, or the "tooth", to indicate that
a sound is dental, rather than alveolar 
.
/, we have chosen to ignore much
detail as irrelevant to the linguistic content. Much
of what we are doing in phonology is an attempt to separate the
linguistic from the non-linguistic, the systematic from the
accidental, the predictable from the random.
Much of phonology rests on one assumption. That is, that the linguistic content of an utterance can be captured using a small number of basic elements. These elements are called phonemes. The set of phonemes in a language is the smallest possible set required to "spell" out the words of the language. Note that "spell" here means to lay out exactly what sounds the word contains. Different words contain different sequences of sounds.
When making a transcription, we may aim to provide only a
broad or phonemic transcription. In
this case, we only want to note which phonemes are used, without
capturing any phonetic detail which is either random or is predictable
by rules. A narrow transcription shows more phonetic
detail, and there is no limit to the amount of detail that can, in
principle, be captured. Narrow transcriptions frequently make use of
diacritics, such as a little circle to indicate that
a sound which is usually voiced is here voiceless , or the "tooth", to indicate that
a sound is dental, rather than alveolar .
Note in passing that I am not going to distinguish between square brackets [...] and back slashes /.../. Some authors, including Ladefoged use slashes for a phonemic transcription and square brackets when more phonetic detail is involved. I will use slashes everywhere and you are free to do the same.
| p | b | t | d |
| m | n | ||
| f | v | s | z |
Some general patterns are obvious here. Stops and fricatives occur in voiced/voiceless pairs. Stops, nasals and fricatives occur at each place of articulation (where /f,v/ are considered as labials). How might we capture this regularity?
Consider the following way of describing these sounds:
| p [not voiced, not alveolar, no airflow] | b [voiced, not alveolar, no airflow] | t [not voiced, alveolar, no airflow] | d [voiced, alveolar, no airflow] |
| - | m [voiced, not alveolar, nasal airflow] | - | n [voiced, alveolar, nasal airflow] |
| f [not voiced, not alveolar, oral airflow] | v [voiced, not alveolar, oral airflow] | s [not voiced, alveolar, oral airflow] | z [voiced, alveolar, oral airflow] |
or, more economically,
| p [-V, -A, S] | b [V, -A, S] | t [-V, A, S] | d [V, A, S] |
| - | m [V, -A, N] | - | n [V, A, N] |
| f [-V, -A, O] | v [V, -A, O] | s [-V, A, O] | z [V, A, O] |
Here we used three features to classify the 10 sounds of Blork. Two of these are binary features, meaning they can just take on two values: voice: [+/-V] and place: [+/-A]. The last feature can take on three values: manner: [S, N, O]. What do we gain by this system?
We can now express the fact that /t/ is different from /s/ in the same way as /p/ is different from /f/ (manner feature only). Likewise, /s/ and /z/ are different in exactly the same way as /p/ and /b/ are different.
These features are rather simple, and they directly express the organization we first observed in the Blork consonant inventory. We could have chosen only binary features, in which case we would need 2 new features to do the work of our one 3-valued manner feature. We could use [+/- Nasal] and [+/- Stop], in which case we would get:
| p [-V, -A, -N, +S] | b [V, -A, -N, +S] | t [-V, A, -N, +S] | d [V, A, -N, +S] |
| - | m [V, -A, +N, +S] | - | n [V, A, -N, +S] |
| f [-V, -A, -N, -S] | v [V, -A, -N, -S] | s [-V, A, -N, -S] | z [V, A, -N, -S] |
Suppose we chose to use these features, we might wounder why this language has [+N,+S], [-N,+S], [-N,-S], but not [+N,-S]. This last class of sound would be not a stop (hence, probably a fricative) and would be a nasal. Nasal fricatives are virtually impossible, as the release of air through the nose prevents the speaker from generating enough pressure to generate a turbulent airflow through a constriction. So we can see a perfectly good reason why the sounds of type [+N,-S] do not occur, but to do so, we had to step outside the formalism for a moment and consider the system which is actually producing the sound.
This is the heart of the matter. Phonology is an attempt to understand the systemic elements of sound. Having a system means having a formalism. But understanding requires knowledge of both the formalism and the phonetic reality. Once you have understood this, you know why it is so hard to draw the line between phonetics and phonology.