Language != thought

Reading: Language is primarily a tool for communication rather than thought, Fedorenko et al., Nature, 19 June 2024.

Do we use language to think? Reviewing research over the last twenty years, there are tests we can apply.

If some forms of thought require language, then linguistic mechanisms should be obligatorily engaged for at least those types of thinking and reasoning, and those types of thought should not be possible without language.

It seems the answer is that language is not “obligatorily engaged” and you don’t need language for thought. This is based on:

• fMRI experiments involving understanding source code (for example) where the language centres of the brain do not engage;
• pre-verbal infant studies; and
• non-human animals, who seem to reason without language.

Related to this, certain kinds of brain damage can leave you with difficulties reasoning, but with full control of language. The argument is that having language doesn’t imply having the ability to think (oh, hello LLMs!).

A second hypothesis:

If language is a tool for communication, then language should show hallmarks of efficient information transfer.

Language does indeed appear to be optimised for information transfer. For example:

The efficiency that characterizes natural languages means that it is typically not possible to make a real language’s kinship system much simpler without having it convey less information, or vice versa. Similar results have been reported for colour terms and season words, and closed-class words and grammatical markers.

The review concludes that “all types of thought tested to date are possible without language”.

Of course: “it is unlikely that our species’ success would have been possible without the cumulative culture that was enabled by the external usage of language.”