Deep dysphasia as a phonetic input deficit: Evidence from a single case

Background: The syndrome of deep dysphasia is characterised by an inability to repeat pseudowords and the production of semantic errors in word repetition. Several single case studies revealed that phonological decoding might be outstandingly impaired. Recovery of deep dysphasia has only been illustrated in detail for patient NC (Martin & Saffran, 1992). Dell, Schwartz, Martin, Saffran, and Gagnon (1997) tried to simulate NC's repetition performance in their connectionist lexical activation model, but it did not fit his error pattern as it assumes perfect recognition of auditory input. Aims: In this new single case study on recovery of deep dysphasia, we intended to collect further evidence for the assumption that impaired input processing is the crucial cause of the impairment. Moreover, we aimed to explain impairment and psycholinguistic parameter effects in the connectionist semantic-phonological model (Foygel & Dell, 2000) by adding a phonetic input level. Methods & Procedures: JR's performance was repeatedly assessed in the course of recovery. Errors in naming and repetition were classified according to the taxonomy of Dell et al. (1997). JR's error patterns were simulated in the semantic-phonological model to determine the naming disorder and to predict word repetition. In addition, we established an error modality analysis to disentangle input and output impairments in repetition. Thus, the source of each error could be subclassified as belonging to either expressive or receptive components of repetition. Outcomes & Results: Initially there was a sharp contrast between severely impaired word and pseudoword repetition and almost unimpaired reading aloud. During recovery, performance in naming and word repetition improved a great deal, while repetition of pseudowords remained impossible. The evolvement of real word repetition was characterised by psycholinguistic parameter effects at different points in time: concreteness before length, before frequency. The connectionist model over-predicted correct responses in word repetition as for NC. There were only few expressive repetition errors; regarding receptive errors, nonwords and null responses decreased significantly while formal errors became the dominant error type in the course of recovery. Conclusions: The development of psycholinguistic parameter effects, dissociations in performance, the computer simulations, and results from error modality analysis as well as changes of error pattern are ample evidence for the primary decoding disorder in JR. We argue that deep dysphasia can be explained by an impairment of phonetic-phonological connections in an extended version of the connectionist one-route model of repetition with four rather than three levels of auditory word processing. The improved real word repetition despite persisting failure on pseudowords is accounted for by an increase of both phonetic-phonological and lexical-phonological connection weights.