Rules vs. analogy in English past tenses: a computational/experimental study

Abstract

Are morphological patterns learned in the form of rules? Some models deny this, attributing all morphology to analogical mechanisms. The dual mechanism model (Pinker, S., & Prince, A. (1998). On language and connectionism: analysis of a parallel distributed processing model of language acquisition. Cognition, 28, 73-193) posits that speakers do internalize rules, but that these rules are few and cover only regular processes; the remaining patterns are attributed to analogy. This article advocates a third approach, which uses multiple stochastic rules and no analogy. We propose a model that employs inductive learning to discover multiple rules, and assigns them confidence scores based on their performance in the lexicon. Our model is supported over the two alternatives by new "wug test" data on English past tenses, which show that participant ratings of novel pasts depend on the phonological shape of the stem, both for irregulars and, surprisingly, also for regulars. The latter observation cannot be explained under the dual mechanism approach, which derives all regulars with a single rule. To evaluate the alternative hypothesis that all morphology is analogical, we implemented a purely analogical model, which evaluates novel pasts based solely on their similarity to existing verbs. Tested against experimental data, this analogical model also failed in key respects: it could not locate patterns that require abstract structural characterizations, and it favored implausible responses based on single, highly similar exemplars. We conclude that speakers extend morphological patterns based on abstract structural properties, of a kind appropriately described with rules.

Aspects of Japanese loanword devoicing

“…On the other hand, since words containing OCP-violating geminates are still small in number (this study contained "only" 28 words with OCP-violating geminates), its phonological property as a group is yet to be stabilized leaving a room for lexical frequencies to affect the phonological variation (see Albright and Hayes 2003;Bybee 2001;Bybee and Pardo 1981;Pierrehumbert 2001 for a relevant discussion). This hypothesis provides an interesting line of approach to model the limited influence of lexical frequencies on the naturalness judgement pattern, but verifying the hypothesis is left for future research.…”

Accent classes in South Kyengsang Korean: Lexical drift, novel words and loanwords

“…Some of the more influential studies include the following. Albright and Hayes (2003) found that several segmental factors correlate with whether an English verb inflects its past tense through suffixation or internal stem change and that subjects project these 'Islands of Reliability' to their responses in novel word experiments. Ernestus and Baayen (2003) report that speakers of Dutch inflect a novel verb terminating in an ambiguous voiceless consonant with a morpheme alternant that reflects the relative type frequency of the particular voiced vs. voiceless alternatives in the lexical inventory of Dutch verbs.…”

Morphology and Language Acquisition

“…This debate has spelled into the morphological processing literature; see Chapter 29 for discussion. Hence, our review highlights a third approach (Yang 2002;Albright and Hayes 2003), briefly considered but not pursued by Pinker and Prince 1988, one which may be of more interest to the linguistic theorists: irregular verbs are also handled by rules albeit lexicalized ones, and they are not learned by direct word association.…”

“…The Wug test is now extensively used in a wide range of studies (see e.g., Bybee and Slobin 1982;Bybee and Moder 1983;Marcus et al 1992;Clahsen 1999;Hahn and Nakisa 2000;Albright and Hayes 2003;Hahn and Nakisa 2000;Hayes et al 2009;Zhang and Lai 2010;Becker et al 2011) which subsequently have had considerable influence in linguistic theorizing (Pinker 1989;Tomasello 2003;Taylor 2003;Hay and Baayen 2005). Yet the Wug test by no means provides a direct window into morphology.…”

“…We do not discuss the use of rating tasks in assessing morphological productivity (e.g., Albright and Hayes 2003). Rating, by definition a gradient task on a scale, guarantees a gradient outcome.…”