Pearsall links to a discussion - well, more of an argument - he had with someone about whether Ebonics - Black American English - is a language or a dialect. The criteria being - where resort to the essentially correct in terms of usage, but hardly very helpful, 'a language is a dialect with an army' is not used - seem to be periodically mutual incomprehensibility and somewhat distinct linguistic roots. Yet those criteria, applied strictly, would make Geordie a language, since it apparently has more Norse in it than is typical of Standard British English, and is periodically rather difficult to understand. But that would seem to collapse the very idea of a dialect altogether, since, by any reasonable standard, Geordie, which is to a very large degree parasitic on Standard British English, is surely only a dialect. Because I am not familiar with Ebonics, I'm in no position to judge whether it is appropriately like Geordie or not. I think, though, that if we are going accept a version of periodic mutual incomprehensibility and somewhat distinct linguistic roots as criteria for deciding on whether forms of speech count as languages or dialects, even if that version requires quite extensive levels of incomprehensibility and distinct linguistic roots, that there are rather a lot more langauges than is conventionally thought.
For example, it would suggest that most Scots are at least bilingual, in Standard British English and their regional Scottish English derivative, and that not all of those Scots are bilingual in the same two languages, that is, that the Scottish English derivatives spoken in different parts of Scotland are not the same language. It would also suggest that not only is the speech typical of Northern Italy a different language from that typical of Southern Italy, but that the speech typical of Naples is a different language from that typical of Bari, just as the speech typical of Florence is a different language from that typical of Venice.
This may sound implausible, I grant. However, the Scots my mother speaks on the phone to her sister is not, I assure you, thick: I have no difficulty understanding it. Yet it is apparently very difficult to understand if you haven't grown up hearing various relatives speak like that all the time (I confess bemusement at this: apparently my mum also has a Scots accent when she speaks Standard British English, but I've never noticed this really). However, despite there being this not-thick version of my mum's Scots which Londoners certainly find periodically incomprehensible which I find as easy to understand as the person reading the news, there are thick versions of the Scots spoken by my mum which I really, really struggle to understand. There are also other forms of Scots I have been known to struggle with. Ergo, criteria of mutual incomprehensibility certainly fulfilled (thin version of the speech periodically incomprehensible to competent speakers of Standard British English; thick version of the speech periodically incomprehensible to some of those who understand the thin version; thick versions of different Scots speech periodically incomprehensible to some of those who understand the thin version of this Scots speech), and presumably, that mutual incomprehensibility comes from somewhere, so there, I would assume, are substantive linguistic differences. So the case for Scots seems strong.
Obviously, I can't offer such strong ancedotal evidence for the claims about Italian, since more or less all Italian is incomprehensible to me, unless it is spoken as one would to a small and particularly unintelligent toddler. However, Wikipedia list 33 langauges spoken in Italy. Whilst not all of them are, as it were, dialects of Italian, most of them are, and whilst perhaps not all of them are different enough to count as full-blown languages, some of them are. Sicilian, for example, is different enough from Standard Italian that thick Sicilian has sometimes to be dubbed or subtitled in Italy, while the three versions of the Lord's Prayer reproduced here are all significantly different. It is not an unconventional opinion amongst scholars of Italian, for example, to say 'there is no such thing as Italian', meaning, teaching someone from a manual of standard Italian and expecting them to understand everyday speech anywhere in Italy would be optimistic. This makes sense if you think that standard Italian was, until the eighteenth century, a more or less solely literary language, based on late medieval poetry, and that the peninsula was not united until the mid-nineteenth century, meaning that the speeches of different parts of it were exposed to quite different influences and had quite different pressures operating on them.
I think perhaps the most helpful way of thinking about whether it would be most helpful to think about whether a form of speech is a dialect or a language is to ask whether someone who had learnt the langauge the speech seems to be a dialect of solely from hearing canonical standard forms would usually understand the speech. In order to remain a dialect, a form of speech must, like Geordie, be generally comprehensible to someone who speaks the language it is a dialect of but is only familiar with the standard form of that language. If it is not generally comprehensible to such a person - as almost all regional dialects of Italian would be to someone with only standard Italian, and as thick Scots is to many Britons - then it is another language. Whether or not that is true of Ebonics is a question I couldn't answer, although any attempt to answer it should bear in mind that, of course, there is no empirical test for it, as there is no-one who only knows a language in its canonical standard form.