Delayed and disordered development of articulation and phonology between four and seven years

Abstract

Some children’s speech impairment resolves spontaneously. Others have persistent problems affecting academic and social development. Identifying early markers that reliably predict long-term outcome would allow better prioritization for preschool intervention. This article evaluates the significance of different types of speech errors, made by 93 four-year-olds in a longitudinal population cohort study, for performance at seven years. At four years, the non-age appropriate speech errors made on standardized assessments were categorized as: phonologically delayed (error patterns typical of younger children); or, some errors atypical of normal development, including consistent errors (e.g. word initial consonant deletion), inconsistent pronunciations of the same word, or lateral distortion of /s, z/). Delayed children, some with occasional interdental /s, z/ articulation errors, were more likely to resolve (67%) than those making atypical errors (35%) by seven years. Qualitative analyses indicated that children making few atypical errors were more likely to resolve, irrespective of total number of errors or whether they received intervention. The findings’ theoretical implications relate to deficits underlying phonological disorders. Clinical implications concern assessment measures and prioritization for intervention.

Keywords

articulation longitudinal study phonological delay phonological disorder predicting persistence

I Introduction

At around nine percent, speech impairment is the most prevalent communication disorder in children aged 3–17 years (NIDCD, 2016). Children’s school achievement is often adversely affected (Hayiou-Thomas et al., 2016) as are their peer relationships, due to social, emotional and behavioural difficulties (Murphy et al., 2014). Persistent impairment in adolescence affects literacy, mental health and employment (Law et al., 2017). While more than two-thirds of children with speech impairment receive intervention in the USA (NIDCD, 2016), limited resourcing often constrains the amount and type of intervention accessed (Law et al., 2017). The problem, then, is how limited resources can best be allocated to provide effective intervention services for the many children with impaired speech. The case data reported here contribute to that debate.

Some authorities argue that most speech difficulties resolve spontaneously by early primary school. Bishop and Leonard (2001: xii) wrote: ‘We need to balance the benefits of starting intervention early …, against wasting intervention resources on children likely to resolve spontaneously.’ Denying intervention to all preschool children with impaired speech is not, however, indicated by research documenting its consequences for children and their families (McLeod, 2015). Other researchers have used longitudinal population studies to seek markers identifying children at risk for persistent speech impairment. Wren et al. (2016) analysed data from the Avon Longitudinal Study of Parents and Children (ALSPAC). Children with low percent consonants correct (PCC) scores in connected speech at eight years were included in a multivariable logistic regression to identify predictors of failure to resolve. Being male and of lower socio-economic status (SES) placed children at risk for persistent difficulties, as did parental report of ‘motor, cognitive, and linguistic (including speech intelligibility)’ abilities (p.647). More specific criteria are required, however, when making clinical decisions about individuals competing for intervention resources.

Another population study (Morgan et al., 2017) examined longitudinal data from the Early Language in Victoria Study (ELVS), a prospective, community study of children recruited before 12 months of age and reassessed almost annually until 13 years. Of 1,494 participants assessed at four years, 164 (11%), made non-age appropriate speech errors. Of these 164 children, 93 were available for reassessment at seven years when a standardized assessment determined whether their speech impairment had resolved or persisted. Over 40% were identified as having persistent errors. A logistic regression evaluated family history, gender, SES, non-verbal intelligence quotient and speech error type at four years as predictors of outcome at seven years. Children were categorized as making errors that were delayed (like younger children) or atypical (of normal development). Type of speech error was the only statistically significant predictor (p = 0.02) of outcome. Children with delayed errors at four years were twice as likely to resolve compared to those making atypical errors.

Morgan et al.’s (2017) findings suggest further exploration of the prognostic value of speech error types. Both Waring and Knight (2013) and Law et al. (2017: 1) identify ‘the heterogeneity of presentation’ of impaired speech as an issue for research into its nature and treatment. Three predominant classification approaches were evaluated by Waring and Knight (2013). Shriberg’s et al. (2010) Speech Disorders Classification System (SDCS) identifies eight subgroups based primarily on aetiology (e.g. hearing). Stackhouse and Well’s (1997) Psycholinguistic framework identifies profiles of deficits in speech processing associated with speech errors (e.g. auditory discrimination). Dodd (1995) distinguished five subgroups by speech error type, describing each subgroup’s speech processing abilities, allowing differential diagnosis linked to specific treatment approaches (for summary, see Dodd, 2014). Waring and Knight (2013: 1) concluded that ‘The differential diagnosis system has a growing body of empirical support across languages … and treatment efficacy studies’. Of the three approaches, it is the only one to describe characteristic error patterns for each subgroup.

Broomfield and Dodd’s (2004) study described all children with impaired speech referred to one UK National Health (Speech Language Therapy) Service over 15 months. Assessment batteries examined confrontation naming (75 words), phonological awareness, expressive and receptive language, and oro-motor skills. Children with intellectual, physical, sensory or psychiatric diagnoses were excluded. Of the 320 assessed: 12.5% had an articulation disorder (e.g. lateral lisp); 57.5% had delayed phonological development, only making errors typical of the normative data of younger children; 20.6% consistently used some speech error patterns atypical of normal development (e.g. backing, affrication); and, 9.4% made inconsistent errors on the same lexical item. No child was diagnosed with childhood apraxia of speech (i.e. inconsistency plus oro-motor signs like syllable segregation). One way of evaluating the clinical usefulness of these subgroups is to test whether types of speech errors can identify children who fail to spontaneously resolve.

Previous research provides evidence that atypical errors are associated with persistent speech impairment. One case study of a toddler, initially aged 21 months (Leahy and Dodd, 1987) indicated that atypical error patterns (e.g. bilabial fricative for all plosive + /l, r, w, j/ clusters, deletion of all syllable final consonants except nasals) persisted until 44 months, when intervention began. Older pre-schoolers participated in Wright’s (2014) longitudinal study that assessed 85 speech impaired children twice, two years apart. Initial assessment was at referral to therapy (aged 36–60 months). Two years later, 57% of children had age appropriate speech, a proportion similar to that reported by Morgan et al. (2017). In Wright’s study, of children initially identified as making only delayed errors, 69% had resolved as opposed to 47% of those initially making atypical speech errors. Forty-six children attended between one and 30 hours of intervention (e.g. home programme, direct individual therapy), making greater PCC gains than untreated children. Comparison between subgroups was confounded, however, as those receiving treatment had made more errors in total, more atypical as opposed to delayed errors, and were younger than the untreated participants at initial assessment. Wright (2014) concluded that clinicians in Ireland use both severity and type of speech error to prioritize cases for treatment.

Speech and literacy difficulties often go together, but not all children with impaired speech have difficulty learning to read and spell (Hesketh 2004). Some studies suggest that speech error type predicts specific literacy difficulties (Hayiou-Thomas et al., 2016). Preston et al. (2013) assessed 25 preschool children at four years and then again at eight years. All had histories of speech impairment but age appropriate receptive language. Different speech error types (articulatory distortion, typical vs. atypical substitutions and syllable structure errors) at four years were used to predict speech, phonological awareness (PA) and literacy outcomes at eight. While the group as a whole showed below average speech performance and low average PA, their reading and spelling was age appropriate. However, children making more than 10% atypical errors performed more poorly on PA and literacy measures than children making fewer than 10% atypical errors. In addition, articulatory distortion in preschool predicted persistent school-age difficulties. Preston et al. (2013: 173) concluded that ‘Different preschool speech error patterns predict different school-age clinical outcomes’.

Speech error types may reflect phonological processing difficulties associated with specific literacy profiles. For example, inconsistent word pronunciation is a marker for later spelling, but not reading, difficulties, even when speech difficulties have resolved (Holm et al., 2007). One study compared the PA abilities of 14 preschool children who made inconsistent speech errors on the same lexical item with 15 age matched typical controls. The inconsistent group performed poorly on syllable segmentation but no differently from controls on the other PA tasks. Another experiment indicated that seven-year-old children with a history of inconsistent phonological disorder performed no differently from controls on measures of phonological awareness and reading, but less well on spelling measures.

The research reviewed suggests that types of speech errors predict long term speech and literacy outcomes. This article presents qualitative analyses of speech data from the 93 ELVS participants whose four- and seven-year-old quantitative assessment results were reported in Morgan et al. (2017). The study investigates the relationship between types of errors at the two assessment points. Children were categorized as making either delayed or atypical errors (including children consistent making non-developmental phonological errors, inconsistent phonological errors, and articulation errors). The study charts changes in types of errors made between four and seven years for each subgroup of speech disorder. The research questions addressed whether diagnosis of speech subgroup at age four years predicts outcome at age seven years. Specifically:

What proportion of children in each of the four subgroups resolve by seven years?

For children with persistent speech disorder, how are the type of errors made at four years related to the number and type of errors made at seven years?

It was hypothesized that atypical errors at four years is a risk factor for persistent difficulties.

II Methods

1 Participants

The ELVS community cohort study participants were recruited in 2003–04 from six local government areas in Melbourne, Australia, representing low, middle and high SES communities (see Reilly et al., 2010). All eligible families were invited to participate. Children with known disabilities were excluded at baseline (e.g. Down syndrome, hearing loss). Parents needed adequate English to complete written questionnaires and provide written informed consent, as determined by primary recruiters, nurses from the Victorian Maternal and Child Health service. Ethics approval was obtained from the Royal Children’s Hospital Human Research Ethics Committee (HREC#23018).

At four years, all participants taking part in the fifth wave of ELVS data collection (n = 1,494) were assessed, with 164 children identified as speech impaired by standardized assessment (i.e. standard scores below the normal range, or inconsistency ≥ 40%). This group’s speech was reassessed at seven years. Loss to follow-up occurred with 54 participants withdrawing after the four-year-old assessment. An additional 17 children’s assessments were not completed or had audio-recordings that could not be fully re-transcribed, leaving 93 participants with both four- and seven-year-old complete speech assessment data. Participant and non-participant groups at seven years were compared on demographic variables of gender, family history and non-verbal IQ. No statistically significant differences were found.

2 Procedures

Assessments were usually conducted at the child’s home with one parent present. At both assessments, participants’ speech production was examined using the standardized Goldman–Fristoe test of articulation 2nd edition (GFTA-2) sounds-in-words subtest (Goldman and Fristoe, 2000). Research assistants were formally trained in the GFTA-2 procedures and followed a set protocol of assessment administration. This tool requires children to name 30 single words in response to picture stimuli. At the four-year-old assessment, children also completed the inconsistency subtest of the Diagnostic Evaluation of Articulation and Phonology (DEAP) (Dodd et al., 2002) where children name 25 pictures three times, each trial separated by another activity.

Video and audio recordings made of one-to-one assessments at four and seven years were transcribed using narrow phonetic transcription. To evaluate reliability, data from 10% of randomly chosen participants were transcribed twice (KB, KN). Point-to-point reliability was 96.8%. To ensure accurate diagnosis of type of speech disorder, AM and BD re-transcribed data from all children, resolving any disagreements.

The following measures were derived from GFTA-2 at both assessments: PCC, percent vowels correct (PVC), phonological error patterns (age-appropriate, delayed and atypical error patterns based on normative data), and, number and type of articulation errors.

Children’s four-year-old data were analysed using the following criteria to determine type of speech impairment:

Articulatory distortion: where occurrence of distortion of a phone was more frequent than correct production. Note that articulatory and phonological errors can co-occur and those children making both were monitored as a separate group at four and seven years.

Phonological delay: consistent use (three examples in different words) of an error pattern used by at least 10% of children of a younger chronological age according to DEAP normative data. Examples are: stopping of fricatives after the age of three years and five months; two-member consonant cluster reduction after 3 years 11 months.

Consistent phonological disorder: consistent use of atypical error patterns (three examples in different words), observed in fewer than 10% of the DEAP normative sample at any age (e.g. sound preference: initial consonants are [h]; backing: marking alveolar consonants with velars).

Inconsistent phonological disorder: production of the same lexical item differently, at least once in three trials, on 40% or more on the DEAP inconsistency subtest (e.g. [lɛlifən], [ælifənt] and [ɛfifənt] for elephant).

At seven years, data analyses determined whether articulatory distortions and/or speech error patterns had resolved (were age appropriate according to DEAP normative data) or persisted. If errors persisted, how they had changed since the child was four years old was described, using the above criteria.

III Results

1 Does the type of errors made at four years predict resolution at seven years?

Longitudinal data are shown in Table 1. Of the 93 children assessed at four years, there were:

67 (72%) making delayed phonological errors: 54 only made delayed phonological errors, and 13 made both delayed phonological errors, plus some interdental distortions of /s, z/. Of these two delayed groups 45 (67%) resolved by seven years.

26 (28%) making atypical errors: 21 children consistently used atypical error patterns (one of these children also made some interdental productions of /s, z/); four children made inconsistent errors; and, one child produced all /s, z/ sounds laterally but made no other errors. Only nine children (35%) making atypical errors resolved.

Table 1.

Children making speech errors at four years: Proportion resolved at seven years, by specific error types (n = 93).

	At four years	At seven years
	At four years	Resolved	Persistent
Delayed errors (72% of total n):
Phonological delay alone	54 (58% of total n)	36 (67% delay alone)	18 (33% delay alone)
Phonological delay + articulation (interdental lisp)	13 (14% of total n)	9 (69% Delay+IL)	4 (31% Delay+IL)
Atypical errors (28% of total n):
Articulation (lateral lisp)	1 (1% of total n)	none	1
Consistent atypical	20 (22% of total n)	9 (45% of CPD)	11 (55% of CPD)
Consistent atypical + articulation (interdental lisp)	1 (1% of total n)	none	1
Inconsistent	4 (4% of total n)	none	4

Notes. CPD = consistent phonological disorder; IL = interdental lisp.

A two-tailed Fishers exact test, used to determine whether children with delayed phonological development (with or without associated articulatory distortions), were more likely to resolve than children who consistently used atypical error patterns, inconsistent errors or lateral /s, z/ articulation alone, was significant (p = 0.005).

2 How many speech errors were made by subgroups at four and seven years?

PCC and PVC for each subgroup at each assessment are shown in Table 2. The data suggest most children made gains in the three years between assessments. Those who resolved, of course, made most progress. Children with persistent speech difficulties varied within and across subgroups:

Unresolved delay showed mean PCC improvement of 17% (sd 13, range 0–38%).

Consistent use of atypical error patterns at four years, that did not resolve, was associated with a mean PCC improvement of 16% (sd 9, range 5–35%).

The four making inconsistent errors had PCC increases of 19, 24, 35, and 51%.

There was no PCC change for the one child with a lateral lisp.

Table 2.

Mean PCC and PVC correct (sd) and range, at four and seven years for children making different types of speech errors according to outcome.

	Four-year-old assessment		Seven-year-old assessment
	PCC	PVC	PCC	PVC
Resolved at seven years:
Delay n = 36	81 (9) 50–97	97 (3) 92–100	98 (2) 93–100	99 (1) 99–100
Delay + IL n = 9	81 (6) 70–87	99 (1) 96–100	97 (2) 95–98	100 (0)
Consistent atypical n = 9	75 (9) 56–83	98 (3) 93–100	97 (2) 96–100	100 (< 1) 99–100
Persistent speech impairment at seven years:
Delay n = 13 At seven years, no change	77 (12) 53–91	96 (4) 90–99	92 (5) 80–98	99 (1) 99–100
Delay n = 5 At seven years, delay + IL	81 (4) 74–85	97 (2) 95–100	86 (3) 82–90	100 (< 1) 99–100
Delay + IL n = 4 At seven years, delay or IL	77 (10) 67–85	100 (0)	89 (2) 87–91	100 (< 1) 99–100
Consistent atypical n = 12 At seven years, mainly delay	74 (9) 58–85	92 (5) 94–100	89 (7) 74–97	99 (2) 95–100
Inconsistent n = 4 At seven years, varied	56 (16) 34–72	81 (13) 65–96	86 (12) 70–97	97 (3) 94–100
Lateral lisp n = 1	89	100	85	100

Notes. IL = interdental lisp; PCC = percent consonants correct; PVC = percent vowels correct.

Severity measures at age four years gave little indication about likelihood of resolution by seven years, apart from children making inconsistent errors (see Table 2).

3 Did the type of errors made by children with persistent speech impairment change?

Of 18 phonologically delayed children who did not resolve, all continued to make delayed errors. Five children, however, acquired an interdental lisp despite producing no distortions of /s, z/ at four years. At seven years, all but one produced sibilants correctly in some phonetic contexts (e.g. final clusters). Four children with delayed phonological errors plus some interdental /s, z/ productions at four years did not resolve by seven years. Two made delayed phonological errors, but no articulation errors. Two made no phonological errors but still made some interdental /s, z/ errors.

Only one of the 11 children consistently using atypical error patterns at four years continued to do so at seven years. Of the remaining 10 children, eight used delayed developmental error patterns, one with an acquired interdental lisp of some /s, z/ productions. Another two children had also acquired some interdental productions of /s, z/ but made no other errors. All three children could accurately articulate /s, z/ in some phonetic contexts. The child who had both consistent atypical errors and some interdental /s, z/ articulation at four years, had delayed phonology plus some interdental /s, z/ errors at seven years.

The four children who initially made inconsistent errors all continued to make speech errors when assessed at seven years. Unfortunately, inconsistency was not re-assessed. One child’s errors included 15 vowel errors as well as other atypical errors. Another child made nine atypical errors (e.g. affrication of clusters). Two made delayed errors, one with an occasional interdental lisp.

The errors made by the one child with a lateral lisp for /s, z/ in all phonetic contexts remained unchanged.

4 Do qualitative analyses of speech at four years predict outcome at seven years?

Data in Table 3 compare speech characteristics at age four for children who resolved and those with persistent speech errors three years later, by subgroup.

For children with delayed phonological development at four years, the 44 children who later resolved did not differ from the 22 who had persistent delay (see Table 3).

Children consistently making atypical errors who resolved, had fewer atypical error patterns and made fewer atypical errors overall at four years than children with persistent speech impairment at seven years. This was confirmed by a two-tailed unpaired t-test comparing number of atypical errors (t (df₂₀) = 2.152, p < 0.05).

Data from children making inconsistent errors, none of whom resolved, also indicate a high number of atypical errors at four years.

Table 3.

Comparing resolved and persistent speech impairment: Mean number (sd) and range of delayed and atypical error patterns and atypical errors by type of speech errors at four years.*

	Resolved	Persistent
Delayed at four years (n = 66)
Delayed error patterns	2.4 (1.1) 1–4	2.6 (1.6) 1–6
Number of atypical errors	3.2 (2.1) 0–9	3.3 (2.4) 0–10
Developmental error patterns used: fronting velars/fricatives, stopping, cluster reduction, weak syllable deletion, prevocalic voicing, and postvocalic devoicing:
Atypical phonology at four years (n = 26): consistently atypical n = 22; inconsistent n = 4
Consistent atypical: delayed error patterns Consistent atypical: atypical error patterns	2.1 (1.3) 0–5 1 (0)	1.8 (2.1) 0–3 1.5 (0.7) 1–2
Consistent atypical number of atypical errors	10 (3.2) 5–16	17.3 (10.4) 4–47
Atypical error patterns used: favourite sound (/h/, /n/ /j/), initial consonant deletion, backing, affrication/frication of clusters, reduplication, glides/z/ →[n], syllable initial velars → [h], all fricatives marked by [s]/ [Ɵ], /f/ →[w], affricates →[j], all final consonants except nasals delete.
Inconsistent: % inconsistency**	None resolved
Inconsistent: % inconsistency**	55% (14.4) 44–76%
Inconsistent: number of atypical errors	30 (17.5) 15–54

Notes. * One child with a lateral lisp not included. ** Diagnostic criterion for inconsistency: 40%.

5 Effect of intervention

Eleven children (12%) received some intervention from a speech pathologist between the four and seven years. This was determined by pre-planned telephone interview at the time of the seven-year-old assessment. No information was collected about the amount, timing or type of intervention received. Table 4 shows that five of the children’s speech difficulties resolved. All had delayed phonological development, with one also making some articulation errors at four years. The one child with delay who did not resolve had a low PCC score and was using six non-age appropriate developmental error patterns at four years. The children making atypical errors who did not resolve, made some progress.

Table 4.

Characteristics of children receiving therapy.

At four years		At seven years
Diagnosis	PCC	Diagnosis	PCC
Consistent atypical	85	Articulation	88
Consistent atypical + articulation	79	Delay + articulation	82
Consistent atypical	57	Delay	92
Consistent atypical	82	Articulation	91
Inconsistent	72	Delay	91
Delay	75	Delay	89
Delay + articulation	85	Resolved	98
Delay	87	Resolved	97
Delay	86	Resolved	99
Delay	80	Resolved	99
Delay	85	Resolved	97
Mean (sd)	72 (9)		93 (6)

Notes. AD = articulation disorder; CPD = consistent phonological disorder; PCC = percent consonants correct.

IV Discussion

Most longitudinal population studies focus on epidemiology to identify markers of persistent speech impairment. In contrast, this report explored the number and type of speech errors at four and seven years of 93 children making non-age appropriate errors at four years of age on standardized assessments. Speech errors were categorized as: phonologically delayed (typical of younger children’s normative data); or, atypical (non-development error patterns like backing; inconsistent realization of the same lexical items; or, articulatory distortions). Sixty-seven percent of phonologically delayed children (with or without occasional articulatory distortions) had resolved by seven years compared to 35% of children making atypical speech errors. Those with delayed phonology, who did not resolve, continued to make delayed errors at seven years, although five acquired an occasional interdental production of /s, z/. Children who had consistently used atypical phonological error patterns at four years, were most likely to make delayed errors at seven years, if they had not resolved. No other children’s speech impairment resolved, those making inconsistent errors at four years had variable outcomes, and the one child with a lateral articulation of all /s, z/ sounds, remained unchanged. Children who made fewest atypical errors were more likely to resolve, irrespective of total number of errors or whether they received intervention. Early identification of persistent speech impairment might, then, best rely on speech symptomatology.

Two-thirds of children with delayed phonology at four years had resolved three years later. This finding is consistent with previous research. Wright (2014) reported that 69% of children identified as delayed at initial speech therapy assessment had resolved two years later, irrespective of whether they received intervention. Another study compared groups of 22 delayed and 15 disordered three-to-seven-year-old children on a waiting list for therapy (mean wait 12 months, sd 4.9) showed no differences in age, time on waiting list, PCC at initial assessment, language, phoneme isolation, non-word imitation or cognition measures (Dodd et al., 2000). Significant differences between groups were found for rhyme awareness and PCC at Assessment 2 and change in PCC between the two assessments, with delayed children performing better than those making atypical errors. Experimental investigation (for summary, see Dodd, 2014) suggests that delayed children usually perform within the normal range on standardized assessment measures of vocabulary, receptive language, literacy and executive function, though sometimes less well than typically developing control groups. In general, children with delay perform better than those making atypical phonological errors. Phonological delay might reflect slower cognitive development or limited language learning opportunities associated with low SES (Head Start, 2010).

Only 35% of children making atypical errors resolved over the three years between assessments. Other research indicates that children making atypical errors are more likely to develop literacy difficulties (Preston et al., 2013). The dual difficulties in spoken and written phonology observed in children making atypical speech errors have been attributed to an impaired ability to derive phonological constraints: the system of phoneme contrasts and how phones may be legally combined to form words for speech (Crosbie et al., 2009); and, phoneme-grapheme relationships for literacy (Leitão, and Fletcher, 2004).

Children with delayed phonological development might, then, respond positively to preschool /school intervention (Head Start Impact Study, 2010; McIntosh et al., 2007). In contrast, evidence suggests that children making atypical speech errors have specific deficits in phonological processing. Children consistently using atypical error patterns do poorly on tasks of executive function with deficits in rule abstraction and cognitive flexibility (Crosbie et al., 2009). Children with inconsistent speech errors, in the absence of apraxia, do poorly on tasks evaluating phonological assembly (Bradford and Dodd, 1996). Such deficits predict persistent speech difficulties.

While fewer disordered children resolved by seven years, there was a change in error types towards consistent use of developmental errors suggesting emerging phonological processing abilities. This finding is not unexpected, given the three year gap between assessments. Possible explanations include exposure to early literacy teaching, better speech models/feedback at preschool/school and peers providing greater motivation to achieve intelligibility. That is, some children with a history of atypical speech errors slowly acquire a better understanding of the constraints of the spoken phonological system over time. Those children making the fewest atypical errors were most likely resolve or make most positive change. Type of errors, as a measure of phonological processing capacity and change, was a better predictor of outcome than number of errors.

The articulation error data provided surprising findings. Only one child in the cohort had lateral production of /s, z/ and that remained unchanged at the second assessment, consistent with findings that distortions tend not to spontaneously resolve without intervention (Flipsen, 2015). In contrast, 14 children who also made phonological errors, articulated more than half of /s, z/ occurrences interdentally at four years. This way of identifying interdental lisps may be misleading. All speech errors of nine of these children spontaneously resolved. Another unexpected finding was that five delayed children who had accurate /s, z/ articulation at four years had acquired some interdental productions of /s, z/ by seven years. In all cases, phonetically accurate production of /s, z/ occurred in some phonetic contexts.

Bowen (2015) suggests that interdental lisps might be considered a developmental error since they are so common, although this has not been observed in normative data (Dodd et al., 2002). The current data indicated that 19% of children with delayed phonology at four years also produced /s, z/ interdentally in some phonetic contexts. The result fits with Rvachew and Andrew’s (2002: 196) conclusion: ‘position-dependent variation in the production of an unmastered phoneme was more likely than a consistent production pattern across syllable positions’ in children described as being phonologically delayed. Future investigations might consider whether similar findings are better described as a dialectal difference, articulation disorder, phonetic variability associated with phonetic context, or a phonological error.

Only 11 of 93 children (12%) with impaired speech at four years received intervention. Parents were informed that impaired speech had been identified and assessment results released if requested. Telephone interview data provided no information about content or amount of intervention, precluding comment apart from noting how few children received speech therapy. Given the debate about need for speech and language intervention, future population studies might plan to include more detailed information about children, in context.

One problem with speech error types as a marker for persistent speech impairment is whether children at high risk can be identified for early intervention. Studies evaluating a standardized speech assessment for two-year olds (Toddler phonology test; McIntosh and Dodd, 2011) established its reliability and predictive validity (Claessen et al., 2017). In addition, Eadie et al. (2015) reported that speech and spoken vocabulary at 24 months forecast performance at four years on a standardized speech assessment.

The findings reported need replicating using research designs that better test speech error types as a useful marker for persistent speech impairment. The GFTA-2 samples few words of restricted phonological complexity and inconsistency was only assessed initially. Information gathered about intervention was inadequate. These shortcomings reflect the limitations of large cohort studies where in depth assessment is limited by large participant numbers.

V Implications

Nevertheless, the findings have two significant clinical implications. One concerns assessment practice for the clinical description of impaired speech. The most common metric for description is PCC. It ignores vowels and phonotactic constraints, makes no distinction between omission, substitution and distortion of speech sounds, and does not discriminate between age appropriate, delayed and developmentally atypical speech errors. Consequently, reliance on PCC risks misidentification of impaired speech. In contrast, clinical description of the types of speech errors made informs prognosis as well as intervention best practice (Dodd, 2014).

Another implication concerns prioritization for therapy. A randomized control trial (Broomfield and Dodd, 2005) reported that age of specific types of intervention affected outcome: phonological contrast therapy was most effective for consistent atypical errors at four years and for delayed errors patterns at five years. Core vocabulary therapy was most effective at age three for children making inconsistent errors. That RCT and the current findings suggest that preschool children making atypical speech errors might have priority for access to therapy.

The findings also contribute to our theoretical understanding of the nature of impaired speech. Waring and Knight (2013) argued that ‘speech sound disorder’ is a generic label for children with differing underlying deficits, linguistic symptomatology and natural history. The findings reported provide support for that position. The natural history (i.e. change in speech errors over time to eventual resolution) differs for subgroups of children with differing patterns of speech errors which have been shown to reflect distinct underlying deficits (Dodd, 2014). Future research studies might seek to investigate homogenous groups of children with speech difficulties, to extend our understanding of the nature of underlying deficits, their symptomatology, natural history and treatment.

VI Conclusions

A descriptive analysis compared the speech errors made at four and seven years by children initially identified as having phonological and/or articulation difficulties. Overall, 42% of children identified at four years had persistent speech difficulties at seven years of age. A significant number do not, then, ‘grow out of’ a speech disorder. Over the three years between assessments, two-thirds of children with delayed phonological development had spontaneously resolved, in contrast to only one-third of children who made errors atypical of normative data for speech development. The findings have clinical implications for assessment practice and service policy decisions as well as demonstrating the need for research that focuses on subgroups of speech impaired children, rather than a single heterogeneous population.

Footnotes

Authors’ Note

Katherine Brommeyer is now affiliated to Kids Plus, Geelong and Kelly Ng is now affiliated to Total Communication, Singapore.

Declaration of conflicting interest

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Bishop

Leonard

(2001) Foreword. In: Bishop

Leonard

(eds) Speech and language impairments in children: Causes, characteristics, intervention and outcome (pp. i–xiii). Oxford: Psychology Press.

Bowen

(2015) When /s/ and /z/ are hard to say. Available at: http://www.speech-language-therapy.com/index.php?option=com_content&view=article&id=86:lisp (accessed September 2017).

Bradford

Dodd

(1996) Do all speech-disordered children have motor deficits? Clinical Linguistic and Phonetics 10: 77–101.

Broomfield

Dodd

(2004) The nature of referred subtypes of primary speech disability. Child Language Teaching and Therapy 20: 135–151.

Broomfield

Dodd

(2005) Clinical effectiveness. In: Dodd

(ed) Differential diagnosis and treatment of children with speech disorder (pp. 211–30). London: Whurr.

Claessen

Beattie

Roberts

. (2017) Is two too early? Assessing toddlers’ phonology. Speech Language and Hearing 20: 91–101.

Crosbie

Holm

Dodd

(2009) Cognitive flexibility in children with and without speech disorder. Child Language Teaching and Therapy 25: 250–70.

Dodd

(1995) A procedure for classification of subgroups of speech disorder. In: Dodd

(ed) Differential diagnosis and treatment of children with speech disorder (pp. 49–64). London: Whurr.

Dodd

(2014) Differential diagnosis of pediatric speech sound disorder. Current Developmental Disorders Reports 1: 189–96.

10.

Dodd

Hua

Zhu

Shatford

(2000) Does speech disorder spontaneously resolve? In: Child Language Seminar 1999 Proceedings (pp. 1–20). London: City University Press.

11.

Dodd

Hua

Zhu

Crosbie

. (2002) Diagnostic Evaluation of Articulation and Phonology. London: Psychological Corporation.

12.

Eadie

Morgan

Ukoumunne

. (2015) Speech sound disorder at 4 years: Prevalence, comorbidities, and predictors in a community cohort of children. Developmental Medicine and Child Neurology 57: 578–84.

13.

Flipsen

(2015) Emergence and prevalence of persistent and residual speech errors. Seminars in Speech and Language 36: 217–23.

14.

Goldman

Fristoe

(2000) Goldman–Fristoe test of articulation. London: Pearson.

15.

Head Start (2010) Head Start impact study: Final report. Washington, DC: US Department of Health and Human Services, Administration for Children and Families. Available at: https://www.acf.hhs.gov/sites/default/files/opre/hs_impact_study_final.pdf (accessed September 2017).

16.

Hesketh

(2004) Early literacy achievement of children with a history of speech problems. International Journal of Language and Communication Disorders 39:453–68.

17.

Hayiou-Thomas

Carroll

Leavett

Hulme

Snowling

(2016) When does speech sound disorder matter for literacy? The role of disordered speech errors, co-occurring language impairment and family risk of dyslexia. Journal of Child Psychology and Psychiatry 58: 197–205.

18.

Holm

Farrier

Dodd

(2007) The phonological awareness, reading accuracy and spelling ability of children with inconsistent phonological disorder. International Journal of Language and Communication Disorders 42: 467–86.

19.

Law

Dennis

Charlton

(2017) Speech and language therapy interventions for children with primary speech and/or language disorders. Cochrane Database of Systematic Reviews: 1. Electronic publication. Published online: 9 January 2017 (doi: 10.1002/14651858.CD012490).

20.

Leahy

Dodd

(1987) The development of disordered phonology: A case study. Language and Cognitive Processes 2: 115–132.

21.

Leitão

Fletcher

(2004) Literacy outcomes for students with speech impairment: Long-term follow up. International Journal of Language and Communication Disorders 39: 245–56.

22.

McIntosh

Dodd

(2011) Toddler phonology test. London: Pearson.

23.

McIntosh

Thomas

Crosbie

. (2007) Enhancing the phonological awareness and language skills of socially disadvantaged preschoolers. Child Language Teaching and Therapy 23: 267–86.

24.

McLeod

(2015) The contribution of the ICF-CY to working with children with SSD. In: Bowen

(ed) Children’s speech sound disorders 2nd edition. Oxford: Wiley, 5–7.

25.

Morgan

Ttofari-Eecen

Brommeyer

. (2017) Who to refer to speech therapy versus who to ‘watch and wait’? The Journal of Pediatrics 185: 200–4.

26.

Murphy

Faulkner

Farley

(2014) The behaviour of young children with social communication disorders during dyadic interaction with peers. Journal of Abnormal Child Psychology 42: 277–89.

27.

NIDCD (2016) Quick statistics about voice, speech and language. Available at: https://www.nidcd.nih.gov/health/statistics/quick-statistics-voice-speech-language (accessed September 2017).

28.

Preston

Hull

Edwards

(2013) Preschool speech error patterns predict articulation and phonological awareness outcomes in children with histories of speech sound disorders. American Journal of Speech-Language Pathology 22: 173–84.

29.

Reilly

Bavin

Bretherton

. (2009) The Early Language in Victoria study (ELVS). A prospective, longitudinal study of communication skills and expressive vocabulary development at 8, 12 and 24 months. International Journal of Speech-Language Pathology 11: 344–57.

30.

Rvachew

Andrews

(2002) The influence of syllable position on children’s production of consonants. Clinical Linguistics and Phonetics 16: 183–98.

31.

Shriberg

Fourakis

Hall

. (2010) Extensions to the speech disorders classification system (SDCS). Clinical Linguistics and Phonetics 24: 795–824.

32.

Stackhouse

Wells

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

33.

Waring

Knight

(2013) How should children with speech sound disorders be classified? A review and critical evaluation of current classification systems. International Journal of Communication Disorders 8: 25–40.

34.

Wren

Miller

Peters

. (2016) Prevalence and predictors of persistent speech sound disorder at eight years old: Findings from a population cohort study. Journal of Speech Language and Hearing Research 59: 647–673.

35.

Wright

(2014) To treat or not to treat: Which children with SSD get better without help? Unpublished paper presented at the Speech Pathology Australia Conference, May, Melbourne, Australia.