According to the philosophy of language, linguistic competence is com-posed of algorithmic skills governed by rules, which account for the learn-able, creative nature of language. Nevertheless, this idea has been put under pressure due to ample proof of sensitivity of linguistic understand-ing in context. I state that cognitive scientists should resort to anti-computationalist suggestions so that they can develop non-algorithmic views of language use which do not turn communication and learning into mysterious. There are some fields, like ecological psychology, enactivism and dynamical systems theory, which have demonstrated that we do not need to turn to rule-guided actions in order to explain complex, open-ended competences. To do so, they turn to complex interactions and couplings brought about among agents and their environment. These other opinions on linguistic competences open up the possibility to deal with metaphor and other areas of speech that have been seen as developed from others and usually considered as not being taking part in the linguistic meaning domain, owing to their non-algorithmic nature. To interpret metaphors, the ability to see significant parallels and connections between different subjects is sine qua non. It is considered a highly context-sensitive, embodied ability, and is much of the time based on relationships between people.