Professional Documents
Culture Documents
net/publication/232538868
CITATIONS READS
40 387
2 authors, including:
Wendelyn Shore
Pacific Lutheran University
18 PUBLICATIONS 436 CITATIONS
SEE PROFILE
All content following this page was uploaded by Wendelyn Shore on 11 August 2014.
The authors investigated the relationship of partial word knowledge and instruction by asking
subjects to choose, from a pair of sentences, the correct use of a target word that they had or had
not looked up in a dictionary. When the correct sentence of the pair could be chosen on the
basis of general semantic constraints, subjects performed well even for words they had dented
were part of their language, and instruction had no effect. Studying dictionary definitions did,
however, increase the amount of specific, detailed information available in memory.
Researchers have speculated on the role of partial word Levels of word knowledge have received some empirical
knowledge in the area of reading comprehension and vocab- attention. While investigating incidental learning from con-
ulary acquisition and on the exact characteristics of known text, Jenkins, Stein, and Wysocki (1984) manipulated sub-
words (e.g., Beck, McKeown, & Omanson, 1987; Dale, 1965; jects' prior experiences with the meanings of the words to be
Loewenthal, 1971; Loewenthal & Gibbs, 1974; Trembly, learned. They speculated that many of the target words used
1966). Three general levels of word knowledge can be identi- in vocabulary acquisition studies may be frontier words rather
fied: One may know nothing at all about the word's meaning; than unknown words as has been assumed. Subjects per-
one may have complete knowledge of the word's meaning; or formed significantly better on tests of vocabulary knowledge
knowledge may fall between these two extremes. Different after context learning of preexposed words than they did after
researchers have operationalized these levels of word knowl- context learning of unknown words.
edge in different ways, especially the intermediate level of When others measured familiarity using an assessment of
familiarity where some, but not complete, knowledge is prior experience rather than a laboratory manipulation, the
assumed. results disagreed with Jenkins et al. (1984). Nagy, Herman,
Like Loewenthal (1971), we identified words about which and Anderson (1985) assessed the subjects' preexperimental
the reader has no knowledge as "unknown*' words. According knowledge of the words. Subjects then read a passage that
to Dale (1965), these are words to which the reader responds, contained the target words. Their results indicated that sub-
"I never saw the word before.* At the other extreme, words jects benefited more from context for unknown words than
that constitute the expressive vocabulary were classified as for words about which they had partial knowledge
"known" words (Dale, 1965; Loewenthal, 1971). Whether or not instruction has an impact on familiar words
In addition to these two extremes of word knowledge, there may depend on the extent to which that instruction augments
is a somewhat nebulous intermediate level that has been the information that already exists in the lexicon. Instruction
defined in a variety of ways. One of the clearest conceptuali- used by Nagy et al. (1985) may have had little impact on
zations of this level is given by Dale (1965) and again by familiar words because the context instruction duplicated the
Trembly (1966). Dale characterized this level by the state- preexperimental experiences that placed the word on the
ment, "I know there is such a word, but I don't know what it frontier. Frontier words are familiar because of prior experi-
means." Such words have been encountered in some form, ence in context, and additional training in context may not
most likely in the context of reading or listening. Trembly supply the type of information essential to make the word
labeled this type as frontier words because they exist at the part of the subject's active vocabulary. Words at different
"frontier" of the learner's knowledge. We have adopted Trem- levels may differ in a qualitative rather than a quantitative
bly's nomenclature in our discussion of this intermediate level way; thus, effective instruction may require the introduction
of word knowledge. Hence, our study focuses on three levels of information other than that normally constituting partial
of word knowledge: unknown, frontier, and known. knowledge.
This difference between levels of word knowledge may be
one of implicit versus explicit knowledge (e.g., Schacter, 1987)
This project was part of the first author's master thesis under the of the word's meaning. In fact, Loewenthal (1971) found that
direction of the second author. subjects performed surprisingly accurately on items that used
Support from the National Science Foundation Grant IST- unknown words, and she hypothesized that "it may be that
8418853 to the second author is gratefully acknowledged.
We thank Larry E. Toothaker, Richard Reardon, Joe Rodgers, and subjects have some knowledge of the word without being
Tom Dayton for their help with the manuscript, and Baron Van aware of it" (p. 227). Readers may have implicit memories
Burkleo and Jon Wright for their help with data collection. for the meanings of their frontier words and perhaps also for
Correspondence concerning this article should be addressed to their unknown words, whereas words in levels beyond the
Francis T. Durso, Department of Psychology, University of Okla- frontier are represented by more explicit memories. Further,
homa, Norman, Oklahoma 73019, or Bitnet: AM1067@UOKMVSA. it is possible that the information that makes meaning explicit
315
316 WENDELYN J. SHORE AND FRANCIS T. DURSO
Procedure
Method
Precheck. Upon entering the laboratory, subjects were asked to
Design complete the precheck. Responses were subject paced. On their first
pass through the precheck, they were told to define the words whose
This experiment was a 2 (sentence type: constraint only or detail) definitions they knew by writing a short definition or by giving a
x 2 (study: studied or not studied) x 3 (word level: unknown, frontier, synonym of the word. Subjects were told that it was acceptable to
or known) mixed factorial, with repeated measures on word level. leave a word undefined if they were not sure of its meaning.
Sentence type and study condition were manipulated between
After completing the definition phase, subjects were asked to go
subjects.
through the precheck again and attempt to use each of the remaining
words in a sentence. It was acceptable to leave a word blank if the
subject was not confident enough of his or her knowledge of its
Subjects meaning to use it in a sentence. Words for which the subject gave a
One hundred thirty-two volunteers from the introductory psychol- correct definition or wrote a sentence that used the word correctly
ogy subject pool at the University of Oklahoma served as subjects. formed the set of known words for that subject.
They were tested in pairs randomly assigned to study condition and Upon completion of the sentence phase, subjects were instructed
sentence type. All subjects were native English speakers and were to make another pass through the precheck. For any words they could
comparable on a measure of preexperimental verbal ability: Overall not define or use in a sentence, subjects were asked to place a check
mean on the vocabulary subscale of the Shipley-Hartford Intelligence by the word if it was familiar to them. They were told that this would
Test was 72%. include any word that they had heard or seen before, even if they had
PARTIAL WORD KNOWLEDGE 317
no real idea what it meant. Words that were checked, but for which Table 1
no attempt at a definition or sentence was made, formed the set of Means and Standard Deviations of Proportion Correct on the
frontier words for that subject. Sentence Decision Task
Finally, subjects made one more pass through the precheck. On
this final pass, they were told that if there was anything left on the Word level
list that they thought was a word, they should circle it. Subjects were Sentence type Unknown Frontier Known
informed that not all of the words on the precheck were valid English
words. They were instructed that if they did not circle a letter string, Constraint
it meant they thought the letter string was not a word in the English Nonstudied
language. The words left uncircled formed the set of unknown words M .70 .76 .98
for that subject. SD .24 .20 .04
Definition presentation. Following the precheck, subjects ran- Studied
domly assigned to the study condition were shown definitions of the M .71 .75 .95
60 English words used in the precheck. Each definition was presented SD .23 .22 .08
only once to approximate a reader's exposure to a definition when it
is simply looked up in a dictionary. Definitions were typed on 3 X 5 Detail
index cards; the target word was centered above its definition. Defi- Nonstudied
nitions from the Oxford English Dictionary (1971) or Webster's New M .49 .60 .93
World Dictionary, Second College Edition (Guralnik, 1982) were SD .18 .18 .06
used. Subjects were told that they would have 10 seconds to study Studied
each of the definitions. Subjects turned to the next card when they M .72 .72 .91
heard a tone on a cassette tape. Order of the 60 cards was randomized SD .27 .19 .09
across subjects.
In lieu of studying, subjects randomly assigned to the no-definition
condition completed the Group Embedded Figures Test (GEFT; .030, p < .001. Second, overall, studying the definition helped,
Oltman, Raskin, &. Witkin, 1971) as a distractor task. Standard F(l, 115) = 7.22, MSe = .030, p < .01. Third, the more
administration of the GEFT was followed.
familiar the word, the better subjects performed, F(2, 114) =
Sentence decision task. Following study of the words or comple-
tion of the distractor task, subjects were randomly assigned to either 3.77, MSe = .033, p < .05. However, these factors were
the constraint-only sentence group or the detail sentence group. They involved in three interactions: Sentence Type x Study, F{ 1,
were instructed to read each pair of sentences and to mark the 115) = 9.68, MS, = .030, p < .01; Word Level x Study, F(2,
sentence that used the underlined word appropriately. Subjects were 114) = 6.53, MSe = -033, p < .01; and the triple interaction
told that there would always be one sentence that used the word of Sentence Type x Study x Word Level, F(2, 114) = 3.57,
correctly, and that they should guess if they were not sure which MSe = .033, p < .05. The Study x Word Level interaction
sentence contained the correct usage. Subjects were given as much suggests that studying dictionary definitions produced the
time as needed to complete this sentence decision task. greatest benefit for unknown words and no benefit for known
Shipley-Hartford Intelligence Test. Upon completion of the sen- words; frontier words benefited an intermediate amount. The
tence decision task, subjects were administered the vocabulary section Study x Sentence Type interaction suggests that studying
of the Shipley-Hartford Intelligence Test (Sines, 1958). This test was dictionary definitions aids the learner only by supplying spe-
used to measure subjects' preexperimental vocabulary knowledge and
was included as a covariate in the statistical analyses. None of the
cific information.
target words were included on the Shipley-Hartford Intelligence Test. These two conclusions were also apparent in the triple
interaction. When the advantage of studying (study perform-
ance minus nonstudy performance) is considered as a function
Results of word level, it is clear that reading a dictionary definition
supplies almost no constraint information beyond what the
Frequencies of unknown, frontier, and known words were subject already has; even if the subject denied the letter string
computed for each subject. Data of 12 of the subjects were was an English word, studying provided only a 1 percentage
deleted because they did not have words in all of these point advantage (.71 vs. .70) in the constraint condition. On
categories. The proportion correct within each of the cate- the other hand, in the specific sentence condition, studying
gories was computed for each subject, and these means appear adds considerable specific information to unknown words
in Table 1. Planned comparisons indicated that performance (+.23), some to frontier words (+.12), and, of course, nothing
in all conditions was reliably above chance (.50) except when to known words (-.02).
subjects made judgments that required detailed knowledge of
unknown words they had not studied. The data were submit- Discussion
ted to a Sentence Type (2) x Study (2) x Word Level (3)
multivariate analysis of covariance (MANCOVA). A MANCOVA It appears that dictionary definitions impart specific infor-
was used to account for violations of sphericity (Winer, 1971) mation about a word rather than general constraint informa-
that may occur across the repeated measure. tion. Accuracy on the constraint sentences was not improved
The MANCOVA revealed three significant main effects. First, by presentation of the dictionary definition. In fact, subjects
distinguishing between correct and incorrect sentences in the entered the experiment with quite a bit of information about
constraint-only sentence condition was generally easier than the constraints of words, even words they denied were part of
in the detail sentence condition, F(\, 115) = 16.34, MSK = their language. In keeping with Loewenthal (1971), we found
318 WENDELYN J. SHORE AND FRANCIS T. DURSO