Let's start with a small dictionary of Saussurean terminology. We have to distinguish three levels of analysis:
- Saussure's langage means something similar to 'the faculty of language' or 'competence' in Chomsky's terms. In Saussurean analysis it is never really relevant, so we won't mention it here anymore.
- Saussure's langue refers to either an unspecified and abstract 'linguistic system' or to a concrete historically attested language (e.g. English, French etc). The langue is, therefore, an abstraction provided/constructed by the linguist, not a really existing. Notice that English term language is closer to Saussure's langue rather than langage.
- Saussure's parole is to be understood as a term covering all the phenomenal manifestations of a langue. The acts of parole are, therefore, concrete and physical and present variability. No two acts of parole are ever identical to each other. However, parole is the manifestation of the langue, and cannot persist without the latter. The reverse is also true: diachronically (i.e. historically) a langue grows up from the "sedimentation" of a huge series of acts of parole. Step by step, tiny changes in parole make the whole system (i.e. langue) drift towards a systemic change.
Now, the equivalency of Phonology with langue and of Phonetics with parole is not an idea of Saussure's, but was suggested by Trubeckoj (who obviously read Saussure's Cours but especially was friends with Karcevskij, another member of the Prague circle who, moreover, personally attended Saussure's classes some 20 years earlier).
Therefore, we must consider this equivalency within Trubeckoj's framework rather than the Saussurean one. Trubeckoj strongly privileged systematic phenomena to the accidental ones. He considered linguistically relevant only the phonemes, meant as abstract entities needed for distinguishing meanings of the words. Whatever is abstract automatically belongs to the langue. Hence comes the first part of the equivalency:
Phonology (meant as a set of phonemes) → langue.
Trubeckoj disregarded any type of variability as irrelevant in Phonology. Variability is concrete and physical and therefore belongs to the realm of Phonetics. In Saussure's terms, unsystematic variability is automatically reckoned as pertaining to parole. Hence the second part of the equivalency:
Phonetics (meant as an open set of physical realisations) → parole.
Notice that post-Trubeckojan Phonology is less strict with respect to variability. Let's distinguish three levels of variability:
Accidental variability: no two utterance of "the same" word or sound can be ever perfectly identical, due to the fact that humans are biological organisms rather than machines.
Even if we do our best trying to pronounce the sound [i] identically, every act of pronunciation will differ at least slightly from all the others. Accidental variants are non systematic and therefore phonetic, rather than phonological.
Free variability: quite often phonological systems allows two, or more, possible phonetic realisations of one and the same phoneme, and these variants can be freely chosen by the speaker.
Thus, in a number of European languages one can freely choose between two realisations of the phoneme /r/: alveolar [r] and uvular [ʀ]. The choice does not involve any change in meaning and is provided already at the level of the abstract system (because every speaker "knows" that both pronunciations are possible in his/her langue).
Contextual variability: some variant realisations depend on the adjacent phonemes, or context. These variants have been termed allophones by Trubeckoj. They are systematic and predictable and can be described through formal rules.
For example, English voiceless stops /p/, /t/ and /k/ are realised with aspirated allophones [pʰ], [tʰ] and [kʰ], when they occur immediately before (no sound in between) a stressed vowel, and are not preceded by /s/; otherwise the non-aspirated allophone is realised. See the realisations of /p/ in pot [pʰɑt] vs. spot [spɑt] vs. potato [pəˈteɪtoʊ] vs. plop [plɑp].
For Trubeckoj no variability belongs to Phonology since Phonology is an abstract construction made by the linguist in order to describe a language with the smallest means: the phonemes are effectively needed in such respect, while the allophones and the accidental variants can be dismissed. For other linguists, contextual variability and perhaps even free variability, are also part of Phonology. This approach is consistent with the view of Phonology as an attempt of representing the real linguistic knowledge of the speaker.