Language families
Language or Dialect: Who Actually Decides
Mutual intelligibility, army-and-navy politics and how classification schemes such as the Linguasphere Register handle the boundary.
There is no universally accepted criterion for distinguishing between a language and a dialect. Mutual intelligibility - whether speakers of one variety can understand speakers of another - is a common test. But it isn’t always a clear call.
Mutual Intelligibility Is the Best-Known Test, but Not the Final One
Mutual intelligibility is the most common measure linguists use to separate languages from dialects.
For example, many linguists consider mainland Scandinavian varieties to be mutually intelligible dialects of one language, but the speakers view them as different languages. Conversely, the Chinese varieties they speak are not mutually intelligible, but are considered dialects of one language.
The University of Groningen’s linguistic research highlights how this test can falter. It notes that mutual intelligibility does not work well in a dialect continuum, where varieties grade into each other. Nor does mutual intelligibility always give a clear answer about how divergent dialects must be to be considered separate languages.
Why the Test Breaks in Real Speech Communities
Dialect continua are common in many language families, breaking the test’s ability to form a clean boundary between languages and dialects. As anyone who has spoken to their friends knows, there’s plenty of room for a great deal of variation inside what most people consider the same language. That becomes a particular problem for mutual intelligibility as a test, as it’s not always a precise, objective measurement.
The Leiden University study shows there are different ways to gauge speakers’ mutual intelligibility. One is judgment testing, in which listeners judge the comprehensibility of the items. The other is functional testing, in which speakers use a variety to carry out language functions, such as reading, retelling, and translation.
The Leiden study defines mutual intelligibility as separate from both of those, as the mean of intelligibility from any variety A to any variety B, and from B to A. But it’s hard to know what to make of that mean, given the different results you can get with judgment and functional testing.
The EMNLP/Computational Linguistics paper proposes a universal cut-off of LDN = 0.51, as a possible way to distinguish between dialects and languages. But that’s not the clean boundary many people imagine.
The UTEP technical report notes that many dialect-language decisions are commonly made with an intuition of a 70% mutual intelligibility threshold.
Linguasphere’s Solution: Classification by Layers, not a Single Border
Linguasphere’s Register of the World’s Languages and Speech Communities has a more layered approach. It calls varieties with no or little mutual intelligibility “outer languages,” diverging dialects or closely related languages “inner languages,” and minor variants “dialects.”
In Linguasphere’s schema, there are 4,994 outer languages, of which 4,557 are still spoken, with 13,840 inner languages with 8,881 constituent dialects within them, subdivided into 694 linguistic sets.
“The Linguasphere Register acknowledges the coexistence of different viewpoints. It does not pass a value judgment on any of the inner languages it lists.” Indeed, it does not even claim that an outer language is not a dialect for another outer language.
The Boundary is Also Institutional, not Just Linguistic
If mutual intelligibility is not the answer, and Linguasphere is something of a patchwork of different systems, then what decides where to draw the line? Well, it depends on who’s making the decision, and why.
The Oxford University Press chapter excerpt notes that scholars commonly use “dialect” to refer to a local variety of a language.
By that definition, most languages are collections of dialects. What makes a dialect a language, rather than a variety of some larger language? Sometimes it’s politics.
Catalan and Castilian, for example, are not mutually intelligible. But Castilian is generally counted as the Spanish language, and Catalan is counted as a separate language. That’s an unusual case, but many decisions are as much political as linguistic.
Where the Article Should Stop Short
There is no single international or government organization with the power to label something as a language or a dialect. Many sources classify according to whether they consider a speech variety to be a language or to be a dialect, but that’s based on the source’s internal rules and purposes.
So sometimes languages are counted as dialects, and sometimes dialects are languages, depending on the source. That’s an unsettling answer to the question of what makes a language a language, rather than a dialect. But it’s the honest one.
Where this sits in the Register
Written by the Linguasphere desk. Classification data referred to here comes from the Linguasphere Register — see sources and attribution.