From a learning-based perspective, manoeuvres designed to resolve competition are a symptom of analytic ‘overextraction’. Individual contrasts are isolated from the larger system of contrasts in which they function and the ‘grammar’ serves an essentially remedial role in correcting the original overgeneralization.
Rules that isolate form alternations from the system of contrasts in which they function appear to be a relic of an essentially pedagogical perspective. Such rules are useful for drawing grammatical generalizations to the attention of the language learner in an explicit learning context. But they have no established relevance to psychologically plausible models of language processing or acquisition. Many of the other idealizations of formal grammars discussed above can also be traced to the pedagogical origins of contemporary grammatical traditions. The notion of “scientific compactness” invoked by Bloomfield refers to the length of an orthographic or phonemic description of a pattern or system. A compact description that draws recurrent patterns to the attention to the reader may be of pedagogical value, though this economy has never been shown to have any psychological relevance. The same is true of‘redundancy-free’ models of the lexicon, which adopt an organizational property that is useful for dictionaries, but of no proven relevance to human lexical knowledge.
In sum, a fully regular grammatical description and a redundancy-free lexicon both contribute to the concision of a pedagogical grammar. Yet the economy achieved by a maximally concise grammatical description does not translate into advantages for the speaker or hearer. The problem with a fully irregular language is not that it violates a theoretical economy principle but that the lack of patterns inhibits the generalizations required to learn the language. A perfectly regular language would facilitate generalization but also incur costs in acquisition and processing due to cue competition between similar forms (Ramscar et al. 2010; Arnon and Ramscar 2012). For the speaker or learner, there is no ‘perfect’ (or even perfectly economical) language design, just different trade-offs between communicative strategies. Various structural or social factors (perhaps including some of the factors suggested by Trudgill 2011) may influence the way that different languages resolve these conflicts.