IOURNAL
OF FLUENCY
DISORDERS
6 (1981)
341
341-350
Disfluency?*
Are There Constraints on Childhood Nan E. Bernstein University
Boston
of the spontaneous
Analysis
children
reveals
moments
that
is constrained
children
experience
and eight
stuttering
constituent
normally normal
and
fluent
disfluent
(b) stuttering
type of constituent
(verb
A view of early stuttering
by normals.
disorder
and
structure,
with a particular
produced
and integration
stuttering
of both
breakdown
disfluently
planning
of eight
location
by sentential
fluency
phrase) not usually a sentence
speech
(a) the
as
is present.
INTRODUCTION
Bloodstein their
examined
(1974)
ers and noted utterances
stuttering However,
syntactic
be the result
motor
planning
in early
Earlier
components.
to be more
have investigated
as place in utterance (Soderberg, stuttered
1967;
Helmreich
number
of
and Bloodstein,
limitations,
generalizations
including
about sentence-level
1967)
their
Such
processing
inability
to discriminate
sentence-generation
between
process and linguistic
Soderberg,
suffer
from
a
to make significant constraints
be at play in the precipitation of stuttering on individual their
locus such a
parameters of
1967;
studies
inability
(p 3901.”
or word type most
and Gantwerk, 1973).
that
to evaluate
such linguistic
frequently
such
the syntactic
it difficult
disfluency
that
structures
or laborious
investigated
stutter-
to fragment
theorized
of syntactic
difficult
making
on (Bloodstein
appeared
He
of “attributes
stuttering,
studies
speech of preschool
of these children
few studies have systematically
of disfluency claim.
into
might
cause their
the spontaneous
that the stuttering
by-products
which
words,
may
as well as
of the
normal
behavior peculiar to stutter-
ing.
‘Paper Convention,
presented at the 1979 Atlanta,
annual
American
Speech
and Hearing
Association
November.
Address all correspondence to Nan E. Bernstein, Progray in Applied Psycholinguistics, Boston University, 605 Commonwealth Avenue, Boston, MA 02215.
OElsevier North Holland, 52
Vanderbilt
Ave.,
New
Inc., 1981 York NY
10017
0094-730X181104341-1082.50
342
Nan E. Bernstein
In order to construct more descriptively models
of stuttering,
it would
be useful
and explanatorily to more fully
adequate
describe
the
parameters governing the fragmentation of the young child’s utterance by stuttered incidents, and to compare such behaviors with those observed in the speech of normal children. designed to systematically stuttered
and fluent
hypothesis
For these reasons, the present study was
describe the distribution
speech. Specifically,
that the distribution
regulated,
as well
of such disfluency
as to account for fluency
sentence generation constraints
of disfluency
in early
the study sought to test the was syntactically
breakdown
in terms
of
of a temporal nature.
PROCEDURE Subjects through
were divided
into two groups-one
second-grade children
comprised
referred by Philadelphia
preschool screening programs for their stuttering of normally
fluent children,
from each stuttering
child’s
of preschool public school
behavior, and the other
matched for age and sex, randomly selected class. Children
suspected of expressive
or
receptive language delay or disorder were excluded from participation in the study. From an original group of 20 children followed, eight matched pairs were selected for the study described here. Each group consisted of two girls and six boys. The mean age of the stuttering group was 6-3, of the control group was 6-4 Children
that
yrs.
participated in an elicited
language procedure in which
they were required to assign captions to cartoon pictures. The task was embedded in a language Lotto format, in which children were required to label the pieces they drew in order to place them on their game board. Cartoon
scenes,
culled
characters familiar
from
comic
to all children
Tom and Jerry, and Dennis was random within
book covers,
depicted common
in the study, such as Mickey
the Menace. Whereas
Mouse,
picture presentation
the child’s set, the game itself was preset so that the
child always won the game. Sessions,
conducted
in
both
home
and school
settings,
were
taperecorded using a Nagra portable tape recorder and Beyer microphone. Transcriptions with hesitations,
were then made of all the children’s
repetitions,
prolongations,
Only full sentences-defined an identifiable
NP-Aux-VP
utterances,
and filled pauses noted.
for these purposes as those containing
sequence-generated
during the course of
Constraints on Disfluency
343
the game were analyzed. assistance or clarification.
Included
in the analysis
Instances of disfluency
were requests
for
for both groups were
considered in terms of their syntactic locus and were classified in terms of the word
involved,
prolongations, hesitations
in the case of full
and part-word
or the one that followed
and filled pauses, following
the incident,
the guidelines
repetitions
and
in the case of
used by Hawkins
(1971). Constituent
structures
phrase (NP,), auxiliary conjunction,
isolated for consideration
were subject noun
(Aux), verb phrase (VP), object noun phrase (NP,),
and complement.
Disfluencies
on the first word of such structures
occurring directly
(constituent-initial)
tallied. All other locations of disfluency
before or
were noted and
were termed midconstituent
lumped together as a single category for analysis,
and
regardless of part of
speech or position in the utterance, for purposes of statistical analysis. After the original transcripts
were compiled, a reliability
made in which an uniformed graduate student-not pathology-was scriptions
provided with the original,
check was
in the field of speech
unnotated tapes and tran-
of each child’s conversation. The student was asked to mark the
presence of all disfluencies test yielded a correlation
described above. Results
of the Spearman r
factor of 0.93 for repetitions,
tions, and 0.62 for prolongations,
all statistically
0.78 for hesita-
significant values for the
sample size under analysis. RESULTS An examination of the relationship the presence of stuttering
between initiation
or disfluency
of a constituent and
required some systemization
the varied types and quantities of verbalization
of
obtained from each child.
To provide a baseline on which to ascertain prevalence of disfluency or before constituent words
on
boundaries, tallies were made of the total number of
per corpus for each child and of the total number of possible
constituent
boundaries
constituent-initial hypothetical
words
present
in using
in corpus
the baseline
divided
obtained-total
by total words-against
corpus length of 100 words, to obtain disfluency
ments for both constituent-initial
a
measure-
occurrence and midconstituent
occur-
rence, which were then comparable across subjects. Two-way
ANOVA
in degree of disfluency
with repeated measures indicated that differences between the two groups were significant
(F =
344
Nan E. Bernstein
4.87; ~~0.05). disfluency
However, group placement failed to influence the types of
observed. There was no significant difference between stutter-
ers and their controls
in the proportion
(words or parts of words), use of filled
of hesitations,
repeated items
pauses, prolonged segments, or
revised sentences. A test of the behaviors themselves level of 0.01 (F = 12.46).
was significant
Scheffe multiple
comparisons
at a confidence were computed
and indicated that the single most frequent form of disfluency groups was hesitation.
For the control
behaviors was displayed significantly for the stutterers, significantly
of an F_,,,
each child’s quently,
more frequently than the others. But and word-level
more often than did other disfluent
Results indicating
both hesitations
test for homogeneity
proportioned
for both
group, none of the remaining
disfluency
repetitions
occurred
behaviors. of variance performed on
values yielded a value of 13.19,
that the variances were not, in fact, homogeneous. preliminary
assessment
the locus of disfluency nonparametric
of the effect of constituent
eliminated
earlier plans to use
tool, the Wilcoxon
Sign Test
Related Samples, was administered
Conse-
structure
for Differences
Between
in order to determine the significance
of the difference between the stutterers’
disfluency
words
tendency for disfluency
and midconstituent
words.
The
on constituent-initial
located on or just prior to the first word of a constituent, interspersed
among constituent
level of significance.
on
Instead, a
ANOVA.
components,
to be
instead of being
was significant
at the 0.01
The same procedure performed on the data of the
control group produced similar findings, at the 0.05 level of significance. Results of Mann-Whitney for stuttering
children
aries was significantly p
U-test analysis indicate that the tendency
to fragment utterances along constituent different
The disfluency
from
of these stuttering
strongly confined to a constituent-initial children.
However,
that of their children
bound(U
=
4;
was much more
locus than was found for normal
there was no significant
groups on frequency of disfluency
controls
difference between the two
located in the midconstituent
position
(U = 18; p>O.O8). Hence, the major difference between the two groups appears to be in the strength and quantity of constituent-initial A two-way specifically
ANOVA
the role played by constituent
ency locus. Results
disfluency.
with repeated measures was used to assess more structure
indicated an effect of constituent
in regulating disflutype, for example,
Constraints on Disfluency
345
NP,, Aux, VP (f = 16.81; p
However, there did not appear to be
between patterns of locus of fluency
speech group (F =
1.25;
p>O.O5).
appeared to experience disfluency
That
breakdown
is, both groups
and
of children
while producing the same select group
of sentence constituents. Scheffe analysis was used to examine the nature of this distribution of disfluency
across the constituent types (Table 1). Results of the analyses
indicate that the first noun phrase of the utterance (NP,) is the favored locus of disfluency
for both groups, followed
by and and verb phrase
(VP). None of the other possible loci were significantly frequent involvement of and primarily
represented. The
in connection with disfluency can be seen as
overlapping with the initiation
the tendency of the children
of the utterance, as in NP,, given
in the study to link full utterances with and.
In fact, the frequency with which and occurred as a sentence lead-in was so common that it was calculated independently
of all other conjunc-
tions. Thus, for the most part, it is difficult to separate disfluency occurring either
on and or on NP,
disfluency
as being indicative
of anything
more than
in attempting to utter the sentence.
However,
VP attracted a significantly
the young stutterers.
Disfluency
significantly
represented
implications
of this surprising
high degree of disfluency
on this particular
constituent
in the speech of the normal
for
was not
children.
The
finding are discussed below.
DISCUSSION The two purposes of this study were to explore (1) the hypothesis disfluency
in young stuttering
than randomly,
distributed
children
might be systematically,
among major
utterance; and (2) similarities
components
and differences
that rather
of the child’s
between productions
of
fluent and nonfluent utterances by young normal and stuttering children, to help isolate characteristics of normal sentence-planning Bloodstein’s
length, stuttering
selections
processes.
appears to be essentially
from
and structure children
incident than
hypothesis (1974) that there are syntactic constraints on
the locus of disfluency examine
more typical of the stuttering
spontaneous
correct, even when we
speech in which
of utterance differ.
Both
appear to chop up their
normal
utterances
vocabulary, speaking
with
and
disfluent
1
6.0
5.1
2.5
1.1
0.4 0.5 0.7 0.7
NP, VP and
Comple Other Aux PrepP Conj
Test Value p 0.05 (*) p 0.01 (**) Means
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
0.25 -
0.38
0.13
-
-
-
-
-
-
-
1.75
1.75
2.00
2.12
1.38 -
0.37
0.37
0.62
0.75
NP, 2.50
1.12
5.72 6.65
5.72 6.65 Conj
-
-
0.00 -
0.25
0.38
5.72 6.65 PrepP 0.75
Across Constituents
5.72 6.65 Aux 0.75
of Disfluency
5.72 6.65 Other 0.50
Comple 0.37
Standard Error = 0.98
NORMALS
Scheff;? Analysis of Distribution
TABLE
-
2.62 -
4.00
4.37
4.37
4.62
4.75
5.12
5.72 6.65 VP
-
0.88
3.50
4.88
5.25
5.25
5.50
5.62
6.00
5.72 6.65 and
1.07
2.75
5.37
6.75+*
7.12**
7.12**
7.50** 7.37**
7.87
NP,
5.72 6.65
2.5 3.7 4.5 4.6 4.7 6.2 12.1 12.4
Means
Comple Other Aux PrepP Conj NP, VP and
Comple 2.50 -
Standard Error = 1.59 Test Value p 0.05 (*I p 0.01 (**I
STUTTERERS
9.24 10.76 Other 3.75 1.25 9.24 10.76 Aux 4.50 2.00 0.75 9.24 10.76 PrepP 4.62 2.12 0.87 0.12 9.24 10.76 Conj 4.75 2.25 1.00 0.25 0.13 -
1.50
NP2
6.25 3.75 2.50 1.75 1.63
9.24 10.76
9.24 10.76 VP 12.12 9.62* 8.37 7.62 7.50 7.37 5.87 9.24 10.76 and 12.37 9.87* 8.62 7.87 7.75 7.62 6.12 0.25 -
9.24 10.76 NP, 16.37 13.87** 12.62** 11.87** 11.75** 11.62** 10.12* 4.25 4.00
348
Nan E. Bernstein
episodes at constituent counter fluency constituents The
boundaries,
breakdown
and neither
group appears to en-
the internal
structure
rank ordering
of constituents
noun
most frequently
phrase and the linking
children-that
represented as
behavior was identical for both groups, with conjunction
affected. It is solely the additional significant stuttering
of sentence
with any significant frequency.
locations for such disfluent the initial
within
of initiation
the pattern and degree of disfluent
and most often
point of disfluency
of verb phrase-that
for the
distinguishes
behavior observed in this structural
analysis. Similarity
can also be found in the types of disfluency
observed
between the two groups. When they did encounter periods of disfluency, both groups manifested essentially
similar
categories.
difference was often heard on the
However,
a qualitative
behavior in terms of rough
tapes between certain normal easy repetitions children’s analysis
and some of the stuttering
more tense and extended repetitions. of the type and distribution
similarities
than differences
Nevertheless,
of disfluency
careful
demonstrated
between young stutterers
more
and normal chil-
dren. The findings of this study indicate that the major point of disfluency for both groups of children unit. Thus, on linguistic
is the constituent
boundary, a grammatical
the theoretical notion of nodes, labeled NP, VP, and so forth, tree-structure
diagrams, appears to constrain the location of
disfluent episodes. It is the initiation
of the node under which lower-level
specific lexical items are subsumed that attracts speech difficulty. highest possible level of analysis and formulation, all-S is,
or sentence -that for
both groups
breakdown synonymous
attracts the greatest degree of disfluency.
of young
children,
in connected discourse with the initiation
Results
the major
is the first
point
noun phrase, which This
is
is a
as we would assume that major
demands occur just prior to the initiation
of previous studies of both children
that processing
That
of fluency
of the utterance as a whole.
perfectly logical and expected finding, processing
At the
it is the highest node of
of the sentence.
and adults strongly
suggest
of spoken language does appear to be by the clause,
rather than word by word, as was suggested in early associationist
models
of language (Maclay and Osgood, 1959; MacKay, 1970; Lindsley,
1975).
However,
speaking is often initiated
completed, especially
before all planning has been
planning of the VP (Lindsley,
1975,
1976). Thus,
on Disfluency
Constraints
points
349
of indecision
or,
less often, error
observed between such points of indecision groups
in this
planning
study strongly
of syntactic
suggests
strings,
are reached. The
similarity
reached within and between
levels
of strain
inherent
rather than any “attributes
in the
of syntactic
structures that cause their motor planning to be more difficult or laborious (Bloodstein,
1974,
p. 390)”
constant in the disfluency
That the constituent
categories remained
of both groups, while the parts of speech that
composed them and surrounding
constituents
varied from child to child
and sentence to sentence, strongly suggests that the process of integrating such structures
into meaningful
selves, is responsible
wholes,
and not the structures
them-
for the difficulty.
Such a view is given support by the special behavior observed in the stuttering
children-that
demonstrated
of disfluent
that, for adults,
before all information
initiation
sentence
of VPs.
production
Lindsley
(1975)
may be initiated
about the verb has been processed, and that “the
final stage of verb processing-the
retrieval of the name for the action-
will be assumed to occur after the initiation of the utterance (p. 333).” subsequent study, Lindsley
In a
(1976) attempted to parcel out the processing
demands for subjects and verbs in simple sentences produced by adults. He concluded that “subject phrases can always be counted on to provide at least a two-syllable
delay by their overt utterance, a delay which may
be generally sufficient
to permit enough verb information
assure fluency (p. 351).” forthcoming.
For the young stutterers,
We might hypothesize
processing to
such fluency was not
that these children require more time
to integrate the components of the various sentence constituents. then, should VPs be especially the simplest
troublesome?
utterance, the constraints
They
Why,
may reflect, even in
of number agreement imposed by
the preceding noun, the necessity of encoding semantic notions such as tense and aspect to be carried by the sentence as a whole. choice of verb will forcing within
Similarly,
constrain the succeeding elements of the utterance,
accommodations
to notions
of transitivity/intransitivity
the verb, as well as complement
given verb selection. Alternatively,
structures
coded
accepted or barred
the VP may simply
represent another
stepping stone to planning the completion of the utterance extending far beyond the scope of the verb. The advantage of midsentential
children
may simply
be taking
stopping points to contemplate the rest of the
utterance as a whole. The fact that stuttered incidents appear to be syntactically
governed
350
Nan E. Bernstein
in the speech of young children planning
suggests that it is higher-level
processes, and not word-,
considerations,
sound-,
that may act to precipitate
sentence-
or motor-gesture-specific the stuttered
moment.
In
addition, the finding that such patterns of disfluency are largely congruent with patterns of disfluency the preceding common
view,
noted in normal children
reinforces not only
but further suggests that fluency
childhood
process and
children and normal children
that differences
breakdown
between
may end up best explained
is a
stuttering
in terms of the
and not in terms of its precipitation.
manifest form of the disfluency,
REFERENCES Bloodstein, 0.
The
Disorders, Bloodstein,
rules
of early
O., and Gantwerk,
young
stuttering.
/ournal
of Speech
and
Hearing
1974, 39, 379-394.
children.
/ournal
6. Grammatical function in relation to stuttering in of
and
Speech
Hearing
Research,
1967,
10,
786-789. Hawkins,
P.R. The syntactic location of hesitation pauses. Language and Speech,
1971,14,277-288. Helmreich,
H.G.,
disfluency
and Bloodstein,
Hearing Research, Lindsley,
The
grammatical hypothesis.
factor
in childhood
/ourna/ of Speech
and
1973, 16, 73 1 - 738.
J.R. Producing simple utterances: How far ahead do we plan? Cognitive
Psychology, Lindsley,
J.R.
journal
1975,
7, l-19.
Producing
simple
of Psycholinguistic
H.,
and Osgood,
speech. Word, Soderberg,
G.A.
Research,
1959,
Linguistic
1967,
utterances:
Research,
McKay, D.C. Context-dependent Maclay,
0.
in relation to the continuity
stuttering.
C. Hesitation
Details
1976,
of the planning
process.
5, 331-353.
Kybernetik, phenomena
1970, 7, l-9. in spontaneous
English
15, 19-44. factors in stuttering.
lO, BOl-810.
/ournal
of Speech
and Hearing