What it exactly means to learn something depends on the nature of this something. In the case of a language, proficiency in semiosis using a specific language system is at stake.
General principles of learning
Learning an activity presupposes the ability to imitate one's peers. This may be based on the mirror-neuron system. The latter is, at any rate, active when a primate observes his peer gesturing. Articulatory gestures, too, are learnt by imitation.
A number of factors condition the ease, speed and quality of a learning process:
- In a behavioristic perspective, the relation between a linguistic expression and its meaning is a relation of stimulus to response. Such relations are learnt the easier the more consistent they are, i.e. the more reliably the response is associated with the stimulus.
- Frequency of the input plays a major role in learnability. A particular item, for instance a word, is ceteris paribus learnt the better the more often it occurs in the input. The same goes for a particular pattern, for instance the construction of [adjective noun] in this order. The latter presupposes that the individual items learnt have been assigned to categories which then form patterns.
- The more complex the input, the slower and harder the learning process and the worse its result. Children learn first simple linguistic units, then more complex ones. Likewise, empathic caretakers first teach the beginner simple expressions.
- All learning is partly conditioned by the feedback that the learner gets on his attempts at practising.
- One important kind of feedback is reinforcement. In the case of language learning, reinforcement generally consists in success of the intended cognitive and communicative operation. In the very early phase, reinforcement includes praise of the learner's successful attempts by his caretakers.
- Another kind is corrective feedback: caretakers take up the child's utterance, replacing an erroneous string by a correct one.
Strategies applied to language learning
Humans have an innate semiotic faculty, rooted in the brain areas responsible for language, that enables them to learn cognitive and communicative operations and a specific semiotic system for their execution (Tomasello 2009: 79). It also enables them to find patterns in the input they receive by drawing an analogy between two complex units and schematizing their common structural traits.
Of the two elementary relations, contiguity provides the basis for semiosis, while similarity provides the basis for categorization. From an age of 21 months on, children develop semantic categories like ‘animal’. As a next step, they develop word classes like ‘noun’. These are distributional categories, i.e. such a category is constituted by a distribution pattern displayed by its members. Semantic categories bear some relationship to word classes. For instance, words designating animals are nouns. Children assign a word to a word class both on the basis of contexts in which they experience the word and on the basis of its semantic category, mapping the former onto the latter to the extent possible. This is functionally-based distributional analysis (Tomasello).
A major problem in learning an operation is to determine the scope of its applicability. One of the advantages of the analytic access to a complex unit consists in the fact that it frees one of the necessity to store and retrieve the unit as a whole. In language learning, the holistic access is applied first. In getting subsequently used to the analytic access, the child sounds the limits of its applicability. This results in overgeneralizations. They are corrected – explicitly or implicitly – by preemption, i.e., by use of the irregular form which show that the rule does not apply.