Clahsen and Almazán interpreted this double dissociation as support for the dual-route assumption that grammar and lexicon are two separate systems, because this allows grammar to be selectively spared in WS and lexical knowledge to be selectively spared in SLI.7

Clahsen (1999) points out that the assumed default forms for English past tense and noun plurals are also the most frequent forms (= majority default system) and that, in order to show that default rules are independent of frequency, it is important to investigate languages other than English to inform the single–dual mechanism debate. The relevant test case would be a minority-default system

Lexical strength is the amount of processing effort it takes to retrieve a (verb) form; forms with strong lexical representation (e.g. frequent forms) require little processing effort, while forms with weak lexical representations (e.g. infrequent forms) require much processing effort. A highly frequent irregular verb with great lexical strength will be more resistant to the pressure of the regular past tense form compared to a lower-frequency irregular verb, but over time the strength of lower-frequency irregulars will increase, and so will their ability to resist.

Pinker (1984) suggested that children’s early paradigms are stored, until a child has had sufficient exposure to contrasting inflectional forms to decompose the inflected form into stem and affix (Booij 2010a). Such early storage effects typically apply to frequent forms, which are picked up early by children.

In sum, a basic distinction is made between two views of processing morphologically complex, inflected forms. Dual-route processing is based on the theoretical assumption that lexical entries are impoverished. Inflected forms are generated with rules that are not part of the lexicon, but part of grammar, and lexicon and grammar are viewed as fundamentally different processing mechanisms. Single-route processing departs from the full lexical entry assumption and abandons the fundamental split between lexicon and grammar. It is relevant to note that more recent single-route models embrace a parallel dual processing route with simultaneous lookup and computation (see Gagné and Spalding, Chapter 27 this volume). However, to our knowledge there are no observations from first language acquisition that shed light on these more subtle theoretical distinctions.