This study was supported by a clinical research center grant P50-DC002746,
from the NIH. This study was aided considerably by a research team comprised
of:Marlea O’Brien, Connie Ferguson, Marsh St. Clair, and Amy Schminke.
Is there a Matthew Effect?
d Vocabulary Growth in School Aged Children
Dawna Duff, J. Bruce Tomblin
Dept. of Speech Pathology and Audiology,
University of Iowa
Dept. of Speech-Language-Hearing: Sciences and Disorders,
University of Kansas
This study investigated the impact of early decoding skill on vocabulary growth.
Since reading is expected to have a greater impact on vocabulary acquisition
after fourth grade, this study focuses on changes in the rate of vocabulary
acquisition after fourth grade. A large longitudinal sample, including subjects with
and without language disorders, was tested at kindergarten, second, fourth,
eighth, and tenth grades. Data on vocabulary skills were converted to
developmental ability (W) scores. The data were analyzed using multilevel
modeling of vocabulary growth over time. The rate of vocabulary growth after
fourth grade was significantly related to fourth grade word reading after
controlling for kindergarten vocabulary level. These results support the
hypothesis that word reading ability may influence vocabulary growth in older
children and thus supports the Matthew Effect hypothesis. The magnitude of the
effect on 10th grade vocabulary was shown to be substantial. The data also
showed that the Matthew Effect was greatest for children with above average
reading. Thus, the very rich do indeed become richer.
What is the “Matthew Effect”?
“The very children who are reading well and have good
vocabularies will read more, learn more word meanings, and
hence read even better.” (Stanovich, 1986)
Background for Current Study
Current Status of Matthew Effect (ME)
The Matthew Effect in reading is widely cited. Over 900 citations of Stanovich’s
original paper were found on the Web of Science, however, the evidence in
support of the ME has been mixed.
• Juel (1988) reported poor early reading (1st grade) resulted in poor later reading
(4th grade); however, this study did not directly test for a fan effect.
• Bast and Reitsma (1998) demonstrated increased in the magnitude of individual
differences in word recognition during the early school years, but not for reading
comprehension. Shaywitz, et al (1995) shown no ME on reading, but an effect
was found for IQ.
• Scarborough (2000), Aarnouste and van Leeuwe (2000) Thompson (2003),
Parilla et al, (2005) found little support for a ME within reading
• Scarborough (2003) concluded, “Matthew effects were elusive, despite the
plausability and widespread acceptance of that well-reasoned hypothesis” (p.8).
• Studies have mostly focused on the ME effect within reading whereas Stanovich
(1986) described language as the critical mediating variable in the reciprocal
Therefore, oral language especially vocabulary should be most sensitive to the
• Early reading experience does not expose the child to many new words. The
impact of reading should therefore be expected after the child begins to be
exposed to new words via reading.
Therefore, the ME on vocabulary should be most evident in the later grades and
in particular for those children with good reading skills because these children
are exposed to more new words
Current Study Design
Does oral vocabulary growth from fourth grade (age 10) to 10th grade (age 16)
differ in accord with word reading skills at 4th grade?
Vocabulary as Outcome Variable
Rationale: Vocabulary growth is one of the variables in the reciprocal
relationship described by Stanovich (1986). In contrast, decoding is a
constrained skill (Paris, 2005) that would not be expected to show
substantial variation at all age levels.
Growth Rates after 4th Grade
Rationale: Vocabulary growth through written language will only begin
once new vocabulary is encountered in text, and decoding skills are
adequate for reading new words. This typically happens after grade three.
Word Reading as Predictor Variable
Rationale: It is assumed that word reading skill will govern the reading
volume and the nature of the reading material that the child is exposed to.
In contrast, reading comprehension as a dependent variable would be
expected to be confounded with vocabulary knowledge.
Participants and Tasks
• 604 children participated, 504 completed the study
• Sample was collected from a larger sample of 1,794 kindergarten students
• Data collected in kindergarten, second, fourth, eighth and tenth grades
• 276 subjects had language and cognitive skills in the average range
• 328 subjects were identified with language impairment or nonverbal impairment
• Composite developmental ability scores (W scores) were constructed from the vocabulary
• Oral Vocabulary Measures:
Receptive Vocabulary Expressive Vocabulary
Kindergarten TOLD-2:P TOLD-2:P
Picture ID Oral Vocabulary
Grades 4, 8 10 PPVT-R CREVT-E
• Reading Measures: Woodcock Reading Mastery Test R (Word Identification and Word Atta
Results: Vocabulary Growth
Mean Vocabulary Growth
Vocabulary Developmental Score
rte n Second Fourth Eighth Tenth
Figure 2. Distribution of developmental ability scores for vocabulary at
each observational interval from kindergarten through tenth grade for all
children in the longitudinal study.
Individual Differences in Vocabulary Growth
High Word Readers
4.0 Middle Word Readers
Low Word Readers
3 4 5 6 7 8 9 10 11
Figure N VocabularySLI
3. scores of subjects grouped by word reading
skill (High >1 SD, Low<-1 SD, Middle -1 to +1 SD).
Figure 3 presents the growth patterns of averaged vocabulary scores for children
grouped by word reading ability.
• Overall vocabulary levels differed for the three groups of readers.
• A “fan effect” is evident revealing a divergence in vocabulary over grades.
• There does appear to be a greater effect of 4th grade reading on vocabulary growth
among the students with the highest reading abilities.
Table 1. Tests of random (level 1) and fixed
effects for vocabulary growth using
kindergarten vocabulary as a covariate
Effects Parameter F Value
Kindergarten Voc 0.20 25.01*
4th Grade Reading 0.11 22,68*
Rate of Change
Grade 0.23 2046.20*
Kindergarten Voc. 0.009 2.50
4th Grade Reading 0.03 28.68*
Proc Mixed (SAS) was used to test to obtain the parameters of intercept and slope
for each child’s vocabulary across the three observation intervals. Table 1 shows
that the average starting level (intercept) was 1.89 and the average growth rate
was 0.23 points per year. After controlling for their kindergarten vocabulary level,
the student’s vocabulary in fourth grade was related to their word reading. Also, the
direct test of the ME is found in the degree to which 4th grade reading is associated
with vocabulary development. This effect was significant F(1,485)=28.68, p<.0001.
provides plots SLI
Figure 4 N based on the linear growth modeling of random samples of
children at different levels of 4th grade reading levels. These growth functions look
very similar to the obtained data in Figure 3, with exception of the absence of the
nonlinear quality. Otherwise the fit appears very good.
The parameter of .03 means that increase of 1 standard score unit in reading
would result in an increase of 0.03 points per year in vocabulary. A difference of 5
points (1/3 of an SD) would result in a 0.15 differential gain in vocabulary which
multiplied by the 6 years of the study would result in a 0.9 point gain across the
study. 0.9 points at 10th grade represents 1.5 SD.
A separate analysis was performed comparing the three groups of 4th grade
readers to determine if growth rates in vocabulary differed for the three groups.
This analysis showed that “high” level readers had significantly better growth than
the “mid” level readers t(485)=3.51, p<.0005. However, the “mid” and “low”
readers were not different with respect to growth rate in vocabulary t=(485) -1.65,
Evidence for an effect of word reading ability at 4th grade on growth of oral
vocabulary was found. Thus, support for the Matthew Effect of reading on
vocabulary was found.
The fact that vocabulary in 4th grade was associated with reading levels at this grade
despite having removed vocabulary ability differences measured in kindergarten
suggests that these two variables influence each other during earlier reading
development as well. The direction of this influence is not known.
The Matthew Effect was found to be greatest for children with above average reading
and no effect was found among average and poor readers. Thus, it appears that the
very rich in reading get richer in vocabulary, but the middle class and impoverished
are less affected.
These findings would support the notion that good word recognition skills serve as a
gateway to the development of more advanced vocabulary learning. This assumes
that with with these good skills the child is also being exposed to literature that
provides exposure to these advanced lexical items.