I don't think that reports of mutual intelligibility are a valid basis for considering two speech varieties to be the same language, for two reasons. One is that such reports do not necessarily reflect the intrinsic intelligibility of the two varieties. All too often, they mean that so-and-so, a speaker of A, also understands B. This may, however, be due to at least passive bilingualism. Only if one really knows that the report means that a speaker of A can understand B without prior exposure are such reports to be taken seriously.<br>
<br>More importantly, intelligibility is not suited for this purpose because it is not an equivalence relation: it is not transitive. That is, for purposes of classification, it should be true that if A and B are mutually intelligible and B and C are mutually intelligible then A and C are mutually intelligible. It is, however, false as one can easily find chains where mutual intelligibility obtains between neighbors but more distant varieties are not mutually intelligible.<br>
<br>If two varieties are not mutually intelligible, they are arguably not varieties of the same language, but the mere fact that two varieties are mutually intelligible is not a valid basis for concluding that they are varieties of the same language.<br>
<br>