significant difference on the L1 SIWMC-test (B)
makel it understandable that no difference is obtained on the EFL
SIWMC-test (B) and (C), as these two tests were meant as measures of
near transfer. No difference was found either, on the measures of far
transfer (L1 and EFL Reading Comprehension).
The think-aloud sequel to investigate the inference process
To assess the process effects of the training program, a think-aloud
sequel was conducted, in which we gathered think-aloud protocols from
trained and control students. This sequel is described in the next section.
Subjects
We selected 16 trained students from the whole range of scores on
the L1 SIWMC (B) test. Four students came from quartile 1, three from
quartile 2, four from quartile 3 and five from quartile 4. These students,
and their 16 controls, were asked to take part in an individual think-aloud
session. The sessions were run a few weeks after the training program
had ended.
The words and contexts used
Ten difficult L1 (Dutch) words were selected (see Appendix A),
that we expected to be unknown to the students. We chose ten words
(eight nouns and two verbs) that refer to relatively complex concepts, and
for which no simple synonyms exist, because we expected these words to
give rise to rich inference processes. The students would have to find out
to what concept each unknown word refers and give a description of its
meaning by using more than one word. Three difficult EFL (English)
words (one noun and two verbs) were selected in the same way as was
done for the lessons.
The contexts for the unknown words were constructed in the same
way as described for the training material, that is as 'pedagogical
contexts' (Beck, McKeown & McCaslin, 1983), meant to give relevant
information about the unknown word. Each context described a situation
in which the unknown word was fairly typically used.
15
word learning strategy
Procedure
The trained students were given the instruction to infer the
meaning of the underlined word in each piece of text, while thinking
aloud. They were given one or two practice items to ger used to the
think-aloud situation. The material that they had used during training was
available on their desk: the Strategy Card and practice schemas. The
experimenter told the students that they could use this material if they
wanted, but that they didn't have to. The control students did not have
material from the training program at their disposal, but in all other
respects the procedure for them was the same as for the trained students.
A session per student consisted of two rounds, each round lasting
about half an hour. In round 1 the items were handled in the following
way. For each item the experimenter would ask: "do you know the
meaning of (the target word)?". If the meaning was known (which was
the exception) the item would be skipped. If the meaning was unknown,
the student read the piece of text aloud and tried to infer the word
meaning from it while thinking aloud. The role of the experimenter was
limited to occasional encouragement to the student to keep thinking aloud.
If the student was satisfied with what he or she had figured out about the
word meaning, he or she told the experimenter to write down the resuh
on the word list that the experimenter had before her. This moment
marked the end of the item. After all items were worked through in this
way, round 1 was concluded. The think-aloud material, gathered during
this first round, is considered as the protocols proper.
Round 2 consisted of a retrospection of what had been written
down on the word list in round 1. The material from round 2 was used to
solve ambiguities in the round 1 material.
Sessions were recorded on audiotape for each student, resulting in
32 think-aloud protocols for 8 to 10 items in L 1 and 3 in EFL. The L 1
material resulting from round I was analysed in the way described in the
next section. The material concerning EFL will not be reported here.
Analysis of the protocols, resulting in process codes
The protocols of the 32 students were transcribed from tape, in
order to be coded for the presence or absence of elements from the
training program. To code the occurrence and the quality of elements
from the training in the protocols a coding manual was developed.
Thirteen training elements were coded as present or absent; for the
elements present a quality code was given of 0, 1 or 2 for the cases that
the element was used wrongly, half correctly or correctly, respectively.
Note that the elements do not exclude one another. Theoretically
speaking, each of the items could contain all elements, because the
students were free to produce as many utterances and guesses while
thinking aloud about an item as they wanted. Nevertheless, in most
protocols only a number of elements is present.
Coded were parts of protocols within an item that form a
meaningful unit, being made up of one or more related utterances. The
codes assigned are listed below and illustrated by examples. The contexts
16
word leaming strategy
used for the ten target words are given in Appendix A.
01
02
03
04
05
06
07
Using a cloze sentence. A protocol part was coded 01
when the literal sentence in which the target word occurred was
used as a storting point to get an idea about the target's meaning,
by deleting the target and considering the gap, or by filling out a
word that might mean the same as the target. An example
utterance is: "Dilemma, yes, it is a real dot dot dot ..."
Using a Wh-question. A protocol part was coded 02 when
a Wh-question was asked concerning the content of the situation
described in relation to the target word. An example is the way in
which the student, quoted above, continues: "'What is happening
with dilemma?', well, she does not know what to choose!"
Using the word form to get information about the word
meaning. Code 03 was assigned to protocol parts in which
attention was given to the form of the target word, for instance to
a constituent and its meaning. An example is given by the student
who, looking at the target word 'personalia' remarks: "well, the
word already tells it a bit, of a person, is already in it a bit ..."
Checking the word meaning in the cloze sentence. A very
clear example of a code 04 utterance is given by the student who
has inferred the meaning of 'receipt' as: "a kind of proof that you
have returned the videotape". He or she continues to substitute this
meaning in the sentence in which the word 'receipt' is presented:
"yes, that fits in it, I think, 'In most video hire shops you can ask
for a proof that you have returned it properly if you are returning
a videotape' .. yes, so that fits".
Checking the word meaning in the whole fragment of text.
An example of this more global checking is given by the student
who infers that a dilemma is a problem and says: "well a problem
.. I don't know, it is a problem that you don't know what to do,
with whom to go and things like that .. I think, well, you can
write down 'problem'".
Using one's own experience. This code was assigned to
parts in which a comparison was made between the situation
described in the context and a situation known from one's own
experience. This is an example for dilemma: "That happens to me
too, sometimes. I want to go and sleep over with my friend, and
that very evening at home there are pancakes for dinner!"
Elaborating the contextual information without
decontextualization. This code was assigned to protocol parts in
which links were made within the context itself, between different
parts of it, without abstracting from the context in any way. An
elaboration utterance for dilemma is: "I think that it, say it is
very difficult what she has to do. You have to make a decision
yes dilemma .. you think you want to go on a holiday, but still
you want very much to go to your sister's wedding day .. but on
the one hand you think: my sister's wedding day is only one day,
17
08
09
10
11
12
word learning strategy
say, and then on the other hand a holiday is a whole week, so you
would miss a week's holidays".
Decontextualization. A protocol part was coded 08 when it
gave evidence of abstraction from the context in any way, as in
the following protocol part: ".. I think when you have two things
at the same time, that then you .. that then it is called a dilemma
or something ..". This student has abstracted from the specific
alternatives mentioned in the context to: 'two things at the same
time'. Decontextualization is also evident here in the change from
the 'I' in the context to the more general 'you' in what the student
says.
Using meta-knowledge. A protocol part was coded 09
when it gave evidence of knowledge relevant to the process of
finding the target word's meaning. An example for 'personalia' is:
"I do get what they mean, but I don't know how to explain, your
.. a collective word for .. for .. where you are bom, your first
name, your family name and things like that".
Producing a superordinate category. Code 10 was assigned
when a superordinate was mentioned as a candidate for the target
word's meaning. An example is given by the student who infers
that a receipt is 'a kind of proof that you have returned the
videotape', in which 'a kind of proof' is a clear.indication of a
superordinate.
Producing a specification of the superordinate. Code 11
was assigned when a specification of the target word's meaning
was mentioned, that could be attached to the superordinate. In the
piece of protocol quoted above for code 10, 'that you have
returned the videotape' is coded as a specification; the student
preserves this specification in a more general form in the resulting
word meaning as 'that you have returned it'.
Producing a synonym. Code 12 was assigned when a
synonym was mentioned as a candidate for the target word's
meaning, as done by this student for dilemma: "well .. a bit of a
choice, a 'two-choice' or something ..".
Producing a circumscription of the word meaning in a nonconventional form. Code 13 was assigned to protocol parts in
which the meaning of the target word was described in a form
different from synonyms, superordinates and specifications. An
example is the circumscription 'that you don't very well know
what to do' for dilemma.
Rereading the context or part of it. The fourtéenth element,
rereading, was added during the try-out of the coding manual, as
this activity (one which had not been part of the training program)
proved to occur quite often. Only pure rereading activities were
coded as 'rereading', apart from other activities (like substitution
or checking activities) that involved rereading with a special
purpose.
.
13
14
18
word learning strategy
To distinguish between producing a synonym (12) and producing a
superordinate category (10) it was decided to code one-word suggestions
without specifications as a synonym, unless there was a clear indication
that the student wanted to specify this one word but did not know how to
do it. For instance, for 'dilemma', one of the students states that it is a
kind of problem, but he can't really say what kind; this case is coded as
the production of a superordinate category. But when a student says that
'dilemma is a problem', this is coded as the production of a synonym,
although strictly speaking this student might mean 'a problem' as a
superordinate rather than a synonym.
All protocols were coded by one judge, and two to three items per
subject (in total a quarter of all protocols) were coded independently by a
second judge. Percentage agreement between the judges ranged from .80
to 1.00 for the different coding categories, with a mean percentage of
agreement across categories of .94. The codes given by judge 1 are the
data which are analysed in the following.
Apart from the process codes, we evaluated the word meanings
produced by the students at the end of the think-aloud process (round 1).
The results concerning these definitions are presented after those for the
process codes.
Protocol differences between trained and control students
We compared the frequency and quality of the process codes and
four aspects of the resulting definitions of the trained and control
students. In Table 3 and 4 the results are given for the process codes.
Table 5 and 6 show the results for the definition aspects.
Analysis of the process codes
Frequency of occurrence of the process codes
In Table 3 results are given regarding the frequency of the process
elements coded for the trained and control students.
[Table 3 here]
The codes given were counted per code over items per student and
corrected for the items that were missing for the student concerned (the
minimum score being 0, the maximum 10).
We expected the trained students to use the elements 01 to 09
more often than the control students. The elements 01 to 05 have been
part of the training program as given, and elements 06 and 07 are
possible consequences of the Wh-questions as used in the progam.
Occurrences of element 08 and 09 are possible consequences of the
training program as a whole, as the need to decontextualize and to think
about what it takes to find a word's meaning have been the subject of
training throughout. As for elements 10 to 13 we expected the trained
19
word learning strategy
students to suggest superordinates and specifications more often than
synonyms in comparison with the control students. Moreover, we
expected them to give fewer circumscriptions in a non-conventional form.
This expectation was based on the presentation of the definition format in
the training program, that was meant to do two things: remind the
students of the possibility of giving the meaning of the target word with a
synonym; and making them familiar with the possibility of giving it in
the form of a definition (superordinate and specification).
In sum: we predicted that the trained students would show a more
frequent use of elements 01 to 11 than the control students, so these
differences were tested one-tailed. As we expected the trained students to
use the elements 12 and 13 less often than the controls, these differences
were also tested one-tailed, in the opposite direction. For rereading (14),
added during the try-out of the coding manual, we had no prediction so
this element was tested two-tailed.
In view of the small sample sizes the differences between the
groups were tested by the non-parametric Kolmogorov-Smirnov test,
which compares the frequency distributions of two samples (Siegel 8z.
Castellan (1988). Dni n is the relevant test statistic. The critica' value of
mnD n, n with m=16 and n=16, at alpha= .05 is 112 in a one-tailed test and
128 in a two-tailed test. Values exceeding the critica' value in Table 3 are
marked with an asterisk. Below, the results are considered in order of
process element.
Using a cloze sentence (01)
The trained students use a cloze sentence in about one third of the
opportunities, which is significantly more often than their controls. This
means that the use of a cloze sentence does occur spontaneously with
students of this age, but that the training program has succeeded in
enhancing the use of this element.
Using a Wh-question (02)
Explicitly stated Wh-questions are used with a mean of less than 1
occasion per 10 items by the trained group. Because of this low frequency
the difference with the control group is not significant. Stil!, it is
interesting to note that the control students did not once use a Whquestion explicitly. Asking such questions is probably not part of the
spontaneous repertoire of students of this age. The implication of this is
that the trained students who use Wh-questions have learned to do so
during training.
Using the word form to get information about the word meaning (03)
Even when taking into account that only 2 of the 10 items had a
word form that gave information about the word meaning ('personalia': a
word derived from 'person', and 'hand- en spandiensten': a Dutch
compound word), word form is used very little. There is no difference
between trained and control students in this respect.
20
word learning strategy
Checking the word meaning in the cloze sentence (04) and in the whole
fragment of text (05)
The trained students check significantly more often what they have
inferred from the context than do the control students. This applies to
both methods of checking discerned in the protocols: by substitution of
the inferred meaning in the cloze sentence, and by considering the fit with
the fragment of text as a whole.
Using own experiences (06) and elaboration of the text (07)
Using own experiences (06) and elaboration of the text (07) were
not explicitly trained, but both can be seen as the consequences of
implicitly asking Wh-questions. The trained students use their own
experiences significantly more often than does the control group.
With regard to elaborations, the difference between the observed
frequencies in the two groups is in the predicted direction, but is not
significant.
Decontextualization (08)
All students, trained as well as control, show decontextualization
activities with almost all items. So all students demonstrate that they
performed the task of inferring the meaning of the unknown word while
thinking aloud as a decontextualization task.
Using meta-knowledge (09)
The frequency of use of meta-knowledge by the trained students is
very low, so the difference with the control group is not significant. Still,
as is the case with the Wh-questions, it is interesting to note that the
control students do not once use meta-knowledge. It seems that using
such knowledge does not occur spontaneously with students of this age in
this type of task.
Producing a superordinate category (10), a specification of the
superordinate category (11) and/Or a synonym (12)
We expected the trained students to use fewer synonyms (12) than
the control students and more superordinates (10) and specifications (11),
as the use of the last two elements could indicate the use of the definition
format in finding the new word meaning. The differences between the
observed frequencies in the trained versus the control group for
superordinates, specifications and synonyms are in the opposite direction
to that predicted. However, none of these differences is significant. We
shail return to this point in the next section.
Producing a circumscription in a non-conventional foren (13)
We expected the trained students to use fewer circumscriptions
(13) than the control students, in favor of the use of more superordinates
and specifications. The difference between the observed frequencies in the
two groups for circumscriptions is in the predicted direction, but is not
significant.
21
word learning strategy
Rereading the context or part of it (14)
As stated above, the coding of pure rereading activities was added
during the try-out of the coding manual, as this process element proved to
occur quite often. The difference between the two groups was therefore
tested two-tailed. Pure rereading turned out to occur significantly more
often in the trained group.
In sum, we have found a significant difference between the trained
and control students in the direction predicted, concerning the frequency
of foor elements:
1. the use of a cloze sentence
2. checking of what is inferred from the context in a cloze sentence
3. checking of what is inferred in the text as a whole, and
4. the use of their own experiences.
At least some of the trained students also turned out to use Wh-questions
and metacognitive knowledge, whereas none of the control students did.
Although the use of metacognitive knowledge is not very frequent (less
than once per 10 items) in the trained group, this may still be an
important result. We illustrate this by the way the student in the following
protocol part deals with 'dilemma'. He says: "(...) it is a real dilemma,
yes, a bit of a rotten situation I think, yes, but I don't know if it is álways
so that you have to choose between two things, that I don't know (..) it is
a real choosing problem a bit, you have to choose between two things,
well: problem with choosing". This student is, as it were, negotiating with
himself about the status of 'between two things', and as a result decides
to leave it out of the word meaning for now. He thereby demonstrates
that he knows that infoimation about the meaning of a word from one
context has to be confirrned in others. If he encounters the word
'dilemma' again in the context of having to choose between two things,
he may remember his earlier weighing-process and then add the aspect
'between two things' to the meaning of 'dilemma' in his semantic
memory.
Finally, the trained students turned out to reread the context (or
part of it) significantly more often than the control students did, although
rereading was not taught as such in the training program. For no element
was a significant difference found in the direction opposite to that
predicted.
Quality of the process codes
Table 4 gives the results on the quality of the process elements
codedo
[Table 4 here]
Each process element, as it occurs in the protocols, was given a score of
0 (if it was used wrongly), 1 (if it was used half-correctly) or 2 (if it was
correctly used). For each student the quality score per element was
determined by taking the sum of the quality scores obtained and
22
word learning strategy
extrapolating this from the number of items that the student concerned
had completed to the total number of 10 items (the minimum score being
0, the maximum score 20). For each element, the number of students for
which the element was found in the protocols varies; in Table 4 this
number is therefore given per process code. Again, D m , is the
Kolmogorov-Smirnov test statistic used. The critical value of mnD ri,
varies with m and n. We expected the trained students to use all elements
with more quality than the control students, so we tested all differences in
the predicted direction (one-tailed) at alpha= .05.
For none of the training elements were the differences in quality
between trained and control students significant. But the difference in
quality of elaboration (84, for which alpha= .10) indicates a trend for the
trained students to make better elaborations than their controls. As was
stated above, an utterance by the students was coded as an elaboration
when a part of the context was reformulated in a context-bound way, and
without the use of the students' own experiences. Elaborations can be
seen as the consequence of implicitly asking Wh-questions (like 'who
does this?', 'what does he do?') and are very useful as a bottom-up
activity to start the decontextualization process.
Conclusions concerning the process codes
We conclude that the trained students have gathered a number of
process supports from the training program that might help them to infer
the meaning of an unknown word. These are the use of a cloze sentence,
the use of their own experiences, checking activities and rereading. Some
of the trained students have also come to use Wh-questions to search the
context for information' and some of them use meta-knowledge about
word meanings. We shall look now at these supports in turn.
A cloze sentence 'can help to start up the inference process, by
steering into the direction of a suitable word meaning, that can serve as a
model for the target word meaning. Later in the process such a sentence
can be used to check whether the inferred word meaning fits. In the next
section we will see that checking in this way has its limitations, and that
checking more globally by considering an entire segment of text may be
the preferable way. That trained students check more often is an
important result in itself, as checking is notoriously difficult to train (cf.
Veenman, in press).
Students' use of their own experiences can support the inference
process by triggering ideas about what the unknown word may mean. We
think this use is a consequence of explicitly or implicitly asking Whquestions such as: 'What is the matter in this text?', 'What do I know
about this kind of situation?' and the like.
Rereading the context as such was not an element of the training
program. Nevertheless the fact that the trained students reread more often
can be interpreted as a more global consequence of the program, as the
strategy trained continually stressed the context as a source of information
about the meaning of the unknown word. Clearly, rereading is a good
preparation for other steps in the inference process. (It is relevant to note
23
word learning strategy
here that a recent instructional intervention to influence the process of
deriving word meaning from context by Goerss, Beek & McKeown (in
preparation) includes rereading of the context as a separate component.)
As for the quality of the process codes, no significant difference emerges
between the two groups. However, there is a trend for the trained students
to make better elaborations, which is possibly a result of the higher
frequency of rereading.
That some of the trained students have come to use metaknowledge is a promising finding, because knowledge about what it takes
to find the meaning of an unknown word can stimulate inference
activities and thereby future performance.
Finally, it is important to note that there is no evidence from the
protocols that the trained students unlearned the relevant process elements
that the control students spontaneously use.
Analysis of the definitions as the products of the think-aloud process
In addition to the process codes, we analyzed the word meanings
as stated by the trained and control students at the end of the think-aloud
process. The instruction for the students had been to give the meaning of
the target word "as it might be found in a dictionary". In the following,
we refer to the answers of the students by the term 'definitions',
However, the reader should keep in mind that these definitions can have
different formats, including a one-word format in the case of a synonym.
In fact, the definitions were scored on four aspects: quality,
decontextualizedness, part of speech and format.
Quality
Quality was considered in terms of 'communicative adequacy',
regardless of the format in which the definition was stated. So what
mattered was the content of the information conveyed. The format of the
definitiOn was coded separately (see below). Quality was judged on a
scale from zero (inadequate) to 6 (fully adequate). The maximum score
per student is 60. Quality was evaluated by two independent judges.
Differences in the scores given were solved by discussion.
Decontextualizedness
Decontextualizedness, that is the degree to which the definition is
abstracted from the context, was judged on a scale from zero (very
context-bound) to 4 (highly context-free). The maximum score per student
is 40. This aspect was also judged by two independent judges, and
differences in the scores given were solved by discussion.
Part of speech
Coded here depended on whether the part of speech of the
definition as stated (noun or verb) was the same as or different from that
of the target word. Per item a score of 1 (same) or 0 (different) was
given, so the maximum score per student is 10. Part of speech could be
24
word learning strategy
coded objectively.
Format
The definition as stated could have different forms: the form of
one word; of one word plus (one or more) specifications; or a form
different from these. The format of the definitions could be coded
objectively in these three categories. (Note that the three formats are
mutually exclusive, because the students had to state the definition of the
word they had arrived at after thinking aloud.)
Results concerning the four definition aspects
Results concerning the definition aspects are presented in Table 5
and 6. In Table 5 the results are given on quality, decontextualizedness
and part of speech of the definitions produced by the trained and control
students.
[Table 5 here]
The differences between the two groups are tested non-parametrically.
D„,„ is the Kolmogorov-Smirnov test statistic used. The critical value of
mnD mn at alpha= .05 in a one-tailed test is 112 with m=16 and n=16. For
none of the comparisons made is the difference significant.
The quality scores indicate that the experimental task was a
difficult one for all students, the mean score being 29.03 and 25.81 for
the trained and control group respectively, which is less than half of the
maximum possible score of 60.
Table 6 shows the data on the format variable.
[Table 6 here]
In Table 6 the frequency of occurrence of the definition formats is given.
The difference between the distributions of numbers over the three
formats is tested by the Chi-Square statistic. For the comparison between
trained and control students the value of Chi-Square is 7.35, for which p<
.06 (DF=3). So in comparison to their controls, trained students tend to
give more one-word definitions and fewer definitions in the form of oneword-plus-specifications.
A one-word answer can be meant by the student concerned as a
synonym or as a superordinate that has to be specified later. Throughout
the training program the use of a synonym was presented as one of two
ways of stating the meaning of the unknown word, the alternative being
the formulation of a definition in the form of a superordinate category
with one or more specifications. It was also made clear that often no
exact synonym exists for an unknown word. We expected the training
program to teach the students not only to look for the meaning of the new
word in the form of a synonym, but also in the form of a definition. So
the trend in the results is contrary to what we expected.
25
word learning strategy
To gain more insight into this format difference between trained
and control students, we have looked at the format of the definitions
given at the L1 SIWMC-test version A, taken before the training, and at
the same test version B, taken about ten days after the eight L I lessons.
The results are given in Tables 7.1 and 7.2.
[Tables 7.1 and 7,2 here]
A salient fact revealed by Table 7.1 and 7.2 is that the majority of the
answers is in the foren of one word, for control as well as trained
students, both before and after training. The observed frequency of oneword answers given by the trained group is slightly lower than that for
the control group before training, and somewhat higher after training, but
the values of Chi-Square (.45 for which p< .80 and 3.88 for which p< .14
respectively) indicate that the differences are not significant.
If we compare the figuren in Tables 7.1 and 7.2 with those in
Table 6 we can conclude that the paper-and-pentil tests elicited mainly
one-word answers (91% in both groups before training, after training 93%
in the trained group and 89% in the control group). The think-aloud
situation clearly elicited more diversity in format of definition for both
groups (one-word answers making up 41% for the trained students and
30% for the controls).
The trained students' relative preference for one-word answers
may have been related to the use of a cloze sentence. We explored this
possibility by inspecting cross tables with raw frequenties of use of a
cloze sentence and of one-word definitions for the 16 trained and 16
control students, with 10 items. Within the control group 6 'out of 14
incidents of cloze sentence use (43%) go together with a one-word
formulation, and this figure is 27 out of 46 (or 59%) within the trained
group. In both groups, then, there is a relation between the use of a cloze
sentence and the use of one-word definitions.
This observation brings us to a problem inherent in the process of
inferring word meanings from context. In using the sentence with the
target word as a cloze sentence, a student can get off to a good start. For
one thing the position of the unknown word in the cloze sentence gives
information about the part of speech; further, it can give the student a
good idea about a word meaning that could be helpful as a model for the
meaning of the unknown word. When a one-word solution is found, it is
easy to check this solution in the sentence. If this fits syntactically and
semantically, the inference process is satisfactorily finished. A problem
arises when no one-word solution exists, as is the case for words that
refer to complex concepts. In this case the cloze sentence can still be
helpful in generating a model word meaning, but when it is clear (for
instante because of information in the sentences surrounding the cloze
sentence) that this model needs specification, it then becomes more
difficult to check this more-word meaning in the sentence. It is easier to
check whether 'problem' fits the cloze sentence for dilemma than it is to
check 'problem of choice between two things, both of which you want to
26
word learning strategy
do'. Some students solve this difficulty by looking for a compound word
in which superordinate and specification are merged. An example is given
by the student who tried to integrate two separate aspects inferred for
'receipt', that it is a proof and that it is a voucher, into 'proof-voucher'
('bewijsbon' in Dutch). All the same, it is not impossible to use a cloze
sentence to check a rather complex word meaning, as is illustrated by the
example for the target word 'receipt' that we gave earlier in this paper to
illustrate the process code 04, and that we repeat here: "yes, that fits in it,
I think, 'In most video hire shops you can ask for a proof that you have
returned it properly if you are returning a videotape' .. yes, so that fits".
For a training program in inferring word meanings from context,
the reasoning above implicates that the cloze sentence should be
presented to the students as a stepping stone: helpful in getting started
with the inference process, but not to be used too rigidly in the checking
phase.
Summary of the differences between trained and control students
Here we summarize the results mentioned above in terms of the
theoretical basis of the training program. As was described earlier in this
paper, the program was meant to teach the students to use a number of
text-driven, schema-driven and self-regulatory process elements. We shall
now discuss these in turn.
As for the text-driven elements we saw that the trained students
learned to use the cloze sentence, both to get the inference process started
and to check the result. At least some of them have also learned to ask
Wh-questions with respect to the unknown word, a phenornenon that is
altogether absent in the control group. Wh-questions, although not
spontaneously used at all, may be a useful instrument in inference
processes, and worth teaching to students of this age, because these
questions can result in linking operations within the context as well as
between context and stored knowledge. The implicit use of Wh-questions
may underly trained students' more frequent use of their own experiences.
This implicit use may also have resulted in the better quality of the
elaborations made, that is, in the better quality of the connections
explored within the context itself.
Word form, an element that could be called word-driven rather
than text-driven, was used so linie that we cannot interpret the findings
pertaining to it. The 'Zoom-tactic' that was meant to make students look
at word form was the third tactic presented during training, and it was
probably not given enough attention. It is also possible that the tactic did
not appeal to students because with many unknown words it does not
yield much information. However, it is probably worth teaching the skili
of using word form in its own right, as the research by Anglin (1993)
indicates. Anglin's empirical results show that children add a certain
amount of words to their vocabularies every day by learning words as
such, but also a second amount by applying morphological analysis and
composition to potentially knowable words. This second amount becomes
27
word learning strategy
more important with age during the elementary school years.
As for the schema-driven elements, we must look at the
consequences of the Definition Schema, as presented in the training
program. This schema was intended to teach the students two ways of
stating the meaning of the target word: one being a synonym, and the
other a definition composed of a superordinate category with
specifications. As we have seen, during the inference process the trained
students did not use more superordinates or specifications than did the
controls. We have the impression that they preferred to use synonyms,
because they found the definition form too complicated. But the finding
that at least some of the trained students use meta-knowledge during the
inference process, that is, knowledge about what it takes to find the
meaning of an unknown word, may be a more genera' result of the
practice that these students had with the definition format.
The aim of the use of text- and schema-driven elements was to
stimulate the decontextualization process. No differences were found
between the two groups of students in this respect. On the one hand, the
task of inferring the meaning of the unknown word while thinking aloud
has probably stimulated all students, both trained and control, to
decontextualize to some degree. On the other hand, the trained student§
practiced decontextualization during training in the form of dialogues,
while during the think-aloud task no dialogue with the experimenter was
allowed. Probably the students had too little practice during training to
internalize the decontextualization activities and have the dialogue, as it
were, within themselves.
A last ingredient of the training program was formed by its selfregulatory elements. These were meant to make the students aware of the
context as a valuable source of information about the unknown word, to
make them aware of the status of the information inferred, and of the
need to check this information. That the training program succeeded in
heightening this awareness or bringing it about is indicated by the higher
frequency of rereading and checking operations by the trained students.
An example of this awareness is given by the student who, asked about
what she thought she had learned from the training program, answered:
"(I learned that) if I don't know a word, what I have to do, that I don't
have to look in a dictionary right away, but that I simply .. can see, first
read on, and then see if I can find it out for myself". This result is not
unimportant in view of the many opportunities that will arise for this
student to learn something in this way. (See also Nist & Olejnik, 1995, p.
188, who mention the fact that even college students don't seem to spend
much time looking into context where this would be a sensible thing to
do.)
The higher frequency with which trained students used the process
elements mentioned above does not have significant consequences for the
quality of these activities (except for the quality of elaborations). The
same goes for the quality of the new word's definition as formulated at
the end of the think-aloud process. If the training program had consisted
of more practice, with a larger variety of items, the frequency of the
28
word learning strategy
operations performed might have resulted in a better quality of these
operations and consequently in more adequate definitions. As to the
format of the definitions, we found that the trained students showed a
preferente for one-word definitions over more complex ones, while for
the control students a reverse trend is visible. As we have shown above,
this phenomenon is probably related to the use of cloze sentences.
General conclusions and discussion
On the basis of the protocol analyses we conciude that the strategy
as trained has had effect on the frequency of a number of inference
activities executed by the trained students. This effect is promising in
view of the future performance of these students. If they actually go on to
perform some of these activities on encountering unknown words, this
means that an inference process has started and may be maintained long
enough to yield information about the word meaning that can be
elaborated further at every subsequent confrontation with the same word.
Hardly any effect was achieved with regard to the quality of the
activities performed by the trained students, nor on the quality of the
resulting definitions. This last finding is in accordance with the fact that
no significant effect of the training program was found on the paper-andpencil tests administered prior to the think-aloud sessions.
To sum up, the trained students performed more relevant inference
activities but the quality of these activities left room for improvement.
Our endeavor to teach students to deal with the task of inferring word
meanings from context in its full complexity, by training them to use the
Word Learning Strategy as a whole, would probably ask for more than
the limited number of lessons that we had available. The protocol
analyses provide more insight into the specific difficulties the students
were confronted with during training. In the following sections we look at
the strategy as trained in relation to the task difficulties; the Jatter are
discussed first.
The inherent difficulties of inferring word meaning from context
The inherent difficulties of inferring word meaning from context
once again become clear from the think-aloud protocols of the
experimental task. Let us look more closely at these difficulties, which
earlier in this paper were compared to those of a bootstrapping process.
The instruction given to the students was to infer, as well as they
could, the meaning of the target word from the context given, and to
arrive at the word meaning "as it might be found in a dictionary". This
instruction aims at a decontextualization process, starting from the context
and working towards an end product in the form of either a suitable
synonym or a (more or less) formal definition. In order to succeed at this
task the students must come to understand the context by making
connections within the text itself (what we have called 'elaborations') and
by making connections between the text and information already known
(for instante by looking for an analogous situation from their own
experience). By doing this the students must decontextualize the
29
word learning strategy
information that is related to the unknown word from the text and state it
in the form of a more or less conventional definition. In the protocols we
saw that both one-word solutions and definitions have their own
difficulties. As the target words refer to complex concepts, no simple
well-known synonyms exist for these words. Looking for a suitable
synonym therefore means trying to construct a compound word that
covers different aspects of the word meaning as inferred from the context.
1f such a compound word cannot be found or constructed, the student can
opt to settle for a 'synonym' that captures only part of the target word's
meaning, often an overly general term that would be more suitable as a
superordi nate.
A formal definition may be even more difficult to draw up, as this
entails finding a suitable superordinate as well as specification.
Superordinates for complex concepts tend to be abstract ('mark' for
'stigma', 'data' for 'personalia'). Without a superordinate it is allo
difficult to find a specification, unless the student uses an 'empty'
superordinate such as 'something'. Students who are not aware of the
difficulties of finding a synonym or a definition, or who try to circumvent
them, tend to give circumscriptions of the target word's meaning in a
different form such as: 'dilemma means that you don't know what to
choose'. Although these circumscriptions can be fairly adequate from a
short-term communicative point of view, they often cover only part of the
target word's meaning, and in a form that is not fully decontextualized.
This form may not lend itself very well to integrating the new word
meaning into the structure of the semantic network in due time, nor to
future formal operations to be executed on parts of the word meaning,
such as differentiating dilemmas from other kinds of problems (see Van
Daalen-Kapteijns & Elshout-Mohr, 1981).
The Word Learning Strategy as it was trained
In the training program the two main means for the
decontextualization process were presented consecutively in the 'Tracktactic'. The students were taught to follow up the text-driven help first
(cloze sentence and Wh-questions) and then the schenia-driven help
(filling out the definition schema with either a synonym, or a general and
specific characteristic). The results indicate that the students learned to
use the text-driven elements more than the schema-driven elements. The
fact that the students found the Jatter more difficult to use may have
contributed to this outcome.
Looking at the strategy now, we think that it was too complex to
be taught in a limited number of lessons. It might have been wiser to
present the students with one help method at a time, either the text-driven
help to work from the text upwards, or the schema-driven help to
approach the context with the definition schema as a search plan. This
thought is in line with the finding by Hattie, Biggs & Purdie (1996, p.
129), who carried out a meta-analysis on the effects of learning skilis
interventions. They found that single-component interventions, aiming at
near transfer of a specific task-related skill, are more effective than
.
30
word learning strategy
multiple-component interventions. On the other hand, as was mentioned
in the introduction to this paper, a limited training program reduces the
generalizability of the effects attained beforehand, and leaves the task of
integrating the components still to be done.
Below we sketch three outlines for simpler and perhaps more
successful training procedures. In order to compensate for the potentially
limited effects of each simple procedure as such, a more complex training
program could consist of a combination of simple procedures.
Suggestions for more simple training procedures
A text-driven procedure
A 'text-driven' training procedure would consist of a short series
of lessons presenting elements like: starting from the sentence in which
the target word occurs and asking Wh-questions in order to understand
the context in relation to the target word. The lessons would have to
include instruction in how to reformulate the information gathered into a
more or less decontextualized form. Note that the use of questions is also
useful for other verbal tanks, such as reading comprehension and for
studying text (see McKeown, Beck & Worthy, 1993; Overmaat, 1996, p.
131-134).
A schema-driven procedure
A 'schema-driven' procedure would consist of a short series of
lessons that would acquaint students with schema-driven elements. First
the students would practice with word meanings very well known to
them, by formulating aspects of meaning in a conventional form which
fitted the definition format. This could be done by practicing with the
Definition Schema as used in the Word Learning Strategy or by
practicing with the Word Map (see Schwartz & Raphael, 1985). They
would then learn to use this format as a couple of empty slots to be filled
by searching the context for the required information. Students would also
practice the use of the schema as a means for encoding and remembering
selected information about the meaning of a word in 'a simple,
verbalizable form. The schema could then function as an aide to forming
and retaining the word meaning as a stable unit, a unit that is suitable as
a constituent in a semantic memory structure.
A model-driven procedure
Still another training procedure is feasible. This would start from
the Tinding that most students seem to prefer the simplest case: that of
trying to find a synonym, or at least a one-word solution that more or less
captures the meaning of the unknown word. Put differently, this series
would start from the 'primitive routines that get the job done' (Garner,
1990, p. 519).
The strategy to be taught to the students would have this for step
1: look for one word that you think has a meaning close to that of the
unknown word and use this as a model word (see Van Daalen-Kapteijns
31
word learning strategy
& Elshout-Mohr, 1981). Step 2 would be: see if you can add something
to this model word, to make the meaning more precise. In the lessons
concerned something would have to be said about the phenomenon that
not many words have exact synonyms, and that in most cases a different
word has a different or at least partially different meaning. Step 3 would
consist of checking operations. First, the model word plus addition would
have to be checked in the context from which it was inferred. (In view of
what was said above about the limitations of a cloze sentence, this check
would have to be done in the entire segment of text rather than in the
sentence in which the target occurs.) Ideally, the plausibility of the word
meaning would also have to be evaluated in the light of the broader
semantic field. A study by Nagy
Scott (1990) showed that students
from different grade levels (7th and 10th graders and undergraduates) do
indeed apply word schemas in evaluating hypotheses about possible
meanings for unknown words.
This strategy, starting from a model word, differs from what we
have tried to do in the training program as it was given. The program
offered the students the definition format as an empty model to be filled
from the context, but students might find it easier to start with an already
filled model that they then adapt with help from the context.
The training procedures, the core of which is sketched above,
would have to be enriched by elements of meta-knowledge, i.e.
knowledge about what it takes to arrive at a more or less
decontextualized, conventional word meaning, and by elements of selfregulation, i.e. awareness of the need to use the context and check the
information inferred.
In our view, each of these procedures is an interesting candidate
for future research. If one or more of these should prove to be successful,
these are better candidates for use in the classroom than a more complex
program. Note again that the three procedures, embedding different strategies, would not have to exclude one another. It is conceivable that they
are all part of a wider curriculum in which teaching students to become
strategie in the development of their vocabulary has an important place.
The different strategies can be offered to students as alternative rather
than complementary strategies to help decontextualize information.
Moreover, students would have to learn when the use of which strategy
was most likely to be successful. In a curriculum as proposed here, the
notion of the 'brake-tactic' (to stop and pay attention to an unknown
word) and the 'zoom-tactic' (to pay attention to the word itself) would
also have to be incorporated.
32
word learning strategy
References
Anglin, J.M. (1993). Vocabulary development: a morphological analysis.
Monographs of the Society for Research in Child Development, 58 (10,
Serial No. 238).
Beck, I.L. & McKeown, M.G. (1991). Conditions of vocabulary acquisition. In
R. Barr, M.L. Kamil, P.B. Mosenthal & P.D. Pearson (Eds.), Handbook
of Reading Research, Volume II (pp. 789-814). New York: Longman.
Beck, I.L., McKeown, M.G. & McCaslin, E. (1983). All contexts are not created
equal. Elementary School Journal, 83, 177-181.
Beck, I.L., Perfetti, C.A. & McKeown, M.G. (1982). Effects of long-term
vocabulary instruction on lexica' access and reading comprehension.
Journal of Educational Psychology, 74, 506-521.
Buikema, J.L. & Graves, M.F. (1993). Teaching students to use context cues to
infer word meanings. Journal of Reading, 36, 450-457.
Campbell, D.T. & Stanley, J.C. (1963). Experimental and quasi-experimental
designs for research. In: N.L. Gage (Ed.), Handbook of research on
teaching (pp. 171-246). Chicago, Rand McNally.
Carnine, D., Kameenui, E.J. & Coyle, G. (1984). Utilization of contextual
information in determining the meaning of unfamiliar words. Reading
Research Quarterly, 19, 188-204.
Daalen-Kapteijns, M.M. van & Elshout-Mohr, M. (1981). The acquisition of
word meanings as a cognitive learning process. Journal of Verbal
Learning and Verbal Behavior, 20, 386-399.
Daalen-Kapteijns, M. M. van, Schouten-Van Parreren, C. & Glopper, K. de
(1993). Het afleiden van woordbetekenissen uit de context. (The
derivation of word meanings from context). Levende Talen, 485, 589-593.
Garner, R. (1990). When children and adults do not use learning strategies:
Toward a theory of settings. Review of Educational Research, 60, 517.529.
Glopper, K. de, Daalen-Kapteijns, M. M. van & Schouten-van Parreren, C.
(1997). Vocabulary knowledge and skill in inferring word meaning from
context, Amsterdam: SCO-report 462.
Goerss, B.L. (1995). Study to train elementary students to become more sensitive
to context clues. Paper presented at the Combined Meetings of the Great
Lakes and Southeast International Reading Association, november 11-15,
1995, ED392033.
Goerss, B.L., Beck, I.L. & McKeown, M.G. (in preparation). The development
and assessment of an instructional intervention to influence the process of
deriving word meaning from context.
Graves, M.F. (1987). The roles of instruction in fostering vocabulary
development. In M.G. McKeown & M.E. Curtis (Eds.), The nature of
vocabulary acquisition (pp. 89-105). Hillsdale, N.J.: Lawrence Erlbaum
Ass.
Hattie, J., Biggs, J. & Purdie, N. (1996). Effects of learning skills interventions
on student learning: a meta-analysis. Review of Educational Research, 66,
99-136.
Hedges, L.V. & Olkin, I. (1985). Statistical methods for meta-analysis. New
33
word learning strategy
York, London: Academie Press,
Huijgen, M. & Verburg, M. (1987). Van Dale Basiswoordenboek Nederlands
[Van Dale Basic Dutch Vocabulary]. Utrecht/Antwerpen: Van Dale
Lexicografie.
Jenkins, J.R., Matlock, B. & Slocum, T.A. (1989). Two approaches to vocabulary
instruction: the teaching of individual word meanings and practice in
deriving word meaning from context. Reading Research Quarterly, 24,
215-235.
Kameenui, E.J., Carnine, D.W. & Freschi, R. (1982). Effects of text construction
and instructional procedures for teaching word meaning on
comprehension and recall. Reading Research Quarterly, 17, 367-388.
McKeown, M.G. & Curtis, M.E. (Eds.) (1987). The nature of vocabulary acquisition. Hillsdale, N.J.: Lawrence Erlbaum Ass.
McKeown, M.G., Beck, I.L. & Worthy, M.J. (1993). Grappling with text ideas:
Questioning the author. The Reading Teacher, 46, 560-566.
Nagy, W.E., Anderson, R.C. & Herman, P.A. (1987). Learning word meanings
from context during normal reading. American Educational Research
Journal, 24, 237-270.
Nagy, W.E. & Scott, J.A. (1990). Word schemas: expectations about the form
and meaning of new words. Cognition and Instruction, 7, 105-127.
Nation, I.S.P. (1990). Teaching and learning vocabulary. New York: Newbury
House.
Nist, S.L. & Olejnik, S. (1995). The role of context and dictionary definitions on
varying levels of word knowledge. Reading Research Quarterly, 30, 172193.
Overmaat, M. (1996). Schrijven en lezen met tekstschema's [Writing and reading
with text schemas]. Arnsterdam: SCO-KIJPOW Universiteit van
Amsterdam.
Palincsar, A.S. & Brown, A.L. (1989). Instruction for self-regulated reading. In
L.B. Resnick & L.E. Klopfer (Eds.), Toward the thinking curriculum:
current cognitive research (pp. 19-39). 1989 Yearbook of the Association
for Supervision and Curriculum Development.
Patberg, J.P., Graves, M.F. & Stibbe, M.A. (1984). Effects of active teaching and
practice in facilitating students' use of context clues. In J.A. Niles & L.A.
Harris (Eds.), Changing pers ectives in research in reading/language
processing and instruction (pp. 146-151). Rochester, NY: National
Reading Conference.
Rosenshine, B. & Meister, C. (1994). Reciprocal teaching: a review of the
research. Review of Educational Research, 64, 479-530.
Schwartz, R.M. & Raphael, T.E. (1985). Concept of definition: a key to improving students' vocabulary. The Reading Teacher, 198-205.
Shefelbine, J.L. (1990). Student factors related to variability in learning word
meanings from context. Journal of Reading Behavior, 22, 71-97.
Shu, H., Anderson, R.C. & Zhang, H. (1995). Incidental learning of word
meanings while reading: A Chinese and American cross-cultural study.
Reading Research Quarterly, 30, 76-95.
Siegel, S. & Castellan, Jr. N.J. (1988). Nonparametric statistics for the behavioral
sciences, 2nd edition, New York: McGraw-Hill.
34
word learning strategy
Sternberg, R.J. (1987). Most vocabulary is learned from context. In M.G.
McKeown & M.E. Curtis (Eds.), The nature of vocabulary acquisition
(pp. 89-105). Hillsdale, N.J.: Lawrence Eribaum Ass.
Sternberg, R.J. & Powell, J.S. (1983). Comprehending verbal comprehension.
American Psychologist, 38, 878-893.
Veenman, M. V. J. (in press). Metacognitieve vaardigheden tijdens rekenen en
wiskunde. In H. W. Bakker-Renes, & C. M. Fennis-Poort (Eds.),
Remedial teaching hulp bij leerproblemen [Remedial teaching help with
learning problems]. Houten: Bohn, Stafleu, Van Loghum.
35
Appendix A
word learning strwegy
The contexts used for the 10 target words in the
think-aloud sequel (translated from Dutch)
Item 1 articuleren')
You have to articulate better. If you mutter like that nobody understands
you. And speaking louder doesn't help, if you still swallow half your
words.
Item 2 ('reçu')
In most video hire shops you can ask for a receipt if you are returning a
videotape. That is useful, as sometimes a videotape gets lost. And if they
think that you have lost it, then at least you can prove that you had
returned the tape properly.
Item 3 ('dilemma')
I don't know exactly what to do. My friend and her parents have asked
me to go with them for the holidays. But my sister is getting married at
the same time. If I join my friend, I will miss the wedding. And if I stay
at home for the wedding, I can't go on holiday. It is a real dilemma!
Item 4 ('stigma')
Long ago this man robbed a bank. Since then he has the stigma of being
a criminal. He has suffered his punishment, and has never stolen anything
again. But still people point him out as a bank robber.
Item 5 ('insinueren')
What are you insinuating? Why don't you say just what you mean? If
you really think he bas stolen something, you'd better say so openly.
Then at least he can defend himself. Just spreading vague stores isn't fair
to him.
Item 6 'traditie')
In our family we have a nice tradition. Every child that is bom comes
into the 'family cradle', the cradle in which grandma herself slept as a
baby. My mother has been in that 'family cradle', and me too. And my
cousins. But when there are two babies at the same time, then we have a
problem!
Item 7 ('personalia')
When you apply for a passport, of course they have to know your
personalia. The man or woman who helps you asks when you were bom,
what your first name and family name is, where you live, and so on. That
is all fed into the computer.
Item 8 ('hand- en spandiensten')
When we have a party we can count on John to do a number of odd jobs.
He is always willing to help out, and he never refuses to roll up his
sleeves, if needed.
36
word learning strategy
Item 9 ('faciliteiten')
This sports hall has all facilities: baskets, balts and nets, but also tribunes.
toilets and a canteen. So you can easily spend a whole day in there.
Item 10 ('veto')
Many people in the United States want taxes to be raised. But the
president does not want that at all. If a proposal to raise taxes is accepted,
he will put in his veto. So for the time being taxes will remain the same.
37
Figure 1. Definition Schema
The unknown word:
general characteristic:
specific characteristic:
Definition:
Figure 20 Cloze sentence, possible Wh-questions and answers for 'anecdote'
Cloze sentence: "This is the that my father likes best"
possible Wh-questions
possible answers
What is it that my father likes best?
That story about my grandfather
It makes us laugh
About my grandfather falling
through the ice
Why does he like that best?
What is it about?
Figure 3. The definition schema for 'anecdote'
The unknown word: 'anecdote' (no synonym)
general characteristic:
specific characteristic:
a story
it is funny
definition: 'a story that is funny'
38
Table 1
Time of administration, Means and Standard Deviations for all measures for
the Experimental (E) and Control (C) group
coeffidient
alpha
Time
E (n=32)
M
C (n=32)
SD
M
SD
Ll Vocabulary
.88
ml
38.94
9.43
37.44
9.49
EFL Vocabulary
.97
ml
22.02
13.15
23.90
15.06
L1 SIWMC (A)
.71
ml
6.83
2.93
7.23
3.18
Ll SIWMC (B)
.78
m2
8.52
3.58
8.24
3.94
EFL SIWMC (A)
.88
ml
3.29
3.27
3.68
3.36
EFL SIWMC (B)
.82
m2
2.34
2.71
1.98
2.41
EFL S1WMC (C)
.89
m3
3.83
3.36
3.95
3.23
Ll Reading Comprehension
.68
m2
14.52
3.73
13.93
3.52
EFL Reading Comprehension
.72
m3
10.98
3.60
10.85
4.12
39
Table 2a
Analyses of Covariance of the scores on LI SIWMC (B) with covariates Ll
SIWMC (A) and Ll Vocabulary
SS
DF
MS
F
Significance
of F
1.365
1
1.365
.150
.700
School
41.953
3
13.984
1.533
.217
2-way interaction
Treatment x School
7.948
3
2.649
.290
.832
Source of Variation
Treatment
Table 2b
Analysis of Covariance of the scores on EFL SIWMC (B), with covariates
EFL SIWMC (A) and EFL Vocabulary
Source of Variation
SS
DF
MS
F
Significance
of F
Treatment
7.137
1
7.137
2.449
.124
School
38.130
3
12.710
4.362
.008
2-way interaction
Treatment x School
2.374
3
.791
.272
.846
Table 2c
Analysis of Covariance of the scores on EFL SIWMC (C) with covariates
EFL SIWMC (B), EFL SIWMC (A) and EFL Vocabulary
Source of Variation
SS
DF
MS
F
Significance
of F
Treatment
.134
1
.134
.039
.845
School
11.679
3
3.893
1.123
.350
2-way interaction
Treatment x School
5.915
3
1.972
.569
.638
40
Table 2d
Analysis of Covariance of the scores on L1 Reading Comprehension with
covariates L1 SIWMC (A) and L 1 Vocabulary
Source of Variation
SS
DF
MS
F
Significance
of F
Treatment
.744
1
.744
.120
.731
School
28.423
3
9.474
1.524
.220
2-way interaction
Treatment x School
20.406
3
6.802
1.094
.360
Table 2e
Analysis of Covariance of the scores on EFL Reading Comprehension with
covariates EFL SIWMC (B), EFL SIWMC (A) and EFL Vocabulary
SS
DF
MS
F
Significance
of F
Treatment
3.156
1
3.156
.532
.470
School
43.324
3
14.441
2.433
.077
2-way interaction
Treatment x School
4.846
3
1.615
.272
.845
Source of Variation
41
Table 3
Mean frequeney (and standard deviation) of use of the process
codes with ten items by the control group (C) (m=16) and the
experimental group (E) (n=16)
C
m=16
E
n.16
1.03 (1.44)
3.51 (2.85)
.438*
02 Wh-question
.00 ( .00)
.81 (1.59)
.250
03 word form
.29 ( .65)
.24 ( .51)
.000
04 checking in sentence
.45 ( .76)
1.66 (1.59)
.500*
05 checking in text
.59 (1.07)
1.99 (1.61)
.500*
06 own experiences
.26 ( .76)
1.33 (1.66)
.438*
07 elaboration
2.59 (1.91)
2.78 (2.23)
.188
08 decontextualization
9.23 ( .90)
9.45 ( .85)
.188
09 metacognition
.00 ( .00)
.63 ( .92)
.375
10 superordinate
4.44 (2.50)
4.09 (1.36)
.125
11 specification
3.83 (2.10)
3.65 (1.44)
.188
12 synonym
3.40 (2.31)
4.00 (101)
.063
13 circumscription
5.04 (2.35)
3.77 (2.65)
.250
.73 (1.54)
2.45 (1.69)
.688*
01 cloze sentence
14 rereading
Dmn
42
Table 4
Means (and standard deviations) of quality of the process codes
used with ten items by the control group (C) and the experimental group (E).
The critical value of the Kolmogorov-Smirnov test statistic
mnD mn varies with m and n. The value for which p< .10 is
marked with a plus sign.
E
C
01 cloze sentence
Dr„,„
M
m
M
n
19.05 (1.63)
7
19.08 (2.01)
13
0
19.38 (1.25)
4
02 Wh-question
.132
03 word form
10.00 (10.00)
3
13.33 (11.55)
3
.333
04 checking in sentence
19.00 (2.24)
5
14.62 (7.76)
13
.385
05 checking in text
15.33 (6.50)
5
13.85 (6.18)
13
.215
06 own experiences
17.50 (3.54)
2
14.54 (6.91)
9
.444
07 elaboration
11.31 (4.38)
14
14.62 (4.52)
14
.429'
08 decontextualization
11.30 (2.59)
16
12.47 (2.24)
16
.250
0
13.33 (5.16)
6
09 metacognition
10 superordinate
14.33 (3.11)
14
15.11 (2.49)
16
.286
11 specification
11.07 (4.17)
14
10.15 (5.30)
16
A25
12 synonym
7.17 (3.37)
14
9.42 (4.00)
13
.346
13 circumscription
9.53 (3.24)
16
9.31 (1.92),
13
.236
43
Table 5
Quality, decontextualizedness, and part of speech of the
definitions produced by the control group (C) (m=16) and
the experimental group (E) (n=16)
M
C
E
quality
25.81 (7.05)
29.03 (5.21)
.375
decontextualizedness
30.54 (5.74)
31.74 (5.82)
.188
part of speech
7.04 (2.54)
6.66 (2.83)
.188
Table 6
Number of times a definition is stated in the formats
discerned, by the control group (C) (m=16) and the experimental group (E) (n=16)
format of word mea•ning
C
E
row total
one word
40 (30%)
57 (41%)
97
one word plus specifications
50 (37%)
34 (25%)
84
other format
44 (33%)
47 (34%)
91
no'answer (item not
unknown)
(23)
(16)
(39)
column total
157
154
311
44
Table 7.1
Number of times a word meaning is stated in the formats discerned by
trained and control students with 25 paper-and-pentil items in the L 1
Contextual Word Learning Test version A, taken before the training.
Maximum column total per group is 400.
format of word
meaning in L1
Contextual Word
Learning Test
version A
trained
group
n=16
one word
287 (91%)
298 (91%)
585
one word plus
specifications
15 ( 5%)
16 ( 5%)
31
other format
15 ( 5%)
12 ( 4%)
27
column total
317
control
group
n=16
326
row total
643
Value Chi-Square .45 (DF=2) p< .80
Table 7.2
Number of times a word meaning is stated in the formats discerned by
trained and control students with 25 paper-and-pencil items in the Ll
Contextual Word Learning Test version B, taken about ten days after the
eight Ll lessons. Maximum column total per group is 400.
format Of word
meaning in L 1
Contextual Word,
Learning Test
version B
trained
group
n=16
control
group
n=16
row total
One word
300 (93%)
274 (89%)
574
20 ( 6%)
26 ( 8%)
46
3 ( 1%)
8 ( 3%)
11
one word plus
specifications
other format
column total
323
Value of Chi-Square is 3.88 (DF=2) (p< .14)
308
631
SCO-Kohnstamm Instituut
voor Onderzoek van Opvoeding en Onderwijs
van de Faculteit der Pedagogische en
Onderwijskundige Wetenschappen
Universiteit van Amsterdam
Wibautstraat 4
1091 GM Amsterdam
Telefoon 020 -525 1201