Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Hypothesis
Author(s): Stephen Krashen
Source: The Modern Language Journal, Vol. 73, No. 4 (Winter, 1989), pp. 440-464
Published by: Blackwell Publishing on behalf of the National Federation of Modern Language
Teachers Associations
Stable URL: http://www.jstor.org/stable/326879
Accessed: 25/08/2009 00:57
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=black.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact support@jstor.org.
Blackwell Publishing and National Federation of Modern Language Teachers Associations are collaborating
with JSTOR to digitize, preserve and extend access to The Modern Language Journal.
http://www.jstor.org
We Acquire Vocabulary and Spelling
by Reading: Additional Evidence for
the Input Hypothesis
STEPHEN KRASHEN
University of Southern California
EXCELLENT REASONS EXIST FOR DEVOTING comprehensible input in the form of reading,
attention to vocabulary and spelling.' First, a position argued by several others (e.g., 19,
there are practical reasons. A large vocabulary 122, 123).2
is, of course, essential for mastery of a lan- According to IH, when the Language Acqui-
guage. Second language acquirers know this; sition Device is involved, language is subcon-
they carry dictionaries with them, not gram- sciously acquired-while you are acquiring,
mar books, and regularly report that lack of you don't know you are acquiring; your con-
vocabulary is a major problem. Spelling, espe- scious focus is on the message, not form. Thus,
cially for treacherous languages such as Eng- the acquisition process is identical to what has
lish, is also a problem. Our standards in been termed "incidental learning." Also, ac-
spelling are 100%; a single spelling error in quired knowledge is represented subconsciously
public can mean humiliation. in the brain -it is what Chomsky has termed
On the theoretical level, the study of the ac- "tacit knowledge."
quisition of vocabulary and spelling ability can IH allows that the development of some lin-
help us understand language acquisition in gen- guistic knowledge may occur in other ways,
eral. In my view, the most promising hypothe- outside the language acquisition device, using
sis is that vocabulary and spelling are acquired other mental faculties (16, 17, 42, 75, 120).
in fundamentally the same way the rest of lan- This knowledge is deliberately and consciously
guage is acquired. If this supposition is true, learned, and is represented consciously in the
these areas can be useful laboratories for the brain. Linguistic competence developed this
study of language acquisition. way is highly limited, since it utilizes mental
In this paper, I review some research in vo- faculties that are not specialized for language.
cabulary and spelling and suggest that the Severe limits exist on how much can be
results of this research are, so far, consistent learned, as well as how this knowledge is used
with a central hypothesis that has been pro- (the Monitor Hypothesis, 73, 75).
posed for language acquisition in general, the IH has several competitors. Two of the most
Input Hypothesis, and inconsistent with two popular are the Skill-Building Hypothesis
alternative hypotheses. (SBH) and the Output Hypothesis (OH).
The Input Hypothesis (IH) assumes that we The Skill-Building Hypothesis. According to
acquire language by understanding messages SBH, we learn language by first consciously
(73, 75). More precisely, comprehensible input learning individual rules or items, and gradu-
is the essential environmental ingredient-a ally, through drills and exercises, make these
richly specified internal language acquisition rules "automatic." In terms of the theoretical
device (16) also makes a significant contribu- framework developed in Krashen (72, 75),
tion to language acquisition. I argue that the SBH is the hypothesis that "learning becomes
best hypothesis is that competence in spelling acquisition," also known (121) as the Interface
and vocabulary is most efficiently attained by Hypothesis.3
In vocabulary learning, the skill-building
view involves learning words one at a time, by
The Modern Language Journal, 73, iv (1989) deliberate study, and may include analyzing
0026-7902/89/0004/440 $1.50/0 their parts, their prefixes, suffixes, and roots,
?1989 The Modern Language Journal
and exercises (e.g., draw a line from the word
Acquiring Vocabulary& Spelling 441
to the definition, fill in the blank). The skill- EVIDENCE FOR THE INPUT HYPOTHESIS
building approach to spelling is through word If a hypothesis is correct, it will make only
lists, spelling rules, and exercises. correct predictions. I attempt to show here that
To be precise, two versions of SBH or Inter- the Input Hypothesis makes correct predictions
face Hypothesis exist. The strong version insists
concerning vocabulary and spelling develop-
that all our competence in language comes from ment. This step is necessary to demonstrate the
skill-building, that the only route to acquired correctness of a hypothesis, but it is not suffi-
competence is through conscious learning made cient. It must also be shown that alternative ex-
automatic by some kind of "practice."A weaker
planations for the phenomena predicted by the
version of SBH is that skill-building is a pos-
hypothesis are not correct. I attempt to show
sible route. Other routes exist, such as com- here as well that strong versions of SBH and
prehensible input. OH are not able to explain phenomena pre-
The OutputHypothesis. According to OH, we dicted by IH, or do so awkwardly. Skill-build-
learn language by producing it. I consider only
ing and output plus feedback also have
one version of OH here, one I believe to be
difficulty with phenomena that IH does not
false: we learn rules and items by trying them
predict directly, but that it handles easily.
out in production. If we experience communi-
cative success, our (conscious) hypothesis about MORE COMPREHENSIBLE INPUT, MORE
the rule or item is confirmed. If we experience LANGUAGE ACQUISITION
communicative failure, or correction, our hy- If IH is correct, it predicts, first, that more
pothesis is disconfirmed and we alter it. OH comprehensible input, aural and written,
can exist alone (the strong version), or in com- results in more language acquisition.5 This pre-
bination with other hypotheses.4 diction has been confirmed for other aspects of
Skill-building and output with feedback, it is linguistic competence. Chomsky (13) reported
hypothesized, may produce some competence, that children who grew up in richer print envi-
but the competence is learned, not acquired, ronments displayed more grammatical compe-
and thus is very limited. When skill-building tence. Several studies show that better writers
or output/feedback based classes "work," when read more outside of school (74). Better second
they produce language development, it can be language acquisition, as measured by a variety
due either to conscious learning or to the pres- of tests, is associated with more comprehensible
ence of comprehensible input (sometimes con- input in the second language outside of school
sidered to be "practice"). When the latter is (studies reviewed in Krashen, 73, 75; see also
responsible for student improvement, skill- Hafiz and Tudor).
building usually gets the credit. Good evidence exists that this assertion is
Both SBH and OH are, in this view, closely also true for vocabulary and spelling: more
related. In skill-building, the student is given comprehensible input, in the form of reading,
an explicit rule, then "practices" it. In output is associated with greater competence in vocab-
plus feedback, the student "discovers" the cor- ulary and spelling.
rect explicit rule. Skill-building is thus similar Vocabulary.Children who perform better on
to what is known as "deductive learning," while vocabulary tests report more free voluntary
output plus feedback is similar to "inductive reading. Anderson, Wilson, and Fielding asked
learning." Despite this relationship, and despite fifth graders to record their activities outside
the fact that many programs do both, in this of school, and reported that "among all the ways
paper I consider skill-building and output plus children spend their time, reading books was
feedback to represent independent hypotheses. the best predictor of several measures of read-
A substantial amount of first language re- ing achievement" (2: p. 285), including vocabu-
search on vocabulary and spelling bears on lary. Greaney (52) and Greaney and Hegarty
these hypotheses, but only scraps of second lan- (54) found modest but significant positive cor-
guage research address them. Both first and relations between the amount of leisure read-
second language acquisition results, in my ing reported by fifth graders and performance
opinion, support the view that comprehensible on tests that included a vocabulary measure.
input is the major source of vocabulary and In addition, Rice (111) reported that adults
spelling competence. This evidence is quite who said they spent more time doing leisure
similar to that supporting IH for other aspects reading scored higher on a vocabulary test.
of language. The results of in-school free reading pro-
442 StephenKrashen
grams also demonstrate that more comprehen- tary reading. While this explanation can apply
sible input results in more vocabulary acqui- to the studies showing a relationship between
sition. In Sustained Silent Reading (SSR), a vocabulary and spelling proficiency and re-
certain amount of time, usually five to fifteen ported free voluntary reading, it is hard to see
minutes, is set aside for free voluntary reading, how it applies to Wells' two findings: a relation-
with no book reports or tests on the reading. ship between vocabulary knowledge and pre-
When SSR supplements regular language arts school listening to stories and preschool book
instruction, it typically results in superior vo- ownership; it is unlikely that preschool children
cabulary development (Table I).6 engaged in much vocabulary study. One could
More comprehensible input in the form of hypothesize that these preschool experiences
listening to stories is also associated with better inspired more study in school or more self-study
vocabulary development. Wells (143) reported later on (far-fetched, but possible). To account
that children who heard more stories during for the effects of in-school free voluntary read-
their preschool years were judged by their ing on vocabulary development, skill-building
teachers to have better vocabularies at age ten.7 needs to hypothesize that these programs
Children who grow up in print-rich environ- inspired more study, or contained extra skill-
ments also have better vocabularies. Wells (142) building. While these scenarios are possible,
found that children who owned more books at they seem to me to be extremely unlikely. They
age five, before starting school, did significantly are, however, unexplored possibilities.
better on tests of vocabulary (as well as read- A strong version of OH can account for the
ing comprehension and math) two years later.8 spelling-reported free voluntary reading rela-
Spelling. The relationship between reported tionship by positing that those who read more
leisure reading and spelling has not been inves- outside of school also wrote more, and received
tigated extensively. To my knowledge, only one appropriate feedback on their efforts. To
study exists: Polak and Krashen found that col- account for the vocabulary data, OH can posit
lege ESL students who reported more leisure that those who read more outside of school, who
reading did better on a spelling test. grew up in print-rich environments, who par-
Supplementary free reading in school should ticipated in supplementary free reading pro-
also result in better spelling. In Pfau's study, grams, and who heard more stories also wrote
SSR was done in addition to the regular lan- more, or used new words more in conversation
guage arts program. Pfau found, however, no and received appropriate feedback. These
difference between experimental and compari- scenarios might be partly true; it would be no
son subjects after two years of SSR. While the surprise to learn that more output did take
readers were no worse off, the hypothesis that place in these cases. It is doubtful, however,
more comprehensible input results in more ac- that enough output occurred, or that feedback
quisition was not supported. As we shall see, was frequent enough or precise enough to
such apparent counterevidence is rare. account for the tremendous amount of vocabu-
Alternative Hypotheses. A strong version of lary and spelling development that takes place
SBH, one that claims that skill-building is the (see "Complexity/Size of Language" below).9
only route to competence, can account for these
studies by hypothesizing that deliberate study
of vocabulary and spelling (either in school or ACQUISITION WITHOUT LEARNING
self-study) leads to better reading comprehen- A second prediction that IH makes is that
sion, which in turn leads to more free volun- acquisition can occur without learning. Re-
TABLE I
Impact of Sustained Silent Reading on Vocabulary Acquisition (SSR as Supplement to Regular Program)
search strongly suggests that this is so. Two using a technical vocabulary away and above that ordi-
kinds of evidence for acquisition without learn- narilyemployedby pupils at this level. Our pupilstalked
and wrote easily of "demolition,""spillways,""replenish-
ing are presented here: 1) studies that show that
ing," "fauna,""utilization,""agrarian,""reclamation,"
competence can develop without instruction
"depletion,"and other such terms not common to junior
(defined here as a program based on skill-build- high school vocabularies.
ing); 2) "Read and Test" studies that show that The faculty was impressed. At no time in the study
acquisition occurs after a small amount of com- of the unit had we so much as mentioned vocabulary.
prehensible input. We had been concerned with content. Yet now we were
sure that in the extensive reading done on the subject,
COMPETENCE WITHOUT INSTRUCTION our pupilshad incidentallyaccumulatedan unusualstore
of conservation terms ....
Competence without instruction exists in Our conclusion is this: extensive reading by pupils
other areas of linguistic competence. It has been having definite information goals ahead is most condu-
shown that first and second language acquirers cive to vocabulary growth. We believe that this method
acquire rules of grammar that have never been of vocabulary enlargement, motivated by a special in-
terest and immediate need, is probablyless painful and
taught (22). Many documented cases exist of
adult immigrants acquiring second languages more challengingthan the direct methods of feeding re-
without instruction, some attaining high levels quired, selected lists of words. The latter method sug-
of competence (e.g., 29, 75). The success of gests drill, requirements, uniformity, memory work;
while the former allows choice, individuality, selection,
language teaching methods that rely nearly association, self-activity, experimentation, which pro-
completely on comprehensible input, such as cedures in the secondary school we find more effective
Total Physical Response (4) and The Natural in the main than compulsion (92: pp. 665-66).
Approach (59, 78) suggest that acquisition
without learning exists. Kramer and Palmer re- Miller estimated, on the basis of a vocabulary
test administered after the unit, that about a
cently described a college German class that
relied exclusively on comprehensible input, and year's growth in vocabulary had taken place.
There was, however, no pre-testing and no
reported that a significant amount of language
acquisition took place. The success of "sheltered control group.
subject matter teaching," the finding that sub- Many of those with large vocabularies do not
stantial amounts of language acquisition can claim to have developed them through vocabu-
take place when students learn subject matter lary programs. Smith and Supanich (127)
tested 456 company presidents and reported
through another language is an additional ex-
that they had significantly larger vocabulary
ample of acquisition without learning (31, 60,
scores than a comparison group of adults did.
80).
Evidence exists for competence without in- When asked if they had made an effort to in-
struction in first language literacy development crease their vocabulary since leaving school,
as well. As noted earlier, those who report more only 54.5% of the sample said they had. Of
pleasure reading outside of school write better those 54.5%, when asked what they did to in-
(74), and students who participate in free read- crease their vocabulary, about half mentioned
ing programs in school do better on tests of reading. About a third of this group mentioned
writing, reading, and grammar (76, 77). use of only a dictionary; fourteen (six percent
of those who tried to increase their vocabulary
Vocabulary.Miller (92) informally observed
vocabulary acquisition without instruction in and three percent of the entire sample) men-
junior high school students who completed a tioned vocabulary books.
month's study on conservation of natural re- It has also been shown that children know
sources, which included a great deal of read- enormous numbers of words and acquire vo-
ing, including reading in a phase of conserva- cabulary at an incredible rate, and that vocabu-
tion selected by the student. At the end of the lary teaching programs cannot be a source of
unit, students wrote an editorial essay. Miller this knowledge. Miller (91) has estimated that
remarks: children between the ages of six and eight pick
Only near completion of the unit, after hearing the spe- up an average of fourteen basic words a day.
cific discussions in the science and social science classes, Nagy, Herman, and Anderson (97; see also 96)
afterreadingthe effectiveeditorialson conservation,did have estimated that school-age children acquire
we realize some of the concomitant values of the plan several thousand words per year. Nagy argues
(of study). For one thing, in the study of and writing that direct teaching of vocabulary cannot be the
on these subjects, our junior high school pupils were source of these gains, since even the most am-
444 StephenKrashen
bitious vocabulary teaching programs typically What this table tells us is that children can
do not cover more than a few hundred words spell words that they have not yet been taught;
0 for example, fourth graders could spell 56.8%
per year.
The fact that those who report more free of the fifth grade words, and sixth graders could
reading outside of school have better vocabu- spell 71.2% of the seventh grade words. In
laries (see above, "More Comprehensible addition, the children continued to improve on
Input, More Language Acquisition") and the words on spelling lists taught in previous years.
fact that those who participated in in-school free For example, the fifth graders scored 83.7%
reading programs show significant vocabulary on the grade three list, compared to 79.6% for
gains (see previous section and the next sec- fourth graders and 71.8% for third graders.
tion), are also evidence for acquisition without Unless all spelling words are regularly re-
learning. viewed, this is additional evidence for the ac-
Spelling. Abundant evidence exists that chil- quisition of spelling competence without in-
dren can learn to spell without instruction. One struction.
interesting case history was provided by Good- The last column in the table ("not taught")
man and Goodman, who reported that their are words that none of the students had studied,
daughter Kay, at six-and-a-half years of age, since they were assigned in the last half of grade
could independently read materials written at eight (testing took place in the middle of the
the fifth grade level, even though she had had year). Curtiss and Dolch found that each grade
no formal instruction in reading. Kay's spell- was able to spell some of these words, and
ing ability was impressive. On a spelling test eighth graders scored 82.3%.
based on words taken from a third-grade Nine years before Curtiss and Dolch's study,
reader, Kay was able to spell fifty-eight percent Thompson had reached a similar conclusion:
of the words correctly; on a multiple-choice test, most words explicitly taught in grades two
she recognized ninety-one percent of the cor- through eight "are correctly spelled by upwards
rect spellings of the words. They concluded, of fifty percent of the pupils before they are
quite correctly in my view, that "spelling can given particular attention in the spelling period.
be learned naturally without instruction. At Many words are spelled correctly by ninety per-
least one child has learned to spell without cent or more of the pupils before special study"
studying lists of words in isolation and without (135, p. 39). Twenty-seven years after Curtiss
learning rules or generalizations" (49: p. 226). and Dolch's paper appeared, Hughes reported
Not only can preschoolers learn to spell with- similar results: two-thirds of the fourth grade
out instruction, but older children can as well. children he studied could "already spell cor-
Curtiss and Dolch provide one of the most in- rectly seventy-five per cent of the words sug-
teresting studies showing this. In a simple ex- gested for study for the year" (65: p. 54). See
periment, they administered a test of 500 spell- also Guiles (55).
ing words taught in grades two through eight The results of several other studies suggest
(fifty words at a time) to children in these that spelling competence can develop without
grades, in order to see how well the children formal instruction. Tyler's report is similar to
could spell words they had not yet studied, as Miller's observation on vocabulary growth.
well as how well they could spell words they had Tyler reported that sixth grade students im-
already studied. Table II gives their results. proved significantly in their ability to spell
TABLE II
Average Scores of Each Grade on Words Taught to Each of Grades II-VIII
specialized words frequently used in social free reading programs. Conceivably, company
studies after a fifteen-week unit on history and presidents write and engage in conversation
geography "involving wide reading and a great more, and good readers excused from spelling
variety of learning activities" (140: p. 110). instruction (79) wrote more. But all this re-
Typical pretest scores were around eighty out mains to be demonstrated. Also, it is not likely
of 260; after the unit, average scores rose to that exact feedback was provided in any of these
about 107 correct. Tyler interpreted his data situations.
as suggesting that "individual spelling vocabu- Skill-builders and output advocates could
laries are acquired from pursuit of various ac- also dismiss Miller's junior high school results
tivities relating to specialized fields" (p. 110). by pointing out that no scientifically valid test-
Kyte studied students who were excellent ing took place, and that her observations could
spellers and who were excused from spelling in- simply have been due to the students' greater
struction. He found that these excellent spell- use of words they already knew.
ers, all of whom were good or excellent readers, It could be argued that students consciously
continued to improve their spelling without learn rules and generalizations in class, either
instruction. deductively or inductively, and that they apply
Cornman studied the effect of dropping all these rules to untaught words. This could ac-
spelling instruction in two elementary schools count for some of the results presented here
for three years (spelling errors in compositions (e.g., 23, 24, 79). Rule-learning could account
were still corrected by teachers, however). for some improvement, but not much, due to
Cornman concluded that the effects of spelling the complexity of the rules (see "Complexity/
instruction were "negligible," that uninstructed Size of Language").
students continued to improve in spelling and Similarly, one could argue that students sub-
did just as well as previous years' classes and consciously acquire rules for word-formation
just as well as students in other schools. Corn- and spelling from lists and exercises and apply
man's results were replicated by Richards, who these acquired rules to untaught words. While
studied seventy-eight children in grades six, possible, it is doubtful that much acquisition
seven, and eight who went without spelling in- occurs under these conditions, due to the im-
struction for one year. Richards reported that poverished nature of the input.
67.5% of the children improved more than one
year in spelling during this time, while 20.4% INCIDENTAL READ AND TEST STUDIES
made no change, and only twelve percent got
worse. A number of studies using a similar para-
Alternative Hypothses. To account for these digm confirm that both vocabulary and spelling
results, a strong version of SBH must assume can be acquired by reading for meaning. In
that self-study of vocabulary and spelling took each case, subjects are asked to read something,
place, that those who improved without instruc- usually a short passage, but occasionally some-
tion studied self-help vocabulary-building thing longer (in one study, an entire novel was
books, listened to vocabulary tapes, and dili- used). The text to be read contains words sub-
gently looked words up in the dictionary on jects are unfamiliar with or words they cannot
their own. This is denied by most of the com- spell. After reading the passage, subjects are
pany presidents in Smith and Supanich, is tested on these words. In the studies I report
simply impossible for the Goodmans' Kay (who here, subjects were not focused on vocabulary
developed spelling competence before starting or spelling, that is, they were unaware that spell-
school), and is unlikely in the other studies dis- ing or vocabulary would be tested. Rather, they
cussed here. were focused on the meaning of the passage.
A strong version of OH must make the as- IncidentalRead and Test: Vocabulary.The most
sumption that acquirers, in each case, had careful Read and Test studies probing vocabu-
ample opportunities to try out new words in lary acquisition were done by Nagy, Herman,
writing, and received feedback on their efforts, and Anderson (98), using elementary school
and/or tried out new vocabulary in conversa- students as subjects and passages from school
tion. Increased production opportunities may textbooks as texts. Nagy's team concluded from
have occurred in cases of increased vocabulary its data that when an unfamiliar word was seen
and spelling competence accompanying subject in print, "a small but statistically reliable in-
matter instruction (91, 140) and with in-school crease in word knowledge" typically occurred
446 StephenKrashen
(96: p. 26). They found that the chance of a Results showed that considerable vocabulary
subject's acquiring a word from one exposure acquisition had taken place. Scores ranged from
was between five to twenty percent, depending fifty to ninety-six percent correct, with an aver-
on the testing method used. This percentage age of seventy-six percent. Subjects had picked
may not seem high, but when we consider the up at least forty-five words simply by reading
amount of reading children do, even this small a novel! Saragi also observed some relationship
effect results in considerable vocabulary acqui- between frequency of occurrence and acquisi-
sition. Nagy (97) calculated that if children read tion, noting that words that appeared less than
about a million words per year, given only a ten times were typically not consistently ac-
five percent chance of acquiring a word from quired, a conclusion that matches Nagy, Her-
context with each exposure will result in man, and Anderson's results closely.
vocabulary growth of about 1,000 words per SecondLanguage Studies. A series of recently
year, "well enough to pass fairly discriminating completed projects done at the University of
M Southern California has confirmed that second
multiple-choice tests" (97: p. 262).
This research suggests that words are not language acquirers can also increase their
learned all at once when they are seen in con- vocabulary by reading. In Dana Ferris' study,
text; rather, word knowledge grows in "small university level students of English as a second
increments." At any given time, there are words language (international students) read George
we know well, words we do not know, and Orwell's novel Animal Farm. They were given
words in-between. To increase our vocabulary a multiple-choice test of seventy-five words, in-
we need to follow Twadell's advice and learn cluding fifty words used in Animal Farm, before
to tolerate some vagueness, vagueness that is and after reading the book, words that students
reduced bit by bit as we read more and encoun- at their level typically do not know. The stu-
ter unfamiliar words more. At any given mo- dents made significantly better gains on the test
ment, Twadell notes, "we may 'know' a very than control students who did not read the book
large number of words with various degrees of (Table III).
vagueness- words which are in a twilight zone
between the darkness of entire unfamiliarity TABLE III
and the brightness of complete familiarity" Vocabulary Gains After Reading AnimalFarm
(139: p. 73). Pretest Posttest
The Clockwork Orange Study. Saragi,
Group Score Score Gain
Nation, and Meister asked adults to read Bur-
Experimental (read Animal
gess' novel A ClockworkOrange.As readers who 16.13 27.63 11.50
have read this novel know, it contains a num- Farm)n= 30
Control (did not read
ber of words from a Russian-based slang called AnimalFarm)n=21 12.10 16.14 4.14
nadsat.It can be assumed that few readers know
From Ferris.
these words before reading the book. There are
241 nadsatwords in A ClockworkOrange,and they
are repeated an average of fifteen times each. Emiko Kiyochi asked nine advanced ac-
The version of A ClockworkOrangesold in book- quirers of Japanese as a second language to
stores has a dictionary in the back, so readers read a science fiction story in Japanese con-
can look up the meanings of the nadsat words. taining twelve invented words representing
In Saragi, Nation, and Meister's study, sub- concepts assumed to be familiar to the readers.
jects were simply asked to read the book, and (Subjects were told they were participating in
were told that after they finished it they would market research and were evaluating the text
be given a test of comprehension and literary for its entertainment value and comprehensi-
criticism. They were not told to try to learn or bility.) After reading the story, the subjects
remember the nadsat words. What is crucial is were given a multiple-choice test on the mean-
that they were given copies of the book with- ings of the invented words; the average num-
out the dictionary. The subjects read the book ber acquired was about seven, significantly
on their own time, and reported finishing the better than chance.
book in three days or less. Within a few days Pitts, White, and Krashen performed a suc-
of their finishing the book, subjects were given cessful partial replication of Saragi's A Clock-
a multiple choice test covering ninety nadsat work Orangeexperiment with adult students of
words. English as a second language. The students
Acquiring Vocabulary& Spelling 447
read only the first two chapters of the novel, reading took place "a few days" after the pre-
and, like Saragi's subjects, were deliberately not test.) After reading the passage, they could spell
told to try to learn the nadsat words, but were an average of about one of these words out of
asked simply to read the chapters for meaning. twenty-five. Nisbet found this figure unimpres-
After reading the chapters, students averaged sive and concluded that "intensive reading and
about two words right on a multiple-choice test study of a passage . .. does lead to some learn-
of thirty nadsatwords, after correcting for guess- ing of spelling, but this gain is not sufficient
ing, nearly a seven percent gain. ... to justify the neglect of specific spelling
These three studies confirm that second lan- instruction" (99: p. 11). As we just saw, how-
guage acquirers can acquire vocabulary by ever, Nagy, Herman, and Anderson found that
reading, and that they can do so with "authen- vocabulary acquisition from reading occurred
tic" texts (A ClockworkOrange,Animal Farm) and with similar efficiency. Thus, one out of
with different second languages (English, Japa- twenty-five may be enough, if readers do
nese). enough reading.
ReadingAloud. Using a variation of the Read In Ormrod (101) college students who read
and Test procedure, Eller, Pappas, and Brown for meaning learned to spell an average of 2.8
demonstrated that kindergarteners acquire "pseudo-words" out of a possible eight (an ex-
word meanings from stories read to them; chil- ample of a pseudo-word is Aerodern,used as a
dren were read two stories containing words person's name in the passage; each pseudo-
judged to be unknown to the children. Each word appeared six times). Like Nisbet, Ormrod
story was read three times, usually on consecu- was not pleased with these results and con-
tive days. After each reading, the pupils were cluded that "college students clearly have
asked to "read it in their own way" (35: p. 9). trouble learning to spell the words they see
Analysis of the children's readings showed that within the context of a reading passage .. ."
with each reading, there was evidence of in- (pp. 653-54). I think that her data show just
creased and more accurate use of the target the opposite: after a brief exposure, subjects
words, providing evidence of acquisition with- made progress in learning to spell unfamiliar
out instruction. Elley has reported similar re- words, even when not asked to do so.
sults for seventh and eighth graders using Neither SBH nor OH, in their strong forms,
multiple-choice tests. has any way to deal with incidental Read and
Incidental Read and Test: Spelling. Read and Test studies. In these studies, there was no
Test studies in spelling come to very similar chance to do any deliberate study nor any kind
conclusions. In a series of reports, Gilbert dem- of output. OH supporters could argue that the
onstrated that high school (45, 46) and univer- kindergarten children in Eller (35) searched out
sity students (47) can improve their spelling by meanings for the words used in the stories they
reading. Gilbert presented his subjects with a heard between readings (e.g., asked their
spelling pretest, a reading passage containing parents), and tried out these words at home and
some of the words on the pretest, and a posttest. got feedback. (The children apparently got no
Subjects did much better on the posttest on those feedback on their uses of the words in their re-
words that appeared in the reading passage. tellings.) One might also insist that Ferris' sub-
In Gilbert's studies, the pretest was usually jects looked up unknown words in Animal Farm
given immediately before the reading. Never- on their own, and that subjects in Saragi (114)
theless, Gilbert argued that his subjects were somehow found copies of A ClockworkOrange
focused on the meaning of the passages and not with dictionaries of nadsat words. These expla-
spelling, since they were told they would be nations seem to me to be implausible, but are
tested on the content of the passage. In two of possible.
Gilbert's studies, one experimental group took Intentional Read and Test Studies. In "inten-
the pretest well before doing the reading-three tional" Read and Test studies, subjects are de-
weeks for one of three groups in Gilbert (47), liberately focused on new vocabulary and
four weeks in Gilbert (45). Gains were smaller, spelling words (5, 32, 67, 70, 101) and in some
but were still evident - about 14.5 % in one (47) cases do additional skill-building exercises using
and four percent in the other (45). the target words (32, 70). Intentional Read and
In Nisbet's study, children ages eleven to Test subjects do consistently better in vocabu-
fourteen read passages containing words they lary and spelling than incidental subjects; this
could not spell correctly on a pretest. (The shows that focusing on form will result in addi-
448 StephenKrashen
tional development, and supports the view that for vocabulary); 3) comparing the effects of
some language development can take place formal instruction to normal growth without in-
outside of the Language Acquisition Device struction (spelling only); 4) determining
(LAD). whether more formal instruction results in
We cannot, however, from intentional Read more acquisition (spelling only).
and Test studies alone, ascertain just how
strong the effect of learning is, since we are not VOCABULARY
informed how much extra time and effort went
In-School Free Reading Comparedto Traditional
into skill-building and focusing on form. We Instruction.Previously, I discussed in-school free
do not know, in other words, whether the extra
time would have been better spent reading. reading programs that supplement regular lan-
guage arts programs. Some in-school free read-
(For an attempt to compare time efficiency of ing programs replace some or all of regular
reading and direct vocabulary instruction language arts instruction, and can thus be con-
across studies, see the next section.) The results
sidered competing methods.
of incidental studies suggest that compre- Table IV covers Sustained Silent Reading
hensible input alone can do the entire job for
programs that replace part of regular language
vocabulary and nearly the entire job for spell- arts instruction. Table V covers Self-Selected
ing (see Appendix I). This is, I think, good Reading. As in Sustained Silent Reading, chil-
news for readers. Many people, I am sure, dren in Self-Selected Reading choose their own
would not read at all if they were compelled to
work on their spelling and vocabulary while reading material. They also have regular, brief
conferences with teachers to discuss their read-
trying to enjoy a good book.12 ing and deal with problems.
In Table Va, Self-Selected readers are com-
THE EFFECTS OF INSTRUCTION
pared to students who follow the regular pro-
IH insists that the only way of stimulating gram. The design in Table Vb is weaker
the operation of the language acquisition device because no control group is used; instead, stu-
is comprehensible input. Other means of at- dents' progress is compared with expected
tempting to internalize language, such as skill- growth.
building, rely on other mental faculties outside Tables IV and V show that when free volun-
the language acquisition device. These attempts tary reading, in the form of either Sustained
may result in language-like behavior, but not Silent Reading or Self-Selected Reading, sub-
real language. If this is so, we would expect that stitutes for all or part of the regular language
competence produced outside the LAD makes arts program, free readers do at least as well,
little contribution to language performance. and often better, than students in the regular
This view leads to the prediction that the program on vocabulary tests, suggesting that
effect of traditional, conscious, learning-based free reading is at least as effective as traditional
instruction will be small, compared to the effect instruction. As noted earlier, free reading pro-
of comprehensible input (reading). grams look even better when they are allowed
Comprehensible input has done very well in to run for seven months or longer (see Table
method comparisons in the past: beginning VI; the difference between long- and short-
second language students in comprehensible term programs is not, however, statistically
input-based methods consistently outperform significant).
students in "traditional," usually skill-building- The results of two studies suggest that
based methods (studies reviewed in Krashen, vocabulary development is better served when
73; see also 37 and 59). At the intermediate at least some language arts time is devoted to
level, "sheltered" subject matter teaching has reading aloud. Cohen (19) reported that second
been shown to be very effective (31, 60, 80). graders who were read to every day in school
Previously published research allows several made better gains in vocabulary than second
ways of evaluating the effect of instruction on graders who were not read to regularly. Feitel-
vocabulary and spelling development: 1) com- son, Kita, and Goldstein found that first
paring the progress of students in in-school free graders who were read to daily outscored com-
reading programs with the progress of compari- parison students on every language test given
son groups participating in traditional pro- to them, including vocabulary diversity on a
grams; 2) comparing the time efficiency of free story-telling task.13
reading and formal instruction (possible only Time Efficiency.Nagy's calculations (97) allow
Acquiring Vocabulary& Spelling 449
TABLE IV
Impact of Sustained Silent Reading on Vocabulary Growth (SSR Replaces Part of Regular Program)
TABLE Va
Self-Selected Reading and Vocabulary Growth Compared to Regular (Basal) Programs
TABLE Vb
Self-Selected Reading and Vocabulary Growth (Without Control Group)
TABLE VII
Mean Scores on the "Testof Written Spelling" at Different Grade Levels
the most widely used series in grades three and TABLE VIII
four. Table VII shows only small differences Yearly Gains on Spelling Lists (from Table II)
among the three series (the first three rows); Spelling List
other commercial series (row four) produced
Grade Tested 3 4 5 6 7 8
similar results.
Of more interest to us, Hammill (58) also 3 48.9 29.6 27.4 17.0 13.6 10.9
4 7.8 36.0 19.9 19.8 13.9 13.9
tested the spelling of a group of children who
5 4.1 -1.9 19.0 14.6 15.6 9.5
had had no spelling instruction at all. These
6 11.4 15.0 13.3 31.6 23.9 25.3
uninstructed children were behind at grades 7 -1.5 -4.9 -4.6 -6.1 9.0 1.1
three and four, which seems to indicate that 8 3.8 7.6 7.4 6.7 6.9 2.1
spelling instruction in the other groups was
effective. By grade five, however, there was no Original data from Curtiss & Dolch.
difference among the groups; the uninstructed
children had caught up. Also, the gains during the year of instruction
What has happened here, I suggest, is that are not simply due to instruction; they include
the spelling test used covered many words that the effect of incidental acquisition as well.
children eventually learn to spell from reading. Finally, students' steady progress on word lists
Spelling instruction, for these words, was a during "off years" and their progress in words
wasted effort; it only succeeded in helping chil- that were never instructed (Table II) strongly
dren learn to spell words that they would have suggest that incidental acquisition will even-
learned to spell anyway on their own! The tually take care of nearly all of their spelling
tragedy is that time devoted to spelling could needs. Just as Hammill, Larsen, and McNutt's
have been used for activities we know are good data indicate, instruction may simply teach the
for language development, activities that are spelling of words that children will soon acquire
more pleasant for both children and teachers. on their own anyway.
Thompson studied the effect of spelling Does More InstructionResult in MoreAcquisition?
instruction on 1,528 students in grades two While, in Thompson's study, considerable in-
through eight, and concluded that formal in- struction apparently resulted in significant
struction resulted in a "permanent improve- growth in spelling, according to other studies
ment equivalent to one-half year's gain in more instruction does not consistently lead to
excess of that to be expected as a function of better spelling.
general maturity and incidental learning" (135: Rice (110) administered a variety of spelling
p. 71). This effect is large. The reason may be tests in different school districts "in various sec-
the fact that the children Thompson studied tions of the United States" (110: p. 460). He
spent considerable time on spelling instruction: tested well over 10,000 students, and concluded
in the schools Thompson studied, spelling was that there was no relationship between time de-
covered every day, with a pretest on Mondays voted to spelling and spelling proficiency. Rice
that was followed by student correction and list- interpreted his results as showing "the futility
ing of misspelled words, intensive individual of the spelling grind" and recommended that
study on Tuesdays, a second test on Wednes- no more than fifteen minutes per day be
day, more intensive study on Thursday, and devoted to spelling instruction. Cornman
still another test on Friday. Since twenty new claimed similar results after surveying thirteen
words were covered each week, it is likely that elementary schools in the Philadelphia area.
Thompson's students devoted quite a bit of time In Hillerich, fifth and sixth graders partici-
to spelling. pated in a program in which time devoted to
Curtiss and Dolch Revisited. One could argue spelling was reduced. Their compositions, how-
that the Curtiss and Dolch data (presented in ever, contained fewer spelling errors than com-
Table II), support the effectiveness of instruc- positions written by fifth and sixth graders in
tion. While their data clearly show that spell- a traditional program. The experimental pro-
ing development can occur without instruction, gram also emphasized "many experiences in
the best progress appears to be in the year the writing and focus on the clarity and interest of
target words were taught (in italics in Table that writing, as opposed to its mechanical cor-
VIII). Note, however, that overall, most of the rectness" (63: p. 839). This raises the possibility
progress for any list occurs outside of instruc- that it was writing, not reading, that led to
tion, in years in which the list is not taught. spelling improvement.
452 StephenKrashen
Clarke compared first grade children who spired more word study. It could also be argued
were allowed to use invented spelling with chil- that reading aloud had a similar effect. Skill-
dren whose teachers emphasized correct spell- builders could also argue that the instructional
ing in their writing. Both groups had traditional programs utilized were incorrectly planned or
(basal) reading instruction, including phonics. executed, or that the children didn't apply
The invented spellers scored significantly themselves. A few more repetitions, a little
higher on two out of three spelling tests at the more drill, might have made the difference.
end of the school year (they also did better on OH does better. It needs only to hypothesize
the third test, but the difference was not sig- that in-school free reading and reading aloud
nificant). Traditional spellers spelled more were accompanied by more writing and/or oral
words correctly in their writing, but had access language use, which certainly happens in
to dictionaries and word lists, and were encour- "whole language" programs, or that children in
aged to consult the teacher for help. The in- these programs were stimulated to do more
vented spelling group, as expected, wrote much writing and speaking, and that they received
more (on the final writing assignment it aver- accurate feedback on their attempts to use new
aged 40.9 words, compared to 13.2 for the words.
traditional group), which means that these re- To account for the finding that traditional
sults, like those of Hillerich, are consistent with programs have little or no effect, OH needs
both OH and IH.14 only posit that these programs included little
Conclusions and Alternative Hypotheses. IH opportunity for language production and feed-
handles the results of the studies reviewed in back. OH even finds some support here, since
this section quite well. two programs found to be better for developing
* In-school free reading programs typically spelling ability provided more writing (18, 63).
result in equivalent, and often better vocabu-
lary development, when compared to tradi- COMPLEXITY/SIZE OF LANGUAGE
tional programs (Tables IV, V, VI), and in the
only study comparing free reading and tradi- Native speakers and advanced second lan-
tional instruction in spelling (51, 53), no dif- guage performers of a language have mastered
ference was found. Reading alone is thus at large and complex systems of vocabulary and
least as effective as formal instruction for spell- spelling. This fact alone is devastating for
ing and vocabulary development. strong versions of both SBH and OH, since in-
* Picking up vocabulary from reading is more spection of these systems shows they are too
time-efficient than methods that aim to give stu- complex to learn consciously -so complex, in
dents a thorough knowledge of words. Meth- fact, that professional linguists have not yet
ods that are more time-efficient than reading succeeded in describing them. This fact is not
do not appear to provide as deep a knowledge predicted by IH, but IH has no trouble deal-
of words as reading does. ing with it. The complexity argument has also
* Some studies show spelling instruction to be been used for grammar (73), for phonics (123),
no more effective than incidental acquisition and for the acquisition of"planned discourse,"
alone (23, 112), while others show that instruc- or expository prose style (74).
tion produces measurable gains (24, 55, 58, Vocabulary Size. Using the technique of testing
135). In one (55), gains were modest; in subjects on samples of words chosen randomly
another (135), they required extraordinary from a dictionary, Seashore and Eckerson esti-
effort. In Hammill (58), children without in- mated that the average college undergraduate
struction caught up with instructed children by knows about 156,000 words (58,000 "basic"
grade five. words, 96,000 "derived"words, and 2,000 "rare"
* The results of several studies suggest that words; an example of a basic word is "legal";
more instruction in spelling does not result in "legally" is a derived word). As Smith (123) has
more proficiency. In some cases, in fact, more noted, this knowledge could not have come
instruction appears to result in less spelling pro- from 156,000 trips to the dictionary, 156,000
ficiency (18, 63). flash cards, or 156,000 fill-in-the-blank exer-
SBH must work very hard to explain these cises.
results. In the case of in-school free reading, Not everyone has agreed with this estimate.
it could be hypothesized that free reading in- Lorge and Chall criticized Seashore and Ecker-
Acquiring Vocabulary& Spelling 453
son's methodology, and concluded that Sea- the social and affectivelevels, the terms are not synony-
shore and Eckerson's estimate "based on their mous (41: p. 184).
test may be reduced by at least one-half to When we acquire a word, we also acquire
three-fourths" (87: p. 153). Taking Lorge and considerable knowledge about its grammatical
Chall's conservative figure, reducing 156,000 properties. With verbs, for example, this in-
by three-fourths, leaves about 39,000 words, cludes knowing whether they are transitive or
still far too many to do by direct teaching.15 intransitive, what kinds of complements they
What about second language performers? can be used with, etc. Very little of this is de-
While this research remains to be done, it liberately taught.
seems obvious that many second language ac- Spelling. The complexity of the English spell-
quirers have huge vocabularies, and it is quite ing system is well known. This fact was noted
doubtful that they developed them exclusively by Townsend, who commented that "English
from exercises and dictionary work. orthography seems to demand the mental
VocabularyComplexity. Not only are there prowess required of a student in astronomy"
many words to acquire, there are also subtle (138: p. 470). The difficulties in English spell-
and complex properties of words that compe- ing have been documented repeatedly. Horn,
tent language users have acquired. As noted for example, after analyzing a corpus of 10,000
above ("The Effects of Instruction"), the mean- words, pointed out that single sounds often are
ing of a word is nowhere near adequately repre- represented many different ways in spelling.
sented by a synonym. Indeed, it is often very The sound /sh/, for example, was spelled seven-
hard to come up with a precise definition of a teen different ways, including "ti"(attention),
common word-that is, it is hard to describe "sh"(she), "ci"(ancient), and "ssi"(admission).
the knowledge that advanced language users /k/ was spelled 11 ways, including "c"(call), "k"
have acquired. (keep), "x" (expect, luxury), "ck"(black), and
Finegan and Besnier provide some good ex- "qu" (quite, bouquet).
amples of the subtleties of word meaning that Researchers have formulated rules relating
we acquire. They note: sounds and spelling (e.g., spell /sh/ as "ti" in
initial position of unaccented syllables), but
... we often find that words that appear to be synony- there are a lot of rules, some rules are quite
mous at first glance are used to refer to slightly different
complex, and many rules don't work very well.
sets of concepts or occur in different situations of use.
Smith, citing research by Hanna, Hodges, and
The adjectivesfast, quick, and rapid may be used inter-
Hanna, pointed out that the rules of spelling
changeably in reference to someone's running speed, for
developed by researchers "will lead to an incor-
example. But only fast can be used to describe some-
one's talking speed (as in a fast talker -which is different
rect spelling more than 50 percent of the time,
from a "quick talker") or in the expression life in thefast and they are particularly unreliable for the
lane; quickis the most appropriate term to describe a mind words we want most to spell, the common
or a glance; and rapid is more usual than the other two words of our language. The 'regular' words that
terms if reference is made to a person's strides . . . (41: conform most to the rules tend to be infrequent
p. 184). words for which we rarely have a need" (122:
p. 153).
Words that refer to the same thing often have The complexity of English spelling and our
different social and affective meanings. Finegan failure to describe it adequately lead to the hy-
and Besnier point out that: pothesis that competent spellers acquire spell-
ing knowledge by reading, and not from the
? . . even though movie,film, flick, and motionpicture all
study of spelling rules. This hypothesis cor-
have the same referential meaning, the terms differ in
rectly predicts Henderson's observation, that
social and affective meaning. Film strikes (North Ameri-
good spellers do not know the rules and bad
cans) as a British word or a word that applies more to
movie classics or art movies. Flick is recognized as a term spellers remember them incorrectly.
that can be used only in the most informal contexts. The
AlternativeHypotheses. As noted throughout
term motionpictureis quaintly outdated; if someone were this section, the complexity and size of spelling
to use it in a conversation, it would be for its connota- and vocabulary make any strong version of
tion as a term from the thirties or forties. Thus we can SBH untenable. If the rules are too complex
consider the terms synonymous if we specify that we are for professionals to describe, how can they be
taking only referential meaning into consideration. At learned by language students?
454 StephenKrashen
OH has similar difficulties. For the strong Thus, while both acquisition and learning exist,
version of OH to be correct for vocabulary, they are not equal partners. This conclusion is
each new word, or at least each basic word (but consistent with the hypothesis that the acquisi-
see note 15), must be used in speech or writing, tion of vocabulary and spelling involves the lan-
with each shade of meaning, exhibiting each guage faculty, the mental organ specialized for
of its grammatical properties. In addition, the language. When vocabulary and spelling are
language user must receive feedback on his or consciously learned, mental faculties outside
her use of the word, must notice the feedback, the language faculty are used, and only a
and must be able to use the feedback to arrive limited amount of "language-like" competence
at a correct meaning of the word, or a meaning can be developed (16, 17, 75, 120).
closer to the correct meaning. For spelling, OH Nearly all of the research I have reviewed
must assume that each spelling rule is tested here is from first language acquisition--the
out in writing, that writers receive feedback on small amount of second language research we
each rule, that they notice the feedback, and have, however, points in the same direction.
make a correct analysis of the spelling rule from A hypothesis that spelling and vocabulary are
this feedback, or at least an analysis leading to developed in second languages as they are in
a better version of the rule. If we must learn the first language, by reading, is thus at least
to spell each individual word this way, the reasonable.
burden, of course, is greater. Given the com- If these conclusions are correct, the peda-
plexity and size of the tasks, and the infre- gogical implications are obvious: we are teach-
quency with which children write, these ing vocabulary and spelling, as well as the rest
scenarios are highly unlikely. of language, the hard way. (Even if it were
shown that conscious learning was as good as
acquisition, or even twice as efficient, I would
SUMMARY AND CONCLUSIONS still prefer comprehensible input: an hour of
In some cases, the strong versions of SBH pleasure reading is far preferable to thirty min-
and OH struggle to account for the results that utes of drill.) The easy way is to encourage a
IH handles with ease. In other cases, they fail lot of reading, especially free voluntary read-
completely. 16 According to accepted scientific ing. Nearly everyone in the language teaching
method, just one failure would be enough to profession agrees that reading is beneficial,
eliminate a competing hypothesis. Here, the even without research evidence. Yet, few first
competing hypotheses fall short again and language and nearly no second or foreign lan-
again. guage teaching programs do anything to
Nevertheless, the data do not support a pure encourage it.
IH. Despite the presence of a few studies that We have, of course, tried nearly every other
suggest that learning has no effect (23, 110) or teaching device but the interesting book or
a negative one (18, 63), learning appears to magazine. If reading is so effective, we need
have some impact, as evidenced by the find- much better libraries, filled with books in first
ing that students in traditional methods do and second languages that our students will
make some gains. While a substantial part of read, and we need to provide students help in
these gains is probably due to incidental com- finding books-courses in popular literature,
prehensible input, there are gains over and newspapers, magazines, etc.
above what one would expect from comprehen- Problems.Whenever I suggest a comprehen-
sible input alone (24, 135). Also, subjects who sible input-based approach, certain problems
participate in intentional Read and Test studies are pointed out: 1) we lack books and money
make gains superior to incidental readers, for them; 2) our students need to pass discrete-
showing that focusing on form has some effect. point, form-based tests; 3) next year's teacher
A combined position does fit the data: CI will expect them to know certain words, cer-
results in acquired competence, while skill- tain rules, etc.; 4) reading in school should
building and output plus feedback can lead to focus on works of proven worth; 5) parents,
consciously learned competence. school boards, and administrators expect to see
But conscious language learning does not ap- vocabulary lists and spelling drills.
pear to be as efficient as acquisition from input. Here are my answers:
Acquiring Vocabulary& Spelling 455
* If you lack books, get them. My suspicion * Just because next year's teacher does it wrong
is that reading is not simply a way to develop doesn't mean this year's teacher should also do
vocabulary, spelling, and other important it wrong. Except for the situations noted pre-
aspects of competence, it is the only way. We viously (tests based on particular texts, and
have no choice. The problem is not always grammar tests that emphasize late-acquired
money; often it is a matter of priorities. Just rules), students will be better prepared for any
think how many books can be bought for the situation if they have acquired more language.
price of one computer. * Massive quantities of pleasure reading in the
* Students who read more will do well on dis- students' own areas of interest may, ironically,
crete-point exams of spelling and vocabulary. be the best way to prepare them for the serious
Nevertheless, we do need to get rid of the study of literature. While it is desirable to have
exams. The problem has nothing to do with the students read works of "great literary merit"
validity of discrete-point exams; some of them from the start, it is not always possible to pro-
are valid, that is, they test what they are sup- vide enough books of literary merit that are
posed to test. Students who know more vocabu- comprehensible to the less advanced reader
lary, for example, will do better on a valid (136). Large quantities of light, "low risk"read-
discrete-point vocabulary test than those who ing, in which students are not held responsible
know less. for content, in which they can skip words with-
There are exceptions: when vocabulary tests out fear of missing anything that affects their
are based on particular texts. Kingsley (see grade, will result in vocabulary growth and
Table Vb) argued that her self-selected readers overall language competence that will make
did not excel on vocabulary for this reason- reading the classics easier.17
the test was based on words from a basal series. Light reading may also provide the founda-
Her readers did exceed expected growth in tion for the appreciation of more sophisticated
reading comprehension, but not in vocabulary. literature. Britton argues that ". .. a taste for
Also, students in comprehensible input-based the stereotyped, the second-rate, may at times
classes who have not "learned" will be at a dis- be the first rung of a ladder and not the first
advantage when tests focus on late-acquired step to damnation" (8: p. 268).
rules. * The problem here is that the teaching pro-
Even valid discrete-point language arts tests fession is controlled by outsiders, by amateurs,
cause problems, however, because of their a situation unheard of in other professions
impact on teachers and students. Teachers will (125). It is as if hospital administrators dictated
teach to the exam, and students will study for to surgeons how to operate. We simply need
the exam, and no force is likely to change this. to take control.
Even if free reading is the best way to build Even if we do control our profession, it will
vocabulary, it is hard to convince teachers and certainly be helpful if the public understands
students to throw away the vocabulary lists and what we are doing and why. Part of the long-
read. term solution lies in informing students about
Ironically, dumping vocabulary and spelling the language acquisition process. As Sternfeld
tests can result in better vocabulary and spell- has pointed out, this information will help stu-
ing development. No testing means no time de- dents continue their language acquisition prog-
voted to studying word lists, time that can be ress after the class ends. It will also eventually
devoted to reading, and better vocabulary and produce an informed citizenry that understands
spelling development. language acquisition.
building, with little or no comprehensible input (reading the word, for reasons just given; 2) if you can't skip the
for meaning). word, if it appears to be essential to the meaning of the
15Do we need to acquire derived words? It could be text, guess (more accurately,allow yourself to guess before
argued that once basic words are acquired, all we need to you reach for the dictionary; guessing the meaning of un-
do is learn or acquirerules of affixation.Even if this is true, known words in context occurs automaticallyand rapidly).
the task of vocabulary acquisition is still gigantic. If Sea- As you keep reading, the text itself will tell you whether
shore and Eckerson are correct, there are still 58,000 basic you guessed right or wrong; if the subsequent text makes
words to be acquired-less if Lorge and Chall are correct, sense, your guess was probably right, or at least correct
but an impressive number nevertheless. enough to allow you to interpret the text correctly and
Also, rulesof affixationare themselvescomplex. Teaching acquire at least part of the meaning of the word. If the sub-
students the meanings of roots and affixes are not of much sequent text doesn't make sense, you guessed wrong. Try
help when word meanings are not obviously derived from again. (It is possible, of course, that the text will make sense
component parts. Levin, Carney, and Pressley reportthat even with a wrong guess, and that a reader may carry
direct instruction on roots and affixes helped in clear cases arounda falsehypothesisaboutwhat a wordmeans [7, 116].
such as "exsect"(to cut off, from "ex"[out] and "sect"[cut]), This situation will not last for a long time, however, since
but not in less clear cases, such as "premit"(to introduce, contexts that lead to the right interpretationwill predomi-
from "pre"[before] and "mit"[send]). nate. The cure for wrong hypotheses is more reading!); 3)
16Notethat a competing hypothesis need not fail com- if guessing doesn't work, if you keep guessing incorrectly
pletely.As Chomsky(14) points out, the factthat a hypothe- and cannotcome up with a reasonableidea of what the word
sis can handle data only awkwardly is reason enough to means that makes sense in the text, and the word appears
doubt it. to be essential to the meaning of the passage, look it up
7Ironically, skipping words is an excellent way of (or ask someone). Going to the dictionary is thus the last
building vocabulary. When we skip unessential words, we resort.
read more, and acquirethe meanings (or partsof the mean- The system presented here is for low-risk free voluntary
ings) of other words. If we stop to look up every strange reading. There are, of course, times "when the compre-
word, we read less and acquireless vocabulary.I personally hension must be total whatever the cost in time" (139: p.
know people who refuseto read in otherlanguages, because 77), when readerswill need to know every wordof a passage
they have the mistaken notion that they should never skip for precise interpretation. This is often the case in reading
wordsthey do not know, and they dreadthe workthat look- greatliterature,as well as in technicalreading. But pleasure
ing up every word entails. reading is excellent preparation for this kind of intensive
Here is a three-stepplan for dealing with new words (in- reading, and will make appeals to the dictionary less
spired by Smith and Goodman's ideas): 1) if possible, skip necessary.
APPENDIX I
Good Reader, Poor Speller?
APPENDIX II
Time Efficiency of Reading and Instruction for Vocabulary Acquisition
The effectiveness of different approaches to vocabulary for meaning. (For a thorough meta-analysis of vocabulary
acquisitionis difficultto comparedue to the many different instruction studies based on effect size, see Stahl &
experimental conditions used by researchers, differences Fairbanks.)
in tests, etc. Here, I attempt a crude comparison by calcu- To achieve some uniformity, I have limited this com-
lating time efficiency,the averagenumberof wordsacquired parison to studies using a multiple choice test format. Effi-
per minute. This allowsa comparisonof deliberatevocabu- ciency was calculated by multiplying the rate of presenta-
lary teachingmethodswith Nagy, Herman, and Anderson's tion by the percent of target words acquired.
figure of .25 words per minute acquired through reading
Rate of
Presentation Efficiency
Study/Subjects Method (words/minute) Percent Acquired (words/minute)
Studies with Children.
Beck et al. (7), grade 4 Rich instruction .05 82 .04
McKeown et al., grade 4 Rich instruction .05 75 .04
Levin et al., grade 4 Keyword .48 85 .41
Contextual analysis .48 72 .35
Semantic mapping .48 81 .39
Bull & Wittrock, grade 5 Imagery (1) .67 57 .38
Eeds & Cockrum, grade 5 Teacher interaction .07 84 .06
Dictionary .07 64 .05
Read .07 55 .04
Stahl & Vancil, grade 6 Semantic mapping .04 67 .03
Studies with Adults:
Anderson & Kulhavy Generation 10 65 6.5
(2nd trial)a
Read word & defini- 10 73 7.3
tion 3 times (2nd trial)a
Pressley et al. (exp 4) Keyword 6 79 4.7
Imagery (2) 6 55 3.3
Synonym (1) 6 27 1.6
No strategy 6 50 3.0
Johnson & Stratton Definitions .67 87 .58
Sentences .67 90 .60
Classification .67 90 .60
Synonym (2) .67 86 .58
"Corrected for guessing.
Efficiency in adult studies is significantly greater than efficiency in studies using children as subjects(adult mean = 2.876,
sd=2.555; child mean= .179, sd= .176; t=3.330, df=9, p<.01).