Consonant production and overall speech characteristics in school-aged children with cerebral palsy and speech impairment

Abstract

The aim of the present study was to investigate the speech characteristics of school-aged children with cerebral palsy (CP) and speech impairment at various cognitive levels. Nineteen children with a mean age of 11;2 years (9;2–12;9 years) with spastic, dyskinetic, and ataxic CP and speech impairment participated. Phonetic transcription of oral consonants, ratings of hypernasality, and severity of overall dysarthria, together with free field descriptions of respiration, voice quality, and prosody, were performed independently by two speech-language pathologists. The non-verbal cognitive level was also studied. More than half of the children had large problems with the articulation of consonants, and the children with ataxic CP were most affected. The majority was rated as having dysarthria, mostly mild, but hypernasality was rare. Gross motor problems were not significantly associated with the articulation of consonants or the severity of dysarthria, whereas non-verbal cognitive level was. This underlines the importance of taking non-verbal cognitive level into account, when designing individual speech treatment programs for this group of children. Finally, a careful examination of the articulation of consonants is recommended in order to study speech production thoroughly in children with CP.

Keywords::

• Cerebral palsy
• children
• speech

Introduction

Research on the nature of speech impairment in children with cerebral palsy (CP) is limited and primarily related to the underlying nature of the gross-motor function, such as the type of CP (Hustad, 2010; Pennington, 1999). Motor speech impairment affect ∼ 20% of children with CP and are present in all types of CP (Andersen, Irgens, Haagaas, Skranes, Meberg, & Vik, 2008; Himmelmann, Hagberg, Wiklund, Eek, & Uvebrant, 2007; Nordberg, Miniscalco, Lohmander, & Himmelmann, 2013; Sigurdardottir & Vik, 2011; Watson, Blair, & Stanley, 2006). Spastic CP is the most common type of CP (70–80%), followed by dyskinetic CP (10–20%), and ataxic CP (5–10%) (Stanley, Blair, & Alberman, 2000). Studies of speech characteristics in CP have been carried out on children with spastic and dyskinetic CP (Hardy, 1983; Workinger, & Kent, 1991), whereas, to our knowledge, no studies have investigated the speech of children with ataxic CP. Workinger and Kent (1991) compared perceptual characteristics in 18 children with CP (nine spastic and nine dyskinetic) with severe gross motor problems and a mean age of 12;3 years (6;11–14;9 years), and found only one child with cognitive deficits. In that study, the most common perceptual speech characteristics for the children with spastic CP were consistent hypernasality, breathy voice, and change of voice quality (Workinger & Kent, 1991). For the children with dyskinetic CP, the most frequent speech deficits included reduced stress, inappropriate voice stoppage/release, and slow rate. Dyskinetic CP was related to more severe problems with articulation and the co-ordination of articulation movements or timing than spastic CP. Similarities between the speech profiles of the two groups were found in terms of phonation (Workinger & Kent, 1991). However, it is not obvious that generalization can be made based on the type of gross-motor disorder (such as the type of CP) and that they can be applied to speech motor movements (Hardy, 1983; Workinger & Kent, 1991).

Speech analyses in individuals with CP have primarily been auditory perceptual and have been carried out using rating scales. In the study by Workinger and Kent (1991), two listeners rated each of the 18 children with spastic and dyskinetic CP on the overall type and severity of dysarthria, as well as 22 perceptual speech dimensions. Definitions of the dimensions were given to each listener. A seven-point ordinal scale was used, with “1” indicating performance within normal limits and “7” indicating extremely severe involvement. The rating procedure and the chosen speech dimensions were carried out according to the conventional Mayo system for the classification of dysarthria in adults (Darley, Aronson, & Brown, 1969, 1975). This includes hypernasality, which is the most common symptom reported on velopharyngeal impairment of different origin. Hypernasality was the highest ranked perceptual speech dimensions for the children with spastic CP in the study by Workinger and Kent (1991), and there has been some interest of the velopharyngeal sub-system in individuals with CP. For example, were different patterns of velopharyngeal dysfunction presented in five children with CP and speech impairment in the late 1960s (Netsell, 1969). However, the main speech symptom, hypernasality, has received little attention in recent research.

Only a few studies of children with CP have attempted to give a detailed description of consonant production. In an early study by Byrne (1959), the speech of 61 children with CP was investigated. Sixty realizations of consonants were examined (22 initial, 19 medial, and 19 final), but it is not clear which speech material or which method was used. From the presentation of the results, the analysis appears to have been based on an assessment of correctness. The majority of the children produced /w/, /b/, /j/, and /d/ correctly, and at least 60% of the children produced /n/, /h/, /g/, /k/, and /t/ correctly in the initial word position. Consonants in the final and medial positions were the most difficult. Workinger and Kent (1991) also made an attempt to describe articulatory errors using a broad transcription of the speech material of the 18 children with CP. They found omissions of speech sounds, vowel errors, and substitutions, and noted that speakers with dyskinetic CP produced more errors than children with spastic CP.

The type of speech material and method used in the assessment could possibly influence the results. For example, Klintö, Salameh, Svensson, and Lohmander (2010) showed that word naming was the most reliable speech material in the speech assessment of children with a cleft palate. The speech material used in the studies of children with CP differs, or has not been clearly described. For example, Workinger and Kent (1991) used the repetition of five sentences and, in the early report by Byrne (1959), the speech material was not specified. In some of the few published studies on speech intervention in children with CP, speech was elicited by picture naming of single words and connected speech by picture description (Nordberg, Carlsson, & Lohmander, 2011; Pennington, Roelant, Thompson, Robson, Steen, & Miller, 2013).

Cognitive deficits appear to be common in children with CP and speech disorders, including deficits in a range of verbal and non-verbal processes, including language, learning, memory, reasoning, problem-solving, and attention. In a recent population-based Swedish study, ∼ 40% of the children with CP and speech disorders were reported to have cognitive deficits (Himmelmann, Hagberg, & Uvebrant, 2010; Nordberg et al., 2013). Moreover, Hustad (2010) reported data suggesting that almost two thirds of children with CP and speech motor disorders had cognitive deficits. Interestingly, Kent (2004) described recent theories pointing to an inter-relationship between speech and cognition, with links between speech motor processes and linguistic processes. Given these potential links, it is important to examine speech features in children with CP, with and without cognitive deficits, especially given that studies of speech characteristics in children with CP have mainly targeted children without cognitive deficits (Hustad, Gorton, & Lee, 2010).

Based on the findings in the literature, it can be concluded that there is a dearth of both overall and detailed descriptions of speech production in children with CP. It is, therefore, important to examine speech features in children with CP, regardless of cognitive level.

The overall aim of the present study was to investigate some speech characteristics of a consecutive group of school-aged children with CP and speech impairment, as well as various cognitive levels.

The specific research questions were:

• How does the consonant production of school-aged children with different types of CP and speech deficits compare?

• What is the occurrence of hypernasality and severity of the overall dysarthria in school-aged children with different types of CP and speech deficits?

• What is the relationship between speech production, gross motor function, and non-verbal cognitive level?

• Are there any specific speech characteristics noted in children with spastic, dyskinetic respective ataxic CP?

Methods

Participants

Nineteen children, nine girls and 10 boys with a mean age of 11;2 years (9;2–12;9 years) born in the Västra Götaland Region of Sweden between 1999–2002 and with a diagnosis of CP defined according to Bax, Goldstein, Rosenbaum, Leviton, Paneth, Dan, Jacobsson, & Damiano. (2005), participated in the present study. The CP diagnosis was verified by the local neuropaediatrician. Information about the children's gross motor function and hearing were taken from notes in the medical records of the children (Himmelmann et al., 2010). The inclusion criteria for the 19 children were speech impairment documented in the medical and speech and language therapy records (Himmelmann et al., 2010; Nordberg et al., 2013). Demographic data on the 19 children are presented in Table I. All but three of the children were monolingual Swedish speakers. The three children (S3 (Assyrian), S10 (Albanian), and S12 (Sorani)) were judged by their teacher to be bilingual, with Swedish at the same level as their native tongue. Eight of the children had been diagnosed with unilateral spastic CP, five children with bilateral spastic CP, two with dyskinetic CP, and four children with ataxic CP. Cognitive impairment had previously been reported as being present in nine of the 19 children (Himmelmann et al., 2010). Gross motor function had been classified into the five levels of the Gross Motor Function Classification System (GMFCS) (Palisano, Rosenbaum, Walter, Russell, Wood, & Galuppi, 1997). Eleven of the children in the study walked without limitations (level I), one child walked with limitations (level II), three children walked with a hand-held mobility device (level III), four children had self-mobility with limitations (level IV), and none of them was transported in a manual wheelchair (level V) (Himmelmann et al., 2010).

Table I. Demographic data for 19 children with CP and speech impairment.

Child	Gender	Age	Hearing deficit**	Type of CP**	Gross motor function level**	Receptive language measured with*** TROG-2 (percentiles)	Non-verbal cognitive level measured with**** RAVEN (percentiles)
S1	F	9:6	No	Unilateral Spastic	I	< 1^st	< 5^th
S2	M	9:9	No	Unilateral Spastic	I	10^th	25^th
S3*	M	12:8	No	Unilateral Spastic	I	34^th	< 5^th
S4	M	11:11	Hearing loss one ear	Unilateral Spastic	I	1^st	10^th–25^th
S5	F	12:9	No	Unilateral Spastic	I	34^th	25^th–50^th
S6	F	12:8	No	Unilateral Spastic	I	1^st	50^th–75^th
S7	F	9:8	No	Bilateral Spastic	IV	77^th	< 5^th
S8	M	10:7	No	Bilateral Spastic	IV	< 1^st	< 5^th
S9	M	11:3	No	Unilateral Spastic	III	2^nd	10^th–25^th
S10*	M	11:8	Monoaural deafness	Unilateral Spastic	I	1^st	< 5^th
S11	M	11:8	No	Bilateral Spastic	1	< 1^st	< 5^th
S12*	F	12:8	No	Bilateral Spastic	IV	< 1^st	< 5^th
S13	F	11:9	No	Bilateral Spastic	III	< 1^st	< 5^th
D1	M	11:10	No	Dyskinetic	III	34^th	25^th
D2	F	9:2	No	Dyskinetic	II	18^th	5^th
A1	F	10:1	No	Ataxic	I	14^th	5^th–10^th
A2	F	9:2	No	Ataxic	IV	< 1^st	< 5^th
A3	M	10:7	No	Ataxic	I	< 1^st	< 5^th
A4	M	12:01	No	Ataxic	I	1^st	< 5^th

F, Female; M, Male; S, spastic CP; D, dyskinetic CP; A, ataxic CP; GMFCS, Gross Motor Function System (GMFCS) has five levels for describing severity of gross motor impairment with level I—walks without limitations, level II—walks with limitations, level III—walks with adaptive equipment assistance, level IV—self-mobility is limited, likely to be transported in a wheelchair, level V—transported in a wheelchair.

*Bilingual children. **Himmelmann et al. (2010). ***Test for Reception of Grammar Version 2 (TROG-2) (Bishop, 2003, Swedish translation, 2009). Norms available for children 4–16.11 years. ****Raven Coloured Progressive Matrices (Raven et al., 1998). Norms available for children 5.6–11.6 years. For children > 11.6 we used the 11.6 norms.

Procedure

Test instrument

The standardized Swedish Articulation and Nasality Test (SVANTE) (Lohmander, Borell, Henningsson, Havstam, Lundeborg, & Persson, 2005) was used for the collection of speech data. SVANTE consists of 74 pictures designed to elicit one- and two-syllable single words. Sixty-four of these are isolated target consonants with 59 pictures for assessing the articulation of oral consonants and five with nasal consonants to evaluate hyponasality. The remaining 10 pictures are used to elicit s-clusters. The SVANTE test also includes the repetition of sentences and the elicitation of connected speech. In this study, the 59-single-word sample was used to assess oral consonant articulation. The articulation of the following consonants were assessed: /p/, /t/, /k/, /b/, /d/, /g/, /f/, /s/, /ç/. The sample, thus, includes all Swedish oral stops, the two sibilants, and one labiodental voiceless fricative. These consonants are considered particularly vulnerable for articulation deficits in contrast to the glides and voiced fricatives or approximants, which are therefore not included in the SVANTE test. For each included consonant, there are three possible realizations in the initial position and, for all except for /ç/, two in the medial and final position, respectively, with possibilities for decision on stability of the production. There are norm values for different ages, and the closest to the mean age in the current study are the 10-year norms, recently updated on a larger group of children (A. Lohmander, personal communication, n.d., 2014). The children's speech was audio recorded with a portable stereo audio recorder (TASCAM Frontier HD-P2) using a condenser microphone (Sony ECM-MS 957). The microphone was placed centrally on a table in front of the child during the assessment. The audio recordings were saved as separate wav. files. All data collection was conducted by a speech-language pathologist, with more than 20 years experience in the field (first author), and took place in a separate room at the children's schools, except for one child (S6), who was assessed at home. The words were elicited by picture naming. In cases in which a child was unable to find the target word and semantic prompting failed, imitation of the target word was used. Six of the audio recordings (31%) were randomly selected and duplicated for subsequent measurements of the consistency of the assessment. The total number of 25 audio recordings were randomized and coded from 1–25.

Current non-verbal cognitive level was measured in this study using progressive matrices, coloured version (RCPM) (Raven, Raven, & Court, 1998). The RCPM test consists of 36 items, and each of them contains a picture with a pattern with one part removed. There are six pictured inserts, one of which contains the correct pattern, and the child has to point to the pattern piece he/she considers to be the correct one. The maximum score is 36. This test is frequently used in studies of children with speech and language disorders. Since there are no standardized norms for the RCPM test on Swedish children, the British norms were used. Receptive language ability was assessed with the standardized Test for the Reception Of Grammar Version 2 (TROG-2) (Bishop, 2003; Swedish version, 2009). TROG-2 involves matching 80 orally presented sentences of increasing complexity to the correct picture out of a choice of four. The maximum score is 80. The information about receptive language was only used for description of the participants in the present study (Table I).

Analysis

Two experienced certified speech-language pathologists (SLPs) served as listeners for the perceptual analysis. The first author and an external SLP with more than 10 years’ experience as a SLP related to CP but not involved in the study and without any knowledge of the included children took part. Neither of the two SLPs had treated the children. Both SLPs assessed the speech material from all the children and six randomly selected recordings (32%) were duplicated for re-assessment and the calculation of intra-judge reliability. The assessments were performed within 2 months. Scale rating of dysarthria and hypernasality was performed from the whole speech material. The listeners rated each variable for each child separately using ordinal four-point scales (no, mild, moderate, and severe). First, the ratings of dysarthria took place and the listeners used the definition put forward by Darley, Aronson, and Brown (1969, 1975): “Dysarthria is characterized by slow, weak, imprecise and/or uncoordinated movements of the speech musculature”. Second, the ratings of hypernasality took place. In addition, the listeners were asked to give five free field descriptions of other speech characteristics.

Then, the phonetic transcription of the speech material was performed preceded by 9 hours of transcription calibration, whereby the two listeners transcribed recordings of four other children with CP and speech impairment. Narrow transcription based on the International Phonetic Alphabet (IPA) and the extended IPA for disordered speech [ExtIPA] conventions (IPA 1999, 2005) was used. Diacritics for laryngeal and respiratory function were also used. The use of different phonetic symbols and diacritics was discussed until consensus was reached. Every target sequence could be listened to repeatedly and all the speech sounds in each word were transcribed.

The two listeners then independently transcribed each sample for each child. Most of the transcribed single word samples comprised 58 or 59 words, but one child produced 52 words and another child (S13) 15 words. The transcriptions of the target consonants were analysed as correct or not correct, based on the scoring rules for the PCC (Shriberg & Kwiatkowski, 1982; Shriberg, Austin, Lewis, McSweeny, & Wilson, 1997) (Appendix). The total number of intended oral consonants (target consonants) correctly articulated was then divided by the total number elicited, giving the percentage of correctly articulated oral consonants. The child with only 15 realizations was excluded from the analysis of the percentage of correctly articulated oral consonants. In addition to the analysis of correctness, an analysis of articulation error patterns for each speaker was performed according to the procedure used by Workinger and Kent (1991), with the calculation of frequency of articulation errors. The type of consonant errors searched for from the phonetic transcriptions of the target consonants were substitutions, omissions, additions, voicing, and nasalization.

The descriptive overall comments of the speech characteristics in free field were classified into respiration and speech breathing, voice quality, and prosody.

Reliability

The intra- and inter-transcriber reliability of the consonant transcriptions using the IPA and of scale ratings of speech variables (dysarthria and hypernasality) were calculated using percentage agreement, point by point. The median values of intra-transcriber agreement from the re-transcription of 32% of the material was excellent for both transcribers; 95 and 100%, respectively. The intra-rater agreement of the ratings of dysarthria and hypernasality was 100% for both raters. The median values of inter-transcriber agreement were good (80–90%), although for three children the agreement was below 70% and regarded as not acceptable. A third experienced speech-language pathologist and researcher in speech-language pathology (third author) therefore performed complementary transcriptions for six of the 19 children (32%). It was the transcriptions of the three children the two first transcribers did not agree on. Another three transcriptions were randomly added from the sample.

The inter-rater reliability for ratings was poor, with 68% for dysarthria and hypernasality. The third author, therefore, performed complementary ratings for the speech material for 13 of the 19 children (68%) and the median value was used. When the complementary ratings by the third author were completed, the inter-rater reliability for dysarthria was 100%, while it was 95% for hypernasality.

Statistical methods

Statistical analyses of the data were carried out using the SPSS (Statistical Package for Social Sciences) version 19 for Windows (SPSS Inc., Chicago, IL). The analyses contained descriptive statistics for the speech results. Associations between variables were evaluated with Spearman's rank correlation coefficient. The level of significance was set at 0.05.

Results

Consonant articulation

The mean value for the percentage of correct oral consonants was 83.7% (49.6–100) for the group of 18 children (child S13 was excluded, due to small speech material), which is below −2 SD for typically-developing 10-year-olds (92.3%) or below the 5^th percentile (88.2%) (A. Lohmander, personal communication, n.d., 2014). The percentage of correct oral consonants and distribution by type of CP is shown in Table II. Division by type of CP revealed mean values (range) for 12 children with spastic CP (S13, excluded) of 90.6% (49.6–98.3) and ataxic CP of 72.8% (50–91.4). One child with dyskinetic CP (D1) produced all the oral consonants correctly and the other (D2) had 91.5% correct oral consonants.

Table II. Correct consonant articulation, frequency of consonant errors, and results of overall listener ratings of speech characteristics for the children with spastic, dyskinetic, and ataxic CP. A four-point scale was used for the listener ratings (0 = no disorder, 1 = mild, 2 = moderate, 3 = severe). Standard scores for the SVANTE-test are: −1 SD = 95.5%, −2 SD = 92.3%, 5^th percentile = 88.2 %, 10^th percentile = 95.3%.

	Consonant articulation SVANTE-test (%) Correct	Type and numbers of consonant errors							Listener ratings dysarthria	Listener ratings hyper-nasality	Verbal and non-verbal profiles***
Type of CP		Substitution	Omission	Addition	Distortion	Voicing	Nasalization	Total
Spastic CP
S1	78.0	8		4		1		13	1	0	V+ N-V
S2	89.3	2				4		6	0	0	0
S3	98.3					1		1	0	0	N-V
S4	93.1				2	2		4	1	0	V
S5	98.3					1		1	0	0	0
S6	89.8					8		8	1	0	V
S7	93.1				2		4	6	1	2	N-V
S8	96.5		1			1		2	1	1	V+ N-V
S9	79.7	1	1	1		9		12	1	0	V
S10	91.4			1		4		5	1	0	V+ N-V
S11	49.6	4	21	2		4		31	3**	1	V+ N-V
S12	66.1	14	2		1	2		19	2	1	V+ N-V
S13	*	17	35			2		54	3	2	V+ N-V
	Mean (range) 90.6 (49.6–98.3)	Total (%) 29 (26.9)	Total (%) 25 (23.1)	Total (%) 8 (7.4)	Total (%) 5 (4.6)	Total (%) 37 (34.3)	Total (%) 4 (3.7)
Dyskinetic CP
D1	100							0	2	0	0
D2	91.5				2	3	5	10	2	2	N-V
Ataxic CP
A1	91.4	4				1		5	1	0	N-V
A2	74.1	6	2			7		15	2	0	V+ N-V
A3	50	23		1		2		26	3**	0	V+ N-V
A4	75.8	4		1	3	7		15	1	0	V+ N-V
	Mean (range) 72.8 (50–91.4)	Total (%) 37 (60.6)	Total (%) 2 (3.3)	Total (%) 2 (3.3)	Total (%) 3 (4.9)	Total (%) 17 (27.9)	Total 0

V, verbal deficits only; V+ N-V, verbal and non-verbal deficits; N-V, non-verbal deficits only.

*S13 had small speech material and was, therefore, not included in the data analysis of correctly articulated consonants.

**S11 and A3 were commented on as having dyspraxia of speech.

***Verbal and non-verbal deficits = cut-off scores below 10^th percentile on TROG-2 (Bishop, 2003) and Raven Coloured Progressive Matrices (Raven et al., 1998).

Type of consonant errors

The consonant error patterns for each child are given in Table II. The most frequent errors for children with spastic CP were voicing errors: 37 (34.3%), substitutions: 29 (26.9%), and omissions: 25 (23.1%). One child with dyskinetic CP had few articulation errors and the other no errors. For the four children with ataxic CP, the most frequent consonant errors were substitutions: 37 (60.6%) and voicing errors: 17 (27.9%).

Severity of dysarthria

The occurrence of dysarthria and distribution by type of CP are shown in Figure 1. Three children (15.8%) were rated as not having dysarthria. Nine children (47.4%) were judged to have mild dysarthria, four moderate dysarthria (21.1%), and three children (15.8%) severe dysarthria. More than half the children with spastic CP were rated as having mild dysarthria, while the two children with dyskinetic CP were rated as having moderate dysarthria. All the children with ataxic CP were rated as having dysarthria.

Figure 1. Number of children (n = 19) by type of CP with degree of overall severity of dysarthria rated on a 4-point scale.

Hypernasality

The occurrence of hypernasality and distribution by type of CP are presented in Figure 2. Thirteen of the 19 children (68.4%) were rated as having no hypernasality, three children (15.8%) mild hypernasality, and three (15.8%) moderate hypernasality. No child had severe hypernasality.

Figure 2. Number of children (n = 19) by type of CP with degree of hypernasality rated on a 4-point scale.

More than half the children with spastic CP were rated as having no hypernasality, three as having mild hypernasality, and two children as having moderate hypernasality. One child with dyskinetic CP was rated as not having hypernasality and one as having moderate hypernasality. All the children with ataxic CP were rated as not having hypernasality.

Descriptive comments on the children's speech

Respiratory and speech breathing problems were commented on for 13 of the 19 children (68.4%): low energy, breathless and timing difficulties when speaking, insufficient air for speech production, low-energy voice, weak voice, excess loudness variation, and loudness decay.

Voice problems were also commented on in 13 of the 19 children (68.4%): deviant voice quality, high pitch level voice, strained-strangled voice, harsh voice, wet or liquid sounding hoarseness, and creaky voice.

Comments on prosody were made for six of the 19 children (31.6%): deviant prosody, fluctuations of deviant prosody, monotonous speech, and deviant equal stress on usually unstressed parts of speech.

Individual consonant articulation, overall speech characteristics, and level of gross motor function

Each child's scores for correct oral consonants and overall listener ratings are presented in Table II. Children S11, S12, and A3 had the lowest score for correct oral consonants (below 70%). Two of them, S11 and A3, were also commented on as having dyspraxia. Child S13 (excluded from the consonant articulation, due to small speech material), with moderate motor problems, was the only child rated as having severe dysarthria, while the hypernasality was rated as moderate. In the free field descriptions made by the listeners, all the speech sub-systems were mentioned for child S13 (the speech sub-systems consist of respiration, laryngeal function, velopharyngeal function, and articulation). Child S12 was rated as having moderate dysarthria but mild hypernasality.

Moreover, the four children, S1, S9, A2, and A4, had low scores for correct oral consonants (below 80%). Three of them were only rated as having mild dysarthria (S1, S9, and A4), and one of them (S9) also had mild hypernasality. Two children, S7 and S8, with severe motor problems, had high scores for correct oral consonants (above 90%). Child S7 was rated as having mild dysarthria and moderate hypernasality, while child S8 was rated as having mild dysarthria and mild hypernasality.

Non-verbal cognitive level

The non-verbal cognitive level measured with Raven's coloured progressive matrices (RCPM) (Raven et al., 1998) for each child is presented in Table I. As there are no norms for the RCPM test on Swedish children, our results were computed from the British norms. More than half, 11 of 19 children (58%), had very low scores on the RCPM test, below the 5^th percentile.

Correlations

Speech production and gross motor function

There was no significant correlation between the articulation score or the ratings for dysarthria and the level of gross motor function (Rho = .082, p = .746; Rho = .381, p = .108), whereas hypernasality correlated significantly with gross motor function (Rho = .484, p = .036).

Speech production and non-verbal cognitive level

There was a significant positive correlation between the articulation score and the children's raw scores on Raven's matrices, i.e., the children with high scores for the correct production of oral consonants had high scores on Raven's matrices (Rho = .561, p = .015). Similarly, there was a significant negative correlation between the ratings of dysarthria as well as hypernasality and the children's raw scores on Raven's matrices, i.e., the children with mild or no dysarthria or mild or no hypernasality had high scores on Raven's matrices (Rho = −.540, p = .017; Rho = −.514, p = .024).

Discussion

In this study, the speech characteristics of a group of 19 school-aged children with CP and speech impairment, as well as current non-verbal cognitive level, were investigated. The mean score for correct consonant articulation was well below −2 SD for typically-developing children or below the 5^th percentile. Three children (S3, S5, D1) scored according to the age norm, and three children (S4, S7, S8) more close to the norm, i.e., 33% had typical or close to typical consonant articulation.

The descriptions of articulation and speech characteristics, in this study, are similar to those used in the study of children with CP by Workinger and Kent (1991). However, there were dissimilar findings between the two studies. First, Workinger and Kent found that children with dyskinetic CP produced more articulation errors than children with the spastic type of CP, but this was not the case in the present study, where one child with dyskinetic CP produced 100% and the other 91.5% of the consonants correctly. Possible explanations for the differences in both occurrence and type of articulation problem found in the present study compared to Workinger and Kent (1991) could be differences in speech materials and methods for perceptual analysis. Workinger and Kent (1991) used short speech material based on imitation and assessment by broad transcription, whereas a comprehensive single-word naming test and narrow transcription were used in the present study. Narrow transcription is preferred when the aim is to record which phonetic variant a speaker uses and when it is not possible to predict in what area and how someone's speech will be impaired (Heselwood & Howard, 2008). The result of articulation of consonants in the present study is based on a large number of realizations of each target consonant in different positions, making measurements of occasional realizations unlikely. Second, the procedure for assessing hypernasality was not described in detail by Workinger and Kent (1991), who presented hypernasality as the highest ranked speech characteristic among the children with spastic CP, whereas more than half the 13 children with spastic CP in the present study were rated without hypernasality. Difficulties in the auditory perceptual assessment of speech are well known (Kent, 1996) and the reliability of perceptual assessments of hypernasality has often been reported as low, with poor agreement between listeners (e.g., Keuning et al., 1999; Brunnegård & Lohmander, 2007). This was also the case in the present study. For this reason, an experienced third SLP performed ratings in all patients when there was disagreement between the first two raters. This enabled more reliable assessments of hypernasality to be presented, which in contrast to Workinger and Kent (1991) revealed no hypernasality in the children with spastic CP. The question of whether the contradictory results could be related to the methodology used or to actual differences should be explored in future studies. Another finding in this study was that gross motor function correlated significantly with hypernasality, which indicates that level of motor function is related to the presence of nasality in children with CP. Third, voicing errors were the most frequent type of articulation problem for the children with spastic CP in this study, whereas omissions were most common in the study by Workinger and Kent (1991). An inability to produce perceptually different voiced and voiceless speech sounds is often a consequence of poorly controlled and timed phonation (Yorkston, Beukelman, Strand, & Hakel, 2010). Interactions of factors involved in the respiratory, laryngeal, and oral articulatory sub-systems need to be controlled when producing the important linguistic contrasts of voiced/voiceless consonants (Löfqvist, 1992), which is probably a challenge for many children with CP. Additional reasons for the dissimilar findings between the two studies may be the heterogeneity of the children with variations in the level of the children's gross motor function and cognition. It has been stated that motor behaviour is associated with cognitive influences (Kent, 2004); one example of this is the auditory-motor linkage for speech. Fadiga, Craighero, Buccino, and Rizzolatti (2002) observed adult subjects during speech listening and, using transcranial magnetic stimulation, they found increased motor-evoked potentials from the listeners’ tongue muscles, as support for this linkage. Cognitive-motoric interaction can also be seen in early infancy. It is well established that normal newborn infants imitate the lip and tongue movements of others (Meltzoff, 1999), a specific capacity to judge auditory-motor correspondence. For this reason, not only motor function but also cognition are important factors to take into account when studying children with CP.

Non-verbal cognitive level was assessed in the present study. Only two children (S5, S6) scored within the typical age range. Nine of the 19 participating children had previously been identified with cognitive impairment (Himmelmann et al., 2010), but 17 children in the current study scored far below their age on Raven's coloured matrices. The non-verbal cognitive level was associated with poor consonant articulation and with dysarthria. This result highlights the importance of considering cognitive level when examining speech characteristics in children with CP (Hustad, 2010; Kent, 2004). Moreover, it seems reasonable to consider both cognitive and linguistic level when planning suitable intervention of the individual child.

The four children with the rare type of ataxic CP displayed more severe problems with the production of consonants than the children with spastic and dyskinetic CP. Substitutions were the most common error type. The high frequency of substitutions may be caused by the specific motor problems of children with ataxic CP, making the articulation of higher complexity consonants, such as fricatives, difficult (Straub & Obrzut, 2009). Notably, the children with the ataxic type of CP were rated as only having articulation difficulties, as no hypernasality was found. To our knowledge, specific data on consonant articulation in children with ataxic CP have not previously been reported.

It is a matter of great priority to find appropriate tools for assessing the speech ability of children with CP. Different procedures are used to identify the children's motor speech impairment, but no agreed clinical paediatric assessment protocol of the speech sub-systems exists. The Verbal Motor Production Assessment for Children (VMPAC) instrument (Hayden & Square, 1999) is an oro-motor investigation procedure with non-speech and speech oro-motor control sub-tests related to the different speech sub-systems to distinguish dysarthria from dyspraxia and determine the severity of these disorders. In addition, a test of articulation would be appropriate to study speech production in children with CP in order to detect speech symptoms that might be relevant for intervention. A majority of the children in the present study were rated to have mild dysarthria. The perceptual cues for the listeners who rated the severity of dysarthria were probably not linked to the articulation sub-system. Instead, the children's deviant voice quality (including hypernasality) and respiration appear to have been relevant for the listeners’ perception of dysarthria in these children. This observation was based on free field comments in the present study and needs to be investigated further in a larger group.

Limitations

In this study, we focused on the speech production difficulties of children with CP. Children's speech impairment may sometimes result from phonological difficulties and sometimes from phonological/articulatory difficulties in combination. Further analyses of speech and language deficits in the present group may help to disentangle whether the speech difficulties are due to articulation or phonological problems. Moreover, although these children represented a population-representative consecutive group, the present sample size was small, which calls for caution when interpreting the results.

Conclusion

This study demonstrated that children with CP constitute a very heterogeneous group of individuals, even within each type of CP, with a range of difficulties in diverse areas, unique to each individual child. More than half the children had large problems with the articulation of consonants and the children with ataxic CP had the most severe problems. Interestingly, only hypernasality was statistically associated with gross motor problems, whereas the other speech impairments were significantly linked to non-verbal cognitive level. There is a need to conduct comprehensive assessments that include cognitive and linguistic functioning, not just assessments of speech sound production. It is also important taking cognitive and linguistic level into account. Finally, a careful examination of the articulation of consonants is recommended in order to study speech production thoroughly in children with CP. More research and investigations are important to obtain knowledge to promote a greater understanding of speech production in children with CP.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.

Appendix: Modified PCC scoring rules

• Certain deletions (e.g., deletion of final /r/) and co-articulations common in spoken Swedish (e.g., /n/ in connection with /k/ changes to /ŋ/) are scored as correct.

Scored as incorrect:

• Deletions or substitutions of a target consonant;

• Diacritic signs, such as for laminal/lateral/inter-dental production, de-voicing (with exception of aphonic speech), voicing, aspiration, nasal escape, slightly nasalised oral consonants, weak articulation;

• Changed place or manner of articulation;

• Additions, such as when /f/ becomes /vf/, if not self-corrected in the subsequent utterance. If it is not obvious to which of the closest consonants the added phoneme is related, it is excluded;

• Deletions, except for deletion of a word ending, if it is a common deletion in spoken language; and

• In metatheses, every consonant that does not match the right place in the target word is scored as incorrect.