| |
Children who have speech sound errors that have persisted past the preschool years are considered to have residual errors (Shriberg, 1994). Typically, these school-age children have substitution and distortion errors rather than deletions, and intelligibility is not usually a primary issue. Children with residual errors generally have acquired the sound system of their language, but they have errors that draw attention to the speaking pattern. The assumption is usually made that they are having difficulty with the articulatory movements needed to produce the acceptable sound and with embedding that sound into the stream of speech. It should be noted, however, that in the early days of studying communication disorders, authorities such as Van Riper (1978) assumed that the child's difficulty was first of all perceptual.
Until recently, there has been little research about how residual speech sound errors develop, even though some of the earliest research and intervention in communication disorders focused on children with residual errors. The profession has uncovered bits and pieces of information about development after the preschool period, but no coherent picture has emerged to assist in predicting which children will actually make needed changes without intervention. The question is an important one, because if we fail to treat a child at age 6 who is not going to change spontaneously, and instead we wait until age 9, the child has 3 additional years of practice on an error phoneme, and remediation will likely be more difficult. Certain information that can be obtained from speech sampling may provide insight into the prediction question:
• Most residual errors affect a subset of the phonemes of English (“the big 10”): // (Winitz, 1969). These are typically late-acquired phonemes, but most of them are used correctly by 90% of children by age 8 (Smit et al., 1990).
• The phoneme in error may make a difference. Reanalysis of the Smit et al. (1990) cross-sectional data suggests that children may be less likely to self-correct the alveolar and palatal fricatives and affricates than they are the /r/ and the /θ ð/ (Smit, unpublished).
• The nature and allophonic distribution of the error may make a difference. Stephens, Hoffman, and Daniloff (1986) showed that children with lateral productions of alveolopalatal fricatives and children who substituted back sounds for these fricatives generally did not improve spontaneously, whereas about half of the children with dental errors corrected them. Hoffman, Schuckers, and Daniloff (1980) showed that children who produced the consonantal /r/ allophone correctly some of the time were likely to achieve the other /r/-allophones spontaneously.
• The length of time that the child has made the error may make a difference. It is reasonable to assume that if a child's production of a phoneme has not changed at all in several years, then spontaneous change is unlikely.
• The child's developmental history may make a difference. Shriberg (1994) has pointed out that some children had phonological errors as preschoolers, while others did not. On logical grounds, children who had phonological problems earlier are less likely to change without intervention because they have already demonstrated difficulties in learning the sound system of their native language.
• The pattern of change in the child's errors may be important. Recent research by Gruber (1999) into the time taken for children who are receiving intervention to normalize (achieve age-appropriate phonology) has provided some clues about prognosis. For example, it appears that if the child reduces substitutions and omissions while increasing distortions, that child will take longer to normalize than the child who decreases all types of errors.
Speech Sampling
Speech-language pathologists typically elicit a speech sample using a published test of articulation, supplemented with a conversational speech sample. For a school-age child, the conversational sample should be audio-or video-recorded, with careful attention to the quality of the recording. This sample should include at least 100 different words and 3 minutes of child talking time. If the child has a relatively large number of errors, this speech sample can be transcribed phonetically for further analysis. If the child has just a few sounds in error, then the clinician may decide not to transcribe the entire speech sample. Instead, he tallies all instances of a target phoneme in the sample, determines how many were acceptably produced, and derives a percentage of correctly produced sounds. When these counts are based on a 3-minute sample, this procedure results in a TALK sample (Diedrich, 1971), which is a probe of conversational speech.
Articulation Tests
Some of the first tests that were commercially available in communication disorders were tests of articulation and were designed to assess the development of speech sounds. Typically, tests of this type are based on the single-word naming of pictures without a model from the examiner. Most tests of articulation assess production of all English consonant phonemes in word-initial and word-final position, and possibly English consonant clusters as well. Scoring sheets usually allow explicit comparisons between adult targets and the children's productions. Most articulation tests result in a summary numerical score that can be compared to normative data. Some currently used tests of articulation include the Templin-Darley Tests of Articulation (Templin and Darley, 1969), the Goldman-Fristoe Test of Articulation—2 (Goldman and Fristoe, 2000), the Smit-Hand Articulation and Phonology Evaluation (SHAPE; Smit and Hand, 1997), and the Photo Articulation Test—Third Edition (Lippke et al., 1987).
Most of the inventory tests do not require that the clinician use narrow transcription, the exception being SHAPE. Rather, broad transcription is generally used, even when the test requires only a notation of “correct,” “substitution,” “omission,” or “distortion.”
The error sounds identified on an inventory test are often examined in light of “ages of acquisition” for those sounds. The age of acquisition is the age at which 75% or 90% of children typically say the sound correctly (Templin, 1957; Smit et al., 1990). The guidelines used by many school districts to determine caseload make reference to these ages of acquisition.
There are other types of tests of articulation besides inventory tests. In particular, there are several tests of contextual variation, among them the McDonald Deep Test of Articulation (McDonald, 1964), the Contextual Test of Articulation (Aase et al., 2000), and the Secord Contextual Articulation Tests (S-CAT; Secord and Shine, 1997). Contextual variation is a way of manipulating the phonetic environment of a target sound in order to see if the client can produce the target sound acceptably in one or more of these novel phonetic environments. If the child is able to do so, then the clinician can use these facilitating contexts in the first few treatment sessions.
Still another kind of assessment involves determining stimulability. Stimulability refers to the ability to elicit an acceptable production of a speech sound or structure, such as a consonant cluster, from the child by presenting instructions, cues, and models. A systematic way to assess stimulability has been proposed by Perrine, Bain, and Weston (2000). The implications of stimulability have been addressed by numerous researchers, but that research can be summed up in a few statements:
1. The child who is not stimulable for a specific phoneme target is the child who should have the highest priority for intervention.
2. If a child is stimulable for a target phoneme, the child may or may not improve without intervention.
3. Stimulable phonemes are likely to bring about quick success in intervention.
Finally, the clinician may assess inconsistency. Inconsistency refers to variations in the child's productions of a given phoneme. If the child's production is characterized by “inconsistency with hits” (correct productions), then the context of the hits can be determined. Just as in the case for contextual variation, a hit can serve as an entrée into intervention. To look for inconsistencies, the clinician may catalogue the productions of the target that are heard in the conversational speech sample. Alternatively, she can administer the Story-Telling Probes of Articulation Competence from the S-CAT (Secord and Shine, 1997).
Intelligibility
Intelligibility is defined as a listener's ability to understand a speaker's words. Although intelligibility of speech can be reduced in children who have residual errors, the reduction often is not substantial. One exception is the child who may have a few distortions of phonemes but who has particular difficulty in stringing sounds together in multisyllabic utterances. This difficulty may manifest itself as weak or imprecise articulations of sounds, along with deletions of some consonants. In such cases, the examining clinician may want to repeat what she understood the child to say immediately afterward, so that her gloss is also recorded on the tape when the conversational speech sample is recorded.
The standard way to assess intelligibility in a numerical way is to have a person who is not familiar with the child listen to the audio recording of the conversational sample, but without hearing the examiner's speech. This person writes down the child's words. This gloss is compared to the one generated by the examining clinician, and a Percent of Intelligible Words is calculated (Shriberg and Kwiatkowski, 1982).
For children with residual speech sound errors, a more salient issue than intelligibility may be that their errors call attention to their speech, that is, to the medium rather than the message. Listeners may consider their speech to be babyish, bizarre, or odd. The clinician can develop questionnaires and rating scales and can ask persons familiar with the child to fill them out in order to document this perception, in addition to asking the child about the content of any teasing that may occur.
Interpreting the Data
Decisions about whether to provide intervention are often based on multiple factors. These include the child's age relative to the age of acquisition for the child's error phonemes, whether or not the child is stimulable for correct production, intelligibility, and the degree to which the child and significant others consider the speech to be a problem. There is little research other than that of Gruber (1999) to go by in establishing prognosis. However, a reasonable assumption is that the older the child who has residual errors, the longer it will take to achieve normalization in a treatment program.
| |
