SHORT-TERM MEMORY AND iMITATIO‘N: -‘ .-

EFFECTS or SENTENCE LENGTH, »  

SENTENCE TYRE WORD TYRE; .
STRESS. AND GRADE *

Thesis for the Degree of M. A.
MICHIGAN STATE UNIVERSITY
ANNETTE SMITH

' 1972 e

lllllllllllllllllllllllllllllll

WWIIlIIITHI‘ITUTllllllhl"ll*l"lllIHIIHHHIIUI LIBRARY

31293 00618 4950 .-. ' Michigan State
Univex. 7"y

 

 

!
I

 

 

 

 

 

ABSTRACT

SHORT-TERM MEMORY AND IMITATION: EFFECTS
OF SENTENCE LENGTH, SENTENCE TYPE,
WORD TYPE, STRESS, AND GRADE

BY
Annette Smith

A review of the literature suggested the need for
further analytical work in the area of language learning.
Previous research, in an effort to be descriptive about a
particular variable under study, had failed to consider
the interaction of the variables involved and, further,
failed to adequately experimentally control extraneous
variables.

The purpose of this study was to investigate the
responses of first, third, and fifth grade children to a
sentence imitation task involving simultaneously the
factors of sequence length, sentence type, word type, and
stress.upon a sentence repetition task.

Ninety first, third, and fifth grade children from
an elementary school in southern Michigan served as
subjects. All had normal speech, language, and hearing
as assessed from the school screening records. Fifty-four

sentences were presented via a tape recorder in a sound

Annette Smith

field to each of the subjects. The four factors of sentence
length, sentence type, word type, and stress were completely
counter-balanced and then randomized for the presentation.
All ninety subjects received three—, five-, and seven-word
sentences of ill-ordered, anomalous, and well-formed con-
struction for two conditions of stress (stress and unstress)
falling on both contentive and functor word types. Using
standardized instructions subjects were asked to repeat
exactly what they heard on the tape.

Errors were scored on a mean per cent error basis.
Any words deleted or substituted were considered an error.

The results of this study indicated that sentence
length, sentence type, word type, stress, and the grade
level of the child did interact at various levels. There-
fore, the main effects depicted by a separate analysis of
the variables became incomplete as a description of what
cues may be used for language encoding and performance.
The results showed that as length increased, error rate
increased. Further, ill—ordered sequences had greater
error rates than anomalous and well-formed types, re-
spectively. The effect for word type indicated that more
errors occurred on functor word types than on contentive
word types. The effect for stress indicated that unstressed
words had greater error rates than stressed words. Finally,
error rate decreased as grade level increased.

Two major interactions occurred between sentence

length and sentence type and between stress and word type.

Annette Smith

Ill-ordered sequences had higher error rates at all lengths
than any other length-type combinations. The differences
between stressed and unstressed error rates were greater
than the differences between functor and contentive error
rates. This indicated that the use of stress as a diff-
ential cue in retention of words was correctly utilized
more often than were word type cues.

The results of this study, as depicted by the
subjects' differential retention of words in this sentence
imitation task, indicated that sentence length, sentence
type, word type, stress, and grade level of the child did
interact at various levels. This study also suggested
that future research be conducted in order to further

substantiate the trends and investigate the interactions.

SHORT-TERM MEMORY AND IMITATION: EFFECTS
OF SENTENCE LENGTH, SENTENCE TYPE,
WORD TYPE, STRESS, AND GRADE

BY

Annette Smith

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF ARTS

Department of Audiology and Speech Sciences

1972

Accepted by the faculty of the Department of Audiology and
Speech Sciences, College of Communication Arts, Michigan State
University, in partial fulfillment of the requirements for the

Master of Arts Degree.

4/4’Z-Cé4f—7 77c dds k
‘\_‘
Director of Thesis

Guidance Committee: »{ j/ZL,’ [[5 (KW ‘ Chairman
I

 

ii

AC KNOWLEDGMENT S

I wish to eXpress my sincerest gratitude to
Dr. Daniel S. Beasley, my thesis director, and to
Dr. William F. Rintelmann and to Dr. Julia Falk, the
members of my guidance committee, for their personal and
intellectual contributions of time and assistance in the
preparation of this thesis.

I would also like to express appreciation to
Dr. David Haarer, Director of Special Education for Ingham
Intermediate School District, for his cooperation and
support in this study. Special thanks are extended to
Mr. D. Macgrayne, Principal of North Aurelius School, and
to the students who so willingly participated as subjects
in this study. I further thank the teachers and staff
who gave their valuable time for the benefit of this study.

I graciously thank Marlene Cosgrove, my family,
and particularly my husband, Gaylord, who so willingly and
unselfishly gave their time, encouragement, and support
throughout this endeavor.

Finally, I would like to express my sincere personal

gratitude to Dr. Beasley who, as a friend, instructor,

iii

teacher, and advisor, contributed endlessly to the

development of this thesis and this student.

iv

Chapter

I.

II.

III.

IV.

V.

TABLE OF CONTENTS

INTRODUCTION . . . . . . . .
Short-Term Memory . . . . .
Repetition Tasks . . . . . . .
Stressing in Sentence Imitation Tasks
Age as a Factor . . . . . .
Statement of Problem . . . .

EXPERIMENTAL PROCEDURES . . . .

subjects 0 O I O O O O 0
Design and Stimuli . . . . .
Presentation Procedures . . .
Analysis . . . . . . . .
RESULTS 0 O O O O O O O 0

Effect of Sentence Length . . .
Effect of Sentence Type . . .

Effect of Word Type . . . . .
Effect of Stressing . . . . .
Effect of Grade Level . . . .
Summary of Results . . . . .
DISCUSSION . . . . . . . .
Trends Related to Prior Work . .
Sentence Length . . . . . .
Sentence Type . . . . . . .

Word Type . . . . . . . .
Stress . . . . . . . . .
Grade Level -. . .

Implications for Therapy and Future

Research . . . . . .

Implications for Further Analysis of

Present Data . . . . . .

CONCLUSION . . . . . . . .

Page

H

15
15
16
17

19
19
37
46
49
51
53
56
56
58
61
62
65
65
67

69

Page

LIST OF REFERENCES . . . . . . . . . . . . 71
APPENDICES
APPENDIX A - SENTENCES LISTED ACCORDING TO SENTENCE

TYPE 0 O O O O I O C O O O O O O O O 73
APPENDIX B - SENTENCES CHANGED FROM ORIGINAL TAPE

USED BY BEASLEY AND ACKER . . . . . . . . . 75
APPENDIX C - RANDOMIZATION OF SENTENCE TYPE,

SENTENCE LENGTH, AND STRESS . . . . . . . . 76
APPENDIX D - INSTRUCTIONS GIVEN TO SUBJECTS . . . 77
APPENDIX E - ANSWER FORM USED TO TRANSCRIBE

SUBJECTS' RESPONSES . . . . . . . . . . . 78
APPENDIX F - TABLES OF MEAN PERCENTAGE ERROR SCORES . 79
APPENDIX G ~ COMPUTATION PROCEDURES . . . . . . 82

Vi

L I ST OF TABLES

Table Page

1. Terms and Sources Used to Describe the
Processing Systems for Memory . . . . . . 2

2. Mean Percentage Error Scores for Well-Formed
Sentences by Grade, Word Type, Sentence
Length, and Stress . . . . . . . . . 79

3. Mean Percentage Error Scores for Anomalous
Sentences by Grade, Word Type, Sentence
Length, and Stress . . . . . . . . . 80

4. Mean Percentage Error Scores for Ill-ordered

Sentences by Grade, Word Type, Sentence
Length, and Stress . . . . . . . . . 81

vii

Figure

1.

10.

11.

LIST OF FIGURES

Memory Model Employing Various Concepts
Associated with Memory Systems . .

Mean Percentage Error Scores: Sentence
by Sentence Type and Word Type . .

Mean Percentage Error Scores: Sentence
by Sentence Type and Stress . . .

Mean Percentage Error Scores: Sentence
by Sentence Type and Grade Level . .

Mean Percentage Error Scores: Sentence
by Word Type and Stress . . . . .

Mean Percentage Error Scores: Sentence
by Word Type and Grade Level . . .

Mean Percentage Error Scores: Sentence
by Stress and Grade Level . . . .

Mean Percentage Error Scores: Sentence
by Word Type and Stress . . . . .

Mean Percentage Error Scores: Sentence
by Word Type and Grade Level . . .

Mean Percentage Error Scores: Sentence
by Stress and Grade Level . . . .

Mean Percentage Error Scores: Word Type
by Stress and Grade Level . . . .

viii

Page

21

23

25

27

29

31

39

41

43

48

CHAPTER I

INTRODUCTION

There has been increasing interest in recent years
in determining how children acquire and utilize various
aspects of language behavior. One method often used in
such investigations involves repetition tasks. Further,
recent theoretical considerations require that such tasks
and resultant behavior be related to memory storage systems

for language.

Short-Term Memory,

Several theoretical descriptions and models of
short-term memory are available in the literature (Aaronson,
1967; Broadbent, 1957; Pollack, 1959; Sperling, 1963).

Most of these models View memory as at least a two-stage
process. Table 1 lists several terms associated with models
of memory. Figure 1 depicts a composite view of these
several models for purposes of the present investigation.
The first stage is characterized by a large capacity,
parallel processing system whereby more than one stimulus
item can be dealt with simultaneously; however, the stimuli

decays, i.e., is forgotten rapidly in Stage I. Stage II, at

TABLE l.-—Terms and Sources Used to Describe the Processing
Systems for Memory.

 

Short—Term Memory — Names and Characteristics

 

1. Stage I (Aaronson, 1967)

2. Sensory System (Broadbent, 1957)

3. Performance (Chomsky and Halle, 1968)
4. Primary Memory (Aaronson, 1967)

5. Automatic (Kirk, 1968)

6. Large Capacity (Aaronson, 1967)

7. Rapid Decay (Aaronson, 1967)

8. Parallel System (Aaronson, 1967)

 

Long-Term Memory - Names and Characteristics

 

1. Stage II (Aaronson, 1967)

2. Perceptual System (Broadbent, 1957)
3. Competence (Chomsky and Halle, 1968)
4. Secondary Memory (Aaronson, 1967)

S. Representational (Kirk, 1968)

6. Small Capacity (Aaronson, 1967)

7. Slow Decay (Aaronson, 1967)

8. Single Channel (Aaronson, 1967)

 

 

 

 

 

 

STAGE II REPRESENTATIONAL
SECONDARY MEMORY small capacity

' PERCEPTUAL SLOW DECAY

' COMPETENCE SINGLE CHANNEL

 

 

 

 

INPUT I I

 

Parallel System
Stage I Rapid Decay
Primary Memory LARGE CAPACITY
Sensory System Automatic
Performance

 

 

 

OUTPUT

 

 

FIGURE l.--Memory Model Employing Various Concepts
Associated with Memory Systems.

 

least in the model used for this study, is characterized

by a single channel entrance to a small capacity but slower
decaying processor. While Stage I may be most closely
related to performance, Stage II is most likely related to
competence as defined by Chomsky and Halle (1968).

Studies described in the following section have
been involved in investigating specific aspects of language
acquisition. It is important that if any conclusions are
drawn from the results of such studies, that consideration
be given to the relationship of memory storage, particularly
short-term memory. It seems reasonable to postulate that
various aspects of language behavior be reflective of
short-term memory performance, particularly on relatively
simple repetition tasks. That is, behaviors associated
with language and perception may be developmentally bound,
and such developmental behavior may very well be related
to performance characteristics of short-term memory. Thus,
as the child matures, the burden upon memory processes
becomes less difficult as a result of the child's increasing
ability to handle several characteristics of language
simultaneously. -Because of its relative simplicity, a
repetition task serves as a very useful method for in-

vestigating these processes in children.

Repetition Tasks

 

Brown and Bellugi (1964) found that children seem

to respond to repetition in their original learning of

language. Often, the child's imitation is a reduction of
the model utterance. According to Brown and Bellugi, this
reduction is usually expanded again by the mother (the
model). In spite of the model's expansion or the length
of the utterance, the child would repeat only two to four
morphemes which Brown and Bellugi noted as characteristic
of this age (eighteen to thirty-six months). It would
therefore seem reasonable that some type of repetition task
might be used to analyze which characteristics of speech
input (i.e., stimuli) serve as cues to the child when he
is learning to produce language. There is, however, a
problem with such study. It is possible that children would
repeat a string of words or a sentence without actually
having processed the stimuli through the secondary process-
ing system. (See Figure 1.) If this should occur, then
the results obtained from a repetition task may not reflect
secondary memory language skills (or, moreover, application
of language rules, i.e., competence), but may merely be a
reflection of short-term memory, or, as noted by Schuckers
(1971), a "parroting" effect. Such repetition tasks not
controlling for short—term memory effects may not yield
any differentiating information about the higher processes
of language learning. It is therefore necessary to control
in some way for the child's short-term memory capacity.

One possible control is to use a variety of sequence

compositions such as well-formed, ill-ordered, and anomalous

sentential strings (Martin, 1968; Scholes, 1970). A well-
formed sentence is characterized by a semantically and
syntactically correct utterance. An anomalous sequence
approaches a sentence because it is syntactically correct
but its meaning has been destroyed. An ill-ordered sequence
is neither syntactically nor semantically correct, but all
the words necessary for a well-formed sentence are present.
(See Appendix A.) If a child cannot repeat a five-word
ill-ordered sequence without omissions (i.e., carrot the
bunny eats the), then perhaps it can be assumed, at least
to a first approximation, that his memory span has been
exceeded.

According to Menyuk (1963), a child does not easily
produce or reproduce a non-sentence. (Menyuk's example of
a non-sentence is one which is ill-ordered.) Omissions of
words from such a sequence may indicate that the limits
of short-term memory have been reached (Miller, 1956).

When these short-term memory limits have been defined for

an individual (through an ill-ordered sequence repetition
task), then the attempted repetition of either well-formed
or anomalous sequences (of the same length or longer than
the ill-ordered variety) may be analyzed in terms of re-
flecting higher, secondary memory processes. Thus, on

the most difficult ill-ordered sequence of a reptition task,
the memory system's limit can be defined as that ill-ordered

sequence length at which words are omitted from the string.

If the individual's immediate memory span has not been
exceeded, then there is a possibility that the imitative
response is simply a mimic.

Anomalous sentence types have been included in
imitation tasks andcare included in the present study for
purposes of comparing the results of this study with
previous studies. Anomalous sequences may also yield
information about imitation task processing. If error rate
on anomalous sequences approaches that of well-formed
sentences, then perhaps syntactic structure is as important
to selective retention ability as is semantic form.

Another possible way of controlling for short-term
memory parroting effects would be to insert pause time
between the words of the stimuli (Schuckers, 1971). The
longer time might allow for each word to enter the single
channel system (long-term memory) and be utilized for
perceptual and, subsequently, encoded processing, thus
reflecting upon language competence. (See Figure 2.) The
problem with this approach is that it is undoubtedly
unrealistic to normal language acquisition patterns.
Further, the results of such a task may not lead to infor-
mation concerning what is stored and how it is stored, since
according to the model, the stimuli could either be stored
or could decay depending upon the subject's processing
strategies (such as rehearsal) and/or the amount of pause

time inserted between the words. Further, the prosody

(those elements of stress, accent, and inflection which
accompany the human voice) of the sentential sequences
would be distorted (Schuckers, 1971). This distortion
results from the methodical pause time which destroys
elements of coarticulation and meaning or emphasis which

may be communicated through a particular phrasing pattern.

Stressing in Sentence Imitation Tasks

 

Brown and Bellugi (1964) suggested that the element
of stress may be an important factor in children's acqui-
sition of language. This observation was made when they
were transcribing tapes of mother-child conversations. The
words that the mother said which they could hear most
clearly were usually the words that the child reproduced.
They had trouble hearing the mother's "weakly stressed
functors" (p. l39)(inflections, auxiliary verbs, articles,
prepositions, and conjunctives) and "the child usually
failed to reproduce them" (p. 139). If indeed as Brown
and Bellugi suggest, "English is a well-designed language
that places heavier stress . . . on concentives" (p. 139)
(nouns, verbs, and adjectives), then stress is another
factor which must be controlled in language acquisition
studies.

Blasdell and Jensen (1970) studied the effects of
stress and word-position on thirty- to thirty-six-month-old
children. Each string presented to the children for

repetition consisted of four nonsense syllables which had

one word receiving primary stress, another intermediate
stress, and the third and fourth being least intense, as
measured acoustically and perceptually. Perceptual
judgments were made by five untrained listeners. There
was 100 per cent agreement on the primary stress items

and 96 per cent agreement on the intermediate stress items.
Blasdell and Jensen found that the subjects imitated the
syllables with primary stress and those syllables that
occurred in the final position of the string significantly
more often than the other syllables.

Scholes (1970) investigated the strategy used by
children in their differential retention and deletion of
items during a verbal imitation task. His subjects con-
sisted of eleven children from three years, one month to
four years, six months. He presented sentential strings
of three, four, and five words in length. The length
factor was covaried with a sentence-form factor involving
well-formed, ill-ordered, and anomolous strings. He did
not report the results of length effect upon retention.
However, he did find that deletion of content words by the
children was unaffected by the sentence form (well-formed,
ill-ordered, or anomalous). He also found that functors
were deleted more often from well-formed sentences than
from anomalous or ill-ordered types. He indicated that all

the sentential strings in the experiment were made from

10

citation form readings, which he concluded as resulting

in non-differential word-stress. He concluded that stress-
ing could not be the reason for retention of contentives
because no stressing occurred. However, such a conclusion
is unwarranted since he reported no analysis to support

the contention that stressing did not occur. Also, it is
dangerous to conclude that because a factor is not studied,
its effect upon the experiment is minimal. His results do
not permit the reader to observe whether or not the short-
term memory span of the children involved had been exceeded,
and if so, at what point this might have occurred. Because
he also used nonsense words with real contentives and
functors, it would seem that a three-year-old might only
recognize contentives, which would account for higher
retention of such words.

Beasley and Acker (1971) used a repetition task
with four- and five-year-old children. Sentential sequences
were either well-formed, ill-ordered, or anomalous, and the
lengths of the sequences were three, five, and seven words.
Real words were used in all sequences. In some sequences
the functor was stressed and in other sequences the con-
tentive was stressed. No significant differences in the
number of words deleted were found for the three sequence
types of the three-word length. This would indicate the
possibility of a parroting effect in operation for this

sequence length (Schuckers, 1971). However, in five and

11

seven word sequences, the ill-ordered sentences had more
words deleted than did the anomalous and well-formed
sequences. Unstressed functors were deleted more often
than stressed functors and unstressed contentives were
deleted more often than stressed contentives. The differ-
ence was not as great between stressed and unstressed
contentives as it was for stressed and unstressed functors.
Thus far, the only studies discussed which have controlled
for short-term memory effects in verbal repetition tasks
indicate that stressing appears to be a major factor in
retention of sentential sequences (Beasley and Acker, 1971;
Blasdell and Jensen, 1970).

How could stressing be interpreted in light of the
model (see Figure 2)? The words receiving heavier stress
also probably have a longer duration in relationship to
the other unstressed words of the same length (Fry, 1955).
This additional time could allow for their entrance into
the long-term memory processing system and thus they are
more permanently stored. Also the patterns of stress may
allow for "chunking" to occur, i.e., large amounts of
information may be processed simultaneously (Miller, 1956).
Even though the short-term memory system may be limited,
because of chunking, we can process, understand, and respond
to verbal sequences which may exceed our immediate memory
limits.

In early studies, digit memory spans were measured,

and it was found that as the length of sequence increased,

12

the errors of substitution and omission increased. Miller
(1956) discussed the subject of short-term memory and
sequence length relative to information processing. He
noted that unless some mnemonic technique is employed, such
as chunking, the average sequence memory length of an adult
is seven (words, digits, categories) plus or minus two.
What Miller means by chunking is a method of seemingly
expanding the immediate memory by grouping words or digits
into categories which could be thought of as one. For
example, the sequence 423679485957134 could be repeated by
individuals who could group the digits as 423 (four hundred
twenty-three), 679, 485, 957, 134, thus making a five item
sequence. Perhaps stressed words in a sequence could also
serve as a chunking device. Since Brown and Bellugi have
stated that stressed words are often contentives which, by
definition, are "high information" words, perhaps the only
words necessary to remember or process are the contentives.
The other words are discarded or at least recognized only

briefly or in a scanning fashion.

Age as a Factor

 

Earlier it was noted that several factors may be
operating during language acquisition to enhance the child's
performance on repetition tasks. It is interesting to note
that in the previous studies discussed the subjects were
under five years of age and that no attempt was made by the

reseearchers to test subjects of school age. Do children in

l3

elementary school rely on the same "cues" for repetition
tasks as do children who are just acquiring language? Or,
does their experience with a learned language change their
performance? Is there an age level at which sequence length,
sequence formation (well-formed, ill-ordered, anomalous),
word type (contentive or functor), and/or stressing no longer

affect their responses?

Statement of Problem
When only one or two factors are examined at a

time, the important characteristics of language acquisition
may be obscured and the studies may perhaps lead to erroneous
conclusions. It was therefore the purpose of this study to
investigate the responses of first, third, and fifth grade
children to an imitation task involving simultaneously the
factors of sequence length, word type, sequence composition,
and stressing upon sentence repetition tasks. Specifically,
the following questions were asked relative to an imitation

task:

1. What will be the recall accuracy of first, third,
and fifth grade children for three, five, and seven

word sequences?

2. What will be the recall accuracy of first, third,
and fifth grade children for ill-ordered, anomalous,

and well-formed verbal sequences?

14

What will be the recall accuracy of first, third,
and fifth grade children for contentive and functor

word-types?

What will be the recall accuracy of first, third,
and fifth grade children for stressed and unstressed

verbal stimuli?

Will the above factors and their various levels
interact to produce differential recall accuracy

for first, third, and fifth grade children?

CHAPTER II

EXPERIMENTAL PROCEDURES

Subjects
Ninety first, third, and fifth grade children--

thirty at each grade level--from Mason Public Schools served
as subjects. All children had normal articulation, language,
and hearing as determined by the school's screening records

for speech and hearing.

Design and Stimuli

 

The stimulus materials were the same used in a
previous investigation of verbal imitation behavior by
Beasley and Acker (1971). Eighteen three-, five-, and
seven-word sequences, respectively, were recorded by an
experienced phonetician on an Ampex AG44-4 tape recorder
(frequency response of 50 to 15000 Hz). Six of each set
of eighteen sequences were well-formed, six were anomalous,
and six were ill-ordered in nature. In eighteen of the
fifty-four sequences, one contentive per sentence was
heavily stressed (six within each sentence type and two
per like sentence length); in another eighteen of the

sequences, one functor per sentence was heavily stressed

15

16

(six within each sentence type and two per like sentence
length); and in another eighteen sequences, no lexical item
was heavily stressed. The presence of stress was determined
by listener judgment. The judges were fifteen graduate
students in the department of Audiology and Speech Sciences
at Michigan State University. Beasley and Acker (1971)
found that the judges were able to correctly identify
stressed words in the sequences with a better than 90 per
cent accuracy. For the present study, sentences described
in Appendix B were modified from the original sentences
used by Beasley and Acker. Perceptual judgments on these
sentences also revealed a better than 90 per cent accuracy

of stress judgments.

Presentation Procedures

All sentences were presented in a randomized order
(see Appendix C) via a Sony tape recorder TC-106A. All
subjects heard all sentences and were asked to repeat
exactly what they heard. Each subject sat approximately
twenty-six inches directly in front of the speaker. The
intensity level was set at 70 to 75 dB sound pressure level
(SPL); re: .0002 dynes/cmz. There was occasional peaking
at 80 dB SPL. The ambient noise level in the test room
*was measured at 60 to 65 dB SPL on the C scale of a Brfiel
and Kjaer type 2204 sound level meter using a type 4145

sound field condensor michr0phone. This ambient noise

17

level was sufficiently low so as not to interfere with the
subjects' listening task.

After a brief greeting period, the examiner read
a set of standardized instructions to each subject (see
Appendix D). In addition to the three practice items given
in the standardized instructions, an example of each type
of sentence was given at the beginning of the experimental
tape (see Appendix D). If any practice items were missed,
they were repeated a second time. No subject required
more than one repetition. Each subject was tested indi-
vidually and was allowed fifteen seconds in response time
between stimulus items. Beasley and Acker (1971) found
fifteen seconds to be sufficient time to respond.

The responses of the subjects were written by the
examiner on the answer sheet (see Appendix E). Recordings
were also made of the subject's responses and played back
at a later time. The examiner would check her original
written record against the playback tape. In all but two
words, there was 100 per cent agreement between the original
data and the transcribed data. In those two cases the

error was not scored.

Analysis

 

An error was defined as any word deleted or substi-
tuted. Order errors and addition errors, although tran-

scribed, were not counted as errors.

18

All words in every sentence were classified into
one of four categories: stressed contentive, unstressed
contentive, stressed functor, and unstressed functor. The
raw score errors were then converted into per cent scores

(see Appendix G). The results were graphically displayed.

CHAPTER I I I

RESULTS

This study investigated the effects of sentence
length, sentence type, word type, and stressing upon the
imitation behavior of first, third, and fifth grade children.
The results revealed consistent differences in imitation
scores (i.e., per cent error) between the three grade levels
as a result of the manipulation of the several factors under
study, as well as their respective interactions. These
results, discussed below, can be found labeled in Appendix F
and graphically illustrated in Figures 2-11. A summary of

the results has been included at the end of this chapter.

Effect of Sentence Length

Figures 2-7 indicate that, generally, as the number
of words within the imitated unit increased, the number of
errors in imitation increased. Further, there was a greater
ixucrease in the percentage of error made between the five-
tx> seven-word sentences than between the three- to five-word
sentences .

Figures 2-4 reveal the effect of sentence length as

assc>ciated with sentence type (well-formed, anomalous, and

19

20

FIGURE 2.-—Mean Percentage Error Scores: Sentence Length
(three words long, five words long, seven words
long) by Sentence Type (I = ill-ordered, A =
anomalous, W = well-formed) and Word Type (F =
functor, C = contentives).

MEAN PERCENTAGE OF ERRORS

zzzzgeeexausassw

O'-Nonadhowaodbs

21

I meme
[:1 commwls

IF IC N‘ACWWG IPIGAFAGWWO IFICAFACWFWO
SEVEN WORD FWI WORD TH"! MID

 

FIGURE 3

22

FIGURE 3.--Mean Percentage Error Scores: Sentence Length
(three words long, five words long, seven words
long) by Sentence Type (I = ill-ordered, A =
Anomalous, W = well-formed) and Stress (S =
stressed, U = unstressed).

23

A W

I

"WHWND

     

I wetness»
[j snasszb

A VV
HWWWM

I

 

- -

 

I A \N
NMWWMD

 

 

DannasflnnmwNNWMwMfluuw00765431to
9.9.5 to was»; 3‘

FIGURE 3

24

FIGURE 4.--Mean Percentage Error Scores: Sentence Length
(three words long, five words long, seven words
long) by Sentence Type (I = ill-ordered, A =
anomalous, W = well-formed) and Grade Level
(first grade, third grade, fifth grade).

MEAN PEKENTAGE OF ERRORS

0"” unseat-Ion

833828338882BHSH$3838

 

 

    

25

 

 

 

SEVEN FIVE THREE
FIRST GRADE

 

I ILL-ORDERED
E] mucus
ES WELL-romeo

SEVEN FIVE THREE SEVEN FIVE THREE

THIRD GRADE

FIGURE 4

FIFTH GRADE

26

.Awwmmmuumcc

N D .cmmmmuum N mv wmmuum cam

Auouocsm N m .e>wucwucoo N UV mama who: ma Amcoa

mpuoz ce>mm.wcoH
numcmq mocmucmm

mpuo3 m>flw .mcoH mcuo3 wmunuv
”meuoom Honum emmuceoumm cmszT.m mmeHm

27

 

I Pancreas

 

R382922919§=29¢90mvnN-0
SHOWS :IO 33%|)!an WEN

 

FIGURE 5

28

.Ampmum

numfim .memuo saws» .memum umuwmo H0>0q menus one

Amw>flucmucoo N O .muouocSM N my mama cuoz ma Amcoa

mpuo3 cmbmm .mcoa mpuo3 0>Hm .mcoa mcuo3 mounuv
summed mocmucmw "meuoom uonum mmwpceouem :mmzll.m mmeHm

29

m HMDUHW

35.0 IhuE 355 3:1...
Hut... m2”. zw>mm mums. m2“. zw>mm

 

 

 

 

325958 B
93.52:“. I

 

 

 

mo<ac .55“—

mm‘:.-. w):— E>mw "mac;
.0
f .
I.“
r n
. ¢
r.h
:0
lb
r m
.. V
. o.
u_.
.. N.
u n.
r S
. m.
r 0—

Rub.
L. 0-
inc.
Eon

 

suouua :IO abuNaaaad Nvaw

30

.Ampmnm

nuwflm .mcmum puflzu .mpmum umufimv Hm>mq mpmuw paw

Acmmmmnumcs N D

.pmmmmuum N my mmmuum >3 flmcoH

mpuoz cm>em .mcoH mpuo3 e>wm .mcoH mcuo3 mounuv

numcmq mocmpcmm

"mmuoom Hounm emmucmoumm :mszT.h mmDon

31

h MMDUHm

836 .5: 35.0 2.1...
ARE ms: 53m ”3.: “21 23mm

 

 

 

 
 

 

 

3825:: D
Sage '

35cc 5.:
35:. m2“. 53m ”883

 

 

 

I l l l

l

 

 

l
[I
L

1 l l l
T I U I

fivmbomvnfi- 0

5308213 10 abvmaouad mm

1 J 1 1
I I

I.

I,

f
N:

in.
.3
.m.
1 0.
.b.
.0.
w a.
.0"

 

32

ill-ordered). As can be seen, the error rate increased as
the number of words in the sentence increase for all three
sentence types. In addition, there was a trend for well-
formed sentences to be least affected by sentence length,
followed by anomalous and ill-ordered, respectively. The
single exception to this trend occurred in three-word
sentences. In this case, more errors occurred in three-
word well-formed than in anomalous sequences. This will
be discussed in Chapter IV. However, the greatest error
rate occurred in three-word ill-ordered sentences.

Figures 2, 5, and 6 reveal the effect of sentence
length as associated with word type (functors and con-
tentives). As illustrated, the functors tended to show an
increased error rate as sentence length increased. How-
ever, the contentives showed a similar error rate for
three- and four-word sequences but an increased error rate
for seven-word sequences.

Figures 3, 5, and 7 show the effect of sentence
length in conjunction with stressing. Generally, error
rate increased for both the stressed and unstressed
stimuli as sentence length increased, with the greatest
increase occurring in seven-word sequences. Further, as
sentence length increased, the difference between the
stressed and unstressed stimuli error rate decreased.

Figures 4, 6, and 7 depict the effect of sentence

length as it interacts with grade. As a general trend,

33

error rate decreased as grade level increased for all three
sentence lengths. This effect was most prominent for seven-
word sequences. On three- and five-word sequences, differ-
ences between third and fifth graders were minimal.

Figure 2 depicts the three way interaction of
sentence length with sentence type and word type. The
trend indicated that as sentence length increased, error
rate increased within each sentence type (ill-ordered,
anomalous, well-formed) for both functors and contentives.
The one exception was that the error rate for three-word
well-formed contentive errors was higher than the error
rate for five- and seven-word well-formed contentive
sequences. In seven-word ill-ordered sequences, there was
a higher error rate in functor word types than in contentive
types. A similar relationship between functors and con-
tentives existed for five-word and three-word ill-ordered
sequences. Five-word anomalous error rate was higher for
functor types than for contentive. Little difference
existed between contentive and functor error rates for
seven- and three-word anomalous and for seven-word well-
formed sequences. In three-word well-formed sentences,
contentive errors occurred at a higher rate than functor
errors. There was a trend for ill-ordered sequence error
rates to be higher than anomalous sequence error rates for
functors. Functor error rates associated with anomalous

sequences were higher than well-formed sequences with the

34

exception that three-word anomalous were not higher than
seven-word well-formed sentences.

The same sentence type trends existed for contentives
with the following exceptions: seven word anomalous had
higher error rate than three-word ill-ordered (same as
functor), seven-word well-formed sentences had more errors
than three-word anomalous sequences, and three-word well-
formed sequences demonstrated more errors than seven- and
five-word well-formed or three-word anomalous sequences.

Figure 3 depicts the interaction of sentence length
with sentence type and stressing. Generally, as sentence
length increased within each sentence type, the error rate
increased for unstressed words. An exception occurred in
three-word well-formed sentences for the unstressed con-
dition, where the error rate was higher than for either
seven- or five-word well-formed sentences. As sentence
length increased the error rate increased in ill-ordered
sequences under the stressed condition. No trend could
be seen for length among anomalous sequences for the
stressed condition. Seven-word well-formed sentences had
more errors than five- and three-word well-formed sentences
under the stressed condition.

Figure 4 shows the relationship of sentence length,
sentence type, and grade level. Seven-word ill-ordered
sentences had the highest error rate in each grade level.

For seven-word ill-ordered sequences the error rate

35

decreased as grade increased. Error rate decreased in the
following order for all grades: seven-word ill-ordered
types, five-word ill-ordered types, seven-word anomalous,
three-word ill-ordered. For these conditions, error rate
was higher for first graders than third or fifth graders.
The differences between third and fifth grades were minimal,
except perhaps for seven-word sequences. No grade associ-
ated trend was observed for seven- and three-word well-
formed sentences, or five- and three-word anomalous
sequences. The differences between first, third, and fifth
graders were minimal for all of these conditions. The
lowest error rate for all grades occurred on the five-word
well-formed sentences, and again the differences between
grade levels were minimal.

Figure 5 shows the interaction of sentence length
with word type and stress. The trend showed that as
sentence length decreased, error rate decreased within each
word type stress combination. The exception to this trend
occurred in the stressed contentive combination where it was
found that as sentence length decreased from five words to
three words, the error rate increased. Combinations in-
cluding functors comprised the greatest error rate in
seven-word and five-word sequences. Combinations including
unstressing had the greatest error rate in three-word
sequences. The words with the lowest error rate for each

sentence length were stressed.

36

Figure 6 depicts the results of sequence length by
word type and grade level. For all grade levels, as
sentence length decreased, error rates decreased for
functors. Seven-word sentences with contentives had a
higher error rate than did five- and three-word sentences
for all grade levels. The differences between contentives
in three- and five-word sentences were minimal. For all
grade levels and within each sentence length, there was a
trend for a higher error rate to occur with functors than
with contentives. The exception to this trend was found
in the three-word sentence type for fith graders.

Figure 7 shows the relationship of sentence length
and stressing by grade level. As sentence length decreased,
error rates decreased for stressed and unstressed words at
all grade levels. Stressed words had fewer errors at all
sentence lengths for all grade levels than did unstressed
words. Error rates for first graders were higher for all
sentence lengths and conditions of stressing than were
error rates for third and fifth graders. Error rates for
third graders were greater than error rates for first
graders for seven-word sentences but the differences in
error rates diminished (and in fact were reversed for
three-word sentences with conditions of stress) on five-

and three-word sequences.

37

Effect of Sentence Type

 

Figures 2, 3, 4, 8, 9, and 10 indicate that error
rate was highest for ill-ordered sentences, lower for
anomalous types, and lowest for well-formed sentences.

The relationship between sentence type and sentence
length as shown in Figures 2-4 was revealed earlier. The
trend indicated that error rates increased as sentence
length increased, that more errors occurred in ill-ordered
than anomalous sequences, and that errors in anomalous
sentence types were greater than well-formed types.

Figures 2, 8, and 9 show the influence of sentence
type on word form. The greatest error rate for functors
and contentives were made in ill-ordered sentences. Fewer
word type errors were made on anomalous sentences and well-
formed sequences, respectively. There appeared to be
little difference between contentive and functor errors.
Functor error rates were greater than contentive error
rates on anomalous and ill-ordered sentences, but contentive
error rate was higher than functor error rate on well-formed
sequences.

Figures 3, 8, and 10 show the relationship between
sentence type and stressing. More unstressed word errors
occurred than stressed word errors. The greatest error
rate for both stressed and unstressed conditions occurred
in ill-ordered sentence types. A lower error rate was
demonstrated in anomalous types, and the lowest error rate

occurred in well-formed sentences.

38

.Acemmmnum

N m .Ummmmuumcs N 99 wmmuum paw Amnouocdm
N m .mm>Hucmucoo N UV mama puoz an AcmEHOM
THHD3 N m3 .msonEocm N d .UDHDUHOIHHfl N OHV

mama mucoucmm

umeOUm .HOHHH OOMHCOUHOQ Gmwzll.w mmDOHm

39

 

I FUNCTORs
[:1 CONTENTNES

 

 

 

:E::4:4:::::4+1fi':::;.
822:2£:_§:200b0vaN-0

$110383 :IO SWINE?!” NVBW

 

FIGURE 8

40

.Amomum
numflm .memnm snag» .memum umuflmv H0>0q menus
cam Amm>wusmucoo N O .mnouoc5m N my saws cuoz
an AUmEuowTHHmz N 3 ~msonEocm N < .peumpuouaaw
N Hy mama mocmucmm "mmuoom uouum mmmucmouwm cmwzll.m mmeHm

41

 

 

 

 

 

 

 

 

 

 

Rhflflxflflaflzszsuxndz QVQPODVDN-O
SNOW! :IO abvmaauad MBN

 

A w I A w
FIRST-GRADE THIRD GRADE
FIGURE 9

I

FIFTH GRADE

42

.Ampmum

sumam .mpmum than» .oomum umuHmV Hm>mq mtmuu

Dam Apmmmmnumcs

N D .Dmmmmnum N my mmmnam ma

ADDEDOMTHHD3 N 37.msonEocm N e .DmumDHOTHHH N Hg

mmwe mocmucmm

"mmuoom uouum mmmucmoumm cmszT.OH MMDDHD

SI:

26"

I wsmessso
[:1 STRESSED

 

ANNE-.5

43

 

A VV
FWWGMfl

I

    

A VV

 

TMWQWE

I
FIGURE 10

   

W
W

A
HKTQM

 

I

 

ILILIJ nnnnnnn 111111
u-vuvwliilI1I1IIIIl—1

22:2£IflflzevwbomvnN-°

swwmnflmmwuuuw

44

In Figures 4, 9, and 10 sentence type and grade
level are depicted. All grade levels had the highest error
rate in ill-ordered sentence types, followed by anomalous
sentences and well-formed sentences, respectively. As
grade level increased, there was a trend for error rates to
decrease. The exception to this trend involved anomalous 4
type sentences where third graders made fewer errors than I
fifth graders. 4%
Three three-way interactions involving the effect I
of sentence type with sentence length and word type,
sentence length and stress, and sentence length and grade
level were described earlier (see Figure 2). Generally,
the trends indicated that ill-ordered sequences had higher
error rates than anomalous or well-formed sequences of the
same length, regardless of the condition of stress, word
type, or grade level. Anomalous sequences had more errors
than well-formed sentences for the same length regardless
of stress, word type, or grade level, with the following
exceptions: three-word well-formed sentences had higher
error rates than three-word anomalous sentences if con-
tentive errors only, or errors of unstressing, were
considered.
Figure 8 depicts the interaction of sentence type
with word type and the two conditions of stress. The main
effect showed a higher error rate for ill-ordered sentence

types than for anomalous or well-formed types, regardless

45

of word type or stress. Anomalous sequences had a higher
error rate than well-formed sentences within each combi-
nation of stress and word type. These differences were
less than the differences between ill-ordered and anomalous
sequences. For all three sentence types, the lowest error

rate occurred with stressed contentive type words. No

*1

trends could be seen for the other error rates across the
three sentence types for like word type and stress patterns.
Figure 9 depicts the three-way interaction of
sentence type with word type and grade. Within ill-ordered
and anomalous sentences the functor error rates were higher
than the contentive error rates for all grade levels.
However, for well-formed sentences contentive error rates
were higher than functor error rates for the first grade
level, and little differences appeared for the third and
fifth grade levels. Error rates for ill-ordered sequences
were higher than error rates for anomalous types for both
word types and all grades. Anomalous error rates were
higher than well-formed error rates for both word types in
all grades. Figure 10 also shows that first graders had
higher error rates than third and fifth graders for corre-
sponding sentence types and word types. Third graders had
more errors on contentives and functors in the ill-ordered
sequences than fifth graders. Error rate differences
between third and fifth graders for anomalous and well-

formed sequences and word types were minimal. In

46

well-formed sentences there was little difference between
number of contentive and functor errors for all grades.
Figure 10 shows the relationship of sentence type,
stress, and grade level. There was a trend for more errors
to occur on unstressed words than stressed words over all
grade levels and within all sequence types. More errors
occurred in ill-ordered than in anomalous sequences within
the same condtions of stress for all grade levels. Further,
more errors occurred in anomalous sequences than in well- I
formed sentences within the same conditions for stressing
in all grade levels. Fifth graders had higher error rates
than third graders on anomalous sequences for both stress

and unstressed conditions.

Effect of Word Type

 

Figures 2, 5, 6, 8, 9, and 11 show that the error
rate was greater for functors than contentives. The inter-
action of word type with sentence length and sentence type
has been discussed earlier (see Figure 2).

Figures 5, 8, and 11 depict the interaction between
word type and stress. Generally, unstressed error rates
were higher than stressed error rates for both functors and
contentives. The difference between the error rate for
stressed and unstressed words was greater for contentives
than for functors.

Figures 6, 9, and 11 show the interaction of word

type and grade level. As can be seen, error rate for both

47

.Amvmum sunfim .mpmum DHHED .mpmum

umuflmv Hm>mq mpmuw
N my mmmuum
N mv came who:

can Apmmmeuumcs N D .Dmmmmuum
>9 Amm>wucmucoo N U .muouocsm
"mmuoom uouum mmmucmoumm cmmzll.aa DMDOHm

48

m u

T . ‘0‘
I r‘..illti_

HH MMDme

mag FEE 356 9:1... 355 .52...
m. 3 m 3 m 3

 

 

  
 
 

 

 

3>§Hzoo D
30.52:... I

 

r

U

T 1

I

40t~~0tn<fN1¢|-O

snows so abvmaouad NVHW

 

49

word types decreased as grade level increased. Functors
had a higher error rate than contentives for all three grade
levels.

Three-way interactions combining word type with
sentence length, sentence type, and grade level have been
discussed in detail (see Figure 6). Generally, the functor
error rate exceeded the contentive error rate. The ex-
ceptions, when they occurred, fell in the three-word
sequences and well-formed types.

Figure ll depicts the relationship of word type
stress, and grade level. As can be seen, the unstressed
contentives had a higher error rate than the stressed
contentives over all grade levels. The same relationship
existed in functor word types for first grade. Differences
between stressed and unstressed functors were minimal for
third and fifth graders. There was a trend for error rate
to decrease as grade level increased within each combination
of stress and word type. The exception was in the stressed-
contentive combinations between third and fifth grades,
whereby an inverse relationship occurred. The difference

between third and fifth graders were again minimal.

Effect of Stressing
Figures 3, 5, 7, 8, 10, and 11 show a higher error
rate for unstressed words than for stressed words. The
interaction of stressing with sentence length, sentence

type, and word type has already been discussed (see Figures

1a..
" 1

50

3 and 5). Generally, error rate increased for both the
stressed and unstressed words as sentence length increased.
The error rate for stressed and unstressed words was lowest
for well-formed sentences, followed by anomalous and ill-
ordered sentences, respectively. Usually, unstressed

functors and contentives had higher error rates than

i

stressed functors and contentives.

Figures 7, 10, and 11 show the effect of stressing

by grade level. As grade level increased, errors on
stressed and unstressed words decreased. The percentage
of error on unstressed words was greater than stressed words
at all grade levels.

All possible three-way interactions involving stress
and the other combinations of sentence length, sequence
type word forms, and grade level have been discussed earlier
(see Figures 3, 5, 7, 8, 10, and 11). In looking at stress
as it interacts with sentence length and sentence type,
exceptions to the trend for higher error rates among un-
stressed words occurred in seven-word and three-word
anomalous type sequences, and seven-word and five-word
well-formed type sequences (see Figure 3). In viewing
the interaction of stress, sentence length, and word type,
there was a trend for higher error rates to occur on
functors regardless of stress pattern, with the exception
of the stressed functor word type in three-word sequences

(see Figure 5). In looking at stress as it related to

51

sentence length and grade level, consistent patterns were
found in error rates of stressing associated with sentence
length and grade. All unstressed words had a higher error
rate than the stressed words within the same sequence
length at all grade levels (see Figure 7). As stress
interacts with sentence type and word type, one consistent
trend indicated that stressed contentives had the lowest

error rate for all sequence types. More errors were made

‘4

in ill-ordered sequences for all combinations of word type
and stress than in anomalous and well-formed sequences (see
Figure 8).

Finally, when stress interacts with sentence type
and grade level, all unstressed words had higher error
rates than stressed words within all sentence types for all
grades. Error rate again decreased as grade level increased

(see Figure 10).

Effect of Grade Level

Figures 4, 6, 7, 9, 10, and 11 indicate that the
number of errors decreased as the grade level increased.
The greatest decrease in errors occurred between the first
and third grade level.

As reported earlier, grade level interactions with
sentence length, sentence type, word type, and stress were
shown in Figures 4, 6, 7, 9, 10, and 11.

Briefly, as can be seen in Figures 5, 7, and 8,

errors increased as sentence length increased over all grade

52

levels. The increase was greatest for three- and seven-word
sentences. Third and fifth graders showed little difference
in percentage of errors for three- and five-word sentences.
Figures 4, 9, and 10 show that first grade children
had more errors on every sentence type than did third and
fifth graders. Third graders tended to do better on
anomalous types than did fifth graders. (

Figures 6, 9, and 11 show that functors had a higher Eu

 

error rate than contentives for all grade levels. As level
increased error rate decreased for both contentives and
functors.

Figures 7, lO, and 11 show a higher error rate on
unstressed words than on stressed words for all grade levels.

Three-way interactions involving all possible com-
binations of sentence length, sentence type, word type,
stress, and grade have also been discussed above (see
Figures 4, 6, 7, 9, 10, and 11).

The main effect present in all three-way inter-
actions involving grade level showed that first graders made
more errors than third and fifth graders within each
specific combination of variables. The differences between
third and fifth graders were minimal with the following
exceptions: seven-word ill-ordered sequences (Figures 4, 9,
and 10), seven-word contentives and functors (Figure 11),
seven-word stressed and unstressed conditions (Figure 7),
ill-ordered functors and contentives (Figure 9), and ill-

ordered stressed and unstressed conditions (Figure 10). In

53

these cases more than minimal differences existed and
followed the general trend for all grade level comparisons:

as grade level increased, errors decreased.

Summary of Results

 

The main effects which seemed most prominent were
sentence length, sentence type, word type, and grade level.
Briefly, error rates increased as sentence length increased.
Error rates were greatest for ill-ordered sequences, followed
by anomalous and well—formed sequences, respectively. Con-
tentives had lower error rates than functors. Finally,
error rates decreased as grade level increased.

The two-way interactions which seemed most profound
were: sentence length by sentence type, sentence length
by word type, stress by word type, sentence length by grade
level, and sentence type by grade level.

Sentence length and sentence type generally followed
this pattern: seven-word sequences had the highest error
rate over all sentence types, followed by five-word and
three-word sequences, respectively. Within each sentence
length, ill-ordered sequences had the highest error rate,
followed by anomalous and well-formed sequences, respective-
ly. The exception occurred in well—formed, three-word
sequences where the error rate was higher than five- and
seven-word, well-formed sequences and than three-word,

anomalous sequences.

54

Sentence length interacted with word type in three-
word sequences where error rates for contentives and
functors were similar. In seven- and five-word sequences,
functor error rates were higher than contentive error rates.

The difference between stressed and unstressed error
rates were minimal for functor word types and greater for
contentive word types. Unstressed error rates remained
greater than stressed error rates.

First graders had greater error rates than third
and fifth graders, respectively, on seven-word sequences.

On five- and three-word sequences, however, the differences
in error rate between third and fifth graders were minimal.

Generally, error rate decreased as grade level
increased for all sequence types; however, in anomalous
sequences, fifth graders had a higher error rate than did
third graders.

Three-way interactions which seemed most evident
were: sentence length by sentence type and stress, sentence
length by sentence type and word type, and sentence length
by word type and stress.

In seven-word sequences, unstressed error rates
were lower than stressed error rates for ill-ordered and
well-formed sequences. While in anomalous sequences, the
error rate for stressed words was lower than for unstressed
words. In five-word sequences, unstressed error rates were
higher than stressed error rates in ill-ordered and anomalous

sequences, but in well-formed sequences, the stressed and

 

55

unstressed error rates were about the same. In three-word
sequences, unstressed error rates were higher than stressed
error rates for ill—ordered and well-formed sequences, but
in anomalous sequences, unstressed error rates were lower
than stressed error rates.
An interaction of sentence length, sentence type, FA
and word type occurred in three—word, well-formed sequences. i!
In this case, contentive error rates were higher than i
functor error rates. Usually, functor error rates were
higher than contentive error rates within all three sequence
lengths and across all three sequence types.
Finally, sentence length, word type, and stress
interact. In seven-word sequences, stressed functors
showed higher error rate than unstressed functors. Also,
stressed contentives had higher error rates in three-word
sequences than in five-word sequences.
While other interactions were reported in the body
of the results chapter, this summary mentions the most

striking data.

CHAPTER IV

DISCUSSION

The results indicated that sentence length, sentence
type, word type, stress, and grade level interacted at
various levels. Further, specific variables in certain
cases took precedence over others in affecting the recall
accuracy of children on this imitation task. Earlier
studies had not investigated the simultaneous interactions
of these several variables upon imitation behavior, thus
limiting possible conclusions and generalizations about
imitation behavior. The present study has, to a large
extent, overcome that problem. However, in order to
generalize about the behavior of individuals, trends must
be found and, further, any exceptions to those trends must
be explained in the same context as the trends. The trends
of the present study have been identified and the exceptions
mentioned. Now, the trends will be related to prior work,

and the exceptions will be discussed.

Trends Related to Prior Work
In viewing the variables independently of each

other, the results of the present study support earlier

56

 

57

investigations. The trend for error rates to increase as
sentence length increased supports the findings of earlier
studies of sentence length, item length, and short-term

memory (Aaronson, 1967; Miller, 1956; Schuckers, Shriner,

and Daniloff, 1971). When a string of items (words or

 

E:- El.

digits) has exceeded the immediate or short-term memory

span of an individual, that individual tends to delete

words or digits from the string. Miller (1956) found that
the short-term memory span was usually seven items long,
plus or minus two. This present study found that more

words were deleted or incorrectly remembered from seven-word
sequences, followed by five- and three-word sequences,
respectively.

The fact that more errors occurred in ill-ordered
sequences, followed by anomalous and well-formed sequences,
respectively, supports the studies of Beasley and Acker
(1971), Brown and Bellugi (1964), and Scholes (1970). The
trend for higher rates to occur on unstressed words than on
stressed words also agrees with earlier studies (Beasley
and Acker, 1971; Blasdell and Jensen, 1970). Finally,
error rated decreased as grade level increased. Previous
studies dealing with some combination of the five factors
under investigation in the present study did not use
school age children as subjects. However, the same trends
had been found with pre-schoolers: As age increased, error

rate decreased (Beasley and Acker, 1971; Scholes, 1970).

58

Sentence Length

As the results indicated, error rate increased as
sentence length increased. It was evident that there was
a greater error rate increase between the five- and seven-
word sequences than there was between the three- and five-
word sequences. When word length increased to seven, then
the increase in percentage of errors may have indicated
that short-term memory storage capacity and/or processing
capability had been exceeded. When immediate memory limits
have been exceeded, then recall becomes selective and must
reflect strategies used by the listener in processing the
information he must repeat or imitate. Further support
that seven-word sequences exceeded short-term memory span
can be seen when total sequence deletions were counted.
Twenty-seven seven-word sentences were completely deleted
while only four sentences were completely deleted in both
five- and three-word sequences, respectively. The type of
words retained by the subjects on seven-word sequences may
very well reflect upon the strategies used by individuals
to select words to repeat or imitate. If a trend emerges
in the selection of words to be retained, it is plausible
that individuals use similar cues in other tasks which
make use of imitation.

Sentence length as it interrelates with sentence
type presents an interesting phenomenon. Within five- and

seven-word sentences, more errors were found in ill-ordered

59

types, followed by anomalous and well-formed, respectively.

This does seem to indicate that there is a distinction

between syntactically correct versus semantically meaningful

sentences in terms of word retention. That is, the im-

portance of syntactic completeness for recall accuracy is

emphasized. An exception to the trend occurred in three- F4
word sequences where well—formed sentences had more errors

than the anomalous types. It appears unreasonable that the

 

length/type combination which should be the easiest to I
imitate fails to support the trend. Although, children have
been found to respond better to commands above their own
level of production (Shipley, Smith, and Gleitman, 1969);
responses of children in an imitation task may or may not
follow the same pattern. In the present study, the follow-
ing offers one possible explanation for 75 per cent of the
errors: There are six, three-word well-formed sentences,
and upon closer examination of these six sentences it was
found that, over all, fifty—six errors were made in these
sentences. Out of the fiftyesix errors, forty-one occurred
in the same word in one sentence. The sentence was, "Her

doll cries,‘ and the substitution was, "Her dog cries."
Further, no other errors occurred in that particular
sentence. Due to the characteristics of the consonant
cluster llgg ([lk]) in doll_g£ies, and the manner in which

such a cluster is coarticulated, it is probable that the

point of articulation for the [l] is closer to the point

60 i

of articulation for [k]. Thus, the perception of the sound
at the end of dgll is influenced by the beginning of the
word cries. The voiced element of [1] remains, and with
the placement of the tongue for [k], the new sound becomes
perceived as a [g] (the sound which assimilates some
features of both [1] and [k]). This explanation for the
effect of coarticulation is supported by the findings of
Manning (1972). Generally, the point of articulation of
consonant sounds are modified by neighboring sounds in a
consonant cluster, whether these clusters occur between
words or within words.

Sentence length interacted with functor word types
differently than contentive word types. In the case of
functors, error rate increased with length, whereas with
contentives, error rate showed a marked increase only in
seven-word sentences. When sentence length approached
and/or exceeded the short-term memory capacity (as may be
the case in seven-word sentences), then it appeared that
word form was not a differential cue in the retention of
words.

The relationship between grade and sentence length
suggests a developmental process for short-term memory.
Short-term memory capacity imposed greater limits upon
first graders than on third and fifth graders. That is,
it appears that short-term memory use increased with age.

It appears from the data that short-term memory processing

61

is limited at first grade level, and that a leveling of this
processing ability may occur between third and fifth grade.
However, further research on seventh and ninth graders would
be necessary before it could be determined if a leveling

or merely a plateau for the development of short-term memory

 

had been achieved by fifth grade. A

Sentence Type

 

When sentence type, sentence length, and word type 4
were compared (see Figure 2), the interaction between the
three variables indicated that sentence type appeared to
cause more difficulty than sentence length on the imitation
task. The three-word ill-ordered sequence had a higher
error rate than the seven-word anomalous types. The same
trend existed when sentence type, sentence length, and
stress were compared. Since ill-ordered sequences were
more difficult to reproduce in this imitation task study,
followed by anomalous and well-formed, respectively, then
some cues for differential retention of words must come
from the sequence structure itself. Since syntactically
correct but semantically meaningless sequences were harder
to imitate than well-formed types but easier than ill-
ordered sequences, then some cues must come from syntactic
structures as well as semantic ones. To decide what these
cues are, further analysis of the types of errors made,

based on order of sequence, would have to be carried out.

62

Wordeype

 

It would seem reasonable that children rely upon
"educated" guessing from context and/or that they learn
to expect certain elements of a sentence to be present
without having to really process every word separately.
Perhaps certain words are perceived from the physical
setting rather than the auditory signal. This helps ex-
plain why children tend to make more errors on function
words than content words. Functors tend to need a specific
context in order to be meaningful. There is little differ-
ence whether one says "Take the book" or "Take a book"
unless there is a specific meaning attached to the sentence
by the context in which the sentence is used (i.e., 932
book on a shelf as opposed to several books on a shelf).
This context was not a variable under study in the present
investigation. Therefore, the functors were harder to
guess. The word take could not be guessed from context,
as a choice between leaving the book or taking the book
would have to rely on an auditory signal. B295, on the
other hand, may or may not be guessed from context. If
the only available item in sight were a book, then the
word could be guessed from context; but if there were other
items on the shelf, such as a cup or a ball, then the word
2295 could not be as easily guessed from context. A
reasonable question to ask at this point would be: Is

there any differentiating cue made by the speaker when

63 5

pronouncing words that cannot be guessed from context? The
answer may suggest that stressing plays an important role
in this process. If, for example, there was only one book
on the shelf, the speaker may have said, "Take the book,"
with natural stress falling on the word take. If there
were a book, a cup, and a ball on the shelf, then the
speaker may have said, "Take the 222k," with primary stress
falling on bggk and secondary stress falling on take.

Further research is needed to support this hypothesis.

Stress

Perhaps stress could be put into better perspective
by further examination of the interaction between sentence
length, word type, and stress (see Figure 5). When stress
and word type were compared for all three lengths, functor
word types had the highest error rate (two functors were
unstressed and one functor was stressed). The lowest
error rate occurred on stressed words for all three lengths
(two were stressed contentives and one was a stressed
functor).

When stress and word type are considered together
as shown in Figures 5, 8, and 11, it was found the differ—
ences.between stressed and unstressed error rates were
greater than the differences between contentive and functor
error rates. This data was contrary to statements made
about word types and stress by Scholes (1970). It was

apparent that the differences in error rate were greater

64

between stressed and unstressed words within the same
sequence length and sequence type than the error rate
between contentives and functors for the same sequence
length and sequence type. This data indicated, then, that
stress was primary to word form as a basis for differential
retention of words.

The interaction of sentence type and word type
(see Figure 9) as compared with the interaction of sentence
type and stressing (see Figure 11) further supported the
contention that stressing played a more important part than
word type, at least in the present imitation task. The
error rate difference between functors and contentives was
less than the difference between stressed and unstressed
words. The greatest difference in error rate for stressed
and unstressed words occurred in ill-ordered sequences and
in seven-word length sentences. It was at these points
that stress appeared to serve their most important function
in the retention ability of the subjects. When short-term
memory had been exceeded (increased sentence length) and
task difficulty increased (ill-ordered sequence types),
as indicated by an over-all increase in error rate, the
error rate differences between stressed and unstressed
words were the greatest. This would seem to indicate that
stress played a role in the retention ability of the
subjects for non-normal sentences. In first language

learning, children deal with adult models which usually

65

exceed the child's memory span. It seems reasonable that
stress would also aid the child in the reduction of the

adult model (Brown and Bellugi, 1964). Stress may not,
however, have a significant function in well-formed sentences

(Lieberman, 1965).

Grade Level 4

 

When interactions involving grade level were
examined, the only exception to the trend occurred in '
anomalous sentence types than did third graders. Perhaps
fifth graders recognized the syntactic correctness of
anomalous sentences and tried to make them more well-formed.
anomalous sentences and tried to make them more well-formed.

Implications for Therapy
and Future Research

As the data showed, stress was an important cue in
the retention of words on the present imitation task study.
Therapy, making use of imitation techniques, should include
the use of stress in its program to help cue individuals to
remember and produce words usually deleted or in error.

Further research should be conducted in clinical
settings to see whether or not language and speech habili-
tation programs which make conscious use of stress as a
cue in correction and expansion models of therapy are
successful in increasing short-term memory span. Further,
the role of stress in improvement of speech and language

production should be examined.

66

Implications for therapy and for the classroom
teacher arise from the first grade limitation on short-term
memory. It seems that in any learning task, the child is
only processing shorter units effectively. This may ex-
plain why it is so difficult for first graders to "follow
directions" and why many first grade teachers devote the
whole year to this task. If a child is limited in his
processing of longer sequences, then it may follow that
with repetition, the child can eventually acquire the
meaning out of the longer sentence, particularly if there
is a way of directing his attention to the words he missed
on preceding trials.

Results of grade level trends indicated a need for
future research including seventh and ninth grade levels
in this present study. Although the error rate differences
of third and fifth graders were minimal, the general trend
for error rate to decrease as grade level increased was
still visible. This leveling of error rate needs to be
further stabilized before statements can be made about the
processing abilities and differential retention abilities
of third and fifth graders as compared to the adults'
abilities.

The area of mental retardation is in great need of
some guidance in the area of speech and language programs.
This present study should be carried out with a population

of mentally retarded to see how the trends develop. A

67

comparison should be made between the trends found in the
normal population and the trends depicted by the mentally
retarded population. If similar patterns emerged, then
programs for the retarded could parallel regular language
programs. If, however, the trends were non-existent or
very different from the normal population, then language
programs for the mentally retarded should be based on cues
which aid the retarded individual's optimal retention
behavior. The results of the present study (if presented
to a mentally retarded population) may yield implications
for therapy or may only yield implications for further
research. In either case, something definite could be said
about the mentally retarded population with regard to short-
term memory and differential cues utilized in imitation
behavior as affected by mental age. This in itself would
contribute considerably to the literature in the area of
mental retardation and language development.

Implications for Further Analysis
of Present Data

 

 

In order to obtain trends for the interactions of
the variables under study, it was necessary to score words
as either correct or incorrect. However, information about
type of phonetic, phonemic, syntactic, and semantic mis-
perception could very probably be gleaned from further
analysis of the various substitution and deletion errors.

For example, information regarding the development of the

 

68

phonological rule for plurals could be obtained by analysis
of the words where the plural case was changed by the
subjects. The same type of analysis could be conducted

for coarticulation effects, comparing universal sound
segments with non-universal sound segments, order errors

in relationship to syntax especially in ill-ordered sentence
types, and word substitutions in relationship to semantics
in anomalous sentence types. Phonetic features of sound
substitutions could be compared with those of the sounds
they replaced to see if the distinctive features were

modified.

 

CHAPTER V

CONCLUSION

The results of this study, as demonstrated by the
subject's differential retention of words in a sentence
imitation task indicated that sentence length, sentence
type, word type, stress, and grade level of the child did
interact at various levels.

The results showed that error rate increased as
sentence length increased. Further, ill-ordered sequences
had greater error rates than anomalous and well-formed types,
respectively. The main effect for word type indicated that
a higher percentage of errors occurred on functor word
types more than on contentive word types. The effect for
stress indicated that unstressed words had greater error
rates than stressed words. Finally, error rate decreased
as grade level increased.

Exceptions to these trends occurred when inter-
actions between sentence length and sentence type were
examined. Ill-ordered sequences had higher error rates
at all lengths than any other length-type combination.

Another exception to the trend involving sentence length

69

70

and sentence type was found in three-word well-formed
sequences where upon closer examination, it appeared that
coarticulation might have played an important role in the
processing of that length-type combination.

Sentence type interacted with grade level where
fifth graders made more errors on anomalous sequences than
on ill-ordered sequences.

Interactions involving word type and stress
indicated that in ill-ordered and well-formed sequences,
stressed functor error rates were higher than unstressed
functor error rates. Further, unstressed contentive error
rates were higher than stressed error rates.

Results of grade level trends indicated a need
for further research. Although the error rate differences
of third and fifth graders were minimal, the general trend
for continued decreasing error rates needs to be stabilized
by testing the recall ability of seventh and ninth graders

on this present imitation task study.

LIST OF REFERENCES

 

Aaronson, D. "Temporal Factors in Perception and Short-Term
Memory." Psychological Bulletin, LXVII (1967), 130-
44.

I“ ..x“‘. .8-E
4:.
-

Beasley, D., and Acker, J. "The Effect of Sentence Length,
Sentence Type, Word Type and Stress on Four and Five
Year Olds." Unpublished study, Michigan State
University, 1971.

Blasdell, R., and Jensen, P. "Stress and Word-Postion as
Determinants of Imitation in First-Language Learners."
Journal of Speech and Hearing Research, XIII (1970).

Broadbent, D. "A Mechanical Model for Human Attention and
Immediate Memory." Psychological Review, LXIV (1957),

 

205-15.
Brown and Bellugi. "Three Processes in the Child's Acqui-
sition of Syntax." Harvard Educational Review,

 

XXXIV (1964), 133-51.

Chomsky, N., and Halle, M. The Sound Pattern of English.
New York: Harper and Row, 1968.

 

Fry, D. B. "Duration and Intensity as Physical Correlates
of Linguistic Stress." Journal of the Acoustical
Society of America, XXVII, No. 4 (1955), 765-69.

 

Lieberman, P. "On the Acoustic Basis of the Perception of
Intonation by Linguists." Word, XXI, No. l (1965),
40-54.

Manning, W. H. "Coarticulation as an Adjunct to Speech
Perception." Unpublished study, Michigan State
University, 1972.

Martin, J. "Temporal Word Spacing and the Perception of
Ordinary, Anomalous, and Scrambled Strings."
Jougpal of_ygrbal Learning and Verbal Behavior, IX
(1970), 167-70.

71

72

Menyuk, P. "A Preliminary Evaluation of Grammatical Capacity
in Children." Journal of Verbal Learningiand Verbal
Behavior, II (1963), 429-39.

Miller, G. "The Magical Number Seven, Plus or Minus Two:
Some Limits on Our Capacity for Processing Infor-
mation." Psychological Review, LXIII (1956), 81-97.

Pollack, I. "Message Uncertainty and Message Reception."
Journal of the Acoustical Societypof America, XXXI
(1959), 1500-08.

Scholes, R. "On Functors and Contentives in Children's
Imitations of Word Strings." Journal of Verbal
Learning and Verbal Behavior, IX (1970), 167-70.

Schuckers, G.; Shriner, T.; and Daniloff, R. "Auditory
Reassembly of Segmented Sentences by Children."
Submitted to Journal of Speech and Hearinngesearch,
(1971).

Shipley, Smith, and Gleitman. "A Study in the Acquisition
of Language." Language, XLV, No. 2 (1969), 332-42.

Sperling, G. "A Model for Visual Memory Tasks." Human

.m...\u_—.v—s an“

APPENDICES

APPENDIX A

SENTENCES LISTED ACCORDING TO SENTENCE TYPE

APPENDIX A

SENTENCES LISTED ACCORDING TO SENTENCE TYPE

Well-Formed

 

My mother washes the dishes.

My dog runs.

The boy plays.

Her doll cries.

The little boy pulls his red wagon.
His brother climbs a tree.

My black dog likes the new bone.
Her big sister wants a new dress.
The pretty teacher reads a funny story.
The other girl lost her blue ball.
Your father drives a car.

Take the book.

The boy rides a bike.

Wash your face.

My cat drinks the milk.

Throw the ball.

The old man takes a long walk.

The girl wears a hat.

Ill-Ordered

 

His purrs cat.

Eats bear the fresh honey the brown.
Man dirty nice cleans shoes the his.
Balloon happy the clown carries funny a.
Hits car train the a.

My bring bike.

Milk spills the.

Choose friend a.

Door close the.

Carrot the bunny eats the.

Kitten chases my her dog.

Covers blanket baby the yellow the bed.
Washes a the man big green car.

73

74

Spins toy your.

Brings a flowers lady the.

Baby his calls the mother.
Sister my the spills water.
Girl red cow silly draws the a.

Anomalous

 

The tall tree washes the dirty dog.

The chair flies.

My yellow dress cries a large tear.

Her big brother sings his black shoe.
Read a paint.

Drink the table.

An airplane chops the wood.

The dog purrs his bone.

A leaf walks.

Comb your teeth.

A nice man wears his new horse.

Her bike eats an apple.

The happy children drink a green spoon.
The tree hops.

His sister bakes a book.

The pretty picture colors a little boy.
The chair plays a drum.

The lady walks her coat.

 

APPENDIX B

SENTENCES CHANGED FROM ORIGINAL TAPE
USED BY BEASLEY AND ACKER

APPENDIX B

SENTENCES CHANGED FROM ORIGINAL TAPE

USED BY BEASLEY AND ACKER

Sentence Number 20
Was: Write letter a.

Changed to: Choose friend a.

Sentence Number 34
Was: A bird bounces.

Changed to: A leaf walks.

Sentence Number 44
Was: The fish hops.

Changed to: The tree hops.

75

APPENDIX C

RANDOMIZATION OF SENTENCE TYPE,
SENTENCE LENGTH, AND STRESS

APPENDIX C

RANDOMIZATION OF SENTENCE TYPE,
SENTENCE LENGTH, AND STRESS

Sentence Type No. of the Sentence

well-Formed 1, 4, 5, 6, 9, 17, 18, 23, 25, 30,
35, 36, 39, 45, 46, 50, 53, 54.

Ill—ordered 2, 10, 11, 14, 15, 16, 19, 20, 21,
22, 26, 27, 28, 29, 31, 37, 38, 52.

Anomalous 2, 6, 8, 12, 13, 24, 32, 33, 34,
40, 41, 42, 43, 44, 47, 48, 49, 51.

Sentence Length

 

Three-word 2, 4, 6, 7, 9, 13, 16, 19, 20, 21,
24, 29, 34, 39, 40, 44, 46, 53.

Five-word l, 5, 15, 18, 22, 26, 31, 33, 36,
37, 38, 42, 45, 47, 49, 50, 51.

Seven-Word 3, 8, 10, ll, 12, l4, 17, 23, 25,
27, 28, 30, 35, 41, 43, 48, 52, 54.

Stressing

Contentives Stressed 2, 7, 10, l3, 14, 18, 21, 23, 26,
36, 37, 39, 42, 43, 44, 48, 49, 54.

Functors Stressed l, 3, 5, 9, 15, 16, 24, 28, 30, 34,
35, 38, 41, 46, 47, 51, 52.

No Stress 4, 6, 8, ll, 12, l7, 19, 20, 22,
25, 27, 29, 31, 32, 33, 40, 45, 50,
53.

76

APPENDIX D

INSTRUCTIONS GIVEN TO SUBJECTS

APPENDIX D

INSTRUCTIONS GIVEN TO SUBJECTS

I want you to listen to what the man is saying on
the tape recorder and then tell me exactly what you heard
him say. If you can't remember everything, tell me as much
of it as you can. For example, if the man said, "He likes
food," what would you say? .
Good! Let's try another. If the man said, "Find go him,"
what would you say? . Good!
Let's do one more. If the man said, "The door plays," what
would you say? . Fine, let's

begin.

77

APPENDIX E

ANSWER FORM USED TO TRANSCRIBE
SUBJECTS' RESPONSES

78

.xamz mcoH m mwxmu awe 0H0 one .em .pwn may soaam» may anon uwxcman mum>ou .bm

 

 

.Hamn may songs .mm .606 go: as mmmmao cmuuas .mm

.m on» m3mup madam 300 pmu Huwu .Nm .mmwup 3mg m mucmz,umumwm man aw: .mm

.umoo um: mxams moms one .Hm .maamu man sauna .vm

.xHHE may mxcwup you a: .om .mcoa so: on» mmxwa moo Roman >2 .mm

.Esup m wwwam uwwco may .mv .mnu mums wanna may uouumo .NN

.mon mauuwa m muoaoo mucuoam Suumum one .mv .mnu mmoao Moon .HN
.xoon m moxmn umumwm mam .bv .m pcmwuw mmoonu .ON

.006“ uses ammz .64 .mau mHHHmm gas: .ma

.oxua.m much“ son was .ms .mmuu m mnsuao umnuoun mum .mH

.mmoa 00H» mna .wv .comw: 60H man madam hon mauuwa was .ba

.coomm cmwum m xcwup coupawco humus one .me .waQ ucwhn m2 .wH
.mammm cm mung mwa Hum .Nv .m ms» cash» you muwm .ma

.mmuon 36: man mumms cue wows m .HV .0 macaw pmwuumo csoao may mama: cooHHmm .vH

.nummu use» £800 .ov .ucwmm m comm .MH

.xoon on» mxma .am .mosm soman mun macaw nonuoun mun um: .NH

.kums maawmm on» ma umumfim .mm .mwn ms» muonm mcmwao wows muuwp cm: .HH

.Hmcuos on» maamo mac anon .hm .czoun may mmcoc anon“ on» Hump mumm .OH

.umo m mm>wup Honumm “sow .om .mowuo Haop um: .m

.Hamn msan no: umoa Huwm nonuo one .mm .ummu wmumH m mmwuo mmoup scaamu MS .m
.mxams mama a .4m .mmmam son was .5

.mcon mac manna mow 0:9 .mm .mmwam Hanna was .w

.poo3 may muono occamuwm c< .Nm .umn m musm3 Huwm was .m

.05» Sosa mumsoam m mmcflum .Hm .mcsu moo as .q

.auoum macaw m mummy uwcomwu muumum one .om .moo muuwp on» mmcmms 00H» Ham» one .m
.Hsom mou mcamm .m~ .umo manna mam .m

.umo cmwum man cos on» m mmsmms .wm .mmnmflp on» mmnmw3 “dance as .H

.mxamu xmmo was .mam wan mH .mumm Hugo page ”mzuaH moHaoamm
.moz .mmw med mmmodme madam *m -m24z.

mmmzommmm .mBUfihflDm mmHmumZ¢M9 OE DmmD ZMOh mm3m2¢

m Xanmmmd

APPENDIX F

TABLES OF MEAN PERCENTAGE ERROR SCORES

'79

 

 

o.~ v.H H.H o.~ m.o m.n m.m Hmuoe

n.H m.o m.H o.o H.H ~.~ o.o «.m H.v ~.~ unwon

.m.~ w.o H.H o.c n.o v.H c.o m.o m.o n.o n

H. o.o o.o o.o m.o m.o o.o o.c o.o c.o m

m.~ ~.~ m.q o.o q.~ 5.. c.o o.m o.m o.o m mm>wucmucoo
m.~ m.H o.H ~.N m.H m.H R.H m.m m.~ 4.4 ucfion

n.m m.~ h.a m.m m.H m.H R.H ~.> m.v o.o~ a

~.H m.H m.H p.H m.m >.H m.m ~.H m.~ o.o m

¢.H m.c c.o n.H m.H m.~ o.o H.~ m.o m.m m muouocsm

vegan pummouumca pummmuum unwon vwmmouumco 00mmmuum ucMOH pmmmmuumca Gammouum

Hmuos guano; moms cues

 

mcmuw nuuwm

menus cause

menus umuum

 

mpuo3 cm>wm .qcoH mono: w>wm .mcoH mpuo3 owucuv Sumcmq wocmucwm .amo>wucmucoo can mucuocsmv mmha who:

.Apmmmmuumcs can pommmuumv mmmuum can xAOCOa

.awcmuo numam .opmum pagan .mcmum umufluv upmuu Sn mmocoucom ousuonuaamz How mwuoom Houum wmmucmuuum cmwzlu.~ mqmme

8C)

 

 

0.0 m.m 0.N v.m 0.N H.b 0.m HMUOB

0.N m.N m.m h.H @.H m.N 0.0 m.v h.m m.m “GHOH

0.v 0.N m.m 0.0 h.N ¢.m 0.0 0.0 m.NH 0.m h

N.H N.H m.N 0.0 0.0 m.H 0.0 m.H H.m 0.0 m

0.N m.m 0.N 0.m N.H h.0 b.H 0.m 0.H 0.m m m0>wUCOUGOU
0.v H.v m.v 0.m 0.m m.¢ m.m m.w v.0 m.v uCflOb

h.m m.m 0.0 h.H h.m m.h 0.0 m.m M.NH 5.0 h

0.0 0.0 m.m 0.0 v.m 0.v 5.0 N.m M.HH 0.m m

0.N h.H h.H h.H m.N h.H m.m h.H h.H h.H m mHOHOQfih

unw0h oommmuumco pmwmuuum ucfion nonmouumco commouum unwed commuuuuca nanomuum
Hmuoa numcwg mmha who:

 

mucus chasm momma ensue

ocmuu uuuwm

 

.AcOmmuuumca can pwmmwuumv umouam can .AmcoH
mpuoz c0>wm .wcoH mono: w>wm .mcoH mpuoz mounuv cumsoq oucmucmw .amw>wacmucoo can nuouocdwv wmha who:
.Ampwum sumau .mpmum pugs» .mpmum umuwwv mpmuu an moocmucom macawabnd How auuoum uouum mmmucwoumm cumzll.m mam¢a

81

 

 

 

0.GH 0.0 0.5 w.ma m.NH m.NN 5.5M HMHOB

0.HH 5.0 0.0 ¢.¢ 0.0 N.NH m.m 0.5H 0.0N 0.¢H “GMOH

0.MN H.mH N.mH 0.0H m.NN m.5N 5.0M v.mn 5.Dm 0.0m 5

N.m m.m m.5 0.0 0.0 m.5 0.0 0.0 5.NH m.m m

0.m m.m m.m m.m 5.0 m.H 0.0 0.HH n.0H 5.HH m mwbwficflucoo
0.0H 0.0H H.0H H.HH 0.0H m.vH 0.0H H.MN 5.mN m.0N HGHOH

H.¢m m.m~ 0.5a 0.0m 0.0m 0.5m 0.mv 0.Nv m.NV m.mv 5

H.HH 0.0 5.0 m.m 5.0 0.0 m.m m.0H 5.0M n.0H m

m.m v.N 5.v 0.0 0.0 5.0 m.n 0.0 0.0M 0.0 m mHOUOQSh

unwed 00mmmuumca pmmmmuum unwon cummouumca pommmuum usuoo 60mmouumcb pwmnmuum
Haves numcoq 0959 GHOS

menus gamma

mumuo snags

mocha umuflm

 

.Apmunouumcd can powmuuumv amouum can .AucoH
mpuos co>om .mcoH mpuos m>wu .chH mpH03 mmunuv sumcwq oocmucmm .Am0>wuaoucoo can muouoaduv make ouoz
.Aopmuv auuwm .mpmuv buds» .opmum umuwmv opmuw in mwocoucom canopHOIHHH you mouoom Houum «unaccouom amoxrl.v wands

APPENDIX G

COMPUTATION PROCEDURES

APPENDIX G

COMPUTATION PROCEDURES

Computation procedures for per cent error scores
were made as follows: Every word was classified as either
a stressed contentive, a stressed functor, an unstressed
contentive, or an unstressed functor. The total number of
word-type/stress combinations for each sentence type,
sentence length, and grade level, respectively, was divided
into the number of errors made in each condition of stress

and word-type. From this came the percentage error scores.

82

   

HICHIGQN STQTE UNIV. LIBRQRIES

I‘ll
8

HI?” llll W|l||||H
9

III |||||||||llH|
312 30061 5

49 0