This article was downloaded by: [Curtin University of Technology]

On: 23 October 2011, At: 19:21

Publisher: Psychology Press
Informa Ltd Registered in England and Wales Registered Number: 1072954
Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Aphasiology
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/paph20

Therapy for phonological assembly

difficulties: A case series
a b a a
Heather Waldron , Anne Whitworth & David Howard
a
Newcastle University, Newcastle, UK
b
City Hospitals Sunderland NHS Foundation Trust, Sunderland,
UK

Available online: 30 Sep 2010

To cite this article: Heather Waldron, Anne Whitworth & David Howard (2011): Therapy for
phonological assembly difficulties: A case series, Aphasiology, 25:4, 434-455

To link to this article: http://dx.doi.org/10.1080/02687038.2010.494830

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-

conditions

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any
representation that the contents will be complete or accurate or up to date. The
accuracy of any instructions, formulae, and drug doses should be independently
verified with primary sources. The publisher shall not be liable for any loss, actions,
claims, proceedings, demand, or costs or damages whatsoever or howsoever caused
arising directly or indirectly in connection with or arising out of the use of this
material.
 APHASIOLOGY, 2011, 25 (4), 434–455

Therapy for phonological assembly difficulties: A case series

Heather Waldron
Newcastle University, Newcastle, UK, and City Hospitals Sunderland NHS Foundation
Trust, Sunderland, UK

Anne Whitworth and David Howard

Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

Newcastle University, Newcastle, UK

Background: One of the few studies to describe therapy specifically for phonological
assembly difficulties in aphasia is a single-case study by Franklin, Buerk, and Howard
(2002). Their client made a significant improvement in picture naming for both treated
and untreated words after therapy comprising two phases, the first aimed at improving
auditory awareness and the second aimed at improving self-monitoring.
Aims: This study aimed to determine whether the treatment programme used by Franklin
et al. (2002) is replicable with other people with impaired phonological assembly as the
main part of their aphasia, and to explore any differences in the outcomes for partici-
pants.
Methods & Procedures: A case series of four participants with aphasia with mixed impair-
ments including phonological assembly difficulties received a replication of the treatment
protocol described by Franklin et al. (2002).
Outcomes & Results: None of the participants responded to therapy in the same way as
Franklin et al.’s (2002) original client. Three participants improved on naming after the
first, auditory discrimination, phase of therapy, but only one participant made further
gains in naming following the second, monitoring, phase of therapy, and all improve-
ments seen were for treated items only. One participant did not show any significant
improvement on naming of treated or untreated items after either phase of therapy.
Conclusions: Whereas Franklin et al.’s (2002) original client had a relatively pure post-
lexical phonological assembly impairment, the three participants in the current study
whose speech improved after therapy had a combination of lexical and post-lexical
phonological impairments, and it is proposed that their item-specific improvements in
picture naming occurred as a result of improved mapping between semantics and the
phonological output lexicon. The participant in this study whose speech did not improve
following therapy had a combination of phonological assembly difficulties and apraxia
of speech. This study demonstrates that the same therapy can work at different levels for
different individuals, depending on many factors, including their profile of linguistic and
cognitive impairments.
Keywords: Aphasia; Phonological assembly; Therapy; Case series; Monitoring.

Address correspondence to: Heather Waldron, Newcastle University, Speech and Language Sciences
Section, King George VI Building, Queen Victoria Road, Newcastle upon Tyne, NE1 7RU, UK. E-mail:
heather.waldron@ncl.ac.uk
This research was funded by a bursary from The Stroke Association (ref TSAB 2006/02).
© 2010 Psychology Press, an imprint of the Taylor & Francis Group, an Informa business
http://www.psypress.com/aphasiology DOI: 10.1080/02687038.2010.494830
 PHONOLOGICAL ASSEMBLY THERAPY 435

A common symptom of aphasia is the production of phonological errors in speech,

i.e., errors where some or all of the phonemes in a word are substituted by others
or omitted. Understanding the nature of this symptom is complicated, however, as
the phonological system contains many different components (Schwartz, Wilshire,
Gagnon, & Polansky, 2004), which have been underspecified in early cognitive
neuropsychological models. In Levelt, Roelofs, and Meyer’s (1999) WEAVER ++
model of normal speech production, phonological processing is described in more
detail. Levelt et al. (1999) proposed that when a spoken word is produced, a lexical
concept, or semantic representation, must first be activated, which then triggers the
retrieval of an abstract lemma (containing grammatical information) from the men-
tal lexicon. Next, the word’s phonological form is retrieved from the mental lexicon,
containing information about the word’s metrical structure and segmental proper-
Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

ties. Following lexical retrieval, in the process of phonological encoding, segmental

information is first inserted into the metrical frame, and the word is then divided into
syllables according to context in a process called syllabification. After this, a process
of phonetic encoding takes place, where the phonetic gestures for common syllables
are retrieved from a store (the mental syllabary), before articulation of the word. An
alternative model of speech production, devised by Dell, Schwartz, Martin, Saffran,
and Gagnon (1997), proposed that the two stages of lexical access, the lemma level and
phonological level, are linked by a bidirectional network of spreading activation. Dell
et al. (1997) argued that this interaction between levels explains why speakers with
aphasia can produce real-word errors that are both semantically and phonologically
related to the target. A weakness of Dell et al.’s (1997) model, however, is the lack of
detail about post-lexical phonological and phonetic encoding.
Using Levelt et al.’s (1999) model, phonological errors in speech can arise either
due to retrieval of an incorrect or incomplete phonological representation from the
mental lexicon, or from a difficulty within the process of phonological encoding, as in
conduction aphasia. People with aphasia whose phonological errors are due to lexical
retrieval difficulties often have unimpaired speech in repetition, may show an effect
of word frequency on naming (Goldrick & Rapp, 2007) and may use circumlocu-
tions or benefit from phonemic cueing (Whitworth, Webster, & Howard 2005). People
whose phonological errors are caused by a post-lexical phonological assembly deficit
may show a word length effect and make target related phonological errors across all
modalities of naming, reading aloud and repetition (Butterworth, 1992). People with
apraxia of speech (AOS) can also make phonological errors. In its purest form, AOS
is a motor programming disorder that, in Levelt et al.’s model, could be caused by
a deficit in the retrieval of stored phonetic gestures (Ziegler, 2002). AOS frequently
co-occurs with phonological assembly impairments (McNeil, Doyle, & Wambaugh,
2000) which can make it difficult, clinically, to distinguish between them.
Many people with phonological assembly difficulties have shown “conduite
d’approche”. Kohn (1984) argued that these repeated attempts at the target demon-
strate an awareness of errors and therefore intact monitoring, although once an error
has been detected, it is not necessarily self-corrected (Miller & Ellis, 1987). In Levelt
et al.’s (1999) model, monitoring is carried out by the speech perception system, either
through overt monitoring of speech output after articulation, or covert monitoring of
inner speech via an internal feedback loop. According to Whitworth et al. (2005), the
speech perception system has three stages. First the speech sounds in a heard word
are identified by an auditory analysis module; familiar words are then recognised by
the auditory input lexicon, before their meaning is activated in the semantic system.
 436 WALDRON

A deficit in any of these processes could, then, be responsible for a difficulty in self-
monitoring. However, Nickels and Howard (1995) found no link between auditory
comprehension abilities and proportion of attempted self-corrections in a group of
15 people with aphasia, instead they argued for a production based pre-articulatory
monitor.

THERAPY FOR PHONOLOGICAL ASSEMBLY IMPAIRMENTS

In a review of naming therapy studies Nickels (2002) drew a distinction between
phonological therapy tasks and therapy for phonological impairments. Many stud-
ies have shown positive effects of using phonological tasks in therapy (e.g., Raymer,
Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

Thompson, Jacobs, & Le Grand, 1993; Hickin, Best, Herbert, Howard, & Osborne,
2002) but these have usually involved participants with a range of underlying causes
underpinning their word-finding difficulties, mostly either within or in accessing the
phonological representation from the lexicon.
One of the few treatment studies to use a well-designed format specifically targeting
phonological assembly impairments is that of Franklin, Buerk, and Howard (2002).
Therapy was carried out with participant MB in two phases, the first aimed at improv-
ing MB’s auditory awareness while the second aimed at improving her self-monitoring
skills. A significant improvement in picture naming for both treated and untreated
words was seen after both phases, and these improvements were maintained 2 months
later. Gains in repetition, reading aloud, and self-correction of errors were also found.
However, the authors acknowledged that the reasons why this therapy caused MB’s
speech to improve were unclear as, although their original aim was to teach a self-
monitoring strategy, it was actually the process of phonological encoding that had
improved.
The finding by Franklin et al. (2002) that MB showed generalisation of improve-
ments from treated to untreated words is unusual because many word retrieval treat-
ment studies have found only item specific effects (e.g., Miceli, Amitrano, Capasso, &
Caramazza, 1996). Best et al. (2006) suggested that therapy for phonological assembly
impairments is more likely to achieve generalisation to untreated items than therapy
aimed at the phonological output lexicon because it is targeting a process or strat-
egy, rather than specific lexical entries. Best et al. (2006) reported that 4 out of 16
participants with anomia who received a phonological cueing therapy showed gener-
alisation to untreated items. They analysed the assessment data of these participants
and found that the people who showed generalisation were those with a pattern of
impairment consistent with a phonological assembly deficit, shown by phonological
errors in output with less accurate production of longer targets.

AIMS OF THE STUDY

Using a case series design, this study had the following aims:
1. To determine whether the treatment programme used by Franklin et al. (2002) is
replicable with other people with impaired phonological assembly as the main part
of their aphasia.
2. To explore any differences in the outcomes for participants.
3. To identify any factors that suggest which people will benefit most from this
therapy.
 PHONOLOGICAL ASSEMBLY THERAPY 437
TABLE 1
Participant details

Participant SD BB HS PL

Gender Female Female Male Male

Age 75 years 76 years 63 years 82 years
Medical history Left parietal lobe Left middle Left parietal lobe Left middle
infarct cerebral artery infarct cerebral artery
infarct infarct
Months 5 32 45 5
post-onset
Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

METHOD
Participants
Four participants were recruited through local speech and language therapy services.
Background information is shown in Table 1. All participants were right-handed,
monolingual English speakers, and were deemed suitable for twice weekly assessment
and therapy, with sufficiently good comprehension to give consent. All had aphasia
following a stroke more than 3 months prior to the beginning of the study, although
there was variation in time post onset, with two participants in a chronic stage of
recovery, and two still relatively acute. Information about lesion site is limited to that
provided by the referrer. Each participant presented with the same primary symptom
of phonological errors in picture naming, word repetition, and reading aloud, consid-
ered to reflect a post-lexical phonological impairment, but they had different patterns
of co-occurring impairments, with the view that information would be gained about
which factors are important in identifying the best candidates for Franklin et al.’s ther-
apy. People with dementia, or any speech or language difficulties that were not a result
of their stroke, were excluded from the study.

Study design
Franklin et al.’s (2002) therapy protocol was replicated with all four participants, each
undergoing two consecutive therapy phases. The two phases of therapy were given in
the same order to all participants, first in order to achieve a true replication of Franklin
et al.’s (2002) therapy, and second because the first auditory discrimination phase may
be a necessary precursor to the second monitoring phase. Language assessment was
carried out with each participant on five occasions. Two assessment periods took place
1 month apart prior to the intervention period, the third and fourth after each ther-
apy phase, and a final assessment was carried out 2 months after therapy. During the
first assessment period, a selection of subtests from the Psycholinguistic Assessments
of Language Processing in Aphasia (PALPA) (Kay, Lesser, & Coltheart, 1992) and
several other cognitive assessments (see Table 2) were administered, both to gain a
detailed picture of each participant’s language and cognition, and to hypothesise the
level, or levels, of breakdown in their language-processing abilities. Spoken naming
was then reassessed after a break of 1 month, during which no therapy or assessment
was carried out, to establish a stable baseline and control for any spontaneous recov-
ery that might influence results. Selected assessments, including spoken naming, were
re-tested after each phase of therapy, with spoken naming then reassessed 2 months
 438 WALDRON

after the end of therapy to establish whether any gains had been maintained. The
written version of the Test for Reception of Grammar (TROG) (Bishop, 1982) was
administered before and after therapy as a control task, as it was not anticipated
that written sentence comprehension would improve after therapy targeting auditory
discrimination or monitoring.
Spoken picture naming was chosen as the primary outcome measure, and was
assessed using the Nickels naming test (Nickels, 1992) at all five assessment periods.
This test, used in the Franklin et al. (2002) study, would permit direct comparisons
to be made with the earlier study and, owing to the large number of items (130)
orthogonally varied by number of syllables and word frequency, would allow both
measurement of any change and an examination of whether either of these two vari-
ables had an impact on word retrieval. After the initial assessment, the items from the
Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

Nickels naming test were divided into two sets, one to be used in treatment, the other
to be left untreated, each containing equal numbers of items named correctly on the
first attempt (not including self-corrections) and matched for syllable length and word
frequency. Reading aloud and repetition of the same words used in the Nickels nam-
ing test was also tested before therapy, after the first phase of therapy and after the
second phase of therapy, to look for any difference between patterns of improvement
on naming, reading and repetition, which may inform discussion as to how therapy
worked.
In addition, all four participants had their hearing tested at the start of the study
using pure tone audiometry. This was done to assess the degree to which any hearing
impairment may have been impacting on participants’ auditory processing skills, and
was particularly important given that many of the therapy tasks focused on listening.
Audiometric testing was carried out according to the British Society of Audiology’s
recommended procedure and took place in participants’ homes, with the room made
as quiet as practically possible. SD, BB, and HS were shown to have either normal
hearing or a mild loss (threshold of 30dB or less) for the frequencies required to
hear most speech sounds (500–2000Hz). This was not considered a significant hearing
impairment as all participants were aged over 60 and testing took place in ambient
noise. PL wore a hearing aid in his left ear for all assessment and therapy, including
the audiometry test. Even when aided, PL had a moderate hearing loss (threshold of
70dB or less) in his left ear between 500 and 2000 Hz, but only a mild loss for these
frequencies in the right ear. All participants had a greater degree of loss at the highest
frequencies (4000 and 8000 Hz), which was expected for their ages. To compensate for
PL’s mild to moderate hearing loss, all auditory input assessments were presented with
a raised voice and the therapist sat to his better, right-hand side.
All assessment and therapy was administered by the first author, who is a qualified
speech and language therapist. Participants’ spoken responses were recorded using a
Roland Edirol R-09 digital audio recorder.

Background assessments
The results of each participant’s pre-therapy assessment are set out in Table 2.
Raw scores and proportions correct are given for each participant, as well as mean
performance from normal controls, where published in the assessment manual.

Assessment of auditory input processing. On the tests of minimal pair discrimina-

tion, HS scored within normal limits, indicating intact auditory analysis. Similarly, HS
 Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

TABLE 2
Background assessments

SD BB HS PL Normal mean

PALPA 1 Auditory Discrimination 19/36 (.53) 21/36 (.58) 35/36 (.97) 53/72 (.74) 70/72 (.97)
of Nonword Minimal Pairs
PALPA 2 Auditory Discrimination 38/72 (.53) 61/72 (.85) 36/36 (1.00) 62/72 (.86) 70/72 (.97)
of Word Minimal Pairs
PALPA 5 Auditory Lexical 129/160 (.81) 132/160 (.83) 154/160 (.96) 141/160 (.88) 155/160 (.97)
Decision
PALPA 15 Auditory Rhyme 37/58 (.64) 39/58 (.67) 57/58 (.98) 55/58 (.95) not available
Judgement
PALPA 14 Picture Rhyme 24/40 (.60) 16/40 (.40) 21/40 (.53) 22/40 (.55) not available
Judgement
PALPA 28 30/60 (.50) 47/60 (.78) 41/60 (.68) 43/60 (.72) 54/60 (.92)
Homophone Decision Regular: 12/20 Regular: 17/20 Regular: 16/20 Regular: 17/20 Regular: 18/20
Exception: 10/20 Exception: 18/20 Exception: 15/20 Exception:17/20 Exception: 18/20
Nonword: 8/20 Nonword: 12/20 Nonword: 10/20 Nonword: 9/20 Nonword: 18/20
(Nickels &
Cole-Virtue, 2004)
PALPA 13 Digit Matching Span 3.5 4.9 4.5 5.5 not available
PALPA 47 Spoken Word-Picture 38/40 (.95) 39/40 (.98) 39/40 (.98) 39/40 (.98) 39/40 (.98)
Match
PALPA 48 Written Word-Picture 33/40 (.83) 37/40 (.93) 39/40 (.98) 37/40 (.93) 39/40 (.98)
Match
Pyramids and Palm Trees 36/52 (.69) 45/52 (.87) 49/52 (.94) 46/52 (.88) 51/52 (.98)
(3 picture version)
Camden Short Recognition 15/25 (.60) 21/25 (.84) 25/25 (1.00) 19/25 (.76) see centile result
Memory Test for Faces <5th centile 25th centile 90th centile 10th centile
Wisconsin Card Sorting Test 0 categories correct 0 categories correct 3 categories completed 2 categories completed see centile result
6–10 centile 6–10 centile > 16th centile > 16th centile
PHONOLOGICAL ASSEMBLY THERAPY

69% error responses 66% error responses 46% error responses 56% error responses
6th centile 8th centile 12th centile 27th centile
439
 440 WALDRON

scored near to normal on auditory lexical decision, indicating an intact auditory input
lexicon. BB and PL were impaired compared to normal performance on discrimina-
tion of both word and nonword minimal pairs, but they were better at discriminating
real words, suggesting they may have been using some intact lexical processing to help
make this decision. Further evidence for BB and PL having access to the auditory
input lexicon comes from their scores on auditory lexical decision, which were simi-
lar to their real-word minimal pair discrimination scores (above 80%). SD was severely
impaired at discrimination of both word and nonword minimal pairs, in fact her scores
of around 50% on both tasks were no better than would be predicted by chance. SD’s
superior performance on auditory lexical decision, however, shows that her auditory
discrimination abilities may have been better than her minimal pair discrimination
scores suggest. Interestingly, participants’ scores on assessment of auditory processing
Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

did not appear related to their pure tone audiometry results. PL had the most impaired
hearing but still performed relatively well on minimal pair discrimination. SD, BB and
HS all had a similar mild hearing loss but HS still performed within normal limits on
these tasks.

Assessment of phonological processing. Phonological processing was further

explored through auditory rhyme judgement, picture rhyme judgement (which
involved looking at two pictures and, without naming them aloud, deciding whether
the two words rhyme), and written homophone decision (which involved reading two
words silently then deciding whether the two words would sound the same if they were
said out loud). The results in Table 2 show that BB, HS, and PL tended to perform
better on homophone decision than on picture rhyme judgement, although the differ-
ence was only statistically significant for BB (Fisher exact p = .0001), and homophone
judgement for nonwords was still impaired. BB, HS, and PL also performed better on
auditory rhyme judgement than picture rhyme judgement, although BB still showed
some difficulty with the auditory task, whereas PL and HS scored above 90% on
the auditory version. SD was similarly impaired on both auditory and picture rhyme
judgements and homophone judgement.
Howard and Franklin (1990) argued that homophone judgement can be performed
by accessing output phonology alone, without access to input phonological processes,
whereas rhyme judgements require access to auditory input information as well as
output phonology, because segmentation of the word into onset and rime is required.
Therefore, the pattern shown by BB, HS, and PL of relatively good performance on
homophone judgement and auditory rhyme judgement compared with poor perfor-
mance on picture rhyme judgement is compatible with a deficit in the sublexical link
between phonological assembly and auditory analysis, a pattern also seen in five par-
ticipants studied by Nickels, Howard, and Best (1997). Alternatively, the difficulties
with picture rhyme judgement faced by BB, HS, and PL could be due to a problem in
lexical retrieval. Both the Howard and Franklin (1990) and Nickels et al. (1997) stud-
ies used written rhyme judgement, which does not require lexical retrieval. It is also
plausible that the picture rhyme judgement task is more difficult than the written ver-
sion. To confirm a deficit in the sublexical output–input link, scores on both written
and picture rhyme judgements would be needed. However, the low scores of BB, HS,
and PL on picture rhyme judgement were not due to impaired short-term memory
because, as Table 2 shows, all three scored relatively highly on the digit matching span
assessment.
 PHONOLOGICAL ASSEMBLY THERAPY 441

Assessment of semantics and cognitive skills. Table 2 shows that all four partici-
pants scored above 90% correct on the spoken word to picture matching task, and
all participants, except SD who scored 83%, also scored above 90% on written word
to picture matching. This indicates that no participant had a severe impairment in
accessing the semantic representations of concrete words from either spoken or written
input. HS performed within the normal cut-off of 49/52 on the three-picture version
of the Pyramids and Palm Trees test (Howard & Patterson, 1992), indicating that his
access to the semantic system from pictures was largely intact. BB and PL scored
just below normal limits on this test, and SD scored well below normal performance
(69%), showing that her nonverbal semantic system may have been impaired, despite
good access to semantics on the easier word picture matching tasks.
Participants’ non-linguistic cognitive abilities were also assessed using the Camden
Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

Short Recognition Memory Test for Faces (Warrington, 1996) and the Wisconsin Card
Sorting Test (Grant & Berg, 1993) assessments of executive function and problem
solving. In a study on the use of errorless learning to treat anomia, Fillingham, Sage,
and Lambon Ralph (2006) found that participants who performed well on these tests
also made the greatest gains in therapy. These scores may therefore be relevant when
explaining any differences in treatment outcomes. On the recognition memory test, HS
had no difficulty, BB and PL both scored toward the lower end of the normal range
for elderly controls, and SD performed well below the norm. On the card-sorting test,
SD and BB were both unable to perform the task. HS and PL had some success,
falling at the 12th and 27th percentile respectively for percentage error responses, but
both were below average. It should be noted that the percentile results are determined
according to the participant’s age, so HS made fewer error responses than PL but fell
at a lower percentile because of his younger age. Interestingly, HS’s performance was
low in comparison to his near normal language performance and good performance
on the face recognition test.

Assessment of spoken word production. The results from tests of single-word spoken
production for each participant at the first assessment period are shown in Table 3.
Raw scores and proportions correct are given. The scoring uses the final response,
i.e., including self-corrections. All participants showed a deficit in producing spoken
words. HS had the highest scores across all tasks, followed by BB, then SD, with
PL scoring lowest across all tasks. All participants were impaired across all modal-
ities to some degree, and were impaired for nonwords as well as real words. Using
the model of spoken word production described earlier, a deficit involving all spoken
output tasks, i.e., a post-lexical impairment in phonological assembly, is suggested
for all participants. In addition, PL showed some of the characteristics of apraxia of
speech (AOS) suggested by McNeil, Pratt, and Fossett (2004), including vowel distor-
tions, slowed speech rate and articulatory groping. Further assessment on the Apraxia
Battery for Adults (Dabul, 1979) supported a diagnosis of moderate to severe AOS. On
the Dysarthria Profile (Robertson, 1982), PL scored either normal or good on 19/20
dimensions for facial musculature, ruling out any muscular weakness.

Length and frequency effects. Table 3 shows participants’ performance on high- and
low-frequency and one-, two-, and three-syllable words on the pre-therapy spoken
word production assessments. A Fisher exact test (one-tailed) revealed a significant
effect of word frequency on spoken naming for SD (p = .034) and on reading aloud
 442 WALDRON

TABLE 3
Pre-therapy assessment of spoken word production

SD BB HS PL

Spoken picture Total correct (n = 130) 32 (.25) 48 (.37) 73 (.62) 4 (.03)

naming: High Frequency (n = 65) 21 (.32) 20 (.31) 37 (.57) 3 (.05)
Nickels Low Frequency (n = 65) 11 (.17) 28 (.43) 36 (.55) 1 (.02)
naming test 1 syllable (n = 50) 16 (.32) 20 (.40) 35 (.70) 1 (.02)
2 syllable (n = 50) 12 (.24) 19 (.38) 27 (.54) 2 (.04)
3 syllable (n = 30) 4 (.13) 9 (.30) 11 (.37) 1 (.03)
Repetition of Total correct 43 (.33) 79 (.61) 110 (.85) 17 (.13)
Nickels High Frequency 24 (.37) 36 (.55) 56 (.86) 11 (.17)
naming test Low Frequency 19 (.29) 43 (.66) 54 (.83) 6 (.09)
Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

words 1 syllable 19 (.38) 29 (.58) 44 (.88) 9 (.18)

2 syllable 20 (.40) 31 (.62) 46 (.92) 6 (.12)
3 syllable 4 (.13) 19 (.63) 20 (.67) 2 (.07)
Reading Aloud Total correct 56 (.43) 69 (.53) 107 (.82) 12 (.09)
of Nickels High Frequency 32 (.49) 37 (.57) 58 (.89) 6 (.09)
naming test Low Frequency 24 (.37) 32 (.49) 49 (.75) 6 (.09)
words 1 syllable 29 (.58) 25 (.50) 46 (.92) 6 (.12)
2 syllable 22 (.44) 30 (.60) 43 (.86) 6 (.12)
3 syllable 5 (.17) 14 (.47) 18 (.60) 0
PALPA 8 Nonword Repetition (n = 30) 1 (.03) 11 (.37) 12 (.40) 0
PALPA 8 Nonword Reading (n = 30) 0 2 (.07) 5 (.17) 0

for HS (p = .034), with more high-frequency words produced correctly than low-
frequency words, but there were no significant effects of frequency on any other tasks,
or for any other participant. On a Jonckheere Trend Test (one-tailed), SD and HS
both showed significant effects of syllable length on all speech production tasks, with
more one-syllable words produced correctly than three-syllable words (for SD, nam-
ing p = .043, repetition p = .037, and reading p < .001, and for HS, naming p = .003,
repetition p = .028, and reading p = .001). BB and PL did not show a significant
effect of syllable length on any task (for BB, naming p = .239, repetition p = .346,
and reading p = .455, and for PL, naming p = .457, repetition p = .095, and reading
p = .088).

Speech errors. Table 4 shows the number of different error types, as a proportion
of the total number of incorrect first responses, made by each participant on the three
tasks of naming, real-word repetition, and reading aloud pre-therapy. Proportions are
given for naming both the full 130 items from the Nickels naming test and for an
edited set of 63 of the items that gained 90% naming agreement with elderly controls,
as the remaining 67 items achieved 90% naming agreement only with younger con-
trols (Nickels & Howard, 1994). Errors were classified as phonologically related, if
they shared 50% or more of their phonemes with the target in any order; unrelated,
if they shared less than 50% of their phonemes with the target; other, which included
semantically related or semantically and phonologically related real words as well as
nonwords that shared 50% or more of their phonemes with a semantically related
word; or no response. The first response containing both a consonant and a vowel was
analysed, i.e., those false starts containing only a single phoneme were not counted
as a response. Items produced correctly immediately were excluded from this analysis,
but error responses that were subsequently self-corrected were included.
 PHONOLOGICAL ASSEMBLY THERAPY 443
TABLE 4
Types of speech error on pre therapy naming, repetition, and reading aloud

SD BB HS PL

Naming full set (n = 108) (n = 91) (n = 80) (n = 127)

Phonologically related 31 (.29) 33 (.36) 33 (.41) 26 (.20)
Unrelated 23 (.21) 24 (.26) 7 (.09) 31 (.24)
Other 27 (.25) 32 (.35) 36 (.45) 5 (.04)
No response 27 (.25) 2 (.02) 4 (.05) 65 (.51)
Naming edited set (n = 48) (n = 38) (n = 31) (n = 60)
Phonologically related 21 (.44) 16 (.42) 15 (.48) 16 (.27)
Unrelated 12 (.25) 9 (.24) 4 (.13) 15 (.25)
Other 8 (.17) 13 (.34) 10 (.32) 2 (.03)
Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

No response 7 (.15) 0 2 (.06) 27 (.45)

Repetition (n = 95) (n = 62) (n = 21) (n = 116)
Phonologically related 70 (.74) 55 (.89) 21 (1.0) 47 (.40)
Unrelated 21 (.22) 6 (.09) 0 17 (.15)
Other 4 (.04) 1 (.02) 0 0
No response 0 0 0 52 (.45)
Reading aloud (n = 91) (n = 71) (n = 35) (n = 119)
Phonologically related 66 (.73) 66 (.93) 35 (1.0) 44 (.37)
Unrelated 23 (.25) 4 (.06) 0 44 (.37)
Other 2 (.02) 1 (.01) 0 0
No response 0 0 0 31 (.26)

All four participants made high numbers of phonologically related errors on all
three tasks, supporting the diagnosis of a post-lexical phonological assembly impair-
ment. In addition, SD, BB, and HS made some errors in the “other” category on
naming, even when the edited set with better naming agreement was used, suggest-
ing they also had some lexical retrieval difficulties. PL made very few of these lexical
errors, but had high numbers of no responses across all three tasks. The unrelated
nonword errors produced by all four participants could arise either because of a lexi-
cal retrieval difficulty, as in jargon aphasia (Marshall, 2006), or a severe phonological
output impairment (Franklin et al., 2002).

Summary
Using Levelt et al.’s (1999) model of spoken word production, all four participants
showed evidence of a post-lexical phonological assembly deficit, as all spoken output
modalities were impaired, phonological errors occurred on all output tasks, and SD
and HS showed effects of word length on spoken production. In addition, SD, BB, and
HS had co-occurring lexical retrieval difficulties, shown by semantic errors on picture
naming and lower scores on naming compared with repetition and reading, plus an
effect of word frequency on naming for SD. In contrast, PL had co-occurring AOS,
shown by many no responses on all tasks, and features such as vowel distortions and
articulatory groping.

Therapy procedure
Participants were seen for therapy twice a week in their homes. Sessions lasted
approximately 45 minutes. The first phase of therapy aimed to improve auditory
 444 WALDRON

discrimination and was administered over six sessions. Tasks were single sound to let-
ter matching, choosing the initial or final sound for a spoken word, deciding whether
two heard words had the same or different final sound, and choosing a written word
that rhymes with a spoken word. The second phase of therapy aimed to improve
monitoring of speech errors and took place over 14 sessions. There were three stages
within the second phase. For the first six sessions of the monitoring therapy partic-
ipants listened to the therapist naming a picture and were required to decide if the
word sounded right or wrong. If the therapist had named the picture correctly, partic-
ipants repeated the word back before moving on to the next item. If the therapist had
made an error, participants were asked to decide whether the phonological error was
at the beginning, middle, or end of the word (by pointing to a written prompt sheet),
and then they were asked to produce the word correctly. In the next four sessions
Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

the participants were audio recorded while they named the pictures themselves. They
then heard their responses played back and were asked to decide if they had said the
word correctly. If they had made an error, they had to decide on the location of their
error and then correct themselves. In the final four sessions this was done without the
audio recording and participants were asked to make the judgements about their own
errors online. The only change made to the procedure used by Franklin et al. (2002)
was that participants were given homework, based on the therapy tasks, to carry out
between sessions. This was done in order to maximise therapy effects and because all
participants were keen to do work independently.

THERAPY OUTCOMES
Test scores for each participant on naming, repetition, and reading aloud of the items
from the Nickels naming test, as well as scores from nonword repetition and read-
ing, minimal pair discrimination, auditory lexical decision, and the written TROG,
from the various assessment points during the study, are shown in Table 5. Raw scores
and proportions correct are given. The final response is shown for all tasks, including
self-corrections. Results for naming of treated and untreated items from the Nickels
naming test for each participant are also shown in Figure 1. McNemar’s test (one-
tailed) was used to assess the significance of any improvement seen, and a Fisher
exact test (one-tailed) was used to compare treated and untreated sets. No participant
showed any significant change in spoken naming between the repeated baseline of the
two pre-therapy assessments, and no participant showed any significant improvement
on the control task of the written TROG between the start and end of therapy. Taken
together, this supports the hypothesis that any positive effects seen were due to therapy.
Individual participants’ results are discussed in detail below.

Participant SD
There was a significant improvement in the number of treated items named correctly
by SD after phase 1 of therapy (30/65) compared with her previous pre-therapy 2 score
(16/65) (p = .001) but naming of untreated items did not improve (p = .073 compar-
ing score after phase 1 of therapy with score at pre-therapy 2, p = .598 compared with
score at pre-therapy 1, and p = .005 comparing scores on treated and untreated items
after phase 1 of therapy). There was no change in SD’s spoken naming scores after
phase 2 of therapy, compared with her previous scores after phase 1 (p = .598 and
p = .500 for treated and untreated items respectively). On assessment 2 months after
 PHONOLOGICAL ASSEMBLY THERAPY 445
TABLE 5
Results of assessments after therapy

SD BB HS PL

Spoken Naming
Pre-therapy 1 Total correct (n = 130) 32 (.25) 48 (.37) 73 (.62) 4 (.03)
Treated items (n = 65) 17 (.26) 26 (.40) 41 (.63) 2 (.03)
Untreated items (n = 65) 15 (.23) 22 (.34) 32 (.49) 2 (.03)
Pre-therapy 2 Total correct 25 (.20) 42 (.32) 82 (.63) 6 (.05)
Treated items 16 (.25) 26 (.40) 43 (.66) 3 (.05)
Untreated items 9 (.14) 16 (.25) 39 (.60) 3 (.05)
Post phase 1 therapy Total correct 45 (.35) 54 (.42) 89 (.68) 10 (.08)
Treated items 30 (.46) ∗ 35 (.54) ∗ 52 (.80) ∗ 4 (.06)
Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

Untreated items 15 (.23) 19 (.29) 37 (.57) 6 (.09)

Post phase 2 therapy Total correct 46 (.35) 54 (.42) 104 (.80) 9 (.07)
Treated items 30 (.46) 40 (.61) 62 (.95) ∗ 8 (.12)
Untreated items 16 (.25) 14 (.22) 42 (.65) 1 (.02)
2 months post-therapy Total correct 42 (.32) 48 (.37) 100 (.77) 13 (.10)
Treated items 25 (.38) 33 (.51) 54 (.83) 6 (.09)
Untreated items 17 (.26) 15 (.23) 46 (.71) 7 (.11)
Repetition
Pre-therapy Total correct 43 (.33) 79 (.61) 110 (.85) 17 (.13)
Treated items 20 (.31) 44 (.68) 58 (.89) 9 (.14)
Untreated items 23 (.35) 35 (.54) 52 (.80) 8 (.12
Post phase 1 therapy Total correct 58 (.45) 83 (.64) 104 (.80) 12 (.10)
Treated items 30 (.46) ∗ 41 (.63) 54 (.83) 6 (.09)
Untreated items 28 (.43) 42 (.65) 50 (.77) 6 (.09)
Post phase 2 therapy Total correct 54 (.42) 87 (.67) 122 (.94) 9 (.07)
Treated items 31 (.48) 50 (.77) 64 (.98) ∗ 5 (.07)
Untreated items 23 (.35) 37 (.57) 58 (.89) ∗ 4 (.06)
Reading Aloud
Pre-therapy Total correct 56 (.43) 69 (.53) 107 (.82) 12 (.09)
Treated items 31 (.48) 39 (.60) 55 (.85) 5 (.07)
Untreated items 25 (.38) 30 (.46) 52 (.80) 7 (.11)
Post phase 1 therapy Total correct 66 (.51) 64 (.49) 121 (.93) 20 (.15)
Treated items 36 (.55) 34 (.52) 60 (.92) 13 (20) ∗
Untreated items 30 (.46) 30 (.46) 61 (.94) ∗ 7 (.11)
Post phase 2 therapy Total correct 63 (.48) 68 (.52) 120 (.92) 11 (.08)
Treated items 37 (.57) 35 (.54) 62 (.95) 7 (.11)
Untreated items 26 (.40) 33 (.51) 58 (.89) 4 (.06)
PALPA 8 Nonword Pre-therapy 1 (.03) 12 (.40) 12 (.40) 0
Repetition (n = 30) Post phase 2 therapy 0 12 (.40) 18 (.60) 0

PALPA 8 Nonword Pre-therapy 0 3 (.10) 5 (.17) 0

Reading (n = 30) Post phase 2 therapy 0 7 (.23) 9 (.30) 0

PALPA 2 Auditory Pre-therapy 38 (.53) 61 (.85) 36/36 (1.00) 62 (.86)

Discrimination of Word Post phase 2 therapy 56 (.78) ∗ 66 (.92) ∗∗ 62 (.86)
Minimal Pairs (n = 72)

PALPA 5 Auditory Lexical Pre-therapy 129 (.81) 69/80 (.86) 154 (.96) 141 (.88)
Decision (n = 160) Post phase 2 therapy 138 (.86) 67/80 (.84) ∗∗ 146 (.91)

Written TROG Pre-therapy 25/44 (.56) 45/60 (.75) 63/80 (.79) 40/52 (.77)
Post phase 2 therapy 29/44 (.66) 46/60 (.77) 67/80 (.84) 41/52 (.79)
∗ = statistically significant change compared with the previous assessment (McNemar’s test). ∗∗ HS was not
re-assessed on minimal pair discrimination or auditory lexical decision post therapy because his scores were
within normal limits pre-therapy.
 446 WALDRON

Proportion correct
0.8

0.6
SD
0.4

0.2

0
Pre therapy 1 Pre therapy 2 Post phase 1 Post phase 2 2 months post
therapy therapy therapy

1
Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

Proportion correct

0.8

0.6
BB
0.4

0.2

0
Pre therapy 1 Pre therapy 2 Post phase 1 Post phase 2 2 months post
therapy therapy therapy

1
Proportion correct

0.8

0.6
HS
0.4

0.2

0
Pre therapy 1 Pre therapy 2 Post phase 1 Post phase 2 2 months post
therapy therapy therapy

1
Proportion correct

0.8

0.6
PL
0.4

0.2

0
Pre therapy 1 Pre therapy 2 Post phase 1 Post phase 2 2 months post
therapy therapy therapy

Treated items Untreated items

Figure 1. Naming of treated and untreated items after therapy.

 PHONOLOGICAL ASSEMBLY THERAPY 447

the end of therapy, SD’s score on naming of treated items had fallen slightly and the
difference between naming of treated and untreated items was no longer significant
(p = .095). Her score on naming of treated items was still significantly higher than
it had been before therapy (p = .038 comparing pre-therapy 1 score with 2 months
post therapy score). SD also made a significant improvement in repetition of treated
items after phase 1 of therapy (p = .038) but repetition of untreated items did not
improve significantly (p = .192) and there was no further improvement in repetition
after phase 2 of therapy (p = .500 and p = .895 for treated and untreated items respec-
tively). SD showed no significant change in reading aloud after phase 1 (p = .192 and
p = .166 for treated and untreated items respectively) or phase 2 of therapy (p = .500
and p = .910 for treated and untreated items respectively). Repetition and reading of
nonwords remained at floor levels post therapy. After therapy, SD showed a significant
Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

improvement in auditory discrimination of word minimal pairs (p = .001) but not in

auditory lexical decision (p = .094).

Participant BB
Like SD, the number of treated items named correctly by BB improved significantly
after phase 1 of therapy (35/65), compared with the previous pre-therapy 2 score
(26/65) (p = .047) but naming of untreated items did not improve significantly
(p = .324 comparing naming of untreated items after phase 1 of therapy, 19/65, with
the pre-therapy 2 score, 16/65, and p = .004 comparing treated and untreated sets after
phase 1 of therapy) and there was no significant change in her naming after phase 2 of
therapy compared with previous scores after phase 1 (p = .212 and p = .895 for treated
and untreated items respectively). However, the improvement in naming of treated
items was not maintained 2 months after the end of the study (p = .141 comparing
pre therapy 1 score with 2 months post therapy score) although the difference between
treated and untreated sets remained significant (p = .001). BB showed no significant
change in repetition or reading aloud of treated or untreated items after either therapy
phase. BB’s nonword repetition score was unchanged after therapy and her nonword
reading score had not increased significantly (p = .145). After therapy, BB’s auditory
discrimination of word minimal pairs had not changed significantly (p = .166), and
neither had her auditory lexical decision score (p = .387).

Participant HS
HS showed a significant improvement in the number of treated items named cor-
rectly after phase 1 of therapy (52/65) compared with his previous pre-therapy 2 score
(43/65) (p = .032). Unlike SD or BB, there was a further significant improvement in
naming of treated items after phase 2 of therapy (62/65) compared with after phase
1 (p = .006). He did not show any significant improvement in naming of untreated
items (p = .773 comparing pre-therapy 2 score with score after phase 1, p = .113 com-
paring score after phase 1 with score after phase 2, and comparing scores on treated
and untreated sets, p = .004 after phase 1 and p = .001 after phase 2). As with SD, on
reassessment 2 months after the end of therapy, HS’s score on naming of treated items
had fallen slightly but was still significantly higher than it had been before therapy
(p = .001 comparing pre-therapy 1 score with 2 months post therapy score). In addi-
tion, although his naming of untreated items had not shown a significant improvement
after either therapy phase, HS’s score on naming of untreated items 2 months after the
 448 WALDRON

end of therapy was also significantly higher than before therapy (p < .001) and the
difference between treated and untreated sets at the final assessment was not signif-
icant (p = .073). HS did not show any improvement in repetition after phase 1 of
therapy (p = .927 and p = .788 for treated and untreated items respectively), but he
did show a significant improvement in repetition of treated (p = .001) and untreated
items (p = .029) after phase 2 of therapy in comparison with his previous scores after
phase 1. HS showed no improvement in reading of treated items after either therapy
phase (p = .113 and p = .313 after phase 1 and phase 2 respectively). He did show a
significant improvement in reading of untreated items after phase 1 (p = .011), but not
after phase 2 (p = .969). An increase in nonword repetition after therapy approached
statistical significance (p = .055), but nonword reading did not improve significantly
(p = .172). HS did not show any difficulty with minimal pair discrimination or lexical
Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

decision at the first assessment so he was not re-tested on these after therapy.

Participant PL
PL’s scores on spoken naming over successive tests showed no significant changes after
either therapy phase for treated or untreated items. After phase 2 of therapy there
was a significant difference between naming of treated and untreated sets (p = .019)
due to a small increase in the score for treated items and a decrease in the score for
untreated items, but on all other assessments the difference between sets was not sig-
nificant. There was a significant overall improvement across all five naming trials for
both treated (Wilcoxon one sample one-tailed p = .028) and untreated (Wilcoxon one
sample one-tailed p = .033) items, but there was a similar amount of change during the
untreated periods as during the treatment phases. Most improvement occurred after
the end of treatment, suggesting that this overall improvement cannot be attributed
reliably to therapy effects, even though his performance on the control task of the
written TROG did not change. PL showed no significant improvement in repetition of
treated or untreated items, in fact his scores on real-word repetition decreased over the
course of the study. His reading aloud of treated words did improve significantly after
phase 1 of therapy (p = .011) but went down again after phase 2, and there was no
change in reading aloud of untreated items (p = .637 and p = .938 after phase 1 and
phase 2 respectively). PL’s reading and repetition of nonwords remained at floor lev-
els after therapy. When PL’s auditory discrimination was re-tested after therapy, there
was no change in his scores on word minimal pair discrimination and no significant
change on auditory lexical decision (McNemar’s test one-tailed p = .192).

Changes in naming errors

As well as examining any changes in the total number of words produced correctly
after therapy, the profile of error types was examined after each therapy phase.
Figure 2 shows the numbers of different error types on naming of treated and
untreated items for each participant as a proportion of total error responses, before
therapy, after phase 1 of therapy, and after phase 2 of therapy. A chi-square analysis
was used to identify any significant changes.
For SD, there was a significant change in error type on naming of treated items
after therapy, χ 2 (6) = 13.21, p = .040, but not for untreated items, χ 2 (6) = 6.22,
p = .399. Figure 2 shows that the proportion of phonologically related errors made
by SD increased for treated items after the second phase of therapy. For BB, there was
 PHONOLOGICAL ASSEMBLY THERAPY 449

Treated items Untreated items

Proportion of total errors

1 1
0.8 0.8
0.6 0.6
SD
0.4 0.4
0.2 0.2
0 0
Pre therapy 1 Post phase 1 Post phase 2 Pre therapy 1 Post phase 1 Post phase 2
therapy therapy therapy therapy
Proportion of total errors

1 1
0.8 0.8
Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

0.6 0.6
BB
0.4 0.4
0.2 0.2
0 0
Pre therapy 1 Post phase 1 Post phase 2 Pre therapy 1 Post phase 1 Post phase 2
therapy therapy therapy therapy
Proportion of total errors

1 1
0.8 0.8
0.6 0.6
HS
0.4 0.4
0.2 0.2
0 0
Pre therapy 1 Post phase 1 Post phase 2 Pre therapy 1 Post phase 1 Post phase 2
therapy therapy therapy therapy
Proportion of total errors

1
1
0.8
0.8
0.6 0.6
PL
0.4 0.4
0.2 0.2
0 0
Pre therapy 1 Post phase 1 Post phase 2 Pre therapy 1 Post phase 1 Post phase 2
therapy therapy therapy therapy

Phon related Unrelated Other No response

Figure 2. Changes in naming errors after therapy.

a significant change in error type on naming of treated items, χ 2 (6) = 20.68, p = .002,
and untreated items, χ 2 (6) = 15.05, p = .020. Figure 2 shows that, like SD, the propor-
tion of phonologically related errors made by BB increased for treated items after the
second phase of therapy, but in addition on naming of untreated items, the proportion
of unrelated errors decreased and the proportion of other errors increased after the
second phase of therapy. For HS, there was no significant difference in error type for
naming of treated items, χ 2 (6) = 2.58, p = .897, or untreated items, χ 2 (6) = 7.19,
p = .310, although Figure 2 shows a trend towards an increase in the proportion of
phonologically related errors and a decrease in the proportion of other errors for both
treated and untreated items, after both phases of therapy and particularly after the
 450 WALDRON

second therapy phase. For PL, there was a significant change in error type on naming
of treated items after therapy, χ 2 (6) = 16.17, p = .011, but not for untreated items,
χ 2 (6) = 7.16, p = .307. Figure 2 shows an increase in phonologically related errors
and a decrease in no responses on naming of treated items after the first phase of
therapy, with the proportions returning to pre-therapy levels after the second phase.
It is not possible to say whether this reflects an improvement in PL’s phonological
processing after the first therapy phase, or whether he was just more willing to try
and respond, and the proportion of phonologically related responses simply increased
because of this.

Changes in monitoring
Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

Given the focus of the second therapy phase on monitoring of speech errors,
self-correction behaviour was also examined for evidence of any change following
intervention.
Participants’ responses on spoken naming of the 130 item Nickels naming test pre-
therapy, and after therapy phases 1 and 2 were analysed, and the results for treated and
untreated items are shown in Table 6. The first row of data for each participant shows
the proportion of incorrect responses; that is, excluding items named correctly straight
away and items with no response, which were followed by at least one further attempt
at the target. This figure is taken as an indication of monitoring ability because it
shows that participants realised their first response was wrong and knew they needed
to try to correct it. It is likely to be an underestimate of participants’ awareness of
their errors, however, as there were several occasions when all four participants indi-
cated they knew they were wrong (e.g., saying “no”) but then did not produce another
response. Further evidence of monitoring ability comes from the finding that there
were no occasions when any participant named a picture correctly straight away but
then went on to have a further, incorrect attempt. The second row of data for each
participant in Table 6 gives the proportion of those items with more than one attempt
where the final response was correct, showing how successful each participant was at
self-correcting their errors.
It is interesting, first, to compare the patterns of monitoring behaviour for each
participant before therapy. All participants had one or more further attempts at the
target after producing an error response at least 50% of the time, suggesting that even
before therapy they were aware of more than half of their speech errors. In the cases
of SD, BB, and PL, however, few of these resulted in the correct target. This was
especially marked for PL who, before therapy, had more than one attempt at over 70%
of his errors but only one of these ended with the correct response. HS made more self-
corrections pre-therapy than the other participants, but he still successfully corrected
less than half of his repeated attempts.
A Jonckheere trend test was performed on the data shown in Table 6, to identify
whether there was any significant change in any participant’s monitoring behaviour
after either therapy phase. Across the three occasions of testing, SD and BB showed
no significant change in any of the monitoring behaviours, for treated or untreated
items. PL showed a significant increase in both the proportion of errors with more than
one attempt (p = .022) and in the proportion of these errors that were self-corrected
(p = .047) on naming of treated items after therapy. This improvement in monitoring
may account for the overall increase in naming scores reported earlier. HS showed
a significant increase in the proportion of errors with more than one attempt that
 Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

TABLE 6
Multiple attempts at naming of treated and untreated items

Naming of treated items Naming of untreated items

Pre-therapy1 Post Phase 1 Post Phase 2 Pre-therapy1 Post Phase 1 Post Phase 2

SD Incorrect responses with > 25/38 (.66) 13/32 (.41) 18/34 (.53) 22/43 (.51) 11/36 (.31) 12/33 (.36)
1 attempt
Items with > 1 attempt that 5/25 (.20) 3/13 (.23) 5/18 (.28) 5/22 (.23) 2/11 (.18) 3/12 (.25)
were self-corrected
BB > 1 attempt 24/45 (.53) 9/30 (.30) 18/33 (.55) 23/44 (.52) 23/46 (.50) 17/46 (.37)
self-corrected 6/24 (.25) 5/9 (.55) 8/18 (.44) 3/23 (.13) 6/23 (.26) 4/17 (.24)
HS > 1 attempt 34/38 (.89) 13/19 (.68) 9/11 (.82) 19/38 (.50) 16/30 (.53) 13/25 (.52)
self-corrected 16/34 (.47) 7/13 (.54) 8/9 (.89)∗ 7/19 (.37) 6/16 (.37) 5/13 (.38)
PL > 1 attempt 22/30 (.73) 39/46 (.85) 30/32 (.94)∗ 25/32 (.78) 31/38 (.82) 33/39 (.85)
self-corrected 0/22 (.0) 2/39 (.05) 4/30 (.13)∗ 1/25 (.04) 1/31 (.03) 1/33 (.03)
∗ = statistically significant change across three test occasions (Jonckheere trend test).
PHONOLOGICAL ASSEMBLY THERAPY
451
 452 WALDRON

were self-corrected on naming of treated items (p = .036) after therapy, but it is not
possible to ascribe the change to one of the two treatments in particular as it was
a significant trend that started after the first therapy phase and continued after the
second.

DISCUSSION
The aims of this study were to investigate whether the gains described by Franklin
et al. (2002) for a person with phonological assembly difficulties could be replicated
in a case series of four participants, and to look for any differences in the outcomes
for each participant, with a view to exploring which factors might identify the people
Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

who will benefit most from this therapy.

None of the four participants in this study responded to the therapy programme
in the same way as MB, Franklin et al.’s original client. MB improved on naming
of treated and untreated items after the first and second phases of therapy. This
was explained as a generalised improvement in phoneme activation. In the current
study, SD and BB improved on naming of treated items after the first, auditory dis-
crimination, phase of therapy, but made no further improvements after the second,
monitoring, phase of therapy, with no improvement on untreated items, and the gains
seen were more limited than for MB. HS improved on naming after both the first and
the second phase of therapy but only for treated items. PL did not show any significant
improvement on naming of treated or untreated items after either phase of therapy,
although he did show a significant improvement in naming over the course of the
whole study. Possible reasons for the patterns of performance seen and the differences
between participants will now be explored.
MB, Franklin et al.’s (2002) original client, had a relatively pure post-lexical phono-
logical assembly deficit. She was impaired in all modalities of spoken output and
showed an effect of phoneme length on all speech production tasks, with no effects
of word frequency or imageability, made few semantic errors and many phonemic
errors on naming, and was good at self-correcting her errors. In contrast, SD, BB,
and HS had a combination of impairments in both phonological assembly and lex-
ical retrieval, which may explain why they responded in a different way from MB
to the treatment described. Howard (2000) proposed that most treatments for word
retrieval, whether the tasks are semantic or phonological, work through activating
both the semantic representation and the output lexical phonology of the target word.
Furthermore, Howard suggested that treatments that work in this way are likely to
produce item-specific effects, because the mappings from word meaning to word form
are arbitrary. True generalisation to untreated items, he argued, can only be expected
when a strategy is taught, or when the target of therapy is a post-lexical process. The
finding, in the current study, that the improvements in spoken naming made by SD,
BB, and HS were for treated items only, suggests that both phases of therapy worked
by improving the mapping between semantics and lexical phonology. Although ther-
apy had been aimed at self-monitoring, the nature of the tasks meant that lexical
processing was an integral part; when participants had repeated opportunities to hear
the target word in the first auditory discrimination therapy phase, and to see the
picture and name the target word in the second monitoring therapy phase, the tar-
get word’s semantic and lexical representation would have been activated. Therefore,
although the first therapy phase did not entail seeing pictures or producing words
 PHONOLOGICAL ASSEMBLY THERAPY 453

aloud, the tasks still involved lexical and semantic access, simply through exposure to
real-word stimuli.
An improvement in the link between semantics and lexical phonology could also
explain the changes in speech errors and self-correction seen after therapy. SD and
BB showed a significant increase in phonologically related errors on naming of treated
items after therapy, and HS showed a significant increase in successful self-corrections
on naming of treated items after therapy. These findings may be explained by par-
ticipants having a strengthened phonological lexical representation and therefore a
clearer idea of the target being aimed for. The monitoring element of the second
therapy phase does not seem to have been responsible for HS’s improvement in self-
correction as the change started in the first and continued into the second therapy
phase.
Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

HS was the only participant in the current study who improved in the number
of treated words named correctly after the second monitoring therapy phase. There
are many potential reasons for this. One possibility is that SD and BB had already
achieved their maximum potential for improved naming after the first therapy phase
as the same set of words was treated in both phases. Even though SD and BB’s naming
scores after the first therapy phase remained relatively low, this may have been the max-
imum they were able to achieve. Alternatively, the pre-therapy linguistic and cognitive
assessment data reveal many differences between HS and the other three participants.
He scored higher on the auditory processing assessments and on the Pyramids and
Palm Trees and Camden recognition memory tests. He also had the least severe speech
production difficulties, gaining the highest scores on all output tasks pre-therapy, and
made most self-corrections pre-therapy. HS was also the youngest participant, and
the longest time post stroke, although based on the small number of participants in
this study it is not possible to be conclusive about which of these factors, if any, are
responsible for the pattern of results.
Similarly, there are many possible reasons why PL showed such limited improve-
ment after therapy, but again we cannot be sure which, if any, were influential. PL
was the oldest participant, and due to having the most severe mobility problems fol-
lowing his stroke, he was wheelchair bound and lived in a residential care home,
which reduced his opportunities for social communication. PL also had a mild to
moderate hearing loss and had the lowest scores on all spoken output tests before
therapy. In addition, PL did not have any lexical retrieval difficulties; rather he had
phonological assembly difficulties combined with AOS. Therefore, if the effects of
therapy seen in HS, SD, and BB were due to improved activation between semantics
and lexical phonology, then this would not be expected to cause any improvements
for PL.
This study was an exact replication of the therapy reported by Franklin et al. (2002).
To reduce the number of potential confounding variables, the procedure was kept as
close as possible to the original, including the amount of therapy given to each par-
ticipant, i.e., 45 minutes twice a week, or 11/2 hours per week in total. This is typical
of what is offered by many NHS speech and language therapy services in the United
Kingdom and it was sufficient to achieve significant improvements in MB’s speech
production. However, a review by Boghal, Teasell, Foley, and Speachley (2003) rec-
ommended that more intense therapy input is most likely to be effective; therefore if
participants in this study had received more hours of therapy the changes might have
been greater—indeed it may be informative for future studies to consider using more
intensive input.
 454 WALDRON

Conclusions
This study has added to the evidence base of therapy studies for people with phonolog-
ical assembly difficulties by replicating a successful single-case study by Franklin et al.
(2002) with a case series of four participants. A relatively short period of auditory dis-
crimination therapy was successful in improving spoken naming in three out of four
participants, but only one participant made further gains following the monitoring
therapy, and all improvements seen were for treated items only. Many differences
between the participants have been identified, both in their pre-therapy pattern of
linguistic deficit and in their response to therapy, demonstrating that this group of
clients is far from homogeneous, and that the question of identifying which treat-
ment is effective for whom is complex, especially as it seems the same therapy can
Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

work at different levels for different individuals, and does not always work in the
way we might predict. The therapy described in this paper was originally devised by
Franklin et al. (2002) with the aim of teaching a self-monitoring strategy; but instead
MB, who had a relatively pure post-lexical phonological impairment, showed a gener-
alised improvement in phonological assembly. In contrast, the three participants in the
current study whose speech improved following therapy had a combination of lexical
and post-lexical phonological impairments and showed item-specific improvements in
the mapping between semantics and the phonological output lexicon. Further treat-
ment studies are clearly needed to continue to understand this challenging client
group.

Manuscript received 5 November 2009

Manuscript accepted 17 May 2010
First published online 30 September 2010

REFERENCES
Best, W., Grassly, J., Greenwood, A., Herbert, R., Howard, D., & Hickin, J. (2006). Generalisation
to untreated items in anomia therapy. Presented at the 12th International Aphasia Rehabilitation
Conference, Sheffield, UK.
Bishop, D. V. M. (1982). TROG Test for Reception of Grammar. University of Newcastle upon Tyne, UK.
Boghal, S. K., Teasell, R. W., Foley, N. C., & Speachley, M. R. (2003). Rehabilitation of aphasia: More is
better. Topics in Stroke Rehabilitation, 10, 66–76.
Butterworth, B. (1992). Disorders of phonological encoding. Cognition, 42, 261–286.
Dabul, B. (1979). Apraxia Battery for Adults. New York: C.C. Publications, Inc.
Dell, G. S., Schwartz, M. F., Martin, N., Saffran, E. M., & Gagnon, D. A. (1997). Lexical access in aphasic
and nonaphasic speakers. Psychological Review, 104(4), 801–838.
Fillingham, J. K., Sage, K., & Lambon Ralph, M. (2006). The treatment of anomia using errorless learning.
Neuropsychological Rehabilitation, 16(2), 129–154.
Franklin, S., Buerk, F., & Howard, D. (2002). Generalised improvement in speech production for a subject
with reproduction conduction aphasia. Aphasiology, 16(10/11), 1087–1114.
Goldrick, M., & Rapp, B. (2007). Lexical and post-lexical phonological representations in spoken
production. Cognition, 102, 219–260.
Grant, D. A. & Berg, E. A. (1993) Wisconsin Card Sorting Test. New York: Western Psychological Services.
Hickin, J., Best, W., Herbert, R., Howard, D., & Osborne, F. (2002). Phonological therapy for word-finding
difficulties: A re-evaluation. Aphasiology, 16(10/11), 981–999.
Howard, D. (2000). Cognitive neuropsychology and aphasia therapy: The case of word retrieval. In I.
Papathanasiou (Ed.), Acquired neurogenic communication disorders. A clinical perspective. Oxford, UK:
Blackwell.
Howard, D., & Franklin, S. (1990). Memory without rehearsal. In G. Vallar & T. Shallice (Eds.),
Neuropsychological impairments of short term memory. Cambridge, UK: Cambridge University Press.
 PHONOLOGICAL ASSEMBLY THERAPY 455

Howard, D., & Patterson, K. (1992) The Pyramids and Palm Trees Test. Windsor, UK: Thames Valley Test
Company.
Kay, J., Lesser, R., & Coltheart, M. (1992) PALPA. Psycholinguistic Assessments of Language Processing in
Aphasia. Hove, UK: Lawrence Erlbaum Associates Ltd.
Kohn, S. E. (1984). The nature of the phonological disorder in conduction aphasia. Brain and Language,
23, 97–115.
Levelt, W. J. M., Roelofs, A., & Meyer, A. S. (1999). A theory of lexical access in speech production.
Behavioral and Brain Sciences, 22, 1–75.
Marshall, J. (2006). Jargon aphasia: What have we learned? Aphasiology, 20(5), 387–410.
McNeil, M. R., Doyle, P. J., & Wambaugh, J. (2000). Apraxia of speech: A treatable disorder of motor
planning and programming, In S. E. Nadeau, L. J. Gonzalez Rothi, & B. Crosson (Eds.), Aphasia and
language: Theory to practice (pp. 221–265). New York: Guilford Press.
McNeil, M. R., Pratt, S. R., & Fossett, T. R. D. (2004). The differential diagnosis of apraxia of speech. In
B. Maassen, R. D. Kent, H. F. M. Peters, P. H. H. M. van Lieshout, & W. Hulstijn (Eds.), Speech motor
Downloaded by [Curtin University of Technology] at 19:21 23 October 2011

control in normal and disordered speech (pp. 389–413). Oxford, UK: Oxford University Press.
Miceli, G., Amitrano, A., Capasso, R., & Caramazza, A. (1996). The remediation of anomia resulting from
output lexical damage: analysis of two cases. Brain and Language, 52, 150–174.
Miller, D., & Ellis, A.W. (1987). Speech and writing errors in “neologistic jargonaphasia”: A lexical activa-
tion hypothesis. In M. Coltheart, R. Job, & G. Sartori (Eds), The cognitive neuropsychology of language
(pp. 253–271). London. Lawrence Erlbaum Associates Ltd.
Nickels, L. (1992) Spoken word production and its breakdown in aphasia. Unpublished PhD thesis, University
of London, UK.
Nickels, L. (2002). Therapy for naming disorders: Revisiting, revising, and reviewing. Aphasiology,
16(10/11), 935–979.
Nickels, L., & Howard, D. (1994). A frequent occurrence? Factors affecting the production of semantic
errors in aphasic naming. Cognitive Neuropsychology, 11(3), 289–320.
Nickels, L., & Howard, D. (1995). Phonological errors in aphasic naming: Comprehension, monitoring and
lexicality. Cortex, 31, 209–237.
Nickels, L., Howard, D., & Best, W. (1997). Fractioning the articulatory loop: Dissociations and associa-
tions in phonological recoding in aphasia. Brain and Language, 56, 161–182.
Nickels, L., & Cole-Virtue, J. (2004). Reading tasks from PALPA: How do controls perform on visual lexical
decision, homophony, rhyme and synonym judgements? Aphasiology, 18(2), 103–126.
Raymer, A. M., Thompson, C. K., Jacobs, B., & Le Grand, H. R. (1993). Phonological treatment of naming
deficits in aphasia: Model-based generalization analysis. Aphasiology, 7(1), 27–53.
Robertson, S. J. (1982). Dysarthria profile. Tuscon, AZ: Communication Skill Builders.
Schwartz, M. F., Wilshire, C. E., Gagnon, D. A., & Polansky, M. (2004). Origins of nonword phonological
errors in aphasic picture naming. Cognitive Neuropsychology, 21(2/3/4), 159–186.
Warrington, E. K. (1996). The Camden Memory Tests. Hove, UK: Psychology Press
Whitworth, A., Webster, J., & Howard, D. (2005). A cognitive neuropsychological approach to assessment
and intervention in aphasia. Hove, UK: Psychology Press.
Ziegler, W. (2002). Psycholinguistic and motor theories of apraxia of speech. Seminars in Speech and
Language, 23(4), 231–243.