Phonological disorders in the light of constraints

The appearance of Optimality Theory (OT) in 1993 has not yet been able to change the focus of research on phonological disorders since they basically continue to use the model based on rules of autosegmental phonology. Some works, however, have effectively analyzed data of phonologically disordered children by employing the constraint-based model proposed by OT. Among them are Bernhardt & Stemberger (1998), Dinnsen & O ́Connor (2001), Barlow (2001) and Lamprecht (1999) for Brazilian Portuguese data. Although some papers are already employing OT in the description and analysis of data with phonological impairment, it remains necessary to take up some aspects of disordered speech and attempt to clarify them through the theory. The advantages attributed to OT do not appear to lie merely in the rereading fostered by the tableaux which show interactions not yet evidenced by the use of derivational models but they seem especially to reside in its capacity to explain specific aspects of disordered speech from within its own architecture. From the perspective of OT, the characteristics attributed to disordered speech can be viewed through the workings of the grammatical system itself, expressed by means of functions such as Gen and Eval, of learning algorithms and, basically, of the constraint hierarchies which constitute grammar. The present paper seeks to ponder on how OT understands speech disorders and on the contributions that this constraint-based model can make to the description, analysis and treatment of them. It also seeks to comment on the efficacy of OT in making explicit the implicational relations proposed by Implicational Model od Features Complexity (Mota, 1996). Through the analysis of the data of 25 Brazilian phonologically disordered children, Mota (1996) found implicational relations among the distinctive features which lead to different possibilities in the development of consonantal segments during the acquisition process. Considering the subjects' performance in what concerns the distinctive features involved in each of the consonantal segments observed, the author proposed a table of implicational relationships among features. By means of the application of OT to some aspects of Mota's work, we have tried to demonstrate that the theory is able to effect a rereading of the analyzed data in a more satisfactorily way, allowing for the "visualization" of phonological disorders from "within" the linguistic system. According to Optimality Theory, the hierarchization of features does not occur in a treelike structure, but through the ranking of constraints. From this perspective, it is also necessary to consider other features that militate in the early stages of acquisition, such as [+approximant], [+vocoid] and [dorsal], which would account for the early emergence of the vocalic segments /a/, /i/ and /u/. Therefore, the 0 level proposed by Mota (1996) would be formed by the features [+/-voice], [+/-vocoid], [+/-approximant], [+/-continuant], [coronal], [labial], [dorsal], [+/-anterior] and [+/-sonorant]. This alteration is justified because the demotions of feature constraints such as *[+approximant] and *[dorsal], below the faithfulness constraints, allowing for the acquisition os /a/, for example, are not related to the vocalic segments only. At first, it cannot be conceived that the markedness constraints be related to one only kind of segment; actually, it is the totality of such constraints that forms the segments in OT. Differently from Feature Geometry (Clements & Hume, 1995), Optimality Theory does not need different representations for vocalic and consonantal segments; it can also express, within its own architecture, the complexity of segments as it assumes the existence of conjoined constraints. The difference between normal and disordered acquisition resides, basically, in the late demotion of conjoined constraints acting in the domain of the segment and established during the acquisition process. The data seem to indicate that phonologically disordered children do not present difficulty in the demotion of individual constraints which constitute universal subhierarchies, for they evidence in their system the several distinctive features constituting the segments of Brazilian Portuguese. As it was also observed that the excess of variation presented by the phonologically disordered learner seems to be related to the late acquisition of the segments, since according to the architecture of OT the slow demotion of constraints leads to the construction constraint-sharing strata. The importance that should be given by the analyst to the constraint sharing strata must, therefore, be emphasized although it must be stressed that the best speech therapy is not always the one which seeks to dismember them, for the demotion of fixed constraints, ranked a bit higher, may be more efficient in the treatment, by means of generalization. It must be also stressed that the discussion of certain aspects which characterize disordered speech, and which used to be discussed "on the side of" theoretical models, can now be supported in the own architecture of the theory.