J . COMMUN . DISORD .
19 (1986),133-160

MARKEDNESS ANALYSIS OF CONSONANT
ERROR PRODUCTIONS IN APRAXIA OF SPEECH
LESLEY WOLK
University of the Witwatersrand, Johannesburg, South Africa

This study applied a markedness analysis of consonant error productions to three adult
aphasic patients with apraxia of speech . Single consonants and initial consonant clusters
were investigated, employing both naming tasks and spontaneous narrative tasks . Results
showed that phonemes of lower complexity values tended to substitute for phonemes of
higher complexity values more than the reverse, and successive attempts at self-correction
tended to reveal a progression toward the target segment in terms of markedness complexity .
Theoretical and clinical implications are discussed .

INTRODUCTION
Both researchers and clinicians have struggled for many years with the
conceptual issues in apraxia of speech as well as with the development
of suitable clinical regimes . Several investigators have described the disorder referred to as "apraxia of speech" as a motor-programming disturbance not primarily due to neuromuscular or higher language functioning deficits and have suggested that the programming of phonemes
can be selectively impaired without impairment of language (Johns and
Darley, 1970 ; Deal and Darley, 1972) . Martin (1974), however, objected
to the term "apraxia of speech" and argued that motor-programming
disturbances are inextricably tied to language deficits and cannot be defined except within the framework of aphasia . Several studies in apraxia
of speech have employed taxonomic methods of phonemic analysis .
Based on these analyses, some investigators have concluded that consonant substitutions are random and unrelated to the target sound (Critchley, 1952 ; Deal and Darley, 1972) . By contrast, investigators employing
distinctive feature analyses have indicated that consonant substitutions
are systematic and rule-governed (Blumstein, 1973 ; Martin and RigrodThis study formed part of the author's M .A . dissertation entitled "Phonological Impairment in Aphasia" submitted to the Department of Speech Pathology and Audiology, University of the Witwatersrand, Johannesburg .
Address correspondence to Lesley walk, Department of Communication Sciences and
Disorders, Syracuse University, 805 S . Crouse Ave ., Syracuse, NY 13244 .

C 1986 by Elsevier science Publishing Co ., Inc .
52 Vanderbilt Ave ., New York, NY 10017

133
0021-9924/86/$03 .50

134

L . WOLK

sky, 1974a . 1974b ; Hatfield and Walton, 1975) . Recently there have been
a few preliminary attempts to apply a markedness analysis to consonant
error productions in apraxia of speech (Dunlop and Marquardt, 1977 ;
Welk, 1978 ; Marquardt, Reinhart, and Peterson, 1979) . 'The concept of
markedness in apraxia is discussed in some detail in Wertz, LaPointe,
and Rosenbek (1984) . Before considering these studies, however, a brief
discussion of markedness is relevant .
Markedness theory is based on the underlying theoretical assumptions
of natural phonology (Hyman, 1975) . The fundamental principle of natural
phonology is that certain segments are more natural than others and that
the "naturalness" of segments can be captured through the notion of
"markedness ." Within this conceptual framework, two phonemes or phoneme sequences are differentiated by considering one of them unmarked
(U) for a particular feature and the other marked (M) for that feature .
The basic assumption is that the (U) member represents the more natural
state and will never contribute to the complexity of a segment . In addition,
there is a set of marking conventions that are sensitive to other feature
values, both in the same segment and in adjacent segments, so that the
complexity of a phoneme is equal to the sum of its marked features . It is
expected that (U) segments at any level are common to the languages of
the world, are reflected in early language acquisition, are preferred in
historical language change, have greater articulatory ease, and are perceptually salient (Hyman, 1975) .
In addition to natural segment types, there are also natural phonological
rules that lead to preferred (simpler) syllable structure and less marked
segments or sequences of segments . Following the same principle as that
discussed for (M) and (U) phonemes, the most unmarked syllable type is
CV (an initial consonant followed by a vowel) . Other syllable types such
as CVC are relatively complex, and CCVC and CVCC syllables are still
more complex . It has been suggested that different consonant types should
be assigned strength values to capture phonological relations between
segments, particularly as they function in syllables (Hyman, 1975) . The
major criticisms of markedness are based on a lack of agreement in the
assignment of consistent marked and unmarked values to phonemes
(Hooper, 1976 ; Sommerstein, 1977) . In spite of this, Sommerstein (1977)
has cited Venneman and Hooper, who feel that "natural phonology must
be regarded as one of the most important developments in phonological
theory, if not the most important, since GP [generative phonology] itself"
(Sommerstein, 1977, p . 237) . Furthermore, markedness may indeed have
some relevance for the speech clinician .
Klich, Ireland, and Weidner (1979) compared distinctive feature analysis and markedness analysis on nine adult patients with apraxia of
speech and found a markedness analysis to be superior in reflecting the
nature of the consonant errors . According to Blumstein (1973), a knowl-

MARKEDNESS IN APRAXIA

135

edge of both feature specifications and markedness relationships is essential to describe the phonological system of a language .
Marquardt, Reinhart, and Peterson (1979) have carried out a markedness analysis of phonemic substitution errors in apraxia of speech . They
found that their patients produced more errors on phonemes high in
markedness and that more changes occurred from marked to unmarked,
suggesting that the patients' articulatory behavior reveals a systematic
effort to reduce the complexity of the articulation required to approximate
target phonemes . Marquardt, Reinhart, and Peterson (1979) make a significant comment that at a given time children may not have mastered
the skills to produce more difficult phonemic features, while the apraxic
adults have lost some degree of ability to coordinate or plan motor speech
output . Their study may be considered incomplete, however, in that they
used an imitation task only, they investigated only part of the total consonant system of English, thus precluding generalization to the entire
phonemic system, and they provided little explanation of how they arrived
at their markedness complexity values . Further investigations of articulatory difficulties in adult apraxic patients, employing the framework of
natural phonology in terms of markedness, was believed to be needed .
This study was undertaken to explore the applicability of a markedness
analysis of consonant error productions to three adult aphasic patients
with apraxia of speech, in terms of markedness complexity of errors and
successive attempts at self-correction . The author's intention was not to
evaluate a particular theory or to determine its validity from a linguist's
point of view, but rather to investigate its potential applicability for the
speech pathologist concerned with adult aphasic phonological impairment . This article is thus concerned specifically with the area of perceptual investigations in apraxia of speech and will not deal with issues pertaining to acoustic or physiological areas of investigation .

METHOD
Subjects
Three adult male aphasic patients were the subjects for this study . All
cases were chosen from Johannesburg hospitals and were diagnosed as
aphasic by neurologists and speech pathologists . They ranged from age
58 to 73 years with a mean age of 63 .3 years . All cases presented a history
of left (L) hemisphere cerebrovascular accident (CVA) . They demonstrated various degrees of concomitant Broca's aphasia, but in each case
phonological deficit predominated . They ranged in months post onset
(MPO) from 9 to 24, with a mean of 15 .7 months . All cases were native
speakers of English and had no previous history of speech or hearing
problems, premorbid psychological problems, senility, or mental retar-

Case 1
(A . M .)
Case 2
(F .T .)
Case 3
(E .J .)

Male

Male

Male

73

59

Sex

58

Age

CVA

CVA

CVA

Etiology

14

9

24

Months post
onset (MPO)

Table 1 . Description of Cases Used in This Study

Broca

Broca

Broca

Type of
aphasia

English

English

English

Home
language

Std . 9

Std . 8

Std . 6

Educational
level

Dispatch
manager
Salesman

Plumber

Premorbid
occcupation

(R)

(R)

(R)

Handedness

12 mos .

4 mos .

18 mos .

Speech
therapy
received

MARKEDNESS IN APRAXIA

137

dation . Years of education ranged from 8 to I 1 years, with a mean of 9 .7
years . Case 1 had the most severe form of apraxia of speech, and case 3
had the least severe disorder, but they were all believed to be representative of mild motor aphasics . Subjects had to have suffered a left (L)
cerebral infarct and were required to have reached a point of neurological
stability considered as 6 months after the neurological insult (Sarno and
Levita, 1971) . The cases are summarized in Table 1 . Computerized tomographic (CT) brain scans for each case are in Figures 1-3 .

Procedure
The preliminary tests to determine whether the subjects satisfied the selection criteria are summarized in Table 2 . Two approaches were used
for elicitation of speech for linguistic analysis : naming tasks and spontaneous narrative tasks .

Table 2 . Results of Preliminary Tests
Preliminary test

(A .M .)

Case 2
(F .T .)

Case 3
(E .J .)

Boston Diagnostic
Aphasia
Examination
(BDAE)
Goodglass and
Kaplan (1972)
Token Test
De Renzi and
Faglioni (1978)
Oral Non-verbal
Gesture Battery
Moore, Rosenbek,
and LaPointe
(1976)
Screening PureTone
Audiometric Test
Goldman-Fristoe
Test of Auditory
Discrimination
(Goldman,
Fristoe, and
Woodcock, 1970)
Discrimination of
CVC and CCVC
words (designed
by the writer)

Score on severity
rating scale = 2

Score on severity
rating scale = 2

Score on severity
rating = 2

34/36 = 94%

34/36 = 94%

36/36 = 100%

56/56 = 100%

56/56 = 100%

56/56 = 100%

Normal

Normal

100%

Normal except
45dB at 4000 Hz
bilaterally
100%

100%

100%

100%

Case I

100%

1 38

(a)

(b)
Figure 1 . Case I (A .M .) : CT Scan Showing Cerebral Infarction Involving Mainly
the (L) Paricto-Occipital Region and Extending Toward the Island of Reil (Insula)
on the Left Side

139

(a)

(b)
Figure 2. Case 2 (F .T .) : CT Scan Showing Cerebral Infarcts in the (L) ParietoFrontal Region Involving Broca's Area and in the (R) Parieto-Occipital Region

140

L . WOLK

(a)
Figure 3 . Case 3 (E .J .) : CT Scan Showing Cerebral Infarction Involving the (L)
Parietal Region Extending Toward the Vertex and Inferiorly to the (L) Island of
Reil (Insula) and Upper Temporal Lobe

Naming Tasks
A test in two parts was developed by the writer for use in this study . Part
I was designed to test single consonants in initial (I) and final (F) word
position, and Part II was designed to test initial consonant cluster sequences . These are included in the Appendix .
Criteria for selection of words . Part I of the articulatory test was composed of 89 items carefully selected to elicit all the consonants (C) in
English in (I) and (F) word positions . An attempt was made to select items
in which the desired consonant in (I) and (F) word position could be
elicited simultaneously, thus maximally conserving the total number of
picture stimuli required (e .g ., initial /d/ is elicited by the picture words
"dog" and "duck" at the same time that final /g/ and /k/ are being elicited) .

MARKEDNESS IN APRAXIA

141

(b)
Figure 3 . (continued)

Each consonant was elicited at least twice in (I) and (F) word position,
ranging between two and four different phonemic contexts . The phonemic
environment was varied with respect to both vowels and consonants in
the immediate phonemic vicinity . Thus 81 words represented consonants
in (I) position, and 70 words represented consonants in (F) position, while
the total number of test items was 89 .
Part II of the articulatory test was composed of 99 items . Each (I) cluster
of English was elicited in at least two stimulus words, ranging between
two and four different phonemic contexts .
Experimental manipulation of words . Stimulus items were constrained
in the following ways : (1) Phonological rules governing the combination
and sequencing of phonemes in English were applied . (2) Words that could
be pictorially represented were used, because imitation was to be deliberately avoided . (3) Word length and syllable structure were considered ;
monosyllabic and disyllabic words were included, multisyllabic words
were excluded . (4) Morphological endings were excluded when devising
the list of single consonants in (F) position to ensure that only phonological

142

L . WOLK

capacities were being tapped, and not other levels (e .g ., the morphophonemic level) .
The entire articulatory test was thus comprised of 188 stimulus items
(Part I = 89 ; Part II = 99) . Each response was elicited twice, to provide
an estimate of the subjects' consistency of production of the same sounds
in identical environments . Thus both parts of the naming task, naming
singletons (NS) and naming clusters (NC), were administered twice to all
subjects .

Spontaneous Narrative Tasks
(1) Thirty minutes of an open-ended conversation concerning the patient's
illness, work, hobbies and general issues was elicited . The middle 20
minutes were selected for analysis . (2) Fifty picture cards from the Phonological Process Analysis (PPA) (Weiner, 1979) were selected for the
purpose of picture description . (3) Six large, colored picture cards depicting real-life scenes were chosen in order to elicit conversational
speech . In addition to (1), (2), and (3), many of the spontaneous speech
samples resulted from the subjects' misnaming of test items and interjected spontaneous speech .

Administration of phonological assessment procedure . All testing took
place in a sound-treated room and was carried out by the writer during
several 45-minute sessions . Subjects were tested individually . For the
naming tasks, pictures were presented in random order, but attempts were
made not to present two words with the same consonant or consonant
cluster successively . Part I (NS) and Part II (NC) were administered twice
over four sessions, following the same procedure . For the spontaneous
tasks, each subject had to describe the 50 picture cards selected from the
PPA and to tell a story for each of the six picture cards chosen . Spontaneous narrative tasks were administered once over three or four
sessions .

Response transcription . All responses were recorded on a high-quality
Revox tape recorder (Model 1132, Dolby version) . Wherever possible,
on-site transcriptions were carried out by the writer, utilizing both visual
cues and contextual information . Tape recordings were transcribed by
the tester in broad phonetic transcriptions and were used as a comparison
with on-site transcriptions . A second transcription of all data was carried
out using three transcribers who served as judges . Judges worked independently and then met jointly to resolve any discrepancies, discarding
any instances in which no consensus could be reached, providing maximum objectivity and accuracy of recorded data .

MARKEDNESS IN APRAXIA

143

Markedness Analysis
Analysis of Phonemic Errors in Terms of Complexity
Consonant error productions were subjected to a markedness analysis by
applying marked (M) and unmarked (U) values and total complexity values to every consonant and consonant cluster sequence (error and target
segments) . A "9 Distinctive Feature (DF)" matrix was used, with values
available for the following features : consonantal, sonorant, anterior, coronal, continuant, strident, voice, lateral, and nasal . A complexity score
was computed by adding the (M) features in each consonant .
The markedness values used were based on those suggested by Cairns
(1969) and Chomsky and Halle (1968) . They are discussed in an earlier
report (Wolk, 1978) and presented imAppendixes A and B .
The following symbols were used to refer to the direction of change in
terms of complexity values (p = phoneme ; H = high complexity ; L =
low complexity) :
(p = p) The target segment, or a substitution of equal complexity value
to that of the target .
(H-~L) A substituted phoneme or phoneme sequence that has a lower
. error
(L) complexity value than that of the target i .e ., target (H-*L) .
(L.H) A substituted phoneme or phoneme sequence that has a higher
(H) complexity value than that of the target . It reflects a change from
less complex to more complex, i.e ., target -* error (L->H) .
Analysis of Self-Correction

For each target segment the subjects' successive approximations toward
the target were carefully observed, i .e ., whether the self-correction process revealed a progression toward the target in terms of markedness
complexity . The following three categories were developed for this
purpose :
(1) Progression (Pr) : (H-L) -, (p=p), (H-*L) - (p=p) - (L.H),
. (L-'H)
(H-*L) (2) Nonprogression (NPr) : (p=p)~ (H.L), (p=p)-* (L-*H), (L-'H)
-. (H.L)
(3) No Change : (p=p)-* (p=p), (H-*L)-* (H-*L), (L-H)- . (L-.H)
In a progression, the subject begins with a less complex segment, progresses to the target segment or one that is equal to the target complexity
values, and sometimes "overshoots" the target complexity value, resulting in a more complex segment . These progressions will be referred
to as approximations toward the target segment . In a nonprogression,
successive attempts proceed in the opposite direction .

ET

F .T .

A. M.
F .T .
E .J .
A .M
FA .
E .J .
A .M .
F .T .
E .J .
A.M .

Cases

y
71,43
85 .89
89,07
16,03
10,48
7,10
12,54
3,63
3,83
100
100
100

No .

205
213
163
46
26
13
36
9
7
287
248
183

NS-(1)

199
181
127
37
33
39
27
14
14
263
228
180

No .

SS-(I)

100

100

75,67
79,39
70,56
14 .06
14 .47
21,66
10,27
6 .14
7,78
100

%

0
275
222
164

186
156
142
72
61
22
17

No .

c,
67,64
70,27
86 .59
26 .18
27 .47
13 .41
6 .18
2 .25
0
100,
100
100

NS-(F)

237
197
154

129
98
108
99
94
44
9

No .

SS-(F)

54,43
49,75
70
.13
41,77
47,71
28 .57
3 .80
2,54
1,30
100
100
100

%n

Key : NS =Naming Singletons ; SS =Spontaneous Singletons ; p = phoneme : L = low complexity : H = high complexity : (p- p) = target segment or substitution
of equal complexity value to that of target : (H-sL) = substitution of lower complexity value than that of target ; (L-'H) = substitution of higher complexity value than
that of target .

Total (N)

L-*H

H-L

P= p

Complexity
Value

Table 3 . A Distribution of Phonemic Errors on Target Singletons According to Markedness Complexity . (p= p), H-L),
and (L.H)

145

MARKEDNESS IN APRAXIA
Scoring

The scoring procedure for the markedness analysis, including the model
for assigning complexity values to single consonants and initial consonant
cluster sequences, is described in an earlier report (Wolk, 1978, p . 88) .

RESULTS AND DISCUSSION
Analysis of Phonemic Errors in Terms of Markedness Complexity
Table 3 and Table 4 show that in each case a large number of utterances
on all tasks consisted of phonemes and phoneme sequences in which the
value of the features in the target segments were retained (p=p) . It is
clear, however, that when a change occurred, target segments were more
commonly replaced by segments of lower complexity rather than segments of higher complexity (H.L) . For example, in Table 3 on SS in the
(F) word position for Case 1 (A .M .), out of a total of 237 utterances, 99
(41 .77%) were replaced by less complex segments (H.L) and only 9
(3 .80%) were replaced by segments of higher complexity (L- H) . The
remainder (54 .43%) revealed no change (p =p) . Similar proportions can
be seen on singletons in the (I) and (F) word positions for all cases . In
Table 4, on NC for Case 2 (F .T .), out of a total of 347 utterances, 116
(33 .42%) were replaced by less complex segments (H-.Q and only 8
(2 .31%) were replaced by segments of higher complexity (L-*H) . The
remainder (64 .27%) revealed no change (p=p) . Similar proportions can
be seen on NC and SC for all cases . A larger number of total utterances

Table 4 . A Distribution of Phonemic Errors on Target Consonant Clusters
According to Direction of Change, (p=p), (H-.L), and (L-.H)
Complexity
value
P=p

H-.L

L-.H

Total (N)

NC

SC

Cases

No .

%

No .

A .M .
F .T .
E .J .
A .M .
F .T .
E .J .
A . M.
F .T .
E .J .
A .M .
F .T .
E .J .

282
223
193
198
116
56
54
8
9
534
347
258

52,81
64,27
74,81
37,08
33,42
21,71
10,11
2,31
3,49
100
100
100

185
204
155
86
51
64
24
6
5
295
261
224

62,71
78,16
69,20
29,15
19,54
28,57
8,14
2,30
2,23
100
100
100

Key: NC = Naming Clusters ; SC = Spontaneous Clusters ; p = phoneme ; L = low complexity ;
H = high complexity .

L . WOLK

146

Table 5 . Example from Case I (A .M .) of a 9 Distinctive Feature (DF)
Matrix Computed for the Singleton /d3/ in Initial Word Position,
and Its Various Substitutions
/d3/ = (4) total cnmjIrxity
Feature
cons
son
ant

cor
coot
stri
voice

lat
nas
Complexity

d3

k

if

r

m

9

U
M
M
U
U
M
M

U
M
M
U
U
U
U

U
M
M
U
U
N1

U
M
U
11
M
U
U

U
U
U
M
U
U

U
M
M
U
U
U
M

U
M
U
U
M
M
U

U
M
U
U
U
U
M

U

M

U

I1
3

2

U

U

4

2

d

3

(N) occurred on naming tasks (NS and NC) than on their respective spontaneous tasks (SS and SC) . This shows that more attempts at self-correction occurred on a voluntary-purposive naming task and fewer in connected speech .
Table 5 presents an example from Case I (A .M .) of a 9 Distinctive
Feature (DF) matrix computed for singletons and clusters to show analyses of phonemic errors in terms of markedness complexity . Table 6
shows that Case I (A .M .) showed a tendency for segments of lower complexity to replace target segments . For example, the phonemes /r, m, d,
k/ with complexity values equal to 2 . and /s, g, I fl with complexity values
equal to 3, were frequently substituted for the target singleton /d3/ with
a complexity value equal to 4 .
Table 6. /d3/ in Initial Word Position, and its Various
Substitutions
"target word
Utterance
Total C
M/U
I . giraffe

r
m

2
l

2 . juice

d
d3
3 . jug

k

4 . jersey

if
d3

4

H~L
H.L
H-*L

Pr

ll-HL'
H +L
H-.L
P°p ,

Pr

II-s[_
H-.L

PI

p=p

Key : Total C = lulal complexity ; p = phoneme : H - high complexity ; L
= low cumplexlly : Pr = progression : NPr = No iprugresswn .
From . Naming Tasks NS .

MARKEDNESS IN APRAXIA

147

It is therefore evident that when Case 1 (A .M .), Case 2 (F .T .), and Case
3 (E .J .) phonological errors reflected a change in markedness complexity,
they were likely to reflect a change in the direction from more complex
to less complex . The data does not support a statement that these cases'
articulation errors always reflect a change from more to less complex,
but they do suggest that when a change occurs it is more likely that the
direction will be from more complex to less complex . The total number
(N) and percentage of times that features change from (H-*L) far exceeds
the number of times the errors reflect changes in the opposite direction
(see Table 6) . This tendency is seen to occur consistently on all tasks and
corroborates previous reports, although the complexity values used in
these studies all varied (Blumstein, 1973 ; Wolk, 1978 ; Klich, Ireland, and
Weidner, 1979 ; Marquardt, Reinhart, and Peterson, 1979) . Marquardt et
al . (1979) propose a correlation between directional changes in markedness with ease of articulation . They interpreted their findings as a systematic effort to reduce the complexity of articulatory gestures required
to produce phonemes .
Though not the major tendency, there were changes in each case from
lower to higher complexity, which does not support the general hypothesis . This was predominant for the feature voice (v), a finding consistent
with Klich et al . (1979) and Marquardt et al . (1979) . Marquardtet al . (1979)
propose that possible contextual effects may be one source for this finding . This is explained by Smith and Ruder (1975), who have suggested
that context may significantly affect segmental markedness for voicing .
For example, sequences with voiced -> voiceless clusters are more complex and require more physical effort because two adjustments of the
larynx are required for production of the cluster . Other explanations for
the change (L-*H) in this study may be offered, such as consonant harmony and the interchange of /1/ and /r/ where /I/ has a complexity value
equal to I and /r/ has a complexity value equal to 2 .

Analysis of Self-Correction
Table 7 and Table 8 summarize successive attempts at the target segments
on singletons and clusters respectively . Totals (N) in these tables reflect
the total number of test items on which the cases produced successive
attempts at self-correction . In each case a large number of approximations
reflected no change in complexity . Either another segment of equal complexity value to that of the target was substituted, or the same segment
was merely repeated several times . When a change occurred, however,
there was a consistent tendency for a progression (Pr) to occur rather
than a nonprogression (NPr) . That is, initially a segment of lower complexity was replaced, followed by successive attempts at self-correction,
which revealed a steady increase in complexity until the target was

A .M .
F .T .
E .J .
A .M
F .T .
E .J .
A .M .
F .T .
E .J .
A .M .
F .T .
E .J .

Progression (Pr)

14
7
7
12
7
3
16
14
1
42
28
11

No .
33,33
25,00
63,63
28,57
25,00
27,27
38,10
50,00
9,09
100
100
100

%

Key : NS = Naming Singletons ; SS = Spontaneous Singletons .

Total (N)

No change

Nonprogression
(NPr)

Cases

Direction

NS-(q

11
9
6
8
3
1
20
22
6
39
34
13

No .

SS-(1)

28,21
26,47
46,15
20,51
8,82
7,69
51,28
64,71
46,15
100
100
100

%
23
21
II
8
4
0
14
14
0
45
39
11

No .

%
51,11
53,85
100,00
17,78
10,26
0
31,11
35,89
0
100
100
100

NS-(F)

Table 7. Successive Attempts at the Target Singleton in Terms of Markedness Complexity

31
17
8
5
5
1
9
9
2
45
31
11

No .

SS-(F)

68,89
54,84
72,74
11,11
16,13
9,09
20,00
29,03
18,18
100
100
100

MARKEDNESS IN APRAXIA

149

Table 8. Successive Attempts at the Target Cluster in Terms of Markedness
Complexity
NC
Direction
Progression (Pr)

Nonprogression (NPr)

No change

Total (N)

SC

Cases

No .

%

No .

%

A .M .
F .T .
E .J .
A .M .
F .T .
E .J .
A.M .
F .T .
E .J .
A .M .
F .T .
E .J .

48
44
23
17
5
5
35
29
9
100
78
37

48,00
56,41
62,16
17,00
6,41
13,51
35,00
37,18
24,33
100
100
100

23
17
6
10
4
2
18
13
4
51
34
12

45,09
50,00
50,00
19,61
11,76
16,67
35,30
38,24
33,33
100
100
100

Key : NC = Naming Clusters ; SC = Spontaneous Clusters .

reached (H-*L) -> (p=p) . Sometimes the substitutions reflected a segment of higher complexity overshooting the target complexity value
.H) . This tendency was consistent over all tasks except for
(H.L) - . (LCase 2 (F .T .) on NS in the (I) word position, where no tendency in one
direction over another could be observed . The trend was more pronounced for clusters than for singletons .
A striking example of the tendency for a progression to occur rather
than a nonprogression is illustrated in Table 7 for Case 3 (E .J .) on NS in
the (F) word position . Here out of a total of I I test items representing
singletons in the (F) word position in which successive attempts at selfcorrection were made, 100% revealed a progression toward the target
segment . Similarly, in Table 8 for Case 2 (F .T .) on NC, there was a total
of 78 target words representing initial consonant clusters on which successive attempts were made at self-correction . Out of these, 44 (56 .41%)
represented a progression toward the target, while 5 (6 .41%) represented
a nonprogression . The remainder, 29 (37 .18%), reflected no change in
complexity . Thus in all cases successive attempts at self-correction do
not appear to be random . They commonly reveal a progression toward
the target segment in terms of markedness complexity (H- .L) - (p=p) .
Some examples from Case I (A .M .) of initial consonant clusters and
their various substitutions are provided in Table 9 . The first example of
the word "spread," which clearly demonstrates the process of self-correction for the cluster /spr/, can be examined to highlight the present
findings . The initial attempt is /pr/ with a complexity value equal to 4,
the second attempt is /fr/ with a complexity value equal to 6, and on the

150

L . WOLK

Table 9 . Some Examples from Case 1 (A .M .) of Initial Consonant Clusters and
Their Various Substitutions, to Show the Process of Self-Correction
Target word

Error segment

spread

scratch

screw

Total complexity

M/U

Isprl = (8) Total Complexity°
pr
4
fr
6
spr
8

H'-Ll
H,L
Pr
p=p

Iskrl = (9) Total Complexity`
kr
5
7
str
skr
9

H-+L
H-*L~ Pr
p=p

sl

str
skr
skr

7
7
9
9

H- .L
H.L
p=p
p=p

Pr

Ifr/ _ (9) Total Complexity"

shrub

sp

br
hr
br
skr

fly

suing

strap

3
5
5
5
9

H-*L'
H, 1,
H,L Pr
H, L
P=P

ff11 = (5) Total Complexity"
w
1
3
pl

H-. L
H,L

lstrl = (7) Total Complexity"
st
3
tr
4
str
7

H--*L
H-L~ Pr
p=p

so
so
si

sir

4
4
7
7

Pr

H-L
H- 4 . Pr
P=p
P=P ,

°= From : Spontaneous Tasks .
From : Naming Tasks .

third attempt the target sement /spr/, which has a total complexity value
equal to 8, is reached .
The findings for all cases thus provide evidence for a pattern of successive attempts at self-correction, showing a definite link or "relatedness" between error production and desired production . These cases appear to feed back their own phonological productions, compare them to

MARKEDNESS IN APRAXIA

151

the original targets, and then induce an improved approximation to the
target . This corroborates the findings of Joanette, Keller, and Roch-Lecours (1980), who further suggest that the targets must be adequately
represented to permit these operations . The present findings do not lend
support to Johns and Darley (1970), who characterize the variability of
phonemic productions in apraxia of speech as unrelated, additive substitutions, repetitions, and blocks, groping through repeated efforts toward the correct production . These results of self-correction are believed
to be extremely significant for the following reasons :
(1) Successive attempts at self-correction provides an opportunity to
observe dynamic linguistic processing or the "internal functioning"
of the phonological system operating in each case .
(2) They provide further information about perception . The fact that
these patients are able to monitor their own output shows that to
some extent the perceptual mechanism is functioning . But why does
this process of self-correction result in further substitution errors?
That is, why do these patients follow an initial attempt at a target
production with further error utterances rather than stop? It might
be plausible to suggest that proper auditory feedback for monitoring
ongoing speech is disturbed, a hypothesis supported by Ireland,
Klich, and Panagos (1977) . That is, there may be a subtle deficit in
the sensorimotor feedback loop .
It is important to bear in mind the level of reference previously identified
by the investigator, which in the analysis of self-correction, for example,
was the progression or nonprogression toward the target segment in terms
of markedness complexity . That is, the third category, "no change," was
developed to accommodate responses that did not fit into the body of
reference previously identified . In each case the subject did not stop .
Either another segment of equal complexity value to that of the target
was substituted, or the same segment was merely repeated several times .
Therefore, one may infer from this behavior that the patient must be
attending to something but that possibly for this task the attention of the
investigator did not parallel the attention of the speaker because of the
previously identified level of reference . Thus one can conclude only that
there was a greater tendency in each case to show a progression rather
than a nonprogression toward the target segment, and that further study
would be needed in order to fully understand the area of "no change"
which occurred in each case . A review of initial consonant clusters from
the data of all cases shows that similar substitution errors tended to occur
on different clusters comprising the same elements . Table 10 illustrates
an example of this phenomenon from Case 3's (E.J .) data on the following
three cluster segments : /st/, /tr/, /str/ .
The data in this study refer to the speech samples of three adult aphasic

152

L . WOLK

Table 10. Example from Case 3 (E .J_) of Three Different Initial Consonant
Clusters Comprising the Same Phonemes [/st/ _ (3), /it/ _ ( 4), /str/ _ (7)1
Target word
/st/

stove

Error segment

Total complexity

k
st

H-L,
Pr
p=p
H,L
H-Ll Pr
P=p

stamps
stove
/tr/

tried
trunk
tractor
try
trying
tram
tractor
train
trees

/str/

strap

tJ
If
If
t
t
pr
tr
d
tr
s

tr
dr
str
J
sk

Direction of change

3
3

H->L
HL
H- •1 .
TI-+L
H, L

Pr

4
4
2
3
3
4

p=pl

Pr

5

H-L

p- p J

H,L
H.L
H-xL1 Pr
p=p

p=p

I

Pr

straw
H-L
straw
H,L
straight
s
3
H,I .
Key : p = phoneme ; H = high complexity ; L - low complexity ; Pr - progression ; NM
nonprogression .
In summary, the errors are lstl : s/st, t/st, k/st : ltrl: Or, t/tr, tJ/tr, pr/tr, d/tr ; and lstrl : s/ s tr. st/str. sk/
str, f/str, dr/str .
From : Naming Tasks .

patients . They cannot form the basis for any extensive comparison with
the development of speech in children, but it is interesting to note that
the findings do correlate with Cairns (1975), who found that children's
substitution errors may be characterized by a shift toward a lower complexity value for a particular feature . The findings concerning the aphasics
in this study further correspond with those of McReynolds, Engmann,
and Dimmitt (1974), who characterized substitutions of children with articulation errors in terms of marked and unmarked values . Their results
imply that when a change occurs it is more likely that the direction will
be from more complex to less complex . (It should be noted, however,
that some of the overall complexity values in the present study differed
from those used in the study by McReynolds et al .) Panagos, Quine, and
Klich (1979) contend that structural complexity may contribute to overall
hierarchical complexity, in turn causing children to simplify their speech .
Similarly, the finding that these aphasic adults substitute segments and
syllable structure of lower complexity value for those of higher complexity

MARKEDNESS IN APRAXIA

153

value shows some link with various theories postulated about articulatory
effort (Locke, 1972 ; Parnell and Amerman, 1977) . Before one could make
any statement about this with confidence, however, future research to
generate an ease-of-articulation hierarchy is needed .
The overall results seem to contribute to previous support for Stampe's
hypothesis of "centrality of processes" (Macari, 1978) . An important limitation of these findings is that Stampe's argument is based primarily on
evidence from speech production with very little that can be said about
perception . Recently, however, Macari (1978) illustrated that these processes do seem to be true mental operations, which he suggests might
also be functional in perception .

CONCLUSION
A markedness analysis of consonant error productions showed that there
is systematic rule-governed behavior to the apraxic impairment in these
three adult aphasic patients . For all cases many error utterances consisted
of phonemes and phoneme sequences in which the values of the features
in the target segments were retained (p=p) . When a change occurred,
however, substitution errors were more likely to reflect a change from
more complex to less complex . That is, segments of lower complexity
tended to replace those of higher complexity more than the reverse
(H-- •L) . Successive attempts at self-correction did not appear to be random . In each case it was possible to establish patterns of self-corrections,
reflecting a progression toward the target segment in terms of markedness
complexity values .
This study attempted to extrapolate from theories within the broad
framework of natural phonology for these adult aphasic cases, as it was
originally developed to explain acquisition of phonology in the child . From
the conclusion that these aphasic cases manifest a general tendency to
replace segments and syllable structure of lower complexity for those of
higher complexity, it may be possible to speculate about the nature of
phonological breakdown for the patients of this study . Where the child
learns to inhibit natural segments and to expand the repertoire of feature
contrasts in the acquisition of his system, these adult aphasics may fail
to suppress the earlier tendencies as a result of neurological disruption .
This may be viewed as "disinhibition," whereby the aphasic adult is
unable to inhibit natural phonological processes . In a sense this may be
seen as related to the regression hypothesis commonly discussed and
originally proposed by Jakobson (1968) . However, it must be stressed
that it is seen in terms of general patterns or trends, not in terms of specific
phoneme production . This explanation bears some relation to many of
the early theorists, particularly Pick (cited in Critchley, 1952), who considered that the essential cause of the general defect in aphasic patients

154

L . WOLK

was a lack of inhibition . He also draws a parallel with normal states of
disinhibition caused by exhaustion and stress, which tend to produce similar lapses in speech .
A markedness analysis may facilitate an assessment of phonemic errors
in terms of their marked and unmarked segments, graded on a continuum
of relative complexity values . Here phonological errors are assessed
within a contextual framework, thus encouraging a certain amount of
flexibility in relation to different phonemic environments . Furthermore,
an evaluation of successive attempts at the target segment in the process
of self-correction may, first, enable the clinician to determine the relative
point at which the patient is in terms of recovery, and, second, provide
insight into predictable phonemic error patterns and vulnerability to
change . The relative complexity values (the scales provided in the Appendix) may facilitate the hierarchical arrangement of speech sounds and
clusters on which remediation of feature contrast processes can be based .
It may be useful to highlight the production of phonemes within a contrastive framework, as discussed by Luria (1970), yet this may be defined
within the scale of complexity postulated in markedness theory . Central
to the concept of "naturalness" is the discoverable "path" from input to
output, within complex sound changes (Darden, 1974) . For example, in
the case of a reduced cluster b/br, distinctive theory would require a jump
from br-> b, whereas markedness theory provides an explanation for the
entire process, possibly from br-s bl -* b in terms of complexity values .
Markedness may thus provide the intermediate steps necessary for prediction of error patterns, evaluation of progress, and accurate planning
of steps toward rehabilitation . Further research is needed to explore the
value of markedness, aimed at improving both diagnostic and therapeutic
efficacy in apraxia of speech .
The writer wishes to offer her sincere thanks to Professor M . L . Aron, Head of the Department of Speech Pathology and Audiology, and Dr . C . Penn, Senior Lecturer in the
Department of Speech Pathology and Audiology, University of the Witwatersrand, Johannesburg, for their guidance and support in carrying out this research . The writer also wishes
to thank Dr . Andrew Spencer for his valuable comments in reviewing the manuscript . For
financial assistance, the writer is grateful to the Witwatersrand University Senior Bursary
Fund and the Human Sciences Research Council, Johannesburg, South Africa .

REFERENCES
Blumstein, S . E . (1973) . A Phonological Investigation ofAphasic Speech . Boston :
Brown University and Aphasia Research Centre ; The Hague : Mouton .
Cairns, C . E . (1969) . Markedness, neutralization, and universal redundancy rules .
Language 45 :863-885 .
Cairns, H . S . (1975) . Phonetic feature theory : The linguist, the speech scientist,
and the speech pathologist . J . Commun . Disord. 8 :157-170 .

MARKEDNESS IN APRAXIA

155

Chomsky, N., and Halle, M . (1968) . The Sound Pattern of English . New York :
Harper & Row .
Critchley, M . (1952) . Articulatory defects in aphasia . J . Laryngol . Otol . 66 :1-17 .
Darden, B . J . (1974) . Introduction in Parasession on Natural Phonology, ed . A .
Bruck, R . A. Fox, and M . W . La Galy . Chicago Linguistic Society, April 18 .
Deal, J . L ., and Darley, F . L . (1972) . The influence of linguistic and situational
variables on phonemic accuracy in apraxia of speech . J . Speech Hear . Res .
15(3) :639-653 .
De Renzi, E ., and Faglioni, P . (1978) . Normative data and screening power of a
shortened version of the token test . Cortex 14(1) :41-49 .
Dunlop, J . M ., and Marquardt, T . P . (1977) . Linguistic and articulatory aspects
of single word production in apraxia of speech . Cortex 13(1) :17-29 .
Goldman, R ., Fristoe, M . S ., and Woodcock, R . W . (1970) . Goldman-Fristoe
Test of Auditory Discrimination . Minnesota : American Guidance Service .
Goodglass, H ., and Kaplan, E . (1972) . The Assessment of Aphasia and Related
Disorders . Boston : Boston Veterans Administration Hospital and Aphasia Research Centre, Department of Neurology, Boston University .
Hatfield, L . G ., and Walton, K . (1975) . Phonological patterns in a case of aphasia .
Lang . Speech 18 :341-357 .
Hooper, J . B . (1976) . An Introduction to Natural Generative Phonology . New
York : Academic Press .
Hyman, L . M . (1975) . Phonology Theory and Analysis. New York : Holt, Rinehart
& Winston .
Ireland, J . V ., Klich, R . .,
J and Panagos, J . M . (1977) . Is there phonemic regression in aphasic speech? Perceptual and Motor Skills 44 :1042 .
Jakobson, R. (1968) . Child Language, Aphasia and Phonological Universals . The
Hague : Mouton .
Joanette, Y ., Keller, E ., and Roch-Lecours, A . (1980) . Sequences of phonemic
approximations in aphasia . Brain Lang . 11 :30-44,
Johns, D . F ., and Darley, F. L . (1970) . Phonemic variability in apraxia of speech .
J . Speech Hear. Res . 13 :556-583 .
Klich, R. .,
J Ireland, J . V ., and Weidner, W . E . (1979) . Articulatory and phonological aspects of consonant substitutions in apraxia of speech . Cortex
15(3) :451-470.
Locke, J . L . (1972) . Ease of Articulation . J . Speech Hear . Res . 15 :194-200 .
Luria, A . R . (1970) . Traumatic Aphasia : Its Syndromes, Psychology, and Treatment. The Hague : Mouton .
Macari, N . J . (1978) . Some psychological evidence for natural phonological processes . J . Phonet . 6 :103-126.
Marquardt, T. P ., Reinhart, J . B ., and Peterson, H . A . (1979) . Markedness analysis of phonemic substitution errors in apraxia of speech . J. Commun . Disord .
12 :481-494 .

156

L . WOLK

Martin, A . D . (1974) . Some Objections to the Term "Apraxia of Speech ." J .
Speech Hear . Dis . 39(1) :53-64 .
Martin, A . D ., and Rigrodsky, S . (1974a) . An investigation of phonological impairment in aphasia, part 1 . Cortex 10(4) :317-328 .
Martin, A . D ., and Rigrodsky, S . (1974b) . An investigation of phonological impairment in aphasia, part 2 : Distinctive Feature Analysis of Phonemic Commutation Errors in Aphasia. Cortex 10(4) :329-346 .
McReynolds, L . V ., Engmann, D ., and Dimmitt, K . (1974) . Markedness theory
and articulation errors . J . Speech Hear . Dis . 39(1) :93-103 .
Moore, W . M ., Rosenbek, J . C ., and La Pointe, L . L . (1976) . Assessment of Oral
Apraxia in Brain-Injured Adults . Proceedings of Clinical Aphasiology Conference, May, Portland, Oregon .
Panagos, J . M ., Quine, M . E ., and Klich, R . J . (1979) . Syntactic and phonological
influences on children's articulation . J . Speech Hear . Res . 22(4) :841-848 .
Parnell, M ., and Amerman, J . D . (1977) . Subjective evaluation of articulatory
effort . J . Speech . Hear . Res . 20 :644-652 .
Sarno, M . T ., and Levita, E . (1971) . Natural course of recovery in severe aphasia .
Arch . Phys . Med. Rehabil . 52 :175-178 .
Smith, M . D ., and Ruder, K . F. (1975) . The Relevance of Markedness . J . Speech
Hear . Dis . 40 :546-548 .
Sommerstein, A . H . (1977) . Modern Phonology . London : Edward Arnold .
Weiner, F . F. (1979) . Phonological Process Analysis . Baltimore : University Park
Press .
Wertz, R . T., LaPointe, L . L ., and Rosenbek, J . C . (1984) . Apraxia of Speech
in Adults : The Disorder and Its Management . New York : Grune & Stratton .
Wolk, L . (1978) . A markedness analysis of initial consonant clusters in aphasic
phonological impairment: A case study . South African Journal of Communication Disorders 25 :81-100 .

Complexity

5U

6U

2

U

U

1

U

M

2

5U

M
U
U

U
U

3

4U

U

M
U
M

U
M
U

U

M
U

v

f

U
U
U

U
M
U

U
M
U

Consonantal
Sonorant
Anterior
Corona)
Continuant
Strident
Voiced
Lateral
Nasal

b

p

Feature

2

6U

U

M
U
U

U
M
U
U

B

3

5U

U

M

U
M
U
U
M
U

$

1

7U

U

U
M
U
U
U
U
U

t

2

6U

U

U
M
U
U
U
U
M

d

4

4U

U

U
M
M
U
M
M
U

f

5

3U

U

M

U
M
M
U
M
M

3

3

5U

U

U
M
M
U
U
M
U

tf

4

4U

U

U
M
M
U
U
M
M

d3

2

5U

M

M
U
U

U
U
U

m

2

6U

M

M
U
U

U
U
U
U

n

3

5U

M

U
U
M
U
M
U
U

n

1

7U

M
U
U
+
U

U
U
U
U

I

2

6U

U

U
M
U
U
M
U
U

r

Appendix A . MfU Values for All Consonants in English, Using a 9 Distinctive Feature Matrix

1

6U

U

M
U
U

U
U
U

w

2

6U

U

U
U
M
U
M
U
U

y

2

5U

U

M
U
U

U
U
M

h

2

6U

U

M
M
U
U
U
U

U

k

3

5U

U

M

M
U
U
U

U
M

g

3

5U

U

U
M
U
U
M
M
U

s

4

4U

U

M

M
U
U
M
M

U

z

158

L . WOLK

Appendix B . Complexity Values for Each
Initial Consonant Cluster in English
Cluster
r blends

I blends

s blends

Complexity value
pr
br
tr
dr
kr
gr
fr
Or
fr
PI
bl
kl
gl
tl
Sill

so
sp

4
5

7
6
7
9
3
4
4
5
5
4
4

SI

Other 2-element blends
3-element blends

sk
sl
sw
tw
kw
spl
spr
Sir
skr
skw

7
7
3
4
8
7
9
7

MARKEDNESS IN APRAXIA

159

Appendix C . Naming Task-Part 1 : Stimulus Items
to Test Single Consonants of English in the Initial and
Final Word Position

n

Initial

Final

man
milk
mouth
monkey
knife
nose
nurse

thumb
tram
drum
pram
pen
train
gun
ring
swing
string
cup
zip
rope
sheep
crib
scrub
*tub
*bib
hat
shirt
pot
rabbit
bird
bride
spade
salad
duck
book
block
sock
pig
dog
frog
jug

7

p

b

t

d

g

r

I

f

pig
pot
pen
pipe
hall
bird
butter
book
tie
torch
teeth
table
dog
duck
doll
dice
cat
cage
cup
keys
gun
girl
garage
game
ring
radio
rabbit
lamp
lemon
leaf
lorry
fork
fire

0

f

Final

feather
fish
vase
vest
violin
thumb
thimble
thin
thick
these/those
this/that
sock
soap
silver
salad
zebra
zoo
zip
sheep
shirt
shoe
ship

giraffe

3
h

j

if

ball
wool
snail
girl
knif
leaf

Initial

d3

hat
house
horse
hammer
wool
whistle
watch
wash
yacht
yellow
chair
cheese
chain
church
jug
giraffe
juice
jersey

* Words not included previously under initial position .
Total = 89 .

*five
store
twelve
teeth
mouth
cloth
-lathe
*bathe
house
horse
dress
nurse
cheese
nose
vase
fish
squash
splash
brush
*beige
*rouge

watch
torch
church
scratch
cage
*badge
fridge
bridge

160

L . WOLK

Appendix D . Naming Task-Pan II : Stimulus Items to Test
Initial Consonant Clusters of English
pr

br

tr

dr

kr

gr

fr

PI

bl

prize
pram
present
bread
brush
bride
bridge
tree
train
tractor
tram
dress
drill
drum
drink
crib
crown
cricket
crayon
grass
green
grapes
fridge
frog
fruit
freckle(s)
plate
plant
plane
plaster
blouse
block(s)
blanket
blazer

kl

gl

fl

sl

sn

Or

fr
tw

kw
sw

clock
clown
cloth
climb
gloves
glass
glasses
flag
fly
flour
flower
slacks
slipper(s)
sleep
snow
snake
snap(s)
snail
three
thread
throw
shred
shrub
twig
twin
twelve
twenty
queen
quickly
swing
sweet(s)
sweep
swim

Total = 99 .
Naming Task- Pan l = 89
Naming Task- Part II = 99
Total
- 188 Stimulus items

sm

sp

at

sit

spl

spr

str

skr

spl

skw

smoke
smell
smile
smarties
spoon
spade
spider
spanner
stove
stains)
stamp
steak
scale
skirt
skittle(s)
ski
split
splash
splinter
spread
spray
sprinkle
stripes
straw
string
strap
scratch
screw
scrub
split
splash
splinter
squirrel
squash
squeezer