Connectionist model do not formulate autonomous symbolic rules for regular inflection, but rather treat all inflection in the same way: a network of mappings from base form to past-tense form is constructed. The network memories individual patterns and their transformations and generalises on the basis of regularities found in the network.
- Bybee (2004), Pag. 43

For the purposes of the present paper, connectionist models and the network model make essentially the same claim: type frequency is a major determinant of productivity.
- Bybee (2004), Pag. 48

My model and the connectionist model claim that type frequency is an important determinant of productivity, whereas the dual-processing model denies any role of type frequency in productivity.
- Bybee (2004), Pag. 42

The connectionist model and the network model are similar in not postulating a discrete cut-off point between regular and irregular morphology, as the dual-processing model does. Connectionist and network models attribute differences between regular and irregular morphology to quantitative differences, in particular to the relatively greater type frequency of regular patterns. [...] In my network model (and in connectionist models), regular and irregular forms and patterns are treated in the same way and there is no separate component for morphological rules. Differences in degrees of productivity are not attributed to different processing types, but to differences in type frequency and the openess of defining schemas.
- Bybee (2004), Pag. 47