Developmental Phonological Disorders III: Long Term Speech ...€¦ · Developmental Phonological Disorders III: Long Term Speech-Sound Normalization Lawrence D. Shriberg Frederic

Journal of Speech and Hearing Research, Volume 37, 1151-1177, October 1994

Developmental PhonologicalDisorders III: Long TermSpeech-Sound Normalization

Lawrence D. ShribergFrederic A. GruberJoan Kwiatkowski

University of Wisconsin-Madison

Prior articles in this series provide a descriptive profile of 178 children with developmentalphonological disorders (Shriberg & Kwiatkowski, 1994) and predictive correlates of short-termspeech-sound normalization in 54 children (Shriberg, Kwiatkowski, & Gruber, 1994). Thepresent article reports findings from a study of 10 children with developmental phonologicaldisorders whose progress was followed at least once yearly for 7 years. Analyses characterizethe sequence, rates, and error patterns of long-term speech-sound normalization in relation todevelopmental perspectives on the nature of children's phonological disorders. Findings areinterpreted to support the hypothesis of a critical period for speech-sound development, withlong-term normalization of significant speech delay reaching a chronological age boundary atapproximately 8.5 years.

KEY WORDS: phonology, disorders, prosody, development, prediction

Speech-sound normalization in developmental phonological disorders has beendefined as the behaviors and processes by which speech becomes normallyarticulate over time (Shriberg et al., 1994). Two periods for the achievement ofspeech-sound normalization with or without intervention have been proposed: short-term normalization, in which normally articulate speech is achieved by 6 years of age,and long-term normalization, in which normally articulate speech is achieved anytimeafter 6 years. Short-term and long-term speech-sound normalization are assumed toinvolve both similar and unique linguistic processes, with long-term normalizationassociated with additional academic and psychosocial correlates and consequences.That is, children whose speech errors persist after 6 years of age are likely to differ intheoretically and clinically important ways from children whose speech errors havenormalized by 6 years. A prior article in this series addressed prediction issues in theshort-term speech-sound normalization of a group of 54 children with developmentalphonological disorders (Shriberg et al., 1994). The purpose of the present paper is toconsider in some detail the long-term speech-sound normalization profiles of 10children with significant developmental phonological disorders whose progress wasfollowed for 7 years.

Research in Long-Term Speech-Sound Normalization

The following review is organized to provide information about the sequence, rate,and error patterns observed in normal speech acquisition and in normalization ofdelayed speech. Specifically, (a) Do children follow a similar sequence in acquiringthe sounds of their language, and, if so, does this sequence also characterize thelong-term normalization of a developmental phonological disorder? (b) Do the rates ofnormalization for children with speech disorders parallel the rates of normal speech-

0 1994, American Speech-Language-Hearing Association 1151 0022-4685/94/3705-1151

1152 Journal of Speech and Hearing Research

sound acquisition, albeit several years later? and (c) Do theerror patterns of children with speech disorders match theerror patterns observed within the notably brief period inwhich most children acquire articulate speech without diffi-culty?

The Sequence of Speech-Sound Normalization

The concept of a developmental speech disorder is pred-icated on the assumption that there is a normal sequencethat children follow as they acquire the consonants andvowels-diphthongs of their ambient language. This assump-tion follows from the observation that, within each language,children produce some sounds earlier than others and thatsounds observed to be acquired early occur with greaterfrequency in the phonetic inventories of the world's lan-guages. Correspondingly, those sounds observed to beacquired late occur with lower frequency in the world'slanguages (Jakobson, 1941/1968; Locke, 1983; Slobin,1977). From these observations, phonological acquisitionsequences are claimed to reflect outcomes of a variety ofmechanisms such as perceptual development (Winitz, 1969,1975), ease of articulation (Locke, 1983, p. 185), or the lessontologically committed markedness (Trubetzkoy, 1958/1969; cf., Chomsky & Halle, 1968). Information on thesequence of speech-sound mastery in normal and disor-dered phonology is found within each of the three majormethodological periods in child language study describedmost recently by Ingram (1989): the period of diary studies(1876-1926), the period of large sample studies (1926-1957), and the period of longitudinal studies (1957-present).

Diary studies. Diary studies describing the sequence ofboth normal phonological development (e.g., Albright & Al-bright, 1956; Bateman, 1916; Chamberlain & Chamberlain,1904, 1905; Gr6goire, 1937; Hills, 1914; Holmes, 1927;Humphreys, 1880; Jegi, 1901; Leopold, 1947; Lewis, 1936,1951; Menn, 1971; Moskowitz, 1970; Nice, 1917; Pollock,1878; Smith, 1973; Velten, 1943; Weir, 1962) and disorderedphonological development (e.g., Applegate, 1961; Cross,1950; Edwards & Bernhardt, 1973; Haas, 1963; Hinckley,1915; Lorentz, 1974) are characterized by major differencesin methodologies and theoretical investments. Reviews byEdwards and Shriberg (1983), Ferguson (1973), and Ingram(1989) emphasize the rich diversity and individual variabilityfound in the order of speech-sound acquisition, which aretypically described from a system-internal developmentalperspective rather than against an adult standard. It is notsurprising that a primary conclusion gleaned from the infor-mation in these studies is that phonological development inchildren is internally consistent, but externally variable (In-gram, 1989). For example, in an examination of the acquisi-tion sequences of fricatives and affricates implied by Jakob-son's (1941/1968) laws of irreversible solidarity, considerablevariability from the predicted universal sequence was ob-served across children (Ferguson, 1973; Ingram, 1989).

Cross-sectional studies. The large-scale cross-sectionalstudies constituting Ingram's second period of languagestudies include those providing age-of-acquisition data (e.g.,ArIt & Goodban, 1976; Bricker, 1967; Olmsted, 1971; Poole,

1934; Prather, Hedrick, & Kern, 1975; Smit, Hand, Freilinger,Bernthal, & Bird, 1990; Templin, 1957; and Wellman, Case,Mengert, & Bradbury, 1931; cf. reviews by Bernthal & Bank-son, 1993; Edwards & Shriberg, 1983; Ingram, 1989; Sander,1972; and Smit, 1986). The normative studies are based onjudgments of correct/incorrect phonological realization ac-cording to an adult standard, with considerable variationamong studies as to how this adult standard is defined,applied, and reported. For these and other reasons, studieswithin this period are not directly comparable, nor do theyyield a consistent phonological acquisition sequence.

There is ample evidence for a stable sequence of speech-sound mastery at levels superordinate to the phoneme.Menyuk (1968) constructed feature hierarchies for both nor-mal children and children with articulation disorders, conclud-ing that these orders represent the sequence of acquisitionsand substitutions. Singh, Hayden, and Toombs (1981) con-ducted a cross-sectional study of 1,077 children receivingspeech treatment, using a somewhat different set of featuresbut arriving at a similar hierarchy. Since Chomsky and Halle's(1968) adoption of markedness as both a representation ofthe order of speech-sound acquisition and the likelihood of asound occurring in a language inventory, it has been evidentthat the concept may apply to error frequency as well. Forexample, Toombs, Singh, and Hayden (1981) constructed amarkedness matrix for seven nontraditional features of En-glish consonant phonemes and then ranked the phonemesby the number of marked features for each. Applying theirmarkedness metric to the articulatory substitutions of 801children evidencing articulation disorders, they concludedthere was a definite hierarchy in which more marked soundsare replaced by less marked sounds. More recently,Dinnsen, Chin, and Elbert (1992) and Dinnsen, Chin, Elbert,and Powell (1989) have proposed that feature-level pro-cesses describe the pattern of growth for phonetic invento-ries. Based on analyses of cross-sectional developmentalliterature and children's progress in speech treatment, theseauthors propose a five-stage developmental progressionbeginning with the major class distinctions [consonantal][sonorant] [syllabic], which are required for [voice]; which inturn is required for the [continuant] [delayed release] mannerdistinction; which is required for [nasal]; and finally [strident][lateral]. These levels yield a limited hierarchy of possiblephonetic inventories that Dinnsen and colleagues claimcharacterizes both normal and delayed phonological devel-opment, whether acquired naturally or through intervention.

Longitudinal studies. The most relevant data addressingthe sequence of speech-sound normalization are found inlongitudinal studies, which include prospective, retrospec-tive, and follow-up studies. The latter two designs are notusually considered appropriate for establishing normalizationsequences because of the many methodological difficulties inassembling a phonological sequence from the number andtypes of data points available. Longitudinal studies of normalchildren (e.g., Dyson & Paden, 1983; Ferguson & Farwell,1975; Hoffman, Schuckers, & Daniloff, 1980; Klein, 1985;Leonard, Mesalam, & Newhoff, 1980; Rockman, Elbert, &Saltzman, 1979; Shibamoto & Olmsted, 1978; Stoel-Gam-mon & Cooper, 1984; Vihman & Greenlee, 1987) have allemphasized between-subject differences. However, in the

37 1151-1177 October 1994

Shriberg et aL: Long-Term Normalization in Developmental Phonological Disorders 1153

discussion of a study of 19 boys and 15 girls assessed at3-month intervals from the age of 15 months to 2 years,Stoel-Gammon (1985) noted that the individual differencesamong subjects were relatively minor. Stoel-Gammon at-tributes this homogeneity to the age of the subjects, theirstage of vocabulary acquisition, and the fact that productionswere not related to an adult standard.

Among the more widely cited prospective studies describ-ing the speech-sound normalization of speech-disorderedchildren, reports by McDonald and Shine (1969) and byDiedrich and Bangert (1980) were concerned with a re-stricted number of error sounds, and a study by Weiner andWacker (1982) provides limited phonetic data on 10 speech-delayed compared to 10 speech-normal children. Only thelarge-scale prospective study by Sax (1972) provides longi-tudinal self-correction data on the sequence of speech-soundnormalization in children identified as having a speech dis-order. Sax's data were based on a sample of 266 male and269 female children who were on a waiting list for speechtreatment services in a school district near Detroit. Thechildren were tested at the beginning and end of their firstschool year and then yearly to the end of fourth or fifth grade.An articulation test was used in kindergarten through firstgrade, but in successive years responses were based onsentence imitation. The group-averaged sequence of nor-malization without intervention during this period was inter-preted as mirroring the developmental sequence of normalspeech-sound acquisition, supporting the concept of aspeech delay.

Summary. When viewed at the level of longitudinal subjectprofiles, the normative studies indicate that individual speechsounds cannot each be assigned a rank-ordered position ona sequence of developmental mastery because childrenacquiring speech normally do not progress through an ex-actly similar sequence of speech-sound learning. Rather, theso-called normal sequence of speech-sound acquisition is aprobabilistic ideal that is only grossly sequential at a higher-order featural level. Moreover, speech-sound acquisitiondoes not always continue in a positive direction. A number ofnormative studies report reversals in mastery levels, espe-cially for/s/ (Kenney & Prather, 1986; Poole, 1934; Prather etal., 1975; Smit et al., 1990; Templin, 1957), with Sax (1972)specifically noting the tendency of some children to adopt anerror variant for /sl and /zl for a short time after a period ofcorrect production. Such individual differences and inconsis-tencies are reflected in the typically large ranges and stan-dard deviation values included in the tabular and graphicrepresentations of normal speech acquisition. Lacking evi-dence to the contrary in the few studies of developmentalphonological disorders, considerable individual differences inthe sequence of each speech-disordered child's long-termspeech-sound normalization are expected in the prospective7-year study reported below.

Rate of Speech-Sound Normalization

The concept of rate of speech-sound normalization re-quires a measure of correctness by time, with a procedurealso needed to estimate correctness values between ob-

tained data points. Normative studies such as those citedpreviously have typically used a normalization criterion of75% or 90% of children producing a sound correctly at ageintervals of several months to a few years. From these datapoints, rate is roughly estimated by measuring the slope ofthe straight line between data points. Because the actualdevelopmental function may not be linear, such assumptionscan lead to considerable error in estimates of rates ofspeech-sound acquisition. In the most recent and compre-hensive of the cross-sectional normative studies, Smit et al.(1990) do not compute rates of acquisition because growthcurve modeling requires that data be collected longitudinally(cf. Wohlwill, 1973).

The most relevant methodological considerations and sub-stantive data on rates of speech-sound acquisition werereported by Burchinal and Appelbaum (1991). These authorsestimated individual developmental functions in a variety ofways for the total number of articulation errors (reported assimplification processes, cf. Stampe, 1973) recorded for 4330- to 96-month-old children studied by Roberts, Footo, andBurchinal (1988). In the staggered longitudinal study, articu-lation test data were collected from each child four to sixtimes. Five growth curve models-individual nonlinear, indi-vidual polynomial, population nonlinear, population polyno-mial, and prototypic-were estimated, and indices of theestimated growth curves were correlated with independentestimates of intelligibility. Results demonstrated that curveselection involves compromise between the amount of rateinformation gained and the ability of the curve to accuratelydescribe a variety of sound changes. Crucial to the selectionof a model was whether or not the underlying growth pro-cesses were viewed as similar or different among individualsubjects-that is, whether the articulatory differences ob-tained were viewed alternatively as error variance or truescore variance. Within the normally developing childrentested with the Goldman-Fristoe Test of Articulation (Gold-man & Fristoe, 1969), preschool-aged process dissolutionwas characterized by significant individual differences.

Summary. A number of estimation techniques for model-ing developmental functions are available (e.g., McArdle &Epstein, 1987; Nesselroade & Baltes, 1979; Nunnally, 1962;Rogosa, Brandt, & Zimowski, 1982; Wohlwill, 1973). How-ever, with the exception of the study by Burchinal andAppelbaum (1991), there are no statistical studies modelingthe rates of phonological development. Bernhardt (1990,1992) and Von Bremen (1990) suggest that, in a nonlinearphonology, higher-level features have a faster rate of acqui-sition and may also have precedence developmentally.These authors acknowledge that supportive data for suchassumptions are as yet unavailable (cf. Bernhardt, 1992,Figure 1 for an example of a feature hierarchy in nonlinearphonology). As suggested previously, normalization rates ona variety of putative phonological units should be informativein determining whether (a) so-called speech-delayed childrenare, in fact, delayed only in the temporal markers for onsetand normalization, or (b) their patterns of normalizationprovide counter evidence for the notion of simple delay (cf.Bishop & Edmundson, 1987; Locke, 1994; Wohlwill, 1973).This question will be addressed in the study to be reported.

1154 Joumal of Speech and Hearing Research

Error Patterns In Speech-Sound Normalization

Analyses of the types of errors children make when theymisarticulate speech sounds have long been central totheoretical perspectives on the nature of normal and disor-dered phonological acquisition. For example, Locke (1983)quotes Noble's (1888) observations that because childrenmake the same "mispronunciations" in a wide variety of theworld's languages, there must be "... some law as theirinciting cause .... " Jakobson (1941/1968) proposed his lawsand principles to account not only for developmental se-quence, but also for the types of replacement patterns. Todate, primary descriptive information on the error patterns inspeech-sound acquisition and normalization can be dividedinto studies based on speech sounds and those based on theconstructs of phonological processes.

Speech sounds. Few normative studies of phonologicaldevelopment based on speech sounds also report frequen-cies of error types. Templin (1957) summarized the threemajor error types (omission, substitution, and distortion) byage level; and Snow (1963) itemized frequencies of omis-sions, substitutions, and moderate and severe distortions foreach phoneme and provided frequencies for specific substi-tuted phonemes. Prins (1962) and Snow and Milisen (1954)proposed an error-pattern normalization sequence fromomissions to substitutions to distortions within and amongsounds. This normalization sequence appears to be quitestable across children, although not all speech-error targetsgo through a period reflecting each error type. For example,the normalization data from five children in treatment for /s/errors reported by Elbert and McReynolds (1978, 1979) aregenerally consistent with this sequence, but not all childrenexperienced each of the three error types. As indicated insuch fine-grained studies, word position interacts stronglywith error type, a factor accounted for in at least some of theso-called phonological processes (e.g., Final ConsonantDeletion) discussed below.

Phonological processes. Stampe (1969, 1973) andDonegan and Stampe (1979) claim that the process ofspeech-sound development proceeds by suppression ordissolution of certain natural processes that interfere withotherwise adult-like child phonological competence. Thesenatural processes, which partially capture context-sensitiverelationships in phonology, have been tabulated and se-quenced in the speech of children with developmental pho-nological disorders reflecting many languages and dialects(e.g., Bortolini & Leonard, 1991; Goldstein & Iglesias, 1991;Hodson & Paden, 1981; Iglesias & Anderson, 1993; Ingram,1974; Magnusson & Naucler, 1990; Ries, 1987; Shriberg,Kwiatkowski, Best, Hengst, & Terselic-Weber, 1986). Thefindings in Grunwell's (1982) study of normal process disso-lution from first word use to 4 years of age have beenreplicated in studies by several other investigators (Dyson &Paden, 1983; Khan & Lewis, 1986; Vihman & Greenlee,1987). In addition to the many descriptive studies, a numberof studies have attempted to understand variables associ-ated with the error patterns captured by phonological pro-cesses (e.g., Anderson & Smith, 1986; Dunn, 1982; Dunn &Davis, 1983; Paden, Novak, & Beiter, 1987; Shriberg &Smith, 1983; Smith & Stoel-Gammon, 1983). For example,

reduced rates of language development and frequent dele-tion of final consonants have been related to early frequentuse of reduplication (Fee & Ingram, 1982; Ferguson, Peizer,& Weeks, 1973; Ingram, 1989; Klein, 1981; Lahey, Flax, &Schlisselberg, 1985; Schwartz & Leonard, 1983; Schwartz,Leonard, Wilcox, & Folger, 1980). Phonological processeshave been widely adopted as the analytic units for descrip-tion and explanation of speech-sound acquisition and nor-malization; however, the many systems of process analysishave also been the source of considerable critique (cf.Bernthal & Bankson, 1993). For reasons reviewed elsewhere(Shriberg, 1991a, 1993), the study reported here uses alter-native methods to describe long-term speech-sound normal-ization.

Summary

A fairly broad literature review indicates that, although thegroup-averaged sequence of speech-sound mastery ap-pears to be essentially similar for speech-normal andspeech-involved children, there are significant differences inindividual sequences and error-type patterns and there areessentially no data on rates of normalization. Such findingsprovide only tentative support for the perspective that adevelopmental speech disorder actually reflects only a delayin the rate of speech-sound mastery. Rather, findings amongthe large number of studies of normal and disordered childphonology appear to confirm only that speech sounds differon some multidimensional metric reflecting complexity ofarticulatory-acoustic features. Clearly, unlike children withnondevelopmental speech disorders, who may articulatedevelopmentally earlier sounds less well than developmen-tally later sounds (e.g., difficulty with stop consonant articu-lation in the presence of velopharyngeal insufficiency ormotor-speech deficits), children with speech disorders ofcurrently unknown origin have difficulty with the same soundsas children acquiring speech normally. Moreover, the con-cept of a speech delay is not particularly informative inaccounting for certain error types termed "non-natural"(Shriberg & Kwiatkowski, 1980), "unusual" (Dodd & lacono,1989; Leonard, 1985), or "uncommon" (Smit, 1991). A goalof the present study is to explore these and related issues bydescribing the long-term normalization patterns of 10 chil-dren originally referred for intelligibility problems of unknownorigin.

Method

Subjects and Procedures

A longitudinal study was initiated in 1975 with 10 children(9 males, 1 female) who were referred to a preschoolprogram at the University of Wisconsin for intelligibility prob-lems of unknown origin. Subjects ranged in age from 3:8 to5:4 (years:months) at first testing and from 8:11 to 11:4 atfinal testing. Nine subjects were seen nine times each; onemale subject was tested only five times, after which his familyleft the city. Each child was tested at approximately 6-month

37 1151-1177 October 199

Shriberg et al.: Long-Term Normalization in Developmental Phonobgical Disorders 1155

2AGE (YEARS)

FIGURE 1. The 12 age sets derived from the ages at which each of the 10 subjects were retestedduring the 7-year longitudinal study. See text for grouping criteria.

intervals for five sessions and then at approximately yearlyintervals for four additional sessions.

Age sets and age levels. Because testing began atdifferent ages for the subjects, it was necessary to collapsethe individual session data into arbitrary time periods termedage sets. The criteria used to determine subject assignmentto age sets were (a) only one assessment session from eachchild per set, (b) a minimum of three subject assessmentsper set, and (c) an attempt to make sets of comparable size.Figure 1 shows the 12 age sets derived from the threecriteria. Each age set (Age Set 1 through Age Set 12)included assessment data from 3 to 10 children, with 8 (67%)of the age sets including scores from 8 (80%) or more of thechildren. To allow for the range of subjects per age set (3-10children) the median age in months of each age set, termedthe age level, was used as the age reference. Thus, asshown in Figure 1, Age Set 1 was assigned age level 3:11,and so forth. The mean difference between the resulting agelevels and the actual ages of subjects included in each setwas 2.6 months (SD = 1.7). The maximum age discrepancybetween age levels and the actual age of any one subjectwas 6 months, which occurred in one of the older age sets.The analyses used actual ages whenever possible and agesets/age levels elsewhere.

Procedures. The examiners for the first and secondsessions were five graduate students trained to administerthe procedures by one of the authors (JK). The third throughfinal sessions for all children were conducted by the sameauthor, who maintained communication with each child dur-ing the 7-year period of the study. Each assessment sessionwas simultaneously audiotaped on a Nagra audiotape re-corder and either an Audiotronics Model PVR-708 4" B/Wvideotape system or a Sony AVC-3400 3/4" B/W videotapesystem coupled to a JVC camera. All sessions included an

administration of the Photo Articulation Test (Pendergast,Dickey, Selmar, & Soder, 1969) and a conversational speechsample. The conversational speech samples were used toobtain data on language production (Miller, 1981). Theprotocol completed during the first or second session in-cluded a hearing screening by staff audiologists, with somechildren receiving threshold audiometry and tympanometryand an oral peripheral examination. Case history data, phy-sician records, parental reports, and school records werealso assembled and coded, using procedures described inthe two prior reports in this series (Shriberg & Kwiatkowski,1994; Shriberg et al., 1994).

Subject description. Table 1 is a summary of demo-graphic and severity of involvement information for eachsubject. These data were obtained from the first or secondassessment session, depending on which session yieldedthe richest information for speech, prosody-voice, and lan-guage analyses. This sample of children had more severephonological involvement compared to the demographicallysimilar children with developmental phonological disordersreported in several descriptive studies (Shriberg & Kwiat-kowski, 1982, 1994; Shriberg et al., 1986). As classified bythe Percentage of Consonants Correct (PCC) metric, threechildren had mild-moderate speech involvement, five hadmoderate-severe involvement, and two had severe involve-ment. All subjects were, at various times, enrolled in speechtreatment in the schools during the period of study, with mostalso receiving other special educational assistance. Their long-term normalization patterns were undoubtedly influenced byintervention, but there is no clear way to delineate specificeffects from this source of variance. Because they receivedservices differing in form and intensity during preschool andelementary years, the children in this sample are representativeof children in the current educational environment.

UJm

I-O

Ctco,


TABLE 1. Subject description at the onset of the longitudinal study.

Percentage PercentageAge In Intelllgibility vowels consonants PCC severity

Subject Gender months percentage correct correct (PCC) level

1 M 45 60.9 80.2 42.8 Severe2 M 47 37.5 80.3 52.2 Moderate-Severe3 M 49 70.8 88.8 62.0 Moderate-Severe4 F 56 69.5 88.5 49.8 Severe5 M 57 90.5 94.1 69.2 Mild-Moderate6 M 58 83.7 94.3 59.5 Moderate-Severe7 M 59 51.8 87.1 74.4 Mild-Moderate8 M 60 84.6 91.3 62.1 Moderate-Severe9 M 62 81.3 93.0 62.0 Moderate-Severe

10 M 64 73.4 90.9 68.3 Mild-Moderate

M 55.7 70.4 88.9 60.2SD 6.5 16.4 5.1 9.6

Phonetic transcription and reliability. The conversa-tional speech samples and all sounds in the articulation testswere transcribed by a team of two experienced transcription-ists using the narrow phonetic transcription system describedin Shriberg and Kent (1982) and the transcription teamconsensus procedures described in Shriberg, Kwiatkowski,and Hoffmann (1984). Procedures for glossing, segmenting,and formatting transcripts for computer-assisted linguisticanalysis have been described in prior reports (Shriberg,1986; Shriberg & Kwiatkowski, 1980; 1983).

Utterances from four randomly selected subjects wereused to assess the reliability of phonetic transcription. A totalof 78 utterances consisting of 231 words were randomlyselected from the conversational samples, and a total of 74words were randomly selected from the articulation testtranscripts. Percentage of intrateam agreement was calcu-lated by means of a computer program developed for a largerstudy of phonetic transcription reliability (Shriberg & Lof,1991; Shriberg & Olson, 1988). Intrateam agreement forbroad transcription of the conversational speech samplesranged from 63.6% to 100% (see Shriberg & Lof, Data SetB). Only the reliability estimates for //, , SI, and /6/ fellbelow 90%. Broad transcription reliability estimates for thearticulation test data ranged from 75% to 100%. Only tS/, /s/,and /f/ fell below 90%. These reliability percentages areconsidered acceptable for the primary data of the presentstudy, which uses information at the level of broad phonetictranscription.

Stability of the Conversational Speech Samples

The citation-form data to be presented are spontaneous(i.e., not imitative) responses to the Photo Articulation Test(Pendergast et al., 1969) for all 86 assessment sessions. Toensure that the 86 conversational speech samples were alsostructurally comparable to one another and to availablereference data for conversational speech samples, descrip-tive analyses were completed on all samples for (a) averagewords per utterance, (b) percentage of occurrence of each of10 intended word forms (e.g., CVC, CV, VC targets, etc.),and (c) percentage of occurrence of each of 23 target English

consonants (percentages were not calculated for the infre-quent /3/).

Words per utterance. The entries in Table 2 indicate thechange in average words per utterance during the period ofthe study and comparison data for 3- to 6-year-old childrenacquiring speech normally (Hoffmann, 1982). Average utter-ance length for the first five age sets was slightly less thanthe average utterance lengths from samples of childrenacquiring speech normally. From the first to the last session,average utterance length for the children with developmentalphonological disorders more than doubled.

Word forms. Data presented in Shriberg and Kwiatkowski(1983) and Shriberg et al. (1986) indicate that the percent-ages of intended word forms in continuous conversationalspeech samples are stable. The entries in Table 3 indicatethe distribution of word forms per utterance during the periodof the study and include comparison data for both childrenwith delayed and with normally developing speech. Thesimilarity of percentages and rank-ordering among sessionsand between these data and comparison sources supports

TABLE 2. Average words per utterance for the 10 childrenduring the period of study. Comparison data are from 72 3- to6-year-old children acquiring speech normally (Hoffmann,1982).

Words perutterance

Words per HoffmannAge utterance males (1982)

Set Level M SD M SD

1 3:11 2.4 .4 4.2 .92 4:5 3.1 .4 4.7 .53 4:9 3.3 1.0 4.0 .84 5:2 3.9 1.5 4.4 .75 5:8 4.8 .6 4.7 1.26 6:2 5.1 1.87 6:8 5.0 1.48 7:6 5.5 1.79 8:4 5.2 1.3

10 9:4 6.7 2.711 10:4 7.0 2.212 11:2 6.5 1.2

37 1151-1177 October 1994

Shriberg et al.: Long-Term Normalization in Developmental Phonological Disorders 1157

N st _ XCD( cD O 9CDtCM - - - DFt_

N - ) I. C C i N -0

- N t CI U) U) I' U) U) 0

-N,*Mwwr)cOw00

NCOO0WONN0NO

N r - - b 0 4 D N-

- N st CO) U) 1- CD U) 0O

4-N vm F 0 w o-

N N _ -

- N 't 10) U) U) CD r- 0O

NUwOr-wM0)

U) 0 N 0 0) U CI CD 0_

NO N - 0 0 N O

N Q N q q 0 q00

- N t CO) U) U) Ib U) 0 O)

CDJ 1D ) U) CN N-- s

- N 't CI U) <O Ib U) U) 0

Nd N- _ - U) N U) -0O

- N t CI U U) It- U 0 O)

NN-_- 0O

- N U ) U) U) Z U) U 00O U) CDN U) ON

N~~r.r

- N U)CI t U) I U) 0 U)

O) N N ' U) X -'00 s

C N - -

- N CI) t U) U) - U) U) 0

_nooawooouo

N 00)U) ) N O OO

- N ) t U) CO I- 0) 0) 0

CM -~

- IOtO)U)U)_U)OOU)

U) N ) N 0 U0 N - 0

N NM - -_ N CI * U) (D) 1 0O

e

UN)U)U)) r Orrj~cI) N.-.-cr

OON» )Dc00> 0)

C

c0

E

a

I

'i

C

C

c

U

3

.Ig

c

EC

I

e

am

S

a

S

S

E

C A

c

S

w

I

P

U

e I

c W0=

.

US

4

O

c

U

w)

N

-S

3 U

aa0

'I

a

M

a

M

acc

acc

a

M

a

M

a

U

a

U

U

C

0

U)

1z

|e

ie

M

e

e

e

e

e

a

e

le

le

CDj CO O: 6 S (D (D 6 U) 0) COi -: N IO 0) s 0N D_

___1: 11 IO 4 t U) 06 Xt N _ Ch 6 N,-_ _ N _ _~ O U)0U)ON )U)N-ou oo~~ o r---o

0o 0 U) ) 0 U) 0 N U) _ - CO -_ _ _ N _N N N N

N 0) U 0 _D 0) Nt0 WD .t U U 1

_c__ t _ N N N N a N

000000000000000000 U) U 0000

N Mr,-t rNw . r N NO N N

OOOOO0OOOOOOOOO NCO DNU).O

N_)CI)U) ) -dUU)d .- i)CI) ..d 000

000000000000000000000000

r ____N N N N N

00000000000000U)OSONOO

q cq q 7 'It LQ q q q q q q

~~~000rm~~~~O ~ c Om V~O O U O)O OR': O O I O O - N U) O OR 0 U) N 0

00000000000000000U00000

-Nr -0 rr CU 'CI-tOO-@e+O-ttN^XON____OOOO

000000000000000 00000000-N O N.-U)UNU).-__ 0_OOOO

ItN t N N N NltV- N

CO U)bNCI) CI) CI) CI)Nr ..-- 0000m000000000000000000000000

_NOFOW0*0XtS¢Nid0dX0O~t

rN0)O(DrrrN06o) (j.U.N~rt0~0r0~fl00)Ca)UNrs

U)0 _)________N-NNNN0O6

ONOOONdddde-^NO-*NNrddd_

000000U00000000000000000

U))U_)-____--NNN000000000000 0U)00)0U0U)OO

_ )U)N00I. .0 w N t w 0N_ w N

r ___NNNN N900 N0 I0 U)U) ) U C) 0 .- '00 N I)U)t U _NN0UWhUNNr -,-i__66660

U)U)oooooouooouonuouoUo,Uo

_-__------NNN N

OONOSOO@X@X+-NOOUNO~tOSO

S000000)000Ut i00Woooo

UNU0OOOOU)UO~U)~uOou)OOOONuNOU

-0 0_10NC _< D <C r <NONN N N) N N

q ,OOq OOOq'qq q P.'": )UI) .-.-. 0000

00000000000000000trH0i0

rN 0)C _ 1U)U 1U0 U) 10) U)0Zi ti i NN N st_ __ -_ -r__N N N N N

0@0000000000000000000000

_N 0i )0IiCO i0N i tiOo_

,,~ ~ ~ ____N

qoo C _O t 0)X 1 tt NN _t-< 1 0 F 0

0000000000000000000000_ N r i (C b C 00 NCf s ldiZt bCi0 _N sr~~~ _ -~ -r _ _ _N NN N N

Ot~r~t~O-Q#-te4Z~eS~d~ib

e dddd i~dd~,dd

,, ~ ~ ~ ~ ~~~~

C2

0

I?U

o

3o

I.C

a

C

i

I.

0

a

1t

a

5

C

S

!f

a

.3iI

i.

g.,

,c

c4Jo

acm

o e:

IE

-1

U

CD

Ix:

a

PI

w

0

t I:

C1

U

I

e

0

e

* h

3 G

S 1

V

U2

iS

0:a.

aiI

ac

e

C

00

0P0o

Z

I


the stability of the conversational speech samples. Specifi-cally, although the number of words per utterance increasedthroughout the study (Table 2), the conversational speechsamples maintained essentially similar and representativedistributions of word forms.

Intended consonants. The entries in Table 4 are thedistributions of intended consonants during the period ofstudy and comparison data for children with normal and withdelayed speech development. Summed over all 86 testsessions, the rank-order correlation of the percentage ofintended consonants in the present data with the means forthe comparison studies was .97. Discussion of factors under-lying the characteristic stability of word forms and intendedconsonant types in conversational speech is presented inShriberg (1982).

Results and Discussion

Findings are organized by the three-part framework usedto review the literature-information on the sequence, rate,and error patterns observed in long-term speech-soundnormalization. Subsections within each of these dependentvariables address effects associated with alternative modesof assessment (articulation tests, conversational speechsamples) and linguistic units (developmental sound classes,individual speech sounds, word position), using comparabledata sets from children with normal speech acquisition anddevelopmental phonological disorders. The goal of theseanalyses is to construct a detailed profile of long-termspeech-sound normalization in children identified in pre-school years as having a significant speech delay of un-known origin.

The Sequence of Long-Term Speech-SoundNormalization

This section begins with a description of individual differ-ences in the sequence of speech-sound normalization foreach subject. These data are followed by a series of group-level analyses of potential interactions of sequence withmode of sampling and word position. The primary question iswhether the temporal sequence of speech-sound normaliza-tion in boys and girls with phonological disorders is similar tothe sequence characterizing normal speech-sound acquisi-tion.

Individual sequences of speech-sound normalization.Table 5 is a summary of the rank-ordered, speech-soundnormalization data for the nine subjects who were tested ninetimes in 7 years (incomplete data for Subject 2 are excludedfrom this table). The entries in each row indicate the order ofspeech-sound normalization based on the earliest session inwhich each of the consonants normalized. The Appendixdescribes the criteria used to classify sounds as normalizedin the articulation test data and the conversational speechsamples of the present study and two comparison normativestudies. Vertical lines between sounds indicate ties at thoseranks (i.e., sounds that met the criteria for normalized at thesame assessment session). The order of entry for all tiedranks was determined by the sequence listed for the Early-8,

Middle-8, and Late-8 sounds in the speech profile format inShriberg (1993); see Figure 2 in the present paper for the24-consonant sequence. These rank-ordered data are pro-posed as the best description of the individual sequences ofspeech-sound normalization for each of the nine children.

Keeping in mind the approximately 6- to 12-month gapsbetween test sessions and the large number of tied ranks,entries in Table 5 indicate notable individual diversity inspeech-sound normalization sequences. No two children hadan exactly similar sequence of speech-sound normalizationbased on either articulation test responses or conversationalspeech samples. The greatest differences in sound se-quences within and between children occurred for the Early-8and Middle-8 sounds. The Late-8 sounds had greater rank-order stability across children. For all children in both sam-pling modes, reversals were most evident on Early-8 andMiddle-8 sounds.

Spearman Rho coefficients were computed from the datain Table 5 to provide quantitative estimates of the similarity ofthe sequences. The interest was in the absolute magnitude(i.e., R2 ) of coefficients, rather than in whether any coeffi-cients reached levels required for statistical significance. Amatrix of coefficients compared each of the nine normaliza-tion sequences in each mode with each other sequence. Forthe articulation test data, the 36 intersubject coefficients (i.e.,each child with every other child) ranged from .44 to .83 (M =.70; SD = .09), which when squared accounts for approxi-mately 19% to 69% of common variance in normalizationsequences. For the conversational speech data, the 36intersubject coefficients ranged from .42 to .88 (M = .67; SD= .12), which accounts for approximately 18% to 77% ofcommon variance. The nine intrasubject coefficients (i.e.,comparing each child's sequence as sampled by articulationtesting versus conversational speech sampling) ranged from.54 to .90 (M = .64; SD = .12), which accounts for approx-imately 29% to 81% of common variance. Subsequentanalysis by word position (see following section) indicatedthat the normalization sequence obtained by the two modeswas highly correlated for sounds in word-initial position(Spearman Rho = .91) and moderately correlated in word-medial (Spearman Rho = .46) and word-final (SpearmanRho = .49) positions.

Position of sound in the word and sampling mode.Table 6 is a group-level summary of the rank-ordered se-quence of speech-sound acquisition by word position andsampling mode. The procedures used to derive the individualdata in Table 5 were used to derive these grouped data inTable 6, with percentages reflecting the sequence averagedacross the nine children. Visual inspection of the eightsequences in Table 6 suggests that the group sequencediffers considerably by position of the sound within the wordand sampling mode.

Quantitative evaluation of the similarity among the eightsequences in Table 6 was estimated by the Spearman Rhocoefficients summarized in Table 7, which range from .01 to.91. Once again, word position and sampling mode areassociated with differences in the magnitudes (and signifi-cance levels) of the coefficients. Over all word positions(total) the correlation between sequences obtained in eachsampling mode is .51. However, when computed by word

37 1151-1177 October 1994


TABLE 5. Speech-sound normalization sequence for 9 of the 10 subjects tested In each sampling mode."

Articulation testing Conversational speech

Sub#: Rank 1 3 4 5 6 7 8 9 10 1 3 4 5 6 7 8 9 10

1 h m b

2 m n h

3 1* p m

4 n b pI I

5 w t j*

6 p d t

7 b k f*

8 t g tS

9 d* ' n

10 k h w*I

11 g 1 S*I I I

12 f w* d3

13 v j* d

14 O tS 1*

15 6 d3 k

16 j f g

17 I# 6 v

18 z# v s

19 r# 0 z

20 s# s 0

21 S# Z 6

22 3# 3 3

23 tS# r* I#

24 d3 I# r#

m

n

e*

w

Pp

b

nI

w

t

g

mI

h mI

m* pI I

J1* b

b h

w n

d* t

t p p k

d b k w*I I Ik f g g

I Ig h n f

f d j* dI I

6* k t 13

h j f* 0

tS v v* 6I I I I

d3* s* 6 s*

Sk Z* S* Z

j 0 d3* SI I Is 6 0 tS

v* tS S d3

o tS

z d3 z# v

3 r* I# 3

1# 3 r# I

r# I# 3# r*

n m* j mI I w W* f 1*l I Ib p z* j*l I It b h* p

d t m g

h d n v*I I

m k w h*

p n p d3*

f 3* b wI I

6 g v* b*I I

k f* g k*I I

g 6* f n

I I Iv* z* * * d

0 h t* f*I I I

S tS d SII I I I

tS d3 k t

d3 S 6* 8*

13 v* d3 t*

i i S s

s* s tS, ZI I

z* 3 1# 6'

3 O r# I#

I# I# s# r#

r# r# 3# 3#

m m m* j w*

j n* n* b* b

b* w w* d* f*I I I I Ih j* b* g* h

p p* d* n* mI I I

w* b* h* w nI I

tS* d* j* f* pI I I I

d3* k* p* h* g

f g* S* tS* OI

S* f* t* m*

g O* k* p dI I

n S 9g* d3 kI It h* v k v*I I I I

d* tS f 0* 0

v d3* 6* v s*I

k v 0 t# z*

0 *j z* 6# S*

s# t* s* s# tSI

z# 6 tS S# t*

6# s# r* z# 6*

91# Z# 3 J# 3

I# I# d3 I# d3

r# r# I# r# r#

3# 3# §# 3# 1#aVertical bar between sounds indicates tied rank-order.*Reversals occurred after the 75% criterion for articulation testing or the 90% criterion for conversational speech was reached.#Never reached the 75% or 90% criterion.

position the coefficients between the two sampling modesare highly correlated in word-initial position (.91) and onlymoderately correlated in word-medial (.46) and word-final(.49) positions. The correlation between sequences derivedfrom the total consonants and any one word position is highfor word-initial (articulation testing = .64, continuousspeech = .73) and word-medial (articulation testing = .69,continuous speech = .60). The correlation between se-quences derived from the total consonants and word-finalposition is only moderate in articulation testing (.46) andwholly uncorrelated in continuous speech sampling (.01).The latter finding is likely explained by the increased

frequency, variety, and instability of speech-sound errorsword-finally (Shriberg & Kwiatkowski, 1994), a thesis sup-ported by the higher number of reversals word-finally (seelater discussion).

Speech-sound normalization sequences compared tonormal speech acquisition. Table 8 and Table 9 providecomparisons of overall speech-sound normalization se-quences for subjects in the present study and subjects in twonormative studies. Table 8 includes the rank-ordered normal-ization sequences (mastery level = 75% correct; see Appen-dix) from the 997 boys and girls in Smit et al. (1990), the 72boys and girls described by Hoffmann (1982), and for 9 of the

n* m*

w nI Ij* w

b* ij*

d* p*I I

f* bI I

O* d*I I

h* g*

m f*

t* h*I I

tS* d3*

P S*Ik t,Ig kI IV V

1* t

S* s

d3 0

s Z*

z 3

6# 6#

I# I#

r# r#

3# O#


TABLE 6. Speech-sound normalization sequence by word po-sition and mode of speech sampling.a b

Initial Medial Finalposition position position Total

Rank AT CS AT CS AT CS AT CS

1 h m p* g* m m* h m*

2 m

3 n

4 w*

5 i

6 p

7 b

8 d

9 t

10 g

11 f

12 k

13 v*

14 tS

15 d3

16 I

17 6

18 s*

19 6

20 1

21

22 r#

23 (no)

24 (3)

Ij*I

P*

I

bIt*

e* m*

f* w*I

b b*

m r* nI I

h

n*

d*

tS*

w*

f

k

9

v

S*

I*

d3*

6#

s#

I#

n p*It 6*

6* n*I

d iI Iv r

w 3*I Ii h

k tS

g v*

0

S

tS

d3

Z*

d

6

0

d'~

s

I3 t

l# z#

r#

z+

(n),v.

(3)

0*I

v* m* j*

0* 0* w*

tS* p* b

p* b h

n* t n*I

P g n

b* d3 dI Id b f

Ig k w

I Iv 6* 6*I I0* S kI I

s* t g

z* f vI

S s# j

d3 z# SI

tS d# tS

l# 0# d3

r# r# 0I

6+ I# s*

3+ 3+ 3

(w) (w) z

3*

P

r# I# ( ) 0)

h+ r# (h) (h)

aVertical bar between sounds indicates tied rank order. bArticulationTesting (AT); Conversational Speech (CS).*A reversal below the 75% or 90% criterion occurred.#The sound never achieved criterion.+Insufficient data to rank.()Does not occur in this position.

10 speech-delayed children (excluding the incomplete datafrom Subject 2) in the present study. Smit et al. assessed 3-to 9-year-old children in a cross-sectional design using apicture articulation test devised for the study. Hoffmanntested 3- to 6-year-old children cross-sectionally using thecontinuous speech samples and the Photo Articulation Test(Pendergast et al., 1969). Both normative studies usedassessment procedures similar to the protocol used in the

TABLE 7. Spearnman-Rho correlation coefficients (corrected forties) corresponding to speech-sound normalization sequencesby word position and mode of speech sampling.

Initial Medial Finalposition position position Total

CS AT CS AT CS AT CS AT

Initial position CS -AT .91t -

Medial position CS .41* .45* -AT .42* .30 .46* -

Final position CS .26 .14 .22 .47* -AT .34 .25 .07 .72t .49* -

Total CS .73t .82t .60t .20 .01 .08 -AT .63t .64t .61t .69t .16 .46* .51* -

*p < .05.tp < .01.

longitudinal study of the 10 speech-disordered children.Entries in Table 8 reflect responses to the two articulationtests that were scored only by target position for the Smit etal. and Hoffmann studies, but included transcription of allconsonant sounds for the present study. Table 9 is a sum-mary of Spearman Rho correlation coefficients for eachrelevant pairwise comparison between acquisition and nor-malization sequences. The information in Tables 8 and 9,which to accommodate the Smit et al. data are based only onarticulation test responses, provides comparisons of normalacquisition and normalization sequences at the levels ofgender and word position.

Gender. As the current study includes 9 boys and 1 girl, aninitial question is whether gender is an important indepen-dent variable in generalizations about the sequence of nor-mal speech acquisition and speech-sound normalization.Although the question cannot be posed directly within thecurrent data set of speech-disordered children, gender infor-mation from the two normal acquisition data sets studies canat least suggest interpretative guidance.

There are seven Spearman Rho coefficients aligned diag-onally in Table 9 that express the similarity in sequence ofnormal speech acquisition for boys and girls within the Smitet al. (1990) and the Hoffmann (1982) studies. The threecoefficients for the Smit et al. data comparing the boys'acquisition sequences to the girls' acquisition sequences inword-initial (.89), word-final (.98), and overall (total) (.94)positions are strong positive. The four coefficients for theHoffmann data, which include word-initial (.95), word-medial(.96), word-final (.83), and total (.95), are also strong positive.These coefficients suggest that the sequences of speechacquisition observed in boys and girls in two studies usingdifferent articulation tests are highly similar in all word posi-tions. These data appear to provide the needed support forcollapsing gender in all subsequent analyses of the 10speech-disordered children.

Word position. A finding reported above was of low tomoderate correlation between the speech-sound normaliza-tion sequences of 9 of the 10 speech-disordered children asassessed at the level of the position of the sound in the word.

37 1151-1177 October 1994

t*

f,

k*

d*

z#

Shriberg et al.: Long-Term Normalization in Developmental Phonlogical Disorders 1161

TABLE 8. Speech-sound acquisition sequence reported by Smit et al. (1990) and Hoffmann (1982) compared to the speech-soundnormalization sequence In the present study as evoked by citation form articulation testing for all three studies. For each study,ranking reflects a 75% correct criterion. The age ranges of subjects in the Smit study are from 3:0 to 9:0 years; for Hoffmann, 3:0 to6:0 years; and, for the current study 3:9 to 11:4 years. The Smit and the Hoffmann studies were cross-sectlonal.8

Smit et al. (1990) Hoffmann (1982) Present studyN = 514 males, 483 females N = 36 males, 36 females N = 9 males, 1 female

Initial Final Total Initial Medial Final Total Initial Medial Final Total

Rank M F M F M F M F M F M F M F Both Both Both Both

1 m m m m m mI I I I I I

2 n n n n n n

3 h h p p p p

4 w w b b b b

5 j p t t t t

6 p b d d d d

7 b t k k k k

8 t d g g g g

9 d k f f f f

10 k g tS s* v s*

11 g f v v tS vI I I I

12 f s* d3 S d3 S

13 v j s* t s* * t

14 d3 v r* S d3

15 S 6 r d3 r 1*

16 tS S 0 I* 0 0

17 s* tS O* Z* z zI I

18 6 d3 z g* I rI I

19 r z I 0 0+ 0+

20 0 0 6+ 6+ h+ h+

21 z 1 3+ 3+ w+ w+

m m m m m m m m h p* m* h

n nI I

w wI I

P P

b b

t t

d d

k* k

g* g

f f

h 0

v h

0 vI I

6* 6*

I I In n nI I I

0 0* 0I I I

w w pI I II bi j bI Ip p t

b b d

t t k

d* d g

k* k f

g g v

f f 6

v v 6*

6* 6 r*

0 O* tS#

I In n nI I

I Ip w wI Ib j iI It p p

d b b

k t t

g* d dI If k* k

v g g

6 f fI I I

6* h vI1* v hI

r* O OI

I* z* I z* d3# d3# 6 6*I I I

r* S* r# I* I# z# I# z*

tS# I S# r s# S# r# I

s# r z# d3# z# 3# tS# r

z# tS# tS# S# S# tS# z# d3#

S# d3 # d3# 3# 3# s# S# S#

22 I r (w) (w) j+ j+ d3# s# s# tS# (w) (w) s# 3#

23 (8) (O) (j) (j) 6+ 6+ (§) (0) 3# s# (j) (j) d3# tS#

m

n*

W*

I

pP

b

d

0*

f*

b

m

n

t

*

t* mI I

f* 0*I

k* p

b* b

n

0*

pI

t d d f

g v g w

f w d3 6*I

k j v kI I I I

v* k O* gI I ItS g s* v

d3 O z* jI I I

S s S S

0 S tS tS

s* tS I# d3

6 d3 r# 0I II Z* 6+ s*

z# 3 3+ 3

r# I* (w) z

(0) r# (j) I#

n

d

24 (3) (3) (h) (h) 3+ 3+ (3) (3) h+ h+ (h) (h) 3# s# (3) h+ (h) r#aVertical bar between sounds indicates tied rank-order.*Reversals occurred after the 75% criterion was reached.#Never reached the 75% criterion.+Insufficient data to rank.()Does not occur in this position.

The coefficients in the last four rows in Table 9 assess the speech-disordered children are remarkably similar to those insimilarities among sequences of normal acquisition and children acquiring speech normally. Comparing the corn-sequences of speech-sound normalization for each of the bined-gender (Both) sequence from the current study withthree word positions. Independent replications are provided those in the two normative studies respectively, coefficientsby the two levels of gender within the two normative studies. are highly positive for word-initial (male = .92, .85; female =As shown in Table 9, the normalization sequences for the .80, .81), word-medial (Hoffmann only: male = .71; female =


TABLE 9. Spearman-Rho correlation coefficients (adjusted for ties) for the rank-order acquisition and normallzamon data In Table 8.

Smit Hoffmann Current Study

Initial Final Total Initial Medial Final Total Initial Medial Final Total

M F M F M F M F M F M F M F Both Both Both Both

Smit Initial M -F .89 -

Final M .61 .63 -F .59 .64 .98 -

Total M .60 .63 .98 .96 -F .56 .69 .95 .96 .94 -

Hoffmann Initial M .90 .80 .51 .51 .49 .46 -F .82 .71 .46 .45 .44 .40 .95 -

Medial M .58 .43 .52 .52 .50 .43 .68 .65 -F .50 .37 .41 .40 .39 .32 .63 .62 .96 -

Final M .33 .34 .80 .78 .77 .71 .39 .35 .71 .68 -F .43 .26 .63 .62 .60 .53 .51 .49 .85 .83 .83 -

Total M .74 .61 .44 .44 .40 .35 .86 .81 .81 .76 .51 .65 -F .71 .54 .40 .38 .38 .30 .81 .79 .84 .81 .53 .70 .95 -

Current Initial Both .92 .80 .45 .42 .45 .41 .85 .81 .46 .39 .17 .29 .72 .68 -Medial Both .36 .39 .55 .51 .57 .51 .36 .34 .71 .72 .74 .69 .54 .54 .32 -Final Both .44 .49 .86 .84 .87 .84 .38 .39 .62 .55 .83 .71 .48 .47 .28 .72 -Total Both .62 .61 .41 .37 .41 .34 .60 .55 .53 .53 .48 .43 .77 .76 .67 .69 .48 -

.72), and word-final (male = .86, .83; female = .84, .71)positions. Once again, as shown in the bottom row of Table9, correlations among the three studies were generally lowerwhen sampled across the total of all word positions. Becausethe Hoffmann (1982) normative study is closer in methodol-ogy to the current study than the Smit et al. (1990) study, thehighly positive correlations provided by these pair-wise com-parisons are particularly supportive of the similarity in normalspeech-sound acquisition and speech-sound normalizationsequences for English consonants. As calculated at the levelof word position in Table 9, normal speech-sound acquisitionand speech-sound normalization share as much as approx-imately 85% common variance word-initially, 52% word-medially, and 74% word-finally.

Comparison to representative cross-sectional studies. Toexamine the methodological and theoretical generality ofthese findings, the speech-sound normalization sequencesfrom the present data were also compared to widely citedcross-sectional data from normally developing children andchildren with developmental phonological disorders.

Table 10 is a summary of normalization sequences for fivewidely cited, but methodologically diverse, cross-sectionalstudies of normal speech acquisition (cf. Hester, Godbold,Lee, & Stephens, 1984). Each of the five studies useddifferent criteria for subject selection, data collection, tran-scription, scoring, analysis, and reporting (cf. Smit, 1986).Importantly, however, each study used a mastery criterion forspeech-sound normalization of 75% of subjects articulatingthe sound correctly. Sequences are averaged across wordposition and gender. Arlt and Goodban (1976) used elicitedimitation to secure responses; all other studies summarizedin Table 10 relied primarily on spontaneous responses toarticulation tests, supplemented with imitation only whennecessary. Although a number of individual sound se-quences appear to be similar to the sequences presented inTable 8, overall agreement among studies is high only atlevels that aggregate above the level of the phoneme.

At the class level the sequences in Tables 8 and 10 arealso consistent with the Early-8, Middle-8, and Late-8 devel-opmental classes derived from cross-sectional data on chil-dren with developmental phonological disorders (cf. Shri-berg, 1991b, 1993). Figure 2 compares the presentnormalization sequence data to findings reported in Shriberg(1993). Sixty-four 3- to 6-year-old children with speechdelays (left ordinate) are compared to the present rank-ordered longitudinal data (right ordinate) summed acrosschildren and word position. Both data sets are from conver-sational speech samples, collected and analyzed with nearlyidentical procedures and personnel. There is clearly closecorrespondence between the percentage correct data asscaled on the left axis and the sequence of speech-soundnormalization rank-ordered from 1-24 on the right axis. TheSpearman Rho coefficient of .88 between sequences indi-cates 78% common variance. The greatest discrepanciesbetween the sequences in the two databases were on threeMiddle-8 sounds, /t/, //, and /k/.

Summary. The findings reviewed to this point indicate thatnormalization sequences at the level of individual sounds arecharacterized by considerable individual differences acrosschildren, sampling mode, and word-position. Notwithstandingthe variance associated with such factors, the descriptivedata and associated correlational findings support a conclu-sion that the group-averaged sequence of speech-soundnormalization is generally similar to sequences documentedin two recent and comparable studies of normal speech-sound acquisition, as well as four other widely cited studies.Characterized at the higher-order level of phonetic mannerfeatures, the sequential order of nasals, glides, stops, frica-tives/affricates, and liquids is generally stable, although notinvariant (see Table 5). Characterized according to the threedevelopmental sound classes used in the speech profilesapproach, which was derived from cross-sectional data onspeech-disordered children, the Late-8 sound class generallyretains its last-ranked position in both the normal acquisition

37 1151-1177 October 1994

Shriberg et aL: Long-Term Nomalizauon in Developmental Phonological Disorders 1163

TABLE 10. Speech-sound normalization sequences averagedacross gender and word position for five major representativecross-sectional studies that use a 75% correct criterion. Thestudies are those of Wellman at al. (1931) for ages 2:0-6:0, N =204; Templln (1957) for ages 3:0-8:0, N = 480; Prather et al.(1975) for ages 2:0-4:0, N = 147; Arlt and Goodban (1976) forages 3:0-5:6, N = 240; and Smit et al. (1990) for ages 3:0-9:0, N= 997. All studies used picture articulation testing except forArlt and Goodban (1976), which used imitation techniques.

Study

Rank Wellman Templln Prather Arlt Smit

1 m

2 n

3 w

4 b

5 h

6 f

7 p

8 j

9 k

10 g

11 1

12 d

13 t

14 v

15 tS

16 s

17 z

18 r

19 d3

20 3

21 +

22 S+

23 O+

24 6+

+Sound not tested or data ii

and normalization data.quence for the Early-8classes is less stable in rfor normal speech-soundof speech-sound normaldered speech is not expeRather, the per-child seqi

m

nI

m m mI In n n

tion presumably reflects the multidimensional effects of per-ceptual, articulatory, distributional, and morphophonemicproperties of the speech sounds of a language in interactionwith each child's individual communicative needs and inter-vention histories.

Ages and Rates of Long-Term Speech-SoundNormalization

The goals of the second series of analyses were toestimate ages associated with the sequence of speech-sound normalization described above and to provide infor-mation on periods of more rapid versus less rapid rates ofnormalization.

U U U w Ages of speech-sound normalization. Figure 3 is a plotI p p p over time of the averaged performance of the 10 speech-

I I I I disordered children on five indices of speech-sound produc-p h b b tion in conversational speech: Intelligibility Index, PercentageI I I of Vowels Correct, Percentage of Consonants Correct-Sin-h i t gletons, Percentage of Consonants Correct-Clusters, andf d d d the weighted average of the latter two variables-the Per-

I I I centage of Consonants Correct. Beginning with the Intelligi-k k 9 bility Index in Figure 3 (defined as the percentage of words

that could be glossed by the transcribers; Shriberg, 1986),b f g h the trend at approximately 5.5 years changes from thed b w f approximately 80% intelligibility range and below to the 90%I I | range and above for succeeding years. The Percentage ofk w h Vowels Correct, which includes all vowels, all diphthongs,g t f k and the rhotic vowels /3/ and /a,/, does not average over 90%I correct until sometime after 8 years. Trends for the Percent-r g v v age of Consonants Correct-Singletons and Percentage of

I I Consonants Correct-Clusters are generally parallel, withtS s tS' S differences at each of the 12 age levels ranging from 3.1% to

S I d3 t 18.8% (M = 10.2%; SD = 5.1%). Notice that the trend for theI I I overall consonant index-the Percentage of Consonantss r s d3 Correct-is closest in magnitude to the trend for Percentage

I of Consonants Correct-Singletons, which is weighted by thet ZtS z S more frequent occurrence of consonant singletons. Havingv S | 6 illustrated the parallel associations among the three conso-I I nant indices, the Percentage of Consonants Correct will be

6 d3 3 0 used in place of the singleton and cluster indices in all furtheranalyses.

The shapes of these consonant normalization trends ind3 3 0o Figure 3 and, to a lesser degree, the intelligibility and theI I I vowel normalization trends suggest that there is a leveling of

z v 6 r the gains in normalization between 6 and 7 years, and againI I z at approximately 8.5 years until the latest available data

I I point. Thus, these group-level data suggest that speech-3 z j+ 3+ sound normalization does not proceed in a linear fashion.

nsufficient to rank. Rather, normalization progress appears to differ within timeperiods. Importantly, when assessed by the 90% masterycriterion, speech-sound normalization is not complete for

The averaged normalization se- some children with developmental phonological disorders byand Middle-8 consonant sound as late as 11:4, the age of the oldest child at the last

elation to the sequences observed assessment session. A more detailed analysis of the ages atacquisition. Thus, the exact order which normalization occurs requires examination of the rates

zation in a given child with disor- of speech-sound normalization.cted to follow a specific sequence. Rates of speech-sound normalization. The concept ofuence of speech-sound normaliza- rate of speech-sound normalization requires that absolute

24 b

1164 Joumal of Speech and Hearing Research

I-i

m FA

0C)

MZCo

4z0Coz0C.)U-0iLC,

iW

0Iz

z0

N

M-J4

0zU-0LU

Z

C)z

U)

w m b j n w d p h t k g f v tr d3 e s z 8 r 3EARLY-8 MIDDLE-8 LATE-8

FIGURE 2. Comparison of the present longitudinal data on the sequence of long-term speech-sound normalization to the cross-sectional data for speech-disordered children reported InShrlberg (1991b, 1993).

gains in the percentage of correct sounds be converted torelative values reflecting point-to-point gains, losses, or pla-teaus per time period. The trends in each of the six panels inFigure 4 include such information. The three top panelsprovide information for the three developmental soundclasses defined previously as the Early-8, Middle-8, andLate-8 sounds. The bottom three panels include rate infor-mation relative to word position-initial, medial, and final.

0

I-ZLU0

LU

The solid line trends in the upper portion of each of the sixpanels are the percentage of correct consonant sounds foreach age set. The trends in the lower portion of each graph,which vacillate around the 0 percent line, reflect a measure ofrate obtained as follows.

Procedure to derive rate of normalization. When the per-centage of consonants correct data are plotted by age level,the line between each data point is a linear approximation of

3 4 5 6 7 8 9 10 11 12AGE (YEARS)

FIGURE 3. Longitudinal data for the 10 speech-disordered children on five Indices of speech-sound production In conversational speech.

37 1151-1177 October 1994

it


t ,a

CL

0.c

c

0- 0

0

I

.5

\i ~CC

J ~~ S

a.

I'~~~~~~!i

IIIII

I!

i-ia0J

p a R o a o 0

III

i i

4

1IC

Ima

Iz-

i 0

I

iNi %

aa U a -g

I - I

* ( SI

O

*

.

a ~~~~~~~~~a.,h

g ss~~seisa i~ig o

IC

cc

0

' i o

/ "i-r~ ~~~~~? I

Ua C

-o

- 2 xi~ w

S

l· . , . . . . . . . . . , - . C_

" | , : I_ W -~U

&

-

s

A ......... , ·

- ^ * - - " - V7 ?


a function y = f(t)l where y equals the cumulative normal-ization in percentage of consonants correct as a function f oftime t. The rate of change of this function was calculated bya differencing procedure. The percentage correct value foreach data point was subtracted from the value immediatelyfollowing. The resulting value was divided by the age differ-ence between the same data points. Thus, for the 12 datapoints representing the age sets, the differencing procedureyielded 11 data points, each representing the time-adjusteddifference between two adjacent data points. The dashedtrend in Figure 4 representing the rate of normalization (i.e.,percent per year) was obtained by plotting these new datapoints, using the midpoint between adjacent ages to deter-mine location on the age axis. Thus, whereas the solid linetrend in each of the six panels represents the point-to-pointprogress towards speech-sound normalization, the dashed-line trend in each panel represents the rate of change of eachline segment in the upper trend. Points above 0 on the lowertrend indicate a positive rate of normalization, with peaksindicating the periods of most rapid change. Conversely,points below 0 reflect regressions in rate of normalization.

Findings: Developmental Sound Class. Beginning with thePCC trend data in the upper left panel of Figure 4, the Early-8sounds were nearly normalized at the onset of the study withjust under 80% correct, gaining approximately 20% duringthe study. As indicated by the Rate trend below, most of thisgain occurred between ages 4:5 and 4:9, when the normal-ization rate peaked at an average rate of 38% per year for abrief 4-month period. After this gain, normalization pro-ceeded at 0/o-6% per year as the 100% ceiling was gradu-ally approached.

For the Middle-8 sounds, the PCC trend indicates thatapproximately 50% of sounds in this class were correct at theonset of the study, with a 40% total gain achieved byapproximately 11 years. Mean normalization rates are high-est between ages 5:2 and 5:8 at 18% per year and from ages6:8 to 8:4, peaking at 20% per year between ages 7:6 and8:4. No gain or a slight regression demarcates these periodsof more rapid normalization in a short 5-month period fromages 5:8 to 6:2. After about 8.5 years, progress towardsnormalization occurs at a rate of about 4% per year.

For the Late-8 sounds at study onset, 5% of consonantswere correctly articulated. By 11 years, this improved to justunder 70%, a gain of roughly 65%. The rate trend in thispanel indicates a mean gain rate of 36% per year in the initial6 months, followed by an equivalent reduction in rate reflect-ing the reversal in the PCC trend for the next 4 months.Because the first two data points represent the mean of only3 of the 10 subjects (see Figure 1) these early data pointsmay be less stable. Mean PCC gains of approximately100/-15% occurred from 4:9 to 6:8; followed by a plateauuntil approximately 7.5 years. The rapid rate of normalizationduring the following year is indicated by the peak in rate of32% during this period. Normalization after approximately 8.5years proceeds at the 2%-6% rate observed for both theEarly-8 and Middle-8 sounds.

These developmental sound class data provide a morefine-grained view of the index data presented in Figure 3.Specifically, they indicate that (a) for the Early-8 sounds, onebrief period of normalization occurs in the latter half of the

fourth year; (b) for the Middle-8 and the Late-8 sounds, twoperiods of more rapid speech-sound normalization occur, thefirst between the fifth and sixth year, and the second justbefore the eighth year; and (c) for all sounds not yet acquiredby about 8.5 years, there is a marked reduction in furthernormalization rates on the order of 00/-6% per year.

Findings: Word position. The lower three panels in Figure4 are profiles of normalization progress and rates of normal-ization collapsed by sound class and displayed by wordposition. As shown in the left-most lower panel, speechsounds in word-initial position are approximately 60% correctat 3:11 and gain about 30% over the duration of the study,with most of this gain occurring in two periods. The firstperiod of more rapid normalization occurred from 4:5 to 4:9 ata mean rate of 18% per year. The second period from 7:6 to8:4 had a mean gain rate of 14%. There is also a small anduncertain normalization increase in the last 10 months,possibly reflecting a small regression in the prefinal sessionor instability associated with the last data point, whichincludes data for only four subjects. Normalization of conso-nant sounds in word-medial position begins at approximately42% correct at study onset, gaining approximately 40%during the period of study. Periods of greatest average rate ofimprovement are from 4:5 to 4:9 at 28% change per year andfrom 7:6 to 8:4 at 20% per year. These are approximately thesame periods and similar rates found for the initial sounds. Inthe right-most lower panel, word-final sounds at study onsetare 40% correct at 3:11 and gain an overall 40% by 11:2.There are three periods of accelerated normalization. Thefirst, from 3:11 to 4:5 is at 18% per year; the second, from 5:2to 5:8 reaches 14% per year; and the third period of high rate,from 6:8 to 8:4, averaged 16% per year also between 7:6 and8:4.

These word position data provide another and compatibleview of the developmental sound class data in the upperthree panels. Word-initial and word-medial sounds had thehighest mean normalization rates from the fourth to fifthyears, whereas the peak normalization period for word-finalsounds is later-at approximately 5.5 years. When collapsedacross word position and weighted by frequency of occur-rence, the peak normalization period occurs from 7.5 years tojust over 8 years of age, the second period of more rapidnormalization described above in the analysis by develop-mental speech-sound classes. Over the entire period of thestudy, average normalization rates for consonants for thesechildren is approximately 5% per year.

Ages and rates of speech-sound normalization com-pared to normal speech acquisition. To compare thespeech-sound normalization of the speech-delayed childrenwith the speech acquisition of their speech-normal agemates, articulation test data were plotted for the Early-8,Middle-8, and Late-8 sounds together with the correspondingdata collected by Smit et al. (1990) reported as the Iowa-Nebraska normative project. Figure 5 is a plot of the sixtrends reflecting percentages of consonants correct by age inyears. The filled symbols are the averaged data for eachsound class for the speech-delayed children in each of the 12age sets that include from 3 to 10 children. These trends arethe same as the solid-line trends in the three upper panels inFigure 4. The open symbols in Figure 5 are data average-

37 1151-1177 October 1994

Shriberg et aL: Long-Term Normalization in Developmental Phonological Disorders 1167

2 3 4 5 6 7 8 9 10 11 12

=. AGE (YEARS)FIGURE 5. Comparison of the longitudinal normalization data for the present speech-delayedsubjects with the cross-sectional acquisition data from Smit, Hand, Frellinger, Bernthai, andBird (1990).

dacross each sound class from the cross-sectional datareported by Smit et al. using only the male data to mostclosely approximate the gender distribution (9 to 1) of thespeech-delayed children. Each plot point in the Smit et al.data represents the responses of 22-73 children.

Focusing first on speech-sound mastery at the earliestages reported in each study, the speech-delayed children(at 3:11) have approximately 45% of the Early-8 soundscorrect, compared to 90% or twice that percentage correctfor the Smit et al. (1990) study's male children (at 3:0),meeting Sander's (1972) highest criterion for speech-soundmastery. At this same age, the speech-delayed childrenhave approximately 30% of the Middle-8 sounds correct,compared to over twice that percentage (70%) correct forthe normally speaking children. For the Late-8 sounds, therelevant figures are 5% correct for the speech-delayedchildren and 35% for the speech-normal children, a seven-fold difference.

Under inspection in Figure 5 is how the normalizationplots for speech-normal and speech-delayed children mightbe developmentally related, if they are. Theoretical posi-tions outlined by Bishop and Edmundson (1987) suggestthree possible alternatives: (a) speech-normal and speech-delayed children display the same course of normalization,but speech-delayed children begin to normalize later; (b)speech-normal and speech-delayed children begin normal-izing at the same time, but speech-delayed children have aslower rate of normalization; or (c) speech-normal andspeech-delayed children begin normalization at the sametime and rate, but at some subsequent age the speech-delayed children cease normalizing. If the first of thesethree alternatives best characterizes the data, appropriate

lagging on the age axis should result in only a singlenormalization course for both groups. Alternatively, if either(b) or (c) more closely characterizes the data, age-laggingshould not result in a coincidence of the normalization plots.

Figure 6 includes the six normalization plots in Figure 5,lagged by the number of years and months indicated in thelegend. To create the coincidence apparent in Figure 6,lagging was based on a fixed age-scale position for theEarly-8 sound plot for normal children so that all adjustmentswere made by subtracting the appropriate years:months (seelegend) from values on the age axis. A consequence of thisprocedure is that the resulting continuous plot in Figure 6extends from a negative 4.5 years to a positive 10 years. Thealignment needed to achieve one composite trend yields thefinding that significantly speech-delayed children are approx-imately 5 years behind speech-normal children in acquiringthe Early-8 sounds and approximately 3 years behind inacquiring the Middle-8 and Late-8 sounds.

To estimate the goodness-of-fit of the shifted normalizationplots to this composite trend, age points were set to positivevalues and regression procedures applied (Weisberg, 1980).Figure 7 is a plot of the resulting fit for the regressionequation along with unconnected plots for the age-shiftedpercentage of consonants correct data from Figure 6. Theresulting equation accounts for a decisively high 93.3% of thevariance, with a standard error of 6.83%. By traditionalstatistical criteria, it appears to be appropriate to claim thatthis equation and its corresponding fit provide a valid char-acterization of speech-sound normalization in both normaland speech-delayed children. The trend in Figure 7 is con-sistent with the position that there is a single course ofnormalization for both groups of children, differing only in

I-O. 100rrn- 90.0(.) so.

Co U)I 70'Z<Z 600oC) 50'Z0o 40'

O 30-

LI 20'

I 10-ZL 0OC.)Ii

_ _ _ _ ~ ~ ~ ~~~. .. . . . . .


QWI-a

CO)o

zz

Z

0Moz0sooLLU-0

w0

zw

a

4l

iS

100o

90-

80-

70'

60

50 -

40

30

20'

10'

0o-5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9

ADJUSTED AGE LEVELS (YEARS)10

FIGURE 6. Plot of the trends in Figure 5 lagged by the number of years and months Indicatedin the legend.

temporal markers among the three speech-sound classesand between group assignment. This finding is markedlyconsistent with the first of the three hypotheses aboutspeech-sound development proposed by Bishop and Ed-mundson (1987) and the findings of Curtiss, Katz, and Tallal(1992) for syntax.

An even closer fit to the normalization equation wouldhave resulted had the normalization trend for the speech-

0iL

0

W

O

z

I-Z

Z0

CU)Z0

0

I-

1llC-

a.

a10

9

8

7

6(

5

4(

3

2

I1

delayed children's Late-8 sounds not flattened at about 8.5years. Inspection of Figures 5-7 indicates that in fact littlenormalization occurs after 8.5 years of age for the speech-delayed children, whereas for the speech-normal childrenall sounds have by this time achieved a normalizationceiling. These additional findings are more consistent withBishop and Edmundson's third hypothesis that speech-delayed children, after following the same course of speech-

0 0

io.,/ o

0 o / 0 PCC = -.92 yrs + 19.4 yrs - 2.67

0' /· r 2 = 93.3%SE = 6.83%

01 0 o

0 e

FIGURE 7. Regression analysis of the age-shifted percentage of consonants correct data InFigure 6.

0 +1 +2 +3 +4 +5 +6 +7 +8 +9 +10 +11 +12 +13 +14RELATIVE AGE (YEARS)

_j

37 1151-1177 October 1994

---

I · r ·- r · I · I-l · 1 I ·r ~ l · i-i · i I


sound acquisition as speech-normal children, might reachan age plateau after which little further normalization oc-curs. Thus, a more appropriate composite interpretation isthat, in comparison with the course of normal speechacquisition, (a) speech-delayed children follow an identical,but temporally delayed, course of normalization, with theamount of delay greater for the characteristically earliestdeveloping sounds; and (b) development is similar untilabout 8.5 years of age, when normalization is either essen-tially complete or it plateaus, leaving residual articulationerrors (cf. Shriberg, 1993).

A final series of age analyses explored the validity of thesecond part of the above claim-that normalization plateausat 8.5 years in speech-delayed children. Figure 8 includesnormalization plots for 7 of the Late-8 sounds for the 10speech-delayed children (there were insufficient conversa-tional speech tokens to plot normalization of /3/). The uppertwo panels include the cognate pairs /01-/6/ and /s/-/z/,respectively; the lower panels include the trends for IS/ andthe trends for the liquids /r/ and /I/. For the //-/6/ cognatesthe most rapid period of normalization occurred from 7:6 to8:4. After 8:4, slight gains in /6/ are observed until 9:4.Maximal normalization rates also occur from 7:6 to 8:4 for/s/ and /z/, with both sounds continuing to improve until thelast age level tested. Maximal normalization rates alsooccur from 7:6 to 8:4 for I/S/, with rather large session-to-session variability, but no appreciable gains after 8:4.Maximal rates of normalization for /I/ occur from 6:8-8:4,after which either a plateau or even a regression occurs inthese group-averaged data. Finally, the pattern of /r/ nor-malization is difficult to characterize. From 3:11 to 8:4normalization progress is from near zero percent correct toapproximately 15% correct, after which the averaged scoresrange from approximately 15% correct to about 30% cor-rect. These sound-level data support the prior overall findingof rapid growth from approximately 7:6 to 8:4, with a clearslowing of normalization after 8:4.

Error-Type Patterns in Long-Term Speech-SoundNormalization

Figure 9 includes group-level analyses of the covariance ofspeech-sound normalization (i.e., percentage consonantscorrect) with the three absolute error types. Thus, the fourtrends sum to 100% in each of the nine panels. Trends in thenine panels illustrate the interaction over time of the threeprimary factors in long-term speech-sound normalization:Word Position (word-initial, word-medial, word-final) x De-velopmental Sound Class (Early-8, Middle-8, Late-8) x Ar-ticulation Type (correct, omissions, substitutions, distortions).The distortion errors include only those classified as clinicallysignificant (cf. Shriberg, 1986, 1993), the most frequent ofwhich are dentalized sibilant fricatives, lateralized sibilantfricatives, derhotacized r-colored vowels and consonant /r/,and labialized liquids.

Early-8 sounds. Beginning with the Early-8 sounds inFigure 9, the upper three panels include for each wordposition the mean percentages of consonants correct andpercentages of each of the error types for each age level.

Omission errors are the most frequent error type for theseeight early-developing sounds, becoming more frequent asthe target sounds occur later in the word. Substitutions arethe next most frequent error type for both word-initial andword-medial Early-8 consonants. Notice that, in word-finalposition, the only remaining errors after approximately 7.5-8.5 years are omissions. Distortion errors are infrequent in allword positions for the Early-8 sounds.

Middle-8 sounds. The pattern of errors for the Middle-8sounds across word position also consisted of omission andsubstitution errors, with 0% or very low percent distortionerrors over time. Omission errors are relatively frequentonly up to approximately 5 years of age in word-initialposition, up to just over 8 years in word-medial position, andup to approximately 9 years in word-final position. Substi-tution errors are overall the most frequent error-type forMiddle-8 sounds. Particularly notable is the crossover in thefrequency of omission and substitution errors at just over 5years of age for Middle-8 sounds occurring in word-finalposition.

Late-8 sounds. Error-type patterns for the Late-8 soundsare generally similar in each of the three word positions.Substitution errors are most frequent in word-initial positionuntil a sharp increase in correct sound production at about7.5 to just over 8 years, after which distortion errors are mostfrequent. In word-medial position, all three error types occurwith roughly equal frequency until approximately 7.5 years,after which distortion errors also become most frequent.Finally, in word-final position, the three error types occur withapproximately equal frequencies until the sharp increase incorrect production of Late-8 sounds at just over 8 years, afterwhich distortion errors are most frequent. Thus, as Late-8sounds reach their overall levels of approximately 60%-70%correct in each of the three word-positions at just over 8years, the absolute error types also stabilize at approximately25%-30% distortions, 1 0%-15% substitutions, and 0%0/-10%omissions.

Figure 10 is an alternative summary of the information inFigure 9, here collapsed over word position. Averaged overthe word-position effects just described, the differences in theerror-type patterns for the three developmental soundclasses form a generally more coherent picture. Errors onEarly-8 sounds are most frequently omissions, with therelative percentage of omissions increasing with age level.Errors on the Middle-8 sounds are most frequently omissionsand substitutions. Finally, beginning at approximately 7.5years, errors on the Late-8 sounds are most frequentlydistortions.

Summary

To summarize the error pattern data in speech-soundnormalization, the progression of error types from omissionto substitution to distortion interacted with developmentalsound class and word position. The Early-8 sounds weremore often omitted, the Middle-8 more often omitted andsubstituted, and the Late-8 were more often distorted. Ingeneral, omissions and substitutions increased from word-initial to word-final position.


_A >]

I I 'r

S- ~

0 -

D

V

S.- *S

37 1151-1177 October 1994

. ..... ,-,., .. ,.... t. -,e

8 : a 9 9 a 8 ; , N 0

.N

0

w

iJ

r

r >

w

OOVC,W

-W

cn

C,

l

0

rm

ID

C.,

_

2-

0

c -

. oa-

00

': 0

Z

I1

e v0 E

o

.C

z

m C

l1CL

10 a3 a 0 IN 3 N

133Uk00 3ViN3933d

I


4

r

8-s g-gf a v yI a v -

-j0a

Ula2

a:-az

.j

-i ii 8 '--ia -

z

'i

I

C,

-I4

2

r

W

-a

a

i - i , i x - A

o

!

.14a

I

iz

20?

Mi

U

sa R a3R a -

Itr

:N_

0g

-I

4E3

2

2

aGPla-j

3

V

.1-j4z

z

IZ0S

8 a s a A ° -

-lQ

QIU

z

-a

8 a g ig s 3a ai-

-l

t3

98z

8

-I

5 5 5 g 5 E V N N 0

E

SO

oca

-=

-o

C~o

C 0

0X

2E

*:6

C

C.

OCc

W

co

o

o,-

.E

d.

I 00

iN3t d NS aIN3OU3d NV3M

- __ro

·- ,. ....... ._n

rb.- II

_

N _N

cc

o

o

o

m

m

_ N

I I

N

I

I

N

I 1.0

·a

=.

1 8118 1

f 9�

i

I

I

I1;11 o

;. . .. . . m


z

0COC?LU-ja0

i

r

z

-u

LU

II R 8 5 X R ° °

.1I ,E0

0

*co

q0

-TCI

Cnazm0Co

-J

rl

A:

_N

0

LU

LU

0

-a

0·;

2

d0

S

0

0

'SI0

CL

'.a.3

Il

.Ccm0m

C

i0.

0

UJc'

- - s -

37 1151-1177 October 1994

N

i

IN333d 3AIIVf13tl


Conclusions

Six primary findings from this study can be summarized asfollows:

1. There is no one invariant sequence of speech-soundnormalization acquisition or error-pattern followed by allchildren acquiring speech normally or normalizing a devel-opmental phonological disorder. Even when group-aver-aged, there is no one sequence of speech-sound acquisitionor error-patterns that has high stability across studies ofnormal acquisition or short-term or long-term normalization ofa speech disorder.

2. Individual and group-averaged similarities in the se-quence and error patterns of normal speech-sound acquisi-tion and short-term and long-term normalization of disordersbegin to emerge when (a) cross-tabulated by the indepen-dent variables of sampling mode and word-position (andsingleton-cluster status) and (b) subgrouped by dependentvariables based on either developmental sound classes orphonetic features.

3. When examined at the group-averaged level by theindependent and dependent variables in (2) above, long-termspeech-sound normalization of a developmental phonologi-cal disorder follows the same sequence and is characterizedby the same error patterns as occur in normal speechacquisition.

4. For children with moderate-to-severe speech-soundinvolvement when first identified during the preschool period,the average temporal delay in achieving nearly-completelong-term speech-sound normalization is 5 years.

5. The approximately 5-year period of long-term speech-sound normalization is characterized by two periods of morerapid gains from approximately 4 to 6 years and 7 to 8.5years, interspersed with a period of relatively slow growthfrom 6 to 7 and after 8.5 years.

6. The persistence of distortion errors on Is/-/zl, /r/-/3/-/a/,and /I/ (as well as some other sounds) past approximately 9years is likely in children with moderate-to-severe speechdelays identified at preschool ages.

Theoretical Issues

The primary theoretical issues prompted by these findingson long-term speech-sound normalization concern the neu-rodevelopmental correlates of speech acquisition. After ap-proximately 8.5 years, each of the children in this studyevidenced a clear slowing of normalization, with the excep-tion for some children of continued but slower gains in thecorrect articulation of Isl and /z/. Although on a percentagebasis their speech was nearly normalized by 8.5 years, mostchildren continued to have deletion or substitution errors onone or more of the Late-8 sounds in addition to the predom-inant distortion errors.

One compelling theoretical explanation for the slowing ofnormalization at approximately 8.5 years is to invoke someform of the critical period perspective discussed by Lenne-berg (1967) and more recently by Locke (1994). This per-spective presumes that the capacity for language developswithin a maturational envelope that begins to close at 8 to 9

years of age. Support for this critical or maximally sensitiveperiod for language acquisition includes studies of foreignlanguage accent and plateaus in language development,with retardation beginning at approximately 8 years. Follow-ing this perspective, one interpretation of the present data isthat (a) the speech acquisition of these children was, forsome reason, literally delayed; (b) normalization processesreached a species-general terminus period, past which littleor no further normalization occurred; and (c) if not correctedby the end of this critical period, remaining errors persistedas residual speech-sound errors.

Although the present data are consistent with the generalconcept of a developmental window, the concept wouldbecome even more attractive if it could provide specificexplanation for two other findings. First, stronger support forthe relevant cognitive versus speech-motor correlates of acritical period would involve associations with some metricunderlying the sequence of speech-sound acquisition andnormalization-or at least some ontogenetically principledscale underlying the developmentally stable classes termedthe Early-8, Middle-8, Late-8 sounds. Second, unaddressedis the question of the developmental processes that mayunderlie the two periods of accelerated normalization or,alternatively, of the plateaus in normalization from approxi-mately 6-7 years and after 8.5 years. Overall, however, theprimary finding that normalization processes appear to slowdown before all errors are corrected provides strong supportfor some form of an age-based developmental window forspeech acquisition. Specifically, regardless of the absolutelevels of cognitive and speech involvements, normalization ofat least some speech-sound distortions appears to be mark-edly less probable after approximately 8.5 years. Furtherspeculation in relation to alternative theoretical views ofphonological acquisition would be inconsistent with the de-scriptive purposes of this prospective study. For example,although it might be appealing to interpret the observedplateaus as support for stage-based models of normal pho-nological acquisition, such argument would exceed the meth-odological constraints of these data.

Clinical Issues

The findings summarized at the beginning of this sectionmight provide some perspective on the long-term clinicalexpectations for children with moderate to severe develop-mental phonological disorders. Three issues warrant com-ment.

First, it should be emphasized that the children in this studyconstituted a group with significant speech-language involve-ment. This longitudinal study was initiated nearly two dec-ades ago, predating the major shifts in phonological theory,assessment procedures, and service delivery perspectivesthat characterize contemporary activities in developmentalphonological disorders. For example, children selected forparticipation had significantly more speech involvement thanthe children studied by Shriberg and Kwiatkowski (1994) and,as expected, most later received special educational serv-ices in schools for associated language and academicneeds. None of these children had histories of short-term


normalization, defined in the prior paper in this series asnormalization by age 6 years. Thus, it is appropriate to limitclinical generalizations to the percentage of children from thissector of the speech-delayed population. Based on Percent-age of Consonant Correct status (including only children withmoderate-severe and severe involvement considering theages of these children at first assessment), this sector isestimated to comprise approximately 42% or less than half ofall children identified as having a developmental phonologi-cal disorder (Shriberg & Kwiatkowski, 1994).

Second, for counseling caregivers of children with thisdegree of involvement it seems clear that speech-soundnormalization should not be expected to be essentiallycomplete until approximately third grade. This endpoint fornormalization precisely coincides with the traditional "waitand see" period used by generations of speech-languagepathologists who have observed some children normalizewithout intervention during this period. In the present situa-tion, these children did receive intervention programming ofmany types, yet their speech errors persisted until approx-imately 8.5 years when most, but not all, speech-sounderrors were resolved. The types of intervention program-ming they received reflected the state of clinical practice inthe late 1970s and early 1980s, raising the question ofwhether current clinical techniques in child phonology mightyield more effective and specifically more efficient gains.Perhaps the best use of these clinical data from theperspective of caregiver expectations is as a benchmark forthe outer boundaries of speech-sound normalization. Bar-ring other complications in the individual or his or herservice delivery history, children similar to those describedin this study might be expected to follow the same timecourse of normalization of the Early-8, Middle-8, and Late-8sounds described in this report.

Finally, these findings for rates of speech-sound normal-ization in relation to the sequence and error-type findingswould seem to challenge contemporary speech-languagepathologists and researchers to examine the timing of theirintervention programs. Hodson and Paden (1991) reportthat for most children an intensive cyclical program ofintervention provided during preschool years results inintelligible speech warranting clinical dismissal after approx-imately 30 hours of clinical work, with a maximum reportedto be 72-91 hours over an 18-month period. Other contem-porary treatments based on a diversity of cognitive, motor-speech, and descriptive linguistic frameworks also claim tobe effective in rapid remediation of children having speech-language disorders (cf. Bernthal & Bankson, 1993). In lightof the periods of more accelerated versus flat normalizationgrowth observed in the present data, two relevant clinicalquestions arise: (a) Does intensive early speech interven-tion, especially within certain felicitous time periods, in-crease the probability of short-term rather than long-termnormalization? and (b) Is the selection and sequencing ofphonological targets more crucially tied to a child's currentphonological system or, alternatively, to the child's chrono-logical age? A maximally efficient clinical technology mightsomeday be based on readily calculated trajectories ofspeech-sound normalization, with intervention services ac-celerated during optimum growth periods.

Acknowledgments

We thank each of the persons listed in the acknowledgments in thefirst article in this series (Shriberg & Kwiatkowski, 1994, p. 1117) fortheir contributions to the present article as well. We also wish toacknowledge the capable and affable long-term assistance of WilliamHome, who made simultaneous audio-video tapes of each assessmentsession during the entire 7 years of the study. Finally, our sincerethanks to the children and their parents, who so agreeably fulfilledtheir parts in a logistically difficult long-term commitment. This workwas supported by grants from the U.S. Department of Education(G008400633) and the U.S. Public Health Service, NIDCD No.DC00496.

References

Albright, R. W., & Albright, J. B. (1956). The phonology of a twoyear old child. Word, 12, 257-261.

Anderson, R., & Smith, B. L (1986). Phonological development ofa two-year-old monolingual Puerto Rican Spanish-speaking child.Journal of Child Language, 14, 57-78.

Applegate, J. R. (1961). Phonological rules of a subdialect ofEnglish. Word, 14, 150-177.

Arit, P. B., & Goodban, M. T. (1976). A comparative study ofarticulation acquisition as based on a study of 240 normals agedthree to six. Language, Speech, and Hearing Services in Schools,7, 173-180.

Bateman, W. (1916). The language status of three children at thesame ages. Pedagogical Seminary, 23, 211-240.

Bernhardt, B. (1990). Application of nonlinear phonological theoryto intervention with six phonologically disordered children. Unpub-lished doctoral dissertation, University of British Columbia.

Bernhardt, B. (1992). The application of nonlinear phonologicaltheory to intervention with one phonologically disordered child.Clinical Linguistics & Phonetics, 6, 259-281.

Bernthal, J. E., & Bankson, N. W. (1993). Articulation and phono-logical disorders (3rd ed.). Englewood Cliffs, NJ: Prentice Hall.

Bishop, D. V. M., & Edmundson, A. (1987). Language-impaired4-year-olds: Distinguishing transient from persistent impairment.Journal of Speech and Hearing Disorders, 52, 156-173.

Bortolln, U., & Leonard, L. B. (1991). The speech of phonologicallydisordered children acquiring Italian. Clinical Linguistics & Phonet-ics, 5, 1-12.

Bricker, W. A. (1967). Errors in the echoic behavior of pre-schoolchildren. Journal of Speech and Hearing Research, 10, 67-76.

Burchinal, M., & Appelbaum, M. I. (1991). Estimating individualdevelopmental functions: Methods and their assumptions. ChildDevelopment, 62, 23-43.

Chamberlain, A., & Chamberlain, I. (1904). Studies of a child.Pedagogical Seminary, 11, 264-291, 452-483.

Chamberlain, A., & Chamberlain, I. (1905). Studies of a child.Pedagogical Seminary, 12, 427-453.

Chomsky, N., & Halle, M. (1968). The sound pattern of English.New York: Harper & Row.

Cross, E. (1950). Some features of the phonology of a four-year-oldboy. Word, 6, 137-140.

Curtiss, S., Katz, W., & Tallal, P. (1992). Delay versus deviance inthe language acquisition of language-impaired children. Journal ofSpeech and Hearing Research, 35, 373-383.

Diedrich, W. M., & Bangert, J. (1980). Articulation learning. Hous-ton, TX: College-Hill Press.

Dlnnsen, D. A., Chin, S. B., & Elbert, M. (1992). On the lawfulnessof change in phonetic inventories. Lingua, 86, 207-222.

Dlnnsen, D. A., Chin, S. B., Elbert, M., & Powell, T. W. (1989).Some constraints on functionally disordered phonologies: Pho-netic inventories and phonotactics. Journal of Speech and HearingResearch, 33, 28-37.

Dodd, B., & lacono, T. (1989). Phonological disorders in children:Changes in phonological process use during treatment. BritishJournal of Disorders of Communication, 24, 333-351.

37 1151-1177 October 1994


Donegan, P. J., & Stamps, D. (1979). The study of natural phonol-ogy. In D. A. Dinnsen (Ed.), Current approaches to phonologicaltheory (pp. 126-173). Bloomington: Indiana University Press.

Dunn, C. (1982). Phonological process analysis: Contributions toassessing phonological disorders. Communicative Disorders, 7,46-58.

Dunn, C., & Davis, B. L. (1983). Phonological process occurrence inphonologically disordered children. Applied Psycholinguistics, 4,187-207.

Dyson, A. T., & Paden, E. (1983). Some phonological acquisitionstrategies used by two-year-olds. Journal of Childhood Commu-nication Disorders, 7, 6-18.

Edwards, M. L., & Bernhardt, B. H. (1973). Phonological analysesof the speech of four children with language disorders. Unpub-lished manuscript, Stanford University, Institute for ChildhoodAphasia, Stanford, CA.

Edwards, M. L, & Shrlberg, L D. (1983). Phonology: Applicationsin communicative disorders. San Diego: College-Hill.

Elbert, M., & McReynolds, L. V. (1978). An experimental analysis ofmisarticulating children's generalization. Journal of Speech andHearing Research, 21, 136-150.

Elbert, M., & McReynolds, L. V. (1979). Aspects of phonologicalacquisition during articulation training. Journal of Speech andHearing Disorders, 44, 459-471.

Fee, J., & Ingram, D. (1982). Reduplication as a strategy ofphonological development. Journal of Child Language, 9, 41-54.

Ferguson, C. A. (1973). The acquisition of fricatives. Papers andReports on Child Language Development, 6, 61-86.

Ferguson, C. A., & Farwell, C. B. (1975). Words and sounds inearly language acquisition. Language, 51, 419-439.

Ferguson, C. A., Peizer, D. B., & Weeks, T. E. (1973). Model-and-replica phonological grammar of a child's first words. Lingua, 31,35-65.

Gibson, D. (1966). Early developmental staging as a prophecy indexin Down's Syndrome. American Journal of Mental Deficiency, 70,825-828.

Goldman, R., & Fristoe, M. (1969). Goldman-Fristoe test of articu-lation. Circle Pines, MN: American Guidance Service.

Goldstein, B., & Igleslas, A. (1991, November). The developmentand treatment of phonological processes in Spanish speakingchildren. Paper presented at the annual Convention of the Amer-ican Speech-Language-Hearing Association, Atlanta, GA.

Gr6golre, A. (1937). L'apprentissage du langage (Vol. 1). Liege,Paris: Droz.

Grunwell, P. (1982). Clinical phonology. Rockville, MD: Aspen.Haas, W. (1963). Phonological analysis of a case of dyslalia. Journal

of Speech and Hearing Disorders, 28, 239-246.Hester, E. J., Godbold, D., Lee, L. M., & Stephens, S. (1984,

November). Articulation norms used by practicing speech-lan-guage pathologists. Paper presented at the annual Convention ofthe American Speech-Language-Hearing Association, San Fran-cisco, CA.

Hills, E. (1914). The speech of a child two years of age. DialectNotes, 4, 634-639.

Hlnckley, A. (1915). A case of retarded speech development.Pedagogical Seminary, 22, 121-146.

Hodson, B. W., & Paden, E. P. (1981). Phonological processeswhich characterize unintelligible and intelligible speech in earlychildhood. Journal of Speech and Hearing Disorders, 46, 369-373.

Hodson, B. W., & Paden, E. P. (1991). Targeting intelligible speech:A phonological approach to remediation (2nd ed.). Austin, TX:Pro-ed.

Hoffman, P., Schuckers, G., & Daniloff, R. G. (1980). Develop-mental trends in correct I/r/ articulation as a function of allophone.Journal of Speech and Hearing Research, 23, 746-756.

Hoffmann, K. A. (1982). Speech sound acquisition and naturalprocess occurrence in the continuous speech of three-to-six yearold children. Unpublished master's thesis, University of Wisconsin-Madison.

Holmes, U. (1927). The phonology of an English-speaking child.American Speech, 2, 219-225.

Humphreys, M. (1880). A contribution to infantile linguistics. Trans-

actions of the American Philological Association, 11, 5-17.Iglesias, A., & Anderson, N. (1993). Dialectal variations. In J. E.

Bernthal & N. W. Bankson (Eds.), Articulation and phonologicaldisorders (3rd ed., pp. 147-161). Englewood Cliffs, NJ: PrenticeHall.

Ingram, D. (1974). Phonological rules in young children. Journal ofChild Language, 1, 49-64.

Ingram, D. (1989). First language acquisition: Method, descriptionand explanation. New York: Cambridge University Press.

Jakobson, R. (1968). Child language, aphasia, and phonologicaluniversals (R. Keiler, Trans.). The Hague: Mouton. (Original workpublished 1941.)

Jegi, J. (1901). The vocabulary of a two-year-old child. Child StudyMonthly, 6, 241-261.

Kenney, K. W., & Prather, E. M. (1986). Articulation development inpreschool: Consistency of productions. Journal of Speech andHearing Research, 29, 29-36.

Khan, L. M., & Lewis, N. P. (1986). Khan-Lewis phonologicalanalysis. Circle Pines, MN: American Guidance Service.

Klein, H. B. (1981). Early perceptual strategies for the replication ofconsonants from polysyllabic lexical models. Journal of Speechand Hearing Research, 24, 535-551.

Klein, H. B. (1985). Relationship between early pronunciation pro-cesses and later pronunciation skill. Journal of Speech andHearing Disorders, 50, 156-165.

Lahey, M., Flax, J., & Schlisselberg, G. (1985). A preliminaryinvestigation of reduplication in children with specific languageimpairment. Journal of Speech and Hearing Disorders, 50, 186-194.

Lenneberg, E. H. (1967). Biological foundations of language. NewYork: Wiley.

Leonard, L. B. (1985). Unusual and subtle phonological behavior inthe speech of phonologically disordered children. Journal ofSpeech and Hearing Research, 22, 4-13.

Leonard, L. B., Mesalam, L, & Newhoff, M. (1980). Individualdifferences in early child phonology. Applied Psycholinguistics, 1,7-30.

Leopold, W. (1947). Speech development of a bilingual child: Alinguist's record. Vol. 2: Sound learning in the first two years.Evanston, IL: Northwestern University Press.

Lewis, M. M. (1936). Infant speech: A study of the beginnings oflanguage. New York: Harcourt Brace.

Lewis, M. M. (1951). Infant speech: A study of the beginnings oflanguage (2nd ed.). New York: Harcourt Brace.

Locke, J. L. (1983). Phonological acquisition and change. NewYork: Academic Press.

Locke, J. L. (1994). Gradual emergence of developmental languagedisorders. Journal of Speech and Hearing Research, 37, 608-616.

Lorentz, J. P. (1974). A deviant phonological system of English.Papers and Reports on Child Language Development, 8, 55-64.

Magnusson, E., & Naucler, K. (1990). Reading and spelling inlanguage-disordered pre-school children-linguistic and metalin-guistic prerequisites: Report on a longitudinal study. ClinicalLinguistics & Phonetics, 4, 49-61.

McArdle, J. J., & Epstein, D. (1987). Latent growth curves withdevelopmental equations models. Child Development, 58, 110-133.

McDonald, E. T., & Shine, R. E. (1969). Comparison of phoneticprofiles of kindergarten and first grade children. Unpublishedmanuscript, Pennsylvania State University, University Park.

Menn, L. (1971). Phonotactic rules in beginning speech. Lingua, 26,225-251.

Menyuk, P. (1968). The role of distinctive features in children'sacquisition of phonology. Journal of Speech and Hearing Re-search, 11, 138-146.

Miller, J. F. (1981). Assessing language production in children.Austin, TX: Pro-Ed.

Moskowltz, A. (1970). The two-year-old stage in the acquisition ofEnglish phonology. Language, 46, 426-441.

Nesselroade, J. R., & Baltes, P. B. (1979). Longitudinal research inthe study of behavior and development. New York: AcademicPress.

Nice, M. (1917). The speech development of a child from eighteen


months to six years. Pedagogical Seminary, 24, 204-243.Noble, E. (1888). Child speech, and the law of mispronunciation.

Education, 9, 44-52, 117-121, 188-194.Nunnally, J. C. (1962). The analysis of profile data. Psychological

Bulletin, 59, 311-319.Olmsted, D. (1971). Out of the mouths of babes. The Hague:

Mouton.Paden, E. P., Novak, M., & Belter, A. (1987). Predictors of

phonologic inadequacy in young children prone to otitis media.Journal of Speech and Hearing Disorders, 52, 232-242.

Pendergast, K., Dickey, S., Selmar, J., & Soder, A. (1969). ThePhoto Articulation Test Dansville, IL: Interstate Printers andPublishers.

Pollock, F. (1878). An infant's progress in language. Mind, 3,392-401.

Poole, . (1934). Genetic development of articulation of consonantsounds in speech. Elementary English Review, 11, 159-161.

Prather, E. M., Hedrick, D. L., & Kern, C. A. (1975). Articulationdevelopment in children aged two to four years. Journal of Speechand Hearing Disorders, 40, 179-191.

Prins, T. D. (1962). Analysis of correlations among various articu-latory deviations. Journal of Speech and Hearing Research, 5,152-160.

Rles, J. E. (1987, November). Phonological development: Koreanand English. Paper presented at the annual Convention of theAmerican Speech-Language-Hearing Association, New Orleans,LA.

Roberts, J. E., Footo, M., & Burchlnal, M. R. (1988, November).Normal phonological development from 2-1/2 to 8 years: A growthcurve analysis. Paper presented at the annual Convention of theAmerican Speech-Language-Hearing Association, Boston, MA.

Rockman, B., Elbert, M., & Saltzman, D. (1979, November).Diagnostic prediction and patterns of articulation change in pre-schoolers. Paper presented at the annual Convention of theAmerican Speech-Language-Hearing Association, Atlanta, GA.

Rogosa, D., Brandt, D., & Zlmowskl, M. (1982). A growth curveapproach to measurement of change. Psychological Bulletin, 92,726-748.

Sander, E. K. (1972). When are speech sounds learned? Journal ofSpeech and Hearing Disorders, 37, 55-63.

Sax, M. R. (1972). A longitudinal study of articulation change.Language, Speech, and Hearing Services in Schools, 3, 41-48.

Schwartz, R. G., & Leonard, L. B. (1983). Some further commentson reduplication in child phonology. Journal of Child Language, 10,441-448.

Schwartz, R. G., Leonard, L. G., Wilcox, M. J., & Folger, M. K.(1980). Again and again: Reduplication in child phonology. Journalof Child Language, 7, 75-87.

Shibamoto, J. S., & Olmted, D. (1978). Lexical and syllabicpatterns in phonological acquisition. Journal of Child Language, 5,417-457.

Shrlberg, L. D. (1982). Toward classification of developmentalphonological disorders. In N. Lass (Ed.), Speech and language:Advances in basic research and practice (Vol. 8, pp. 2-18). NewYork: Academic Press.

Shrlberg, L. D. (1986). PEPPER: Programs to examine phoneticand phonologic evaluation records. Hillsdale, NJ: Lawrence Erl-baum.

Shrlberg, L. D. (1991a). Directions for research in developmentalphonological disorders. In J. F. Miller (Ed.), Research on childlanguage disorders: A decade of progress (pp. 267-276). Austin,TX: Pro-ed.

Shrlberg, L. D. (1991 b, August). Speech-prosody characteristics ofyoung children with speech disorders of unknown origin. Paperpresented at the Twelfth International Congress of Phonetic Sci-ences, Aix-en-Provence, France.

Shrlberg, L. D. (1993). Four new speech and prosody-voice mea-sures for genetics research and other studies in developmentalphonological disorders. Journal of Speech and Hearing Research,36, 105-140.

Shriberg, L. D., & Kent, R. D. (1982). Clinical phonetics. New York:Macmillan.

Shrlberg, L. D., & Kwlatkowski, J. (1980). Natural Process Analy-

sis: A procedure for phonological analysis of continuous speechsamples. New York: Macmillan.

Shriberg, L. D., & Kwlatkowski, J. (1982). Phonological disordersI: A diagnostic classification system. Journal of Speech andHearing Disorders, 47, 226-241.

Shriberg, L. D., & Kwlatkowski, J. (1983). Computer-assistedNatural Process Analysis (NPA): Recent issues and data. In J.Locke (Ed.), Assessing and treating phonological disorders: Cur-rent approaches. Seminars in Speech and Language, 4, 389-406.

Shriberg, L D., & Kwlatkowski, J. (1994). Developmental phono-logical disorders I: A clinical profile. Journal of Speech andHearing Research, 37, 1100-1126.

Shriberg, L. D., Kwiatkowski, J., Best, S., Hengst, J., & Terselic-Weber, B. (1986). Characteristics of children with phonologicdisorders of unknown origin. Journal of Speech and HearingDisorders, 51, 140-161.

Shriberg, L. D., Kwiatkowski, J., & Gruber, F. A. (1994). Devel-opmental phonological disorders II: Short-term speech-sound nor-malization. Journal of Speech and Hearing Research, 37, 1127-1150.

Shrlberg, L. D., Kwiatkowskl, J., & Hoffmann, K. A. (1984). Aprocedure for phonetic transcription by consensus. Journal ofSpeech and Hearing Research, 27, 456-465.

Shriberg, L. D., & Lof, G. L. (1991). Reliability studies in broad andnarrow phonetic transcription. Clinical Linguistics and Phonetics,5, 225-279.

Shriberg, L. D., & Olson, D. (1988). PEPAGREE: A Program toCompute Transcription Reliability. Waisman Center ResearchComputing Facility, University of Wisconsin-Madison.

Shriberg, L. D., & Smith, A. J. (1983). Phonological correlates ofmiddle-ear involvement in speech-delayed children: A method-ological note. Journal of Speech and Hearing Research, 26,293-297.

Singh, S., Hayden, M., & Toombs, M. (1981). The role of distinctivefeatures in articulation errors. Journal of Speech and HearingDisorders, 46, 174-183.

Slobin, D. (1977). Language change in childhood and history. In J.Macnamara (Ed.), Language learning and thought (pp. 185-214).New York: Academic Press.

Smit, A. B. (1986). Ages of speech sound acquisition: Comparisonsand critiques of several normative studies. Language, Speech,and Hearing Services in Schools, 17, 175-187.

Smit, A. B. (1991, November). Common and uncommon phonologicerrors in the speech of children: Word-initial clusters. Paperpresented at the annual Convention of the American Speech-Language-Hearing Association, Atlanta, GA.

Smit, A. B., Hand, L., Frellinger, J. J., Bernthal, J. E., & Bird, A.(1990). The Iowa articulation norms project and its Nebraska repli-cation. Joumal of Speech and Hearing Disorders, 55, 779-798.

Smith, N. V. (1973). The acquisition of phonology: A case study.Cambridge: Cambridge University Press.

Smith, B. L., & Stoel-Gammon, C. (1983). A longitudinal study ofthe development of stop consonant production in normal andDown's Syndrome children. Journal of Speech and Hearing Dis-orders, 48, 114-118.

Snow, K. (1963). A detailed analysis of articulation responses of'normal' first grade children. Journal of Speech and HearingResearch, 6, 277-290.

Snow, K., & Milisen, R. (1954). Spontaneous improvement inarticulation as related to differential responses to oral and picturearticulation tests. Journal of Speech and Hearing Disorders Mono-graph Supplement, 4, 46-49.

Stampe, D. (1969). The acquisition of phonemic representation.Proceedings of the Fifth Regional Meeting of the Chicago Linguis-tic Society (pp. 433-44). Chicago, IL: Chicago Linguistic Society.

Stampe, D. (1973). A dissertation on natural phonology. Unpub-lished doctoral dissertation, University of Chicago.

Stoel-Gammon, C. (1985). Phonetic inventories, 15-24 months: Alongitudinal study. Journal of Speech and Hearing Research, 28,505-512.

Stoel-Gammon, C., & Cooper, J. (1984). Patterns of early lexicaland phonological development. Journal of Child Language, 11,247-271.

37 1151-1177 October 1994


Templin, M. C. (1957). Certain language skills in children: Theirdevelopment and interrelationships. Minneapolis: University ofMinnesota Press.

Toombs, M., Slngh, S., & Hayden, M. (1981). Markedness offeatures in the articulatory substitutions of children. Journal ofSpeech and Hearing Disorders, 46, 184-190.

Trubetzkoy, N. S. (1969). Principles of phonology (C. A. M. Baltaxe,Trans.). Berkeley: University of California Press. (Original workpublished 1958.)

Velten, H. (1943). The growth of phonemic and lexical patterns ininfant speech. Language, 19, 281-292.

Vlhman, M. M., & Greenlee, M. (1987). Individual differences inphonological development: Ages one and three years. Journal ofSpeech and Hearing Research, 30, 503-521.

Von Bremen, V. (1990). A nonlinear phonological approach tointervention with severely phonologically disordered twins. Unpub-lished Master's thesis, University of British Columbia, Canada.

Weiner, F. F., & Wacker, R. (1982). The development of phonologyin unintelligible speakers. In N. J. Lass (Ed.), Speech and lan-guage: Advances in basic research and practice (Vol. 8). NewYork: Academic Press.

Weir, R. (1962). Language in the crib. The Hague: Mouton.

Weisberg, S. (1980). Applied linear regression. New York: JohnWiley and Sons.

Wellman, B., Case, I., Mengert, I., & Bradbury, D. (1931). Speechsounds of young children. University of Iowa Studies in ChildWelfare 5.

Winitz, H. (1969). Articulatory acquisition and behavior. EnglewoodCliffs, NJ: Prentice-Hall.

Wlnltz, H. (1975). From syllable to conversation. Baltimore: Univer-sity Park Press.

Wohlwill, J. F. (1973). The study of behavioral development. NewYork: Academic Press.

Received April 6, 1993Accepted March 31, 1994

Contact author: Lawrence D. Shriberg, PhD, The PhonologyProject, The Waisman Center on Mental Retardation and HumanDevelopment, University of Wisconsin-Madison, 1500 Highland Av-enue, Madison, WI 53705. E-mail: [email protected]

Appendix

Criteria for Speech-Sound NormalizationAlthough Sander (1972) argues for viewing customary production

as the point at which at least 50% of children produce a soundcorrectly and mastery as the point at which at least 90% of childrenproduce a sound correctly, most of the widely cited normative studieshave used the intermediate mastery percentage of at least 75% ofchildren (cf. Arlt & Goodban, 1976; Prather et al., 1975; Templin,1957; Wellman et al., 1931). In the present study, calculation ofnormalization sequences for all data sets (i.e., for the Hoffmann,1982, data as well as data for the present subjects) is based on theearliest age at which 75% of responses were judged to be phonet-ically correct, using the response definitions for narrow phonetictranscription described in prior reports (Shriberg, 1986; Shriberg &Kent, 1982; Shriberg et al., 1984). The 75% correct criterion is alsoused for comparisons to the Smit et al. (1990) data, which wereoriginally cast as age-based percentages of correct sounds (i.e., total

correct responses divided by total obtained responses), rather thanby the percentage of children producing the sound correctly.

For conversational speech sample analyses, where direct com-parison with traditional normative studies is less appropriate, thecriteria for a normalized speech sound was set at 90% correct tokens(cf. Sander, 1972). Transcription of all intended consonants (andvowels-diphthongs) in the conversational speech samples yielded anaverage of approximately 135 (76%) more consonant tokens (M =313.4 consonants) per sample than obtained from transcripts of thearticulation test data (M = 178.5 consonants).

When data from either the articulation tests or the conversationalspeech samples are collapsed across independent variables-forexample across word position or across gender-the analyses arebased on recalculation of percentages using the appropriate newnumerators and denominators.

1994;37;1151-1177 J Speech Hear Res Lawrence D. Shriberg, Frederic A. Gruber, and Joan Kwiatkowski

Normalization

Developmental Phonological Disorders III: Long-Term Speech-Sound

This information is current as of June 28, 2012

http://jslhr.asha.org/cgi/content/abstract/37/5/1151located on the World Wide Web at:

This article, along with updated information and services, is

http://jslhr.asha.org/cgi/content/abstract/37/5/1151

Documents

Developmental Phonological Disorders III: Long Term Speech ...€¦ · Developmental Phonological Disorders III: Long Term Speech-Sound Normalization Lawrence D. Shriberg Frederic