Full text of DOI:10.1111/0023-8333.00150

Language Learning 51:1, March 2001, pp. 145–181
Deriving the Meaning of Unknown Words From
Multiple Contexts
Maartje van Daalen-Kapteijns, Marianne Elshout-Mohr,
and Kees de Glopper
University of Amsterdam
Students from 11 to 12 years old were invited to derive
the meaning of five unknown words, each embedded in
three contexts. We focused on the students’ proficiency in
three activities: decontextualization of the target word
meanings from the contexts (decontextualization), testing
initial ideas about the word meanings with subsequent
contexts (cumulative testing), and formulating dictionary-
like definitions (defining). We compared eight students of
high and eight students of low verbal ability. The students
were led individually through the process of deriving the
meaning of the target words, while thinking aloud. The
study showed that even young students of low verbal
ability are capable of performing meaning-derivation ac-
tivities that are in general ascribed to mature students of
high verbal ability.
Maartje van Daalen-Kapteijns and Kees de Glopper, SCO-Kohnstamm Insti-
tute; Marianne Elshout-Mohr, SCO-Kohnstamm Institute and Graduate
School of Teaching and Learning.
The authors thank the Netherlands Organization for Scientific Research
(NWO) for financing the research reported in this article. They are also
grateful for the comments given by the anonymous reviewers of Language
Learning and by the members of the special interest group Higher Cognitive
Processes (Hogcog) on earlier versions of the paper.
Correspondence concerning this article may be sent to the authors care of
SCO-Kohnstamm Institute, P. O. Box 94208, 1090 GE Amsterdam, The
Netherlands. Internet: maartje@educ.uva.nl, marianne@educ.uva.nl, and
kdeglopper@educ.uva.nl
145

1
46
Language Learning
Vol. 51, No. 1
Graves (1986) estimated that students acquire on average
between 1,000 and 5,000 words from context throughout the
course of a school year.His findings also indicated that the vocabu-
laries of students of high and low verbal ability grow at different
rates, with the result that differences in vocabulary growth in-
crease over the years. Several plausible explanations have been
given for differences in acquisition rate. First, students of high
verbal ability tend to become good readers, who encounter larger
amounts of texts and more new words than their classmates who
are experiencing reading difficulties. For this reason, they have
more chances to increase their vocabulary (Stanovich, 1986). Sec-
ond, students of high verbal ability tend to build up a large
vocabulary, with the result that the number of unknown words in
text passages is often lower for these students than for classmates
of low verbal ability. Therefore, the actual starting point for deriv-
ing the meaning of unknown words is more favorable for children
of high verbal ability than for their counterparts of lower verbal
ability (Anderson & Freebody,1981;Carver,1994). Third, students
of high verbal ability are more active in deriving the meaning of
unknown words they encounter. They make better use of multiple
sources of information (Goerss, Beck, & McKeown, 1999; Miller &
Gildea, 1987), construct more abstract, paradigmatic repre-
sentations of words in addition to context-bound representations
(
Bloom & Gleitman, 1999), and are more proficient in basic pro-
cesses, such as recognizing symbols and words (Hunt, Lunneborg,
Lewis, 1975). In sum, being high verbal influences vocabulary
&
growth in various ways.The term vocabulary growth refers,in this
article, to changes in vocabulary knowledge due to incidental and
intentional learning. These changes do not only concern quantita-
tive aspects of vocabulary knowledge,such as the number of words
which are familiar, but qualitative aspects also; examples of the
latter are completeness of the knowledge of word meaning and the
automaticity of retrieval of relevant aspects of meaning (e.g.,
Merrill, Sperber, & McCauley, 1981).
The intrinsic interdependency between students’ reading
behavior, already available word knowledge, and word-derivation

van Daalen-Kapteijns et al.
147
activities on the one hand and vocabulary growth on the other
hand makes it difficult to investigate how vocabulary growth takes
place. For instance, it is difficult to estimate to what extent
“picking up”new words from context contributed to the vocabulary
growth in a particular period in a student’s life and to what extent
instruction in meaning-derivation strategies enhanced that con-
tribution. Fortunately, many research questions are tackled from
the perspectives of practitioners as well as theorists. Much can be
learned from practice-driven, ecologically valid studies into vo-
cabulary growth during free reading and into the effects of instruc-
tion programs directed at improving students’ skills in deriving
word meanings. Studies that are firmly embedded in educational
practice help draw the whole picture, the interplay of all relevant
factors.At the same time,itisworthwhiletoperform theory-driven
investigations, which attempt to describe focal parts of the whole
picture, such as the relationship between verbal ability and spe-
cific meaning-derivation skills. The present research took the
theoretical perspective. We focused on the relationship between
verbal ability and three meaning-derivation activities, i.e., “de-
contextualization,” “cumulative testing,” and “defining.” In the
next section, we explain what these three activities have in com-
mon and why they are focal in our study.In the sections that follow,
we describe the experiment, the results, and the conclusions.
Word Meaning Derivation Activities
Deriving the meaning of unknown words from context occurs
during the learning of the mother tongue and plays an important
role in the learning of foreign languages as well (Huckin, Haynes,
&
Coady, 1993). Because our study concerns derivation of Dutch
words by Dutch students, we focus on word derivation in the
mother tongue. With this restriction, literature on the strategies
used by students and advocated by teachers or researchers is still
rich.
Strategies are, in general, defined as purposeful and situated
(context-related) sequences of (mental) activities. Which strategies

1
48
Language Learning
Vol. 51, No. 1
are adopted for vocabulary learning depends on interactions
among four factors: the text (e.g., number of unknown words in the
text), the reader (e.g., interest in vocabulary knowledge in general
or in specific words occurring in the text), the task (e.g., text
comprehension or word learning), and the context (e.g., language
or science class). In general, mature readers and readers of high
verbal ability use more advanced strategies than younger readers
and readers of low verbal ability. Many college students, for in-
stance, have learned to explore the meaning of unfamiliar words
by searching for the differences and similarities between these
words and other unfamiliar or partly familiar words that occur in
the same text and belong to the same category (e.g.,<connotation>
and <denotation>).This helps them to study the words intensively
and to build a coherent network of word meanings (Flippo &
Caverly, 2000). Beck, Perfetti, and McKeown (1982) advocated a
similar strategy for younger students. Because younger students
are not yet familiar with the strategy of comparing word mean-
ings, instructors grouped target words beforehand and called
similarities and differences to the students’ attention.
The present study concerns the word-derivation activities of
two groups of students from 11 to 12 years old, differing in verbal
ability. We did not investigate the strategies of the students,
however, nor their reasons to prefer one strategy above another.
Rather than leaving students free to adopt a strategy,we arranged
conditions wherein all students performed the same sequence of
activities. This enabled us to research differences and similarities
between students of high and low verbal ability who performed
the same activities.Our interest in the chosen activities was raised
when we reviewed literature on the cognitive processes involved
in vocabulary learning. First, we found that students’ strategies
were mainly composed of three types of activities, occurring in
various sequences and combinations. Second, we found that one
type of activity was only described in studies on word derivation
by mature students of high verbal ability, whereas these activities
seemed to be absent in the strategies of younger students of low
verbal ability.

van Daalen-Kapteijns et al.
149
We discerned the following three activity types (see Table 1
for an overview). The first type is text oriented; it occurs when a
person is mainly concerned with understanding the whole text
passage around the unknown word and only deals with the word’s
meaning to keep the flow of comprehension going. The second type
of activity is word oriented; it occurs when a person is mainly
concerned with detecting the contextual meaning of the unknown
word and deals with the context in order to find out what the word
means here. The third type of activity is vocabulary knowledge
oriented; it occurs when a person is mainly concerned with using
the encounter with the unknown word as an opportunity to im-
prove his or her vocabulary knowledge and deals with the new
word in order to derive knowledge, which can be integrated into
semantic memory. Our study focuses on activities that belong to the
third type. To clarify the framework, we first discuss all three types
by giving typical examples. The examples refer to the target word
<dilemma> as presented in the three text passages given in Table 2.
Text-Oriented Activities
Two text-oriented activities are substitution and checking. A
simple way to deal with an unfamiliar word in context is to
substitute for the word a familiar word or a couple of familiar
words that fit in the sentence. For instance, the word <dilemma>
in the first context in Table 2 might be replaced by the words
“problem” or “difficult problem.” Substituting of unknown words
with familiar words fosters continuity of the reading process and
minimizes the disturbance caused by unknown words in texts (see,
e.g., Harmon, 1999). That is why we call the activity text-oriented.
When a text is rich or redundant, a student might even skip
unknown words altogether, without experiencing any loss of un-
derstanding. Skipping the word <dilemma> in context 1 leaves an
unobtrusive gap and hampers understanding only marginally. In
addition to substituting or skipping, students often check the
plausibility of the reached understanding with their experiential
knowledge (Stahl, Jacobson, Davis, & Davis, 1989). For instance,

1
50
Language Learning
Vol. 51, No. 1
Table 1
Activities involved in deriving the meaning of unknown words from
context
Orientation
Activities
Text orientation
substitution with familiar word(s)
checking with experiential knowledge
using contextual clues
Word orientation
(shifting from text to word)
morphological analysis
rehearsal
Vocabulary knowledge orientation decontextualization
(shifting from contextual word
cumulative testing
meaning to paradigmatic word
meaning)
formulating dictionary-like definitions
morphological analysis (2)
rehearsal (2)
Table 2
The set of three contexts for <dilemma>
Context 1. I don’t know what to do. My friend has invited me to go on
holidays with her and her family. During the same period,
however, my sister is getting married. If I go with my friend,
I will miss the wedding, but if I stay at home to attend the
wedding, I cannot go on holidays with my friend. It is a real
dilemma.
Context 2. Yesterday in the kitchen it was a real dilemma for our dog.
There were chips on the table and a big piece of sausage on
the kitchen unit. He was pacing to and fro between the two
and did not know what to start with. He would miss out on
both chips and sausage if his owner came in.
Context 3. Mary: “What shall I do? I am asked to play with the first team
of my club tomorrow, but I would also like to watch the match
Ajax-Heracles on television . . .” Simone: “I really don’t think
that’s a dilemma. Ajax is going to win for sure, and the match
is not going to be interesting. I think your choice is easy to
make.”

van Daalen-Kapteijns et al.
151
a student who substituted “problem”for the target word <dilemma>
may take his time to check by reasoning as follows: “Yes, ‘problem’
fits the situation. It is indeed problematic when one has to choose
betweena wedding and aninvitation to goonholidays withafriend.”
Although text-oriented activities are primarily meant to
reach understanding of the text at hand, we must assume that
they also increase students’ vocabulary knowledge incidentally. In
the long run, students acquire knowledge about the domain of
meaning of words and of the sort of texts in which they occur,
especially when the same words are encountered repeatedly.
Word-Oriented Activities
In educational practice, word-oriented activities are widely
used to speed up progress in learning new words. Advocated
word-oriented activities are use of contextual clues (also called
cues), morphological analysis of words, and rehearsal.
A precondition for using word-oriented strategies during
reading or studying is that the student notice unknown words and
make a shift from a text-oriented approach to a word-oriented
approach. Noticing and making the shift should not be conceived
of as natural activities for students who encounter unknown
words. For instance, students who read the word <dilemma> in
the first context of Table 2 may feel no need at all to pay special
attention to the word. Once they have understood that the person
in the story “has a problem” or “must make a difficult choice,” they
may find it quite appropriate to just pass on to the next step of the
ongoing reading or studying task. It depends on the person’s task
and interest in words and word meanings whether he or she
decides to halt and to derive information about the word’s meaning
from contextual clues or from other sources of information, such
as the word’s morphology.
Use of contextual clues is much researched in the domain of
vocabulary learning (e.g., Beck & McKeown, 1991; Sternberg,
1
987). Once students start looking for contextual clues about the
meaning of <dilemma> in context 1 (see Table 2), various types of

1
52
Language Learning
Vol. 51, No. 1
clues are used, such as local clues and more distant clues. A local
clue in context 1 is that the person in the story is “going to miss
either a wedding or a chance to go on holidays with her friend”; a
more distant clue is that the person in the story says that “she
doesn’t know what to do.” A student who has located both clues
might try to do justice to both, by reasoning as follows: “I might
say that <dilemma> means ‘problem,’ because she doesn’t know
what to do, but I also want to take into account that she has to
choose. Thus, I think that a dilemma is a special type of problem
which has to do with making a choice.” Activities involved in using
contextual clues have been the focus of many programs aimed at
improving students’ strategies to enhance their word knowledge.
Meta-analyticstudies (Fukkink &de Glopper,1998;Kuhn& Stahl,
1
998) showed that such programs are, in general, moderately
successful.
Morphological analysis of words is an activity that students
engage in when they notice familiar elements in the form or
features of unknown words (Anglin, 1993). In the case of the word
<
dilemma>, a student might, for instance, notice the prefix “di”
which often means “two” and connect it to the “two things” that
the person in the text wishes to do. In that case, the student could
elaborate on the original idea that <dilemma> refers to “the
problem that has to do with making a choice.”The reasoning might
be as follows: “di means two, so di-lemma probably refers to the
special sort of problem that occurs when a person has two contra-
dictory wishes.”
Rehearsal of derived meanings is a third word-oriented ac-
tivity. It is often advocated during periods in which students
encounter large numbers of new words (Mondria, 1996). In educa-
tional practice, rehearsal is stimulated by instructing students to
make notes of recently encountered words and word meanings in
a personal vocabulary booklet.
Unknown words that occur in context and are singled out for
closer inspection become familiar and linked to domains of mean-
ing (e.g., Goerss, Beck, & McKeown, 1999). Encountering words in

van Daalen-Kapteijns et al.
153
multiple contexts also fosters acquaintance with differing mean-
ings in specific contexts (Nagy, 1988).
Vocabulary-Knowledge-Oriented Activities
Both text-oriented and word-oriented activities contribute to
students’ vocabulary knowledge, but are not labeled “vocabulary-
knowledge-oriented”in Table 1.This term has been kept in reserve
for activities that are specifically related to building paradigmatic
word knowledge. Word knowledge is called paradigmatic when
words (lexical items) not only are related to the contexts in which
they are used and to experientially based concepts, but become
organized within domains of related words (Bloom & Gleitman,
1
<
999; Nelson, 1983). For instance, the representation of the word
mother> in relation to other words for family relations, such as
father, aunt, uncle, and so on, is called paradigmatic. This repre-
sentation occurs at a later developmental stage than repre-
sentation of the word <mother> in relation to a domain of meaning
(
such as the domains of “being near” and “taking care”) and
experiential concepts such as “warmth.” Paradigmatic repre-
sentations help build conceptual networks and vocabulary knowl-
edge that is well structured, i.e., effective, coherent, and
parsimonious (Anderson & Nagy, 1992; Graves, 1987). Three ac-
tivities that facilitate shifting from a word-oriented approach to a
vocabulary-knowledge-oriented approach are decontextualiza-
tion, cumulative testing, and defining. We shall discuss these
activities in some detail, because they are focal in the present
study.
Decontextualization.When students encounter an unfamiliar
word such as <dilemma>, they might ask themselves,“What is the
meaning of the word <dilemma> in this passage?” We call this
question word-oriented. However, they may also ask themselves,
“How is the meaning of the word <dilemma> related to the
meaning of other words that I know for various kinds of problems
and choices?” We call this question vocabulary-knowledge-
oriented. To answer the question, students must decontextualize

1
54
Language Learning
Vol. 51, No. 1
the meaning of <dilemma>; i.e., they must formulate the aspects
of meaning in a formal, “conventional” way rather than in terms
of the context in which they happened to encounter the word. The
term conventional was introduced by Werner and Kaplan (1952)
to refer to elements of meaning that are shared by people,whereas
the term idiosyncratic was used to refer to elements of meaning
that reflect the specific contexts in which individuals encountered
the words. In our research, we do not use the terms conventional
and idiosyncratic but rather the words decontextualized and con-
textualized to refer to essentially the same difference between
paradigmatic word meanings and contextual word meanings (van
Daalen-Kapteijns & Elshout-Mohr, 1981).
Decontextualization is a matter of degree. In the first context
in which the word <dilemma> occurs (see Table 2), the word
<
dilemma> summarizes the position of the person who has to
make the choice between attending a wedding and going on
holidays with her friend. A meaning of the word <dilemma> that
is not at all decontextualized is exhibited in the utterance, “Di-
lemma means that she doesn’t know whether to attend the wed-
ding or to go on holidays with her friend.” Students who present
such a meaning attach a “sentence core concept” to the unknown
word (Goerss, Beck, & McKeown, 1999). A much more decontextu-
alized meaning of the word <dilemma> is expressed in the utter-
ance,“A dilemma is a difficult choice between two things.”Whereas
the meaning of the entire passage is absorbed in a contextualized
meaning, the decontextualized meaning is restricted to elements
of meaning, which link the unknown word to familiar words. By
decontextualization, students relate the unknown word <dilemma>
in a formal way to other lexical units, in this case the lexical unit
choice. They identify (or rather hypothesize) two additional ele-
ments of meaning, namely that the choice is difficult and that it
involves two alternatives. The word choice doesn’t specify the
degree of difficulty or the number of choices, whereas the word
<
dilemma> is assumed to be more specific. Earlier we mentioned
the strategy to group unknown words in ordertostudythemeaning
intensively (Beck, Perfetti, & McKeown, 1982; Flippo & Caverly,

van Daalen-Kapteijns et al.
155
2
000). Grouping probably elicits searching for shared meaning
elements, thereby facilitating decontextualization.
Decontextualization activities result in representation of the
word meaning in the students’ semantic memory as part of the
already available vocabulary knowledge. In the given example we
expect that students link <dilemma> to the lexical unit “choice”
and other units in the semantic network (Anderson & Nagy,1992).
For instance, the idea that <dilemma> means “a difficult choice be-
tween two things” facilitates relating <dilemma> to words that refer
to choices which are easy to make, such as the word “preference.”
Cumulative testing. Cumulative testing can be seen as a
sophisticated form of checking. Deriving a word’s meaning from
just one context is always a hazardous enterprise and may lead to
incorrect, partially correct, and incomplete ideas about the mean-
ing of the word, even when initial ideas are checked with the
student’s own experiences. Cumulative testing consists of testing
the evidence and counterevidence for derived elements of meaning
in multiple contexts. If these contexts are not available or if they
do not provide the necessary information, a final decision must be
postponed.
In the given example of the word <dilemma>, a student could
reason as follows: “I think that <dilemma> means that there is a
problem and that the problem is difficult. I further suppose that
the word <dilemma> is used when the problem has to do with
making a choice between two options; . . . or perhaps any number
of options.” The student might keep in mind the provisional status
of her ideas, until she had a chance to test them in other text
passages, such as contexts 2 and 3 in Table 2. Cumulative testing
involves testing original ideas in the contexts that follow, but it
involves generating additional ideas as well. This is expressed in
the word “cumulative.” An example of a new idea that a student
might pick up from reading contexts 2 and 3 in Table 2 is that
<
dilemma> is only used in situations in which a person has to
choose between two or more attractive options, and that there
might be another (unknown) word that refers to difficult choices
between repelling options. Accordingly, cumulative testing is an

1
56
Language Learning
Vol. 51, No. 1
ongoing process of attaching provisional aspects of meaning to a
word, of looking for evidence and counterevidence for these as-
pects, and of searching for refinement of the word meaning. The
activity leads to effective use of information contained in multiple
contexts about the meaning of unknown or partly familiar words
(
van Daalen-Kapteijns & Elshout-Mohr, 1981). Although there is
little, if any, research on the occurrence of this activity in educa-
tional settings and free reading, we assume that cumulative
testing is indispensable for the construction of well-structured
vocabulary knowledge.
Defining. A third type of vocabulary-knowledge-oriented ac-
tivity is constructing dictionary-like definitions (defining, for
short). When an unknown word is repeatedly encountered in
context, a reader may experience the need to know the formal
definition. The word can be looked up in a dictionary, but a reader
may also try to articulate a dictionary-like definition of his or her
own. For instance, after normal reading of the three contexts in
Table 2, a student might say to herself: “I think that I know what
<
dilemma> means, but can I articulate it? Can I give a dictionary-
like definition?” Expert students might come up with excellent
definitions,such as “a dilemma isa difficult choice betweenequally
attractive options” or “a dilemma is a problem related to having
to choose between equally attractive options.” Less schooled stu-
dents or students of lower verbal ability might stick to a more
simple definition, such as “a dilemma is a problem,” or to a still
partly contextualized definition-like statement, such as “a di-
lemma is a problem, like when you have to choose between attend-
ing a wedding or going on holidays.”
We are aware that several researchers have argued that the
knowledge necessary to produce definitions is not actually needed
in order to understand or use words correctly in context (Fodor,
Garrett, Walter, & Parkes, 1980; Kuhn & Stahl, 1998). Although
we agree with this, we also assume that a student could add
quality to word knowledge by complementing contextual knowl-
edge of a word meaning by paradigmatic knowledge, which relates
the word’s meaning to the meaning of other lexical units. Because

van Daalen-Kapteijns et al.
157
this is precisely the effect of defining, we consider defining a
component of word-meaning-derivation strategies.
Morphological analysis (2) and rehearsal (2). Earlier, we
categorized morphological analysis as being word-oriented, but it
can also be used in a vocabulary-knowledge-oriented way. A stu-
dent can search intentionally for word-components, such as the
prefix “di” in <dilemma>, that are shared by different words. Used
in this manner, this activity helps constructing and organizing the
semantic network, i.e., the internal representation of vocabulary
knowledge. In the same vein, rehearsal can be used in a word-ori-
ented but also in a vocabulary-knowledge-oriented way.Rehearsal
is vocabulary-knowledge-oriented when the student wants to re-
peat the meaning of a word to distinguish it from a related word
meaning. The student might, for instance, say: “So an altruist is a
person who is devoted to others rather than to himself, like an
egotist is.”Repetition of the distinctive aspects of meaning is likely
to strengthen the relevant connections in the semantic network.
The Focus of Our Experiment
In the experiment, which is presented in the next section, we
focused on the first three of the five vocabulary-knowledge-
oriented activities: decontextualization, cumulative testing, and
defining.Our aim was to fill the knowledge gap as to the capacities
of younger students and students of low verbal ability to perform
these activities.Also,we wanted to know more about how students
can be stimulated to make a deliberate shift toward vocabulary-
knowledge-oriented activities. Van Daalen-Kapteijns, Schouten-
van Parreren, & de Glopper (1997) attempted to teach young
students (from 11 to 12 years old) to switch from text-oriented to
word-oriented derivation activities and from word-oriented to
vocabulary-knowledge-oriented derivation activities. This at-
tempt was but partly successful. The post-test revealed that most
students still preferred text-oriented substitution-like activities.
Trained students did not differ from untrained students in scores
for vocabulary-knowledge-oriented activities, although their oral

1
58
Language Learning
Vol. 51, No. 1
reports reflected that the instruction had resulted in some emerg-
ing insights (see also van Daalen-Kapteijns, de Glopper, &
Schouten-van Parreren, 1997). The present study is designed to
identify vocabulary-knowledge-oriented activities that young stu-
dents are capable of, under favorable circumstances and with
proper help.
Method
The subjects in our experiment were students 11 and 12
years old. Half of them were chosen for their high verbal ability,
the other half for their low verbal ability. The experimental task
consisted of five target words presented in three text passages
each. The students’ task was to derive the correct meaning of each
target word from the three passages.We used considerate contexts
(
Konopak, 1988), which provided ample information about the
meaning of the target words,because we wanted to make sure that
even students of low verbal ability would be capable of finding at
least some clues in each context. The reason to offer three consid-
erate contexts in succession was to set the stage for cumulative
testing activities.Thethreecontexts for <dilemma>in Table2were
part of the experimental materials.
We did not expect the students to exhibit the focal activities
of their own accord, neither in a natural setting, nor in the
experimental setting. Therefore, one investigator guided all stu-
dents through the task, individually. While doing so, the investi-
gator gave no content help, nor hints about the meaning of the
words. She just evoked attempts to perform the focal activities.
The students worked while thinking aloud.
By designing the experiment in this way, we fulfilled the
conditions that we thought necessary to make the three focal
processes (decontextualization, cumulative testing and defining)
come out clearly and separately in the performance of the students.

van Daalen-Kapteijns et al.
159
Participants
Sixteen Dutch sixth graders aged 11 and 12 years old partici-
pated in this study. Developmentally, students in this age group
have made the transition to a more formal stage of cognitive
functioning in the Piagetian sense (Inhelder, Sinclair, & Bovet,
1
974). The students were selected on the basis of a test for Dutch
vocabulary knowledge (“Woordentoets Nederlands”). The test con-
sists of 60 four-choice items. Coefficient alpha was .90 in an earlier
study with 402 sixth graders (de Glopper, van Daalen-Kapteijns,
&
Schouten-van Parreren, 1997). First the test was administered
to 62 sixth graders. Then, from these 62, eight students of rela-
tively low verbal ability (with a score of 0.75–1.75 standard devia-
tions below the mean) and eight students of relatively high verbal
ability (with a score of 0.75–1.75 standard deviations above the
mean) were invited to participate in a think-aloud experiment. We
will refer to the two groups as the lower and higher verbal ability
group. The mean score for the former group was 24.25 (with a
standard deviation of 4.03), whereas the mean score for the latter
group was 52.13 (with a standard deviation of 4.85).It is important
to note thatthe setup for the actual learning task was doubleblind:
students were not aware that they represented a low or high
verbal ability level and the researchers were not informed about
the verbal ability level of individual students while they guided
students through the task and analyzed protocols and products.
Learning Materials
Target words were ten existing Dutch words with complex
meanings that are not (yet) known by most children 11 or 12 years
old. To design considerate text passages, we first constructed a
“target structure” for each word. A target structure is a schematic
representation of the meaning of the target word. It contains the
decontextualized (conventional) elements of meaning, which de-
fine the target word and link it to other words in semantic memory
(vanDaalen-Kapteijns&Elshout-Mohr,1981).Thetargetstructures

1
60
Language Learning
Vol. 51, No. 1
of <dilemma> and <altruist> are presented in Tables 3a and 3b. A
context for a target word is considerate whenit contains clear clues
about elements of the target structure. The target structures were
also used in the scoring procedures.
Because students were to receive three contexts in succession
for each target word, three pieces of text were written (in Dutch).
Each piece described a situation in which the target word is fairly
typically used and provided accessible clues about the meaning
(
i.e., the target structure) of the target word. For all target words,
one of the three contexts represented a negative instance of typical
usage of the word, and the other two contexts were positive
instances. The length of the contexts varied from 30 to 60 words.
The passages for the word <dilemma> are presented in Table 2;
those for <altruist> are presented in Table 4. The passages are
translated into English for this article.
Although ten target words were selected and ten sets of three
contexts were prepared, only five sets were to form the items for
the actual word-derivation task for each student. The five spare
ones were used for students who happened to know the meaning
of target words fully or partially.
Experimental Procedure
All 16 students of lower and higher verbal ability were asked
to perform the same task, working individually with the first
author. First they were introduced to the task. The investigator
explained that she was interested in students’ ways of figuring out
the meaning of an unknown word from pieces of text. She made it
clear that words had been selected that the students were not
supposed to know.Then the investigator read aloud the instruction
to derive the meaning of the unknown word from the contexts,
including the instruction to think aloud. The students worked
through a practice item, which acquainted them with the experi-
mental task and the role of the investigator. The students com-
pleted five items in about 30 to 45 minutes.

van Daalen-Kapteijns et al.
161
Table 3a
Target structure of the word <dilemma>
Superordinate
concept
Aspect of
meaning 1
Aspect of
meaning 2
Option 1 Choice / to choose Hard / problematic
Between two things
Option 2 Problem
Having to do with choice Between two things
Table 3b
Target structure of the word <altruist>
Superordinate
concept
Aspect of
meaning 1
Aspect of
meaning 2
Option 1 Somebody
Option 2 A person
Willing to do things
Who always/often does things For others
For others
Guiding students through the experimental task. The guiding
procedure was as follows. After a short introduction, the first
context for the first target word was presented. The student was
asked to read it aloud and then try to find information about the
meaning of the unknown word from the text, all the while working
and thinking aloud. When the student was done, the investigator
concluded this part of the item by asking: “What have you found
out now about the meaning of <the target word>?” The investiga-
tor wrote down the answer to this question as result 1. Then the
second context for the same target word was presented, following
the same procedure, ending with result 2. Likewise, processing of
the third context was concluded with result 3. After writing down
result 3,the investigator said:“Now you have seen three texts with
<
the target word>; what do you think the word means?” The
student was to answer this question while still thinking aloud.
Then, the investigator concluded the item in a clear-cut way by
asking: “And how do you think the dictionary would state the

1
62
Language Learning
Vol. 51, No. 1
Table 4
The set of three contexts for <altruist>
Context 1. It is true indeed that my cousin is an altruist. Tired though
she may be, she will always offer her seat to somebody else on
the bus.
Context 2. A boy in my class is very egoistic. Recently we were treated to
candies. He immediately ran to the front of the class, because
he was afraid he would get nothing. He will never share with
others if, for instance, there are not enough cookies for every-
body. He is absolutely no altruist.
Context 3. Can you imagine going to a far and cold country to work for
the Red Cross, if you have a good job and a nice home here?
My doctor is an altruist like that. He would like to do something
for the sick people over there, so he is going to work for two
years in a small village somewhere up north.
meaning of <the target word>, or how would you put it in the
dictionary?” Although we were aware that the students had but
limited experience with dictionaries, we wanted to find out the
extent of students’ proficiency in defining. As the end result of the
item, the investigator wrote down the dictionary-like definition.
The Data
The experimental task yielded two kinds of data: thinking-
aloud protocols, transcribed from the tape-recorded experimental
sessions, and data about derived meaning aspects and dictionary-
like definitions which were written down by the investigator
during the session. An overview is given in Table 5.
Two Research Questions
The first research question is: Do students in the high and
low verbal group differ in a quantifiable way concerning three focal

van Daalen-Kapteijns et al.
163
Table 5
The data resulting from the learning task per item
Think aloud protocols
Data written down by the
experimenter
Context 1
Context 2
Context 3
Protocol part 1
Protocol part 2
Protocol part 3
Result 1
Result 2
Result 3
Definition task Concluding protocol part
Definition: End result
vocabulary-knowledge-oriented activities? Three hypotheses were
formulated:
1
. Students in the high verbal ability group show more decon-
textualization than students in the low verbal ability group.
2
. Students in the high verbal ability group show more cumu-
lative testing than students in the low verbal ability group.
3
. Students in the high verbal ability group provide better
dictionary-like definitions than students in the low verbal
ability group.
To test these hypotheses, the raw data had to be transformed into
quantifiable dependent measures.
The second research question is: Do students in the high and
low verbal group differ in the quality of their skills and emerging
skills in the three vocabulary-knowledge-oriented activities? To
characterize these skills, we analyzed the data qualitatively.
Dependent measures.To test the hypotheses,three dependent
measures were used, two process measures and one product mea-
sure. A description is given below.
1
. “Degree of decontextualization” is the degree to which the
derivations made by the student have some distance from the
situation as described in the context and show development in
the direction of more conventional aspects of meaning.

1
64
Language Learning
Vol. 51, No. 1
2
. The second measure, “degree of cumulative testing,” is the
degree to which the student takes a derivation from an earlier
context along to the next context (from the first to the second
or from the second to the third context). This becomes visible
when the student tests whether a derivation he or she made
earlier is consistent with the information in the current
context.
3
. The third measure concerns the definition for each target
word, given by the students as the end product of their
derivation activities. The “quality of the definition” for the
particular word is established by comparing the definition
with the meaning aspects represented in the target
structure.
Scoring Procedures
The thinking-aloud protocols and the written data were
prepared for scoring without any indication of the verbal ability
level of the student concerned. The scoring was conducted by two
judges who worked independently and were guided by a scoring
manual for each dependent measure. Kappa was computed to
establish inter-rater reliabilities. For the three dependent mea-
sures the values of kappa were .57, .67, and .71, respectively. The
average of the scores given by the two judges per item constituted
the final score. The actual scoring procedure is described below
and illustrated by the data for <dilemma> provided by one of the
students, whom we call Cynthia (see the Appendix).
Decontextualization. Degree of decontextualization was
judged by results 1, 2, and 3, as written down by the investigator
to conclude the processing of each separate context per target
word. Each result was credited a decontextualization score of 0
(
very context-bound), 1 (fairly context-bound), 2 (somewhat loos-
ened from the context),or 3 (fairly context-free).Whether the word
meaning was reached by substitution of a familiar word for the
unknown word or by any other derivation activity was not taken

van Daalen-Kapteijns et al.
165
into consideration. Because the students processed three contexts
for each of five target words, the score per item ranges from 0 to
9
, with a maximum total score (summed over 5 items) of 45.
Example: Cynthia concludes from processing context 1 for
<
dilemma>: “that it is a problem.” (See the Appendix, result 1.)
This result is credited a score of 3. Another example of result 1
given by another student is “that he does not know what to do, the
wedding or the holidays”; this result is context-bound and there-
fore credited a decontextualization score of zero.
Cumulative testing. For judging the degree of cumulative test-
ing, the protocols and written data for the whole item were consid-
ered. Correctness and form of the derivations made by the students
were not taken into consideration. The number of times that a
derivation, made from an earlier context, was tested in a following
context was established. Each tested derivation contributed two
points to the score,and a new derivation made from the third context
added one point, because this might have been tested if a fourth
context had been available. Theoretically there is no fixed maximum
to the cumulative testing score, but the highest score obtained for
any target word was 9 (four tested derivations plus one new deriva-
tion from the third context), which means that for all practical
purposes the maximum total score, summed over five items, was 45.
Example: The cumulative testing score for <dilemma> by
Cynthia (see the Appendix) amounted to 4. This score resulted
from the data in the following way. Cynthia orally derived two
elements from the first context.The derivation“that it isa problem
for something” was not explicitly tested during processing of
context 2 and 3 and was therefore not rewarded. The second
derivation was “you have to choose then.” This was tested when
the student, after having read context 2, said, “Yes again, he again
has to make a choice.” This tested derivation was credited with 2
points. Furthermore, the student derived “between the two” from
context 2 and gave evidence of testing this element while working
on context 3 by saying, “He again has to choose between two
things.” This added another two points to the score. (Actually the
student hereby tested both the “choice” derivation and the

1
66
Language Learning
Vol. 51, No. 1
“
between two things” derivation. Because the “choice” deriva-
tion had already been tested in context 2 and the student had
been given credit for that, this did not add to the score at this
point.) No further derivations were made from context 3, so
the cumulative testing score for this item totaled 4.
Defining. The dictionary-like definition for each target word
was judged by comparing the definition with the target structure
for the word concerned (see Table 3a for the target structure of
<
dilemma>). Per target word, each aspect of meaning was credited
a score of 2 (adequate), 1 (fairly adequate), or 0 (inadequate). The
number of elements in each target structure was arbitrarily lim-
ited to three.However,if a student added one more correct element
to the target structure, this element was credited also. The maxi-
mum score per item is 8, with a maximum total score, summed
over five items, of 40.
Example: The definition of <dilemma> given by Cynthia (see
the Appendix) is “that it has to be a choice.” This definition is
credited a score of 2.
The internal consistency of the three dependent measures,
estimated per measure by computing coefficient alpha over five
items, was .81, .89, and .67, respectively.
Results
The First Research Question
Predictions were that the higher verbal ability group would
outperform the lower verbal group on measures 1 through 3.These
predictions were tested by t test. Beforehand, homogeneity of
variance was tested for the dependent variables by computing
Cochran’s C. The p values for the computed C values were .57, .17,
and .21, respectively, indicating that the condition of homogeneity
of variance was not violated.
Table 6 presents the means and standard deviations of the
scores for the three dependent measures, as well as the t values

van Daalen-Kapteijns et al.
167
and the effect sizes. The latter are computed following Hedges’
unbiased estimator (Hedges & Olkin, 1985, p. 80). The criterion
for statistical significance was set to equal .01. All three t values
in Table 6 reach significance. The effect-sizes are large (Cohen,
1
988). The results indicate that the higher verbal ability group
gained significantly higher scores on the three dependent mea-
sures than their classmates of lower verbal ability. In other words,
students of higher verbal ability, as a group, performed the three
focal meaning-derivation activities more often and with a better
outcome than students of lower verbal ability as a group. These
results were in line with the expectations.
The Second Research Question
The second question concerned the quality of the students’
skills in decontextualization, cumulative testing, and formulating
dictionary-like definitions. To answer this question, we first com-
pared the mean scores of both groups (see Table 6) with the
maximum scores for decontextualization, cumulative testing, and
giving dictionary-like definitions (maximum scores were 45 for
decontextualization, 45 for cumulative testing, and 40 for quality
of definition). Students of the high verbal group reached 77, 67,
and 53 percent of the maximum scores, and students of the low
verbal group reached 45, 47, and 33 percent. These figures showed
that the experimental task was quite a challenge for students from
1
1 to 12 years old. Even those who had been selected for being of
high verbal ability performed less than perfectly. The definition
task proved to be the most difficult task. On the other hand, one
could say that the students performed quite satisfactorily. Even
those students who had beenselected for being of low verbal ability
gathered a considerable number of credit points for each of the
three focal activities. This warrants the conclusion that we actu-
ally succeeded in eliciting vocabulary-knowledge-oriented activi-
ties by offering favorable conditions and strict, individual
guidance. We investigated decontextualization, cumulative test-
ing, and defining activities in more detail.

1
68
Language Learning
Vol. 51, No. 1
Table 6
The effect of vocabulary test score on three dependent measures:
t-values and effect sizes for the comparison between the subgroup of
higher verbal ability (HVA, n = 8) and the subgroup of lower verbal
ability (LVA, n = 8)
HVA
LVA
M
SD
M
SD
t-value
Effect
size
1
*
Cumulative testing
Decontextualization
Quality of definition
30.00
34.50
21.38
5.76
6.61
3.89
21.00
20.13
13.00
7.21
11.53
6.41
2.76
3.06
3.16
1.31
1.45
1.49
*
*
1
Hedges unbiased estimator
*
p < .01
Decontextualization. To identify decontextualization activi-
ties, we first looked at the word meanings that the students
derived from the first context and counted the number of elements
of meaning derived (see Scoring Procedures). Students of the high
verbal group derived 2.43 elements of meaning on average from
the first context, whereas the students of the low verbal group
derived 1.25 elements of meaning on average from the same
context. To illustrate the kind of reasoning used, which typically
resulted in a large and a small number of elements, we present
fragments of the think-aloud protocols of students whom we call
Daphne, Remy, and Quentin. The students were working on the
first text passage for the target word <altruist> (see Table 4).
Daphne (high verbal) tried to go beyond the contextual
meaning of the target word by formulating its meaning in a more
general, formal manner. Her main concern seemed to be to decon-
textualize separate aspects and to formulate them in a proper
manner. Daphne reasoned as follows in reaction to context 1 for
<
altruist>:
D: “I think that that is somebody who always, . . . whatever
is the matter, . . . simply does . . . does it for someone.”

van Daalen-Kapteijns et al.
169
While the investigator is writing “somebody who always, . . .
whatever is the matter,” Daphne repeats the latter part of the
sentence in a somewhat revised form.
D: “. . . who does what is being asked for someone.”
The investigator writes this down, also.
Remy (low verbal) substituted for the target word a familiar
word that fit the context. Remy reasoned as follows in reaction to
context 1 for <altruist>:
R: “. . . eh, kind?”
Investigator: “I am not allowed to tell you whether it is right
or wrong, so it’s up to you.”
R: “OK, gets up in the bus, for somebody; kind, I think.”
Investigator writes down “kind.”
Quentin (low verbal) seemed to be at a loss with the task. He
produced a sentence-core concept, but hesitated about the appro-
priateness. Quentin reasoned as follows in reaction to context 1
for <altruist>:
Q: (saying nothing for about 20 seconds)
Investigator: “So what do you think that this passage tells
you?”
Q: “What the meaning is of the word?”
Investigator: “Yes”
Q: “That tired though she may be, she will get up for some-
body.”
Investigator: “So, when I ask you what this passage tells you
about the meaning of <altruist> what would you say?”
Q: “ehm, . . . (softly) I don’t know . . .”
These fragments showed that the task in which Daphne was
involved was different from the one that Remy put himself to,
while Quentin exhibited a helplessness that the other students
did not. Such observations do not, of course, explain why Daphne
decontextualized the word’s meaning from the context. However,
the protocols did inform us about the kind of differences that could
underlay differences in vocabulary growth (cf. Graves, 1986). In
our mind, there is no doubt, for instance, that Daphne’s approach

1
70
Language Learning
Vol. 51, No. 1
in the given example is more promising for building a large,
well-organized vocabulary than the approaches Remy and
Quentin exhibited.To substantiate this idea,we searched for signs
of decontextualization in the students’ handling of the second and
third contexts for the target words. We found that most of our
students used text- or word-oriented activities, but vocabulary-
knowledge-oriented activities as well, and that they frequently
switched from one approach to another. This was the case for
students of high and low verbal ability alike. Switches among
orientations are illustrated in the following fragments of the
protocols of Daphne and Quentin. The students are now working
on the second text passage for the target word <altruist> (see
Table 4).
Daphne (high verbal) was engaged in grasping the meaning
of <no altruist> in a context-bound manner. She reasoned as
follows in reaction to context 2 for <altruist>:
D: (saying nothing for about 10 seconds)
Investigator: “Yes, try telling me what you are doing.”
D: “Eh, that . . . eh that he is just, that he is just thinking only
of himself.”
Investigator: “So, what did you find out about the meaning of
<
altruist>?”
D: “That he is not an altruist, because he is always thinking
only of himself, because he is afraid that he is not getting any-
thing.”
Investigator writes down this statement.
Quentin (low verbal). In comparison with his reaction to
passage 1, Quentin was now quick in responding to passage 2. He
decontextualized the meaning of the target word and referred to
a general “you” instead of the “boy” or “he” that is described in the
text. However, when he had to repeat what he found, he used the
more context-bound word “he”again. Quentin reasoned as follows:
Investigator: “And what does this passage tell you about
<
altruist>?”
Q: “That altruist, that you do not only think of yourself, that
you also think of others.”

van Daalen-Kapteijns et al.
171
Investigator: “Shall I write that down? Please say it again.”
Q : “That he does not only think of himself.”
Investigator writes down this statement.
Evidently, the protocols add to the finding that the total
number of decontextualization activities is higher in students of
high verbal ability thanin students of low verbalability.They show
that the students of high verbal ability do not always go beyond
the contextual meaning of unknown or partially known words,and
that the students of low verbal ability do not totally lack the
capability to decontextualize.
Cumulative testing. To investigate cumulative testing activi-
ties, we kept track of the word meanings that individual students
derived from successive contexts. The following fragments illus-
trate cumulative testing as found in the protocols.
Marc (high verbal). Marc formulated an element of meaning
for the target word, which was then confirmed (rather than ac-
tively tested!) in the second context. In the third context, he added
a new—still to be tested—element. Marc reasoned as follows in
reaction to contexts 1, 2, and 3 for the word <altruist>:
Context 1.
M: “Well, someone who is rather kind to other people.”
Investigator: “So, what did you find out about the meaning of
<
altruist>?”
M: “someone who is kind to other people.”
Context 2.
M: “eh. . . .”
Investigator: “Can you tell me what you are doing, what you
are thinking of?”
M: “Eh . . . , actually I am thinking the same thing again.”
Investigator: “So, what did this passage tell you about the
meaning of <altruist>?”
M: “Eh . . . , altruist is someone who is kind to other people.”

1
72
Language Learning
Vol. 51, No. 1
Context 3.
M: “. . . Someone who takes care of other people?”
Investigator: “You are the one to decide. I just would like you
to say what you found out about the meaning of <altruist>.”
M. “Again, someone who is kind and takes care of other
people.”
Stephanie (low verbal) elaborated several ideas about the
meaning of <altruist> while processing the first context. Because
the elements of meaning were not fully decontextualized, cumula-
tive testing was difficult. This is shown in Stephanie’s reasoning
in reaction to contexts 1 and 3 for <altruist>:
Context 1.
S: “It means, I think, that you, that you—even when you are
tired—that you will offer your seat to someone else, that that does
not change, that you offer your seat, that you ... what is the proper
way to say it . . . that you just, when someone asks you to . . . even
when you are tired . . . that you just do it.”
Investigator: “So, when I ask you what this passage tells you
about the meaning of <altruist>, what would you say?”
S: “Eh, yes, that you just, . . . that you just do something for
somebody else . . . if you are asked to do so,even when you are tired
or when you do not feel like it, that you just do it.”
Context 3.
S: “Yes, I still think that he . . . just, because he wishes for
somebody else to fare well, that he does something for somebody
else.Not only because of what he does for himself,for he has a good
position and a nice home, but he wants to go to the other country
to help the people over there.”
Investigator: “So, when I ask you what this passage tells you
about the meaning of <altruist>, what would you say?”
S: “That you just, that you .. . for somebody else, that you give
up about everything, . . . that you give up your very home because
you want to help those people in that poor country.”
Like decontextualization, cumulative testing proved to be a
matter of degree. None of the students in our experimental group
of young students stated clear-cut hypotheses or tested ideas

van Daalen-Kapteijns et al.
173
systematically in successive text passages.Still,all of our students
seemed to be aware that they needed to preserve some unity in
the meaning of the target word in the three successive passages.
This was shown, for instance, in their choice of wording. Once a
student had chosen a word or a couple of words to characterize the
<
altruist> in the first text passage, he or she tended to stick to the
chosen phraseology. While some students took along words such
as “kind” (like Marc did) or “friendly,” others felt obliged to
stick—at least for sometime—to words like“polite”or “respectful.”
Probably the repetitive use of the word “just” and the phrase “for
somebody else” in Stephanie’s protocol must be interpreted in the
same vein. We conclude that the protocols indicate that even
students of low verbal ability strive for overlap between the
contextual word meanings or core-sentence concepts, which they
attach to a word that they encounter in multiple contexts. We view
this striving as a precursor of cumulative testing, because the
students are aware that they are not free to attach any meaning
to a target word without caring about the word’s meaning in prior
contexts.
Defining. When the students had finished working on the
third context, they had to collect and summarize what they found
out in order to conclude with a dictionary-like definition. Although
the students had limited experience with dictionaries, all of them
showed awareness of one or two of the following ideas: contextual
details should be deleted, conventional or formal terms should be
preferred, and part of speech should correspond to the target word.
A few defining activities are shown in the following two protocol
fragments.
Marc (high verbal) transformed his original description of the
meaning of <altruist> to a more formal definition.
Investigator: “What is your idea about the meaning of <al-
truist>?”
M: “Somebody who is kind for other people and who takes
good care of them.”
Investigator: “And how do you think the dictionary would put
it? How would you put it in the dictionary?”

1
74
Language Learning
Vol. 51, No. 1
M: “Eh yes . . . ehm, ‘a sympathetic person’ . . . or so; that it
is stated like that.”
Stephanie (low verbal), too, showed some awareness of what
a definition might look like, although she did not use an appropri-
ate part of speech in the following example.
Investigator: “What is your idea about the meaning of
<
altruist>?”
S: “I think that it just means that you don’t keep everything
for yourself and that you do, that you do something for someone
else, also . . . just like that man that goes to the other country to
help those people over there and that he is not only considerate
with himself but also with another person, that he is respectful.”
Investigator:“And how do you think the dictionary would put
it? How would you put it in the dictionary?”
S: “That you do not give everything to yourself, but that you
also give something to others. To take, but to give too.”
Investigator writes the answer down.
For students such as Marc, who perform some kind of cumu-
lative testing,defining was relatively easy.These students already
keep, so to say, a running summary of elements of meaning of the
target word. Other students, such as Stephanie and Cynthia (see
the Appendix), find defining more difficult. Although precursory
forms of cumulative testing facilitate keeping track of emerging
elements of meaning for some time (sustained by phraseology),
earlier elements of meaning tend to slip away in favor of new
elements of meaning.
Summary of the Main Results of the Qualitative Analysis
The purpose of the qualitative analysis was to increase our
knowledge about the skills in vocabulary-knowledge-oriented ac-
tivities of students of both high and low verbal ability in the middle
grades. The first conclusion was that the experimental conditions
elicited such activities and that the task was neither too easy for
students of high verbal ability nor too difficult for students of low
verbal ability.