EFFECTSOF INTERACTIVEVOCABULARY

INSTRUCTIONON THE VOCABULARY

LEARNINGAND READING
COMPREHENSIONOF JUNIOR-HIGH
LEARNINGDISABLED STUDENTS
Candace S. Bos and Patricia L. Anders

Abstract. Drawing upon theory-driven vocabulary instruction and the vocabulary-

reading comprehension connection, this study compared the effectiveness of three
interactive vocabulary strategies derived from the knowledge hypothesis with defini-
tion instruction derived from the access and instrumental hypotheses. Subjects were
61 learning disabled junior-high students. Using content-area texts, students par-
ticipated in one of three interactive strategies - semantic mapping (SM), semantic
feature analysis (SFA),and semantic / syntactic feature analysis (SSFA)- or in defini-
tion instruction (DI). Learning was measured both at short and long term by
vocabulary and comprehension multiple-choice items and written recalls. Results
from the multiple-choice items suggested that students participating in the inter-
active strategies demonstrated greater comprehension and vocabulary learning than
students receiving definition instruction. Results of the written recalls indicated
qualitatively and quantitatively greater recalls at long term for students in the SFA
and SSFA conditions compared with the DI condition. Implications for research and
practice are discussed.

Much currentreading research and theory sup-
port reading as an interactiveprocess: The reader
is characterizedas an active participantwho inter-
acts with the text to construct meaning. From this
perspective, the reader uses prior knowledge as
a frameworkto negotiate the meaning of the text
(Rumelhart,1980). One of the most importantas-
pects of this priorknowledge is the reader'slabels
for experiences or vocabulary knowledge (John-
son & Pearson, 1984). Earlyfactor-analysis(e.g.,
Davis, 1944, 1972) and correlationalstudies (e.g.,
Farr, 1969) highlighted the positive relationship
between vocabularyknowledge and readingcom-
prehension. Stahl and Fairbanks(1986) conducted
a meta-analysis of studies concerned with the ef-
fects of vocabularyinstructionon comprehension
and found a significant effect size for both com-
prehension of passages containingthe vocabulary
and for global measures of comprehension.
Building upon the work of Anderson and Free-
body (1981), Mezynski (1983) suggested four
hypotheses that may explain the relationshipbe-
tween vocabulary and reading comprehension.
One hypothesis, the aptitude hypothesis (e.g.,
Hunt, 1978), purports that the relationship be-
tween vocabulary and reading comprehension is
based on the reader'sunderlyingverbalabilityand
implies that this aptitude is relatively immutable.
CANDACES. BOS, Ph.D., is Associate Professor
of Special Education, University of Arizona.
PATRICIAL. ANDERS, Ph.D., is Associate Pro-
fessor of Language, Reading, and Culture,Univer-
sity of Arizona.

Volume 13, Winter 1990 31

Sage Publications, Inc.

is collaborating with JSTOR to digitize, preserve, and extend access to
Learning Disability Quarterly ®
www.jstor.org
A second hypothesis, the instrumentalhypothesis,
assumes that knowledge of individualword mean-
ings is the majorfactorrelated to reading compre-
hension. Therefore, increasing word knowledge
by direct teaching of word meanings (cf. Becker,
1977) would directlyincreasereadingcomprehen-
sion. Accordingto the thirdhypothesis, the access
hypothesis, vocabularyknowledge is relatedto the
reader'sabilityto access word meanings and use
those meanings in text processing.This hypothesis
drawsheavilyupon notions of automaticityof word
knowledge (LaBerge & Samuels, 1974; Lesgold
& Perfetti,1978). Instructionalimplicationsof this
hypothesis would emphasize automatic decoding
and definitiongenerationthroughsystematicprac-
tice. The fourth hypothesis, the knowledge
hypothesis, assumes that vocabularytermsare sur-
face representationsof underlyingconcepts repre-
sented in the reader's knowledge structures or
schemata (Anderson & Freebody, 1981). Instruc-
tionally, the knowledge hypothesis would imply
that to teach vocabulary, one must teach the
underlyingconcepts and theirrelationshipsto one
another.
An instructionalinterventioncan be generated
by combining the instructionalimplicationsfrom
the access and instrumentalhypotheses whereas
a contrastingintervention can be generated from
the instructional implications of the knowledge
hypothesis. Specifically, an access / instrumental
intervention would focus on definitioninstruction
highlightingstudent learningof the vocabularyand
the related definitions to develop easy access,
while a knowledge-basedinterventionwould focus
on the importance of activating student back-
ground knowledge and developing conceptual
understandingand semantic relationshipsamong
the vocabulary.
The purpose of this investigation was to com-
pare the effectiveness of two types of theoretically
based vocabularyinstruction:(a) interactivevocab-
ulary instructionbased on principlesof the know-
ledge hypothesis and (b) definition instruction
based on the access/instrumental hypotheses
using junior-high students identified as learning
disabledstudyingscience text. Thispopulationwas
selected for two reasons.
First,these students have been characterizedas
not spontaneously generating vocabulary know-
ledge (Simmons & Kameenui, 1988) and not
spontaneously employing comprehension and
comprehension-monitoringstrategies(Bos & Filip,
32 Learning Disability Quarterly
1984; Wong, 1980). Second, previousvocabulary
research with learning disabled students (Bos,
Anders, Filip, & Jaffe, 1989; Pany & Jenkins,
1978; Pany, Jenkins, & Schreck, 1982) has
resulted in conflictingfindings in terms of the ef-
fects of vocabulary instruction on reading com-
prehension. Bos et al. (1989) compared semantic
feature analysis, a knowledge-based instruction,
to a dictionarymethod (Gipe, 1978-79) and found
that the interactiveinstructionresulted in greater
reading comprehension and vocabularylearning.
In contrast, Pany et al. (1982) compared instru-
mental hypothesis-basedinstructionto a no-instru-
ction control condition with learningdisabled stu-
dents. While vocabulary learning was facilitated
by the instruction, comprehension, as measured
by oral retellings and cloze tests of constructed
stories using the vocabulary, was not facilitated.
The only comprehension measure that demon-
strated instructionaleffects was one that required
studentsto answerfactualquestionscontainingthe
vocabulary words.
The current study directly compared know-
ledge-based and instrumental/ access-based vo-
cabularyinstructionon learningdisabledstudents'
reading comprehension. Specifically, this study
compared the effectiveness of three knowledge-
based interactivevocabularystrategies- semantic
mapping, semantic feature analysis, semantic /
syntactic feature analysis--with an access/
instrumental-basedinstruction(definitioninstruc-
tion) on the vocabularylearning,reading compre-
hension, and quality of written recalls of junior-
high learning disabled students learning from
science text.
METHOD
Subjects
Subjects were 61 learning disabled (LD) stu-
dents recruited from middle- and lower-middle
class junior-highschools in two school districtsin
a large, metropolitanarea in the Southwest. The
41 males and 20 females had been identified as
learning disabled and were receiving services in
either resource or self-containedsettings. Eligibility
criteria included the presence of a severe dis-
crepancybetween intellectualfunctioningand one
or more academic areas and evidence of one or
more deficitsin cognitive processing. These learn-
ing problems could not be primarilyrelated to ex-
clusionary factors. Subjects were further defined
by the researchers as having at least normal in-
 Table 1
Means and Standard Deviations for Student Characteristics (N= 61)
Characteristics
Age
Full-ScaleIQ
VerbalIQ
PerformanceIQ
ReadingAchievement
(Woodcock-Johnson)
PriorKnowledgeof Topic
(TotalPossible = 30)
PriorInterestin Topic
(TotalPossible =35)
telligence as indicated by either verbal, perfor-
mance, or full-scale IQ standard scores of at least
85 and reading identified as a remediation focus.
Subject characteristicsare presented in Table 1.
Students were randomlyassigned to one of four
intervention conditions and instructed in groups
of 6 to 12. Analyses of variance on the student-
characteristic data revealed no significant dif-
ferences among the conditions.
Materials
Assessment materials. To measure learning,
a 30-item multiple-choicetest for the experimental
passage and a similartest for the practicepassage
were developed. Test construction was based on
content analyses of the passages (Frayer,Fred-
erick, & Klausmeier, 1969), and each test con-
sisted of 15 vocabulary and 15 comprehension
items. The vocabulary items measured students'
knowledge of the context-relatedmeanings of the
vocabulary presented in the passage; the com-
prehension items, in turn, measured students'
understanding of the passage or their ability to
apply the concepts presented to novel situations.
Each item on the objectivetest offeredfive options,
Mean Standard
Deviation
13.80 .93
91.97 8.95
88.67 10.05
98.00 10.65
81.30 11.20
13.31 2.84
22.90 7.69
with the correct choices balanced across the test.
To determine students' prior knowledge as re-
lated to the experimental passage, a prior-know-
ledge assessment was constructed using the 30
items from the experimental passage test and 10
items fromthe practicepassage test (servingas dis-
tractors). A 7-item topic interest inventory ascer-
tained students' prior interest in the topic. Each
item represented a superordinate or coordinate
concept presented in the experimental passage.
Students rated each statement on a 5-point Likert
scale measuring their interestin learning about it.
Instructional materials. The instructional
reading materials consisted of a practice and an
experimental passage. Passages were selected
from text and trade books used in junior-high
science classes. The experimental passage was a
1,300-word passage on the topic of fossils with an
estimated fifth-to sixth-gradereadabilitylevel (Fry,
1977). The passage's text structure was mixed,
beginningwith a narrativeabout a young boy who
is interested in fossils and becomes a geologist.
Within the narrative,information about fossils is
provided. After the narrative introduction, the
Volume 13, Winter 1990 33
structure switches to descriptive expository. The
practice passage, on the topic of forming ice, was
of similar length and readabilityutilizinga com-
parison/ contrast text structure.
The concept-related vocabularyfrom the read-
ing passages served as the instructionalfocus for
the four interventionconditions. The concept-re-
lated vocabularywas selected using the same con-
tent analysis as for the test construction. Context-
grounded definitionswere generated and verified
by content experts. For the definition instruction
condition, the instructionalmaterialsconsisted of
a writtenlist of the vocabularyand definitions.For
the semanticmappingcondition,instructional mater-
ials consisted of a written list of the vocabulary
used by the teacher and students to generate a
semantic map (Pearson & Johnson, 1978). The
map was a visual representationof the vocabulary
and the relationshipsamong the differentvocab-
ulary.A relationshipmatrix(Anders& Bos, 1986;
Johnson & Pearson, 1984) was constructed for
use in the semantic feature analysis condition. To
prepare the matrix, vocabulary representing the
superordinate concept served as the title, with
vocabulary representing the coordinate concepts
placed along the top of the matrix, and the vo-
cabulary representing the subordinate concepts
along the side. For the semantic-syntacticfeature
analysiscondition, cloze-type sentences were writ-
ten based on the matrix.Both the relationshipma-
trix and the sentences were used during this in-
structional condition.
Procedures
Five researchers served as teachers. All were
trained in the instructionalinterventions used in
the study and randomly assigned to instructional
groups with each researcherteaching at least two
different conditions and groups.
Intervention consisted of eight 50-minute ses-
sions over a span of approximatelyseven weeks.
Firststudents were given the prior-knowledgetest
and topic interest inventory.Two weeks later,stu-
dents and their assigned researcher participated
in three 50-minute practice sessions followed ap-
proximately two weeks later by the three 50-
minute experimentalsessions. Finally,four weeks
later an additional session was held to collect the
follow-up measures.
Duringthe firstday of both the practiceand the
experimental sessions, students were introduced
to the study whereupon they completed a pre-
readingactivitydesigned around theirinstructional
34 Learning Disability Quarterly
intervention.The definitioninstruction(DI)activity
consisted of directlyteaching the definitionsof the
vocabulary terms (Engelmann & Carnine, 1982;
Pany et al., 1982). The method emphasizes oral
recitation,correctand automaticpronunciationof
each vocabulary word or phrase, and memoriza-
tion of concise context-related definitions. The
vocabulary was learned and reviewed in groups
of five with initial intensive practice followed by
spaced practice and review. The three interactive
conditions employed interactive, discussion-ori-
ented strategies(Bos & Anders, in press) designed
to assist students in activatingpriorknowledge, in-
stantiatingknowledge, and predictingand drawing
relationshipsamong the concepts. Forthe seman-
tic mapping (SM) condition, the researcher and
students constructed a hierarchical relationship
map from the vocabularylist. In the semantic fea-
ture analysis (SFA) and the semantic/syntactic
feature analysis (SSFA)conditions, the researcher
and students predicted the relationships among
concepts using the relationshipmatrix.The resear-
cher and students in the SSFA condition also
predictedthe answers for the cloze-typesentences
using their matrix as a guide.
Duringthe second day of instruction,the resear-
cher and students set purposes for readingaccord-
ing to theirinstructionalinterventionand students
then readthe passage. Afterreading,the researcher
and students met as a group in a postreading ac-
tivity and again discussed their matrix and maps
or practicedand reviewedthe vocabularyand their
definitions. During this time, students were en-
couraged to use the text to confirm definitions
and / or relationships.
On the third day of instruction, students were
given 10 minutes with their instructional aide
(definitions,map, matrix,or matrixand sentences)
to study for the posttest.They were then instructed
to complete writtenrecallsby writing"allyou know
about the topic"of the passage includingwhat they
read, what they learned during instruction, or
other informationthey knew about the topic. Stu-
dents were instructedthatthey would not be judged
for spellingor grammar,but on the content.Twenty
minutes were providedfor writing,whereupon the
multiple-choicetest was administered.To measure
long-termlearning,students again completed writ-
ten recalls and the objective test four weeks later.
Data Scoring
Each objective test item was computer scored
as corrector incorrect,and a vocabularyscore and
 a comprehension score were computed for each
student.
The written recalls were scored using a variety
of procedures. First, they were analyzed for the
vocabulary used in the recall. A list of text-related
vocabularywas generated using the listof instruc-
tional vocabulary,their correspondingdefinitions,
and the content words from the passage. Withthis
list as a template, each recall was scored for the
number of text-relatedvocabulary.Content words
in the student recalls that were not text-related
were scored as eitherstudent-relevantvocabulary
or student-irrelevantvocabulary.
Second, the conceptual units recalled were
analyzed and tallied. Conceptual units, adapted
from Frederiksen's(1975) propositional analysis,
were defined as ideas that convey meaning. For
example, in the sentence, "Fossilscan be bones
from animals who lived long ago,"three concep-
tual units are present (i.e., fossils can be bones;
bones are from animals; animals lived long ago).
Similar to the vocabulary scoring, each concep-
tual unit was scored as text-related, student-
relevant, student-irrelevant,or student-inaccurate.
A conceptual unit was judged as inaccurate if it
could be directlydisproven by the text or instruc-
tional materials.
It was hypothesized that, compared to their
peers, students who received interactive instruc-
tion would integrate more elaborated prior or
scriptalknowledge into their written recalls since
these teaching strategiesemphasizedthe activation
of prior knowledge (Anders & Bos, 1986). To in-
vestigatethis hypothesis, the student-relevantcon-
ceptual units were analyzed for quality of scriptal
knowledge generated. This qualityratingwas ob-
tained by first counting the number of student-
relevant conceptual units. To ascertain quality,
each unit was categorized as elaborate, specific,
or restrictive.This categorizationwas adapted from
Langer (1982): elaborate scriptal knowledge
represented creative extensions, linkings, analog-
ies, definitions, enriched information,or a predic-
tion regardingpropositionsin the text (e.g., "Fossils
are the blueprintsof ancient life.");specific scriptal
knowledge included attributes, describing char-
acteristics,examples, processes or actions specific
to one proposition (e.g., "Fossilsare big or small.");
and restrictivescriptalknowledge contained very
limited information (e.g., "Fossils can be very
small"). Based on both the amount and type of
scriptal knowledge integrated into the written
recall, an overall quality rating from 0 to 3 was
assigned, with 3 representing the richest use of
scriptal knowledge. Finally, a holistic rating was
generated using a 6-point scale rangingfrom 0 to
5 (Irwin& Mitchell, 1983).
Written recalls were analyzed and scored by
trainedscorers knowledgeable in discourse analy-
sis. For each score, the scorers were trained until
a .85 reliabilitylevel was reached. Afterinitialscor-
ing, a second scorerrandomlyselected and scored
15% of the scored recalls. Reliabilitiesfor the dif-
ferent scores ranged from .86 to .97. The excep-
tion to this scoring scheme was the holisticrating.
In this case, holisticscoringprocedures were used
(Cooper, 1977; Myers, 1980), includingselection
of anchored recalls,trainingof raters,and multiple
ratings of each recall.
Data Analysis
Data were analyzedto determinethe short-and
long-term learning effects of the four instructional
conditions on LD students'learningusing a 4 x(2)
mixed design (Lindquist,1953). With the four in-
structionalconditions as the between factor, time
of the test (posttest and follow-up) served as the
within factor.The score for the experimental pas-
sage items on the prior-knowledge assessment,
the score for the topic-interestinventory, and the
full-scale IQ standard score (see Table 1) served
as covariates in the analyses. These three factors
were selected due to theirtheoreticallinkingto the
dependent measures (Anderson,Reynolds, Schal-
lert, & Goetz, 1977; Osako & Anders, 1983; Tor-
gesen, 1987).
RESULTS
Covariates
The role the covariates played in each analysis
varied across the reading test and the written re-
calls. Althoughpriorknowledge served as a signifi-
cant covariate for both the vocabulary score, F
(1, 54) = 20.43, p < .001, and the comprehen-
sion score, F(1, 54) = 20.00, p < .001, on the
readingtest, priorknowledge was not a significant
covariatein any of the written-recallanalyses. Prior
interestin the topic and IQ were not significantco-
variates in any of the analyses associated with
either the reading test or the written recalls.
Reading Test
Results from the reading test were analyzed for
both the vocabularyand the comprehensionscore.
The adjusted means and standard deviations at
posttest and follow-up for both scorers are pre-
Volume 13, Winter 1990 35
sented in Table 2.
Vocabulary. For the vocabulary score, results
of the main-effecttests indicateda significanteffect
for condition, F(3, 54) = 5.37, p < .003, and
for time, F(1, 57) = 59.97, p < .001. A signifi-
cant interactionbetween condition and time also
resulted, F(3, 57) = 5.88, p < .002. Post-hoc
analyses using the Tukey method (Glass & Hop-
kins, 1984) with a pooled errorterm (Kirk,1968)
and a critical alpha level of .05 showed that at
posttest students participatingin the SFA and SM
instructionalconditions had learned more vocab-
ularythan participantsin the DI condition.In terms
of long-term learning (follow-up test), students in
the three interactiveinstructionalconditions out-
performedsubjectsreceivingdefinitioninstruction.
Neither short- nor long-term learningshowed dif-
ferences among the three interactiveinstructional
conditions.
Comprehension. Results for the comprehen-
sion items were similar,with both the main-effect
tests for condition, F(3, 54) = 6.29, p < .001,
and time, F(1, 57) = 17.66, p < .001, being sig-
nificant.Although the interactionfollowed the pat-
tern of the vocabularyscores, it was not significant,
F(3, 57) = 2.69, p < .06. At posttest, post-hoc
analyses indicated that students in the interactive
conditionsscored significantlyhigherthan students
in the definitioninstructioncondition.At follow-up,
students in the SSFA conditionscored significantly
higher than those in the DI condition. Neither at
posttestnor follow-upwere theredifferencesamong
the three interactive instructionalconditions.
Written Recalls
Data relatedto the writtenrecallswere analyzed
using the same design as for the readingtests. Ad-
justed means and standarddeviationsfor the rele-
vant vocabulary generated and the relevant con-
ceptual units are presented in Table 3; adjusted
means and standard deviations for the scriptal-
knowledge quality rating and holistic rating are
shown in Table 4.

Table 2
Adjusted Means and Standard Deviations for the Reading Test Vocabulary and
Comprehension Scores (N = 61)
Definition Semantic Semantic Semantic,
Instruction Mapping Feature Syntactic
Analysis Feature
Analysis
(n= 11) (n= 19) (n= 17) (n= 14)
Vocabulary

Posttest 8.02 10.58 10.66 9.46

(2.6) (1.5) (2.4) (1.8)

Follow-Up 5.93 8.11 7.72 9.25

(2.8) (1.9) (2.5) (1.9)

Comprehension

Posttest 7.46 9.98 10.60 10.14

(2.2) (1.6) (2.0) (2.2)

Follow-Up 6.92 8.09 8.19 9.79

(2.1) (2.6) (3.2) (1.9)

36 Learning Disability Quarterly

 Table 3
Adjusted Means and Standard Deviations for the Relevant Vocabularyand Relevant
Conceptual Units Generated on the Written Recalls (N = 61)
Definition Semantic Semantic Semantic /
Instruction Mapping Feature Syntactic
Analysis Feature
Analysis
(n = 11) (n= 19) (n = 17) (n = 14)
RelevantVocabulary

Posttest 24.82 19.77 22.97

22.25
(13.8) (8.6) (6.9) (6.4)

Follow-Up 6.25 11.35 14.30 16.83

(5.0) (5.0) (6.3) (6.2)

RelevantConceptualUnits

Posttest 17.74 19.83 15.27 17.82

(15.7) (8.6) (8.1) (7.1)

Follow-Up 2.56 9.04 9.98 13.46

(2.7) (5.5) (5.4) (6.1)

Vocabulary generated. The recalls were
analyzed in terms of the relevant vocabularygen-
erated including the text-related vocabulary and
the student-relevant vocabulary. The irrelevant
vocabulary was also scored but not analyzed be-
cause the instances of irrelevantvocabularyin the
writtenrecalls averaged less than one per written
recall.
Results of the analysis revealed significant ef-
fects for time, F(1, 57) = 108.92, p < .001, and
the interaction,F(3, 57) = 6.96, p < .001, be-
tween condition and time. Post-hoc analyses on
the adjusted means (see Table 3) indicated that
although differences were not evident among the
four instructionalconditionsimmediatelyfollowing
instruction,a month later students in the DI con-
dition generated significantlyless vocabularythan
students in either the SSFA or SFA conditions. No
statisticaldifferenceswere evident among the three
interactive conditions.
Conceptual units. Recallswere also analyzed
for the relevant, accurate conceptual units gener-
ated, includingthe concepts that were represented
in the instructionalmaterialsand text and those
that were generatedby the student. Student-irrele-
vant conceptual units and student-inaccuratecon-
ceptual units were not analyzed, since the number
of occurrencesper writtenrecallaveraged less than
one.
The main-effects tests showed a significantdif-
ference across posttest and follow-up,F(1, 57) =
63.73, p < .001, as well as a significantinteraction,
F(3, 57) = 4.61, p < .006. Like the vocabulary
generated, post-hoc analyses of the adjusted
means (see Table 3) revealed no significant dif-
ference among the four instructionalconditions at
posttest. At follow-up, the findings also paralleled
the findingsfor relevantvocabularygenerated.The
SSFA and SFA students generated significantly
more conceptual unitsthan subjectsin the DI con-
dition. No differences were evident among the
three interactive conditions.

Volume 13, Winter 1990 37

 Scriptal knowledge. Based on the interactive
model, it was predicted that the quality of scriptal
knowledge would be higher for students partici-
pating in the interactivestrategies. The quality of
scriptalknowledge in the student retellingswas as-
certained by assigning an overall quality rating
based on the amount and type of scriptalknow-
ledge generated: elaborate, specific, or restrictive.
Although a significanteffect was not evident for
condition, there was a significant effect for time,
F(1, 57) = 11.60, p < .001, and for the interac-
tion of condition and time, F(3, 57) = 3.45,
p < .05. Post-hoc analyses for the adjustedmeans
(see Table4) indicatethat no differenceswere evi-
dent among the groups on the ratings measuring
short-termlearning.However, for long-termlearn-
ing, students in all three interactiveconditionsgen-
erated a higher quality of scriptal knowledge in
their recalls than students in the DI condition. No
differences emerged among the interactive
strategies.
Holistic rating. The same pattern is evident
with regard to the traditional holistic rating (see
Table4), with a significanteffect for time, F(1, 57)
= 65.03, p < .001, and for the interaction, F
(3, 57) = 4.34, p < .005. Accordingto post-hoc
analyses, no differences were found among the
four conditions at posttest. However, at follow-up
students participatingin the SSFA and SFA con-
ditions generated significantlyhigher holistic rat-
ings than DI students. No differenceswere evident
among the interactive conditions.
DISCUSSION
Previous research with learning disabled
students has resulted in conflicting findings with
regardto the vocabulary-comprehensionconnec-
tion. Instrumental/ access-based interventions
(Pany et al., 1982) have not demonstrateda clear
effect on reading comprehension compared with
a no-instructioncontrol condition; however, inter-
active or knowledge-based interventions (Bos et

Table 4
Adjusted Means and Standard Deviations for the Quality of Scriptal Knowledge and
the Holistic Rating of the Written Recalls (N-=61)
Definition Semantic Semantic Semantic/
Instruction Mapping Feature Syntactic
Analysis Feature
Analysis
(n= 11) (n= 19) (n= 17) (a = 14)
Qualityof ScriptalKnowledge

Posttest 1.78 2.18 1.68 1.75

(1.0) (1.1) (1.2) (1.2)

Follow-Up .33 1.44 1.56 1.61

(.6) (1.0) (1.3) (1.1)

HolisticRating

Posttest 5.27 6.20 5.17 5.44

(2.9) (1.9) (2.1) (1.9)

Follow-Up 1.73 3.62 4.11 4.09

(1.5) (1.9) (2.2) (1.5)

38 Learning Disability Quarterly

al., 1989) have led to greatercomprehension than
a dictionary method of teaching vocabulary. The
currentstudy compared an instrumental/ access-
based intervention with three knowledge-based
interventions.
The results of the reading test and more specif-
ically the comprehension items indicate that stu-
dents in the interactiveinterventionsscored higher
than students engaged in definition learning.'
These findings parallelthe conclusions drawn by
Stahl and Fairbanks(1986) in their meta-analysis
of vocabularyinstruction.They concluded that the
instructionalmethods producing higher compre-
hension effects involved students in deeper pro-
cessing of definition and contextual information
and provided students more than one or two ex-
posures to the to-be-learned words.
Comparisonof the definitioninstructionand the
interactive instruction reveals clear differences
among the methods in terms of depth of process-
ing and use of definition and contextual informa-
tion. Definition instruction focused on correctly
pronouncingthe vocabularyand accuratelymem-
orizing content-related definitions of the words,
more indicative of surface processing. In this in-
structionalconditionstudentswere not encouraged
to "thinkabout7how the vocabularyrelatedto their
currentunderstandingor how the concepts might
be related one to another. Each vocabulary/
definition was taught as a separate piece of infor-
mation (except that students were told they all
related to the superordinate concept of "fossils").
Students were leftto inferthe relationshipsamong
the vocabularyand their priorknowledge. In con-
trast, the interactiveinterventions highlightedthe
semantic relationshipsamong the vocabularyand
encouraged students to "thinkabout" what they
already knew about the concept.
Differencesare also evident in terms of the type
of information highlighted. Definition instruction
emphasized relatively context-free definitions.
Even though each definitionwas content-related,
it did not provide a richcontext for the definitional
information. This finding is again consistent with
Stahl and Fairbanks'(1986) meta-analysis. These
authors noted that methods providingonly defini-
tional informationdid not produce reliableeffects
on comprehension. In contrast, the interactive
strategies highlighted contextual information.
Through discussion, students were encouraged to
activate,share, and elaborateon theirpriorknow-
ledge concerning the vocabulary,thereby provid-
ing a context within which to predict definitions
and provide definitional information.
Results from the written recalls demonstrated
a consistent pattern across the various measures,
but this patternwas not congruent with the results
of the reading test. For the written recalls, dif-
ferences among the four instructionalconditions
were not evident when the students were asked
to writeon the day followinginstruction.However,
one month later, students in the semantic feature
analysis and semantic/syntactic feature analysis
conditions performed significantlybetterthan stu-
dents in the definition instruction condition on
vocabularygenerated, conceptual unitsgenerated,
and the holistic rating. For the quality of scriptal
knowledge used in the follow-up recall, all three
interactiveconditions resulted in a higher quality
of recall compared to definition instruction.
One explanation for the lack of differences in
short-term recall between the access/instru-
mental-basedinstructionand the knowledge-based
instruction may be the nature of written recalls.
Most studies with elementary and middle-school
students have employed oral retellings (e.g.,
Adams, Carnine, & Gersten, 1982; McKeown,
Beck, Omanson, & Pople, 1985; Pany et al.,
1982), requesting students to retellwhat they just
read. One exception was a study of fifth graders
(Wixson, 1986) whose findings were similar to
those of the present investigation;differenceswere
evident among the vocabulary-instructionme-
thods for the measure employing comprehension
questions but not for the writtenrecalls. The pre-
sent study not only employed written recalls but
also gave students more latitudeto utilizea variety
of information including "informationyou have
read about, we have talked about, and other in-
formation you may know about the topic" Such
broad directions may have served as a cueing
device for students in the definition-instruction
condition, therebystimulatingrecallsthat mirrored
those provided by students in the interactivein-
structional conditions.
At follow-up, results suggest that the quality of
the recalls decreased, but that the decrease dif-
fered across the instructionalconditions. The re-
calls reflect a higher quality of long-term learning
for students in the instructional conditions that
employed a relationship matrixcompared to de-
finition instruction.The use of the matrixguided
the students and teacher to discuss, predict, and
confirmthe definitionfor each vocabularyand the
Volume 13, Winter 1990 39
relationshipof each coordinate concept with each
subordinate concept. Such systematic discussion
utilizingstudent background knowledge and text
informationmay be particularlyfruitfulin terms of
generating long-term learning.
Instructionally,this study lends support to the
notion that rich, elaboratedvocabularyinstruction
based on implicationsfrom the knowledge hypo-
thesis can facilitate reading comprehension for
learningdisabled students. Such findingshold im-
plications for practitionersand researchers alike.
From a practitioner orientation, this study and
those described in the Stahl and Fairbanks(1986)
meta-analysis and the Weisberg (1988) review
challenge the use of practicesthat focus on teach-
ing definitionsisolated from content-area context.
Learning,especially long-term learning, seems to
occur under conditions that provide adequate op-
portunitiesfor students to (a) activate and instan-
tiate prior knowledge, (b) share that knowledge
with each other, (c) make predictions concerning
the relationships among concepts, and (d) read
to confirm and justify their predictions. If new
learning is based on old learning, it is imperative
that practitionersconsider the curriculumavailable
to students in content areas. A curriculumof ideas
based on successive lessons and student experi-
ences is needed to develop sufficient schema for
negotiating meanings from content-area text. The
results of this study seem to promote the notion
of engaging students in ideas, thereby facilitating
long-term learning that becomes schema upon
which new learning can occur.
For researchers, these findings raise several
questions. First, the lack of congruence between
the comprehension and composition measures
during short-term learning poses a challenging
question. To what degree must strategy-based
writinginstruction(e.g., Englert& Raphael, 1988)
be paired with vocabulary instructionto bringthe
two processes closer together? Observationsdur-
ing instruction and teacher reports made it clear
that writtenrecalls did not constitute a frequently
used means of measuring content knowledge.
Would students'comprehension be quantitatively
and /or qualitativelydifferentif composing was a
more conventionalinstructionalexperience?A sec-
ond question focuses on the discussionsthat occur
during instruction.What are the characteristicsof
the discussion duringinteractivestrategiesthat are
distinguishablefrom discussion when definitions
are emphasized ? A third question deals with the
40 Learning Disability Quarterly
tie between research and practice. Can such in-
teractivetechniques be employed by practitioners,
and does their use of the strategies lead to similar
findings? Research-based strategies that lack ef-
ficacyfor teachersin theirclassroomsare of limited
use to students, teachers, and intervention
researchers.
In the last several years there has been an in-
creasing call for theory-driven intervention re-
search in learning disabilities(e.g., Wong, 1987).
The currentresearchprovidesevidence for interac-
tive practices that are theoretically derived. By
more closely examining theory-related practices
such as these, researchersand practitionersalike
should find cause to increase their confidence in
the outcome of instructionand their understand-
ing of the tie between research and practice.
REFERENCES
Adams,A., Carnine,D., & Gersten,R. (1982).Instruc-
tionalstrategiesforstudyingcontentareatextin the
intermediategrades.Reading ResearchQuarterly,18,
27-55.
Anders,P.L., & Bos, C.S. (1986). Semanticfeature
analysis:An interactivestrategyfor vocabularyde-
velopment and text comprehension.Journal of
Reading, 29, 610-616.
Anderson,R.C., & Freebody,P. (1981). Vocabulary
knowledge.InJ.T.Guthrie(Ed.),Comprehension
and
teaching:Researchreviews (pp. 77-117). Newark,DE:
International
ReadingAssociation.
Anderson,R.C., Reynolds,R.E., Schallert,D.L., &
Goetz,E.T.(1977). Frameworksforcomprehending
discourse. American Educational Research Journal,
14, 367-382.
Becker,W.C.(1977). Teachingreadingand language
- Whatwe have learnedfrom
to the disadvantaged
field research. Harvard Educational Review, 46,
518-543.
Bos, C.S.,& Anders,P.L.(inpress).Towardan interac-
tive model:Teachingtext-basedconceptsto learning
disabledstudents. In H.L. Swanson & B. Keogh
(Eds.), Learningdisabilities;Theoreticaland research
issues. Hillsdale,NJ: ErlbaumAssociates.
Bos, C.S., Anders,P.L.,Filip,D., & Jaffe,L.E.(1989).
The effectsof an interactiveinstructionalstrategyfor
enhancinglearningdisabledstudents'readingcom-
prehensionand content area learning.Journalof
Learning Disabilities, 22, 384-390.
Bos, C.S.,& Filip,D. (1984).Comprehension
monitor-
ingin learningdisabledandaveragestudents.Journal
of Learning Disabilities, 17, 229-233.
Cooper,C.R. (1977). Holisticevaluationof writing.In
C.R. Cooper& L. Odell (Eds.),Evaluatingwriting
 (pp.3-31). Buffalo,NY:NationalCouncilof Teachers
of English.
Davis,F.B.(1944).Fundamental factorsof comprehen-
sion in reading.Psychometrika,9, 185-197.
Davis,F.B.(1972). Psychometric researchon compre-
hensionin reading.ReadingResearchQuarterly,7,
628-678.
Engelmann,S., & Carnine,D.W.(1982). Theoryof in-
struction:Principlesand applications.New York:
Irvington.
Englert,C.S., & Raphael,T.E. (1988). Constructing
well-formedprose: Process, structure,and meta-
cognitive knowledge. ExceptionalChildren,54,
513-520.
Farr,R. (1969). Reading: Whatcan be measured?
Newark,DE: International ReadingAssociation.
Frayer,D.A., Frederick,W.C., & Klausmeier,H.J.
(1969). A schema for testingthe level of concept
mastery (WorkingPaper No. 16). Madison:The
Universityof Wisconsin,WisconsinResearchand
DevelopmentCenterfor CognitiveLearning.
Frederiksen,C.H. (1975). Representinglogical and
semanticstructuresof knowledgeacquiredfromdis-
course. CognitivePsychology,7, 371-458.
Fry, E. (1977). Fry'sreadabilitygraph:Clarifications,
validity and extensions to level 17. Journal of
Reading,21, 242-252.
Gipe,J.P.(1978-79).Investigating techniquesforteach-
ingwordmeanings.ReadingResearchQuarterly, 14,
624-644.
Glass,G.V.,& Hopkins,K.D.(1984).Statistical methods
in educationand psychology(2nded.). Englewood
Cliffs,NJ: Prentice-Hall.
Hunt,E. (1978). Mechanicsof verbalability.Psycho-
logicalReview,85, 109-130.
Irwin,P.A., & Mitchell,J.N. (1983). A procedurefor
assessingthe richnessof retellings.Journalof Read-
ing, 26, 391-396.
Johnson,D.D.,& Pearson,P.D.(1984).Teachingread-
ing vocabulary(2nded.). New York:Holt,Rinehart,
& Winston.
Kirk,R.E.(1968).Experimental design:Proceduresfor
the behavioral sciences.Belmont,CA:Brooks/ Cole.
LaBerge,D.,& Samuels,S.J. (1974).Towarda theory
of automaticinformation processingin readingcom-
prehension.CognitivePsychology,6, 293-323.
Langer,J.A. (1982). Facilitating text processing:The
elaboration of priorknowledge.InJ.A.Langer& M.T.
Smith-Burke(Eds.),Readermeetsauthor/bridging
the gap (pp. 149-162). Newark,DE: International
ReadingAssociation.
Lesgold,A.M.,& Perfetti,C.A. (1978).Interactive pro-
cesses in readingcomprehension.DiscoursePro-
cesses, 1, 323-326.
Lindquist,E.R. (1953). Design and analysisof ex-
perimentsin psychology and education. Boston:
HoughtonMifflin.
McKeown,M.G.,Beck, I.L.,Omanson,R.C.,& Pople,
M.T.(1985).Some effectsof the natureandfrequency
of vocabularyinstructionon the knowledgeand use
of words.ReadingResearchQuarterly, 20, 522-535.
Mezynski, K. (1983). Issues the
concerning acquisition
of knowledge: Effectsof vocabularytrainingon
reading comprehension.Review of Educational
Research,53, 253-279.
Myers,M.A.(1980).A procedureforwriting assessment
andholisticscoring.Urbana,IL:ERICClearinghouse,
NationalInstituteof EducationandNationalCouncil
of Teachersof English.
Osako, G.N., & Anders, P.L. (1983). The effect of
reading intereston comprehensionof expository
materialswithcontrolsfor priorknowledge.In J.A.
Niles& L.A. Harris(Eds.),Searchesfor meaningin
reading/languageprocessingandinstruction (Thirty-
second yearbook,pp. 56-60). Rochester,NY:Na-
tionalReadingConference.
Pany,D., & Jenkins,J.J. (1978).Learningwordmean-
ings:A comparisonof instructional proceduresand
effectson measuresof readingcomprehensionwith
learning disabled students. Learning Disability
Quarterly,1(2), 21-32.
Pany,D.,Jenkins,J.J.,& Schreck,J. (1982).Vocabulary
instruction:Effectson wordknowledgeand reading
comprehension.LearningDisabilityQuarterly,5,
202-215.
Pearson, P.D., & Johnson, D.D. (1978). Teaching
readingcomprehension.New York:Holt, Rinehart,
& Winston.
Rumelhart, D.E.(1980).Schemata:Thebuildingblocks
of cognition.InR.J.Spiro,B.C.Bruce,& W.EBrewer
(Eds.),Theoretical issuesin readingcomprehension
(pp. 33-58). Hillsdale,NJ: ErlbaumAssociates.
Simmons,D.C.,& Kameenui,E.J.(1988).Learning dis-
abledand normalachievers'vocabularyknowledge:
A quantitativeand qualitativeanalysis. In J.E.
Readence& R.S.Baldwin(Eds.),Dialoguesin literacy
research (Thirty-seventh yearbook,pp. 133-140).
Chicago:NationalReadingConference.
Stahl,S.A., & Fairbanks, M.M.(1986). The effectsof
vocabulary instruction:A mode-basedmeta-analysis.
Reviewof EducationalResearch,56, 72-110.
Torgesen,J.K. (1987). Thinkingabout the futureby
distinguishingbetweenissues that have resolutions
and those that do not. In S. Vaughn& C.S. Bos
(Eds.),Researchin learningdisabilities: Issues and
futuredirections(pp.55-64). San Diego:College-Hill
Press.
Weisberg,R. (1988). 1980s:A changein focusof read-
ing comprehensionresearch:A reviewof reading/
learningdisabilitiesresearchbasedon an interactive
modelof reading.LearningDisabilityQuarterly,11,
149-159.
Wixson,K.K.(1986).Vocabulary instructionandchild-
ren's comprehension of basal stories. Reading
Volume13, Winter1990 41
 Research Quarterly,21, 317-329.
Wong, (1980). Activatingthe inactivelearner:Use
BY.L.
of questions/prompts to enhance comprehension
and retention of implied informationin learning dis-
abled children. Learning Disability Quarterly,3(1),
20-37.
Wong, B.Y.L.(1987). Conceptual and methodological
issues in interventionswith learning-disabledchildren
and adolescents. In S. Vaughn & C.S. Bos (Eds.),
Research in learning disabilities:Issues and future
directions (pp. 185-196). San Diego: College-Hill
Press.
FOOTNOTES
'The lack of statistical differences between the three
interactive strategies may be related to a Type II error
due to insufficient power associated with the small
number of students in the interventionconditions. This
limitationdoes not allowfor discussionrelatedto instruc-
42 Learning Disability Quarterly
tional differencesbetween the three interactivestrategies.
This research was funded through the Interactive
Teaching Project (G008630125) from the Office of
Special Education and Rehabilitative Services, U.S.
Department of Education.
We wish to thank the special education teachers and
theirstudents at Doolen, FlowingWells, Magee, Secrist,
and UtterbackJunior High Schools for allowing us to
work in their schools. A special thanks to Adela Allen,
Judy Mitchell, Grace Duran, MargaretGallego, Elba
Reyes, and Dave Scanlon for their assistance.
Requests for reprintsshould be addressed to: Candace
Bos, Division of Special Education and Rehabilitation,
College of Education,Universityof Arizona,Tucson, AZ
85721.