Journal of Reading Behavior
1993, Volume 25, No. 2
THE ACQUISITION OF MORPHOLOGY: LEARNING THE
CONTRIBUTION OF SUFFIXES TO THE MEANINGS
OF DERIVATIVES
William E. Nagy, Irene-Anna N. Diakidoy, and
Richard C. Anderson
University of Illinois at Urbana-Champaign
ABSTRACT
This study explored the development of students' knowledge of the meanings of
10 common English suffixes. A test was constructed to assess students' knowledge
of the contribution of suffixes to the meanings of derivatives. Students were asked
to choose which of several sentences correctly used a suffixed word. The suffixed
words consisted of novel combinations of familiar stems and suffixes (e.g., butterless). Students were also tested on parallel items using familiar nonsuffixed
words. The test was administered to 630 fourth-grade, seventh-grade, and high
school students. Knowledge of the meanings of common English suffixes was
found to undergo significant development between fourth grade and high school.
Even in high school, however, there were some students who showed little knowledge of the meanings of these suffixes. The test identified students who have
particular difficulties with English suffixes, and thus it has potential as a diagnostic
tool.
We estimate that every year, average upper elementary students encounter an
estimated 10,000 new words, that is, words they have not previously encountered
in print (Nagy & Anderson, 1984). Avid readers may encounter several times this
number (Anderson, Wilson, & Fielding, 1988). Hence, proficient reading requires
not just knowing large numbers of words, although this is certainly helpful, but
also having strategies for interpreting new words (Mason, Herman, & Au, 1990).
In the initial stages of learning to read, most of the new words students encounter
in print are already in their oral vocabularies, so that decoding strategies alone will
give students access to both the pronunciation and the meaning of the new words.
In the upper elementary grades, students encounter increasing numbers of words
in print that are not in their oral vocabularies, so that decoding alone is not sufficient
155
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 18, 2016
156
Journal of Reading Behavior
to determine their meaning. However, the majority of such words are related to
more familiar ones through prefixation, suffixation, or compounding (Nagy & Anderson, 1984; White, Power, & White, 1989). A student encountering uncourteously, queenlike, or playfellow in print is unlikely to have seen these words before;
but despite their length and low frequency, such words are not necessarily difficult—if the reader can recognize the familiar parts and understand how these parts
contribute to the meaning of the word. Hence, students' knowledge of morphology—their ability to gain information about the meaning, pronunciation, and part
of speech of new words from their prefixes, roots, and suffixes—plays an important
role in determining how well they read new, long words.
Although a number of studies have examined children's acquisition of knowledge of morphology (Berko, 1958; Condry, 1979; Derwing & Baker, 1979; Freyd
& Baron, 1982; Selby, 1972; Shepherd, 1973; Sternberg & Powell, 1983; Tyler &
Nagy, 1989, 1990; Wysocki & Jenkins, 1987), in our judgment this research does
not yet provide a solid basis for making decisions about instructional practice. The
primary reason for this is that children's knowledge of morphology is multifaceted;
different aspects of knowledge of morphology develop at different rates and times
(Tyler & Nagy, 1989). Distinctions must be made between compounding and
affixation; among affixes, between prefixes and suffixes; and among suffixes, between inflectional and derivational suffixes. Inflectional suffixes are those that create different forms of the same word, for example, the endings for tense (walk,
walked, walking) or number (frog, frogs). Derivational suffixes, such as -ness,
ity, or -ful, are added to a word to form a different word, and generally change
part of speech.
The goal of the present study is to gain a clearer picture of students' acquisition
of knowledge about what some common derivational suffixes contribute to the
meanings of derivatives. We have focused on derivational suffixes primarily because we feel that they represent the most abstract and difficult aspect of morphology that students must learn. Children master inflectional suffixes and compounding
before derivational suffixation (Berko, 1958; Derwing & Baker, 1979). Tyler and
Nagy (1989) found that students learn to recognize the stems within suffixed words
before they learn the contribution of the suffixes themselves. The ability to recognize stems in novel suffixed words—for example, to see the repeat in repeatable—
appeared to be already fully developed by fourth grade. On the other hand, students'
knowledge of what suffixes contribute to the meaning of a derivative was found to
continue to increase through high school, and to be correlated with reading ability
in high school (Tyler & Nagy, 1990).
One reason why derivational suffixation is mastered later than compounding
is the relative abstractness of the information conveyed in derivational suffixes.
Compounds typically consist of two parts, each of which is a word in its own right.
The meanings of compounds are therefore relatively easy to explain: a snowman
is a man made out of snow; a goldsmith is a smith who works with gold. Prefix
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 18, 2016
Meanings of Derivatives
157
meanings are somewhat more abstract, but still easily paraphrasable: Something
which is unseen is not, or cannot be, seen; to reread something is to read it over
again; a transcontinental railroad goes across the continent.
Derivational suffixes rarely allow this kind of transparent paraphrasing. Derivational suffixes are used to assign a word to a new part-of-speech category, and thus
change how the word can be used in a sentence. Attempts to convey the contribution
of a suffix to a derived word in the format of traditional definitions are often
convoluted. In Webster's Third New International Dictionary (1961), for example,
the first meaning given for confinement is "the act of confining or state of being
confined." A more recent school dictionary defines the same word as "the condition of being confined" (Houghton Mifflin Student Dictionary, 1989). In one recently published reading program, the suffix -ment was defined as meaning "having, showing, or doing something," with agreement given as an illustration of this
meaning. Whatever the value of such definitions, we consider it unlikely that
someone who was not already familiar with the suffix -ment would learn what it
contributed to the word confinement from such a definition, or be able to use it
effectively in a sentence.
Another reason for the later acquisition of derivational suffixes may be that
such suffixes are far more common in written language, or more formal oral language, than they are in everyday conversation. Some derivational suffixes, such as
the agentive suffix -er in swimmer, are common enough in oral language; and, in
fact, this particular suffix is acquired fairly early. However, Chafe and Danielewicz
(1987) report that nominalizations (nouns formed from verbs by adding suffixes,
such as performance or representation) are twice as frequent in lectures as they
are in conversations, and three times more frequent in academic papers. In general,
then, derivational suffixes are associated with the more complex syntax of written
language and formal discourse.
Measuring Students' Knowledge of Suffixes
Freyd and Baron (1982) and Wysocki and Jenkins (1987) measured students'
knowledge of the contribution of suffixes to the meanings of derivatives by examining the extent to which this contribution was represented in the definitions students
gave for suffixed words. In both studies, students in fourth through eighth grades
were found to have difficulty conveying the contribution of suffixes in the definitions they produced; given the task of defining a derivative such as simplicity,
students were likely to produce the definition of the stem simple.
Tyler and Nagy (1989), on the other hand, found somewhat higher levels of
knowledge of the contribution of suffixes. They measured students' knowledge
with a different kind of task, in which students were asked to choose which of
several words, differing only in their suffixes, best fit into a sentence context. The
difference between the results of Tyler and Nagy (1989) and those of Freyd and
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 18, 2016
158
Journal of Reading Behavior
Baron (1982) and Wysocki and Jenkins (1987) can be attributed to differences in
the nature of the tasks used. As was acknowledged by both Freyd and Baron and
Wysocki and Jenkins, failure to represent the contribution of a suffix in a definition
might reflect either lack of understanding of the suffix or the inability to express
this understanding in a formal definition.
Tyler and Nagy (1989) measured students' knowledge of suffixes with two
tasks. In the first, students were given a sentence containing a blank (e.g., You
can
the effect by turning off the lights), and asked to choose which of
several suffixed words fit best in the blank (intensify, intensification, intensity, or
intensive). One possible weakness of this task is that it may overestimate students'
knowledge of suffixes, because knowledge of the suffix per se is not absolutely
necessary to choose the correct answer. Students might be able to answer correctly
because they know the particular words involved, as unanalyzed wholes. That is,
they might know (at least implicitly) that intensify is a verb (and hence fits the
sentence context) and that intensity is a noun (and hence does not), in the same
way that they know rejoice to be a verb, and joy to be a noun, without any help
from a suffix.
In the second task used by Tyler and Nagy (1989) students were also given
sentences containing a blank (e.g., / wish Dr. Who would just
and get it
over with). However, in this case the words from which they had to choose were all
pseudowords constructed by adding real suffixes to nonsense stems (transumpation,
transupative, transumpate, transumpatic). Using pseudowords avoids the weakness
of the first task, but may lead to an underestimation of students' knowledge of
suffixes; the presence of a nonsense stem may make it difficult for some students
to judge the relative acceptability of the choices. When this task was administered
in a high school classroom, one student was heard to complain, "How can I choose
which word fits best when I don't know what any of them mean?"
Another task used by Tyler and Nagy (1989) illustrates a related problem in
measuring students' knowledge of morphology. To measure students' ability to
recognize familiar stems in novel derivatives, students were given short sentences
containing an underlined word consisting of a familiar stem, but with a suffix
not often used on that stem—for example, "I'm in a celebratory mood," Mary
announced. They were then asked a question which could be answered if the
meaning of the stem were recognized—did Mary feel like (a) having a party,
(b) being alone, (c) going to sleep, or (d) having a fight? Students' performance
on such items was found to increase between fourth and eighth grade, and to be
correlated with standardized measures of reading ability within each grade. This
would suggest that the ability to recognize familiar stems in suffixed words increases between fourth grade, and is correlated with reading ability. However, this
does not take into account possible differences in knowledge of the stems themselves, in test-wiseness, or in other factors not related to knowledge of morphology.
To control for such factors, Tyler and Nagy constructed parallel items in which
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 18, 2016
Meanings of Derivatives
159
the sentence contained the stem without a suffix—"I'm in a mood to celebrate,"
Mary announced, with the same question and choice of answers as the original
item.
If students' scores on the latter type of item are used as a covariate—hence
controlling for knowledge of stems, test-wiseness, and general reading ability—
performance on the items with suffixed words becomes a relatively pure measure
of students' ability to see familiar stems in novel derivatives. In such analyses,
Tyler and Nagy (1989) found no gain in students' ability to find stems in novel
derivatives between fourth grade and high school; nor was this ability correlated
with standardized measures of reading ability. This suggests that the ability to find
a familiar stem in a novel derivative—the most fundamental part of morphological
knowledge—is already largely in place by fourth grade.
Tyler and Nagy's (1989) measure of students' ability to find stems in novel
derivatives thus incorporated a baseline or covariate allowing them to control for
nonmorphological factors such as root vocabulary knowledge and test-wiseness.
Their measure of suffix knowledge, on the other hand, did not include such a
baseline. In the present study we sought to remedy this deficiency by developing
a measure of suffix knowledge with such a baseline, thus giving us a clearer picture
of students' developing knowledge of suffixes. This measure was also designed to
address the problems of using either familiar suffixed words, or pseudowords with
nonsense stems, as were used by Tyler and Nagy (1989).
The suffixes investigated in the present study were chosen from among the
most frequently occurring suffixes in English (using data from Harwood & Wright,
1956). We also limited our choice of suffixes to neutral suffixes (see Tyler & Nagy,
1989), that is, suffixes that are added only to free-standing words, and that make
only regular and predictable changes in the spelling of the words to which they are
added (e.g., the change of y to i in happiness). These suffixes were chosen so that
we could be relatively sure that errors by students reflected problems with using
the syntactic information provided by the suffix, rather than lack of recognition of
the suffix.
METHOD
Subjects
Subjects were 630 students from schools in medium-sized mid western towns.
Two hundred four were 4th graders, from four elementary schools. One hundred
seventeen were 7th graders, from two junior high schools. Ninety students were
10th graders from one high school (High School A). Another 219 were students in
another midwestem high school (High School B). Students from this high school
were from 12 intact reading and literature classes selected to provide a variety of
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 18, 2016
160
Journal of Reading Behavior
ability levels (from remedial reading classes to honors literature); each class contained students from different grade levels. There were a total of 70 students in
9th grade, 81 in 10th, 39 in 11th, and 29 in 12th.
For many of the students in this study, we were able to obtain standardized
test scores from school files, thus giving us more information about our sample of
subjects. For 68 of the fourth graders, we were able to obtain Reading Total scores
on the SRA Level 3S, Form Q, expressed in terms of national percentiles. These
students had a mean percentile of 49.8 (SD = 25.53), with a range of 3 to 95. These
tests had been administered the preceding year. Other fourth graders were in schools
that would not release information on students' test scores. For 92 of the seventh
graders, we obtained Reading Comprehension subtest scores on the Stanford 7 Plus
test, again expressed in terms of national percentiles. The mean percentile for
seventh graders was 68.6 (SZ> = 24.8), with a range from 7 to 99.
For 74 of the students from High School A, we obtained Reading Comprehension subtest scores from the Stanford 7 Plus expressed in terms of national percentiles. For these students, the mean percentile was 47.4 (SZ> = 22.0), with the range
from 5 to 98. For High School B, we were able to obtain 180 students' scores
from the Stanford Diagnostic Reading Test (Blue level). Scores available for these
students included the Comprehension, Vocabulary, and Word Parts subtests. The
mean percentile for students in High School B was 71.4 (SL> = 23.4).
Materials
A test was constructed to measure students' knowledge of the contribution of
common English suffixes to the meanings of derivatives. The contribution of a
derivational suffix lies primarily in the effect it has on a word's part of speech,
and hence, on how the suffixed word can be used in sentences. Since the part of
speech of familiar words might be known apart from any analysis of the word into
a stem and suffix, we based our test on rarely-occurring derivatives of familiar
stems. This approach also was intended to circumvent problems associated with
the use of nonsense stems.
Candidate stems were selected that were judged to be familiar to readers at the
fourth grade level or above. Stems in the final materials ranged in frequency from
8.8 to 37.2 occurrences per million words of text, using frequencies from Carroll,
Davies, and Richman (1971).
A pool of potential items was created by pairing stems with common neutral
suffixes in such a way as to produce novel or rarely occurring derivatives. Derivatives used in the final materials were selected on the basis of the judgments of five
trained raters, all graduate students in education. Raters were asked to decide
whether each derivative in the pool of items was well formed or ill formed, that
is, whether or not a suffix could be used with a particular stem to form a plausible
English word. Whether or not a suffix can be used depends to a large part on the
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 18, 2016
Meanings of Derivatives
161
part of speech of the stem. For example, the suffix -less is normally added to nouns
to form an adjective; so ceilingless should be a well-formed, though rare, word,
since ceiling is a noun, whereas inventless is ill-formed, since invent can only
be a verb in English. Raters were given dogless, chairless, laxness, complainer,
paperless, and findable as examples of well-formed derivatives; some of these
words are relatively rare, but they sound reasonable in an appropriate context
("After three dogless years, he went out and bought a puppy"). As examples of
ill-formed derivatives, the raters were given understandless, imagineless, payness,
doorable, lioner, and incitish.
Twenty rarely occurring suffixed words judged as well-formed by all five
raters were selected: rescuable, powderize, roarer, echoful, activeness, weaponist,
parentism, butterless, cheesish, cutely, mirrorize, barkless, orangeness, froggish,
snappable, evilly, greaseful, profitism, herdist, and repairer. Ten out of these 20
words can be found in Webster's Third New International Dictionary (1961). Only
two were found listed in Carroll, Davies, and Richman (1971)—evilly and roarer.
These two words are low in frequency, occurring less than once per 10 million
words of text. Therefore, it was assumed that most students would not have previously seen any of the suffixed words as units and would have to analyze them into
stem and suffix to understand them.
For each of the 20 words, a multiple-choice item was created to assess students'
knowledge of the contribution of the suffix to the meaning of the derivative. Each
item began with a question of the form "Which sentence uses the word X most
correctly?" Following the question were four sentences, each containing essentially
the same content words, and differing only in their syntactic structure. For the
word powderize, the four sentences were (a) First they had to find a powderize
rock, (b) First they had to powderize find the rock, (c) First they had to find a
powderize for the rock, and (d) First they had to find a way to powderize the rock.
Given that powderize must be a verb, only in the last sentence is it used correctly.
Two versions of each item were created. The example just given is the derivative version of the item. A corresponding stem version was created by replacing
the suffixed word, in this case powderize, with a nonsuffixed word of the same
part of speech, in this case, smash, which fit the sentence contexts as well, or as
poorly, as the word it replaced. The nonsuffixed words had a greater range of
frequencies than did the stems of the suffixed words, but were not significantly
different from them in frequency, F ( l , 39) = 1.1, p>.25.
The derivative version of an item measures the students' knowledge of what
the suffix contributes to the meaning of the derivative—that is, the information it
conveys about the part of speech of the suffixed word. The stem version of the
item measures everything else that contributes to a student's performance on such
an item: ability to understand the context sentence, test-taking skills, possible
effects of the order of choices, and syntactic awareness, or the ability to reflect on
sentence structure. Thus, the stem version of the item provides us with a baseline
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 18, 2016
162
Journal of Reading Behavior
against which to evaluate students' performance on the derivative version. When
stem item performance is controlled for, performance on derivative items therefore
gives us a measure of just that by which the two versions differ—the fact that the
derivative item requires the student to use his or her knowledge about suffixes to
figure out the part of speech for a novel suffixed word.
If a student were able to make full use of the information provided by suffixes,
the score on derivative items should be as high as that on stem items. In fact, one
might even expect the score for derivative items to be higher: Although powderize
is a novel word, it is explicitly marked by the suffix for part of speech—it is
obviously a verb. On the other hand, most nonsuffixed English words are at least
potentially ambiguous as to their part of speech. Smash, for example, is typically
a verb, but it is sometimes used as an adjective (a smash hit), and could conceivably
be used as a noun (as in it was a real smash or the smash of the pumpkin on the
sidewalk). The extent that students score lower on derivative items than stem items,
then, can be taken as an indication of their failure to use the syntactic cues provided
by the suffixes.
Two versions of a multiple-choice test were constructed, with each version
containing 10 derivative and 10 stem items in a random order. Any given subject
saw only one version of an item. Versions were distributed randomly within each
group tested; 311 students received one version of the test, and 319 received the
other.
To measure the reliability of the test, Cronbach's alpha was computed separately for the stem and derivative items in each of the two versions. For the stem
items, alpha was .75 and .76 for the two versions. For derivative items, alpha was
.74 and .79 for the two versions. We consider this to constitute adequate reliability
for a 10 item, experimenter-designed test. If all 20 items in each version are taken
into account, alpha is .83 and .87 for the two versions.
Procedure
Students were tested in intact classroom groups, with the exception of the
students from High School A, who were tested in a single group in the school
cafeteria. The suffix test was administered as the second of two paper-and-pencil
tasks, both given within a single class period of 45-50 minutes. (The first task was
a yes/no vocabulary test; results from that test are not reported here.)
Instructions were read aloud as students followed along in the written instructions on the first page of their test booklet. Students were asked to choose which
of four sentences used a specified word most correctly, and circle the corresponding
letter in their test booklet. A sample item was reviewed briefly. Students were
encouraged to read all four sentences before making a choice, and to guess if
necessary.
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 18, 2016
Meanings of Derivatives
163
Design
The primary question addressed in this study is whether there is growth in
students' knowledge of suffixes after fourth grade, after development of other
factors, for example, syntactic awareness and test taking skills, has been taken into
account. This means finding out whether there is an increase with age in scores for
the derivative items relative to the scores for stem items. A 3 X 2 mixed factorial
design was used, with grade (fourth grade, seventh grade, or high school) as a
between-subject factor, and suffixation (stem version or derivative version of item)
as a within subject factor. We expected to find a main effect of suffixation—scores
for derivative items should be lower than scores for stem items—and a main effect
of grade. Developmental changes in students' knowledge of suffixes would take
the form a Grade x Suffixation interaction; we expected the gap between stem
and derivative items to be smaller at higher grade levels.
RESULTS AND DISCUSSION
Means are given in Table 1. An analysis of variance showed a significant main
effect of grade, F(2, 627) = 41.5, p<.00l; older students performed better. There
was also a main effect of suffixation, F ( l , 627) = 343.6, p<Ml. Students did
better on the stem items than on the derivative items. Furthermore, there was a
significant Grade X Suffixation interaction, F(2, 627) = 10.3, p<.001. As can be
seen in Table 1, not only did fourth graders score lower than older students on
both stem and derivative items, but also, the gap between stem and derivative
scores was greater in fourth grade.
Table 1
Means (and Standard Deviations) for Stem and Derivative Items
Stem Items
Raw scores
Fourth Grade (n = 204)
Seventh Grade («=117)
High School (n = 309)
All Subjects (n-630)
.75
.87
.87
.83
(.22)
(.19)
(.18)
(.20)
Adjusted means, with reading ability as a covariate
Fourth Grade (n = 68)
.76
Seventh Grade (n = 92)
, .88
.87
High School (n = 254)
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 18, 2016
Derivative Items
.55
.72
.75
.68
.54
.72
.75
(.26)
(.25)
(.24)
(.26)
164
Journal of Reading Behavior
Tests of simple effects revealed that fourth grade means were significantly
different from seventh grade means for stem items F ( l , 627) = 12.9, p < . 0 1 , and
derivative items, F ( l , 627) = 27.0, p < . 0 1 . The difference between seventh grade
and high school was not significant for stem items, F ( l , 627) = 0.2, or for derivative items, F ( l , 627) = 2.6, p>.05. The difference between stem and derivative
items was highly significant at fourth grade, F ( l , 627)=209.0, at seventh grade,
F ( l , 627) = 62.7, and in high school, F ( l , 627) = 90.8, all/?s<.001.
In a parallel ANOVA with the item as the unit of analysis, results were the
same: There was a significant main effect of grade, F(2, 38) = 115.63, p<.001, a
significant main effect of suffixation, F ( l , 19) = 27.14, p<.001, and a significant
Grade x Suffixation interaction, F(2, 38) = 6.59, p<.005.
Differences between the three grade levels may reflect differences in the makeup of the student populations sampled, as well as genuinely developmental effects.
To compensate for possible differences in ability between the three grade levels,
an analysis of covariance parallel to the main analysis was performed, using as a
covariate students' reading ability, expressed in terms of national percentile scores,
for those students whose ability was known. Results of this analysis were essentially
the same as those for the first analysis: There was a main effect of grade, F(2,
410) = 29.3,p<.001, a main effect of suffixation, F ( l , 410) = 199.6, p<.001, and
a significant Grade x Suffixation interaction, F(2, 410) = 9.1, /?<.001. Means
adjusted for the covariate are given in Table 1.
Another way of representing the results is in terms of a scatterplot comparing
students' scores on the derivative items with their scores on the stem items. Such
a scatterplot is given in Figure 1. The Y-axis represents students' proportion correct
for Derivative items, corrected for guessing. (Scores less than zero have been set
equal to zero.) The X-axis represents proportion correct on stem items, also corrected for guessing. The scatterplot in Figure 1 includes data from all students at
all grade levels; scatterplots done separately at each grade level were highly similar.
The scatterplot in Figure 1 reveals two things about students' performance.
First, there is a fairly strong correlation between students' performance on derivative and stem items, r = .67, p< .001. This is not surprising, of course, since the
two types of items have quite a bit in common. However, this correlation has
important implications for the evaluation of students' knowledge of suffixes. Almost half of the variance in students' performance on a task measuring their knowledge of suffixes can be accounted for by factors that do not directly involve suffixes
at all. That is, a student represented by a point in the lower left-hand corner of
Figure 1 would do very poorly on a classroom task which depended on using a
suffixed word correctly, or on choosing the correct suffixed word to use in a
sentence. In the case of this student, however, such poor performance would stem
from a general difficulty with reflecting on and manipulating syntactic structures.
On the other hand, it can be seen from the scatterplot that students' performance on derivative items depends on specifically morphological knowledge as
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 18, 2016
Meanings of Derivatives
0.0
0.2
0.4
0.6
Stem Proportion Correct
165
0.8
1.0
Figure 1. Proportion correct for derivative items as a function of proportion correct
for stem items.
well. The majority of the students are below the diagonal/that is, their score for
derivative items is lower than their score for stem items. Many are rather far down
toward the lower right-hand corner, indicating that they do quite well on stem
items, yet poorly on derivative items. Syntactic awareness and other abilities that
contribute to students' performance on both stem and derivative items appear to be
a necessary, but not sufficient, condition for knowing the contribution of familiar
suffixes to the meanings of derivatives. Students represented by points in the lower
right-hand corner of the scatterplot appear to have a problem specifically with
suffixes.
To examine the relationship between morphological knowledge and reading
ability, regression analyses were performed separately for each grade level. The
dependent variable was students' score for derivative items; the score for stem items
was then entered to remove variance not specifically associated with knowledge of
morphology; then reading ability scores were entered. Table 2 summarizes the
results of these analyses. At each grade level, reading scores are significantly
related to performance on derivative items even when performance on stem items
is controlled for.
Using the subtest scores for the Stanford Diagnostic Reading Test for students
from High School B, we found additional evidence that students' performance
on Derivative items reflects specifically morphological knowledge. In a multiple
regression analysis parallel to those just described, after entering stem-item scores
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 18, 2016
166
Journal of Reading Behavior
Table 2
Multiple Regressions Relating Performance on Derivative Items to Standardized
Measures of Reading Ability after Controlling for Performance on Stem Items
R2
Change in R2
F
P
Simple R
Dependent Variable: Derivative Item Score
Fourth Grade
Stem score
Reading Ability
.216
.263
.216
.047
18.2
4.2
.000
.045
.46
.26
Seventh Grade
Stem score
Reading Ability
.276
.413
.276
.138
34.3
20.9
.000
.000
.53
.60
High School
Stem score
Reading Ability
.437
.499
.437
.063
195.5
31.3
.000
.000
.66
.51
as a covariate, we then entered subjects' Reading Comprehension Total and Word
Part subtest scores. The effect of the Word Parts subtest was found to be significant whether this variable was entered before, F ( l , 146) = 21.1, p<.00l, or after,
F ( l , 146) = 10.0, p < . 0 1 , the Reading Comprehension Total score. The latter variable was significant, F ( l , 146) = 10.6, p < . 0 1 , only if entered before the Word
Part subtest score.
On the one hand, this is not a surprising result: Both the Derivative items in
the current study and the Word Parts subtest of the Standard Diagnostic Reading
Test purport to measure students' knowledge of word parts. On the other hand,
there are several important differences between the two tests. Of the 30 items on
the Word Parts subtest, only one specifically involves a suffix. The nature of the
tasks is also different; in the Word Parts subtest, students are asked to choose the
answer closest in meaning to a particular underlined word part; there is no context
sentence involved in the items, and part of speech plays no role. Despite their
differences, however, these two tasks appear to tap a similar ability to deal with
word parts that cannot be reduced to general reading ability.
We examined the data further to determine whether other factors not directly
controlled in the main study might contribute to differences in knowledge of suffixes. Adjusted means for the derivative items were computed, controlling for
performance on the stem items. From the data for High School B, the 32 students
with the lowest adjusted means were identified, and paired with 32 other students
who had high adjusted means. Each pair was matched for performance on the stem
items, and as far as possible on the Reading Comprehension Total subscore for the
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 18, 2016
Meanings of Derivatives
167
Derivative Item Performance
.2
.3
.4
.5
.6
.7
.8
.9
to
Stem Item Performance
Fourth Grade
Seventh Grade
- High School
- • - * - -
Figure 2. Performance on derivative items as a function of performance on stem
items for the three grade levels.
Stanford Diagnostic Reading Test. This procedure resulted in a high suffix knowledge group and a low suffix knowledge group. A teacher familiar with the students
in this high school was then asked to categorize a randomized list of these 64
students by gender, race, and whether the student was a native or nonnative speaker
of English. None of these factors was found to relate significantly to suffix
knowledge.
The purpose of this paper was to get a more accurate picture of students'
developing knowledge of suffixes. In particular, we attempted to do this by developing a test which would allow us to distinguish between differences in performance due to knowledge of what suffixes contribute to the meanings of derivatives,
and differences in performance due to non-morphological factors. We have interpreted our results, and the Grade X Suffixation interaction in particular, as showing
a developmental increase in students' knowledge of derivational suffixes between
fourth grade and high school.
An alternative interpretation of the Grade X Suffixation interaction is possible,
however: It may simply reflect a ceiling on scores on the performance of the seventh
graders and high school students. To determine to what extent the relative size of
the difference between stem and derivative scores for the three grade levels might
be due to a performance ceiling, we computed best-fitting functions relating stem
and derivative scores separately for each grade level. Derivative scores were regressed on stem scores, and the square of stem scores. Figure 2 shows the curves
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 18, 2016
168
Journal of Reading Behavior
representing the derivative score associated with a given stem score according to
the best-fitting function at each grade level.
If there is an increase in knowledge of derivational suffixes between fourth
grade and the later grades, one would expect the lines for seventh grade and high
school to be higher than the line for fourth grade. This is in fact the case for the
areas of the graph representing most of the subjects. Less than 3% of the subjects
in this study had stem scores less than .3. In fact, since the scores in this figure
are not corrected for guessing, values less than .25 represent performance below
the level of chance. Ninety percent of the subjects had stem scores ranging between
.5 and 1.0; in this area of the graph, the best-fitting function for fourth grade is
consistently lower than those for seventh grade and high school. At the overall
sample mean for stem scores, .83, there is a difference of .09 between fourth grade
and high school; the lines for seventh grade and high school, however, are very
close. That is, at this point fourth grade derivative item scores were almost one
full item below those of the older students who had attained the same stem item
scores. Therefore, we conclude that there is a real developmental difference between fourth graders and older students in their knowledge of derivational suffixes.
Conclusion
Significant development in students' knowledge of suffixes was found to occur
between fourth grade and high school, with most of the growth appearing to occur
between fourth and seventh grades. Even in high school, however, most students
did not do as well on derivative items as on stem items, indicating that knowledge
of even common English suffixes was not complete.
The bulk of the improvement in students' performance on derivative items
could be attributed to factors also influencing performance on stem items, rather
than specifically to knowledge of morphology. However, there appeared to be
significant gain in knowledge of morphology, as evidenced by the significant Grade
X Suffixation interaction. Furthermore, there was significant individual variation
at each grade level tested that could be attributed specifically to differential knowledge of suffixes. Thus, knowledge of suffixes constitutes a distinct component of
skilled reading, and was found to be more strongly correlated with other measures
of morphological knowledge than with more general measures of reading comprehension.
Most students at seventh grade or above were able to perform fairly well on
the suffix test, demonstrating that they possessed and were able to apply knowledge
of several common English suffixes. Thus, the inability of students to represent
accurately the contribution of suffixes in the definitions they produce for derived
words, as observed by Freyd and Baron (1982) and Wysocki and Jenkins (1987),
can be attributed in part to the difficulty of conveying this contribution through
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 18, 2016
Meanings of Derivatives
169
verbal definitions. This result calls into question the common practice of teaching
the meanings of derivational suffixes via definitions.
It is important to recognize that the scope of this study was limited to a
small number of familiar, neutral suffixes. Although nonneutral suffixes tend to be
acquired later than neutral suffixes, Tyler and Nagy (1990) found no differences
between neutral and nonneutral suffixes, as far as students' knowledge of their
meanings (i.e., syntactic function) was concerned. Hence, we hypothesize that if
this study were replicated with a broader range of suffixes, one might find slightly
lower levels of overall performance, and larger developmental differences.
REFERENCES
Anderson, R. C., Wilson, P., & Fielding, L. (1988). Growth in reading and how children spend their
time outside of school. Reading Research Quarterly, 23, 285-303.
Berko, J. (1958). The child's learning of English morphology. Word, 14, 150-177.
Carroll, J. B., Davies, P., & Richman, B. (1971). The American Heritage word frequency book.
Boston: Houghton Mifflin.
Chafe, W., & Danielewicz, J. (1987). Properties of spoken and written language. In R. Horowitz &
S. J. Samuels (Eds.), Comprehending oral and written language (pp. 83-113). New York: Academic Press.
Condry, S. (1979). A developmental study of processes of word derivation in elementary school.
Unpublished doctoral dissertation, Cornell University, New York.
Cronbach, L. J. (1951). Coefficient Alpha and the internal structure of tests. Psychometrika, 16,
297-334.
Derwing, B. L., & Baker, W. J. (1979). Recent research on the acquisition of English morphology.
In P. Fletcher & M. Garman (Eds.), Language acquisition (pp. 209-223). Cambridge, U.K.:
Cambridge University Press.
Freyd, P., & Baron, J. (1982). Individual differences in acquisition of derivational morphology. Journal
of Verbal Learning and Verbal Behavior, 21, 282-295.
Houghton Mifflin Student Dictionary. (1989). Boston: Houghton Mifflin.
Harwood, F. W., & Wright, A. M. (1956). Statistical study of English word formation. Language, 32,
260-273.
Mason, J. M., Herman, P. A., & Au, K. H. (1990). Children's developing knowledge of words (Tech.
Rep. No. 513). Urbana-Champaign: University of Illinois, Center for the Study of Reading.
Nagy, W., & Anderson, R. C. (1984). How many words are there in printed school English? Reading
Research Quarterly, 19, 304-330.
Selby, S. (1972). The development of morphological rules in children. British Journal of Educational
Psychology, 42, 293-299.
Shepherd, J. S. (1973). The relations between knowledge of word parts and knowledge of derivatives
among college freshmen. Unpublished doctoral dissertation, New York University.
SRA Survey of Basic Skills, Level 3S, Form Q. (1985). Monterey, CA: CTB McMillan McGraw Hill.
Stanford Achievement Test, 7+. (1987). San Antonio, TX: Psychological Corp.
Sternberg, R., & Powell, J. (1983). Comprehending verbal comprehension. American Psychologist,
38, 878-893.
Tyler, A., & Nagy, W. (1989). The acquisition of English derivational morphology. Journal of Memory
and Language, 28, 649-667.
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 18, 2016
170
Journal of Reading Behavior
Tyler, A., & Nagy, W. (1990). Use of derivational morphology during reading. Cognition, 36, 17-34.
Webster's Third New International Dictionary. (1961). Springfield, MA: Merriam Co.
White, T., Power, M., & White, S. (1989). Morphological analysis: Implications for teaching and
understanding vocabulary growth. Reading Research Quarterly, 24, 283-304.
Wysocki, K., & Jenkins, J. (1987). Deriving word meanings through morphological generalization.
Reading Research Quarterly, 22, 66-81.
AUTHOR NOTE
The work upon which this paper was based was supported in part by the Office
of Educational Research and Improvement under Cooperative Agreement No.
G0087-C1001-90 with the Reading Research and Education Center. Requests for
reprints should be sent to William E. Nagy, Center for the Study of Reading,
University of Illinois at Urbana-Champaign, 174 Children's Research Center, 51
Gerty Drive, Champaign, IL 61820.
Manuscript submitted: July 7, 1992
Conditional acceptance: August 10, 1992
Revision received: September 15, 1992
Accepted for publication: October 23, 1992
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 18, 2016