We compare the use and function of two discourse particles that show many similarities: Canadian English eh and Canadian French hein. Surveys of anglophone students at the University of Toronto and francophone students at Université Laval reveal that these particles have similar discourse functions and that there are many parallels in their patterns of use. However, francophone students report a higher use of hein than do anglophone students of eh. Moreover, francophones have more positive attitudes towards constructions with hein than do their anglophone counterparts with respect to eh. In addition, eh—used both less often and valued less positively—has taken on additional functions as an identity marker: it is used to identify speakers of Canadian English and, in print, to evoke Canadian solidarity. In contrast, hein does not have an identity marking function. We propose that the development of an identity marking function for eh—and the lack of such a function for hein—reflects differences in how linguistic identities are constructed in English and French Canada.

Similarities between varieties of the same languages can be explained in terms of shared retentions or innovations. Conversely, differences can be explained in terms of new vocabulary items, divergent changes, language contact, and dialect contact. The latter has been challenged by proponents of monogenetic theories. Evidence for and against monogenetic hypotheses are considered on the basis of two case studies. First, I demonstrate that the dialect enclave of Lunenberg County, Nova Scotia, is a mixed colonial dialect. Second, I argue that the phenomenon of Canadian Raising is the result of dialect mixture. The Canadian English data provide evidence for a connection between dialect contact, mixture, and genesis. The data support the idea that there is a deterministic outcome in situations where the target language is not spoken by a prior-existing population, which in turn accounts for why widely separated varieties of English are similar.

The notion ‘drift’ plays an important role in the development of the modals have to and must in early Canadian English in relation to British and American English during the late 18th and early 19th centuries. Have to is first found in texts that reflect informal usage, and for the period in question (1750–1849), have to is only attested with deontic readings; the data suggest that its rise was not exclusively conditioned by the defective paradigm of must. Must maintains its epistemic function in relation to its Late Modern English competitors. In early Canadian English, changes progress gradually, with individual variables following different directions. Canadian English epistemic must lags behind, while deontic have to has spread more quickly in North America, with Canadian English more progressive than British English varieties, but less so than American English. Within a more general drift towards have to, Canadian English shows independent development in successive periods.

A socially stratified sample—the Toronto English Corpus—together with the construct of apparent time (with speakers aged 10–90 years) reveal that certain features are declining, including future will, deontic have got to, possessive have got, intensifier very, and the sentence tag you know. On the other hand, some features are on the rise, including future going to, deontic have to, possessive have, intensifiers really and so, and sentences tags such as whatever, so, and stuff like that. The younger generation is pushing these changes forward more rapidly. While some developments date back hundreds of years in the history of English, they are not particular to Canada, and are consistent with research on other English corpora. Other changes appear to be progressing in a unique way in Canada, including deontic and possessive have. I argue that the broader socio-historical context is a critical factor: geographic and economic mobility as well as changes in communication technology may explain the rapid acceleration of certain types of linguistic change.