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

net/publication/225973881

The Simple View of Reading

Article in Reading and Writing · January 1990

DOI: 10.1007/BF00401799

CITATIONS READS

2,734 17,593

2 authors, including:

Wesley A Hoover
Southwest Educational Development Laboratory / American Institutes for Research
34 PUBLICATIONS 3,971 CITATIONS

SEE PROFILE

All content following this page was uploaded by Wesley A Hoover on 04 September 2021.

The user has requested enhancement of the downloaded file.

The Simple View of Reading

WESLEY A. HOOVER’ and PHILIP B. GOUGH*

’ Southwest Educational Development Laboratory Austin, Texas, U.S.A.

2 University of Texas, Austin, Texas, U.S.A.

ABSTRACT: A simple view of reading was outlined that consisted of two components,
decoding and linguistic comprehension, both held to be necessary for skilled reading.
Three predictions drawn from the simple view were assessed in a longitudinal sample of
English-Spanish bilingual children in first through fourth grade. The results supported each
prediction: (a) The linear combination of decoding and listening comprehension made
substantial contributions toward explaining variation in reading comprehension, but the
estimates were significantly improved by inclusion of the product of the two components;
@) the correlations between decoding and listening comprehension tended to become
negative as samples were successively restricted to less skilled readers; and (c) the pattern
of linear relationships between listening and reading comprehension for increasing levels of
decoding skill revealed constant intercept values of zero and positive slope values increas-
ing in magnitude. These results support the view that skill in reading can be simply
characterized as the product of skill in decoding and linguistic comprehension. The paper
concludes with a discussion of the implications of the simple view for the practice of
reading instruction, the definition of reading disability, and the notion of literacy.

KEYWORDS: components of reading, decoding, listening comprehension, literacy, reading

ability, reading comprehension, reading disability, reading instruction

The view of reading as a complex activity has long been a part of experi-
mental psychology. At the turn of the century, Huey (1908/1968) wrote
that to analyze reading would be to describe “very many of the most
intricate workings of the human mind” (p. 6). Four decades later, Gates
(1949) expressed a similar view, stating that reading is “a complex
organization of patterns of higher mental processes . . . [that] . . . can and
should embrace all types of thinking, evaluating, judging, imagining,
reasoning, and problem-solving” (p. 3). Holding to the same position, the
authors of a recent report commissioned by the National Academy of
Education (Anderson, Hiebert, Scott, and Wilkinson 1985) have likened
reading to “the performance of a symphony orchestra” (p. 7).
In counterpoint, a different view of reading has developed. This view
was clearly stated by Fries (1963) who noted that while reading certainly
does involve the host of higher mental processes cited by Gates, “every
one of the abilities listed may be developed and has been achieved by
persons who could not read . . . [as] they are all matters of the uses of

Reading and Writing: An Interdisciplinary Journal 2: 127-160,199O.

0 1990 Kluwer Academic Publishers. Printed in the Netherlands.
128 WESLEY A. HOOVER AND PHILIP B. GOUGH

language and are not limited to the uses of reading” (p. 118). In this
simple view, what distinguishes reading is that the reader is exercising
such abilities in response to graphic rather than acoustic signals, a feat
requiring the reader decode the graphic shapes into linguistic form. Thus,
one central claim of the simple view is that reading consists of only two
components, decoding and linguistic comprehension.
The simple view does not deny that the reading process is complex.
Linguistic comprehension is certainly a complicated process, whether
accomplished in reading or auding;’ and decoding, as evidenced by the
extreme difficulty some have in acquiring it, is also no simple matter. The
simple view simply holds that these complexities can be divided into two
parts.
Moreover, the simple view holds that these two parts are of equal
importance. The simple view does not reduce reading to decoding, but
asserts that reading necessarily involves the full set of linguistic skills, such
as parsing, bridging, and discourse building; decoding in the absence of
these skills is not reading. At the same time, the simple view holds that
decoding is also of central importance in reading, for without it, linguistic
comprehension is of no use. Thus, a second central claim of the simple
view is that both decoding and linguistic comprehension are necessary for
reading success, neither being sufficient by itself.

The Separability of Decoding and Linguistic Comprehension

There is much evidence suggesting that decoding and linguistic compre-
hension are separate components of reading skill. First, decoding and
linguistic comprehension can clearly be dissociated (cf. Gough and
Tunmer 1986). It is quite possible to find average and even superior
linguistic comprehension in the virtual absence of decoding skill, as the
phenomenon of dyslexia demonstrates. One can even meet the more
stringent requirement of double dissociation as evidenced by the superior
decoding skill of individuals with inferior linguistic comprehension in the
syndrome known as hyperlexia (Healy 1982). Nonetheless, these are
extreme cases and they do not demonstrate that decoding and linguistic
comprehension are separable in the range of the ordinary reader.
A number of investigations of normal reading and its relationships to
decoding and linguistic comprehension have recently appeared (Curtis
1980; Jackson and McClelland 1979; Palmer, MacLeod, Hunt, and
Davidson 1985; Singer and Crouse 1981; Stanovich, Cunningham, and
Feeman 1984; Stanovich, Nathan, and Vala-Rossi 1986). The general
correlational trends found in these studies can be summarized succinctly:
in the early school grades, decoding and linguistic comprehension are
unrelated; both skills correlate with reading comprehension, but that with
decoding is substantially stronger (coefficients of about 0.55 for decoding,
 SIMPLE VIEW OF READING 129

about 0.35 for linguistic comprehension). In the latter grades, the strength
of the relationship between decoding and linguistic comprehension in-
creases (with coefficients ranging from 0.30 to 0.65); and while both
remain related to reading comprehension (coefficients of about 0.45 for
decoding, about 0.65 for linguistic comprehension), the relationship with
linguistic comprehension becomes the dominant one (cf. Sticht and James
1984; but for an interesting perturbation see Stanovich et al. 1984;
Stanovich et al. 1986).
In addition to simple correlations, more sophisticated analysis tech-
niques have been used to further explore the contributions of decoding
and linguistic comprehension to reading comprehension. Employing
multiple regression, Curtis (1980) found that in her samples of second-,
third-, and fifth-grade students, only decoding (all grades) and linguistic
comprehension (Grades 3 and 5) consistently made significant, indepen-
dent contributions to reading comprehension. Curtis found that after
removal of the effects of nine other variables, decoding uniquely accounted
for from 3% to 13% of the variance in reading comprehension across the
three grade levels studied, while linguistic comprehension accounted for
from 23% to 35%.
Singer and Crouse (1981) tested relationships within a sixth-grade data
set via path analysis, finding that decoding and linguistic comprehension
(assessed as vocabulary knowledge) were both causally related to reading
comprehension after removal of the effects of nonverbal intelligence
(with standardized coefficients of 0.29 and 0.71, respectively). In a model
including both nonverbal intelligence and phonological awareness, a path
analysis of the first grade data of Stanovich et al. (1984) revealed that only
decoding and linguistic comprehension made significant independent
contributions to reading comprehension (with standardized coefficients of
0.39 and 0.26, respectively). In the same study, a series of hierarchical
multiple regression analyses on the third- and fifth-grade data sets showed
that after the removal of the effects of nonverbal intelligence, decoding
accounted for 19% and 38% (Grades 3 and 5, respectively) of the
variance in reading comprehension; linguistic comprehension accounting
for 14% and 13%, respectively.
While such studies demonstrate that both decoding and linguistic
comprehension are substantially related to reading comprehension (noting
that the pattern of relationship changes over grade levels), the analyses do
not specifically address the issue of how these two variables combine in
that relationship. Whether the relationship can be adequately expressed as
a multiplicative one was a focus of the present study.

Reading Ability and Reading Process

The simple view of reading must be distinguished from a kindred
130 WESLEY A. HOOVER AND PHILIP B. GOUGH

hypothesis, the so-called “bottom-up” model of the reading process, a

hypothesis that has been shown to be inadequate (Rumelhart 1977;
Stanovich 1980). The bottom-up conception holds that reading is a serial
process, with decoding preceding comprehension. On this view, decoding
should take place before, and thus, independently of comprehension, and
it should not be influenced by things taking place at any higher levels. Yet
word recognition can be dramatically influenced by linguistic context (e.g.,
see Stanovich and West 1983), and this falsifies the strictly bottom-up
model.
However, the apparent failure of the bottom-up hypothesis does not
invalidate the simple view of reading, for two reasons. First, while the
bottom-up model is undoubtedly wrong, it is not clear how wrong it is.
Although word recognition can be influenced by linguistic context under
certain conditions, a strong, empirically supported argument can be made
that during normal reading, the more proficient the reader, the less the
reliance on context (Gough 1983, 1984; Stanovich 1980). In short, fluent
reading may best be characterized as a bottom-up process.
Second, the fact that decoding does not necessarily precede linguistic
comprehension in terms of a description of reading process does not
imply that decoding is not separate from linguistic comprehension in
terms of a description of reading ability. Questions concerning the
components of a given process are distinct from questions concerning the
relationships of those components.

Decoding, Linguistic Comprehension, and Reading Comprehension

From the standpoint of the simple view of reading, the terms decoding,
linguistic comprehension, and reading comprehension can be defined as
follows.

Decoding. For the simple view, skilled decoding is simply efficient word
recognition: the ability to rapidly derive a representation from printed
input that allows access to the appropriate entry in the mental lexicon, and
thus, the retrieval of semantic information at the word level.
As argued elsewhere (Gough and Hillinger 1980), for beginning readers
the representational capability that must be acquired is one that is phono-
logically based, for the major task confronting beginning readers is one of
accessing the mental lexicon for known words that have never before been
seen in print. If the novice can derive appropriate phonological repre-
sentations for such novel printed inputs, then a lexicon already accessible
on the basis of phonological codes through the course of language
acquisition, can also begin to be accessed on the basis of print.
Lexical access via phonological codes may not predominate skilled
 SIMPLE VIEW OF READING 131

reading, such being augmented, as practice accumulates, by a more direct

graphemically-based system (for a review, see Henderson 1982). How-
ever, such direct access systems cannot benefit the beginning reader, for
in order to acquire a direct access system, both the printed word and its
pronunciation must be encountered together at least once, and it is
precisely because of the rarity of such provided pronunciations that
acquisition of the phonologically-based system is critical if the non-reader
is to become literate.
Whether the system responsible for generating such phonologically-
based representations in begi~ing reading is found to incorporate letter-
sound correspondence rules (e.g., Gough and Tunmer 1986) analogy (e.g.,
Glushko 1979) or some other process, is not central to the simple view.
The simple view only requires that the acquired system allow efficient (i.e.,
fast and accurate) access of the mental lexicon for proper, arbitrary
orthographic representations.
Concerning assessment, in general, an adequate measure of decoding
skill must tap this ability to access the mental lexicon for arbitrary printed
words (e.g., by assessing the ability to pronounce isolated real words).
However, for beginning readers, who must acquire a phonologically-based
system, an adequate decoding measure must assess skill in deriving
appropriate phonologically-based representations of novel letter strings
(e.g., by assessing the ability to pronounce isolated pseudowords).

Comprehension. In the simple view of reading, linguistic comprehension is

the ability to take lexical information (i.e., semantic information at the
word level) and derive sentence and discourse interpretations. Reading
comprehension involves the same ability, but one that relies on graphic-
based information arriving through the eye.
A measure of linguistic comprehension must assess the ability to
understand language (e.g., by assessing the ability to answer questions
about the contents of a listened to narrative). Similarly, a measure of
reading skill must assess the same ability, but one where the comprehen-
sion process begins with print (e.g., by assessing the ability to answer
questions about the contents of a read narrative).
Relevant to the assessment of comprehension is the debated distinction
between natural and formal language. Some have argued that this distinc-
tion reflects a difference between oral and written language (Olson 1977).
Others, while acknowledging the linguistic differences of text, argue that a
natural-formal language distinction is independent of modality (Freedman
and Calfee 1984). The debate will not be joined in this paper, but it is
important to note that if the simple view of reading is to be adequately
tested, parallel materials must be employed in the assessments of linguistic
comprehension and reading comprehension (e.g., if narrative material is
132 WESLEY A. HOOVER AND PHILIP B. GOUGH

used in assessing linguistic comprehension, then narrative, as opposed

to expository, material must also be used in assessing reading compre-
hension).

Specification of the Simple View

On the simple view, then, reading consists of only two parts, decoding and
linguistic comprehension, both necessary for reading success, neither suffi-
cient by itself. If reading (R), decoding (D), and linguistic comprehension
(L) are each thought of as variables ranging from 0 (nullity) to 1 (perfec-
tion), then the simple view of reading can be expressed as R = D X L.
That decoding and linguistic comprehension are combined multiplica-
tively captures the relationship of necessity coupled with non-sufficiency:
progress in reading requires both components be non-zero. To highlight
the difficulty involved in empirically assessing the multiplicative claim of
the simple view, Figure 1 graphically depicts the relationship between
decoding, linguistic comprehension, and reading comprehension, contrast-
ing additive and multiplicative combinations of the two components. Note
the gross similarities between the two graphs: in general, under each
account, reading skill increases monotonically with increases in skill in
either decoding, linguistic comprehension, or both. The relatively subtle
difference in this context of overwhelming similarity is that no increase is
found in reading skill under the multiplicative combination whenever
decoding or linguistic comprehension is zero.

Predictions of the Simple View

Three testable predictions can be drawn from the simple view. The first
prediction is that skill in decoding and linguistic comprehension will make
substantial contributions toward explaining variation in reading compre-
hension, but that the product of skill in the two components will signifi-
cantly improve the estimate of reading comprehension over that obtained
from the linear combination of the two components. That is, an additive
model, while found to be informative, will not prove to provide a fully
adequate account of the relationship of decoding and linguistic compre-
hension to reading comprehension, showing improvement upon inclusion
of the product of the two components.
Second, the simple view holds that poor reading skill results from one
of three conditions: (a) when decoding skill is adequate but linguistic
comprehension is weak, (b) when linguistic comprehension is adequate but
decoding skill is weak, or (c) when both decoding and linguistic compre-
hension skills are weak. Thus, a second prediction of the simple view is
that for less skilled readers, the relationship between decoding and
linguistic comprehension will be negative.
 SIMPLE VIEW OF READING

Fig. 1. Graphic depiction of the relationships between decoding (D), linguistic comprehen-
sion (L), and reading comprehension (R), where each ranges from nullity (0) to perfection
(1). The top panel displays the additive combination of the components (R = 0.5[D + L]),
the bottom panel, the multiplicative combination (R = D X L) that represents the simple
view of reading.

Third, as the multiplicative notion of the simple view holds reading

comprehension to be proportional to decoding and linguistic comprehen-
sion, the following pattern of linear relationships between linguistic com-
prehension and reading comprehension for increasing levels of decoding
skill is predicted. First, the intercept values are predicted to be zero
because regardless of the level of decoding skill, reading comprehension
will be null if linguistic comprehension is. Second, the slopes are predicted
to increase, as the rate of improvement in reading comprehension over
levels of linguistic comprehension is not constant, but is conditional upon
decoding skill, increasing in magnitude with increases in decoding skill.
134 WESLEY A. HOOVER AND PHILIP B. GOUGH

Third, the slope values are predicted to increase from a floor value of zero
because regardless of skill in linguistic comprehension, if decoding is null,
then reading comprehension will also be null.
As an example, for the student whose decoding skill is perfect, an
improvement in linguistic comprehension from nullity to perfection results
in an identical improvement in reading comprehension. However, for the
student whose decoding skill is only halfway toward perfection, the same
improvement in linguistic comprehension (from nullity to perfection)
leaves, as incomprehensible, half of the material that can be read by the
former student. In sum, for the former student, improvement in reading
comprehension is one-to-one with improvement in linguistic comprehen-
sion; for the latter, this ratio is halved.
Each of these three predictions was assessed through a secondary
analysis of a longitudinal data base. The primary study and the secondary
analyses conducted for this report are described in the next sections.

METHOD

Subjects
As this work constitutes a secondary analysis, the primary study’s sam-
pling plan and the subsample drawn for inclusion in these secondary
analyses are discussed. As further description of reading-related experi-
ences, summaries of the primary sample’s pre-reading skills at school
entry and the reading instruction programs subsequently received are also
provided.

Sampling plan. The Teaching Reading to Bilingual Children Study was

conducted by the Southwest Educational Development Laboratory
(SEDL) during the years of 1978-1985 in five sites in Texas. This natural
variation study tracked 254 English-Spanish bilingual children through the
early elementary grades. Multiple yearly assessments were made of each
student’s development in cognition, language, and reading, employing
multiple instruments within each domain, and assessing both English and
Spanish skill with respect to the latter two. Relatively extensive classroom
observations and teacher interviews were also conducted in order to
document the reading instruction received by each student.
The primary study employed a hierarchical sampling plan that focused
successively at the levels of region, school district, school, classroom, and
student. Once classrooms had been selected for participation, an assess-
ment battery was administered to each of the classrooms’ students. Based
on those assessments, within each classroom, 10 target children were
selected to provide systematic variation in degree of bilingualism, cogni-
 SIMPLE VIEW OF READING 135

tive style, and gender. Once selected, target students were tracked through
subsequent classroom assignments effected by the schools’ normal admin-
istrative procedures without influence from the research team.

The sample. Over the five years of data collection, 254 bilingual students
were tracked, 206 students from the beginning of kindergarten and 48
from the beginning of first grade. All students were followed through
second grade, 101 through third grade, and 6 1 through fourth grade.
Given certain restrictions in the primary study’s reading assessment proce-
dures (discussed later), the secondary analyses were based on data from
210 students at first grade, 206 at second grade, 86 at third grade, and 55
at fourth grade.

Pre-reading skills. Upon entering the study, each student’s pre-reading

skills were assessed employing the Stanford Foundation Skills Test (Calfee
and Associates 1978, 1980). The sample’s performance on the assess-
ments of English skills at kindergarten entry can be summarized as
follows: (a) about half the students knew the alphabet; (b) visual matching
skills were highly developed; (c) sight word recognition was minimal; (d)
phonemic segmentation training, based on the rhyming of familiar words,
was generally successful, but the transfer of this skill to novel items was
difficult for some; and (e) vocabulary knowledge was high relative to the
ability to retell listened to stories.

Reading instruction. Each of the study’s sites had developed both English
and Spanish reading programs. In four of the sites, Spanish-speaking
students with low English oral skills at school entry generally began
reading instruction exclusively in Spanish; in the remaining site, all
students initially received both English and Spanish reading instruction.
Transition to exclusive English reading instruction was a goal of each site,
and transition criteria included both the students’ English oral skills and
achievement in English and/or Spanish reading skills.
The general type of English reading instruction received varied by site
and teacher. The instructional emphasis in two of the sites was one of
“skills development,” where the components of decoding, vocabulary, and
text comprehension were given relatively equal attention during the early
phases of instruction. In two other sites, reading instruction focused
heavily on letter-sound correspondences and word attack skills in the
early stages, with increased attention given to comprehension skills as
decoding facility improved. In the final site, the reading program was
characterized by individualized instruction managed through student
contracts. The orientation was strongly “meaning-based,” with little formal
instruction in letter-sound correspondences until after the child had
gained some reading fluency. Two years into the study at this site, the
136 WESLEY A. HOOVER AND PHILIP B. GOUGH

district shifted to a basal reading program, thus abandoning their individ-

ualized approach for a more traditional small-group instructional proce-
dure. (Detailed accounts of the instruction received by these students,
based on classroom observations and teacher interviews, can be found in
Hoover, Calfee, and Mace-Matluck 1984b.)

Sample suitability. As noted by Gough and Tunmer (1986), an assessment

of the predictions of the simple view of reading requires a data base where
evidenced reading skills are not so highly developed that both the
independence of component skills and their variability are restricted. The
SEDL data base, representing varying degrees of student bilingualism,
decoding skill, and reading comprehension skill, met these criteria.

Tasks, Materials, Scoring, and Administration

The SEDL study employed the Interactive Reading Assessment System
(IRAS) (Calfee and Calfee 1979, 1981) as one set of yearly administered
reading assessments. The instrument consists of nine subtests, individually
administered, requiring from 45 to 90 minutes to complete. Each test was
given by trained personnel at the individual student’s school, audio taped,
and subsequently scored by SEDL staff based on both the tape and
written protocols completed during testing.
Given the program emphasis on English reading in the primary study’s
sites, only English literacy indices were analyzed here. In what follows, the
three subtests relevant to these analyses are discussed, providing details on
the structure of each, the materials, and the scoring procedures employed
in deriving summary indices of performance.

Synthetic word decoding. In this IRAS subtest, the student was presented
with six lists of synthetic words, the first four lists containing 6 items each
and the remaining two lists, 9 items each. All 42 items conformed to
English orthography, and lists were ordered by difficulty ranging from
simple monosyllabic CVC patterns to polysyllabic items containing blends,
digraphs, and vowel variations (e.g., from hin and pame in the lowest
ordered list to rhosmic and conspartuble in the highest ordered list).
Before being asked to read the synthetic words aloud, the student was told
that the items were not real words and had no meaning, but could be
pronounced like English words. Each student began on the easiest list and
proceeded to more difficult lists as long as responses were attempted to at
least half of the list’s items. Once this criterion was not met, performance
on subsequent lists was not assessed, assuming failure.
For the lists of synthetic words, each item was scored as follows. A
value of 3 was assigned to any item that was pronounced without error. A
value of 2 was given to those responses that were mostly correct (i.e.,
 SIMPLE VIEW OF READING 137

completely correct except for a minor letter-sound error such as a vowel

shift within vowel family, a stress variant, or pronunciation of a final e). A
value of 1 was given for responses that were partly correct (i.e., completely
correct except for a single vowel or consonant substitution or deletion). A
value of 0 was assigned for assumed failure, no response, or mispronun-
ciations beyond those tolerated in the above categories. Note that the
scoring was fairly stringent as two major errors within an item resulted in
a score of 0 (e.g., pronouncing ufiemiution as ufiematon).
Student performance over the lists was represented by a critical index
(the rationale for this measure is given in Hoover, Calfee, and Mace-
Matluck 1984a). The integer portion of the index represented the ordinal
value of the list of highest success (ranging from 0 to 6 in this task), where
success on a given list was achieved if half or more of its items were at
least minimally correct (i.e., received a value of 1 or more). To this value a
decimal was added that was the ratio of assigned points to total possible
points on the list of highest success. Thus, scores for this scale were
bounded by 0.00 and 6.99 (completely correct responses to all items at
the level of highest success were given decimal values of 0.99).
In the primary study, reliability analyses of each of the IRAS subscales
were conducted during each of the five years of data collection. Over this
period, the average reliability index (Cronbach’s coefficient alpha) for the
synthetic word decoding task was 0.95.

Reading and listening comprehension. The final IRAS subtests assessed

the students’ comprehension of texts, both for material read by the student
(reading comprehension) and for parallel material read to the student
(listening comprehension). The comprehension component consisted of
nine ordered levels, with each of the first six levels containing two well-
formed narrative passages, parallel in structure, one to be used in the
assessment of reading comprehension, the other in testing listening
comprehension. The final three levels contained only passages for the
assessment of reading comprehension under the assumption that students
who successfully read beyond level six (roughly corresponding to begin-
ning fourth-grade basal material) were only limited by their comprehen-
sion abilities. Passages across levels were ordered in difficulty based on
constituent word frequency, number of words per sentence, number of
sentences, and number of propositions expressed per sentence. Each story
conformed to the principles of story grammar, and associated with each
element (e.g., setting, initiating event, attempt, outcome) was a probe
question.
In reading comprehension, the student was presented with the appro-
priate narrative and asked to read it, aloud in the first four levels and
silently in the last five levels (these being appropriate for the grade-level
equivalents of the passage). Once finished reading, the student was asked
138 WESLEY A. HOOVER AND PHILIP B. GOUGH

to retell as much of the story as could be remembered. After completing

this free recall task, any element that was not adequately recalled was then
probed with the corresponding question. If half or more of the elements
were mentioned under either free or cued recall, then more difficult levels
were presented following the same procedures until this criterion was
failed. For students not successful in reading comprehension at the
seventh level or higher, listening comprehension was assessed using the
same procedures.
Each narrative element was assigned a single value ranging from 0 to 7
based on the quality of response under both free and cued recall: (a) a
value of 7 was given to any element recalled completely under free recall;
(b) a value of 6 was assigned to elements briefly mentioned in free recall
(i.e., some, but not all, of the propositions expressed in the element were
given), but recalled completely in cued recall; (c) a value of 5 was given to
elements not mentioned in free recall, but completely given in cued recall;
(d) a value of 4 was assigned to elements only briefly mentioned in both
free and cued recall; (e) a value of 3 was given to elements briefly
mentioned in free recall, but incorrectly recalled in cued recall; (f) a value
of 2 was assigned to elements not mentioned in free recall, but briefly
mentioned in cued recall; and (g) a value of 0 was given to elements not
mentioned under either free or cued recall.
As final summary indices of performance, critical indices were com-
puted. As before, the integer portion of the index represented the level of
highest success, where success on a given narrative was achieved if half or
more of the narrative’s elements were at least minimally correct (i.e.,
received a value of 2 or more). The decimal portion of the value was the
ratio of assigned points to total possible points at the level of highest
success. Thus, for reading comprehension, the index was bounded by 0.00
and 9.99; for listening comprehension, by 0.00 and 6.99.
For the purposes of the primary study, the reading comprehension
index was used as an estimate of listening comprehension skill for those
students not assessed in the latter because of success in reading at the
seventh or higher levels. This procedure was followed in these secondary
analyses as well. The number of such estimates and their percentage of the
corresponding grade-level sample were as follows: 1 case in Grade 1
(0.5%), 7 cases in Grade 2 (3.4%), 17 cases in Grade 3 (19.8%) and 22
cases in Grade 4 (40.0%). Given the large proportion of such cases in the
fourth-grade set, interpretations of the fourth-grade results must be made
cautiously as these procedures may underestimate the listening compre-
hension skills of these students.
As mentioned earlier, reliability analyses of both listening and reading
comprehension subscales were conducted during each of the five years of
data collection. These analyses were conducted at the level of individual
narratives, and over the five year period, the average reliability index
 SIMPLE VIEW OF READING 139

(Cronbach’s coefficient alpha) for the listening comprehension narratives

was 0.87; for the reading comprehension narratives, 0.84.

Administration procedures. Complex administration procedures linking

the IRAS subtests were developed by the SEDL research team in an effort
to reduce testing time. In certain cases these procedures allowed no testing
of reading comprehension (assuming failure) when decoding skills were
minimal (for full details, see Hoover et al. 1984a). In these secondary
analyses, all such cases were dropped (thus, the reduction in sample size
described earlier).

RESULTS AND DISCUSSION

The descriptive statistics and intercorrelations summarizing the sample’s

performance at each grade level are displayed in Table 1. At the end of
first grade, skill in decoding was relatively low with the average student
succeeding on only the simplest synthetic words presented. For listening
comprehension, average first-grade performance was at the third-level
narrative, but only at the first-level narrative for reading comprehension,
reflecting the average low level of decoding skill. Skill in each of the areas
increased over grade levels with the largest increases in performance
evident in the second grade (though the differences in sampling repre-
sented at the remaining two grade levels, which are also confounded by
site, complicate these grade-level comparisons).
The correlations between components were all positive and highly
significant. The lowest coefficients within each grade level were between
the decoding and listening comprehension indices. The component corre-
lations with their product showed the relatively stronger contribution of
the decoding index, especially in the early grades, reflecting the greater
number of zero values it contained (121 versus 19 in Grade 1, and 32
versus 2 in Grade 2, for decoding and listening comprehension, respec-
tively). Finally, for the correlations with reading comprehension, the
largest values were for the product of the decoding and listening compre-
hension indices, with the next largest values for the decoding index in the
first two grades, but for the listening comprehension index in the last two
grades. This shift in correlational pattern with grade level agrees with that
found in the earlier cited studies. However, the large proportion of
listening comprehension estimates in the fourth-grade data complicates
this interpretation.

Prediction 1
The first prediction drawn from the simple view was that skill in decoding
140 WESLEY A. HOOVER AND PHILIP B. GOUGH

Table 1. Descriptive statistics and intercorrelations of variables

Variables I 2 4

Gradel(N=210)
1. Decoding 0.42 0.95 0.84
2. Listening 0.50 0.46
3. Product 0.84
4. Reading
Mean 1.26 3.01 5.66 0.92
Standard deviation 2.06 2.16 10.79 1.45
Grade 2 (N = 206)
1. Decoding 0.59 0.95 0.80
2. Listening 0.70 0.71
3. Product 0.85
4. Reading
Mean 3.68 5.00 21.11 3.22
Standard deviation 2.40 1.92 15.47 2.11
Grade 3 (N = 86)
1. Decoding 0.54 0.89 0.15
2. Listening 0.81 0.80
3. Product 0.91
4. Reading
Mean 4.94 5.74 30.63 4.44
Standard deviation 2.15 1.98 16.86 2.47
Grade 4 (N = 5 5)
1. Decoding 0.72 0.97 0.84
2. Listening 0.79 0.87
3. Product 0.91
4. Reading
Mean 5.35 6.67 37.58 5.84
Standard deviation 2.06 1.31 15.55 2.00

For all correlation coefficients, p < 0.001.

and linguistic comprehension would make significant contributions toward

explaining variation in reading comprehension, but that the product of the
two components would improve the estimate over that obtained from their
linear combination. To assess this prediction, a hierarchical multiple
regression was run on each of the four grade-level data sets, entering the
decoding and listening comprehension indices first, then the product of
the two on the second step. The results of these analyses are summarized
in Table 2.
As seen in the top panel of the table, at each grade level a substantial
proportion of the variance in reading comprehension was accounted for
 SIMPLE VIEW OF READING 141

Table 2. Summary of regression analyses

Variable Multiple R R square F change P

change

Component Terms Followed by Product Term

Grade 1
Linear 0.849 0.721 267.70 0.000
Product 0.856 0.011 8.80 0.003
Grade 2
Linear 0.853 0.728 271.87 0.000
Product 0.865 0.020 16.16 0.000
Grade 3
Linear 0.884 0.782 148.62 0.000
Product 0.921 0.067 35.95 0.000
Grade 4
Linear 0.922 0.851 148.36 0.000
Product 0.948 0.048 24.34 0.000
Product Term Followed by Component Terms
Grade 1
Product 0.843 0.711 511.11 0.000
Linear 0.856 0.022 8.41 0.000
Grade 2
Product 0.848 0.720 523.27 0.000
Linear 0.865 0.029 11.55 0.000
Grade 3
Product 0.911 0.830 409.91 0.000
Linear 0.921 0.018 4.95 0.009
Grade 4
Product 0.909 0.826 251.29 0.000
Linear 0.948 0.073 18.50 0.000

Linear = Linear combination of the individually weighted decoding and listening compre-
hension indices.
Product = Product of the decoding and Listening comprehension indices.

by the linear combination of the decoding and listening comprehension

indices, ranging from 0.72 at Grade 1 to 0.85 at Grade 4 (though the
value at Grade 4 is inflated due to the estimation procedure discussed
earlier). However, the product of these two indices accounted for an
additional significant proportion of variance, ranging from 0.01 at Grade 1
142 WESLEY A. HOOVER AND PHILIP B. GOUGH

to 0.07 at Grade 3. These findings are consistent with the contentions of

the simple view, namely, that decoding and linguistic comprehension are
substantially related to reading comprehension, but further, that their
relationship is conditional.
While the results show that the additive (i.e., linear) model provides an
inadequate account of the data, and that such a model is improved by
inclusion of the product term, it does not show the superiority of the
multiplicative notion of the simple view. Equity would dictate that to test
this, the regression should be run in the reverse order, entering the
product of decoding and listening comprehension on the first step and the
two individual components on the second step. The results of such an
analysis for each grade-level data set appear in the bottom panel of
Table 2.
As was true in the first regression analysis, at each grade level a
substantial proportion of the variance in reading comprehension was
accounted for by the product of the decoding and listening comprehension
indices, ranging from 0.71 at Grade 1 to 0.83 at Grade 3. On the second
step, the linear components accounted for an additional significant
proportion of variance, ranging from 0.02 at Grades 1 and 3 to 0.07 at
Grade 4. These findings, symmetrical with the earlier regression results,
complicate the interpretation, suggesting that neither the additive nor
multiplicative account is adequate alone, but that components of both are
needed. These outcomes should be interpreted in the following context.
In the additive case, the regression procedure combines the two
components (decoding and listening comprehension) with optimal weights
to maximize the least squares fit to the reading comprehension data. In the
multiplicative case, the individual components can not be weighted in
the same manner: multiplication of either component by a constant is
equivalent to multiplication of the product of the components by that
constant, a manipulation that has no effect on the strength of the
relationship of the product to the criterion variable. This advantage in
weighting is what prevents a fair comparison of the multiplicative and
additive notions by contrasting the simple r of reading comprehension and
the product of the two components with either the simple Y of reading
comprehension and the sum of the two components (a procedure that
combines the components with arbitrary weights) or the multiple R
obtained from the regression on decoding and listening comprehension (a
procedure that combines the components with optimal weights).
However, in maximizing the least squares fit of the product of decoding
and listening comprehension to reading comprehension, adjustments made
to the zero points of the two component terms could significantly alter the
relationship with the criterion variable. Such adjustments to the com-
ponent terms are expressed in the following equation:

R = a, + b&a, + 0) (a3 + L)]

 SIMPLE VIEW OF READING 143

Expanding the terms shows that the equation represents the linear sum of
a constant and the individually weighted component terms and product:

R = (a, + b,a,u,) -t- (b,@) + (&a&) + (b,DL)

In short, if the multiplicative notion of the simple view is true and non-
zero adjustments to the zero points of the component terms are optimally
made, then the regression including both additive and multiplicative terms
will be superior to the regression containing only the additive or the
multiplicative terms. The difficulty, of course, is in finding the optimal
zero-points for the component terms.

Prediction 2
The second prediction drawn from the simple view was that for less
skilled readers an inverse relationship would hold between skill in
decoding and linguistic comprehension. To assess this prediction, the
correlation between decoding and listening comprehension was computed
within each grade level for subsamples of students defined by successively
reduced reading comprehension skill. That is, after computing the correla-
tion over the entire sample within a grade level, the correlation was next
computed for the same students less those successful on the highest level
passage, followed in turn by computing the correlation for all students less
those successful at either of the top two passages, continuing in this
manner until the sample included only those who were not successful on
the first passage presented. The results of these analyses are summarized
in Table 3, providing the sample size at each reduction, the correlation

Table 3. Descriptive statistics and correlations between decoding and listening compre-
hension for successive sample reductions based on decreasing reading comprehension skill

MaxR n Correlation Descriptive statistics

(mean and standard deviation)
r P D L P R

Grade 1
I 210 0.422 0.000 1.26 3.01 5.66 0.92
2.06 2.16 10.79 1.45
2 184 0.257 0.000 0.77 2.73 2.96 0.51
1.57 2.13 7.30 0.98
1 157 0.003 0.487 0.35 2.47 0.86 0.14
1.00 2.14 3.00 0.43
0 144 -0.125 0.067 0.15 2.40 0.21 0.02
0.52 2.20 0.88 0.10
144 WESLEY A. HOOVER AND PHILIP B. GOUGH

Table 3 (Continued)

Max R n Correlation Descriptive statistics

(mean and standard deviation)
r P D L P R

Grade 2
206 0.592 0.000 3.68 5.00 21.11 3.22
2.40 1.92 15.47 2.11
193 0.555 0.000 3.51 4.85 19.49 2.96
2.39 1.88 14.55 1.91
163 0.498 0.000 3.10 4.59 16.48 2.48
2.34 1.93 13.45 1.69
141 0.471 0.000 2.81 4.41 14.55 2.16
2.34 2.00 13.10 1.59
87 0.43 1 0.000 1.89 3.80 9.30 1.19
2.22 2.23 12.03 1.27
57 0.271 0.02 1 0.92 3.07 3.87 0.38
1.71 2.33 8.27 0.72
0 45 -0.087 0.285 0.25 2.52 0.46 0.02
0.83 2.20 2.11 0.05
Grade 3
8 86 0.543 0.000 4.94 5.74 30.63 4.44
2.15 1.98 16.86 2.47
6 69 0.476 0.000 4.61 5.29 26.46 3.67
2.26 1.96 16.13 2.14
5 62 0.440 0.000 4.43 5.11 24.66 3.36
2.31 1.99 15.96 2.03
3 40 0.285 0.037 3.53 4.44 17.14 2.24
2.41 2.17 14.94 1.65
2 22 0.019 0.467 2.50 3.50 8.86 1.02
2.39 2.46 12.32 1.26
0 13 -0.546 0.027 1.49 2.59 1.03 0.04
2.14 2.62 3.00 0.07
Grade 4
8 55 0.717 0.000 5.35 6.67 37.58 5.84
2.06 1.31 15.55 2.00
6 33 0.645 0.000 4.62 6.05 30.01 4.67
2.38 1.36 15.89 1.77
4 12 0.569 0.027 2.59 5.17 16.01 2.75
2.62 1.92 16.29 1.57

Max R = Maximum reading comprehension level included in the subsample

D = Decoding index
L = Listening comprehension index
P = Product of decoding and listening comprehension indices
R = Reading comprehension index
 SIMPLE VIEW OF READING 145

coefficient and its significance level, and the descriptive statistics for each
variable (decoding, listening comprehension, their product, and reading
comprehension).
First, note from the descriptive statistics tabled that within each of the
four grade levels represented, as the sample was reduced based on
decreasing reading comprehension skill, aggregate skills in each of the
other indices (decoding, listening comprehension, and their product) also
decreased. Within each grade level, the correlations were significantly
positive when computed over the entire grade-level sample, with coeffi-
cients ranging from 0.42 in Grade 1 to 0.72 in Grade 4. As the sample
was reduced, the coefficients declined in magnitude to values not signifi-
cantly different from zero, and in Grades 1 and 3, to values that were
negative (approaching significance for the former, and significant for the
latter).
It is possible that these results are artifactual. As the samples are
reduced, a few outliers off the diagonal could lead to negative correlations.
Note, however, that this argument is symmetric: if measurement error
is responsible for negative correlations as the samples are successively
restricted to the weakest readers, then it should similarly constrain the
correlational trends as the samples are successively restricted to include
only the strongest readers. Accordingly, analyses were conducted identical
to the ones above, save that sample reduction was based on increasing (as
opposed to decreasing) reading comprehension skill. The results of these
analyses are summarized in Table 4.
As expected from the simple view, and in contrast to the results
displayed in Table 3, aggregate skills in decoding, listening comprehen-
sion, and their product increased as the sample was reduced based on

Table 4. Descriptive statistics and correlations between decoding and listening compre-
hension for successive sample reductions based on increasing reading comprehension skill

MinR n Correlation Descriptive statistics

(mean and standard deviation)
r P D L P R

Grade I
0 210 0.422 0.000 1.26 3.01 5.66 0.92
2.06 2.16 10.79 1.45
1 66 0.566 0.000 3.70 4.34 17.56 2.89
2.05 1.34 12.77 1.00
2 53 0.647 0.000 3.98 4.61 19.90 3.22
1.98 1.23 12.83 0.80
3 26 0.608 0.001 4.73 5.00 24.80 3.80
1.77 1.10 12.19 0.79
146 WESLEY A. HOOVER AND PHILIP B. GOUGH

Table 4 (Continued)

MinR n Correlation Descriptive statistics

(mean and standard deviation)
r P D L P R
-

Grade 2
0 206 0.592 0.000 3.68 5.00 21.11 3.22
2.40 1.92 15.47 2.11
1 161 0.256 0.001 4.64 5.69 26.88 4.11
1.72 1.09 12.32 1.41
2 149 0.307 0.000 4.74 5.74 27.70 4.30
1.68 1.03 12.13 1.29
3 119 0.304 0.000 4.99 5.88 29.74 4.70
1.54 0.97 11.49 1.14
4 65 0.367 0.001 5.57 6.29 35.33 5.50
1.12 0.78 9.54 0.95
5 43 0.409 0.003 5.87 6.54 38.65 6.00
0.91 0.73 8.48 0.79
6 13 0.457 0.058 6.19 7.27 45.10 7.03
0.54 0.48 5.93 0.70
Grade 3
0 86 0.543 0.000 4.94 5.74 30.63 4.44
2.15 1.98 16.86 2.47
1 73 0.539 0.000 5.56 6.30 35.91 5.22
1.46 1.18 12.16 1.76
3 64 0.43 1 0.000 5.79 6.50 38.12 5.61
1.23 0.97 10.51 1.50
4 46 0.059 0.349 6.17 6.86 42.36 6.36
0.60 0.73 6.42 1.06
6 24 0.270 0.101 6.27 7.34 46.08 7.24
0.55 0.46 5.49 0.60
7 17 0.180 0.244 6.29 7.56 47.56 7.56
0.58 0.31 5.09 0.31
Grade 4
0 55 0.717 0.000 5.35 6.67 37.58 5.84
2.06 1.31 15.55 2.00
5 43 0.358 0.009 6.12 7.09 43.60 6.70
0.93 0.64 8.44 0.99
7 22 0.039 0.432 6.44 7.60 48.92 7.60
0.40 0.26 3.52 0.26
-
Min R = Minimum reading comprehension level included in the subsample
D = Decoding index
L = Listening comprehension index
P = Product of decoding and listening comprehension indices
R = Reading comprehension index
 SIMPLE VIEW OF READING 147

increasing reading comprehension skill. Within each grade level, the

correlations were positive, though at the smaller sample sizes within
Grades 3 and 4, they were not significantly different from zero. These
results fail to support an artifactual interpretation of the correlational
trends found within the weaker readers. Indeed, one finds outliers
distributed within the weaker readers of just the type predicted by the
simple view, namely, children with (a) poor listening comprehension skill
coupled with relatively stronger decoding skill, and @) poor decoding skill
coupled with relatively stronger listening comprehension skill. For the
strongest readers, such a distributional pattern is neither predicted nor
found.
Note that both additive and multiplicative combinations of decoding
and linguistic comprehension predict the obtained trend of positive
correlations in skilled readers becoming negative correlations in unskilled
readers. Thus, this evidence cannot inform any selection between these
conceptions. Nonetheless, the result is important in demonstrating the
separate contributions of decoding and linguistic comprehension to
reading ability, as the trend is consistent with the view that for skilled
reading, skill in both components is required, while a weakness in either
component is sufficient for less skilled reading.

Prediction 3
The third prediction drawn from the simple view was that as the level of
decoding skill increased, the pattern of linear relationships between
linguistic comprehension and reading comprehension would reveal (a)
constant intercept values of zero and (b) positive slope values increasing in
mangnitude from a floor value of zero. Unlike the previous prediction, this
one can distinguish between additive and multiplicative combinations of
decoding and linguistic comprehension, as the additive conception pre-
dicts (a) intercept values that increase from a floor value of zero (rather
than constant zero-valued intercepts), and (b) positive, but constant, slope
values (rather than positive slope values that increase in magnitude from a
floor value of zero).2
In assessing this prediction, intercept and slope values expressing the
linear relationship between listening and reading comprehension were
computed within each grade level for various subsamples of students
equated on decoding skill. In this analysis, decoding skill levels were
defined by the list of highest success (i.e., the integer value of the decoding
index, disregarding the decimal value). Some decoding skill levels were
collapsed in certain instances to obtain sample sizes that would provide
reasonably trustworthy estimates of the linear indices. As expected, in the
early grades such pooling was restricted to the higher decoding skill levels,
while in the latter grades, pooling was confined to the lower levels. The
148 W E S L E Y A. H O O V E R A N D P H I L I P B. G O U G H

results of these anlayses are displayed in Table 5, giving the number of

students contained in each decoding skill subsample, the intercept and
slope values, and their associated significance levels.
For the first-grade data, decoding levels 2, 3, and 4 (containing 2, 5,
and 10 students, respectively) were collapsed into a single group (thus
representing students whose decoding indices ranged from 2.00 to 4.99);
decoding levels 5 and 6 (with 12 and 13 students) were also combined. As
seen in Table 5, for each of the decoding skill groups, none of the inter-
cept values differed significantly from zero. The slope values, however,
were significantly positive, with the exception of that associated with the
lowest decoding skill subsample, which did not differ from zero. Further,
these slopes showed the predicted increases in magnitude with increases in
the level of decoding skill (evidence concerning the statistical significance
of these apparent increases is presented later).
For the second-grade data, only decoding levels 2 and 3 were pooled

Table 5. Intercept and slope for the linear relationship between reading and listening
comprehension at various levels of decoding

Decoding levels n Intercept Slope

Value p Value p

Grade 1
0--0.99 135 0.02 0,223 0.00 0.310
1--1.99 33 0.39 0.173 0.47 0.000
2--4.99 17 0.58 0.102 0.54 0,000
5--6.99 25 0.07 0,468 0.64 0.000
Grade 2
0--0.99 45 -0.14 0,162 0.11 0.003
1--1.99 22 1.20 0,046 0.38 0.004
2--3.99 17 0.76 0,155 0.48 0.001
4--4.99 43 1.32 0.064 0.41 0.006
5--5.99 38 0.07 0.472 0.79 0,000
6--6.99 41 -3.18 0.003 1.31 0.000
Grade 3
0--1.99 13 0.23 0.388 0.17 0.183
2--4.99 15 -0.31 0.285 0.77 0.000
5--5.99 18 -1.06 0.161 0.98 0.000
6--6.99 40 -3.74 0.000 1.44 0.000
Grade 4
0--4.99 12 --1.24 0.139 0.84 0.001
5--5.99 11 --3.26 0,114 1.35 0.003
6--6.99 32 --4.16 0.000 1.53 0.000
 SIMPLE VIEW OF READING 149

(containing 4 and 13 students, respectively). The intercept values were not

significantly different from zero with two exceptions. One value was
positive, but only marginally close to the 0.05 significance level, while a
second value, that associated with the highest decoding skill group, was
substantially negative. The slope values were all significantly positive, and
again, increased in magnitude with decoding skill.
For the third-grade data, decoding levels 0 and 1 were collapsed (with
9 and 4 students, respectively) and levels 2, 3, and 4 (with 1, 4, and 10
students). The intercepts again did not significantly differ from zero with
the exception of the value associated with the highest decoding skill group,
which, as in the second-grade data, was substantially negative. The slope
values again increased in magnitude with the level of decoding skill, each
significantly positive, except for that associated with the group of lowest
decoding skill, which did not differ from zero.
Collapsing decoding levels 0 through 4 (with 5, 2, 1, 1, and 3 students,
respectively), the fourth-grade data followed the pattern found in the
second-grade data: intercepts not significantly different from zero except
for the highest decoding skill group (with a large negative value), and
significantly positive slopes that increased in magnitude with decoding
skill.
Overall, for the 17 intercept values computed over the four grade-level
data sets, none differed significantly from zero, with the exception of one
value within the second-grade data set that was only marginally significant,
and three large negative values that were associated with the highest
decoding skill group within the second-, third-, and fourth-grade data sets,
respectively. These latter values are most likely artifacts reflecting a ceiling
effect in the decoding assessment, influencing the linear estimates as
follows. For the sample of students successful in decoding the highest level
list presented, the best decoders (i.e., those who would have decoded
higher level lists had such been available) were pooled with those of lower
skill (i.e., those who were successful on the most difficult list available, but
who would not have been successful on any more difficult lists had such
been presented). Such pooling results in the (expected) greater reading
comprehension skills of the former group of students contributing large
positive (i.e., outlier) values relative to the latter group, leaving the
intercept values deflated and the slope estimates inflated. Thus, in general,
the obtained intercept values conform to the prediction of the multiplica-
tive notion expressed in the simple view of reading.
For the corresponding 17 slope values, all were significantly positive
with the exception of two zero values, those associated with the lowest
decoding skill groups within the first- and third-grade data sets. As
discussed earlier, the simple view predicts zero-valued slopes for students
completely lacking decoding skills. However, students in the lowest
decoding skill groups in these analyses may, nonetheless, possess some
150 WESLEY A. HOOVER AND PHILIP B. GOUGH

decoding skill. First, the easiest list of synthetic words presented to these
students did not necessarily contain the simplest items that could be
decoded with the most rudimentary of decoding skills. Thus, the lowest
level of material may be relatively insensitive to the lowest levels of
decoding skill. Second, the upper boundary defining the lowest decoding
skill group was 0.99 for the first- and second-grade analyses, and 1.99 and
4.99 for the third- and fourth-grade analyses, respectively. As such, for
each grade level the ranges of skill included in the lowest decoding skill
group allowed for at least some success in decoding, though it may not
have been sufficient to meet the criterion for success on the lowest level
list presented (i.e., not succeeding on the lowest level list, but correctly
decoding one or two of the six items it contained).
Thus, for these groups of low skill decoders, the simple view does not
predict slope values of zero, but rather values that are the lowest of any
associated with a group possessing greater decoding skill. As seen in Table
5, the data followed just this pattern.
To assess the statistical trustworthiness of the slope increases, the
correlation between decoding skill (using each group’s mean decoding
value) and slope value was computed. Cautioning that these estimates are
not independent (given the longitudinal nature of the data base), the
correlation over the 17 relevant pairs was significantly positive (r[16] =
0.85, p < 0.001); the coefficient was only slightly reduced when the
inflated slope estimates from the highest decoding skill groups within each
grade level were dropped (r[12] = 0.81, p < 0.001). Thus, these findings
indicate that the rate of improvement in reading comprehension over
levels of listening comprehension increased with decoding skill, such
predicted by the simple view.
To summarize, in accord with the third prediction of the simple view,
the results generally indicated that as the level of decoding skill increased,
the linear relationships between listening comprehension and reading
comprehension were characterized by constant intercept values of zero
and positive slope values increasing in magnitude.

IMPLICATIONS

A simple view of reading has been proposed holding that reading consists
of only two components, decoding and linguistic comprehension. The
simple view holds that neither component is sufficient for skilled reading,
but rather, that skill in both components is necessary if skill in reading is
to advance.
The simple view does not deny the complexity of reading, but asserts
that such complexities are restricted to either of the two components.
Consider a more complex model proposed by Calfee and Drum (1986),
 SIMPLE VIEW OF READING 151

where reading is held to consist of decoding, vocabulary, sentence

comprehension, paragraph comprehension, and text comprehension. The
simple view, while not denying the added complexity, argues that the last
four components are constituents of linguistic comprehension. Impor-
tantly, such a simplification explicitly states that these skills are not
specific to reading, that advancement in any one represents advancement
in linguistic skill, such also advancing reading skill to the degree allowed
by skill in decoding.
While the simple view provides an adequate account of reading, the
task remains to define the components underlying decoding and linguistic
comprehension. Thus, might the vocabulary and sentence comprehension
skills specified in the Calfee and Drum (1986) model be combined
multiplicatively, and if so, how might their product be combined with the
additional proposed skill areas of paragraph and text comprehension?
Answers to these questions await further research.
While acknowledging these issues, the simple view of reading does have
several important implications, these concerning the practice of reading
instruction, the definition of reading disability, and the notion of literacy.
Each of these is discussed below.

Reading Instruction
The simple view suggests that instruction that advances skill in either
decoding or linguistic comprehension will promote skill in reading as long
as skill in neither component is nil. The instructional implications for both
components are treated in turn below.

Decoding. While language is naturally acquired by the normal child

through exposure in the context of human interaction, the acquisition of
decoding is not, formal instruction generally being required (Calfee and
Drum 1986; Gough and Hillinger 1980; Stanovich 1986). The difficulty in
acquiring decoding skills is that a natural strategy based on selective
association (the pairing of a partial stimulus cue to a response), while
initially successful in linking the printed and spoken word, has limited
utility: selective association will not permit the recognition of novel printed
words, and for the beginning reader, print novelty is ubiquitous (Gough
and Hillinger 1980; Jorm and Share 1983). In alphabetic systems, if the
child can learn the systematic relationship between the units of the printed
and spoken word, then novel printed words can be accessed without the
requirement imposed by the associative process that such novel words be
accompanied by pronunciations.
If the child is to advance from the stage of code reading characterized
by selective association, to that of cipher reading based on knowledge of
the systematic relationship between printed and spoken word, a number of
152 WESLEY A. HOOVER AND PHILIP B. GOUGH

conditions must be met (Cough and Hillinger 1980). The child must have
(a) the intent to discover the print-sound relationship, (b) awareness of the
alphabetic units of the printed word, (c) awareness of the phonemic units
of the spoken word, and (d) access to sufficient data (namely, pairs of
printed and spoken words). If such specific conditions must be met in
order to successfully internalize the cipher, then instruction that efficiently
addresses these should facilitate decoding acquisition.
Though instructional effectiveness is determined by more than just
method (e.g., Barr 1984) some have argued that phonic approaches to
reading instruction represent the most effective schemes known for
teaching decoding (Flesch 1981; Williams 1985). While there is evidence
suggesting that phonic approaches are superior to other methods with
which they have been compared (Bond and Dykstra 1967; Chall 1967;
Guthrie, Martuza, and Seifert 1979; Johnson and Baumann 1984) such
evidence does not imply that phonic approaches are the most effective
and efficient instructional methods for acquiring the cipher. Further, the
relative successes of phonic approaches do not imply that those successes
are attributable to the acquisition and use of the grapheme-phoneme
correspondence rules generally taught in such methods (e.g., when two
vowels go walking, the first does the talking). As argued elsewhere (Gough
and Hillinger 1980) such rules cannot be the basis of lexical access for
they are both too few and too slow.
Rather than having any direct influence, the effects of phonic approaches
in acquiring the cipher may be entirely indirect. As one possibility, phonic
methods may simply facilitate phonological awareness (explicit recognition
that the speech stream can be segmented into discrete phonemic units),
thereby satisfying one of the four conditions specified above for acquiring
the cipher. To take another possibility, instruction in phonics may simply
provide the child with a strategy for generating pronunciations from
printed words, such contributing data to be used in discovering the
systematic relationship embodied in the cipher. In each of these cases, the
effect of phonic approaches would have no direct bearing on the specific
character of the cipher acquired.
To summarize, the simple view holds that skill in decoding must be
acquired for success in reading. Further, it has been argued that in an
alphabetic orthography, decoding acquired as ciphering will allow the
recognition of novel printed words, thus freeing instruction from having to
provide pronunciations for every novel printed word encountered by the
child. While phonics instruction seemingly facilitates acquisition of the
cipher (though the mechanism of influence is unknown), the simple view
does not hold that phonics instruction is necessary in order to acquire the
cipher: what is acquired (the precise content of the cipher) and what
constitutes the best method for acquiring it are independent, and currently
unanswered, questions.
 SIMPLE VIEW OF READING 153

Linguistic comprehension. The simple view also claims that linguistic

comprehension is a necessary component of skilled reading. The simple
view does not claim that exactly the same procedures used in linguistic
comprehension are employed in reading comprehension, for there are
clear differences. To take a few examples, the suprasegmentals repre-
sented in speech are greatly impoverished in written language; the
availability of previous input makes review much easier for written than
for spoken language; and the interpretation of deictic terms may be
derived differently in written than in spoken language (Danks 1980; Rubin
1980). Each of these differences is likely to be reflected in processing
differences.
The simple view, however, argues that these differences are relatively
minor in comparison to the great similarities between linguistic and
reading comprehension. The commonalities argued for in the simple view
suggest that instruction facilitating linguistic comprehension should like-
wise facilitate reading comprehension (if decoding is not nil), and indeed,
a number of studies indicate that improvements in listening comprehen-
sion (effected through a variety of training programs) lead to improve-
ments in reading comprehension (for a review, see Sticht and James
1984).
In addition, the simple view implies that as an individual’s knowledge
base increases (through whatever mechanism), as that knowledge base is
reflected in linguistic comprehension, so should it be reflected in reading
comprehension, and vice-versa. To take one relevant study, Pearson,
Hansen, and Gordon (1979) have shown that specific knowledge of a
given area is positively related to reading comprehension. Taking two
groups of second-grade readers matched on IQ and general reading
comprehension, but differing in specific knowledge of a given domain,
Pearson et al. found that those with greater expertise in the selected
domain better comprehended passages pertinent to that domain than did
those with less expertise. In terms of the simple view, the greater the
knowledge base expressible through linguistic comprehension, the greater
the reading comprehension (assuming non-zero decoding skills).
As pointed out by Perfetti and Curtis (1986), the activation of relevant
knowledge while reading may be problematic for some children, the
difficulty stemming from effortful decoding. For such children, instruction
that leads to more efficient decoding will allow more of limited cognitive
resources to be devoted to comprehension processes, including the activa-
tion of relevant knowledge.

Reading Disability
The simple view of reading also has important implications for reading
disability (Gough and Tunmer 1986). As reviewed earlier, skill in decod-
154 WESLEY A. HOOVER AND PHILIP B. GOUGH

ing and linguistic comprehension have been found to be positively related

in the literate population. Such would be expected given the reciprocal
relationships (Stanovich 1986) between these skills (i.e., as decoding skill
improves, reading improves, such tending to lead to increased amounts of
reading, which further strengthens both skill in decoding and linguistic
comprehension). However, within the illiterate population, the simple view
of reading holds that decoding and linguistic comprehension can be
dissociated, and, most interestingly, must be dissociated if substantial skill
is evidenced in either of the two components.

Dyslexia. There has long been recognition of a specific difficulty restricted

to reading that is not accompanied by difficulty in other cognitive
functioning. Commonly known as dyslexia, the disability is defined by
exclusion, dyslexics being those who have difficulty reading despite normal
intelligence and sensory functioning, an adequate opportunity to learn,
and an absence of severe neurological or physical disability, emotional or
social difficulty, or socioeconomic disadvantage (Vellutino 1979).
The simple view, while silent on the unresolved issue of a singular or
multiple etiological base, claims that the reading difficulties encountered
by all such dyslexic individuals, by definition linguistically competent, can
only stem from a deficiency in decoding skill. Though definitive studies
have not been conducted, there are a number of investigations showing
that individuals selected on the basis of the exclusionary criteria given
above are indeed substantially deficient in decoding skill (Doehring,
Trites, Patel, and Fiedorowicz 1981; Seymour and Porpodas 1980;
Snowling 1980; Vellutino 1979). Thus, while the question of why decod-
ing skills are deficient in dyslexics is unanswered, there is support for the
assertion of the simple view that the commonality of reading failure
represented in dyslexia is a failure in decoding.

Hyperlexiu. A second type of dissociation within the illiterate population

predicted from the simple view concerns superior decoding skill accom-
panied by inferior comprehension skill. Given the nature of language
acquisition, such cases are rare; those that have been identified are known
as hyperlexics.
A study by Healy (1982) identified a small sample of children who
evidenced exceptional skill in decoding but reading age equivalents that
were some two years lower than those expected from chronological age.
While demonstrating that decoding skill is not sufficient for reading skill, it
does not follow from such evidence that decoding is therefore not
necessary for reading. Indeed, as would be predicted by the simple view,
the students in Healy’s study evidenced listening age equivalents that were
also two years lower than their chronological ages. From the simple view,
 SIMPLE VIEW OF READING 155

when decoding skills are highly developed, reading and listening skills will
be equivalent, and in the case of hyperlexia, reduced linguistic comprehen-
sion will be responsible for reduced reading comprehension.

Literacy
A third implication of the simple view of reading concerns the notion of
literacy. Clarity about the meaning of literacy is important not just for
purposes of theory, but also because confusion over what constitutes
literacy will likely result in confusion over the solutions proposed to
reduce illiteracy. In the treatment below, only those aspects of literacy
linked to reading will be discussed, ignoring writing (though the arguments
could be extended to include it).
From the simple view, literacy (limited to reading) may be seen as the
contrast between linguistic comprehension and reading comprehension.
For example, suppose an individual has mastered most of the syntax of
English yet knows only the most frequent English words. Further, suppose
this person can decode the printed versions of precisely the same set of
known words, no more nor fewer. Under the simple view of reading,
linguistic comprehension and reading comprehension in this individual are
equivalent: with respect to current linguistic skill, such a person is fully
literate (for reading) since whatever can be comprehended by ear can
likewise be comprehended by eye, and vice versa.
Simply increasing the decoding skill of such an individual will not
increase reading comprehension as the meaning of any words that can
now be decoded given the newly expanded skill will still be absent from
the internal lexicon. In like fashion, simply increasing linguistic compre-
hension by, for example, expanding the domain of known words, will also
fail to result in increased reading comprehension unless success in
decoding the printed representations of those words in the enlarged
domain is also guaranteed.
Treating literacy as reading ability, expanded linguistic comprehension
(e.g., through expansion of lexical entries) without expanded decoding skill
results in reduced literacy as the difference in skill between oral language
(linguistic comprehension) and written language (reading comprehension)
has been increased. However, this circumstance also results in an
increased potential for literacy as linguistic capacity has been expanded
thereby allowing expanded reading potential (Sticht and James 1984)
should adequate decoding skills be acquired.
Of course, the reading skills of fully literate individuals can be
improved. If decoding skills are adequate to efficiently decode any word
encountered, then the limit on reading is the limit on linguistic compre-
hension, and for each increase in linguistic comprehension, there will be
156 WESLEY A. HOOVER AND PHILIP B. GOUGH

an equal increase in reading comprehension. Similarly, for individuals

whose decoding skill is not adequate to decode new entries to the internal
lexicon, if decoding skill improves to allow recognition of the printed
representation of such new entries, then reading comprehension will
improve in step with linguistic comprehension.
While the reading ability of an individual can be simply viewed as the
difference between linguistic comprehension and reading comprehension,
many argue that the notion of literacy entails a certain degree of concep-
tual understanding or cognitive ability assessed relative to an external
standard, for example, by reference to some set of concepts one is
expected to be familiar with or to some set of tasks one is expected to be
able to perform (Applebee, Langer, and Mullis 1987; Harman 1987;
Venezky, Kaestle, and Sum 1987). A number of points can be made
concerning the relationship between reading and a literacy component
representing conceptual understanding.
First, the acquisition of familiarity with a specified set of culturally
valued literary works or knowledge bases is independent of the mode of
acquisition, as such can be achieved in a number of ways that do not
involve the individual’s own reading. Second, Fries’ (1963) point, cited
earlier, that the skills of thinking, evaluating, judging, imagining, reasoning,
and problem-solving can be found in both illiterates and literates is
important, as it emphasizes that these cognitive abilities are not the
exclusive dependents of reading. Finally, while reading is exclusively
linguistic, many of the commonly claimed literacy tasks are non-linguistic
(e.g., knowing how to carry out arithmetic operations or how to “read” a
map).
These points all argue that the notion of conceptual understanding as a
component of literacy is logically independent of reading. While the
understanding of a particular concept does not have a necessary connec-
tion to reading ability, as understanding develops, linguistic comprehen-
sion with respect to that conceptual arena will also likely expand. As
linguistic comprehension expands, reading potential will increase, this
potential being realized if decoding skills are adequate. That such under-
standing may develop either through oral or written language is a key
point, the important application as follows: Teaching reading (in the
simple sense of achieving equal linguistic comprehension and reading
comprehension) will not reduce the problem of illiteracy if that problem is
mainly seen as one focused on conceptual understanding. While reading
undoubtedly can further conceptual understanding (e.g., learning through
reading), to substantially reduce illiteracy defined with respect to concep-
tual understanding the overall education of individuals must be considered
not just their ability to understand through reading what can be under-
stood through listening.
 SIMPLE VIEW OF READING 157

CONCLUSIONS

The strength of the simple view of reading, in addition to its simplicity, is

that it has allowed a set of non-trivial and testable predictions. Under the
assessments described above these predictions could not be falsified,
thereby supporting the assertion that skill in reading can be simply charac-
terized as the product of skill in decoding and linguistic comprehension.
Implications of the simple view, arguing for the necessity and suffi-
ciency of decoding and linguistic comprehension in skilled reading,
include the following: (a) when skill in both components is non-zero, then
instruction that advances skill in either component will advance skill in
reading; (b) within the illiterate population, skill in decoding and linguistic
comprehension will be inversely related if substantial skill is evidenced in
either component; and (c) advancing literacy by achieving equal linguistic
and reading comprehension will not necessarily have direct impact on the
problem of illiteracy if that problem is mainly seen as one of conceptual
understanding.

ACKNOWLEDGMENTS

This publication was based upon work originally sponsored by the

National Institute of Education (NIE), Department of Education, under
Contract No. 400-83-0007 to the Southwest Educational Development
Laboratory. The content does not necessarily reflect the views of the NIE,
the Department, or any other agency of the U.S. Government. The authors
thank John Loehlin for his thoughtful comments on an earlier draft of the
paper. Correspondence concerning this article should be addressed to
Wesley A. Hoover, Southwest Educational Development Laboratory, 211
East Seventh Street, Austin, Texas, 7870 1.

NOTES

’ Auding, a term coined by Brown (1954), is defined as listening to language for the
purpose of comprehension.
z John Loehhn has pointed out that when measurement error is present, both additive and
multiplicative combinations of decoding and linguistic comprehension predict increases in
intercept values. This is because at the floor, error will result in some obtained zero values
that are associated with greater than zero true values, and at the ceiling, some obtained
maximal values that correspond to less than maximal true values. Fitting lines based on the
products of such true scores results in positive (rather than zero) intercepts, the magnitude
of the intercept increasing with decoding level. Measurement error, however, does not
affect the different predictions of the additive and multiplicative notions with respect to
slope values.
158 WESLEY A. HOOVER AND PHILIP B. GOUGH

REFERENCES

Anderson, R., Hiebert, E., Scott, J., and Wilkinson, I. 1985. Becoming a nation of readers:
The report of the Commission on Reading. Urbana, IL: University of Illinois, Center for
the Study of Reading.
Applebee, A., Langer, J., and Mullis, I. 1987. Learning to be literate in America. Princeton,
NJ: National Assessment of Educational Progress, Educational Testing Service.
Barr, R. 1984. Beginning reading instruction: From debate to reformation. In P. Pearson
(ed.), Handbook of reading research (pp. 545-58 1). New York: Longman.
Bond, G. and Dykstra, R. 1967. The cooperative research program in first-grade reading
instruction. Reading Research Quarterly, 2,5--142.
Brown, D. 1954. Auding as the primary language ability. Unpublished doctoral disserta-
tion, Stanford University.
Calfee, R. and Associates. 1978. Stanford foundation skulls test. Unpublished manuscript,
Stanford University, Stanford, CA.
Calfee, R. and Associates. 1980. Stanford foundation skills test (rev. ed.). Unpublished
manuscript, Stanford University, Stanford, CA.
Calfee, R. and Calfee, K. 1979. Interactive reading assessment system. Unpublished
manuscript, Stanford University, Stanford, CA.
Calfee, R. and Calfee, K. 1981. Interactive reading assessment system (rev. ed.). Unpub-
lished manuscript, Stanford University, Stanford, CA.
Calfee, R. and Drum, P. 1986. Research on teaching reading. In M. Wittrock (ed.),
Handbook of research on teaching (3rd ed., pp. 804-849). New York: Macmillan.
Chall, J. 1967. Learning to read: The great debate. New York: McGraw-Hill.
Curtis, M. 1980. Development of components of reading skill. Journal of Educational
Psychology, 72,656-669.
Danks, J. 1980. Comprehension in listening and reading: Same or different? In F. Murray
(ed.), Reading and understanding (pp. l-39). Newark, DE: International Reading
Association.
Doehring, D., Trites, R., Pate], P., and Fiedorowicz, C. 1981. Reading disubihties: The
interaction of reading, language, and neuropsychological deficits. New York: Academic
Press.
Flesch, R. 198 1. Why Johnny still can’t read. New York: Harper & Row.
Freedman, S. and Calfee, R. 1984. Understanding and comprehending. Written Communi-
cation, 1,459-490.
Fries, C. 1963. Linguistics and reading. New York: Holt, Rinehart & Winston.
Gates, A. 1949. Character and purposes of the yearbook. In N. Henry (ed.), The forty-
eighth yearbook of the National Society for the Study of Education: Part II. Reading in
the elementary school (pp. l-9). Chicago: University of Chicago Press.
Glushko, R. 1979. The organization and activation of orthographic knowledge in reading
aloud. Journal of Experimental Psychology: Human Perception and Peformance, 5,
674-691.
Gough, P. 1983. Context, form, and interaction. In K. Rayner (ed.), Eye movements in
reading: Perceptual and language processes (pp. 203-211). New York: Academic
Press.
Gough, P. 1984. Word recognition. In P. Pearson (ed.), Handbook of reading research (pp.
225-253). New York: Longman.
Gough, P. and Hillinger, M. 1980. Learning to read: an unnatural act. Bulletin of the Orton
Society, 30,1-17.
Gough, P. and Tunmer, W. 1986. Decoding, reading, and reading disability. Remedial and
Special Education, 7,6-10.
 SIMPLE VIEW OF READING 159

Guthrie, J., Martuza, V., and Seifert, M. 1979. Impacts of instructional time in reading. In
L. Resnick and P. Weaver (eds.), Theory and practice of early reading (Vol. 3, pp. 153-
178). Hillsdale, NJ: Erlbaum.
Harman, D. 1987. Illiteracy: A national dilemma. New York: Cambridge Book Company.
Healy, J. 1982. The enigma of hyperlexia. Reading Research Quarterly, 17,319-338.
Henderson, L. 1982. Orthography and word recognition in reading. London: Academic
Press.
Hoover, W., Calfee, R., and Mace-Matluck, B. 1984a. Teaching reading to bilingual
children study: Vol. 5. Reading growth (Final Report, Contract No. 400-83-0007).
Washington, DC: National Institute of Education.
Hoover, W., Calfee, R., and Mace-Matluck, B. 1984b. Teaching reading to bilingual
children study: Vol. 6. Instruction (Final Report, Contract No. 400-83-0007). Washing-
ton, DC: National Institute of Education.
Huey, E. 1968. The psychology and pedagogy of reading. Cambridge: MIT Press. (Original
work published 1908)
Jackson, M. and McClelland, J. 1979. Processing determinants of reading speed. Journal of
Experimental Psychology: General, 108,15 l-l 8 1.
Johnson, D. and Baumann, J. 1984. Word identification. In P. Pearson (ed.), Handbook of
reading research (pp. 583-608). New York: Longman.
Jorm, A. and Share, D. 1983. Phonological recoding and reading acquisition. Applied
Psycholinguistics, 4,103-147.
Olson, D. 1977. From utterance to text: The bias of language in speech and writing.
Harvard Educational Review, 47,257-28 1.
Palmer, J., MacLeod, C., Hunt, E., and Davidson, J. 1985. Information processing
correlates of reading. Journal of Memory and Language, 24,59-88.
Pearson, P., Hansen, J., and Gordon, C. 1979. The effect of background knowledge on
young children’s comprehension of explicit and implicit information. Journal of Reading
Behavior, II, 201-209.
Perfetti, C. and Curtis, M. 1986. Reading. In R. Dillon and R. Sternberg (eds.), Cognition
and instruction (pp. 13-57). Orlando: Academic Press.
Rubin, A. 1980. A theoretical taxonomy of the differences between oral and written
language. In R. Spiro, B. Bruce, and W. Brewer (eds.), Theoretical issues in reading
comprehension (pp. 4 1 l-438). Hillsdale, NJ: Erlbaum.
Rumelhart, D. 1977. Toward an interactive model of reading. In S. Domic (ed.), Attention
andpe$ormance VI (pp. 573-603). HiIlsdale, NJ: Erlbaum.
Seymour, P. and Porpodas, C. 1980. Lexical and non-lexical processing of spelling in
dyslexia. In U. Frith (ed.), Cognitive processes in spelling (pp. 443-473). London:
Academic Press.
Singer, M. and Crouse, J. 198 1. The relationship of context-use skills to reading: A case
for an alternative experimental logic. Child Development, 52, 1326-l 329.
Snowling, M. 1980. The development of grapheme-phoneme correspondence in normal
and dyslexic readers. Journal of Experimental Child Psychology, 29,294-305.
Stanovich, K. 1980. Toward an interactive-compensatory model of individual differences
in the development of reading fluency. Reading Research Quarterly, 16,32-7 1.
Stanovich, K. 1986. Matthew effects in reading: Some consequences of individual differ-
ences in the acquisition of literacy. Reading Research Quarterly, 21,360-407.
Stanovich, K., Cunningham, A., and Feeman, D. 1984. Intelligence, cognitive skills, and
early reading progress. Reading Research Quarterly, 19,278-303.
Stanovich, K., Nathan, R., and Vala-Rossi, M. 1986. Developmental changes in the
cognitive correlates of reading ability and the developmental lag hypothesis. Reading
Research Quarterly, 21,267-283.
 160 WESLEY A. HOOVER AND PHILIP B. GOUGH

Stanovich, K. and West, R. 1983. On priming by a sentence context. Journal of Experi-

mental Psychology: General, 112, I-36.
Sticht, T. and James, J. 1984. Listening and reading. In P. Pearson (ed.), Handbook of
reading research (pp. 293-3 17). New York: Longman.
Vellutino, F. 1979. Dyslexia: Theory and research. Cambridge, MA: MIT Press.
Venezky, R., Kaestle, C., and Sum, A. 1987. The subtle danger: Rejlections on the literacy
abilities of America’s young adults. Princeton, NJ: Educational Testing Service.
Williams, J. 1985. The case for explicit decoding instruction. In J. Osbom, P. Wilson, and
R. Anderson (eds.), Reading education: Foundations for a literate America (pp. 205-
2 13). Lexington, MA: Lexington Books.

View publication stats