Current Issues in Developmental Disorders by Marshall Chloë R.;

Author:Marshall, Chloë R.;
Language: eng
Format: epub
Publisher: Taylor & Francis Group

An illustrative example of deficit spread versus compensation

With respect to developmental dyslexia, the most pertinent result concerns conditions where simulations produced lasting deficits for a single process (corresponding, say, to the GPC component in the DRC model). We found that a large impairment (e.g., a 75 percent reduction of the normal level) to just one parameter (the K parameter, asymptote level) was sufficient to produce this outcome across the range of models. We assessed two additional properties. Compensation was assessed based on whether the performance of the initially damaged process was reliably different to if it had developed in isolation, unconnected to any other process. Deficit spread was assessed based on whether the performances of the initially undamaged processes were reliably different to those same processes in the normal model. Figure 4.5 depicts the developmental profiles for each of the individual processes for each architecture. Black lines show the trajectories of each process in the normally developing models and grey lines show the trajectories of each process in developmentally disordered models. Horizontal dashed lines depict the level of performance that is predicted for the level of damage, were the process to develop in isolation, and against which the action of compensation was gauged.

Unsurprisingly, in the modular architecture, early selective damage to a single process resulted in a dramatic drop in the performance level for that process. Also of no surprise was the fact that in the modular model the initially unaffected processes developed normally. Due to the lack of interconnectivity, the modular network exhibited no spread of deficit and equally, offered no compensation to the damaged process. In this case, an early deficit would result in a truly specific impairment. The pattern was different in the other four architectures. As the process of development unfolded, the effects of early damage to a single process were not isolated. Figure 4.5 illustrates the extent of deficit spread for each of the architectures (shown via the lower-than-normal developmental trajectories for initially unaffected processes) and compensation (where performance for the damaged process was above the level predicted).

Notably, while the performance of the damaged component in the latter architectures was significantly lower than the normal model, there was no reliable difference in performance between the initially undamaged processes and their normally developing counterparts, against the background of variability in the population as a whole. That is, the spread of the deficit over development was masked by the fact that the performance was within the normal range for the population, even though it was below the level it would have been if the starting conditions in each system were normal (see Figure 4.6). Superficially, the behavioural profile suggests that specific impairments, of the sort reported in cases like developmental dyslexia, are possible under a variety of neurocomputational conditions, including those that did not specify a modular functional architecture; in fact, in non-modular, interactive cognitive architectures, the effects of the damage were never truly specific but instead were widespread and subtle, with the system’s dynamics determining the degree of deficit spread and the amount of compensation following early forms of damage.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.