In discussions of stable linguistic variation one of the issues is the diachronic comparability of the data sources analysed. Our studies have been based on personal letters, which previous research has shown to come close to interview data, for example, in the degree to which they show speaker/writer involvement (see 3.2.1). Since even persistent genre features are apt to change with time (Biber 2001, Biber & Finegan 1997), the CEEC data was put to the test by analysing the stability of its part-of-speech (POS) frequencies, notably those of nouns and personal pronouns (Saily, Nevalainen & Siirtola 2011).
The frequency of nouns showed a small but steady falling trend during the 270 years covered by the tagged and parsed version of the CEEC, about two percentage points from the first period to the last. Part-of-speech frequencies also revealed systematic gender-based variation that persisted over time: female writers were found to use more pronouns than male writers, whereas male writers used more nouns than female writers. It was also shown that male writers typically used more pronouns in their letters to female recipients than to their male recipients (Saily, Nevalainen & Siirtola 2011: 182-183).
There are of course other factors besides writer gender that may be reflected in POS frequencies, including the social status of the writers, the kind of education they had received, and their relation to the letter recipients, as well as the topics of the letters. But since variation of this kind is found in spoken and written interaction alike, these observations point to style-shifting based on the writer's (and addressee's) gender.8 The fact that female writers retained multiple negation longer than their educated male contemporaries suggests that female usage may have been, on average, closer to the spoken registers of the time. If this is the case, and since women emerged as leaders of most of the changes we have studied in this book, we can assume that a vast majority of linguistic changes took place in the colloquial language of the time. Changes in 'vernacular universals' appear to have been no exception in this respect.
As the eWAVE statistics indicate, the features used to describe nonstandard varieties of English vary from pervasive to existing but rare. This inherent variation is not surprising considering that the varieties themselves have different histories and range from pidgins and creoles to traditional L1 dialects. It is therefore also not surprising to see many of their features change over time and have varying time depths historically. Moreover, regional varieties can foster linguistic complexity, such as the Northern Subject Rule in subject-verb concord discussed in 8.3. The fact that the Type of Subject Constraint triggered the singular suffix with NP subjects and the zero form with pronoun subjects may be reflected in the subject type hierarchy with default singulars in Early Modern English.
To put 'vernacular universals' on a more concrete typological footing, Szmrecsanyi and Kortmann (2009: 37) show that none of the morphosyntactic features identified as such is attested in more than 80 per cent of the varieties of English worldwide. They find that some of these features are, however, more widespread in the Americas than elsewhere globally and could therefore be labelled as 'areoversals'. We hope that our discussion has helped put the history of some of these features on a more concrete footing by tracing the complexity of the local diachronic processes before the rise of a codified standard language.