The experimental results reported in Clahsen's target article clearly distinguish regular from irregular processes and suggest a basic difference between items that are productively formed and items which are stored in the lexicon. However, these results do not directly implicate any particular combinatory operation (such as affixation), nor do they distinguish inflectional items from other productive formations.

The distinction between regular and irregular morphology is not clear-cut enough to suggest two distinct modular structures. Instead, regularity is tied directly to the type frequency of a pattern. Evidence from experiments as well as from naturally occurring sound change suggests that even regular forms have lexical storage. Finally, the development trajectory entailed by the dual-processing model is much more complex than that entailed by associative network models.

Clahsen shows that “combinatorial” inflection is processed differently from “irregular” inflection. However, combinatorially regular affixes need not coincide with “class-default” affixes, that is, affixes shared by more than one inflection class and all of whose rivals are peculiar to one class. This creates a tension that may help to explain the persistence of inflection class systems.

Clahsen proposes two distinct processing routes, for regularly and irregularly inflected forms, respectively, and thus is apparently making a psychological claim. We argue that his position, which embodies a strictly linguistic perspective, does not constitute a psychological processing model.

Research into entries and rules based on English is hampered by the fact that crucial factors coincide. Clahsen avoids this problem and demonstrates his claims by working on German. Seen against the background of the immense variety of the world's languages, this successful move is potentially the first of many possible ones: Several languages offer promising configurations of the factors relevant to Clahsen, and others present new challenges.

Clahsen's view on language is intimately linked with the Chomskian distinction between competence and performance. He uses performance to verify theoretical assumptions about the underlying structure of competence. Using mostly off-line tasks, he may fail to answer the question of how language is generated and perceived in natural situations.

Clahsen's conception of inflectional rules – as being not only regular, but simultaneously only concatenative (combinatorial), general and productive, representing the default, not occurring in interfixation within German compounds, and identical to the first rules to be acquired in first-language acquisition – involves an unwarranted collapsing of morphological concepts.

Clahsen reports the results from two sets of word-recognition experiments with adult native speakers of German supporting the notion that the processing of regular (or default) inflection differs from the processing of irregular inflection. My commentary points to shortcomings in stimulus selection and inconsistencies in the pattern of results, revealing that the empirical support for the proposed dual mechanism is much weaker than Clahsen suggests.

The received view in historical linguistics is that there is always an inverse relation between token frequency and likelihood of analogical change. I have found evidence, however, of a sharp difference in frequency effects between regularization and nonregularizing analogical change. I argue that this difference can easily be accounted for by dual-mechanism models of inflection but is very problematic for pure associative-memory models.

Clahsen's characterization of nondefault inflection as based exclusively on lexical entries does not capture the full range of empirical data on German inflection. In the verb system differential effects of lexical frequency seem to be input-related rather than affecting morphological production. In the noun system, the generalization properties of -n and -e plurals exceed mere analogy-based productivity.

Clahsen's experimental data from generalization, frequency, and priming fail to support and even conflict with those aspects of his dual-route account that distinguish it from schema theories.

Analogy plays an important role in the production of irregular forms but the proposed Minimalist Morphology (MM) representations do not express this. Recent results also show that the regular forms of strong paradigms can have idiosyncratic properties that cannot be accounted for by MM. Methodological problems with an experiment are discussed and a plea for a processing explanation is made.

Clahsen has added to the body of evidence that, on average, regular and irregular inflected words behave differently. However, the dual-mechanism account he supports predicts a crisp distinction; the empirical data instead suggest a fuzzy one, more in line with single-mechanism connectionist models.

Syntax is better viewed as the dynamics of a morphogenetic field on a semantic universe of “content” words. This may take widely different forms, making the acquisition of any language by an aspiring speaker an entirely new experience. The existence of an underlying “universal syntax” might be illusory.

Not only do grammars have the dual structure that Clahsen discusses but the lexicon contains atomic, unanalyzed items, which would be still more mysterious for single-mechanism models. Forms of be in modern English are listed atomically and this is not a simple function of their morphological richness or of the fact that they move.

When referring to the organization of the mental lexicon, the distinction between combinatorial rules and lexical listing for regularly versus irregularly inflected words should be further developed to account for subregular morphological processes. Moreover, the distinction may be more or less appropriate depending on the lexical component under consideration, and it is subject to interplay with other factors that are relevant in determining the representational structure of the lexical system.

Hungarian data provide support for differences in processing regular and irregular morphologies. Stronger priming was observed with “regular” stem types compared to “irregular” ones. Use of nonwords showed a reliance on the grammatical structure of the nonword: Analogical extension of “irregulars” can be observed only in “root” contexts; in other contexts all types were largely overregularized.

The heated debate over whether there is only a single mechanism or two mechanisms for morphology has diverted valuable research energy away from the more critical questions about the neural computations involved in the comprehension and production of morphologically complex forms. Cognitive neuroscience data implicate many brain areas. All extant models, whether they rely on a connectionist network or espouse two mechanisms, are too underspecified to explain why more than a few brain areas differ in their activity during the processing of regular and irregular forms. No one doubts that the brain treats regular and irregular words differently, but brain data indicate that a simplistic account will not do. It is time for us to search for the critical factors free from theoretical blinders.

For language to function we clearly need two formal ordering principles: lexical entries and rules. Clahsen's target article provides multiple empirical evidence for this distinction, but this may be simply to overconfirm the undeniable and to overlook the hidden motor of language use and language development, namely, function. Since at least 1859, linguists have argued for the primacy of function, and these arguments are worth rediscovering today.

Recent studies of Italian past definite and past participle forms show that human performance with regular and irregular inflections is not dissociated as Clahsen's model would predict. Some performance profiles, accounted for by dual-mechanism models in terms of an underlying symbol-manipulating combinatorial procedure, are generated in Italian by the higher learnability and generalizability of phonologically regular morphological processes.