Recent studies of orthographic variation have demonstrated that ideology plays a central role in determining which spelling variants are adopted by a community. This study examines the role of ideology in diachronic changes in spelling variant usage in Canadian English. Previous research has shown that patriotic Canadians are opposed to American spelling variants. We hypothesized that American spelling variant usage decreased during periods in which the United States was viewed negatively in Canada, such as the Vietnam War era. Furthermore, we also hypothesized that trends set during periods of anti-American sentiment have resulted in an overall decrease in American spelling variant usage in Canada over the last century. We gathered over 30,000 tokens of spelling variants spanning a period of approximately 100 years. Our results corroborate the first hypothesis but reject the second hypothesis, leading to a complex view of the role of ideology in diachronic change in Canadian English.

The sociolinguistic enterprise has demonstrated that speakers manipulate linguistic variants as they construct their speech style. Contrary to this expectation, this study introduces specific cases in which stylistic variation is highly constrained. We examine the verbal negative suffix in Kansai vernacular Japanese. We first demonstrate that this variable indexes speech style. We then show that in a few specific contexts, such as following the verb stem shir- ‘know’, speakers overwhelmingly use a single variant, in this case, shira-n ‘not know’. We point out that the unusual forms such as shira-n all have a high relative frequency compared to the other forms in their paradigms. Complex forms such as the English word insane, which occur more frequently than their parts (in+sane), are difficult to decompose. We claim that this phenomenon also impedes stylistic variation as forms such as shira-n are not readily decomposed into verb stem + negative suffix.

This study presents apparent-time changes in the morphology of the expression mitai-na ‘similar to’. Based on apparent-time data, we argue that the morphological boundary between mitai and the attributive morpheme -na in the phrase mitai-na has disappeared, and that this complex phrase is now processed as a monomorphemic form. We suggest that relative frequency is the key to understanding the results. We further supplement our argument with data on the standardization of the adverbial adjective form in the Kansai dialect. Young speakers overwhelmingly use the standard form of adverbials for all adjectives except two: yō ‘a lot, well’ and hayō ‘quickly, early’ (instead of Standard Japanese yoku and hayaku). The three linguistic forms that display unusual behavior (mitai-na and the adverbial forms of yō and hayō) all have a high relative frequency. We conclude that when a complex form occurs more frequently than its components (high relative frequency), then it behaves as a monomorphemic unit. The irregular adverbial forms are leftover from an obsolete system, in the same way that many English irregular past forms are leftover from the Germanic strong verb system. In contrast, the irregular form mitai-na emerged from and competes with the regular inflection paradigm for mitai, illustrating a previously undocumented path for the diachronic emergence of irregular morphology.

Linguists debate the nature of grammatical knowledge. Many argue it is innate knowledge of syntactic structure that we use when generating utterances; others argue it emerges from linguistic experience, and forms exemplars for modeling novel utterances. Yet, still others argue that grammatical forms are processed in parallel by both types of knowledge (innate or otherwise), and crucially, that these two processing routes compete with each other.Our objective is to support the dual route argument with a corpus study illustrating the interaction of these two types of knowledge. We interpret two proxies as indicators of these two types of knowledge: syntactic complexity for generative knowledge and dispersion for emergent knowledge. Previous psycholinguistic work has shown that increased syntactic complexity correlates with increased judgment reaction times. Conversely, increased dispersion correlates with decreased judgment reaction times. If these two processing mechanisms compete, then we predict an interaction: specifically, we hypothesize that the more dispersed a linguistic form is, the less influence syntactic complexity has.We conducted a mixed-effects logistic regression analysis on case marker omissions in Japanese. Our results show that a casual speech style, a dispersed object-verb pair, and a syntactically simple noun phrase for the object correlate with increased case marker omission. More importantly, syntactic complexity and dispersion interact: as dispersion decreases, the estimated coefficient for syntactic complexity increases. These results support the claim that generative knowledge and emergent knowledge compete during language processing.

This work presents a method of word sense clustering that differentiates homonyms and merge homophones, taking Japanese as an example, where orthographical variation causes problem for language processing. It uses contextualised embeddings (BERT) to cluster tokens into distinct sense groups, and we use these groups to normalise synonymous instances to a single representative form. We see the benefit of this normalisation in language model, as well as in transliteration.