What Is the Difference Between a Language and a Dialect? An Army, a Navy and a Line on a Map

The nearest thing to a scientific test is mutual intelligibility: if two speakers can understand each other without learning the other's speech, they speak dialects of one language, and if they cannot, they speak different languages. The test is useful and it fails at the edges, because intelligibility is a matter of degree, is often one-directional, since Portuguese speakers understand Spanish better than Spanish speakers understand Portuguese, and varies with exposure and goodwill. It also produces a dialect continuum: from Portugal across Spain, southern France and Italy each village understands its neighbours and the ends of the chain are foreign to each other, and the same is true across the Slavic lands and the Arab world, so that where the languages begin and end along the chain is a decision, not a discovery.

The cases in which the linguistic answer and the official one disagree are the instructive ones:

• •Chinese: Mandarin, Cantonese, Shanghainese, Hokkien and the rest are as different as the Romance languages, and are called dialects because they share a writing system and a state that calls them so
• •Scandinavian: Danish, Norwegian and Swedish are largely mutually intelligible and are three languages because they are three countries
• •Hindi and Urdu: one spoken language, called Hindustani, written in two scripts and named for two nations since 1947
• •Serbian, Croatian, Bosnian and Montenegrin: one language, Serbo-Croatian, until Yugoslavia broke up, and four since, each with its own academy and dictionary
• •Arabic: a written standard shared from Morocco to Iraq over spoken varieties that are mutually unintelligible at the extremes, and called one language for religious and political reasons
• •Scots and English, Luxembourgish and German, Moldovan and Romanian: distinctions made or unmade by governments within living memory

Standard languages are dialects that were chosen. The dialect of the capital, the court or the printers becomes the written norm, is taught in schools, is used in law and broadcasting, and comes to be regarded as the language of which the others are deviations; the English of London, the French of Paris, the Tuscan of Florence and the High German of Luther's Bible were each one dialect among many before they were the standard. The process needs institutions, a dictionary, a grammar, an academy, a school system and usually a state, and it is why the Weinreich joke is right: a dialect acquires the status of a language when it acquires the apparatus, and Norwegian, which was Danish in writing until the nineteenth century, became a language by deliberately building one.

The label decides what is taught, printed and protected. A variety classed as a dialect is left out of schools and its speakers are marked as uneducated for speaking as they do; a variety classed as a language gets a grammar, a literature and a claim to territory, which is why separatist movements begin by writing dictionaries and why states resist. UNESCO counts about 7,000 languages in the world and the count depends entirely on how the continua are cut; the Ethnologue catalogue's figure has grown by hundreds as varieties have been reclassified. Linguists mostly avoid the word dialect in their own work and speak of varieties, which is accurate and sidesteps the argument, and they note that every variety, standard or not, has a full grammar and is nobody's failed attempt at anything else.

No linguistic test cleanly separates a language from a dialect: mutual intelligibility is a matter of degree and breaks down along dialect continua, and the varieties called Chinese dialects are more different than the Scandinavian languages. The distinction is made by states, scripts and institutions, a language being a dialect that has acquired a standard, a dictionary and usually a border, which is what the joke about the army and the navy means, and the label decides what is taught and whose speech is treated as correct.