THE CONDITIONS AND COURSE OF CLINICALLY INDUCED PHONOLOGICAL CHANGE

This two part study continued the evaluation of minimal pair treatment in phonological change (Gierut, 1989, 1990, 1991a; Gierut & Neumann, 1992). Three linguistic variables relevant to change were experimentally manipulated within an alternating treatments design to determine specifically the interplay of a maximal number of feature distinctions, feature class, and relationship of treated phonemes to a child's grammar in inducing sound change. The conditions of treatment that were shown to facilitate optimal phonological change in previous research were again experimentally replicated. Specifically, minimal pairs comparing two phonemes previously unknown to a child that also differed by maximal and major class features were found to be the preferred context motivating change. Important individual differences emerged and undescribed the role of a child's pretreatment grammar in phonological change. These differences contributed to descriptions of possible courses of change followed by children with phonological disorders and bear upon the predictability of change and the effectiveness of treatments that may condition change.

The conditions and course of phonological change have long been of concern to those interested in the study of language. Conditions of change refer specifically to those variables that motivate, influence, or induce the occurrence of change in language; whereas, the course of change refers to how change actually takes place, progresses, or is implemented in the phonology. Both aspects of phonological change have received attention from theoretical and applied perspectives.

The theoretical contribution comes from the field of historical linguistics; it provides detailed traces of diachronic sound change in fully developed primary languages of the world. Here, the conditions influencing phonological change have emerged along space-time and/or internal-external dimensions and are most typically reported as language-specific rule-governed phenomena (Chen, 1972; Chen & Wang, 1975; Janson, 1983; Labov, 1981; Wang, 1969). The description of these patterns, in turn, has led to an understanding of the course of phonological change. Two main courses of change have been identified: lexical diffusion (Janson, 1983; Wang, 1969) and across-the-board change (Labov, 1981). Of these, lexical diffusion, or the gradual spreading of sound change from lexical item to lexical item in perception and production, is thought to be the more prevalent in fully developed primary languages (e.g., Janson, 1982, 1983; Phillips, 1983), first-language acquisition (e.g., Ingram, 1989; Moskowitz, 1970), and phonological disorders (e.g., Elbert & McReynolds, 1979; Gierut, 1986).

Applied research on these aspects of phonological change enjoys a unique advantage because of the role of clinical treatment and the use of single-subject experimental designs in such treatment. A phonologically disordered child enrolled in treatment can be viewed as an accelerated (albeit manipulated) version of sound change in progress (e.g., Ferguson, 1977; Jakobson, 1941/1968; Ohala, 1980). Treatment sets up the controlled conditions for sound change and then establishes the relevance of these in inducing change. In addition, it is possible to track the course of change longitudinally. The entire process of clinically induced change can be compressed into a relatively short and experimentally manageable time frame, in comparison with studies of historical sound change that take place over decades.

Perhaps, the single-subject design most compatible with this view of the study of phonological change is the alternating treatments design (ATD) (Barlow & Hayes, 1979; Brady & Smouse, 1978; Ellis Weismer & Murray-Branch, 1989; Gierut, 1990, 1991 a; Gierut & Neumann, 1992; Kazdin & Hartmann, 1978; Thompson & McReynolds, 1986; Ward & Bankson, 1989). The ATD directly compares the effects of two experimental treatments on a single subject's learning. The two treatments experimentally establish different conditions for change and, presumably, a subject reacts differently to the treatments by responding differentially on independent probes associated with each. The ATD is well-suited for examinations of the conditions of change because it can isolate variables that induce relative and differential degrees of change within one subject. Moreover, the basic premise of the ATD is consistent with the most commonly observed course of change, namely, lexical diffusion (Gierut, 1991 b). In parallel to natural language change, it is not only possible but likely that phonemic changes will occur gradually and differentially in words of the subject's grammar as motivated by the independent treatments of the ATD. Thus, differential sound change associated with different treatments can single out the linguistic contexts necessary to motivate change and further identify alternate courses of change for continued empirical study.

In applied fields, the course of phonological change has not been as widely studied as the potential conditions that might influence this change. Conditions of change in treatment have been examined from a variety of perspectives and theoretical orientations (e.g., Costello & Bosler, 1976; Elbert & McReynolds, 1978; Elbert, Shelton, & Arndt, 1967; Hodson & Paden, 1983; Leonard & Brown, 1984; Powell, Elbert, & Dinnsen, 1992). Most recently, a research program has been undertaken to evaluate the framework of minimal pair treatment in inducing phonological change (Gierut, 1989, 1990, 1991a; Gierut & Neumann, 1992). Minimal pair treatment simply involves pairing two words that differ by only one phoneme (Crystal, 1985). It is important to note that the way in which these two words (and hence phonemes) differ may be central to issues of phonological change. The phonemes of a minimal pair may differ along at least three dimensions.

A first way in which phonemes may differ is in the number of distinctive oppositions, that is, the number of unique features that differentiate them (Sommerstein, 1977). Phonemes may be minimally opposed, differing by few features, or they may be maximally opposed, differing by many features. For example, the minimal pair pig-big is minimally opposed, differing only by [voice]; whereas, the pair chop-mop is maximally opposed, differing by [sonorant], [nasal], [voice], [anterior], [coronal], [high], [delayed release], and [strident].(n1) Other examples are bend-mend, minimally opposed by the features [sonorant] and [nasal], and thum-gum, maximally opposed by [voice], [anterior], [coronal], [high], [back], [continuant], and [delayed release].

In addition to (and independent of) number of feature differences, phonemes may also differ in the nature of feature distinctions, either major or nonmajor class. Major class features are those that uniquely describe and distinguish among vowels, liquids and nasals, obstruents, and glides. Major class distinctions refer specifically to differences between phonemes involving the features [sonorant], [consonantal], and/or [syllabic]. Major class distinctions were illustrated, above, in the minimal opposition /b/-/m/ (bend mend) and in the maximal opposition /t * This character cannot be converted in ASCII text) /m/ (chop-mop). Nonmajor class distinctions refer to all other place, voice, or manner feature differences between phonemes; hence, both /p/-/b/ (above, a minimal opposition) and /Theta/-/g/ (a maximal opposition) illustrate nonmajor class distinctions.

A third way in which phonemes of a minimal pair may differ is in their relationship to the child's pretreatment grammar. Specifically, the treated phoneme may be unknown to the child, that is, not functional in his or her phonemic inventory. Assuming a relational analysis, the contrasting phone may be its 1:1 corresponding replacement. This comparison is typical of conventional minimal pair treatment (e.g., Weiner, 1981) and establishes an explicit (and potentially homonymous) association between the treated phoneme and the child's grammar. It is also possible to treat an unknown phoneme paired with another known and unrelated phoneme from the child's grammar (Gierut, 1989). In this case, the treated phoneme is again explicitly contrasted with the child's existing system. Finally, two unknown phonemes may simply be compared to each other (Gierut, 1989, 1991a). Here, there is no explicit contrast between treated phonemes and the child's grammar. Thus, minimal pair treatment paradigms may vary in format along at least three linguistic dimensions: the number of distinctive oppositions, the type of oppositions, and the relationship of treated phonemes to the child's pretreatment grammar.(n2)

Thus far, the results of four experimental studies have suggested that these dimensions may indeed be relevant conditions of phonological change (Gierut, 1989, 1990, 1991 a; Gierut & Neumann, 1992). In particular, the results of these studies have demonstrated that

(1) Teaching minimal pairs differing by many features was as effective as or more effective than teaching pairs differing by few features (Gierut, 1989; 1990);

(2) Teaching minimal pairs differing by major class features was as effective as or more effective than teaching pairs differing by nonmajor class features (Gierut, 1990); and,

(3) Teaching minimal pairs comparing two new phonemes was as effective as or more effective than teaching one new phoneme compared to its substituted phone (Gierut, 1991a; Gierut & Neumann, 1992). It is important to note that the effect of teaching minimal pairs with maximal feature differences has been replicated (1,above); however, further replication of teaching major versus nonmajor class features seems warranted (2,above). Also, the effect of presenting minimal pairs relative to the child's pretreatment grammar has only been examined in a limited way (3,above). For example, teaching minimal pairs comparing two new phonemes has not been evaluated relative to teaching one new phoneme paired with another known and unrelated phoneme of the child's system. Furthermore, the impact of systematically combining these factors on sound change has not been studied. Finally, no attention has been given to the possible courses of phonological change that may occur under these different contexts that seem to condition change.

The purpose of this two-part study was to continue and extend this line of research on minimal pair treatments and phonological change. One study was experimental in nature and examined the interplay of a maximal number of feature differences, feature class, and relationship of treated phonemes to a child's grammar in inducing sound change. The second study was descriptive and identified possible types of change in phonologically disordered sound systems. Thus, both the conditions and course of phonological change were investigated herein. Each of four phonologically disordered children was exposed to two different formats of minimal pair treatment within an ATD. The formats represented all logically possible combinations of the variables: feature class and relationship of treated phonemes to a child's grammar. In all treatments, number of distinctions was held constant. Predictably, the contexts identified previously as facilitating optimal phonological change would be experimentally replicated. That is, treatment comparing two previously unknown phonemes + major class differences + maximal feature differences was expected to be the preferred condition of change. Moreover, it was expected that any observed variations from this preferred condition of change would be traceable to the particular course of change followed by a given child and, further, the course of change would be intimately associated with the structure of the child's pretreatment grammar. The findings to emerge have consequences for the efficacy of minimal pair formats of treatment and the predictability of phonological change for children with phonological disorders.

Three boys and one girl participated as subjects of the two studies. Average age of the children was 3 years: 10 months. Children were identified through the diagnostic program at the Speech and Hearing Center at Indiana University. Each child met the following criteria: (a) exclusion of a minimum of six sounds from both phonetic and phonemic inventories as established by a standard generative phonological analysis (see below), (b) normal hearing as determined by a standard audiometric screening (ASHA, 1985), (c) normal oral and speech motor abilities as determined by performance on the protocol developed by Robbins and Klee (1987), (d) no prior clinical intervention, and (e) residency in a monolingual English-speaking family. Additional subject information, shown in Table 1, was obtained but not used to establish eligibility for participation.(n3)

Standard generative phonological descriptions were necessary for both studies. These descriptions were developed for each child prior to treatment based upon extended samples of spontaneous connected speech and the 198-item Phonological Knowledge Protocol (PKP) (Gierut, 1985). The children's phonemic inventories were of central importance. Further, only those phonemes affected by inventory constraints were examined in treatment. Qualitatively, these were phonemes of the ambient language that were never produced or used (correctly or incorrectly) by the child in any word positions or in any morphemes. Quantitatively, these phonemes were produced with 0% baseline accuracy in all nonimitative contexts. Stimulability of phonemes excluded from the inventory was not directly assessed because it is not standard to the development of generative descriptions (cf. Kenstowicz 8 Kisseberth, 1979). It should be noted, however, that none of the children had difficulty imitating phonemes excluded from the inventory when presented in minimal pairs during treatment. Phonemes occurring in each child's pretreatment inventory are displayed in Table 1.

Because the purpose of this study was to replicate and extend the effects of minimal pair treatment formats in inducing sound change, the experimental procedures were identical to those reported elsewhere (Gierut, 1990, 1991a) and therefore will only be summarized briefly herein.

An ATD was used in combination with a staggered multiple baseline across subjects. Each child was exposed to two different formats of minimal pair treatment in the remediation of two different phonemic pairs. Both formats and hence both phonemic pairs were presented in all teaching sessions. Sessions were 60 min in length and were held three times per week. Because both formats were introduced within each session, order of treatment was randomly varied. In each session, the phonemes associated with one treatment format were first introduced, followed by a 10 min nonspeech-related activity, and then the phonemes associated with the second format were presented. Probes were administered throughout to evaluate changes in the child's phonemic inventory.

As in previous studies, several experimental controls were incorporated to guard against the potential effects of multiple treatment interference. These were (a) counterbalanced order of treatment presentation; (b) instructions signaling a switch in treatment; (c) baselines as explicit comparison points of relative treatment effects; and (d) a staggered multiple baseline across subjects design, with the number of pretreatment baselines increasing by one as successive subjects entered the experimental sequence. In this latter case, if multiple treatment interference were observed for a given subject, then the baseline stability of successive subjects would be called upon for a demonstration of experimental control. For a further discussion of these control procedures, the reader is referred to Barlow and Hayes (1979), Gierut (1990,1991a), Kazdin and Hartmann (1978), and McReynolds and Kearns (1983).

Two pairs of phonemes were selected for treatment for each child. Treated phonemes matched the child's experimental assignment and therefore were different across children. The treated phonemes, shown in Table 2, were equated generally along a number of parameters. First, phonemes that were in error relative to the adult target were described by inventory constraints. Second, phonemes illustrated a specific minimal pair format. One treated pair consisted of two phonemes excluded from the child's pretreatment inventory. This format compared two new phonemes to each other, or unknown phoneme: unknown phoneme. The alternate pair consisted of one phoneme excluded from the child's pretreatment inventory and its comparison was a known phoneme from the child's grammar. Here, the format introduced unknown phoneme: known phoneme. Third, phonemic pairs were differentiated by a maximum number of phonological oppositions, following criteria outlined by Archangeli (1988). Maximal oppositions were determined from the ambient specification of features and included only those nonoverlapping and nonredundant distinctions. Finally, phonemic pairs were differentiated by the nature of phonological oppositions, either major or nonmajor class distinctions. The result was that all logically possible combinations pairing two versus one new phoneme and major versus nonmajor class were represented across children, as in Table 3.

Minimal pairs used in the treatments were nonsense words (NSW) assigned lexical meaning within the context of stories. Sixteen NSW minimal pairs served as treatment items for each child, eight pairs for each minimal pair format. The phonetic composition of the NSW pairs was consistent with English phonotactics and contrasting phonemes were always presented in the word-initial position. NSWs were taught using a variety of conceptually based activities: sorting, matching, informal story-telling, and disambiguation of word pairs (cf. Weiner, 1981). Children were also given worksheets, coloring books, and audiocassettes involving the NSW stimuli to complete or to listen to at home. The NSW stimuli were never differentiated from real words in the course of treatment. Moreover, observation and parental report indicated that all children readily accepted and used the NSWs in their everyday vocabulary (see also Carey, 1978).

Consistent with the ATD, the treatment phases were designed to be completed in a relatively short period of time. Also in accord with the ATD, treatment phases were not necessarily intended to guarantee predetermined levels of phoneme mastery or generalization. The aim was simply to induce differential responding as a way of comparing the relative (and not absolute) effectiveness of the two minimal pair treatment formats for a given subject. Toward this end, treatment was delivered in two phases: imitation and spontaneous production. No direct treatment was ever provided for perception or discrimination of phonological contrasts. In both phases, the child was presented with NSW picture pairs. In the imitative phase, the child repeated the clinician's verbal model of the NSWs. Treatment continued until the child maintained 75% accurate imitative production over two consecutive sessions within one minimal pair format or seven consecutive sessions, whichever occurred first. Treatment then shifted to the spontaneous phase with the child producing the NSWs without a model. This phase continued until the child maintained 90% accurate production over 3 consecutive sessions within one minimal pair format or 12 consecutive sessions, whichever occurred first. The time-based criteria for advancing through the treatment phases were followed by all but one child (Subject 19).

A response was judged correct if it was produced as in the ambient language. Children were verbally praised for accurate responding, following a continuous schedule during the imitative phase and an intermittent schedule during the spontaneous phase. If a response was judged incorrect, the clinician presented a verbal model of the NSW target for the child to imitate. No feedback was provided on corrective trials.

Phonological changes that occurred as a result of treatment were evaluated using two probe measures, each specific to a given minimal pair format. Probes consisted of subsets of items from the PKP and sampled all phonemes excluded from the child's pretreatment inventory, both treated and untreated. More specifically, treated phonemes and their voicing counterparts (if in error) were each sampled in three prevocalic, one intervocalic, and three postvocalic exemplars. Each of the remaining phonemes excluded from the child's inventory was sampled in one prevocalic and one postvocalic exemplar. An illustration of the probe items associated with each of the two treatment formats for Subject 19 is shown in the Appendix.

Probe responses were obtained through a picture-naming task that elicited spontaneous productions. Responses were audiorecorded and phonetically transcribed. A phoneme was judged correct if it was produced accurately as in the ambient language. No feedback was provided for probe responses.

A probe was given immediately following its associated format of treatment. Probes were scheduled, on average, every other session. In addition, probes were administered to each child at least two times prior to treatment as baseline measures. As each successive child progressed through the staggered multiple baseline sequence, the number of pretreatment probes was increased by one. Thus, the first subject received two baselines and the last subject received five baselines.

Interjudge reliability was calculated on 10% of the total number of probes administered to each child by the investigator and independent judges (FS, CS) assigned to given children. Judges phonetically transcribed each child's whole word responses on randomly selected probes. Consonant transcriptions were compared point-to-point, with 658 consonants transcribed. Mean interjudge transcription agreement was 89%, with a range of 87 to 91% agreement.

In addition to transcription reliability, point-to-point comparisons were made between the judges' decisions on response accuracy. It is important to note that in only 28 cases judges differed in their assessment of whether or not a target phoneme was produced accurately. That is, only 28 of the 658 consonants transcribed (i.e., 4%) involved relevant differences between the judges' transcriptions and the determination of phoneme accuracy.

Results are presented for individual children with regard to the relative and differential phonological changes induced under the two minimal pair treatment formats. The discussion is limited to within-subject comparisons of phonological change because the crossing and counterbalancing of factors prohibited across-subject comparisons. Also, across subject comparisons regarding performance on a given treated phoneme were not possible because different feature distinctions were manipulated in each case.

In this study, the differential and relative effects of treatment were determined only from probe data, as measured in spontaneously produced real words. There are, of course, other ways to evaluate relative phonological change, including, for example, accuracy of production of treated NSWs (Gierut & Neumann, 1992), number of treatment trials to criterion, and number of treatment sessions. In this study, these alternative metrics did not seem applicable. Production of NSWs did not distinguish among the conditions. Moreover, the ATD design dictated that the number of treatment trials and sessions be identical across conditions for a given subject. It may be important in subsequent studies to assess relative phonological change using both probe and treatment data.

Differential phonological change was established for treated and untreated phonemes. Change in treated phonemes was defined in two ways: (a) the highest percentage score on any given probe and (b) the final probe percentage score in each of the treatments (cf. Ellis Weismer & Murray-Branch, 1989; Gierut, 1990, 1991a; Winner & Elbert, 1988). Change in untreated phonemes was defined by a 10% mean probe difference score (cf. Elbert, Dinnsen, & Powell, 1984). Differential change in untreated phonemes was observed for only one child (Subject 10) and therefore will be discussed in this case only. This is not to say that changes in untreated phonemes did not occur for the other children of this study, only that these changes did not differentiate among the treatment formats following the 10% criteria defined above.

Summary data regarding relative phonological change for all children are plotted in Figure 1. Longitudinal probe data are also shown in Figures 2, 3, 4, and 5. These results are discussed specific to the linguistic factors that were manipulated: the relationship of treated phonemes to the child's pretreatment grammar and the nature of feature distinctions. For this reason, data are not reported in a conventional multiple baseline sequence, by length of the baseline period.

The learning patterns of Subjects 10 and 19 will be considered together because, for these children, the only linguistic dimension that was manipulated was the pairing of new phonemes relative to the pretreatment grammars. As was reported in Tables 2 and 3, these children were taught two new phonemes in one of the minimal pair formats, and one new phoneme paired with a known phoneme in the other format. For a given child, however, the number and nature of feature distinctions was held constant across treatments.

For Subject 10, treatment of two previously unknown phonemes involved the /k/-/l/ comparison. In the alternate condition, one new phoneme /f/ was compared to the known phoneme /n/ from this child's grammar. Treated phonemes in both pairs differed by the major class feature [sonorant]. As seen in the left panel of Figure 1 (also Figure 2), Subject 10 exhibited the highest probe performance when presented with two unknown phonemes. In this treatment format, the child produced both /k/ and /l/ with 43% accuracy; whereas, A/ was produced with 29% accuracy. Subject 1 0's final probe performance reflected this same general pattern. The greatest probe accuracy was observed for /k/, treated in the format comparing two unknown phonemes (Figure 1, right panel). It is important to note that Subject 10 was also the only child who evidenced differential change in untreated phonemes. Specifically, /g/ was added in that treatment condition comparing two new phonemes. No untreated phonemes were added in the other condition. Thus, treated and untreated phonemes seemed to undergo parallel changes for this child. A similar pattern of learning was observed for Subject 19. This child's treatment compared two unknown phonemes /dz/ and /Theta/ within one format and, in the alternate /v/ (an unknown phoneme) was compared to the known phoneme /V. Both treatment pairs differed only by nonmajor class features. For Subject 19, differential changes in treated phonemes were unequivocally observed on the final probe (Figure 1, right; also Figure 3). This child produced /dz/ and /Theta/ with 71% and 14% accuracy, respectively; in contrast, /v/ was never produced accurately. The child's highest probe responding occurred for /dz/, treated in the format comparing two new phonemes (Figure 1, left).

For both subjects, then, treated phonemes appeared to change differentially under the two formats of minimal pair treatment. Probe results suggested that teaching two new phonemes was as effective as or more effective than teaching one new phoneme compared to a known aspect of the child's grammar. It seemed that when the child was exposed to new, previously unknown aspects of the target system during treatment, relatively greater phonological changes occurred. An alternative treatment that made reference to the child's existing grammar in an explicit and comparative way did not appear to induce phonological change to the same degree in this study. This result suggests that it may be unnecessary to teach children that newly reamed phonemes are in some way related to other existing phonemes in their grammar by setting up explicit minimal pair comparisons (Gierut, 1991 a; Johnston & Smith, 1989; Kornfeld & Goehl, 1974; Locke,1979; Priestly, 1980; Weiner & Ostrowski, 1979). Perhaps children are capable of integrating new phonological information into their grammars without this kind of direct instruction.

It is important to note that this finding--that a treatment format comparing two new phonemes induces greater phonological change--was observed under a number of experimental manipulations in this and other ATD studies (Gierut, 1990,1991 a; Gierut & Neumann, 1992). This result appears to be quite robust because it has been obtained regardless of the nature of the distinctions being taught. As in this study, both major (Subject 10) and nonmajor class differences (Subject 19) among new phonemes resulted in greater probe performances. Further, the result held regardless of the number of distinctions being taught. In this study, maximal feature differences among new phonemes prompted greater probe performances, but, as reported previously (Gierut, 1991 a), few feature differences did also. Finally, the result was consistent regardless of the relationship of treated phonemes to the child's grammar. In this study, teaching two new phonemes was relatively more effective in inducing sound change than teaching one new phoneme compared to known aspects of the grammar. A comparable finding emerged when treatment involving two new phonemes was evaluated relative to conventional minimal pair treatment (Gierut, 1991a). Thus, the evidence implies that treatment comparing two new phonemes may be an important structural variable in conditioning phonological change.

The results also suggest that treatment involving two new phonemes may motivate change in untreated phonemes, but perhaps only when pairs differ by a major class feature, as in the case of Subject 10 (see also Gierut, 1990, for comparable findings). Here, major class in combination with two new phonemes may have been the appropriate context for inducing phonological change.

To summarize, the learning patterns of these two subjects again identify the relationship of treated phonemes to the child's grammar and the nature of distinctions between phonemes as possible factors to consider in treatment and phonological change. The potential relationship between these two variables will be examined in the learning patterns of Subjects 11 and 12, where both the number of new phonemes and the nature of distinctions between phonemes were manipulated in the treatment formats.

In the case of Subject 11, one format of treatment compared two previously unknown phonemes /g/ and /s/, differing only by nonmajor class features. For this same child, the alternate format compared the unknown phoneme /* This character cannot be converted in ASCII text)/ to the known phoneme /n/ from the child's grammar. This pair differed in major class among many other distinctions. Thus, treatment pairing Unknown: Unknown phonemes + nonmajor class + maximal feature differences was evaluated relative to treatment pairing Unknown: Known phonemes + major class + maximal feature differences for this subject (Tables 2 and 3).

Subject 11 evidenced essentially equivalent probe performance in the two treatment conditions (Figures 1 and 4). The child's highest probe performance reached 29% accuracy of treated phonemes in both formats. Similarly, final probe performance was comparable across treatments. The learning pattern of Subject 11 suggested no differential responding between minimal pair formats comparing two new phonemes + nonmajor class and one new phoneme + major class. For this child, there appeared to be a trade-off between the impact of treated phonemes relative to his grammar and the nature of distinctions. The particular combination of factors within each treatment format may have established equivalent or counterbalanced linguistic contexts for motivating phonological change. From these data, the relative strength of these variables in change could not be isolated and seems to warrant further study.

Subject 12 was presented with formats of treatment that were the opposite of Subject 11's. That is, Subject 12 received treatment on two new phonemes, /dz/ and /I/, differing in major class features and one new phoneme /g/ compared to a known phoneme /g/, differing only by nonmajor class features. For this child, Unknown: Unknown phonemes were paired with major class + maximal feature differences in one of the treatments, and Unknown: Known phonemes with nonmajor class + maximal feature differences in the other (Tables 2 and 3).

Differential phonological changes were unique for Subject 12. This child's probe performances seemed to support preferential responding for treatment involving one new phoneme and nonmajor class distinctions (Figures 1 and 5). The highest probe score occurred for /g/, produced with 57% accuracy, which exceeded performance on /dz/ and /I/ at 43% and 29% accuracy,.respectively. Subject 12's final probe performance was essentially equivalent across treatments, although there may have been a slight advantage for the format pairing one new phoneme + nonmajor class given 0% accurate production of /I/ in the alternate condition.

Contrary to the established patterns, a minimal pair format comparing two new phonemes + major class + maximal distinctions did not result in relatively greater phonological change for Subject 12. Instead, a combination of presumably less influential factors--one new phoneme + nonmajor class + maximal distinctions--led to more extensive phonological change. The excrescent patterns of Subject 12's reaming are in direct conflict with the body of data on minimal pair treatment formats that condition phonological change (Gierut, 1989, 1990, 1991a; Gierut & Neumann, 1992). The question arises then for Study Il: What variables besides the linguistic format of treatment may have induced phonological change for Subject 12?

One possible source of individual differences is the interaction (or interference) between format of treatment and the child's pretreatment grammar (Dinnsen & Elbert, 1984; Elbert et al., 1984; Fey & Stalker, 1986; Gierut, 1991a; Gierut, Elbert, & Dinnsen, 1987; Tyler, Edwards, & Saxman, 1987, 1990). One current hypothesis is that children may ignore the specific treated phonemes or even the ambient feature specification of those phonemes as input in treatment. Instead, it has been speculated that children focus on only those feature specifications of treated phonemes that directly affect the distinctive and/or redundant aspects of their own grammars (cf. Dinnsen, 1992; Dinnsen, Chin, Elbert, & Powell, 1990; Gierut, 1991a; Ingram, 1987, 1990b; Jakobson, 1941/1968). Perhaps, as in the case of Subject 12, treatment can be claimed effective and predictive of learning patterns only as it relates to a given child's unique, independent phonological system. For this reason, the phonologies of the children of the study were reexamined with regard to the distinctive and redundant properties of the system and relative to the formats of minimal pair treatment and the course of phonological change. The purpose of the second study then was to identify possible types of phonological change in disordered systems as a means of providing a possible account of individual differences.

Descriptive statements of the distinctive oppositions, along with redundancy statements limiting the range of phonemes associated with these oppositions, were developed for each child's phonemic inventory. The contrastive (i.e., distinctive) and predictable (i.e., redundant) aspects of phonemic inventories are two of the most basic properties of any sound system of language (Chomsky & Halle, 1968; Jakobson, 1941/1968; Trubetzkoy, 195811969). Together, these specify the complete and unique description of every phoneme in the language. They further define what is and what is not a possible phoneme in that language. Such descriptions of the distinctive and predictable properties of sound systems have been at the center of all studies of language, regardless of theoretical orientation.

To illustrate these concepts, consider the phonemic status of the labiodental fricatives /f, v/ in the fully developed sound system of English. A description of the distinctive nature of /f, v/ would include that set of features that contrastively differentiates these phonemes from all others in the language, and that further restricts the occurrence of other phonemes that are similar but not present in the language. A description of the distinctive composition of /f, v/ would therefore minimally include the features [consonantal], [sonorant], [continuant], and [coronal]. Specifically, the feature [+consonantal] distinguishes /f, v/ from vowels, glides, and /?/. The feature [-sonorantj further distinguishes /f, v/ from nasals and liquids. The feature [+continuanq differentiates /f, v/ from stops and affricates. Finally, the feature [-coronal] differentiates /f, v/ from all other fricatives in English. Any other features that describe /f, v/ would be completely predictable (i.e., redundant) from this set. For instance, the feature [nasal] is redundant because all phonemes that are specified as [+consonantal] and [-sonorant] must also be [-nasal]. Similarly, [delayed release] is a redundant property because if a consonant is [+continuant], it will be predictably [+delayed release]. Stridency is another redundant aspect in this example because all noncoronal continuant consonants in English are [+strident]. Here, [-coronal! and [+continuant] taken together predict [+strident]. Thus, the features [consonantal], [sonorant], [continuant], and [coronal] are distinctive for /f, v/, but the features [nasal], [delayed release], and [strident] among others are redundant. This description of the distinctive and redundant properties of /f, v/ also places further limitations on the kinds of fricatives that can occur in English. Because the grammar of English specifies that all continuant noncoronal consonants must be strident, other nonstrident fricatives that meet this same general description are impemmissible in the language. Therefore, /Phi, Beta, x, v/ are not phonemes of English. Other examples of the distinctive and redundant properties of fully developed sound systems in languages of the world have been detailed elsewhere (e.g., Archangeli, 1988; Avery & Rice, 1989; Mester & Ito 1989; Pulleyblank, 1986, 1988a, 1988b; Trubetzkoy, 1958/1969). Applications to normal and phonologically disordered sound systems have also been reported (Dinnsen, 1992; Dinnsen et al., 1990; Gierut & Dinnsen, 1988; Ingram, 1989, 1990a, 1990b).

In this study, the segment-internal redundancies of each child's pretreatment grammar were of particular interest because these captured featural restrictions on phonemes that were contrastive in the child's grammar. Redundancies were initially developed as independent prose statements, and these were then diagrammed to illustrate how each restriction of a given child's grammar was related to every other. The following guidelines were employed in the development of redundancy statements and diagrams. These guidelines are consistent with contemporary theoretical linguistic frameworks such as feature geometry and underspecification, among others (Archangel), 1988; Clements, 1985, 1988; Halle, 1988; Ladefoged & Halle, 1988; Sagey, 1986; Stevens & Keyser, 1989), and have been applied in other research on phonological acquisition and disorders (Chin & Dinnsen, 1991, in press; Dinnsen, 1992; Dinnsen et al., 1990; Guerut & Dinnsen, 1988; Ingram, 1990a, 1990b). First, some distinctive features were taken to be more basic in the organizational structure of the grammar than others. This does not also mean that these features necessarily reflected the acquisition sequence. Major class features were assigned a higher-order ranking than features of manner or place whenever possible. Second, the fewest number of features were taken to be distinctive, with all others predictable or redundant. Third, certain features implied certain others, so that it was possible for a given set of redundancies to describe different phonemic inventories because the distinctive properties of these inventories differed. Fourth, redundancies of a given phonological system were taken to be implicationally related and therefore were traceable in a single unique diagram. In following these guidelines, there was very limited variation in the way that redundancy statements and diagrams could be constructed for any set of phonological data.

Redundancy statements for each child's phonology are summarized in Table 4 and diagrammed in Figure 6. The diagrams can be interpreted by following the classificatory branches down from top-most to terminal nodes in conjunction with the prose statements of Table 4. In Figure 6 left-branching elements of feature nodes are positive in value; right-branching are negative. Solid lines track phonemes permitted in the grammar; dashed are cases of those excluded. To illustrate, for Subject 10, a trace of the redundancy diagram from top to right terminal node limits phonemes that are l+consonantal, -sonorant, -continuant, -delayed release] to [+anterior] (i.e., /p, b, t, d/); those phonemes that meet this same description but are [-anterior] (i.e., /k, g/) are not in the phonemic repertoire of this child. Each redundancy statement of each child can be traced in this same way, with two additional clarifications. First, the redundancy statements are limited to descriptions of the consonantal repertoires of each child's phonemic inventory. Hence,[-consonantal] phonemes, such as vowels or glides, are not described herein. Second, if phonemes are fully contrastive (i.e., distinctive) in the child's system, they cannot also be redundant. Therefore, redundancy statements will not capture such distinctions as, for example, voicing in affricates for Subjects 10 and 11.

Two relevant observations emerged from the redundancy diagrams of Subjects 10,19, and 11. Recall that the learning patterns of these three children were consistent with the established conditions of change. First, from the phonologies of all three children, it was possible to arrange top-most redundancies around major class features, [consonantal] and [sonorant], as shown in Figure 6. Then, within the group of nonsonorant consonants, restrictions involved a further dimension of manner, namely, [continuant]. Thus, the features [consonantal], [sonorant], and [continuant] appeared to be relatively higher-order in the organizational schema of restrictions on these children's pretreatment grammars. This hierarchical organization is generally consistent with the feature geometries of fully developed primary languages, where major class and manner emerge as more basic distinctions and independent of place (Clements, 1985; Halle, 1988; Lauttamus, 1990; Sagey, 1986).

Second, for Subjects 1 0, 19, and 11 , treatment was aimed at expanding the contraposit (or opposite) value of a feature that was already available redundantly, but not yet distinctively, in the child's grammar. This potentially expanded feature was always associated with the lowest level terminal node. This was true regardless of the minimal pair format of treatment used to motivate the change. For instance, referring to Figure 6, the pretreatment grammar of Subject 10 categorized all sonorant consonants as [+nasal]. After the introduction of /l/ in treatment, sonorant consonants that were also [-nasal] were potentially permissible in the grammar. For this same child, the grammar further categorized all continuant nonsonorant consonants as [+cornal]. Treatment of /f/ introduced the contraposit value [-coronal]. Continuing the example, the child's grammar categorized all noncontinuant nonsonorant consonants that were also nondelayed release as [+anterior]. Teaching /k/ illustrated that phonemes meeting the same general description could be [-anterior] as well. Similar illustrations are evident for Subjects 11 and 19 by comparing treated phonemes listed in Table 2 with terminal feature nodes of redundancy diagrams in Figure 6.

Thus, it appeared that the phonologies of these three children were "primed" to receive and incorporate new phonemic information provided in treatment to low-level terminal feature nodes. If, in the course of treatment, new phonemic information were learned, then the corresponding redundant feature values were no longer predictable and thus became distinctive in the child's grammar. This is not to say that all treated phonemes at terminal nodes would necessarily be learned to the same degree of accuracy because the nature of the featural restrictions affecting them were quite different within and across the children's phonological systems. For these children then, treatment may have served as the necessary trigger of phonological change, advancing their pretreatment grammars closer to the ambient target. In these cases, the structural organization of the grammar and the formats of minimal pair treatment were compatible. The grammar and treatment provided the appropriate setting for expansion of the phonological system.

The grammar of Subject 12, on the other hand, presented an entirely different case. This child maintained that all consonants must be [+anterior]. Notice that this restriction was global, affecting every consonant of the system regardless of sonority value. Consequently, in constructing the redundancy diagram, it was not possible to hierarchically organize the child's grammar around the features [consonantal] and [sonorant]. Instead, top-most redundancies had to be [consonantal] and [anterior] (Figure 6). Of further importance was the fact that a place feature [anterior] also had to be assigned a hierarchical position above manner distinctions in the grammar. Thus, Subject 12 apparently considered restrictions of place as more crucial than those involving either major class or manner. It remains to be determined whether this kind of structural organization of the grammar is generally representative of earlier stages of acquisition. However, restrictions of this type did appear to have subsequent consequences for treatment and phonological change.

In particular, treatment may not have served as a trigger for expansion of the phonemic inventory for Subject 12. Basic differences in the structure of this child's grammar prevented her from accommodating input provided in treatment by additions to low-level terminal nodes. Recall that the child was introduced to the phonemes /g/ and /dz/ in treatment.(n4) Both targets are nonanterior, a distinction that would have potentially disrupted the highest order, most global restriction on the grammar. Because the input provided in treatment violated the basic structure of her grammar, Subject 12 was forced to follow an alternative course of phonological change. One of two courses of change was possible.

On the one hand, Subject 12 could have performed a simple copying of contrasts, with increases in the phonemic inventory occurring as a mirror-image to the pretreatment organization. In this case, she may have refined nonanterior aspects of the grammar in parallel with anterior aspects so that the redundancy diagram would be symmetrical on both sides [+ anterior]. This type of phonological change has been hypothesized in other work and has been labeled "node copying" (Gierut & Dinnsen, 1988). An illustration of node copying can be developed from Figure 2. For nonanterior consonants, a [sonorant] distinction would provide for the differentiation of /n, r/ from /k, g, [], t[], dz/. Then, tracing [+sonorant], a [nasal] distinction would further differentiate /al from /r/. Within [-sonorant], a [continuant] distinction would differentiate /[]/ from /k, g, t[], dz/, and so on. The end result of node copying would be that both the left and right sides of Subject 12's redundancy diagram would be identical.

On the other hand, Subject 12 could have completely reorganized her pretreatment grammar. Because treatment directly challenged existing structural constraints on the grammar, the hierarchy of redundant features would have to be reranked and organizational patterns revised. Central to this, reorganization would necessitate a rearranging of sonority and continuancy as higher order featural nodes. Also, there would need to be a relaxing of the [anterior] constraints on the grammar and a demoting of place (i.e., [anterior]) to a lower node in the schema.

Thus, the reaming task facing this child appeared to be very different from that of the other children of this study. Rather than expansion of existing contrasts, a copying of contrasts or a complete reorganization of the phonological system was apparently necessary. It will be important in subsequent studies to track such children longitudinally in order to determine which course of change is followed and whether phonological change is retained. Perhaps, the atypical conditions of phonological change observed for this child are the outcome of precisely the interplay between the pretreatment grammar and treatment format. It may be that the impact of treatment on sound systems may be constrained by the structural nature of these systems themselves. With this potential interaction in mind, an integrated summary of the conditions and course of phonological change is developed in the next section.

Together, the results of five experimental studies have identified three factors that seem to influence phonological change in minimal pair treatment (Gierut, 1989, 1990, 1991 a, herein; Gierut & Neumann, 1992). These are (a) the number of feature differences between treated phonemes, (b) the nature of these feature differences, and (c) the relationship of treated phonemes to the child's pretreatment grammar. These factors can be combined in any number of logically possible ways, with a hierarchy of minimal pair formats emerging, as in Figure 7. This hierarchy plots combinations of variables that may result in more (or less) phonological change. The hierarchy arranges only factors relative to a maximal number of feature distinctions, but a similar relationship also seems to result from treatment of few feature differences (Gierut, 1990, 1991a). Treatment of few feature differences, however, may generally lead to less phonological change.

These conditions of phonological change may hold possible clinical application. For example, it is hypothesized that minimal pair treatment introducing two new phonemes + major class + maximal feature distinctions will lead to the greatest change in a child's phonemic inventory. Other treatment formats combining one new phoneme + major class + maximal feature differences, or two new phonemes + nonmajor class + maximal feature differences may induce relatively less change, but these formats may also be comparable to each other. Lastly, treatment involving one new phoneme + nonmajor class + maximal feature differences may result in the least change in phonemic inventories.

It is noteworthy that the proposed conditions of change seemed to obtain regardless of the specific phonemes being taught or the expected sequence of acquisition of these phonemes (Gierut, 1991a; Gierut & Neumann, 1992). Moreover, it appears that the predicted hierarchical effects may have psychological reality given the striking parallels between these results in the linguistic domain and other findings on the development and organizational structure of categories in the cognitive domain (e.g., Au & Laframboise, 1990; Carey, 1985; Keil, 1988; Markman, 1989; Tversky, 1977). The presumed psychological reality of these results may provide further independent support of their validity.

At this time, however, the predicted conditions of change must be restricted to phonemes excluded from a child's pretreatment inventory. It will be important to continue the evaluation of these hypotheses relative to other types of phonological patterns. Further replication of the hierarchical effects using other research paradigms also seems in order. For example, in the present study, there was no replication of the manipulation and crossing of variables; only one subject was assigned to each of the four experimental conditions. Also, the series of ATD studies conducted thus far has been designed to examine systematically relative differences between treatment formats and to compare these effects within subjects. The resulting hypotheses may now need to be evaluated using more conventional experimental designs that rely on absolute phoneme mastery and across-subject comparisons of learning patterns, as within the multiple baseline design.

The accuracy of the proposed conditions of change may also be restricted further by the structure of a child's pretreatment grammar and the possible courses of change to be induced in treatment. Several questions arise in this regard for further empirical study. For example, two possible types of pretreatment grammars were suggested. One grammar (Subjects 10, 19, and 11) appeared to be consistent with fully developed primary languages, ranking major class and manner features as hierarchically more important than place of articulation; the other grammar (Subject 12) assigned place as superordinate in the geometry. Perhaps, these structural differences in grammars may be important diagnostic considerations in differentiating among children with phonological disorders.

As another example, three distinct types of phonological change were proposed: expansion, node copying, and reorganization of the grammar. Change through expansion maintains the basic structure of the pretreatment grammar, with positive changes induced in treatment aimed at low levels of the organizational scheme. Change through node copying builds on the grammar by applying parallel structure and restrictions to other components of the system. Change through reorganization requires completely altering the structure of the pretreatment grammar, if a target-like system is ever to be achieved. Perhaps, certain types of change may be more difficult to induce in treatment, with some grammars being highly resistant to structural modification. The generality of restrictions on the grammar may be one factor to consider in this regard. Conceivably, different teaching methods may even be required to induce and facilitate these different types of change in disordered systems.

In conclusion, the results of this study underscore the importance and potential interactions of a child's grammar and treatment in phonological change. Linguistic formats of treatment may only achieve expected or predicted results as applied to particular kinds of grammars. This is not to say that it will be possible to predict precisely which phonemes will change and to what degree in treatment, but just that seemingly unpredictable or unusual patterns of reaming may be directly traceable to the particular course of change followed by a given child and that, in turn, may be intimately linked to the structure of the child's pretreatment grammar.

This research was supported in part by grants from the National Institutes of Health (DC 00433) and the Biomedical Research Support Grant Program (S07 RR 7031 K) to Indiana University, Bloomington. I would like especially to thank Steve Chin, Phil Connell, Stuart Davis, and Dan Dinnsen for many productive discussions throughout this project. Carol Stoel-Gammon, Jan Ingham, and an anonymous JSHR reviewer also provided valuable comments. The Clinical Faculty of the Department of Speech and Hearing Sciences at Indiana University assisted with subject identification; Melissa Knoll provided clinical treatment; Faith Salesin and Christina Simmerman served as reliability judges; Christina Simmerman also assisted with data management and analysis.

(n1) The Chomsky-Halle feature system was used herein because it meets the four criteria of an adequate feature system as outlined by Kenstowicz and Kisseberth (1979,pp. 241-242). Alternative feature systems could have been used; and although the details would likely differ, the same general points could be made.

(n2) These three factors are defined as linguistic in nature, as opposed to, for example, cognitive or motor components, because they are substantive properties that are fundamental to descriptions and characterizations of any human language.

(n3) Subjects were part of a larger project and were assigned subject numbers according to their order of entry; hence, Subjects 10, 19, 11, and 12 participated in these studies.

(n4) Subject 12 also received treatment on /l/, a low-level terminal node of the grammar.

Legend for Table

[A] - (This character cannot be converted in ASCII text)
[B] - (This character cannot be converted in ASCII text)
[C] - (This character cannot be converted in ASCII text)

           Age      Phonemic                  GFTA[a]
Subject  (yr:mon)   inventory                 (errors)

10       3:06     m,n,[A],w,j,h                 43
                   p,b,t,d,s,z,[B],t [B],d[C]

19       3:08     m,n,[A],w,j,h                 54
                   p,b,t,d

11       4:12     m,n,[A],w,j,h                 48
                   p,b,t,d,f,v,t [B],d [C]

12       3:08     m,n,w,h                       57
                   p,b,t,d,f

              TELD[b]     PPVT[c]        Leiter[d]
            (language    (standard     (intelligence
Subject     quotient)      score)        quotient)

10            108           98            131

19            102           99            105

11             96           92            110

12            106          100            144

[a]Goldman-Fristoe Test of Articulation (Goldman & Fristoe,
1986).[b]Test of Early Language Development (Hresko, Reid,
& Hammill, 1981). [c]Peabody Picture Vocabulary Test-Revised
(Form L, Dunn & Dunn, 1981). [d]Leiter International
Performance
Scale (Arthur Adaptation, Levine, 1986).

Legend for Table

A - (This character cannot be converted in ASCII text)
B - (This character cannot be converted in ASCII text)

Subject                        10       19        11    12

Format of one minimal pair
  Number of new phonemes      2        2           2     2
  Nature of feature
    distinctions            Major    Nonmajor   Nonmajor  Major
  Unknown: Unknown
    phonemes                k:l    d[B]:Theta   g:s     d[A]:l

Format of alternate
minimal pair
  Number of new phonemes      1        1           1       1
  Nature of feature
    distinctions           Major    Nonmajor     Major   Nonmajor
  Unknown: Known
    phonemes               f:n      v:t          [B]:n    g:f

                                Format of one minimal
Minimal pair treatments              pair

Phoneme1 of the air       Excluded from a child's
                                  pretreatment inventory

Phoneme2 of the air       Excluded from a child's
                                  pretreatment inventory

Number of feature differences   Many
                                  Mean = 6 features
                                  Range = 5 to 8 features

Major class differences         Subject 10
                                Subject 12

Nonmajor class differences      Subject 19
                                subject 11

                                Format of alternate minimal
Minimal pair treatments             pair

Phoneme1 of the air       Excluded from a child's
                                  pretreatment inventory

Phoneme2 of the air       Known phoneme from a child's
                                  pretreatment inventory

Number of feature differences   Many
                                  Mean = 6 features
                                  Range = 5 to 8 features

Major class differences         Subject 10
                                Subject 11

Nonmajor class differences      Subject 19
                                Subject 12

Legend for Table

A - (This character cannot be converted in ASCII text)
B - (This character cannot be converted in ASCII text)
C - Theta
D - (This character cannot be converted in ASCII text)
E - (This character cannot be converted in ASCII text)

Subject  Redundancy rules             Permitted    Excluded

10     [+ cons] arrow right [+ nas]    mn A      lr
       [+ son ]
             (Sonorant consonants are nasal.)

        [+ cons]             [+ cor ]   sz B      fv C D
        [- son ] arrow right [+ stri]
        [+ cont]
            (Nonsonorant continuant consonants are coronal and
            strident.)[a]

        [+ cons   ]                     pbtd     kg
        [- son    ] arrow right [+ ant ]
        [- cont   ]
        [- del rel]
            (Nonsonorant noncontinuant consonants that are also
            nondelayed release are anterior.)

11     [+ cons] arrow right  [= nas]   mn A      lr
       [+ son ]
            (Sonorant consonants are nasal.)

        [+ cons ]                       fv       C D szf
        [- son  ] arrow right [- con]
        [+ cont ]
            (Nonsonorant continuant consonants are noncoronal.)

        [+ cons  ]                     pbtd       kg
        [- son   ] arrow right [+ ant]
        [- cont  ]
        [-del rel]
            (Nonsonorant noncontinuant consonants that are also
            nondelayed release are anterior.)

19     [+ cons ] arrow right [+ nas]    mn A      lr
       [+ son  ]
            (Sonorant consonants are nasal.)

        [+ cons ] arrow right [- cont ]  pbtd    kgt B d E
fv C D
        [- son  ]             [+ ant  ]          sz B
            (Nonsorant consonants are noncontinuant and
            anterior[a])
 12     [+ cons ] arrow right [+ ant ]  mnpbtdf  A rkg B t B d E
            (All consonants are anterior.)

        [+ cons ]             [+ nas ]  mn       l
        [+ son  ] arrow right
            (Sonorant consonants are nasal.)

        [+ cons ]             [- cor ]   f       v C D sz
        [- son  ]             [- voi ]
        [= cont ] arrow right
            (Nonsonorant continuat consonants are noncoronal
            and voiceless.)[a]

[a]When more than one feature is predicted in the output of a
single redundancy statement, no precedence among features is
claimed.

GRAPHS: FIGURE 1. Highest and final probe performances of each subject under each of the two experimental treatment formats. Open bare reflect treatment of one new phoneme; hatched bars reflect treatment of two new phonemes.

GRAPHS: FIGURE 2. Baseline and longitudinal probe data for treated phonemes In each format of treatment for Subject 10.

GRAPHS: FIGURE 3. Baseline and longitudinal probe data for treated phonemes In each format of treatment for Subject 19.

GRAPHS: FIGURE 4. Baseline and longitudinal probe data for treated phonemes In each format of treatment for Subject 11.

GRAPHS: FIGURE 5. Baseline and longitudinal probe data for treated phonemes In each format of treatment for Subject 12.

DIAGRAMS: FIGURE 6. Redundancy diagrams of each subject's pretreatment grammar. Left-branching nodes are positive In feature value, and right-branching nodes are negative. Solid lines trace phonemes permitted In me grammar, whereas dashed lines trace phonemes excluded from the phonological system.

DIAGRAM: FIGURE 7. Hierarchy of minimal pair treatment formats predicting relating degrees of phonological change In the phonemic inventories of disordered systems.

References

ASHA Committee on Audiologic Evaluation. (1985,May). Guidelines for identification audiometry. Asha, pp. 49-52.

Archangeli, D. (1988). Aspects of underspecification theory. Phonology, 5, 183-207.

Au, T. K., & Laframbolse, D. E. (1990). Acquiring color names via linguistic contrast: The influence of contrasting terms. Child Development, 61, 1808-1823.

Avery, P., & Rice, K. (1989). Segment structure and coronal underspecification. Phonology, 6, 179-200.

Barlow, D. H., & Hayes, S. C. (1979). Alternating treatments design: One strategy for comparing the effects of two treatments in a single subject. Journal of Applied Behavior Analysis, 12, 199-210.

Brady, D. O., & Smouse, A. D. (1978). A simultaneous comparison of three methods for language training with an autistic child: An experimental single case analysis. Journal of Autism and Childhood Schizophrenia, 8, 271-279.

Carey, S. (1978). The child as word learner. In M. Halle, J. Bresnan, & G. Miller (Eds.), Linguistic theory and psychological reality (pp. 265-293). Cambridge, MA: MIT.

Carey, S. (1985). Conceptual change in childhood. Cambridge, MA: MIT.

Chen, M. (1972). The time dimension: Contribution toward a theory of sound change. Foundations of Language, 8, 457-498.

Chen, M., & Wang, W. S-Y. (1975). Sound change: Actuation and implementation. Language, 51, 255-281.

Chin, S. B., & Dinnsen, D. A. (1991). Feature geometry in disordered phonologies. Clinical Linguistics and Phonetics, 5, 329-337.

Chin, S. B., & Dinnsen, D. A. (in press). Consonant clusters in disordered speech: Constraints and correspondence patterns. Journal of Child Language.

Chomsky, N., & Halle, M. (1968). The sound pattern of English. New York: Harper 8 Row.

Clements, G. N. (1985). The geometry of phonological features. Phonology, 2, 225-252.

Clements, G. N. (1988). Towards a substantive theory of feature specification Proceedings of NELS (Northeastern Linguistic Society), 18, 79-93.

Costello, J., & Bosler, S. (1976). Generalization and articulation instruction. Journal of Speech and Hearing Disorders, 41, 359373.

Crystal, D. (1985). A dictionary of linguistics and phonetics (2nd ed.). Oxford: Basil Blackwell.

Dinnsen, D. A. (1992). Variation in developing and fully developed phonetic inventories. In C. A. Ferguson, L. Menn, 8 C. Stoel-Gammon (Eds.), Phonological development: Models, research, and implications (pp. 191-210). Parkton, MD: York.

Dinnsen, D. A., Chin, S. B., Elbert, M., & Powell, T. W. (1990). Some constraints on functionally disordered phonologies: Phonetic inventories and phonotactics. Journal of Speech and Hearing Research 33 28-37.

Dinnsen, D. A., & Elbert, M. (1984). On the relationship between phonology and learning. In M. Elbert, D. A. Dinnsen, 8 G. Weismer (Eds.), Phonological theory and the misarticulating child, ASHA Monographs 22, 59-68.

Dunn, L. M., & Dunn, L. M. (1981). Peabody picture vocabulary test-Revised. Circle Pines, MN: American Guidance Service.

Elbert, M., Dinnsen, D. A., & Powell, T. W. (1984). On the prediction of phonologic generalization learning patterns. Journal of Speech and Hearing Disorders, 49, 309-317.

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

Elbert, M., & McReynolds, L. V. (1979). Aspects of phonological acquisition during articulation training. Journal of Speech and Hearing Disorders, 44, 459-471.

Elbert, M., Shelton, R. L., & Arndt, W. B. (1967). A task for education of articulation change. Journal of Speech and Hearing Research, 10, 281-2BB.

Ellis Weismer, S., & Murray-Branch, J. (1989). Modeling versus modeling plus evoked production training: A comparison of two language intervention methods. Journal of Speech and Hearing Disorders, 54, 269-281.

Ferguson, C. A., (1977). New directions in phonological theory: Language acquisition and universals research. In R. W. Cole (Ed.), Current issues in linguistic theory (pp. 247-299). Bloomington: Indiana University Press.

Fey, M., & Stalker, C. H. (1986). A hypothesis-testing approach to treatment of a child with an idiosyncratic (morpho)phonological system. Journal of Speech and Hearing Disorders, 51, 324-336.

Geirut, J. A., (1985). On the relationship between phonological knowledge and generalization reaming in misarticulating children. Bloomington: Indiana University Linguistics Club.

Gierut, J. A., (1986). Sound change: A phonemic split in a misarticulating child. Applied Psycholinguistics, 7, 57-68.

Gierut, J. A., (1989). Maximal opposition approach to phonological treatment. Journal of Speech and Hearing Disorders, 54, 9-19.

Gierut, J. A., (1990). Differential reaming of phonological oppositions. Journal of Speech and Hearing Research, 33, 540-549.

Gierut, J. A., (1991a). Homonymy in phonological change. Clinical Linguistics and Phonetics, 5, 119-137.

Gierut, J. A., (1991b, May). Theoretical interpretation of behavior: Phonology. Paper presented at me First Biennial Symposium on Experimental Treatment Research, LAfayette, IN.

Gierut, J. A., & Dinnsen, D. A. (1988,November). Phonological restructuring and classificatory trees. Paper presented at the American Speech-Language-Hearing Association Convention, Boston.

Gierut, J. A., Elbert, M., & Dinnsen, D. A. (1987). A functional analysis of phonological knowledge and generalization reaming in misarticulating children. Journal of Speech and Hearing Research, 30, 462-479.

Gierut, J. A., & Neumann, H. J. (1992). Teaching and learning /Theta/: A nonconfound. Clinical Linguistics and Phonetics, 6, 191-200.

Goldman, R., & Fristoe, M. (1986). Goldman-Fristoe test of articulation. Circles Pines, MN: American Guidance Service.

Halle, M. (1988). The immanent form of phonemes. In W. Hirst (Ed.) Giving birth to cognitive science: A festschrift for George A. Miller. Cambridge: Cambridge University Press.

Hodson, B., & Paden, E. (1983). Targeting intelligible speech. Boston: College-Hill.

Hresko, W. P., Reid, D. K., Hammill, D. D. (1981). The test of early language development Austin, TX: Pro-ed.

Ingram, D. (1987). Categories of phonological disorder. In Proceedings of the first international symposium on specific speech and language disorders in children (pp.88-99). Surrey: Association for All Speech Impaired Children.

Ingram, D. (1989). Phonological disability in children (2nd ed.). San Diego: Singular.

Ingram, D. (1990a, November). The acquisition of the feature [voice] in normal and phonologically delayed English children. Paper presented at the American Speech-Language-Hearing Association Convention, Seattle.

Ingram, D. (1990b). Toward a theory of phonological acquisition. In J. Miller (Ed.), Research on child language disorders (pp. 55-72). Boston: College-Hill.

Jakobson, R. (1968). Child language, aphasia, and phonological universals (A. R. Keiler, Trans.). The Hague: Mouton. (Original work published 1941).

Janson, T. (1982). Sound change and perceptual compensation. In J. P. Maher, A. R. Bomhard, E. F. Konrad Koemer (Eds.), Papers from the third international conference on historical linguistics (pp. 119-127). Amsterdam: John Benjamins.

Janson, T. (1983). Sound change in perception and production. Language, 59, 18-34.

Johnston, J. R., & Smith, L. e. (1989). Dimensional thinking in language impaired children. Journal of Speech and Hearing Research, 32, 32-38.

Kazdin, A. E., & Hartmann, D. P. (1978). The simultaneous-treatment design. Behavior Therapy 9, 912-922.

Keil, F. C. (1988). Concepts, word meanings, and cognitive development Cambridge, MA: MIT.

Kenstowicz, M., & Kisseberth, C. (1979). Generative phonology: Description and theory. New York: Academic.

Kornfeld, J., & Goehl, H. (1974). A new twist to an old observation: Kids know more than they say. In A. Bruck, R. A. Fox, & M. W. LaGaly (Eds.), Papers from the parasession on natural phonology (pp. 210-219). Chicago: Chicago Linguistic Society.

Labov, W. (1981). Resolving the Neogrammarian controversy. Language, 57, 267-309.

Ladefoged, P., & Halle, M. (1988). Some major features of the international Phonetic Alphabet. Language, 64, 577-582.

Lauttamus, T. (1990). A note on the status of categorial features in the phonological feature hierarchy. Lingua, 81, 285-300.

Leonard, L. B., & Brown, B. L. (1984). Nature and boundaries of phonologic categories: A case study of an unusual phonological pattern in a language-impaired child. Journal of Speech and Hearing Disorders, 49 419-428.

Levine, M. N. (1986). Leiter international performance sea/e: A handbook. Chicago: Stoelting.

Locke, J. L. (1979). Homonymy and sound change in the child's acquisition of phonology. In N. Lass (Ed.), Speech and language: Advances in bask research and practice (Vol. 2, pp. 257-282). New York: Academic.

Markman, E. M. (1989). Categorization and naming in children: Problems of induction. Cambridge, MA: MIT.

McReynolds, L. V., & Kearns, K. P. (1983). Single-subject experimental designs in communicative disorders. Baltimore: University Park. bleater, R. A., & Ito, J. (1989). Feature predictability and underspecification: Palatal prosody in Japanese mimetics. Language, 65, 258-293.

Moskowitz, B. A. (1970). The two-year-old stage in the acquisition of English phonology. Language, 46, 426-441.

Ohala, J. J. (1980). The application of phonological universals in speech pathology. In N. Lass (Ed.), Speech and language: Advances in bask research and practice (Vol. 3, pp. 75-97). New York: Academic.

Phillips, B. S. (1983). Lexical diffusion and function words. Linguistics, 21, 487-499.

Powell, T. W., Elbert, M., & Dinnsen, D. A. (1991). Stimulability as a factor in the phonological generalization of misarticulating preschool children. Journal of Speech and Hearing Research, 34, 1318-1328.

Priestly, T. M. S. (1980). Homonymy in child phonology. Journal of Child Language, 7, 413-427.

Pulleyblank, D. (1986). Underspecification and low vowel harmony in Okpe. Studies in African Linguistics, 17, 119-153.

Pulleyblank, D. (1988a). Underspecification, the feature hierarchy, and Tiv vowels. Phonology, 5, 299-326.

Pulloyblank, D. (1988b). Vocalic underspecification in Yoruba. Linguistic Inquiry, 19, 233-270.

Robbins, J., & Klee, T. (1987). Clinical assessment of oropharyngeal motor development in young children. Journal of Speech and Hearing Disorders, 52, 271-277.

Sagey, E. C. (1986). The representation of features and relations in non-linear phonology. Unpublished doctoral dissertation, Massachusetts Institute of Technology, Boston.

Sommerstein, A. H. (1977). Modem phonology. London: Edward Arnold.

Stevens, K. N., & Keyser, S. J. (1989). Primary features and their enhancement in consonants. Language, 65, 81-106.

Thompson, C. K., & McReynolds, L. V. (1986). Wh interrogative production in agrammatic aphasia: An experimental analysis of auditory-visual stimulation and direct-production treatment. Journal of Speech and Hearing Research, 29, 193-206.

Trubetzkoy, N. S. (1969). Principles of phonology. (C. A. M. Baltaxe, Trans.) Berkeley and Los Angeles: University of California. (Original work published 1958).

Tversky, A. (1977). Features of similarity. Psychological Review, 84, 327-352.

Tyler, A. A., Edwards, M. L., & Saxman, J. H. (1987). Clinical application of two phonologically based treatment procedures. Journal of Speech and Hearing Disorders, 52, 393-409.

Tyler, A. A., Edwards, M. L., & Saxman, J. H. (1990). Acoustic validation of phonological knowledge and its relationship to treatment. Journal of Speech and Hearing Disorders, 55, 251-261.

Wang, W. S-Y. (1969). Competing changes as a cause of residue. Language, 45, 9-25.

Ward, J. M., & Bankson, N. W. (1989,November). Phonological genera/ization associated with the traditional and minimal contrast approaches. Paper presented at the American Speech-LanguageHearing Association Convention, St. Louis.

Welner, F. F. (1981). Treatment of phonological disability using the method of meaningful minimal contrast: Two case studies. Journal of Speech and Hearing Disorders, 46, 97-103.

Weiner, F. F., & Ostrowski, A. A. (1979). Effects of listener uncertainty on articulatory inconsistency. Journal of Speech and Hearing Disorders, 44, 487-493.

Winner, M., & Elbert, AL (1988). Evaluating the treatment effect of repeated probes. Journal of Speech and Hearing Disorders, 53, 211-218.

Received February 15, 1991

Accepted October 24, 1991

Contact author: Judith A. Gierut, Department of Speech and Hearing Sciences, Indiana University, Bloomington, IN 47405.

Appendix

Probe Items for Each Minimal Pair Treatment Format for Subject 19

As described in the body of the paper, each probe sampled
the treated
phonemes and their voicing counterparts (if in error) in
three prevocalic, one
intervocalic, and three postvocalic exemplars. Each
of the remaining phonemes excluded from the child's
inventory was sampled in
one prevocalic and one postvocalic exemplar.

Treatment format: Unknown : Unknown phonemes + Nonmajor
class + Maximal feature differences

/Theta/   /d (This character cannot be     /(This character
                                             cannot be
               converted in ASCII text)/     converted in
                                             ASCII text)

thumb          jump                           there
thirsty        jelly                          these
thunder        juice                          that
nothing        vegetable                      feather
bath           page                            -
mouth          orange                          -
teeth          cage                            -

/t (This character cannot be      /v/         /f/      /g/
 converted in ASCII text)/
cheese                          vegetable    fish      gum
chair                           glove        leaf      pig
chicken
picture
catch
watch
punch

/(This character cannot be   /r/       /l/      /s/    /z/
 converted in ASCII text)

shoe                         run      leaf      sock   zipper
fish                         deer     tail      juice  cheese

Treatment format: Unknown : Known phonemes + Nonmajor class
+ Maximal feature differences

/v/           /t/     /f/     /Theta/    /d/(This character
                                            cannot be
                                            converted in
                                            ASCII text)

van         toes     fish     thumb            juice
vegetable   tail     fat      mouth            page
vanilla     teeth    fire
seven       button   goofy
drive       eat      knife
stove       fat      leaf
glove       boot     roof

/ (This character cannot be    /t (This character cannot be
 converted in ASCII text)        converted in ASCII text)

there                                cheese
feather                              catch

/g/      /(This character cannot be    /r/    /l/    /s/  /z/
           converted in ASCII text)
gum        shoe                        run     leaf  sock  zipper
pig        fish                        deer    tail
                                                     juice cheese

~~~~~~~~

By Judith A. Gierut Indiana University Bloomington