Input Gain (with Daragh Sibley). We built a simple neural network model in which the space of learned and novel items can be visualized as points on a grid. The demos show how the control paramter input gain alters the organization of the grid. At low gain, the grid is organized according to a simple rule learned in the training set. Higher levels of gain amplify the influence of learned exceptions to the rule.

• Kello, C. T., Sibley, D. E., & Plaut, D. C. (in press). Dissociations in performance on novel versus irregular items: Single-route demonstrations with input gain in localist and distributed models. To appear in Cognitive Science.

Forward Modeling (with David Plaut). We trained a neural network model to generate acoustic outputs on the basis of articulatory inputs. Inputs and outputs came from the recordings for one speaker in the MOCHA articulatory speech database. The demos show a few spectrograms of the model targets and outputs. Click on the spectrograms to listen and compare. This forward model is planned to be one component of a larger model of phonological development.

• Kello, C. T., & Plaut, D. C. (2004). A neural network model of the articulatory-acoustic forward mapping trained on recordings of the vocal tract. Journal of the Acoustical Society of America, 116, 2354-2364.