The child learning the sound system of his language is faced with tasks at two levels:

1. He learns the phonological categorization of the acoustic input.
2. He learns the idiomatic pronunciation of expressions of this language.

The first task is the development of the phonological system of the language. This includes the establishment of phonological units and rules. They are abstracted from phonetic percepts which are analyzed and categorized.

For instance, the child learning English understands that the first sound in tear and the last sound in soft, although phonetically different, belong to the same phoneme /t/, while the first sound in dear and the middle consonant in waiter, although phonetically similar, belong to different phonemes, viz. /d/ and /t/ (Ladefoged 1975). At the same time, he learns that the distinctive feature distinguishing /t/ from /d/ may have different phonetic realizations which are determined by context-dependent rules.

The order in which the speech sounds are acquired depends on various factors, some of which are systematic. The markedness relations between sounds that are in opposition determine a system taking the form of a complexity hierarchy governed by the principle of unilateral foundation (Jakobson 1968).

Therefore, /t/ is acquired before /k/ and /e/ is acquired before /ø/ (e.g. in French or German). Likewise, simpler syllable structures as in pot are acquired before more complex ones as in spot. As long as marked members of a paradigm have not been acquired, their unmarked counterparts step in. Thus, the child says [tʌp] instead of [kʌp] (cup) and pot instead of spot.

The child's second task consists in pronouncing expressions of the language the same way as his models do. For instance, if he is learning Spanish, he will learn to pronounce the two different r sounds [r] and [ɾ], while if he is learning English, he will learn to pronounce the retroflex [ɽ]. On the other hand, acquisition of the speech rythm and melody characteristic of the variety spoken by his models does not posit an extra problem because the child was constantly exposed to it before being born. They become so deeply entrenched that they are the hardest part to get rid of in learning a second language.

At the beginning, the child does not yet know which sounds are subsumed under the same or different phonemes. For instance, upon hearing English [tʰ], [t] and [d], the child can distinguish them, but does not know that the difference between [tʰ] and [t] does not matter for the English system. After half a year of age, the child starts categorizing sounds as representing phonemes. Once this is achieved, the child will neglect phonetic differences that do not matter for his language. This attitude persists for the rest of his life unless the person is exposed, early enough, to other languages where other phonetic differences matter for phonemic categorization.