| |
Children who have speech sound disorders can reasonably be separated into two distinct groups. One group comprises children for whom intelligibility is a primary issue and who tend to use many phonological processes, especially deletion processes. These children are generally in the preschool age range. The second group includes children who have residual errors, that is, substitution and distortion errors that are relatively few in number. These children are typically of school age, and intelligibility is better than in the first group. The first group, namely children with phonological difficulties, is the focus of this entry.
In most cases, when we assess speech production in a preschool child, the purposes of the assessment are to describe the child's phonological system and make decisions about management, if needed. A good audio-or video-recorded speech sample from play interactions with parents or the clinician can capture all of the primary data needed to describe the system, or it can be supplemented with a single-word test instrument. Typically, we define an adequate conversational speech sample as one that includes at least 100 different words (Crystal, 1982). Additionally, these words should not be direct or immediate repetitions of an adult model. Finally, for children with poor intelligibility, it is helpful for the examiner to repeat what he believes the child said after each utterance so that this spoken gloss is also recorded on the tape.
Once the recording has been made, the next task is for the examiner to gloss and transcribe the speech sample, ideally using narrow transcription (Edwards, 1986; Shriberg and Kent, 2003). Professionals who frequently do extensive transcription may wish to use a computer-based system. (See Masterson, Long, and Buder, 1998, for an excellent review of such software.) However, such systems are only as good as the clinician's memory for the symbols and diacritics and her memory of their location on the keyboard; consequently, doing transcription by hand may be the more reliable way to go about this task if one does not do it frequently.
With the transcript in hand, the clinician now has a choice of types of analyses. First of all, one can undertake both independent and relational analyses (Stoel-Gammon and Dunn, 1985). Independent analyses treat the child's system as self-contained, that is, with no reference to the adult system. They include a phonetic inventory for consonants and perhaps for vowels, as well as tallies of syllable or word shapes. Relational analyses, on the other hand, explicitly compare the child's production to that of the adult, including a segmental (phonemic) inventory for consonants and perhaps for vowels and a list of the phonological processes that the child uses.
Independent analyses are appropriate for children who are very young, or who have very poor intelligibility, or who appear to use few differentiated speech sounds. Clinicians typically devise their own forms for these analyses, although some of the software mentioned above permits certain of these independent analyses to be done automatically. Frequency of use is an issue in independent analyses, so the various phones and syllable structures that appear in the transcript should be tallied on the inventory form. It is helpful to structure these forms into major consonant classes and major vowel classes, as well as syllable position—syllable-initial, syllable-final, and intervocalic. In addition, for some types of analysis, separate inventories should be done for one-, two-, and three-syllable utterances or words.
One kind of relational analysis, the segmental or phonemic inventory, which compares the child's production to the adult target, is more familiar to clinicians because it resembles typical published tests of articulation and phonology. Phonological process analysis is also considered to be relational in nature. If clinicians are working from a transcript of conversational speech, all of the software mentioned earlier can provide at least a list of phonological processes. Alternatively, clinicians may again devise their own forms for the segmental inventories and the list of phonological processes. Typically, the list of phonological processes will include the 8–10 processes commonly listed in texts and tests of phonology, as well as any unique or idiosyncratic processes that the child uses. The examiner then goes through the transcript noting what the child produces for each adult form. These productions are also tallied.
One other important measure that is relational in nature is the Percentage of Consonants Correct (PCC; Shriberg and Kwiatkowski, 1982). The number of adult targets that the child attempts is tallied, using the standards of colloquial speech, and the number of targets that the child produces acceptably is also tallied. Simple division of these tallies results in the PCC. The PCC is considered to be a measure of severity. It is a useful measure for assessing change over long periods of treatment, such as 6 months.
Tests of Phonology
Several tests published commercially permit analysis of children's use of phonological processes on the basis of single-word naming of objects or pictures. They include the Bernthal-Bankson Test of Phonology (BBTOP; Bernthal and Bankson, 1990), the Khan-Lewis Phonological Analysis (KLPA; Khan and Lewis, 1986), and the Smit-Hand Articulation and Phonology Evaluation (SHAPE; Smit and Hand, 1997), all of which are based on pictures or photos, and the Assessment of Phonological Processes–Revised (APP-R; Hodson, 1986), which is based on object naming.
Tests of phonology can complement analyses of the conversational sample. One of their virtues is that for the phonological processes that are assessed, they incorporate multiple exemplars of each process, so that there is some assurance that the child's use of the process is not happenstance. For some extremely unintelligible children, tests of phonology may be the only way to figure out the child's patterns because the clinician at least knows what the intended word should be. Some of these tests (BBTOP, KLPA, and SHAPE) also permit comparisons to be made between the child's performance and normative data.
On the other hand, if a test of phonology becomes the primary assessment tool, the clinician needs to be aware that measures derived from single-word naming may differ from measures derived from conversation. The conversational speech of disordered children generally includes more nonadult productions than does single-word naming. In addition, tests of phonology tend to deal with a very circumscribed set of phonological processes, and they do not deal at all with vowel productions. Consequently, if the child uses important but idiosyncratic processes, such as glottal replacement, or has systematic vowel errors, they may not be picked up. (However, SHAPE has an extensive list of potential idiosyncratic processes in an appendix, along with instructions for determining the frequency of use of these processes.)
Specialized testing using single-word stimuli may be required for specific treatment orientations. For example, in order to carry out a generative analysis, Elbert and Gierut (1986) have developed a test with more than 300 items in which bound morphemes, such as re- (meaning again) or the diminutive -y or -ie, are added to the word at either end. The purpose is to determine whether the child changes an error production of the first consonant or the last consonant in the presence of the morphological addition and changes it in a way that clarifies the child's underlying representation.
Another example of specialized testing using either single words or connected speech is the elicitation of data needed for nonlinear analysis (Bernhardt and Stoel-Gammon, 1994). Nonlinear approaches deal with the hierarchies of representation of words, for example, at the segmental level, the syllable level, the foot level, and so on. Productions of multisyllabic words with varying stress patterns are needed to complete these analyses. In most cases these multisyllabic words must be elicited by imitation or picture naming.
Finally, an altogether different type of specialized testing is the determination of stimulability. A child is said to be stimulable for an error sound if the clinician can elicit an acceptable production using models, cues, or phonetic placement instructions. This part of the assessment is often performed informally and usually for just a few error sounds. However, Perrine, Bain, and Weston (2000) have devised a systematic way to assess stimulability based on a hierarchy of cues and models that is helpful in planning intervention.
Intelligibility
Intelligibility refers to how well the child's words can be understood by others. There are at least two measures of intelligibility that are based on counts of intelligible words, as well as many scales for making perceptual judgments of intelligibility. The most straightforward numerical measure, Percent of Intelligible Words (PIW), has been described by Shriberg and Kwiatkowski (1982). To determine this measure, a person who does not know the child listens to the conversational speech sample but does not hear the clinician's comments. This person attempts to gloss the sample. The number of words correctly glossed by the listener is divided by the total number of words to obtain the PIW.
A second measure based on counts, the Preschool Intelligibility Measure, has been developed by Morris, Wilcox, and Schooling (1995). The child is asked to imitate a series of one-or two-syllable words that are selected randomly from a large database of words, and her productions are recorded. Then the audiotape is played for listeners, who see 12 foils for each word and circle the one they think the child said. This measure is better suited for documenting changes in intelligibility over time than it is for initial evaluation.
Other scales of intelligibility involve judgments on the part of the clinician or significant others in the child's environment about how well the child communicates. For example, teachers might be asked to rate how well they understand the child on a 6-point scale, with 1 representing “all the time” and 6 representing “never.” Or a parent might be asked to rate the difficulty that family members have in understanding the child.
Treatment Decisions and Prognosis
Decisions about whether to treat and how often the child should be seen are made primarily on the basis of severity and secondarily on the basis of stimulability. Although there is little research on the topic of appropriate treatment decisions, severity appears to have universal acceptance among speech-language pathologists as the most important variable in deciding for or against treatment.
With respect to prognosis, until recently there has been little research about how children normalize (achieve age-appropriate phonology) and how long it takes. However, work by Shriberg, Gruber, and Kwiatkowski (1994) and by Gruber (1999) suggests that some children who receive intervention normalize by about 6 years of age, and that the outer limit for normalization is about 8.5 years. However, the predictors of normalization are not yet known.
| |
