Vowel targeted intervention for children with persisting speech difficulties: Impact on intelligibility

Abstract

Compared to the treatment of consonant segments, the treatment of vowels is infrequently described in the literature on children’s speech difficulties. Vowel difficulties occur less frequently than those with consonants but may have significant impact on intelligibility. In order to evaluate the effectiveness of vowel targeted intervention (VTI) with two 10-year-old children with severe and persisting speech difficulties measures of (a) percentage vowels correct and (b) intelligibility outcomes by peer group listeners were used. Assessment of vowel production was used to design and carry out intervention for each child, the success of which was measured in two ways: comparing (a) percentage of vowels correct before and after the intervention, (b) the percentage of pre- vs. post-intervention utterances understood by a group of typical peer listeners (aged 9 to 11 years). Pre- and post-intervention speech samples (comprising single words, imitated sentences and spontaneous speech) were edited onto a CD for these listeners, who were asked to write down what had been said. The two children with speech difficulties made significant improvement in vowel production as measured by the percentage of vowels correct. The listeners perceived more productions accurately post-intervention than pre-intervention. There was also a reduction in the range of the listeners’ misperceptions of target words. VTI was effective in terms of both increasing PVC and intelligibility outcomes as judged by peer group listeners. It is not more complicated to carry out VTI than consonant targeted intervention; this should be considered more often when planning therapy for children where vowels are affected.

Keywords

vowels intervention persisting speech difficulties intelligibility peer-group listeners treatment outcomes

I Introduction

The study of vowels in children’s speech has received much less attention than consonants (Ball and Gibbon, 2002; Penney et al., 1994; Pollock, 1994; Stoel-Gammon and Herrington, 1990), so much so that Davis and MacNeilage (1990) state that ‘vowels are the poor relations of child phonology’ (p. 16). However, although studies of vowel acquisition are limited in number (Donegan, 2002), there is a consensus that vowel acquisition is fairly well established by three years of age in typically developing children, albeit subject to individual variation. Based on a review of the literature, Stoel-Gammon and Herrington (1990) describe a ‘general order of acquisition’ (p. 148) as follows:

first, the corner vowels /i,ɑ,u/, mid back /o/and central stressed /ʌ/;

second, vowels of ‘intermediate difficulty /æ,ʊ,ɔ,ə/;

third, a ‘late’ group /e,ε,ɪ/.

There have been even fewer studies of vowel production in children with atypical speech development (Davis and MacNeilage, 1990; Gibbon and Beck, 2002; Reynolds, 1990). Studies that have examined the vowel errors of children with atypical speech suggest that vowel errors reflect processes seen in typically developing younger children (Gibbon et al., 1992; Stoel-Gammon and Herrington, 1990), notably lowering, backing and diphthong reduction (Pollock and Hall, 1991). In contrast, Reynolds (1990) identified fronting rather than backing as well as both context-sensitive and idiosyncratic processes. Developmentally immature patterns were also identified by Pollock and Keiser (1990) in their assessment of a group of 15 children with speech difficulties aged between 3;8 and 6;4. The children were rated on a standardized (consonant based) test as having moderate, severe or profound phonological impairments. One of the children was classified as having severely disordered vowel production, and seven others as having mild to moderate difficulties with vowel production. However, only the child with severe difficulties had been identified prior to the study as having vowel errors, suggesting that clinicians may not always attend to difficulties with vowels.

As a result of this possible under-reporting, the percentage of children with speech disorders who make errors in vowel production is not known for sure, but it is likely to be only a small proportion of those children who present with consonant errors. Where vowel errors do exist, however, it may be indicative of more severe speech difficulties, as was found by Pollock and Hall (1991) in their examination of the vowel production of five children with ‘developmental apraxia’. This finding was replicated in the Memphis Vowel Project, which reported on the incidence of vowel problems in a larger group of children (Pollock and Berni, 2003). Two groups of children (aged 18 to 83 months) were assessed in order to measure their ‘percentage of vowels correct’ (PVC). One group (165 children) had normal phonology (NP), and the other (149 children) had disordered phonology (DP). In the NP group the incidence of vowel errors after 36 months of age was ‘minimal’ (0%–4%), but in the DP group 11%–32% of children had vowel errors (depending on whether the standard score cut off was taken as < 95, < 90, or < 85). The exploration of the relationship between the number of vowel and consonant errors confirmed that the occurrence of vowel production errors may be linked with more severe and complex speech disorders including Childhood Apraxia of Speech (CAS).

1 Vowel-targeted intervention

Given the small number of vowel studies in the literature, it is not surprising that there are relatively few accounts of vowel-targeted intervention (VTI). Furthermore, clinicians have been reticent about tackling vowel compared to consonant difficulties, even though ‘there should be no reason to suppose that the principles underlying therapy designed to improve the vowel system should be qualitatively different from those adopted for developing the consonant system’ (Gibbon et al., 1992: 35).

The handful of studies that include VTI generally report positive outcomes (Gibbon et al., 1992; Hargrove et al., 1989; Pollock, 1994) but there have been very few examinations of the specific link between vowel errors and intelligibility. Studies that have done this largely focus on intelligibility of adults with acquired dysarthria (Kim et al., 2011; Liu et al., 2005; Turner et al., 1995;), rather than children with developmental speech difficulties. However, Fletcher et al. (1991) cite vowel production difficulties as a major reason for poor intelligibility in children with profound hearing impairment, and Higgins and Hodge (2002) found a positive correlation between vowels and intelligibility in children with and without dysarthria. This finding was replicated in children with speech disorder by Reese and O’Hanlon (2004).

2 Intelligibility

Intelligibility can be defined simply as ‘the degree to which the child’s speech is understood by the listener’ (Gordon-Brannan, 1994: 17). However, when investigating intelligibility three variables need to be taken into account (Pascoe et al., 2006):

the speech produced by the speaker;

transmission in the environment; and

perception by the listener.

This is potentially even more complex if the participant speakers are children with developmental speech difficulties when the interaction is affected by the presence of immature or atypical segmental and prosodic factors.

Children with typically developing speech are intelligible most of the time to even unfamiliar listeners by the age of four (Hodson and Paden, 1981). This is not always the case for children who have developmental speech difficulties where intelligibility may be of significant concern (Gordon-Brannan and Hodson, 2000). Vowel errors and distortions (where the vowel realization is not typical, perhaps due to length or minor articulatory differences, but still recognizable and contrastive) contribute to this unintelligibility (Pollock and Hall, 1991), particularly as prosodic information, such as intonation, stress and timing, is carried on the vowel as the nucleus of the syllable.

Measuring intelligibility can involve formal and informal assessments (see review by Pascoe et al., 2006). Formal assessments of intelligibility, as distinct from speech sound production assessments, are not typically used when assessing children’s developmental speech difficulties unless these speech difficulties are associated with a specific clinical diagnosis such as cleft palate, cerebral palsy, or hearing loss. Even when used, intelligibility ratings may be ‘impressionistic estimates of intelligibility’ (Gordon-Brannan and Hodson, 2000: 142), which may not be reliable or robust.

In addition to these rating scales, research studies have also used ‘write down’ procedures (e.g. Gordon-Brannan and Hodson, 2000; Kent et al., 1994; Whitehill, 2002). In these, listeners may be given ‘open set tasks’ where they are asked to make judgements about what has been said and to write this down (e.g. Peng et al., 2004). Alternatively they may be given multiple choice or ‘closed set tasks’, where they choose their response from a number of possible targets (e.g. Gordon-Brannan and Hodson, 2000). The stimuli used in such intelligibility tasks range from imitation of single words and sentences to recording of spontaneous speech samples. There seems little doubt that spontaneous utterances, recorded during play and conversation, provide the best opportunities of capturing rich samples of children’s speech in order to ‘classify’, ‘describe’ and ‘explain’ unintelligibility (Kwiatkowski and Shriberg, 1992: 1097). However, in order to quantify listeners’ understanding it is essential to know what the target is, thus picture naming or description, or word and sentence imitation tasks have been used more often.

Typically, studies measuring intelligibility have recruited adult listeners to judge recordings of speech productions. These adults may be ‘naive’ listeners or ‘experienced’ such as speech and language therapists. Findings about how important this variable is when, for example, listening to adults with dysarthric speech have not been clear cut. Liss et al. (2002) found that more experienced listeners tended to give higher intelligibility ratings. In contrast, Ellis and Fucci (1992) found no difference in ratings made by speech and language therapists compared to more naive listeners. A more robust finding in the literature is from studies of listeners rating developmental speech difficulties. In these, familiarity with the speakers was an important factor (Kwiatkowski and Shriberg, 1992). In particular, mothers of children with ‘phonological difficulties’ have been found to understand significantly more of their children’s utterances than have fathers or unfamiliar adults (Flipsen, 1995).

A study by Starr et al. (1984) involved children as listeners. This examined the ratings of nasality and articulation made by clinicians, parents and children ‘aged 8 or older’ with typical or hypernasal speech and concluded that children with typical speech rated articulation in the same way as the adults. However, these children were not rating intelligibility per se so it is still not known how a peer group of child listeners deal with unintelligible speech.

Given the paucity of work on (a) targeted intervention for children with vowel difficulties, and (b) what impact these have on intelligibility as judged by peers, the present study aimed to evaluate whether VTI is effective for children who have vowel difficulties as part of severe and persisting speech difficulties. It addresses the following questions:

Can VTI improve children’s production of vowels?

Does vowel production impact on children’s intelligibility?

II Method and design

To address these research questions, the study was carried out in two parts (one part for each question). Part 1 is two intervention case studies using a pre-test/post-test design. This part of the study was further divided as follows:

An assessment of the vowel productions of two 10-year-old children with severe and persisting speech difficulties. Percentage of vowels correct (PVC) was taken as the baseline for comparison of outcome post-intervention. Percentage of consonants correct (PCC) was chosen as a control measure as these were not being targeted in the intervention. The hypothesis was that if VTI was effective then PVC would increase post-intervention but PCC would not, particularly given the severity of the children’s persisting speech difficulties and previously slow rate of spontaneous change. (PVC and PCC calculations were based on 76 single words designed to assess the number of accurately realized Southern British English vowels and consonants.)

VTI phase

Reassessment of PVC and PCC for comparison with the baseline measures.

Part 2 is an investigation into the impact of vowel difficulties on intelligibility. Recordings of pre- and post-intervention speech productions from Part 1 of the study were randomized and played to a group of typically developing children who were asked to write down what had been said. The hypothesis was that if VTI had been effective in improving intelligibility then the typically developing children would understand more post- than pre-intervention utterances.

Ethics approval for both parts of this study was granted by the Ethics Committee of the University of Sheffield and relevant Local Ethics Committee.

III Part 1: Vowel targeted intervention

1 Participants

The two children with persisting speech difficulties attended a special educational provision in a speech and language unit attached to a mainstream primary school, where they received intensive speech and language therapy. In this unit children’s speech processing strengths and weaknesses are assessed routinely through activities designed to tap different levels of speech input, representation, and output (after Stackhouse and Wells, 1997; 2001). This confirmed that both participants had severe and pervasive speech processing difficulties. Their speech intervention programme prior to the study had focused mainly on consonant production and promoting literacy skills. Vowel work following the approach developed by Reid (2003) had been introduced into their programmes in the year before the study but with the aim of promoting literacy development rather than speech production; this provides children with a metacognitive framework for vowel recognition and the links between spoken and written symbols. Their ability to discriminate and manipulate vowels in a variety of input tasks had improved, but there had been no carryover to speech in spite of vowel production being part of that intervention.

Both the children and their parents were monolingual and spoke with a typical south-eastern English accent (which is non-rhotic; for details see, for example, Wells, 1986); this accent was the same as that of their peer group. Their fictitious names used in this study were chosen by the children themselves: Ryan and Demi. Ryan (a boy, aged 10;7) had a performance IQ of 125 and a verbal IQ of 93; Demi (a girl, aged 10;7) had a performance IQ of 85 and a verbal IQ of 76. Given these cognitive differences the rate of change in response to intervention might be different; Prezas and Hodson (2010) suggest that twice as much time is needed for children who have delayed cognition. Both children had normal hearing.

2 Materials and procedure

A picture-based assessment was designed to elicit each vowel sound in different phonological contexts. The initial approach was informed by that taken by Pollock and Keiser (1990), but to allow repeated testing of each vowel and reduce the length of the test only single syllable words were included. The monosyllabic words used by Pollock and Keiser were vowel-initial (VC), open syllable (CV) and closed (CVC) (p. 165). With this in mind, the CVC words in this assessment were designed where possible (i.e. where a familiar word would result). Each vowel sound was followed in turn by a plosive, fricative, nasal or continuant and preceded by one of a variety of initial consonants. Co-articulation means that consonants affect the quality of adjacent vowels, but in the absence of conclusive findings about the direction of this relationship (i.e. whether vowels that follow or precede consonants are more vulnerable to co articulation), and given the need to manage the length of the world list. Again, to reduce the number of items tested, control in this way of word-initial consonants was not attempted. The contexts assessed were:

CV words with initial plosive, fricative and nasal, e.g. KEY, SHOE, NEAR (12 words);

VC words with final plosive, fricative and nasal, e.g. EAT, ICE, ARM (12 words);

CVC words, e.g. /i/ in SHEEP, TEETH, TEAM and WHEEL; /aɪ/ in BIKE, KNIFE, WINE and TILE (52 words).

In addition, eight imitated sentences, designed to include words with a range of vowels (e.g. charlie left his kite out in the rain), and conversation were recorded. All speech samples were collected on a Sony MiniDisc recorder MZ-NH700 using a Sony external microphone ECM-MS907.

The single word assessment responses were transcribed live and checked against the audio recordings of the children’s speech both before and after the intervention programme. Because the target vowel was always known in the single words, the PVC and PCC scores were calculated on this condition only both pre- and post-intervention.

3 Pre-intervention assessment results

Pre-intervention, Ryan’s PVC was 79.7% and his PCC was 83.8%; Demi’s PVC was 55.6% and her PCC was 78%. Their use of vowels and types of errors differed and they were more similar on PCC than PVC scores. Overall Ryan had more correct vowels than Demi; this difference was mainly in VC and CVC words. In CV words both children scored the same on PVC. Demi’s very low PCC score in CV words is largely accounted for by pre-vocalic voicing errors on 6 of the 12 items. PCC for both children would have been lower if more complex word structures (for example, consonant clusters) had been included.

Both children produced all vowels in words on at least one occasion during the assessment except /ɜ/ as in BIRD /bɜd/, which was not stimulable for either of them in words or in isolation.

a Ryan:

Although Ryan had less overall difficulty with vowels than Demi, his production was variable (especially with diphthongs in CVC contexts) and his speech had an effortful quality. He made many minor vowel errors (distortions) where the vowel was slightly backed or lowered, but not enough to lose contrast. Ryan used diphthongs correctly in open syllable and VC words (see Table 1).

Table 1

Pre-Intervention: Summary of Ryan’s vowel errors (CVC words)

Error	Feature change	Example of error	Example of correct production
/ɜ/→[ɑ]	Backing and lowering	bird→[bɑd]	Never correct
/ʊ/→[ɒ]	Backing and lowering	look→[lɒk]	pull→[pʊl]
/ε/→[æ]	Lowering	ten→[tæŋ]	mess→[mεs]
/u/→[əʊ]	Laxing and diphthongisation	boot→[bəʊt]	choose→[tjuz]
/aɪ/→[ɑ] or [æ:]	Diphthong reduction and lowering or frontingh	bike→[bɑk] knife→[ŋæ:f]	wine→[waɪŋ]
/eɪ/→[æ:]	Diphthong reduction and lowering	cake→[kæ:k]	case→[keɪs]
/əʊ/→[ɔ]	Diphthong reduction and backing	phone→[fɔŋ]	coat→[kəʊt]

Errors in Ryan’s connected speech were the same as in his single word production. For example:

Target sentence: the soup can go through holes

Ryan’s production: [və ˈsəʊp kəŋ ˈgɔ fwu ˈhɔlz]

Target sentence: five add three is eight

Ryan’s production: [ˈfɑv ˈæd ˈfwi iz ˈeɪt]

b Demi:

Demi’s feature change errors were similar to Ryan’s (e.g. backing, lowering and diphthong reduction) and she also showed variability with certain contrasts, e.g. /æ/ and /ɑ/; there was no evidence that this was context sensitive. She used diphthongs correctly in CV words but not in VC or CVC words. In production of VC words the word onset was often /h/; for example, OUT was realized as [hæt], ARM as [hɑm] and ICE as [hæs]. She also used /h/ mid-diphthong, particularly /ɔɪ/ and /aɪ/; for example, foil was realized as [ˈfɒhəl] and TILE as [ˈdɑhəl]. In this respect she is similar to the child described by Harris et al. (1999), who realized diphthongs as bisyllabics; for example, TIRE as [tɔjə] (p. 21). Demi’s vowel errors are summarized in Table 2.

Table 2

Pre-intervention: Summary of Demi’s vowel errors (CV, VC, CVC words)

Error	Feature change	Example of error	Example of correct production
/ɜ/→[ɑ] or [ε:]	Backing and lowering or fronting	purr→[bɑ] bird→[bε:d]	Never correct
V initial words onset [h]		out→ [hæt]	on→[ɒn]
/æ/→[ε]	Raising	add→[hεd]	Usually correct, occasionally /ε/→/æ/
/aɪ/→[ɑ]	Diphthong reduction and lowering	bike→[bɑk]	Only correct in CV words
/eɪ/→[ɑ]	Diphthong reduction and backing	rain→[rɑn]	cake→[geɪk]
/ɔɪ/→[ɔ]	Diphthong reduction	noise→[nɔz]	Only correct in CV words
/əʊ/→[ɑ] or [ɒ]	Diphthong reduction and lowering and backing	goat→[gɑt]coat→[gɒt]	Never correct
/ɑʊ/→[ɑ] or [ɒ] or [æ:]	Diphthong reduction and backing or fronting	loud→[læ:d]	Never correct

Demi’s connected speech also showed the same error types as her single word assessment. For example:

Target sentence: you might do the ironing with an iron

Demi’s production: [hu ˈmɑt du və ˈhɑhənɪŋ wɪv ə ˈhɑhən]

Target sentence: the fish is upsidedown

Demi’s production: [və ˈfɪs hɪs ˈhʌpsæˈdæ:n]

4 Intervention

The vowel targets in the intervention were based on the assessment findings and were therefore different for each child. Ryan’s target vowels were the diphthongs /aɪ/, /eɪ/ and /əʊ/ and the monophthongs /u/, /ʊ/ and /ɜ/. Demi’s were the diphthongs /aɪ/ /eɪ/ and /ɔɪ/ and the monophthong /ɜ/. Therapy for both children combined speech input and output work. Support materials used in all activities were:

A set of pictures of high frequency words containing the vowel targets, e.g. five, game.

Written words plus vowel symbols (after Reid, 2003), e.g. the diphthong /aɪ/ was characterized as /ɑ/ + /i/, with the characters ‘Professor Ah’ (who ‘says’ ah) and ‘Squeak the mouse’ (who ‘says’ ee) making up the two elements of the vowel. Diphthongs were described as ‘vowels that travel’.

The VTI took place over a 6-month period which included a 6-week break for the summer holidays. Intervention sessions were individual and averaged 30 minutes, 3 times a week; Demi had 40 sessions in all and Ryan had 35.

Tasks were introduced in the following order to target auditory discrimination, speech production and metaphonological skills:

minimal pairs for discrimination and confrontation naming, e.g. cart/kite, mark/make; these were based on the children’s realization of the target vowel and the target vowel itself;

rhyme activities (detection and production);

sorting words by vowel sounds, following an adult model;

silent sorting of words by vowel sound;

CVC blending and segmentation (words and non-words);

articulatory drills (CV, VC, CVC);

word/phrase/sentence level production.

Both children attended well to the intervention activities, and it became clear early on that their auditory discrimination of vowels in the range of tasks presented was more accurate than their production. Intervention therefore increasingly targeted their output skills.

5 Post-intervention results

Post-intervention the children’s speech was reassessed to examine any changes in vowel and consonant production; PVC and PCC pre- and post-intervention were compared using the Wilcoxon Signed Ranks test.

a Ryan:

Ryan’s accuracy in the production of vowels post-intervention increased. CV productions were 91.6% correct compared to 83.3% pre-intervention; VCs were 100% correct compared to 91.6%, and CVC were significantly better: 92.7% correct compared 76.3% (p < .003). All targeted vowels showed positive changes (see Table 3). In contrast, consonant productions did not change; PCC 84.5% post-intervention, compared to 83.8% pre-intervention.

Table 3

Ryan’s vowel production pre- and post-intervention: PVC, and PCC in different syllable types

	CV	CV	VC	VC	CVC	CVC
	Pre-	Post-	Pre-	Post-	Pre-	Post-
PVC %	83.3	91.6	91.6	100.0	76.3	92.7^†
PCC %	84.6	84.6	83.3	80.0	92.0	94.0
/i ɑ ɔ ɪ æ ɒ ʌ/	All correct	All correct	All correct	All correct	All correct	All correct
/ɔɪ εə ɑʊ ɪə/	All correct	All correct	All correct	All correct	All correct	All correct
/u/	Correct	Correct	N/A	N/A	1 correct2 [əʊ]	2 correct1 [əʊ]
/ɜ/	[ɑ]	[ɑ]	N/A	N/A	3 [ɑ]1 [ε]	1 correct2 [ɑ]1 [ɑʊ]
/ε/	N/A	N/A	[æ]	All correct	3 correct1 [æ]	All correct
/ʊ/	N/A	N/A	N/A	N/A	1 correct2 [ɒ]	All correct
/aɪ/	Correct	Correct	Correct	Correct	2 correct1 [ɑ]1 [æ]	All correct
/eɪ/	Correct	Correct	Correct	Correct	3 correct1 [æ]	All correct
/əʊ/	[ɔ]	Correct	N/A	N/A	3 correct1 [ɔ]	All correct

Note: ^† p < .003, Wilcoxon Signed Ranks test

b Demi:

Demi’s production of vowels post-intervention showed significant change on VC structures (25% pre-intervention to 100% correct post-intervention: p < 0.003), and on CVCs, which changed from 57.6% to 93.2% correct (p < 0.001). However, CV productions remained the same pre- and post-intervention: 83.3%; this was because the long vowels and diphthongs in CV syllables were already largely accurate pre-intervention. In contrast to Ryan, Demi’s PCC score increased, from a mean of 78% pre-intervention to 87.6% post-intervention. This difference was significant for CVC words only (p < .002), which changed from 80% pre-intervention to 90.8% correct post-intervention. This improvement was in the production of word-initial voiceless plosives, which had previously been voiced. All targeted vowels showed improvement (see Table 4) and, although /ɑʊ/ and /əʊ/ had not been targeted, Demi used them correctly post-intervention, suggesting some generalization in the production of diphthongs.

Table 4

Demi’s vowel production pre- and post-intervention PVC, and PCC in different syllable types

	CV	CV	VC	VC	CVC	CVC
	Pre-	Post-	Pre-	Post-	Pre-	Post-
PVC %	83.3	83.3	25.0	100.0^†	57.6	93.2**
PCC %	46.1	53.8	100.0	100.0	80.0	90.8*
/ɔ u ʊ æ ɒ ʌ/	All correct	All correct	All correct	All correct	All correct	All correct
/εə ɪə/	All correct	All correct	All correct	All correct*	All correct	All correct
/ɪ/	N/A	N/A	[hɪ]	Correct	All correct	All correct
/i/	Correct	Correct	[hi]	Correct	All correct	All correct
/ɑ/	Correct	Correct	[hɑ]	Correct	All correct	All correct
/ɜ/	[ε:]	Correct	N/A	N/A	3 [ε:]1 [æ]	3 correct1 [ε:]
/ε/	N/A	N/A	[hε]	Correct	All correct	All correct
/aɪ/	Correct	Correct	[hæ]	Correct	2 [ɑ]2 [æ]	All correct
/eɪ/	Correct	Correct	[hæ]	Correct	3 correct1 [æ]	All correct
/əʊ/	Correct	Correct	N/A	N/A	3 [ɒ]1 [ɑ]	All correct
/ɔɪ/	Correct	Correct	N/A	N/A	2 [ɔ]1 [ɑhəl]	All correct
/ɑʊ/	[æl]	[ɑʊl]	[hæ]	Correct	4 [æ:]	All correct

Notes: ^† p < .003, Wilcoxon Signed Ranks Test; * p < .002, ** p < .001

Both children struggled with the production of /ɜ/ which had not been stimulable for either child at the start of the study. Even 6 months post-intervention they were not producing this vowel accurately.

6 Discussion

The main aim of Part 1 of this study was to establish if VTI could facilitate more accurate production of vowels in two 10-year-old children with persisting and severe speech difficulties. The baseline assessment established which vowels were problematic for each child.

Examination of Ryan’s and Demi’s vowel errors revealed vowel production errors found in the speech productions of younger typically developing children; for example, backing and lowering, and diphthong reduction (Pollock and Hall, 1991). However, there were also atypical features; for example, Ryan’s diphthongization of the vowel /u/, and Demi’s use of /h/ initially in vowel onset words. Since studies describing vowel disorders are small in number – particularly of children as old as Ryan and Demi – it is difficult to say how idiosyncratic these atypical errors are as Reynolds (1990) also points out. In the study by Pollock and Hall (1991), the most frequent errors made by the five children (age range 8;2 to 10;9) with CAS involved backing, lowering and diphthong reduction. Like Ryan and Demi, two of the five children showed greater accuracy in diphthong production in CV words. Four of the five children did not use /ɝ/ at all (the rhotic equivalent of /ɜ/).

Post-intervention, both children had made more significant progress in vowel production accuracy in single words compared to their non-targeted consonant production. During the intervention, consideration was given to vowels at both syllabic and segmental level to facilitate planning and production in a variety of CV, VC and then CVC contexts. The interaction between vowels and adjacent consonants (both pre- and post-vocalically) may facilitate or hinder production (Bates et al., 2002). For example, there is evidence from clinical practice that production of velar plosives is facilitated by combining /k/ or /ɡ/ with back vowels such as /ɑ/ or /ɔ/ rather than front vowels like /æ/ or /i/, which encourage ‘front’ plosives like /t/ and /d/.

Although, contrary to the initial prediction, Demi’s production of voiceless pre-vocalic plosives in CVC words improved during the period of the study, the percentage improvement (and corresponding significance) was much smaller than that for vowels (9.6% compared with 36.5%). This affected the production of only three consonants (/p, t, k/), whereas the range of vowels that improved was much greater. Given the links between segmental accuracy, severity and intelligibility, it could be argued that the more significant vowel changes would potentially have more impact on how much of Demi’s speech might be understood. The generalization to untreated diphthongs might be explained by an improvement in Demi’s ability to realize complex vowels per se, akin to the generalization that can occur in consonant cluster production. Demi’s response to the intervention was greater overall than Ryan’s; however, he responded well to VTI showing significant improvements in vowel production.

In spite of differences between the cognitive levels of the children (Ryan: performance IQ 125; verbal IQ 93; Demi: performance IQ 85; verbal IQ 76) they both made progress with similar amounts of input and activities (the small difference in number of sessions was because of an unplanned absence). This is contrary to what might have been predicted (by, for example, Prezas and Hodson, 2010) and may reflect the fact that Ryan and Demi were older than children in most other reported studies. Further investigation of the impact of intervention in older children who have cognitive delays is needed.

The programme included frequent opportunities for production practice and, importantly, most progress was made for vowels already established in the children’s phonetic inventories. For example, diphthongs were already present in CV words, whereas /ɜ/ was not stimulable at all at the start of the intervention and was still not produced accurately 6 months later. Although both children developed a vowel that was more like /ɜ/ the accuracy varied, even within the same activity. These findings replicate previous studies: for example, Pollock (1994) reports that the child ‘P’ at 4;9 was not stimulable for the vowel /e/, and discontinued targeting it in intervention after several weeks of trying. Thus, Reynolds (1990) may be right about focusing treatment on listening and recognition of small contrastive differences in sounds whilst looking for methods of facilitating the production of non-stimulable vowels. Visual cues may be of limited value when targeting central vowels (e.g. /ɜ/), or the vowels /æ/ and /e/ which are in close proximity; however, visual information, for example, discussion about mouth shape and jaw height may be of help.

The first part of this study demonstrated a positive change in the PVC following VTI. Part 2 investigates whether this change had any impact on the children’s intelligibility as perceived by a peer group of child-listeners.

IV Part 2: Intelligibility

1 Listener participants

The 19 listeners (8 boys and 11 girls) were aged between 9;1 and 11;2 (mean age 10;0). They attended a mainstream primary school which for ethical reasons was not the same school as the one attended by Ryan and Demi.

2 Materials

The stimuli for the listeners to judge were taken from the pre- and post-intervention vowel assessments in Part 1 and were played to them using a Sony Walkman D-EJ119, through a Yamaha Monitor speaker MS101 II. A response form was designed for the listeners to record their answers; numbers corresponded to heard stimuli; it included written encouragement such as ‘good listening’.

The speech samples were:

12 CVC single words derived from the naming test for each child, giving 2 different sets of words. Each child’s set of words were used pre- and post-intervention plus 3 items from each child, repeated to check for consistency, giving a total of 54 words in all. Words selected included vowel sounds targeted in intervention, which had improved post-intervention as measured by PVC (e.g. Ryan: KNIFE, LOOK; Demi: HIDE, EIGHT).

Four imitated sentences from each child; the same sentences were used pre- and post-intervention (16 sentences in all). The sentences were designed to include vowel sounds targeted in intervention, e.g. BEN LEFT HIS KITE OUT IN THE RAIN.

Four samples of conversational connected speech (CS) from each child, pre- and post-intervention (16 utterances in all). These CS samples were selected on the following criteria:

• It was known what the child was attempting to say, even if the production was not completely accurate.

• Practical considerations, e.g. the utterances that could be edited as cohesive chunks that were not too long for the listeners to remember and write down.

Given these selection constraints vowel content in CS could not be controlled.

The stimuli above (single word naming, imitated and spontaneous sentences) were randomized and edited onto a CD.

3 Procedure

The listeners were recruited through the school, and both they and their parents gave informed consent to participate. They were divided into three groups by their class teacher (to fit with timetable requirements), and each group was seen separately in a quiet room at their school. The task was completed in one session of 40 minutes, with a break in the middle. All listeners heard the same data in the same order and were given instructions to listen carefully even though sometimes it might be hard to understand, and to write down on the response sheet what they thought had been said. It was emphasized that this was not a spelling or writing test. The responses were scored as follows (based on correct identification of whole words):

Single words: The number and percentage written down correctly pre- and post-intervention.

Imitated sentences: The number and percentage of the main information carrying words written down correctly pre- and post-intervention. For example, in the target sentence ben left his kite out in the rain only words underlined were scored.

Spontaneous sentences: As with imitated sentences, the number and percentage of main information carrying words in spontaneous utterances written down correctly pre- and post-intervention.

In addition, single word responses were examined for:

the number of word vs. non-word responses;

the number of different responses for each target;

the phonetic closeness of error responses to their targets.

4 Results

Pre-intervention Ryan’s speech productions (mean of 32.25% across all three sample types) were generally more intelligible to the listeners than Demi’s (mean 8.12%). However, both children’s results showed improvement post-intervention when Ryan’s mean score had increased to 50.76% and Demi’s to 46.8%. There were differences between single words, imitated sentences and spontaneous speech for each child.

a Ryan:

Significantly more of Ryan’s single words and imitated sentences were understood post-intervention than pre-intervention but this was not the case for his spontaneous speech (see Table 5).

Table 5

Ryan number (and percentage and range) of words understood, pre- and post-intervention

	Mean	Standard deviation	Range	Significance (Wilcoxon Signed Ranks Test)
Single words (pre-intervention)	3.48 (28.47)	1.61	1.00–7.00(8.3–58.3)	–
Single words (post-intervention)	6.77 (56.60)	0.78	5.00–8.00(41.6–66.6)	Z = −3.781, p < .0001
Words in imitated sentences (pre-intervention)	4.15 (27.72)	0.47	0.00–9.00(0.00–60.00)	–
Words in imitated sentences (post-intervention)	8.57 (57.19)	4.00	1.00–12.00(6.67–93.33)	Z = −3.775, p < .001
Words in spontaneous speech (pre-intervention)	6.47 (46.21)	2.52	2.00–11.00(14–79)	–
Words in spontaneous speech (post-intervention)	5.39 (38.53)	3.73	0.00–11.00(0–79)	Z = −1.272, p < .203

b Ryan: single words:

Pre-intervention the listeners understood on average 28.47% (3.48 words out of 12) compared to 56.6% (6.77) post-intervention (Z = −3.781, p < .001).

Where the vowel was the only error the ‘frame’ provided by the accurate consonants was correctly written, and the vowel was the only segment altered; e.g. ‘pass’ [pɑs] for PURSE. However, when consonant errors occurred they usually preserved the manner of the target consonant but not necessarily the voicing or place of articulation; e.g. ‘bath’ for PURSE, ‘cat’ for CAKE. Although acoustic similarity might help intelligibility when in context, these decontextualized single words were no closer to being understood. These types of errors did not change qualitatively post-intervention, but their number decreased.

The number of different responses for each target also decreased post-intervention. For example, pre-intervention there were 6 different responses for the target KNIFE: Nine listeners perceived this as ‘love’, 2 as ‘bath’, 2 as ‘glove’, 1 as ‘lath’, 1 as ‘laugh’, and 1 as ‘raff’. Post-intervention, only 1 listener perceived KNIFE wrongly (as ‘nine’), all others made a correct judgement. In contrast, the ‘no response’ rate changed very little and was around 10% both pre- and post-intervention. However, this varied from listener to listener and some showed a much higher rate than others. There were very few non-word responses from the listeners; for example, ‘bine’ for BURN (and ‘lath’ and ‘raff’ quoted above may also fall into this category), but in the main all error responses were alternative real words.

Although the overall error rate dropped significantly post-intervention, some words were not understood even when the target vowel had been produced. For example, CAKE was identified incorrectly as ‘cat’ by 11 listeners pre-intervention but as ‘take’ by 11 listeners post-intervention. Other examples post-intervention include ‘suit’ for SOUP (pre-intervention typically ‘sock’, 8 responses or ‘soap’, 5 responses); and ‘nine’ for KNIFE (pre-intervention most frequent was ‘love’, 8 responses). Importantly, Ryan was judged by the authors to be realizing the consonant targets accurately. It seemed that although the listeners perceived /eɪ/ in CAKE and /u/ in SOUP this was not enough for the words to be consistently understood. This suggested that listeners’ perceptions and ability to match to a word were affected by subtle variations in phonetic realizations of consonants and vowels; i.e. for some listeners the match was ‘right’ but for others it was not.

c Ryan: imitated sentences:

Pre-intervention the listeners understood, on average, 27.72% (4.15 out of 15) words compared to 57.19% (8.57) post-intervention (Z = −3.775, p < .001). Post-intervention the listeners’ responses were generally more accurate; they understood more key words as reflected in the scores but also more of the whole utterance (see Table 6).

Table 6

Examples of listener responses to one of Ryan’s imitated sentences; Target: Tim FOUGHT to get the FIRST PLACE

Listener	Pre-intervention response	Post-intervention response
1	NR	<Fought to get to first place>
4	<Hold_______________>	<Fought to get first place>
5	<Fought to get the first>	<Fought to get the first place>
6	<____________class>	<Fought to get first place>
7	<His first class>	<Fought to get first place>
9	<Place>	<Thought to get his first place>
10	NR	NR
14	NR	<First place>
15	<Bought first place>	<Through to get first place>
16	<Port __________place>	<Thought he would get first place>
18	<Wants to go to a circus>	<Wanted the first place>
19	<To get a first class>	<Fought to get first place>

d Ryan: spontaneous speech:

In contrast to single words and imitated sentences, there was no significant difference between the listeners’ scores on spontaneous speech pre-intervention (mean 46.21%, 6.47) and post-intervention (38.53%, 5.39) (Z = −1.272, p=.203).

e Demi:

Demi’s intelligibility as perceived by the listeners significantly improved across all three conditions (see Table 7).

Table 7

Demi single words: number (and percentage and range) of words understood, pre- and post-intervention

	Mean	Standard deviation	Range	Significance (Wilcoxon Signed Ranks Test)
Single words (pre-intervention)	0.36 (3.06)	0.83	0.00–3.00(0.00–25.00)	–
Single words (post-intervention)	7.00 (58.00)	2.10	3.00–11.00(23.00–91.6)	Z = −3.843, p < .001
Words in imitated sentences (pre-intervention)	1.84 (12.00)	1.64	0.00–5.00(0.00–31.25)	–
Words in imitated sentences (post-intervention)	6.15 (39.00)	2.83	1.00–11.00(6.25–68.75)	Z = −3.838, p < .001
Words in spontaneous speech (pre-intervention)	1.76 (9.31)	1.59	0.00–5.50(0.00–37.00)	–
Words in spontaneous speech (post-intervention)	6.55 (43.68)	1.99	3.50–10.50(23.00–70.00)	Z = −3.737, p < .001

f Demi: single words:

Pre-intervention the listeners understood, on average, 3.06% (0.36 of the 12) words compared to 58% (7.00) post-intervention (Z = −3.843, p < 001). In some responses the vowel was the only error; for example the word BIKE was heard as ‘bark’ by 4 listeners, ‘book’ by 3, and ‘back’ by 1. As with Ryan, in responses where the vowel was correct but the consonants were misperceived, the manner of articulation was typically preserved by the listeners – e.g. ‘nice’ for KNIFE (4 listeners), ‘stuck’ for SPIKE (1 listener) – but perceived voice and place of articulation were variable. In contrast to Ryan, although there was a reduction of vowel based errors in Demi’s word productions post-intervention, the number of different responses recorded by the listeners remained wide.

The overall ‘no response’ rate changed from nearly 20% pre- intervention to around 10% post-intervention suggesting that the listeners were more able to guess the word although again there was individual variation within the listeners. As with Ryan, the listeners only gave very occasional non-word responses (e.g. ‘spurk’ for SPIKE, and ‘zarth’, ‘luft’ and ‘lurth’ for KNIFE).

g Demi: imitated sentences:

Pre-intervention the listeners understood, on average, 12% (1.84 of the 16) words in connected speech compared to 39% (6.15) post-intervention (Z = −3.838, p < 001). Pre-intervention four of the listeners had no response to any of Demi’s imitated sentences; post-intervention no listeners scored zero on the task and scores for individual listeners increased; one listener was able to understand 11 of the 12 words accurately, and two listeners understood 10.

h Demi: spontaneous speech:

Pre-intervention the listeners understood, on average, 9.31% (1.76 of the 15) words compared to 43.68% (6.55) post-intervention (Z = −3.837, p < 001). Post-intervention the listeners’ responses were generally more accurate; examples are given in Table 8.

Table 8

Examples of listener responses to Demi’s spontaneous speech

Pre-intervention target	Because she is moving to a new HOUSE
Post- intervention target	I went to see my new HOUSE
Listener	Response
1	<They have moving hats><I went to see my new house>
2	<I can hear his moving pets><I went to see my new house>
3	<We are moving house><I went to see my new house>
5	<___________ removing hats><I went to see my new house>
8	<Moving I didn’t understand><I went to see my new house>
10	<No response><I went to go and see my new house>
18	<Could I help move the pens><I went to see my new house>
19	<Figure moving hairs><I went to see my new house>

5 Discussion

Findings from Part 2 of the study suggest that improved vowel production following VTI did impact on intelligibility as judged by a peer group of typical listeners. The pre-intervention intelligibility baseline was different for the two children; Ryan’s speech productions were generally more intelligible to the listeners than Demi’s (means of 32.25 % and 8.12% respectively) but post-intervention this gap had narrowed (post-intervention Ryan’s mean score had increased to 50.76% and Demi’s to 46.8 %). Thus, in terms of percentage increase, Demi’s intelligibility improved more than Ryan’s but neither child could be considered to have speech that was easy to understand. For example, on imitated sentences, only 10 of the 19 listeners understood more than 60% of Ryan’s target words post-intervention, the threshold for considering a child intelligible (Gordon-Brannan, 1994). In contrast, only three listeners reached this threshold on Demi’s productions. The listeners’ responses showed that their perceptions varied considerably and several commented that the speech was hard to understand.

Contrary to predictions by Gordon-Brannan and Hodson (2000) and Pascoe (2004) that intelligibility will be judged better in connected speech than single words, Demi’s outcomes were poorer for imitated sentences than for single words both pre- and post-intervention; and there was no significant difference between the judgements made of Ryan’s intelligibility on these conditions. One explanation could be that although as Tye-Murray and Kirk (1993) state that ‘spontaneous speech samples should yield results that are most representative of the child’s productions’ (p. 490), children with speech difficulties may be more variable across the conditions. For example, the most representative ‘best speech’ may be produced on naming and imitation tasks rather than on their connected spontaneous utterances, which may be more typical of their intelligibility in the classroom or playground. There is also a suggestion that the severity of speech difficulties influences the relative success or otherwise in understanding connected speech as opposed to single words (Beukleman and Yorkston, 1979), and Demi, overall, had more difficulty than Ryan as assessed by PVC and PCC.

It may also be that different methods of scoring responses underlie contradictory findings in intelligibility studies. The ‘write-down’ method means that the targets must be known to the scorer in order to decide whether the listeners have correctly understood what was intended. In the present study this was compounded by the need to include examples that could be edited for use that were not too long for child-listeners to remember. Consequently, the vowel content could not be as tightly controlled as planned which may limit the interpretation of these data.

Although other methods of measuring intelligibility outcomes in connected speech could be considered, the value of adopting the ‘write down’ method is that it allows an examination of the range of responses the listeners give to pre- vs. post-intervention targets. The reduction in the range of different responses given to post-intervention items indicate that when the listeners were unsure of a target, they were at least more able to narrow down the field of possible responses which increases the likelihood of more accurate guessing, particularly if contextual cues are present as well. For example, pre-intervention there were no correct responses by the listeners and no response to Ryan’s production of [səʊp] for the target SOUP. Eight listeners heard this as ‘sock’, five as ‘soap’, three as ‘stop’, one as ‘shock’, and one as ‘socks’. This number of different but phonetically similar responses suggests that the listeners matched what they perceived to their own phonological representations of plausible targets. If, for example, the listener perceives the CV onset as /səʊ/ plus a plosive, SOAP is a high frequency possibility. However, if the CV onset is perceived as /sɒ/, SOP is not a typically stored word for this age group unlike the high frequency and similar sounding word SOCK. This would explain why there were very few non-word responses from the listeners overall, and it suggests that more than segmental factors alone need to be taken into account when examining intelligibility, such as the lexical selection and frequency of the listeners’ responses for a target word. This supports the view that ‘intelligibility is not an all-or-nothing phenomenon but is a matter of degree’ (Connolly, 1986: 372), and is influenced both by children’s speech output and by the listeners’ perception and own language experience (Pascoe et al., 2006). This may be particularly the case when the listeners are children rather than adults, and further studies that include children as listeners are needed to explore this interaction further.

V Conclusions and clinical implications

This study aimed to evaluate VTI and was divided into two parts. Part 1 established if VTI could improve vowel production in school-age children with severe and persisting speech disorders, and Part 2 investigated if any improvement had a positive impact on the intelligibility of their speech as judged by a group of typically developing peers. There were significant positive outcomes in both parts of the study. There were significant improvements in vowel production following VTI; comparison of intelligibility scores showed that listeners understood more following this improvement.

This study also demonstrates that vowels can be treated by taking a similar approach to when working on consonants. Clinicians’ anxiety about treating vowel disorders may therefore be unfounded, although facilitation of non-stimulable targets may be more challenging because of the difficulties in explaining and demonstrating vowel placement in comparison with consonants, which generally have more readily demonstrable articulatory placement. Principles when working on vowels or consonants are very similar; for example, production work is aided by discrimination and lexical work, and incorporating meta-skills are essential for progress (Pascoe et al., 2006). Although the literature review indicates that vowels are established by the age of three in typically developing children, VTI is rarely carried out with young children who have developmental speech difficulties with no obvious cause such as hearing loss or cleft palate. If, as this study indicates, successful treatment of vowels can contribute towards intelligibility, there may be a case for suggesting that VTI could be incorporated earlier in intervention programmes for children with significant speech difficulties who present with atypical vowels. However, the results from the two case studies reported here cannot be generalized to the diverse population of children with speech difficulties and further intervention studies focusing on vowels are needed. Both children studied have long-term and complex speech difficulties that would not be resolved by VTI alone. They were still in need of both consonant and vowel targeted intervention focused on connected speech as well as single words.

This study not only highlights the importance of incorporating VTI into children’s intervention programmes but also the importance of considering a range of listeners when working on children’s intelligibility. These listeners will vary in age, familiarity with the children’s speech and in their own language experience. The speech and language therapist who by definition is an ‘expert’ listener may not be the best judge of a child’s intelligibility overall. The views of others, including parents/carers, teachers/assistants and peers, need to be sought in order to establish a child’s functional intelligibility. It is acknowledged that the method adopted here of using groups of peer listeners to do this is not a realistic measure of intervention outcome for everyday practice. However, the finding that increased PVC, a straightforward and easily administered output measure, impacted on intelligibility, suggests that this could be used more routinely alongside PCC as a marker to evaluate progress in intelligibility and intervention.

References

Ball

Gibbon

(eds) (2002) Vowel disorders. Boston, MA: Butterworth-Heinemann.

Bates

SAR

Watson

JMM

Scobbie

(2002) Context-conditioned error patterns in disordered systems. In Ball

Gibbon

(eds) Vowel disorders. Boston, MA: Butterworth-Heinemann, 145–86.

Beukleman

Yorkston

(1979) The relationship between information transfer and speech intelligibility of dysarthric speakers. Journal of Communication Disorders 12: 189–96.

Connolly

(1986) Intelligibility: A linguistic view. The British Journal of Disorders of Communication 21: 371–76.

Davis

MacNeilage

(1990) Acquisition of correct vowel production: A quantitative case study. Journal of Speech and Hearing Research 33: 16–27.

Donegan

(2002) Normal vowel development. In: Ball

Gibbon

(eds) Vowel disorders. Boston, MA: Butterworth-Heinemann, 1–36.

Ellis

Fucci

(1992) Effects of listeners’ experience on two measures of intelligibility. Perceptual and Motor Skills 74: 1099–1104.

Fletcher

Dagenais

Critz-Crosby

(1991). Teaching vowels to profoundly hearing-impaired speakers using glossometry. Journal of Speech and Hearing Research 34: 943–56.

Flipsen

(1995) Speaker-listener familiarity: Parents as judges of delayed speech intelligibility. Journal of Communication Disorders 28: 3–19.

10.

Gibbon

Beck

(2002) Therapy for abnormal vowels in children with phonological impairment. In: Ball

Gibbon

(eds) Vowel disorders. Boston, MA: Butterworth-Heinemann, 217–48.

11.

Gibbon

Shockey

Reid

(1992) Description and treatment of abnormal vowels in a phonologically disordered child. Child Language Teaching and Therapy 8: 31–59.

12.

Gordon-Brannan

(1994) Assessing intelligibility: Children’s expressive phonology. Topics in Language Disorders 14: 17–25.

13.

Gordon-Brannan

Hodson

(2000) Intelligibility/severity measurements of prekindergarten children’s speech. American Journal of Speech-language Pathology 9: 141–50.

14.

Hargrove

Dauer

Montelibano

(1989) Reducing vowel and final consonant prolongations in twin brothers. Child Language Teaching and Therapy 5: 49–63.

15.

Harris

Watson

Bates

(1999) Prosody and melody in vowel disorder. Journal of Linguistics 35: 489–525.

16.

Higgins

Hodge

(2002) Vowel area and intelligibility in children with and without dysarthria. Journal of Medical Speech-language Pathology 10: 271–77.

17.

Hodson

Paden

(1981) Phonological processes which characterise unintelligible and intelligible speech in early childhood. Journal of Speech and Hearing Disorders 46: 369–73.

18.

Kent

Miolo

Bloedel

(1994) Intelligibility in children’s speech: A review of evaluation procedures. American Journal of Speech-language Pathology 3: 81–85.

19.

Kim

Hasegawa-Johnson

Perlman

(2011) Vowel contrast and speech intelligibility in dysarthria. Folia Phoniatrica et Logopaedica 63: 187–94.

20.

Kwiatkowski

Shriberg

(1992) Intelligibility assessment in developmental phonological disorders: Accuracy of caregiver gloss. Journal of Speech and Hearing Disorders 35: 1095–1104.

21.

Liu

Tsao

Kuhl

(2005) The effect of reduced vowel working space on speech intelligibility in mandarin speaking. Journal of the Acoustical Society of America 117: 3879–89.

22.

Liss

Spitzer

Caviness

Adler

(2002) The effects of familiarisation on intelligibility and lexical segmentation in hyperkinetic and ataxic dysarthria. Journal of the Acoustical Society of America 112: 3022–30.

23.

Pascoe

(2004) Investigating intervention: Phonological therapy in a psycholinguistic framework. Unpublished PhD thesis, Department of Human Communication Sciences, University of Sheffield, UK.

24.

Pascoe

Stackhouse

Wells

(2006) Persisting speech difficulties in children. Chichester: Wiley.

25.

Peng

Spencer

Tomblin

(2004) Speech intelligibility of pediatric cochlear implant recipients with 7 years of device experience. Journal of Speech, Language and Hearing Research 47: 1227–37.

26.

Penney

Fee

Dowdle

(1994) Vowel assessment and remediation: A case study. Child Language Teaching and Therapy 10: 47–66.

27.

Pollock

(1994) Assessment and remediation of vowel misarticulations. Clinics in Communication Disorders 4: 23–37.

28.

Pollock

Berni

(2003) Incidence of non-rhotic vowel errors in children: Data from the memphis vowel project. Clinical Linguistics and Phonetics 17: 393–401.

29.

Pollock

Hall

(1991) An analysis of vowel misarticulations of five children with developmental apraxia of speech. Clinical Linguistics and Phonetics 5: 207–224.

30.

Pollock

Keiser

(1990) An examination of vowel errors in phonologically disordered children. Clinical Linguistics and Phonetics 4: 161–78.

31.

Prezas

Hodson

(2010) The cycles phonological remediation approach. In: Williams

McLeod

McCauley

(eds) Interventions for speech sound disorders in children. Baltimore, MD: Brookes.

32.

Reese

O’Hanlon

(2004) Vowel formants and intelligibility in normal and disordered children. Presentation at the ASHA conference, California State University, Sacramento, CA, USA.

33.

Reid

(2003) The vowel house: A cognitive approach to vowels for literacy and speech. Child Language Teaching and Therapy 19: 152–80.

34.

Reynolds

(1990) Abnormal vowel patterns in phonological disorder: Some data and a hypothesis. The British Journal of Disorders of Communication 25: 115–48.

35.

Stackhouse

Wells

(1997) Children’s speech and literacy difficulties: Book 1: A psycholinguistic framework. London: Whurr.

36.

Stackhouse

Wells

(2001) Children’s speech and literacy difficulties: Book 2: Identification and intervention. London: Whurr.

37.

Starr

Moller

Dawson

Graham

Skaar

(1984) Speech ratings by speech clinicians, parents and children. Cleft Palate Journal 21: 286–92.

38.

Stoel-Gammon

Herrington

(1990) Vowel systems of normally developing and phonologically disordered children. Clinical Linguistics and Phonetics 4: 144–60.

39.

Turner

Tjaden

Weismer

(1995) The influence of speaking rate on vowel space and speech intelligibility for individuals with amyotrophic lateral sclerosis. Journal of Speech and Hearing Research 38: 1001–13.

40.

Tye-Murray

Kirk

(1993) Vowel and diphthong production by young users of cochlear implants and the relationship between the phonetic level evaluation and spontaneous speech. Journal of Speech and Hearing Research 36: 488–502.

41.

Wells

(1986) Accents of English 2: The British Isles. Cambridge: Cambridge University Press .

42.

Whitehill

(2002) Assessing intelligibility in speakers with cleft palate: A critical review of the literature. Folio Phoniatrica et Logopaedica 54: 50–58.