Theoretical and methodological difficulties in the use of computational cost as a predictor of processing demands and language impairment
Computational cost and computational complexity are closely related, since complexity is associated with the amount of resources (usually time and memory space) required to solve a problem. In sentence processing, the grammatical encoding of a message to be verbally expressed or the parsing of a sentence to be comprehended involves constructing syntactic objects from lexical items in real time. The cost of this process is expected to be a fraction of the total sentence processing cost and can be informative in the screening of language impairment. The assessment of syntactic processing abilities is not, however, a straightforward task. Computational complexity in natural languages is a far from settled matter and a metric of syntactic complexity is not available. Moreover, in the context of minimalist program it is not clear the extent to which computational complexity can be defined independently from interface demands. Apart from that, the behavioral assessment of the syntactic processes carried out prior to semantic interpretation and referential mapping challenges psycholinguistic methodologies. This presentation aims to discuss which factors can be relevant for a metric of computational complexity in the light of model of on-line computation based on minimalist assumptions, and to discuss methodological options for the assessment of syntactic abilities in the screening of language impairment.
Comments