DISCUSSION

The SLI children in this study were clearly below age level in standardized measures of expressive language. In their spontaneous speech, however, these SLI children did not show the same deficits in either morphology or in syntactic complexity. On most measures, their language when describing their classroom or narrating a cartoon did not differ either from their same-age peers or from younger children matched on standardized expressive language age. Whether the measure was a global measure of morphosyntax, such as MLU or syntactic complexity, or a more specific measure of a particular inflection, such as -ed, no significant differences were found. Only on the irregular past tense did SLI children perform differently from their CA peers, providing the correct form significantly less often when it was required. SLI children also regularized irregular verbs somewhat more than controls, and male SLI children did this more than females (15 different overregularization errors for SLI males vs. 5 for females). Thus, the male SLI children in this sample may have had difficulty remembering irregular forms. As Bishop (1994) points out, the irregular inflection could be correctly produced only if it had a lexical representation. Although she found overregularization errors in the speech samples of twelve 8- to 12-year-old children with SLI, they were rare. However, the finding that SLI children do have difficulty with irregular inflectional morphology means that it is not just a question of learning general rules of verb morphology (Bishop, 1994). In this study, the diversity of verbs that male SLI children used in the regular past tense was also lower than for the female SLI children in this sample, but the main effect of group on this measure was not significant. Our finding of differences in the use of irregular past tense forms by SLI children is congruent with those of Marchman et al. (1999) for older SLI children. Our failure to find differences in the regular past tense form is not in agreement with the results of Rice et al. (1998) for older SLI children.

Why did we not find more deficits in the morphosyntax of these SLI children when the bulk of the evidence cited in the introductory section found such differences? We were particularly surprised to find no differences in MLU (in morphemes) among our groups. We think that the probable reason is that the spontaneous production task was not demanding for the SLI children. They appeared quite relaxed on the videotapes while they described their classroom and recounted the film. They produced more utterances than their peers (an average of 88 vs. 80). In contrast, the standardized language measures may have been more challenging as tasks, while at the same time revealing more subtle deficits. For example, on the Grammatic Completion subtest of the TOLD Primary, the eight SLI children given this test received an average standard score of 6.0, whereas their CA controls received an average standard score of 10.1. For the two matched pairs who were given the TOLD Intermediate, average standard scores on Grammatic Comprehension involving judgments of morphological correctness were 14 for SLI and 22.5 for CA children. Thus, the aspects of these tests directed specifically at morphology show large differences between the SLI children and their peers. The Grammatic Completion subtest probes those morphemes targeted in the analysis of spontaneous speech samples, namely noun plurals, regular and irregular past tense, and pronoun case, as well as possessive “s” and possessive pronouns, third person singular inflection, progressive -ing, comparative -er and superlative -est, and the nominal compound (e.g., drummer). The Grammatic Comprehension test includes all of the morphemes targeted in this study as well as comparative and superlative forms, pronoun case, third person inflection, articles, negative, and adverb -ly endings. We did examine most of these additional forms in our speech samples, specifically possessive, third person singular inflection, progressive, adverb -ly, comparative and superlative, copula and auxiliary verbs. For these forms, there were either very few errors by SLI children or unequivalent obligatory contexts to analyze differences between the groups (for third person inflection and comparative/superlative).

We think that these differences between spontaneous and elicited production tasks make it unlikely, then, that a morphological measure based on spontaneous speech alone will be useful in diagnosing SLI (Bedore & Leonard, 1998). It seems that only elicited production will detect a morphological deficit, at least in elementary school children. Under stressed conditions (e.g., when given a standardized test), it is possible that these children revert to a more holistic phonological representation with insufficient detail for accurate performance on the task (Edwards & Lahey, 1998). This interpretation is in line with the “vulnerable markers hypothesis” according to which SLI children do understand the grammatical function of morphological markers but their accuracy in applying this knowledge will vary according to processing demands (Bishop, 1994).

The importance of task can be seen even when speech is essentially spontaneous. In this study, children in all groups exhibited more complex language when narrating a cartoon than when describing their classroom. Their utterances were longer in terms of morphemes and combined a greater number of syntactic categories. They enjoyed the narrative task more, and this affected the complexity of their descriptions.

Because there were no differences between SLI children and CA controls in the overall number of different verbs used in the regular or irregular past tense nor in the number of different nouns pluralized, it seems unlikely that SLI children are avoiding words in spontaneous speech whose past tense or plural endings they do not know. This study confirmed previous findings (Bishop, 1994; Oetting & Rice, 1993; Rice & Oetting, 1993; Rice et al., 1998) that SLI children do not have great difficulty with noun plurals.

One salient difference did appear that is not captured by any of the measures used in this study or in others but that is probably picked up by clinicians: these children sometimes use deviant syntax. Although such structures are rare (typically one per speech sample), they are very noticeable. They seem to reveal a basic underlying difficulty in generating a complex sentence with a dependent clause. Some of the errors in Table 5 illustrate the complex syntax error patterns described by Dykes and Schuele (2002) in their longitudinal study, specifically the omission of obligatory subject relative markers. Thus, in the elementary school years, when such clauses are typically effortless for ND children, SLI children sometimes express their meaning in quite tortured ways. For coders who were blind to the child's group, these odd sentences were a clear cue. Our measure of syntactic complexity did not capture the problem because it focuses on number of components rather than errors. It might be captured by an elicited production measure focused on sentence combining as in the TOLD-Intermediate or by the more sophisticated measures of sentence structures in spontaneous speech found in Scarborough's (1990) Index of Productive Syntax.

It seems likely to us that the SLI children's difficulty in generating complex sentences in spontaneous speech and their poor performance on standardized language tasks are due in large part to their limited memory capacity. In spontaneous speech, the task demands were apparently within their capacity, except for quite complex sentences. Leonard (1998, p. 266) similarly points out that “when time pressures are reduced, as when spontaneous utterances are the focus of analysis…and the child controls the complexity of the sentence as a whole,” problems with grammatical inflections are reduced. In auditory memory span tasks, we found that these same SLI children significantly underperformed relative to their CA controls (Blake, 2000). The deficit in auditory memory found in our earlier study of preschool 3- to 5-year-old SLI children relative to their peers (Kushnir & Blake, 1996) thus persists in older SLI children. This deficit may underlie their failure to supply as many irregular past tense forms as their peers because they may have difficulty remembering them. Unlike Leonard (1998), who appears to support general processing speed as an important cause of SLI, we would then rather specifically invoke auditory memory. However, it is difficult to devise a pure measure of memory that does not include a processing component (see Blake, 2000), and it is also difficult to distinguish between problems in the formation of phonological representations and their storage (Edwards & Lahey, 1998). Leonard (1998) makes an interesting argument that slow processing will result in incomplete retrieval of an item in storage so that the child can proceed with the rest of the sentence. This argument also applies to phonological encoding. If words are not completely encoded or retrieved, then language reception and production will be deficient. Of course, there are also other sources of SLI children's difficulty with language; and these no doubt vary with the characteristics of each sample.

In conclusion, these results replicated other findings showing difficulties for SLI children in the past tense form of irregular verbs and lack of difficulty in noun plurals. Differences in overall measures of morphosyntax were not significant in this study, however. It may be that the sample sizes were not large enough to detect differences. It seems more likely, however, that older SLI children no longer have the same difficulties with morphology as the younger children studied in most of the previous research reviewed. In addition, some researchers have found that SLI children did not have problems with certain morphemes in spontaneous speech (Leonard et al., 1997; Oetting & Horohov, 1997). Thus, elicited production tasks pose special difficulties for these children but they may also be better at detecting subtle deficits in older SLI children.