Word pronunciation is an overt speech characteristic that readily identifies dialect differences among normal speakers even when other aspects of their spoken language do not. Although regional pronunciation differences in the United States were recognized historically, social dialects were not. Nonprestige social dialects in particular were viewed simply as disordered speech. A case in point is the native English dialect spoken by many African Americans, a populous ethnic minority group. This dialect is labeled in various ways but is referred to here as African American Vernacular English (AAVE). As a result of litigation, legislation, and social changes beginning in the 1960s, best clinical practice now requires speech clinicians to regard social dialect differences in defining speech norms for clinical service delivery. This mandate has created challenges for clinical practices.

One clinical issue is how to identify AAVE speakers. African Americans are racially, ethnically, and linguistically diverse. Not all learn AAVE, and among those who do, the density of use varies. This discussion considers only those African Americans with an indigenous slave history in the United States and ancestral ties to Subsaharan Africa. The native English spoken today is rooted partly in a pidgin-creole origin. Since slavery was abolished, the continuing physical and social segregation of African Americans has sustained large AAVE communities, particularly in southern states.

Contemporary AAVE pronunciation is both like and unlike Standard English (SE). In both dialects, the vowel and consonant sounds are the same (with a few exceptions), but their use in words differs (Wolfram, 1994; Stockman, 1996b). Word-initial single and clustered consonants in AAVE typically match those in SE except for interdental fricatives (e.g., this > /dıs/). The dialects differ in their distributions of word-final consonants. Some final consonants in AAVE are replaced (cf. bath and bathe > /f/ and /v/, respectively). Others are absent as single sounds (e.g., man) or in consonant clusters (test > /tεs/). Yet AAVE is not an open-syllable dialect. Final consonants are variably absent in predictable or rule-governed ways. They are more likely to be absent or reduced in clusters when the following word or syllable begins with another consonant rather than a vowel (Wolfram, 1994) or when a consonant is an alveolar as opposed to a labial or velar stop (Stockman, 1991). In multisyllabic words, unstressed syllables (e.g., away > /-weı/) in any position may be absent, depending on grammatical and semantic factors (Vaughn-Cooke, 1986). Consonants may also be reordered in some words (e.g., ask > /s/), and multiple words may be merged phonetically (e.g., fixing to > finna; suppose to > sposta) to function as separate words. These broadly predictable AAVE pronunciation patterns differ enough from SE to compromise its intelligibility for unfamiliar listeners. Intelligibility can be decreased further by co-occurring dialect differences in prosodic or nonsegmental (rhythmic and vocal pitch) features (Tarone, 1975; Dejarnette and Holland, 1993), coupled with known grammatical, semantic, and pragmatic ones. Consider just the number of grammatical and phonological differences between SE and AAVE in the following example:

Enough is known about the complex perceptual judgments of speech intelligibility to predict that the more work listeners have to do to figure out what is being said, the more likely is speech to be judged as unclear.

Identifying atypical AAVE speakers can be difficult, especially if known causes of disordered speech—hearing loss, brain damage, and so on—are absent, as is often the case. Clinicians must know a lot about the dialect to defend a diagnosis. But most clinicians (95%) are not African American and have little exposure to AAVE (Campbell and Taylor, 1992). Misdiagnosing normal AAVE speakers as abnormal is encouraged further by the similarity of their typical pronunciation patterns (e.g., final consonant deletion, cluster reduction, and interdental fricative substitutions) to those commonly observed among immature or disordered SE speakers. However, typically developing African-American speakers make fewer errors on standardized articulation tests as they get older (Ratusnik and Koenigsknecht, 1976; Simmons, 1988; Haynes and Moran, 1989). Still, they make more errors than their predominantly white, age-matched peers (Ratusnik and Koenigsknecht, 1976; Seymour and Seymour, 1981; Simmons, 1988; Cole and Taylor, 1990), and they do so beyond the age expected for developmental errors (Haynes and Moran, 1989). Therefore it is unknown whether the overrepresentation of African Americans in clinical caseloads is due to practitioner ignorance, test bias, or an actual higher prevalence of speech disorders as a result of economic poverty and its associated risks for development in all areas.

The accuracy in identifying articulation/phonological disorders improves when test scores are adjusted for dialect differences (Cole and Taylor, 1990), or when the pronunciation patterns for a child and caregiver are compared on the same test words (Terrell, Arensberg, and Rosa, 1992). However, tests of isolated word pronunciation are not entirely useful, even when nonstandard dialect use is not penalized. They typically provide no contexts for sampling AAVE's variable pronunciation rules, which can cross word boundaries, as in the case of final consonant absence. Although standardized deep tests of articulation (McDonald, 1968) do elicit paired word combinations, they favor the sampling of abutting consonant sequences (e.g., bus fish), which penalize AAVE speakers even more, given their tendency to delete final consonants that precede other consonants as opposed to vowels (Stockman, 1993, 1996b). These issues have encouraged the use of criterion-referenced evaluations of spontaneous speech samples for assessment (see Stockman, 1996a, and Schraeder et al., 1999).

Despite the assessment challenges, it is readily agreed that some AAVE speakers do have genuine phonological/articulatory disorders (Seymour and Seymour, 1981; Taylor and Peters, 1986). They differ from typically developing community peers in both the frequency and patterning of speech sound error. This is true whether the clinical and nonclinical groups are distinguished by the judgments of community informants, such as Head Start teachers (Bleile and Wallach, 1992), other classroom teachers (Washington and Craig, 1992), or speech-language clinicians (Stockman and Settle, 1991; Wilcox, 1996).

AAVE speakers with disorders can differ from their nondisordered peers on speech sounds that are like SE (Type I error, e.g., word-initial single and clustered consonants). They can also differ on sounds that are not like SE either qualitatively (Type II error, e.g., interdental fricative substitutions) or quantitatively (Type III error, e.g., more frequent final consonant absence in abutting consonant sequences). Wolfram (1994) suggested that these three error categories provide a heuristic for scaling the severity of the pronunciation difficulty and selecting targets for treatment.

Two service delivery tracks are within the scope of practice for speech clinicians. One remediates atypical speech relative to a client's native dialect. The other one expands the pronunciation patterns of normal speakers who want to speak SE when AAVE is judged to be socially or professionally handicapping (Terrell and Terrell, 1983). For both client populations, effective service delivery requires clinician sensitivity to cultural factors that impact (1) verbal and nonverbal interactions with clients, (2) selection of stimuli (e.g., games and objects) for therapy activities, and (3) scheduling of sessions (Seymour, 1986; Proctor, 1994). However, the service delivery goals do differ for these two populations. For abnormal speakers, the goal is to eradicate and replace existing patterns that decrease intelligible speech in the native dialect. This means that the pronunciation of bath/bæθ/ as /baf/ should not be targeted for change, if it conforms to the client's target dialect. But a deviation from this expected pronunciation, such as bath/bæθ/ > /bæt/ or /bæs/, is targeted, if observed at an age when developmental errors are not expected. In contrast, the service delivery goal for normal AAVE speakers is to expand rather than eradicate the existing linguistic repertoire (Taylor, 1986). An additive approach assumes that speakers can learn to switch SE and AAVE codes as the communicative situation demands, just as bilingual speakers switch languages. This means that a speaker's bidialectal repertoire includes both the SE and AAVE pronunciation of “bath” (cf. bath > /baθ/ and /bæf/).

Meeting these two different service goals requires attention to some issues that are not the same. They affect which patterns are targeted and how change is facilitated. For typical AAVE speakers learning SE, second language acquisition principles are relevant. Besides the production practice, service delivery requires contrastive analysis of the two dialects and attention to sociocultural issues that affect code switching. Correct or target productions are judged relative to SE.

In contrast, for speakers with abnormal pronunciation, AAVE should be targeted. Which features to target in therapy and how to model the input become issues, because most clinicians do not speak AAVE. They also may resist modeling a low social prestige dialect because of negative social attitudes towards it. Wolfram (1994) reminded us that AAVE and SE share many of the same target features (e.g., most word-initial consonants). Errors on shared features (Type I) should be targeted first in treatment. They are likely to impair intelligibility even more than the smaller sets of qualitative (Type II) errors, such as stop replacement of interdental fricatives (cf. this /dıs/ > /bıs/), or quantitative (Type III) errors, such as final consonant deletion in more than the allowable context number and types. AAVE features should be targeted for treatment only when pronunciation patterns differ from AAVE norms. Articulatory patterns would not be modified if they differed from the clinician's SE-modeled pattern but matched expected AAVE patterns.

Legitimizing social dialects like AAVE in the United States has required researchers and clinicians to (1) broaden the reference point for normalcy and (2) explore alternative strategies for identifying service needs and modifying word pronunciation. The issues singled out in this entry are not unique to phonological/articulatory problems. However, given their typically higher frequency of occurrence relative to other domains of spoken language in all groups, they may turn up more often in clinical work.

See also dialect speakers; dialect versus disorder; language disorders in african-american children.