Copyright © IEJEE
March 2019, Volume 11, Issue 4, 309-318
The Ineciency of Vocabulary Instruction
Jerey Lawrence McQuillan*
*Correspondence Details: Jerey Lawrence McQuillan, Center for Educational Development in Los Angeles, California. P.O. Box 66577, Los Angeles,
CA, U.S.A 90066. E-mail: firstname.lastname@example.org
Several researchers have advocated explicit instruction of vocabulary in order to help students improve their reading comprehension, especial-
ly low-achieving readers who need to “catch-up” to their age peers. Very few studies, however, have attempted to compare the time eciency of
direct instruction to its alternatives. In this review, I calculate the eciency of vocabulary instruction in 14 studies taken from a recent research
review (Wright & Cervetti, 2017). I then compare those results with estimates of vocabulary acquisition via a likely alternative source of vocab-
ulary growth, free reading. Free reading was found to be 1.7 times more ecient than direct instruction in building vocabulary in short-term
treatments, and 12 times as ecient for long-term treatments.
Keywords: Vocabulary Acquisition, Vocabulary Instruction, Pleasure Reading, Eciency
Success in school rests in signicant measure on the ability
to understand what one reads. Reading comprehension is in
turn strongly inuenced by one’s vocabulary knowledge (An-
derson & Freebody, 1981). Some researchers have conclud-
ed that the best way to help students improve both reading
comprehension and academic achievement is through some
form of direct, systematic vocabulary instruction (Beck, Per-
fetti, & McKeown, 1982; Biemiller & Boote, 2006; National
Reading Panel, 2000; Stahl & Nagy, 2007; Stahl & Fairbanks,
1986). More recently, those emphasizing the importance of
acquiring “academic” vocabulary have recommended teach-
ing these words directly to students (Nagy & Townsend,
2012; Snow, Lawrence, & White, 2009).
While vocabulary instruction typically leads to some gains in
word knowledge, not all instruction improves reading com-
prehension. In particular, vocabulary instruction that is lim-
ited to giving students the denitions of words – “shallow”
instruction – often has little impact on comprehension of
texts that contain those words (e.g. Pany & Jenkins, 1978;
Pany, Jenkins, & Schreck, 1982). In place of shallow instruc-
tion, some researchers have proposed a more time-intensive
“rich” instruction that, they claim, will lead not only to greater
word knowledge but also increased comprehension. Beck,
McKeown, and Omanson (1987), for example, identied sev-
eral elements of what they considered eective rich vocabu-
lary instruction, including:
• Providing clear denitions;
• Having students “manipulate” words in “rich and
varied ways,” describing how words relate to each
• Requiring students to discuss words and give “jus-
tications for the relationships” among words they
• Encountering the words frequently and in dierent
• Encouraging the use of words outside of the vocab-
ulary lessons (p. 149)
Some words are considered better candidates for this more
extensive form of vocabulary instruction than others. Beck et
al. (1987) categorized words into three “tiers” in order to de-
termine appropriate targets for instruction. Basic vocabulary
(“Tier 1”) consists of words that most students will acquire by
the early grades (e.g. cat, mother, talk, chocolate), and are
therefore not good candidates for instruction. “Tier 3” words
are those that are either used rarely or limited to a specic
domain, the latter often referred to as technical vocabulary
(e.g. photosynthesis, tidal pool, cosine). These again would
not be good targets for instruction, since they can be learned
“when the specic need arises, such as presenting nebula
during a lesson or discussion of the solar system” (Beck et al.,
1987, p. 155). Beck et al. recommend that teachers instead
focus their vocabulary instruction on “Tier 2” words, those
that are “of general utility not limited to a specic domain” (p
155). These are also sometimes called sub-technical words
(Cowan, 1974), and can be found in a wide variety of gen-
res and subject matter texts (e.g. inuence, ponder, retort,
Most vocabulary interventions have identied these words
based on teacher or researcher judgment. A few more recent
interventions have used words from the Academic Word List
(AWL) (Coxhead, 2000). The list consists of 570 word fami-
lies thought to be especially important in academic reading.
Ming-Tzu and Nation (2004) found that the AWL word mean-
ings were roughly similar across disciplines, meaning that ac-
quiring an AWL word in one domain will be benecial in other
academic disciplines as well.
Studies of the eects of teaching words on reading compre-
hension have produced mixed results. Stahl and Fairbanks
(1986) found a modest eect of instruction on standardized
reading tests (d= .30) but a much stronger one for research-
er-created passages that contained the words taught in the
intervention (d= .97). Elleman, Lindo, Morphy, and Compton
(2009), reanalyzing several of the same studies included in
Stahl and Fairbank’s review, found the eects of vocabulary
instruction on comprehension were far lower, with no signif-
icant impact on standardized measures (d= .10) and modest
but signicant eects on researcher-created tests (d= .50).
© 2018 Published by T& K Academic. This is an open access article under the CC BY- NC- ND license. (https://creativecommons.org/licenses/by/4.0/)
13 December 2018
12 February 2019
27 February 2019
March 2019, Volume 11, Issue 4, 309-318
Wright and Cervetti’s (2017) narrative review of vocabulary
instruction studies came to a conclusion similar to Elleman
et al.’s (2009). They found that words taught in certain in-
terventions were eective in helping students improve their
reading comprehension of a text containing those words,
but this eect did not generalize to other texts, such as
those found on standardized tests.
Incidental Vocabulary Acquisition via Free Reading
Even if vocabulary instruction can improve reading compre-
hension, it does not appear to be the main source of word
growth for school age children. Nagy and Anderson (1984)
observed that “even the most ruthlessly systematic direct
vocabulary instruction could [not] account for a signicant
proportion of all the words the children actually learn” (p.
304). Evidence for the impact of reading on vocabulary ac-
quisition comes from both experimental and correlational
studies. In “read-and-test” experiments (discussed further
below), subjects are given a text with unknown words in it
and asked to read it for comprehension. They are then given
a (usually surprise) test on the meanings of the new words.
These studies have found that a small but reliable amount
of knowledge is gained from even a single exposure to an
unknown word (Swanborn & de Glopper, 1999). Nagy, Her-
man, and Anderson (1985) estimated that with a sucient
amount of reading, a seemingly low “pick-up” rate could ac-
count for most of the observed growth in vocabulary knowl-
edge among school-age readers.
Additional experimental evidence comes from sustained si-
lent reading (SSR) and extensive reading studies, in which
students are encouraged to read books they select for
themselves. These studies have been conducted with chil-
dren and adults, for both rst and second language readers.
Krashen’s (2004a) narrative review of 54 studies concluded
that SSR and extensive reading treatments were as good as
or better than traditional language arts and reading instruc-
tion in promoting vocabulary and reading comprehension
gains. Two meta-analyses of SSR and extensive reading
studies have found signicant, medium-to-large eects for
free reading. Krashen (2007) examined studies for teens and
young adults and found a large eect (d= .87) on compre-
hension tests. Jeon and Day (2016) found medium eect siz-
es for both vocabulary (d= .47) and comprehension (d= .54)
for studies of both adults and children.1
Time Eciency in Instruction
Many vocabulary interventions are aimed at helping
low-performing students “catch up” to their age peers in
reading, presumably in the most time-ecient way possi-
ble. Carlo, August, McLaughlin, Snow, Dressler, Lippman,
Lively, and White (2004), for example, argued that “gaining
access to the information taught in middle and secondary
content area classes requires all students exit elementary
school with good reading comprehension,” and therefore
“closing this gap has a high priority if the U.S. education sys-
tem is to fulll its goad of reducing inequalities” (p. 188, 190,
emphasis added). Lawrence, Rolland, Branum-Martin, and
Snow (2014) claimed that “without proper intervention, low-
er-skilled students are likely to fall further behind their more
skilled peers in academic domains” (p. 77, emphasis added).
Faw and Waller (1976) noted that despite the presumed goal
of eciency, most educational intervention studies lack any
study or instructional time variable in their analyses. They
It is absurd to think that psychologists and educators can be
content with improving subjects' learning and retention of tex-
tual materials if the altered performance is simply a function
of augmented study time. This would be analogous to attrib-
uting the increased length of a skier's jump to superior coach-
ing when, in fact, the coach had simply provided a steeper and
longer hill from which the jump could be made. (p. 703)
Faw and Waller proposed that researchers distinguish be-
tween absolute performance levels and measures of ef-
ciency. Absolute performance measures look only at the
amount of learning that has taken place during the inter-
vention, such as gain scores from pre-test to post-test. An
eciency measure takes these absolute gains and divides
them by the study time of the intervention, to yield a gains-
per-time estimate. It is then possible to calculate an ecien-
cy index to compare the two approaches by dividing the
eciency score of the experimental group by the eciency
score of the control group. Faw and Waller point out that
methods that produce greater absolute gains may in fact be
less ecient than the alternatives.
Only handful of studies on vocabulary acquisition have
applied the principles laid out by Faw and Waller. Krash-
en (1989) re-analyzed several vocabulary-teaching studies
for word learning eciency. Several studies by Mason and
colleagues also reported the number of words per minute
acquired in a second language classrooms in order to com-
pare the eciency of traditional instruction with compre-
hension-based language teaching methods (Mason, 2007;
Mason & Krashen, 2004; Mason, Vanata, Jander, Borsch, &
My analysis of the relative eciency of direct instruction
and free reading on word knowledge growth is organized
around three questions:
1. What is the average eciency of explicit vocabu-
lary teaching for school-age students as measured in
words learned per minute of instruction?
2. What is the average eciency of free reading in
words acquired per minute of reading, based on
previous studies of incidental word acquisition rates,
the percentage of unknown words in text, and aver-
age reading rates for students?
3. What is the relative eciency of direct instruction
compared to free reading, as measured by an e-
The Eciency of Direct Instruction
The most recent comprehensive review on vocabulary in-
struction is Wright and Cervetti (2017), who reviewed the
results of 36 studies on the eects of vocabulary instruction
on word learning and reading comprehension. They began
with the studies selected by two previous meta-analyses
of vocabulary teaching interventions (Elleman et al., 2009;
Stahl & Fairbanks, 1986), supplementing their pool of stud-
ies with those published after the Ellman et al. review. Their
inclusion criteria diered somewhat from previous reviews.
To be selected for the review, studies had to include PreK-
12 students, a treatment involving the direct instruction of
words, the teaching of “word-solving” strategies, or both,
and a passage comprehension dependent measure.
Their analysis reported estimates of the time spent on vo-
cabulary instruction in each study. These instructional times,
however, are for the amount of time devoted to each word
taught. For an eciency measure, we need the time spent
per word learned (Faw & Waller, 1976). In this analysis, I took
the number of words learned in the intervention divided by
the total instructional time (in minutes) of the vocabulary in-
struction, similar to Krashen (1989) and Mason (2007). For
example, if 100 minutes were spent on instruction, and the
gain score for vocabulary was 5 words, the eciency esti-
mate would be .05 words per minute (wpm) (5 words/100
minutes). An eciency score is calculated for each study.
The Ineciency of Vocabulary Instruction / McQuillan
Gain scores for studies without comparison or control
groups were calculated by subtracting the pre-test vocab-
ulary scores from the post-test scores. Studies that had a
control group but no information on whether the controls
also received vocabulary instruction (“business-as-usual”
or “typical practice” groups) were treated the same as
those with no control groups. If a study had a reading-on-
ly comparison group, the gain scores of the comparison
group were subtracted from gain scores of the treatment
group, resulting in a “net” gain estimate.
Of the 36 studies included in the Wright and Cervet-
ti review, nine studies included interventions aimed at
teaching word learning strategies only, not a specic set
of target words. Of the remaining group of 27 studies, I
excluded 13 studies that failed to meet one of the follow-
ing four criteria: (a) included a measure of target word
knowledge due to the intervention; (b) included a pre-
test or a no-treatment comparison group to control for
pre-treatment knowledge of the target words; (c) provided
sucient data on instructional time to calculate eciency;
and/or (d) included subjects who would likely be able to
read independently (grade 2 or older). This left 14 studies
with sucient data to calculate an eciency estimate, list-
ed in Tables 1 and 2.2
In cases where there was more than one experimen-
tal group and signicant dierences were found among
them, the explicit instruction group that had the highest
vocabulary gain scores was used in the calculations in
order to provide the “best-case scenario” for explicit in-
struction. When there were multiple treatments and no
signicant dierences found in their gain scores, I took the
average gain for all the experimental conditions. Wright
and Cervetti categorized the studies by the length of the
intervention, with a “brief” intervention lasting four weeks
or less, and a “long-term” intervention lasting more than
four weeks (p. 209). Table 1 summarizes the data used to
calculate time eciency from the nine brief interventions.
The data from the ve long-term studies are found in Ta-
I used the “per word” instructional times provided by
Wright and Cervetti (2017) in six of the 14 studies, taken
from their Tables 2 and 4 (pp. 9, 15). For the other eight
experiments, I used a dierent estimate, justied below.
In all cases, my revised estimates were lower than those
given by Wright and Cervetti. Total words learned was cal-
culated by multiplying the raw score increase, pre- to post-
test, by the quotient of total words divided by total words
tested. For example, in Leseaux et al. (2010), there were 72
words taught but only 36 words tested. The raw score was
multiplied by two (72/36) to yield the total words gained.
Bos & Anders (1990). Bos and Anders report that each in-
tervention consisted of six 50-minute sessions over a peri-
od of seven weeks: three “practice” sessions and three ex-
perimental sections. Although Wright and Cervetti (2017)
count both the practice and the experimental sessions for
their time estimates, I have used only the experimental
ones, for a total of 110 minutes of instruction across the
Greene Brabham and Lynch-Brown (2002). Wright and
Cervetti (2017) provided an estimate of 180 minutes for
the experimental treatments (4.5 minutes per word taught
for the 40 words). Greene Brabham and Lynch-Brown
reported that the highest scoring group, the interactive
group, spent 25 minutes on each of two stories. Since the
stories were read three times, I used an estimate of 150
minutes of instruction. I excluded their rst-grade subjects
since it wasn’t clear they were able to read independently.
Hawkins, Musti-Rao, Hale, McGuire, and Hailley (2010).
Hawkins and colleagues studied vocabulary acquisition
using a within-subjects, post-test-only design with a group
of fourth-grade students. The reading-only condition read
three 400-word stories. Since no treatment time was re-
ported for this condition, I began with the average 4th
grade silent reading rate, which is estimated to be around
150 wpm (Carver, 1989; Spichtig, Hiebert, Vorstius, C., Pas-
coe, J., Pearson, P. D., & Radach, R., 2016). However, be-
cause students knew they were going to be quizzed on the
content of the passages, I lowered the reading rate to 100
wpm, as students who know they are to be tested tend
to read more slowly (Carver, 1990). I estimated four min-
utes was spent by the controls on reading the story (400
The listening + vocabulary instruction condition produced
the greatest absolute number of words gained. For the
listening part of the treatment, I doubled the estimate
of the reading-only condition (8 minutes), since students
were asked to repeat each sentence after it was read by
the teacher. For the added pre-reading vocabulary instruc-
tion, I estimated one minute per target word, which is what
Coyne et al. (2009) used as a time estimate for giving word
denitions in storybook reading treatments. Thus the total
listening + vocabulary practice treatment time was 18 min-
utes (10 minutes instruction plus 8 minutes reading and
repeating). Wright and Cervetti’s estimate was “less than
one minute” on each of the 30 target words.
Pany, Jenkins, and Schreck (1982) (Study 1). Only a range
of per-word instructional times (two to ten minutes) was
provided by Wright and Cervetti. Students saw two sets of
target words in each condition. The highest scoring con-
dition was “meanings practiced,” which spent 6.5 minutes
per set of words, for a total of 13 minutes.
Seifert and Espin (2012). Wright and Cervetti estimated that
the researchers spent 12 minutes on each target word
taught, for a total of 120 minutes. But it would appear
from the procedures section of the study (p. 241) that
students spent 30 minutes in each condition for each set
of 10 target words, so the total time by condition was 30
minutes, not 120.
Tuinman and Brady (1974). Wright and Cervetti estimated
10 minutes per word taught (for 660 minutes), although
the time reported for the treatment by the researchers
was 585 minutes (14 instructional sessions of 45 minutes
each, p. 179), so this lower estimate was used.
Beck, Perfetti, & McKeown (1982); McKeown, Beck, Oman-
son, and Perfetti (1983). An estimate of 15 minutes per
word was given by Wright and Cervetti for both of these
studies, for a total of 2,288 minutes for 104 target words.
I used a slightly lower estimate of 2,250 minutes based
upon Beck et al. description’s of the intervention as con-
sisting of 75 30-minute lessons.
Summary of Direct Instruction Studies
Tables 1 and 2 summarize the eciency scores in words
learned per minute for the 14 direct instruction studies
taken from Wright and Cervetti’s (2017) review. Also list-
ed are the grade level, sample size, treatment duration (in
minutes), number of words learned, and type of vocabu-
lary test (meaning recall or meaning recognition) for each
experiment. For short interventions, the average number
of words learned per minute was .07. For long-term inter-
ventions, the average number of words gained per minute
was much smaller, at .01. There was a large standard de-
viation for the short-term studies, indicating considerable
variability in eciency scores.
March 2019, Volume 11, Issue 4, 309-318
Incidental Word Acquisition from Reading
Several studies of K-12 and adult readers have measured
the amount of vocabulary gained incidentally from reading.
As noted above, in these “read-and-test” experimental stud-
ies, subjects are given texts containing unknown words and
told to read the texts for comprehension. They are then giv-
en a surprise vocabulary test on the unknown words, either
immediately or after some delay. We can use these data to
estimate the number of unknown words a typical reader
might acquire through reading for pleasure, given a certain
percentage of unknown words in the text. Combined with
data on reading rate, we can then estimate the number of
words per minute that could have been gained through free
reading for each of the 14 direct instruction studies.
Table 1. Eciency of Direct Instruction of Vocabulary in Long-
Bos & Anders
(Recog.) .07 wpm
(2002) (Grade 3)
(Recog.) .08 wpm
Hawkins et al.
(Recall) .15 wpm
McKeown et al.
(Recog.) .04 wpm
Nash & Snowling
(Recog.) .014 wpm
Pany et al. (1982)
(Recall) .07 wpm
Seifert & Espin
(Recog.) .10 wpm
Stahl (1983) 5
(Recall) .05 wpm
Tuinman & Brady
4, 5, 6
(Recog.) .02 wpm
Average .07 wpm
(Std Dev) (.04)
Recog.= Meaning recognition test.
Table 2. Eciency of Direct Instruction of Vocabulary in Long-
Beck et al. (1982) 4
(Recog.) .02 wpm
Lesaux et al.
(Recog.) .004 wpm
Lesaux et al.
(Recog.) .002 wpm
McKeown et al.
(Recog.) .02 wpm
Simmons et al.
(Recog.) .01 wpm
Average .01 wpm
(Std Dev) (.009)
Recog.= Meaning recognition test.
I have included in Table 3 the studies from Swanborn and
de Glopper’s (1999) meta-analysis on incidental word ac-
quisition among school-age readers. I excluded four un-
published dissertations used by Swanborn and de Glopper,
but added one published study not in their review (Herman,
Anderson, Pearson, & Nagy, 1997). I also excluded studies
in which the researchers deliberately choose or manipulat-
ed the contexts around the target words in order to make
them all either “informative” or “uninformative,” since nei-
ther extreme is representative of natural texts. Only studies
in which words were chosen solely on the basis of whether
they were likely to be unknown to the reader regardless of
context were used.
The studies include readers at every reading ability level,
including less-able readers.3 I report the results by reading
achievement level for those studies that provided a break-
down, taken in part from Swanborn and de Glopper’s me-
ta-analysis (Table 3, p. 273).4 Table 3 lists grade levels tested,
the number of subjects, and the subjects’ reading levels for
each study. The nal column of Table 3 reports the proba-
bility of acquiring an unknown word from a single exposure.
Nagy et al. (1985) denes this probability as “the increase in
the number of words divided by the number of words orig-
inally not known” (p. 248). In some studies, students were
tested on both the target words that appeared in their as-
signed text and on words that appeared in a text they did
not read. This was done instead of a pretest to control for
prior knowledge of the target words. For these studies, I
used Nagy et al.’s (1987) formula to calculate probability:
(Proportion of Target Words Correct – Proportion of Control
Words Correct) / (1 - Proportion of Control Words Correct)
In those studies where a pretest was used instead of control
words, I followed a similar formula, subtracting the propor-
tion of correct pretest words from the proportion of correct
post-test words, and then dividing that result by the propor-
tion of incorrect pretest words.
Table 3. Incidental Acquisition Pickup Rates in 12 Studies
Herman, Anderson, Pearson,
& Nagy (1987)
(81st to 99th Percentile)
(N= 413) H0.26
Herman, Anderson, Pearson,
& Nagy (1987)
(31st to 80th Percentile)
Herman, Anderson, Pearson,
& Nagy (1987)
(3rd to 30th Percentile)
8 L 0.05
Nagy, Anderson, & Herman
3, 5, 7
(N= 352) H, A, L 0.05
Nagy, Herman, & Anderson
(N= 57) H, A 0.11
Schwanenugel, Stahl, &
(N= 43) H, A, L 0.08
Stahl (1989) 6
(N= 182) H, A, L 0.13
Shu, Anderson, & Zhang
(1995) (English experiment)
(N= 170) H, A, L 0.10
a Subjects’ Reading Level: H = High, A = Average, L = Low, taken in part from Swanborn &
de Glopper (1999). Scores from both meaning recall and meaning recognition measures
were averaged to calculate probability if data on both were provided.
The probability of acquiring an unknown word incidentally
through reading ranged from .05 to .26. In their meta-analy-
sis, Swanborn and de Glopper (1999) calculated the average
probability of acquisition to be .15 for the 15 experiments
The Ineciency of Vocabulary Instruction / McQuillan
they included. The average probability of acquisition for
the eight experiments in Table 3 is slightly lower, at .11.
What is the appropriate acquisition rate for comparing
free reading to direct instruction? Since the goal of direct
instruction is often to help low-achieving students, the
most conservative approach is to use one of the lower
estimates. In Table 3, we nd that the lowest probability
estimate is .05. This is the gure used for the eciency
scores calculations below
Percentage of Unknown Words in Text
The number of words that a reader can acquire inciden-
tally from reading depends in part on how many unknown
words are present in the text. Anderson and Freebody
(unpublished, reported in Nagy et al., 1985) estimated the
number of unknown words likely to be encountered by a
“50th percentile fth-grader” in text is between three and
six percent, depending on the criteria used for “knowing”
a word (p. 250). The researchers did not specify the source
of the texts analyzed. Stahl (1990) stated that “a reader
typically encounters between one and a half and three un-
known words per hundred running words” (p. 6), but he
gave no source for his estimate.
Carver’s (1994) attempted to determine the percentage of
unknown words by asking third through sixth grade stu-
dents to circle words they did not know in a set of passag-
es. Students rst were tested to determine their current
reading level, and then given passages to read that were
below, at, or above their grade level. Passages were taken
from both textbooks and library or trade books (p. 416).
Carver noted that all of the percentages are likely to be un-
derestimates, however. A large number (40%) of his initial
sample failed to underline three embedded low-frequency
words and had to be excluded from the study, suggesting
that students had a tendency to under-identify unknown
words in the texts.
Since Carver’s (1994) estimates are the best documented,
I have used for my calculations the average number of un-
known words he found for library books, from two grade
levels above grade level (3.35%), at grade level (1.30%),
and two grade levels below reading level (1.35%), giving
us an estimate of 2%. This number falls at the lower end
of the range given by Stahl (1990), and slightly below the
low end of Anderson and Freebody’s (cited in Nagy et al.,
Eciency calculations for incidental vocabulary acquisi-
tion depend in part on the reader’s reading rate. The most
recent large-scale study of reading rates was a partial rep-
lication of Taylor (1965) by Spichtig et al. (2016). Like Tay-
lor, Spichtig and colleagues measured reading rate along
with comprehension and eye movements for a large sam-
ple (N= 2,203) of students, but limited their study to grades
2, 4, 5, 8, 10, and 12. Unlike Taylor’s study, the researchers
attempted to stratify their sample to reect the current
demographics of U.S. schools, but it was not a random
Table 4 (column 2) lists the mean reading rates, controlling
for comprehension, reported in Spichtig et al. The re-
searchers also reported rates by quartile, so I have taken
the average of the bottom two quartiles in order to provide
an approximate “below average” or “slow” reading rate at
each level (column 3). Since again vocabulary instruction
is often advocated especially for less-able readers, I will
use these lower rates in making eciency estimates when
comparing acquisition rates to direct instruction studies in
the following section, even when it appears the actual sub-
jects in the study were average or above average readers.
Free Reading Eciency Scores
Having established a rate of acquisition (.05), a percentage
of unknown words typically found in text (2%), and aver-
age and low-achieving reading rates for school-age chil-
dren, we can now estimate the number of unknown words
that would likely be acquired from free reading. Table 4
(columns 4 and 5) reports estimates, at various reading
rates, of the number of words per minute likely to be ac-
quired from reading under these assumptions
Table 4. Estimated Incidental Word Acquisition from Free
Reading at Average and Slow Reading Rates
(41.7) 87 .12 wpm .09 wpm
(45.4) 115 .15 wpm .12 wpm
(54.4) 128 .16 wpm .13 wpm
(51.8) 130 .17 wpm .13 wpm
(53.4) 147 .19 wpm .15 wpm
(55.5) 181 .19 wpm .18 wpm
Measured in words per minute of reading, students be-
come more ecient in word acquisition as they age,
although the trend is not perfectly linear due to the
plateauing of reading rates between grades 4 and 8. Stu-
dents reading at an average rate for their grade level will
acquire around .12 words per minute in grade 2, rising
to .19 words per minute by grade 10. For low-achieving
students, the eciency of incidental acquisition goes from
.09 words per minute at grade 2, up to .18 words per min-
ute in grade 12.
Eciency Indexes for Direct Instruction and Free Reading
Having calculated the eciency scores for both direct in-
struction and free reading, we can now provide eciency
indexes to compare the two approaches for each of the
direct instruction studies. Table 5 shows the eciency
scores estimates of direct instruction conditions and our
hypothetical reading-only conditions. I took the estimated
number of words gained and words per minute for each
study in Cervetti and Wright (2017) as reported in Tables
1 and 2 above. For the incidental acquisition estimates,
I used the “slow” reading speed for that grade level as
found in Table 5. For studies that included odd-numbered
grades, I used the estimate from the even-numbered
grade below it (e.g. for grade 5, grade 4 reading speeds
were used.). Shown also in Table 5 is the estimated num-
ber of words one could acquire from reading (an absolute
measure), to compare to those gained in the direct in-
struction experiment (Table 5, column 4, “Reading Words
As was done in Faw and Waller (1996), the eciency in-
dex for a study was calculated by dividing the eciency
score of experimental condition (direct instruction) by the
eciency score of the control condition (reading-only). An
eciency index score of 1.0 means the two approaches
were equally ecient, a number smaller than 1.0 indicates
incidental acquisition was more ecient, and a number
March 2019, Volume 11, Issue 4, 309-318
greater than 1.0 means direct instruction was more ecient.
Results are reported in Table 5, as in Tables 1 and 2 above,
by Wright and Cervetti’s classication of “short-term” and
Short-term direct instruction studies had an average rate of
word learning of .07 wpm. The average rate of word acqui-
sition for free reading was .12 wpm, resulting in an average
eciency index of .56, favorable to free reading. This dier-
ence is statistically signicant (t(16)= 3.36, p< .01), yielding a
large eect size (d= 1.58). Put another way, reading would be
on average 1.7 times (.12 wpm/.07 wpm) more ecient than
direct instruction in word acquisition. Direct instruction was
found more ecient than reading in only one of the nine
For long-term studies, the results favor incidental acquisi-
tion to an even greater degree. The average eciency of
long-term direct instruction treatments was .01 wpm, com-
pared to .12 wpm for incidental acquisition, and the aver-
age eciency index was .10. As with the short-terms stud-
ies, this dierence is statistically signicant (t(8)= 11.83, p<
.0001), with a very large eect size (d= 7.48). This means free
reading would be about 12 times (.12 wpm/.01 wpm) more
ecient than direct instruction in helping children acquire
new vocabulary in long-term treatments.
There was a moderate negative correlation between the
time devoted to instruction and the eciency estimate (r=
-.67, p<.01), meaning that the more time teachers spent on
vocabulary instruction, the fewer number of words per min-
ute their students learned.
Our results indicate that neither short-term nor long-term
instruction is ecient in teaching new words compared
to just reading. Students in short-term direct instruction
treatments learned about four words per hour (.07 wpm),
compared to our estimate of around seven words per hour
(.12 wpm) via free reading. For long-term treatments, stu-
dents learn only about a .5 words per hour of instruction
Greater investments of time into vocabulary instruction ap-
pear to have diminishing returns in terms of the number of
words students learn. It is clear from the estimates in Table
5 that long-term interventions were considerably less e-
cient than short-term ones. The least ecient instruction
was found in those studies (Leseaux et al., 2010; 2014) that
focused on teaching words from the Academic Word List.
Students learned only one new AWL word every ve and a
half hours or so of instruction. At this rate, students would
need to spend roughly 1,600 hours of instruction to learn
just half of the 570 AWL terms, a feat that would take a dec-
ade or more of language arts classes devoted to nothing but
Hypothetical Comparison Groups. Since nearly all of the direct
instruction studies we examined lacked a reading-only con-
dition, we do not have a set of “head-to-head” comparisons
of incidental acquisition versus direct instruction. But there
is no reason to think the subjects in the studies reviewed
here, had they been given the opportunity to read, would
not also have acquired vocabulary at rates found in previous
incidental acquisition studies. Condence in our ndings is
bolstered by the results of two reviews of studies that did
include direct comparisons between a reading-only and a
reading plus explicit instruction condition. McQuillan (2016a)
reviewed eight such experiments with adult second-lan-
guage acquirers, and found that reading-only conditions
had the same or greater eciency than direct instruction in
six experiments, and lower eciency in two treatments. Mc-
Quillan (2019) looked at ve studies that compared direct in-
struction and reading-only conditions in storybook reading
for young children. The reading-only conditions were found
on average to be 66% more ecient than direct instruction
conditions for acquiring new words.
Overly Optimistic Estimates. While I have used conservative
estimates consistent with previous studies of incidental ac-
quisition rates, percentage of unknown words, and reading
Table 5. Eciency of Direct Instruction versus Incidental Acquisition in 14 Studies
Study Duration DI Words
Score E. Indexb
Bos & Anders (1990) 110 7.86 14.3 .07 wpm .13 wpm 0.54
Greene Brabham & Lynch-Brown (2002) 150 11.9 13.5 .08 wpm .09 wpm 0.89
Hawkins et al. (2010) 18 2.67 2.16 .15 wpm .12 wpm 1.25
McKeown et al. (1985) 360 13.17 43.2 .04 wpm .12 wpm 0.33
Nash & Snowling (2006) 360 5.14 32.4 .01 wpm .09 wpm 0.16
Pany et al. (1982) (Study 1) 13 1.42 1.56 .07 wpm .12 wpm 0.58
Seifert & Espin (2012) 30 3.1 4.5 .10 wpm .15 wpm 0.67
Stahl (1983) 75 8.32 9 .05 wpm .12 wpm 0.42
Tuinman & Brady (1974) 585 12.18 70.2 .02 wpm .12 wpm 0.17
Average (SD) .07 wpm (.04) .12 wpm (.019) 0.56 (.35)
Beck et al. (1982) 2250 52.75 270 .02 wpm .12 wpm 0.17
Lesaux et al. (2010) 3240 12.08 421.2 .004 wpm .13 wpm 0.03
Lesaux et al. (2014) 4095 9.62 573.3 .002 wpm .14 wpm 0.01
McKeown et al. (1983) 2250 48.59 270 .02 wpm .12 wpm 0.17
Simmons et al. (2010) 1620 21.98 145.8 .01 wpm .09 wpm 0.11
Average (Standard Deviation) .01 wpm (.009) .12 wpm (.019) 0.10 (.08)
a DI = direct instruction. b E. Index = eciency index.
The Ineciency of Vocabulary Instruction / McQuillan
rate, some might argue they should be even lower. To
provide an even stricter test for my assumptions, I ran a
separate analysis of the eciency data in Table 5 in which
I halved the incidental reading acquisition eciency esti-
mates in all the comparisons, the equivalent of lowering
the estimate of unknown words in a text to 1% or the
probability of acquisition to .025. Even under these very
pessimistic assumptions for incidental acquisition, free
reading was still signicantly more ecient than direct in-
struction for the long-term studies (t(8)= 8.61, p< .0001).
For the short-term studies, the revised average ecien-
cy estimate for free reading was .06 wpm, which was not
signicantly dierent from the direct instruction estimate
of .07 wpm (t(16)= 0.46, p= .665). Free reading, then, was
as good as or better than direct instruction in promoting
word growth regardless of the study length.
In any such “tie” between free reading and direct instruc-
tion, the advantage clearly belongs to free reading, since
reading is less work for the teacher and more enjoyable
for the student. In addition, free reading has important
benets in addition to vocabulary, such as improving
reading comprehension, writing, and grammar (Krashen,
2004a). Nagy et al. (1985) also make this point, noting that
“[a]ny comparison of approaches ought to take account of
the fact that time spent in reading has more benets than
just growth in vocabulary…no doubt the ancillary benets
of vocabulary instruction are less rich” (p. 251).
An argument can be made that the probability estimates
such as those provided by the read-and-test studies
should not be applied in the way done in this analysis. Our
analysis assumes that acquisition is incremental, meaning
that we pick up a small percentage of an unknown word’s
meaning each time we see it in a text. As Stephen Krashen
(personal communication) has pointed out to me, read-
and-test researchers such as Nagy et al. (1985) seem to
assume that a .05 probability means that out of 100 un-
known words in a text, a reader acquires the full meaning
of ve of them. This is not the same as saying that you pick
up partial meanings (say, 5%) of 100 words. If the former
interpretation accurately represents these researchers’
reasoning, then multiplying a probability by the number of
word occurrences may in fact be inappropriate and over-
estimate vocabulary gains.
Still, our nding that free reading is superior to direct in-
struction for word acquisition is consistent with analyses
that used a very dierent approach to the problem, that
of corpus analysis. Nation (2014), for example, looked at a
large corpus of classic novels to determine how much one
would have to read to have a reasonable chance of acquir-
ing new words. Potentially unknown words that occurred
12 or more times in the text were considered “acquired”
in Nation’s study. For texts written with 98% vocabulary
coverage in the 3,000- through 8,000-word-family levels
(the levels at which popular, young adult ction is written
(McQuillan, 2016b)), Nation estimated that on average 6.9
words would be acquired per hour, or 0.12 wpm. This is
identical to our overall incidental acquisition estimate for
school-age readers. Similarly, McQuillan (in press) exam-
ined a corpus of 1.2 million words from young adult popu-
lar ction and, using a similar “cut-o” method as used by
Nation, found that reading was between two and six times
more ecient than direct instruction in acquiring academ-
Short-Term Benets of Instruction. Since Wright and Cervetti
(2017) concluded that “pre-teaching” vocabulary appears
in a text improves comprehension of that text, it could be
argued that direct instruction interventions have real ben-
ets when used in this short-term, “text-by-text” approach.
The weakness of this argument is that such teaching is by
denition only a temporary x, akin to bailing out a sinking
boat with a bucket instead of xing the hole at the bottom.
Stahl (1990) made a similar observation, noting that “[m]
ore intensive instruction is going to take away time from
other activities, including wide reading that will not only
better allow them to solidify their vocabulary gains but
also will itself lead to greater vocabulary growth” (p. 11).
Allington, McCuiston, and Billen (2015) argued that stu-
dents need to read texts they can understand inde-
pendently for real progress in reading to occur, prefera-
bly on topics that are compelling and comprehensible to
them (Krashen et al., 2018). One solution, proposed by
Krashen (2004a) and others, is to give students the oppor-
tunity to read extensively by providing a large number of
interesting and comprehensible books in the school and
classroom library, time to read daily, and a comfortable
environment. Such programs have been found to be as
good as or better than traditional instruction in promot-
ing both vocabulary growth and reading comprehension
(Krashen, 2004a; Krashen & Mason, 2017).
Free Reading as a Bridge to Academic English
I do not claim that free reading alone can give students
100% of the vocabulary or academic language needed for
success in school. Some explicit teaching of terms related
to new concepts, for example, may be required. More im-
portantly, there are characteristics of academic language
that are only found in academic texts (Biber, 1985), and
therefore can only be acquired through academic reading
(Krashen, 2010). Pleasure reading can, however, provide
an important “bridge” to more challenging school read-
ing, including sub-technical vocabulary. Rolls and Rogers
(2017) analyzed a large corpus of science ction and fan-
tasy literature for the presence of sub-technical vocabu-
lary specic to the sciences, based on Coxhead and Hirsh’s
(2007) list of 318 word families (e.g. degrade, module, up-
take). They found that nearly all of the words (92%) oc-
curred at least once in a corpus of one million words, and
majority of those words occurred six times of more, giving
students a good chance to acquire them.
Krashen (2012a; 2012b) advocates a two-stage approach
for helping students advance in both academic vocabulary
and content knowledge via free reading. Stage 1 consists
of “massive, but not necessarily wide, self-selected vol-
untary reading” (Krashen, 2012b, p. 9). Reading done at
this stage builds general vocabulary and knowledge of
the world that will make academic reading more compre-
hensible. Ideally, students read narrowly in order to take
advantage of prior knowledge of a topic or book series
(Krashen, 2004b; Kyungho & Nation, 1989; Schmitt & Cart-
Since general reading will not give students all of the ac-
ademic language they need, Krashen proposes a second
stage called narrow academic reading. This consists of
students reading about an academic topic that they them-
selves are interested in. This sort of reading will give stu-
dents knowledge of academic conventions and language
(such as the words on the AWL) that in turn will help
them across disciplines. Krashen (2012b) gives his own
case history of narrow reading in linguistics and medicine
that gave him sucient knowledge of the academic lan-
guage register to read scholarship in other elds. Indeed,
it would seem that nearly all of us acquired academic
language in this way, and not through direct instruction
(Krashen, 2012a). Providing low-achieving students with
an opportunity to follow that same path should at least
Allington, R. (2014). How reading volume aects both
reading uency and reading achievement. Inter-
national Electronic Journal of Elementary Education,
March 2019, Volume 11, Issue 4, 309-318
Allington, R., McCuiston, K., & Billen, M. (2015). What the re-
search says about text complexity and learning to
read. The Reading Teacher, 68(7), 491-501.
Anderson, R., & Freebody, P. (1981). Vocabulary knowledge.
In J.T. Guthrie (Ed.), Comprehension and teaching: Re-
searcher Reviews (pp. 77-117). Newark, DE: Interna-
tional Reading Association.
Apthorp, H. (2006). Eects of a supplemental vocabulary
program on third-grade reading/language arts. The
Journal of Educational Research, 100(2), 67-79.
Apthorp, H., Randel, B., Cherasaro, T., Clark, T., McKeown,
M., & Beck, I. (2012). Eects of a supplemental vo-
cabulary program on word knowledge and passage
comprehension. Journal of Research on Educational
Eectiveness, 5(2), 160-188.
Beck, I., McKeown, M., & Omanson, R. (1987). The eects and
uses of diverse vocabulary instructional techniques.
In M. McKeown & M. Curtis (Eds.), The nature of vo-
cabulary acquisition (pp. 147-163). Hillsdale, NJ: Erl-
Beck, I., Perfetti, C., & McKeown, M. (1982). Eects of long-
term vocabulary instruction on lexical access and
reading comprehension. Journal of Educational Psy-
chology, 74(4), 506-521.
Biber, D. (1985). Investigating macroscopic textual variation
through multi-feature/multidimensional analyses.
Linguistics, 23(2), 337-360.
Biemiller, A., & Boote, C. (2006). An eective method for
building meaning vocabulary in primary grades.
Journal of Educational Psychology, 98(1), 44-62.
Bos, C., & Anders, P. (1990). Eects of vocabulary instruction
on the vocabulary learning and reading comprehen-
sion of junior-high learning disabled students. Learn-
ing Disability Quarterly, 13(1), 31-42.
Carlo, M., August, D., McLaughlin, B., Snow, C., Dressler, C.,
Lippman, D, Lively, T., & White, C. (2004). Closing the
gap: Addressing the vocabulary needs of English-lan-
guage learners in bilingual and mainstream class-
rooms. Reading Research Quarterly, 39(2), 188-215.
Carver, R. (1989). Silent reading rates in grade equivalents.
Journal of Reading Behavior, 21(2), 155-166.
Carver, R. (1990). Reading rate: A review of research and theo-
ry. New York: Academic Press.
Carver, R. (1994). Percentage of unknown vocabulary words
in text as a function of the relative diculty of the
text: Implications for instruction. Journal of Reading
Behavior, 26(4), 413-437.
Cho, K.S., & Krashen, S. (1994). Acquisition of vocabulary
from Sweet Valley Kids series: Adult ESL acquisition.
Journal of Reading, 37(8), 662-667.
Cowan, J. (1974). Lexical and syntactic research for the de-
sign of EFL reading materials. TESOL Quarterly, 8(4),
Coxhead, A. (2000). A new academic word list. TESOL Quar-
terly, 34(2), 213-238.
Coxhead, A., & Hirsch, D. (2007). A pilot science-specic word
list. Revue Française de Linguistique Appliquée, 12(2),
Coyne, M., McCoach, D., Loftus, S., Zipoli Jr, R., & Kapp, S.
(2009). Direct vocabulary instruction in kindergarten:
Teaching for breadth versus depth. The Elementary
School Journal, 110(1), 1-18.
Elleman, A., Lindo, E., Morphy, P., & Compton, D. (2009). The
impact of vocabulary instruction on passage-level
comprehension of school-age children: A meta-anal-
ysis. Journal of Research on Educational Eectiveness,
Faw, H., & Waller, T., (1976). Mathemagenic behaviors and
eciency in learning from prose. Review of Educa-
tional Research, 46, 691-720.
Greene Babham, E., & Lynch-Brown, C. (2002). Eects of
teachers’ reading-aloud styles on vocabulary acqui-
sition and comprehension of students in the early
elementary grades. Journal of Educational Psychology,
Hawkins, R., Musti-Rao, S., Hale, A., McGuire, S., & Hailley,
J. (2010). Examining listening previewing as a class-
wide strategy to promote reading comprehension
and vocabulary. Psychology in the Schools, 47(9), 903-
Herman, P., Anderson, R., Pearson, P., & Nagy, W. (1987).
Incidental acquisition of word meaning from expo-
sitions with varied text features. Reading Research
Quarterly, 22(3), 263-284.
Jeon, E., & Day, R. (2016). The eectiveness of ER on read-
ing prociency: A meta-analysis. Reading in a Foreign
Language, 28(2), 246-265.
Krashen, S. (1989). We acquire vocabulary and spelling by
reading: Additional evidence for the input hypothe-
sis. The Modern Language Journal, 73(4), 440-464.
Krashen, S. (2001). More smoke and mirrors: A critique of he
National Reading Panel report on uency. Phi Delta
Kappan, 83, 119-123.
Krashen, S. (2004a). The power of reading. 2nd edition. Ports-
mouth, NH: Heinemann.
Krashen, S. (2004b). The case for narrow reading. Language
Magazine, 3, 17–19.
Krashen, S. (2005). Is in-school free reading good for chil-
dren? Why the National Reading Panel Report is
(still) wrong. Phi Delta Kappan, 86, 444-447.
Krashen, S. (2007). Extensive reading in English as a foreign
language by adolescents and young adults: A me-
ta-analysis. International Journal of Foreign Language
Teaching, 3(2), 23-29.
Krashen, S. (2010). Academic language prociency: Acquired
or learned? In Selected papers from the Nineteenth
International Symposium on English Teaching (pp. 34-
43). Taipei: Crane.
Krashen, S. (2012a). Direct instruction of academic vocab-
ulary: What about real reading? Reading Research
Quarterly, 47(3), 223-224.
Krashen, S. (2012b). Developing academic language: Some
hypotheses. International Journal of Foreign Language
Teaching, 7(2), 8-15.
Krashen, S. & Mason, B. (2017). Sustained silent reading in
foreign language education: An update. Turkish On-
line Journal of English Language Teaching (TOJELT), 2(2),
The Ineciency of Vocabulary Instruction / McQuillan
Kyongho, H., & Nation, P. (1989). Reducing the vocabu-
lary load and encouraging vocabulary learning
through reading newspapers. Reading in a Foreign
Language, 6(1), 323-335.
Lawence, J., White, C., & Snow, C. (2010). The words stu-
dents need. Educational Leadership, 68(2), 22-26.
Lawrence, J., Rolland, R., Branum-Martin, L., & Snow, C.
(2014). Generating vocabulary knowledge for at-
risk middle school readers: Contrasting program
eects and growth trajectories. Journal of Educa-
tion for Students Placed at Risk, 19, 76-97.
Lesaux, N., Kieer, M., Faller, S., & Kelley, J. (2010). The
eectiveness and ease of implementation of an
academic vocabulary intervention for linguistically
diverse students in urban middle schools. Reading
Research Quarterly, 45(2), 196-228.
Lesaux, N., Kieer, M., Kelley, J., & Harris, J. (2014). Eects
of academic vocabulary instruction for linguisti-
cally diverse adolescents: Evidence from a rand-
omized eld trial. American Educational Research
Journal, 51(6), 1159-1194.
Mason, B. (2007). The eciency of self-selected reading
and hearing stories on adult second language ac-
quisition. In Selected papers from the sixteenth In-
ternational Symposium on English Teaching (pp. 630-
633). Taipei: Crane Publishing Company.
Mason, B., & Krashen, S. (2004). Is form-focused vocab-
ulary instruction worthwhile? RELC Journal, 35(2),
Mason, B., Vanata, M., Jander, K., Borsch, R., & Krashen, S.
(2008). The eects and eciency of hearing stories
on vocabulary acquisition by students of German
as a second foreign language in Japan. Indonesian
Journal of English Language Teaching, 5(1), 1-14.
McKeown, M., Beck, I., Omanson, R., & Perfetti, C. (1983).
The eects of long-term vocabulary instruction on
reading comprehension: A replication. Journal of
Reading Behavior, 15(1), 3-18.
McKeown, M., Beck, I., Omanson, R., & Pople, M. (1985).
Some eects of the nature and frequency of vo-
cabulary instruction on the knowledge and use of
words. Reading Research Quarterly, 20(5), 522-535.
McQuillan, J. (2016a). Time, texts, and teaching in vocabu-
lary acquisition: A rebuttal to Cobb (2016). Reading
in a Foreign Language, 28(2), 307-318.
McQuillan, J. (2016b). What can readers read after graded
readers? Reading in a Foreign Language, 28(1), 63-
McQuillan, J. (2019). We don’t need no stinkin’ exercises:
The impact of extended instruction and storybook
reading on vocabulary acquisition. Language and
Language Teaching, 8(1), 25-37.
McQuillan, J. (in press). Where do we get our academic
vocabulary? Comparing the eciency of direct in-
struction and free voluntary reading. The Reading
Ming-Tzu, K. W., & Nation, P. (2004). Word meaning in aca-
demic English: Homography in the academic word
list. Applied Linguistics, 25(3), 291-314.
Nagy, W., & Anderson, R. (1984). How many words are
there in printed school English? Reading Research
Quarterly, 19(3), 304-330.
Nagy, W., & Townsend, D. (2012). Words as tools: Learn-
ing academic vocabulary as language acquisition.
Reading Research Quarterly, 47(1), 91-108.
Nagy, W., Anderson, R., & Herman, P. (1987). Learning
word meanings from context during normal read-
ing. American Educational Research Journal, 24(2),
Nagy, W., Herman P., & Anderson, R. (1985). Learning
words from context. Reading Research Quarterly,
Nash, H., & Snowling, M. (2006). Teaching new words to
children with poor existing vocabulary: A con-
trolled evaluation of the denition and context
methods. International Journal of Language Com-
munication Disorders, 41(3), 335-354.
Nation, P. (2014). How much input do you need to learn
the most frequent 9,000 words?. Reading in a For-
eign Language, 26(2), 1-16.
National Reading Panel. (2000). Teaching children to read:
An evidence-based assessment of the scientic
research literature on reading and its implications
for reading instruction. Washington, D.C.: National
Institute of Child Health and Human Development,
National Institutes of Health.
Pany, D., & Jenkins, J. (1978). Learning word meanings: A
comparison of instructional procedures. Learning
Disability Quarterly, 1(2), 21-32.
Pany, D., Jenkins, J., & Schreck, J. (1982). Vocabulary in-
struction: Eects on word knowledge and reading
comprehension. Learning Disability Quarterly, 5(3),
Pellicer-Sánchez, A., & Schmitt, N. (2010). Incidental vo-
cabulary acquisition from an authentic novel: Do
“Things Fall Apart"? Reading in a Foreign Language,
Rolls, H., & Rodgers, M. P. (2017). Science-specic technical
vocabulary in science ction-fantasy texts: A case
for ‘language through literature’. English for Specic
Purposes, 48, 44-56.
Schmitt, N., & Carter, R. (2000). The lexical advantages
of narrow reading for second language learners.
TESOL Journal, 9(1), 4-9.
Schwanenugel, P. J., Stahl, S. A., & Mcfalls, E. L. (1997).
Partial word knowledge and vocabulary growth
during reading comprehension. Journal of Literacy
Research, 29(4), 531-553.
Seifert, K., & Espin, C. (2012). Improving reading of science
text for secondary students with learning disabili-
ties: Reading, vocabulary learning, and combined
approaches to instruction. Learning Disability Quar-
terly, 35(4), 236-247.
Shu, H., Anderson, R., & Zhang, H. (1995). Incidental learn-
ing of word meanings while reading: A Chinese
and American cross-cultural study. Reading Re-
search Quarterly, 30(1), 76-95.
Simmons, D., Hairrell, A., Edmonds, M., Vaughn, S., Lars-
en, R., Willson, V., Rupley, W., & Byrns, G. (2010).
A comparison of multiple-strategy methods:
Eects on fourth-grade students’ general and
content-specic reading comprehension and vo-
cabulary development. Journal of Research on Edu-
cational Eectiveness, 3(2), 121-156.
March 2019, Volume 11, Issue 4, 309-318
Snow, C., Lawrence, J., & White, C. (2009). Generating knowl-
edge of academic language among urban middle
school students. Journal of Research on Educational
Eectiveness, 2(4), 325-344.
Spichtig, A., Hiebert, E., Vorstius, C., Pascoe, J., Pearson, P.
D., & Radach, R. (2016). The decline of compre-
hension-based silent reading eciency the United
States: A comparison of current data with perfor-
mance in 1960. Reading Research Quarterly, 51(2),
Stahl, S. (1983). Dierential word knowledge and reading
comprehension. Journal of Reading Behavior, 15(4),
Stahl, S. (1989). Task variations and prior knowledge in learn-
ing word meanings from context. In S. McCormick, &
J. Zutell (Eds), Cognitive and social perspectives for lit-
eracy research and instruction: Thirty-eighth yearbook
of the National Reading Conference (pp. 197-204). Chi-
cago: National Reading Conference.
Stahl, S. (1990). Beyond the instrumentalist hypothesis: Some
relationships between word meanings and comprehen-
sion. Urbana, IL: Center for the Study of Reading.
Stahl, S., & Fairbanks, M. (1986). The eects of vocabulary
instruction: A model-based meta-analysis. Review of
Educational Research, 56(1), 72-110.
Stahl, S., & Nagy, W. (2007). Teaching word meanings. New
Taylor, S.E. (1965). Eye movements in reading: Facts and fal-
lacies. American Educational Research Journal, 2, 187-
Tuinman, J., & Brady, M. (1974). How does vocabulary ac-
count for variance on reading comprehension tests?
A preliminary instructional analysis. In P. Nacke (Ed.),
Twenty-third Yearbook of the National Reading Confer-
ence (pp. 176-184). Clemson, SC: National Reading
Wright, T., & Cervetti. G. (2017). A systematic review of re-
search on vocabulary instruction that impacts text
comprehension. Reading Research Quarterly, 52(2),
Zahar, R., Cobb, T., & Spada, N. (2001). Acquiring vocabulary
through reading: Eects of frequency and contextu-
al richness. Canadian Modern Language Review, 57(4),
1. The National Reading Panel (2000) concluded that SSR
and extensive reading programs do not help students be-
come better readers, but as Krashen (2001, 2005) pointed
out, the Panel omitted several studies of silent reading pro-
grams in its review, and misreported the results of some of
those they did examine. Lewis and Samuels (2005, report-
ed in Allington, 2014) conducted a review of 49 SSR studies,
and reached a similar conclusion to Krashen’s (2004a): “No
study reported signicant negative results; in no instance
did allowing students time for independent reading result in
a decrease in reading achievement” (p. 17). The eight “true
experiment” studies Lewis and Samuels included in a me-
ta-analysis had an eect size favoring free reading (d= .42).
2. Apthorp et al. (2012) included a comparison group that
received a dierent, less intensive form of vocabulary in-
struction. No raw post-test scores on the target words were
reported, only the hierarchical linear modeling results (Table
7, p. 174). Apthorp et al. claimed that the signicant eects
of the intervention held even when taking into account in-
structional time (p. 173), but this merely indicates that the
experimental form of direct instruction was more ecient
than the comparison form of vocabulary teaching. No read-
ing-only comparison group was used.
3. I was unable to locate a read-and-test study of inciden-
tal acquisition among second language or language minor-
ity K-12 students. However, studies of adult L2 vocabulary
acquisition have reported similar probability of acquisition
estimates as those found in Table 3, ranging from .05 (Za-
har, Cobb, & Spada, 2001) to .17 (Pellicer-Sanchez & Schmitt,
4. Two studies (Nagy, Anderson, & Herman, 1987; Shu, An-
derson, & Zhang, 1995) also reported acquisition rates by
the “conceptual diculty” of the target word. The highest
diculty rating (“Level 4”) was given to words that required
new factual information to be understood. Nagy et al. found
that none of the Level 4 words in the passages read by the
students in their study were acquired, while Shu et al. found
with a similar group of subjects that the probability of ac-
quisition for such words was .07, within the low-end of the
range of probabilities reported in Table 3. Nagy et al.’s Level
4 words appear in part to be technical vocabulary, words
specic to a given discipline (e.g. divide meaning “boundary
between drainage systems” (Nagy et al., 1987, p. 250). These
are not the type of words generally included in vocabulary
teaching programs such as those reviewed by Wright and
Cervetti (2017), most of which used “Tier 2” or sub-technical
In Herman et al.’s (1987) study, nearly half of the target
words could be classied as conceptually dicult or requir-
ing new factual information to understand, terms such as
renal, oodplain, ventricle, oxbow lake, and aorta. Yet the
probability of acquisition in Herman et al. (.10) was compa-
rable to the results from Stahl (1989) (.13), which used only
“dicult synonyms” for the target words. This is additional
reason to suspect that Nagy et al.’s (1987) nding may be
5. There is an obvious problem with determining the rates
of incidental acquisition during self-selected reading condi-
tions: how do you assess word gains when every subject is
reading a dierent text? Cho and Krashen (1994) attempted
in part to do this in their study of a group of adult second
language subjects (N = 4). Each of their subjects read texts
of her own choosing. Three of the subjects also underlined
words they did not know. A fourth subject, Alma, was not
part of the original reading study group, and did not un-
derline any words while reading. She instead was given a
pretest and post-test on 165 words that the other three sub-
jects had consistently marked as unknown from the reading
series all four subjects were using.
Alma was not aware she was going to participate in the read-
ing study when she took the pretest, so there is little chance
she would have attempted to study or memorize the words.
Applying the formula from Nagy et al. (1987) used above to
the data reported by Cho and Krashen (1994, Table 3), we
nd Alma’s rate of acquisition was an impressive .56. Alma
reported she did not use a dictionary at all during her read-
ing. These results suggest that self-selected reading may
yield much higher rates of acquisition than in the laboratory
conditions typically used in other studies, where the texts
are chosen by the researchers.