1. Introduction
2.4 Generalisations
2.4.2 Generalisation 2
It is a widely reported finding from studies of SLI that the children’s errors are primarily omissions of obligatory forms, while commissions are rare (cf. Leonard et al., 1992; Bishop, 1994; Rice et al., 1995). This, if true, has significant consequences for the underlying nature of SLI because it suggests that the elements or processes which are subject to the deficit are still constrained to some degree as they are in ND. As such, their
failure to appear in obligatory contexts cannot be a reflection of a total freedom in their application, with the implication for an explanatory account of the disorder being that the deficit should not relate to the principles which constrain the application of these processes or elements.
The suggestion that SLI children’s errors are uni directional is not uncontroversial, however. A number of studies cite data which appears to be at odds with the predictions of this generalisation. In particular, Myma Gopnik and her colleagues have argued against this view (cf. Gopnik 1990a, 1990b; Gopnik & Crago, 1991). This stance was based upon the observation that their SLI children sometimes produced utterances of the form exemplified in 8;
8 You make one points (Gopnik, 1990b)
In 8 the child has clearly produced a commission error, inserting the plural -s morpheme inappropriately. In addition, Gopnik reports that this form of sentence was produced correctly in almost identical contexts at other times.
According to Gopnik, the SLI child does not grammatically distinguish the bare form (e.g. jump) from either the inflected past tense or plural form (e.g. jumped and jumps). As such, they are free to insert them in any context, irrespective of their perceived functional requirement. If production of the form exemplified in 8 is truly characteristic of SLI children’s language then it would be impossible to maintain Generalisation 2. However, I argue that commission errors of this form are not characteristic of SLI. While Gopnik did not statistically analyse the productivity of the commissions, subsequent studies, while reporting commission errors, suggest that these were no more frequent in the speech of SLI children than they were for the MLU controls (cf. Leonard et al., 1992; Bishop, 1994; Rice et al., 1995). This finding is crucial because it suggests that these errors are not specific to the SLI child’s grammar, but rather a more general developmental characteristic. Table 3 reports the findings from two studies which consider this factor:
Leonard et al. (1992) Rice et al, (1995)
SLI MLU SLI MLU
Articles <1 1 Plural 2 2 Third singular 8 8 0 0 Regular past 0 0 0 0 Copula be 4 5 7 9 Auxiliary be 4 5
Table 3. Percentages of inappropriate use of grammatical morphemes (adapted from Leonard, 1998)
These findings clearly suggest that morphological commission errors are not a characteristic of the SLI child's grammar. This is with the qualification that, although these errors do occur, it is to an extent that is comparable to the children’s language matched controls.
Given these findings, it seems that Generalisation 2 holds at least for the domain of (English) inflectional morphology. However, SLI children’s difficulties are not restricted to this domain, and for the generalisation to be meaningful it must be demonstrated that it holds for all the linguistic domains which are subject to the children’s deficit The generalisation requires further qualification in two other domains before it can be adopted.
The first of these relates to the domain of structural Case. In section 2.4 we discussed findings from this domain, reporting that significant difficulties were observed. SLI children appear to have particular difficulty assigning nominative Case to pronominals, with the accusative form often surfacing in the subject position. These findings are potentially problematic for Generalisation 2 because they clearly represent a commission error, an appropriate form has been substituted by an inappropriate form. Generalisation 2 would predict that where the SLI child’s grammar is unable to access the correct form, the element would be omitted, rather than subject to substitution.
However, a possible explanation for this violation is that the SLI child is simply unable to omit the Case feature. Case, unlike verbal morphology, is intrinsic to its nominal categoiy. As such, there would be no Caseless lexical entry available to the child’s grammar. Omission of the nominal in its entirety would not presumably be an option, and therefore the child has to insert an inappropriate form. This seems a reasonable solution.
In addition, the distribution o f the Case errors also appears to support this proposal. As was noted in section 2.2.4, the Case errors are restricted to subject nominals. Schutze & Wexler do not report problems with accusative Case assignment. Also, they observed that the substitute Case is always accusative, never genitive. These findings would be unexpected if we were to propose that the SLI children could fireely overgenerate Case forms. As discussed, Schutze & Wexler take the distribution o f these errors to indicate that there is a default Case strategy in action. This would also be unusual if the child could fi*eely substitute Case; a default would be pointless. As such, it appears that, while Case production is an apparent exception to Generalisation 2, this exception is principled and predictable.
The second domain in which findings seem to run contrary to Generalisation 2 is wh-question formation. As was observed in section 2.2.3, SLI children regularly produce wh-questions containing an additional wh-element. Consider examples b and c in 9. 9 a. What did John hit?
b. *What did John hit something? c. * What did John hit what?
This form of production clearly represents a commission error. Note, however, that it is not of the same form observed for Case assignment. This type of error is not a substitution error, the inappropriate item is additional to the normally subcategorised arguments of the verb.
As with the Case errors, I feel that this form of production is not necessarily problematic for Generalisation 2. Again, there are characteristics of the distribution of these errors which suggest that they are not simply the reflection of blind over- generation. For example, studies rarely report wh-questions without a sentence initial wh- element, even when this is grammatical. In addition, there is little evidence from other domains that the theta criterion is ever violated by the SLI child, although it is here.
A number of researchers have proposed that these apparent commissions are the result of an explicitly learned strategy, which specifies that a wh-question must have a sentence initial wh-element (cf. van der Lely, 1998). These findings support this view. The strict pattern observed in the distribution of the errors is consistent with a rote
learning strategy. In addition, this explains the apparent violation of the theta criterion. If the child is strategically inserting the wh-element in the sentence initial position, then this leaves the VP internal argument position vacant. As such, the SLI child is free to fill this position with another argument, producing the apparent commission errors observed.
Again, considering the restrictions which appear to hold in other domains, this proposal seems plausible. Standard therapy would promote this type of strategy and the saliency of the objects involved would make them easily formable. I therefore conclude that these commission errors are not characteristic of the SLI child’s granunar.