Nearly two decades ago I wrote a paper on terminological difficulties surrounding the classification of Sinitic languages entitled "What Is a Chinese 'Dialect/Topolect'? Reflections on Some Key Sino-English Linguistic Terms," Sino-Platonic Papers, 29 (September, 1991), 1-31. (Available online at http://www.sino-platonic.org/) In that paper, I did not go deeply into the question of the utility of mutual intelligibility for determining the difference between a language and a dialect, mainly because it is a red hot can of worms, but also because people say such nonsensical things as that "A language is a dialect with an army and a navy." Now, in preparation for updating my 1991 paper, I would like to revisit the matter of mutual intelligibility to see whether it can somehow be salvaged for purposes of taxonomic classification.
One thing is certain: a monolingual speaker of Cantonese cannot understand a monolingual speaker of Mandarin and vice versa. There is zero intelligibility between the two languages. In fact, even within the huge collection of speech forms that fall under the umbrella of "Mandarin," there are many varieties that are more or less mutually unintelligible. On July 4, 1987, I was climbing up Emei Mountain in Sichuan. It was a hot, muggy day, and our small party (my wife, son, sister, and I) were struggling up the steep slopes. We were astonished to see large groups of short, older ladies speeding upward. As we listened to their chatter, we couldn't understand a word of what they were saying. My wife, who grew up in Chengdu, and so speaks Chengdu Sichuanese (Szechwanese) – generally considered to be a variety of Mandarin — suspected that the older ladies were speaking a non-Sinitic language. When we inquired at the little shops along the way, we were informed that the groups of pilgrims were speaking one or another type of Sichuanese from nearby districts. Mind you, Mount Emei is only 150 kilometers (93 miles) southwest of Chengdu City. And even in Chengdu there are expressions that don't sound like Modern Standard Mandarin, such as MO DEI LO ("we don't have any" or "there isn't any"), LANGGE GAO DI ("what's going on?"), ZAGO ("how is it?"), and CHUIZI ("penis"; I'm not sure, but perhaps this originally meant "hammer"). (Forgive me for not recording the tones and sounds exactly right; these are just expressions I have picked up casually — I may not even have the meanings down perfectly.)
It is commonly claimed that there is only one "Chinese" language, and that all of the variants of that language are dialects of it. This conception of there being only one "Chinese" language plays havoc with efforts to classify the countless varieties of Sinitic speech forms into meaningful groups, branches, languages, and dialects, as is normal for other large families or groups of languages.
The old canard that "when the dialects are written down they are the same" is simply untrue, since what gets written down are not the regional variants but standard Mandarin (and in earlier times Classical Chinese, a dead language for at least two thousand years). If one, as a tour de force, does contrive to write unadulterated Cantonese or Taiwanese, for example, they will be as hard for a reader of Mandarin to understand as spoken Cantonese or Taiwanese is for a speaker of Mandarin to understand.
So the question is this: how do we break through the taxonomic impasse presented by the claim that there is only one "Chinese" language for a billion or so speakers? I'm hoping to establish a rational system of classification whereby Sinitic is viewed either as a family or group of languages. The jury is still very much out on whether Sinitic is part of a Sino-Tibetan, Tibeto-Burman, Austronesian, etc. family. I believe that one of the main reasons for the failure to gain acceptance for any of the dozen or so external relationships that have been proposed for Sinitic is that the internal relationships of the family / group remain remain murky.