Chris Weigl at serendipity writes thoughtfully in response to Bill Poser's question about "whether it is possible for ... a reading pronunciation to become so firmly fixed that subsequent intensive exposure to the spoken language does not correct it". Chris notes that "[t]here was a period of about ten years starting at the age of 16 ... during which I read voraciously in English and thereby improved my vocabulary and knowledge of English grammar, collocations and idiomatic expressions while having hardly any need to actually speak it". As a result, she developed a variety of " idiosyncrasies [that] proved relatively difficult to eradicate."
I'd like to point out that many monolingual English speakers have a similar experience in formal or technical registers, which may never come up in conversations with those around them, especially if their real-life environment is not an intellectual one. Geoff Pullum discussed this in his post on "Mispronunciation and Autodidacts".
As Chris and Bill both note, these reading pronunciations and similar mis-analyses are not random. They're the result of of over-generalization of letter-to-sound relationships, or analogy to specific similar words, or intrusion of more natural phonological patterns. This interference can happen within a language (and its orthography) or across languages, but in either case, it's psychologically natural. Therefore, the resulting errors are usually not really idiosyncratic. As with eggcorns, if one person makes the mistake, it's likely that at least a few others have done so as well. This creates a sort of "dialect without borders".
At least one linguist of my acquaintance has used this fact to construct a justification, or perhaps rationalization, for his own (relatively small) set of stubborn spelling pronunciations and malapropisms. "That's my dialect", he says. When he occasionally hears someone else speak in a similar way, he takes this as confirmation.
This brings up some of the issues that Arnold Zwicky has discussed under the heading of "The Thin Line between Error and Mere Variation". All of us believe that sporadic mispronunciation of read words -- like the mis-stressing of attributive with penultimate stress that Chris cites -- are not really dialect differences, they're just mistakes. My "that's my dialect" friend is mis-using the concept of grammar as description. Every once in a while, such mistakes do get picked up by enough members of some speech community to reach the status of a genuine variant. And as Arnold has pointed it, ambiguous cases are common. But sometimes a mistake is simply and clearly just a mistake.
Posted by Mark Liberman at September 15, 2004 07:21 AM