Undeniably members of the species homo sapiens produce and understand speech, and many of them are able to read and write. They do this in very different varieties.  The sound repertoires of the more than 7000 languages that are still around today vary widely, as do their grammatical structures, and the meanings that their lexical items code for. For instance, some languages have a sound repertoire of only a dozen phonemes, whereas others have more than a hundred; some languages have a very elaborate system of morphological markers, whereas others are very limited in their morphological inventory; some languages make semantic distinctions in one domain, others in another domain. Further, sign languages are expressed by movements of hands and face, whereas spoken languages are expressed by movements of the vocal tract. In addition to the variability in the world's languages, there is individual variation in language skills within the population of any given language community. Some people command only a limited vocabulary and simple sentence structures, whereas others are polyglots speaking multiple languages fluently, or can do simultaneous translation between languages.

Moreover, in our daily language use, linguistic tokens never occur in isolation. They go hand in hand with co-speech gestures, facial expressions, etcetera. If there is such a thing as language at all, it is inherently multi-modal. Evolutionary, language is cobbled together from different knowledge sources and computational subroutines.

Despite all this variability and multimodality, there is a certain tendency in both linguistics and the cognitive neuroscience community to try to reveal the underlying essence of language, whereby some aspects are seen as more key than others, and reference is made to ‘the language system’ or ‘the language network in the brain’.

In contrast, I will argue that language is not a natural kind. What we refer to with the term ‘language’ has to be decomposed into contributing knowledge types and subroutines. I will illustrate this for speaking a word, gesturing a verb, composing a sentence interpretation, and inferring speaker meaning. I will advocate a non-essentialistic view of language, based on the conviction that to account for the full picture of human language skills we are not helped by a distinction between essential and non-essential aspects of speech and language.