http://www.economist.com/blogs/economist-explains/2014/02/economist-explains-8

HONG KONG'S education department caused a furore last month by briefly posting on its website the claim that Cantonese was “not an official language” of Hong Kong. After an outcry, officials removed the text. But was the claim correct? The law says that “Chinese and English” are Hong Kong’s official languages. Whereas some people say that Cantonese is a dialect of Chinese, others insist that it is a language in its own right. Who is right—and how do dialects differ from languages in general?

Two kinds of criteria distinguish languages from dialects. The first are social and political: in this view, “languages” are typically prestigious, official and written, whereas “dialects” are mostly spoken, unofficial and looked down upon. In a famous formulation of this view, “a language is a dialect with an army and a navy”. Speakers of mere “dialects” often refer to their speech as “slang”, “patois” or the like. (The Mandarin Chinese term for Cantonese, Shanghaiese and others is fangyan, or “place-speech”.) [#ff0011]Linguists have a different criterion: if two related kinds of speech are so close that speakers can have a conversation and understand each other, they are dialects of a single language. If comprehension is difficult to impossible, they are distinct languages.[/#ff0011] Of course, comprehensibility is not either-or, but a continuum—and it may even be asymmetrical. Nonetheless, mutual comprehensibility is the most objective basis for saying whether two kinds of speech are languages or dialects.

[#ff0011]By the comprehensibility criterion, Cantonese is not a dialect of Chinese. Rather it is a language, as are Shanghaiese, Mandarin and other kinds of Chinese.[/#ff0011] Although the languages are obviously related, a Mandarin-speaker cannot understand Cantonese or Shanghaiese without having learned it as a foreign language (and vice-versa, though most Chinese do learn Mandarin today). [#ff0011]Most Western lingists classify them as “Sinitic languages”, not “dialects of Chinese”.[/#ff0011] (Some languages in China, like Uighur, are not Sinitic at all.) Objective though it may be, this criterion can annoy nationalists—and not just in China. Danes and Norwegians can converse, making some linguists classify the two as dialects of a single language, though few Danes or Norwegians think of it this way.

In China the picture is [#ff0011]further confused by the fact that one written form[/#ff0011] unifies Chinese-language speakers (though mainland Chinese write with a simplified version of the characters used in Hong Kong and Taiwan). [#ff0011]But this written form is not a universal “Chinese”[/#ff0011]: it is based on Mandarin. The confusion arises because many people consider written language to be the “real” language, and speech its poor cousin. The same reasoning can be used to classify Arabic as a single language, though a Moroccan and a Syrian, say, cannot easily understand each other. Ethnologue, a reference guide to the world's languages, calls Chinese and Arabic "macrolanguages", noting both their shared literature and the mutual (spoken) unintelligibility of many local varieties, which it calls languages. [#ff0011]For the most part, linguists consider spoken language primary: speech is universal, whereas only a fraction of the world’s 6,000-7,000 languages are written. This is behind the linguist’s common-sense definition: two people share a language if they can have a conversation without too much trouble.[/#ff0011]