2.

The sound system: phonology

2.1 Phonology: The study of the abstract categories that organise the sound system of a language (in
our minds).

-individual speech sounds

-groupings of speech sounds in syllables or words

2.2 The phoneme

Due to the different ways of pronouncing e.g. r in different words we must conclude that our idea of
the r-sound is an abstract, mental category, rather than a phonetic fact. We call such a category
phoneme /ɹ/, a physical realisation of a phoneme is a phone (when you physically manifest the
previously abstract mental concept of /ɹ/ in an actual word, e.g. rip; [ɹ]). Alternant realisations of the
same phoneme are allophones of the phoneme. E.g. both [ɹ] in rip and [ɹ̥] in trip are allophones of
the phoneme /ɹ/.

- (The dot on the bottom indicates that the sound is voiceless). These sorts of articulatory details in
the transcription are called narrow transcription.

- The reason why we know that [ɹ] and [ɹ̥] are realisations of the same phoneme is because they are
in what is called complementary distribution. This means, that one cannot occur where the other
does. [ɹ̥] only occurs after voiceless consonants (pray, crude, trap) whereas [ɹ] only occurs in the
beginning of words and between two vowels (wrap, room, carrot). The distribution (the different
positions where a speech sound can and cannot occur in the words of a language) of the two clearly
indicates that while they may be both the realisations of the same phoneme, they are nonetheless
distinct in actual usage. (Another argument for why they belong to the same phoneme is simply that
they differ in only one aspect, namely voicing.)

2.3.1 Minimal pairs

- Three strategies to investigate how a language categorises phones into a system of phonemes.

-1. A phoneme being replaced means that the word changes meaning (lap vs tap (minimal pair) the
phones /l/ and /t/ can distinguish different words in English and therefore must belong to different
phonemes), while a change in the allophone of the phoneme simply fucks the pronunciation.
Therefore, a phoneme is the minimal distinctive unit in the sound system of the language.

2.3.2 Distributional characteristics of allophones

-2. Not all allophones are in complementary distribution. E.g. in the word clap, the /p/ can be
expressed as either the released (where the air that was building up due to p being a stop is
released) consonant version [p], or the unreleased (where it is not released) consonant version [p ]̚ .
We conclude that allophones need not be in complementary distribution. Free variation, speakers
can choose which allophone to use.

Notational conventions (word as reference category): #word boundary; ‘_’ position of interest; E.g.
word-final context in which we find unreleased [p ]̚ is symbolized as ‘__#’ as, the sound we are
interested in occurs at the end of the word; C stands for consonant and a subscript specifies what
kind. E.g. C[-voice] _ where [-voice] reads as voiceless. Another important word-medial context is the
position between two vowels in word. ‘V__V’ symbolises our intervocalic position.

2.3.3 Observing allophonic alternations in different word forms.

-Another cue to how language categorises sounds into phonemes and allophones can be found in
looking at different forms of one word. E.g. [klæp ]̚ → [klæpiŋ]. The unreleased allophone of English
/p/ can only occurs word-finally. Thus, if we change the word so that it finds itself in a word-medial
context, the allophonic expression also changes to [p] since it is the only possibility.

-Neutralisation refers to the fact that in a certain context, the contrast between different phonemes
can be neutralised. E.g. /d/ expresses itself as [d] in Räder and [t] in Rad. /t/ expresses itself as [t] in
Räte and [t] in Rat. In the case of comparing Rad and Rat, in which /d/ and /t/ are both expressed as
[t], switching out /d/ or /t/ does not change the meaning of the word, thus neutralising our capacity
to distinguish the word-final phoneme. The two words are immune to the minimal pair test, to thus
prove that /d/ and /t/ are different phonemes we must use another word pair. Dank vs Tank ->
[daŋk] vs [taŋk]. There, done.

2.4.1 Allophones of /l/

- lip [lɪp], clip [kl ̥ɪp], miller [mɪlə], pill [pɪɫ]. You can see that just like /ɹ/, /l/ has a voiced and voiceless
allophone, however, what interests us here is the third realisation of /l/, the [ɫ], which is what is
called a velarised realisation, or dark 1. (The non-velarised version is referred to as clear 1). But wait,
you might think, /l/ is an alveolar sound, what does the velum have to do with anything? Look at the
pictures.

- Now let’s talk about the distribution of /l/. [l ̥] is fully parallel to the distribution of [ɹ̥] in that it
occurs after voiceless consonants. We will concentrate on [l] and [ɫ].

- Testing for word-initial/final and in-between vowels we can see that there is no phonetic context in
which both allophones can occur. Thus, we can say that they are in complementary distribution, this
automatically means that they are not in free variation and that they belong to the same phoneme.
Dark [ɫ] is restricted to the word-final position, whereas clear [l] occurs word-initially and between
two vowels.
- Another test, this one with different word forms,

Shows our findings to be correct. [ɫ] can only occur in word-final positions and changes to [l] if moved
away from there.

-Using this data, we can set up a phonological rule predicting allophone realisations of /l/

Namely: /l/ is realised;

as [l ̥] after word-initial voiceless consonants,

as [ɫ] in world-final positions,

as [l] elsewhere

2.4.2 stop phonemes

- Just like /p/ with its two allophones [p] and [p ]̚ in free variation, the same holds true for /t/ and /k/,
with respectively [t] and [t ]̚ , and [k] and [k ]̚ . [hæt] or [hæt ]̚ . [stæk] or [stæk ]̚ .

- The second type of stop allophony to be discussed manifests itself in stops that occur word-initially
before stressed vowels.
- Both realisations of /p/ are voiceless bilabial stops, but in pin something else happens in addition to
the articulation of the stop. In addition to letting the air escape, they press an additional an
additional amount of air through the larynx, where it produces a friction noise when passing through
the glottis. In other words, speakers articulate at the same time a bilabial stop and a glottal fricative
[h]. As a result, the release burst of the plosive is enhanced. The variant of /p/ in pin is called an
aspirated stop. [ʰ]. [pʰ], [tʰ], [kʰ].

- Let us now look at the distribution of aspirated and non-aspirated stop allophones in English.
(Limited to bilabial stop and specific stress configurations)

- We can see here that the allophones [pʰ] and [p] are in complementary distribution in three
contexts: word-initially, after [s], and between vowels. Only in the word-final context are they in free
variation.

-Let us for the last example look at a strictly General American example and try to determine how it
deviates from what we have learned about the allophony of /t/ in this section. I’m not writing this
down, since its basically just an exercise that is meant to showcase the fact that different dialects
also have different phonologies. In AE for example there is a flap in-between two vocals, changing
butter into budder for example. You can read the part again if you ever need, or transcribe it if it
becomes necessary.

- Phonological rule predicting allophonic realisations of /p/ in RP

/p/ is realised

-as [pʰ] in word-initial position before vowels

-as between [s] and a vowel and between two vowels

-as [pʰ] or [p] or [p ]̚ in word-final position

2.4.3 A slightly more complex case: /ɹ/

- This data illustrates one of the most well-known phonological features of RP. Namely, that r-sounds
do not occur in some varieties of English in the word-final position. Such varieties are termed non-
rhotic. Conversely, rhotic varieties of English are those in which r-sounds can occur in word-final
position. E.g. North-American as well as Irish or Scottish varieties. However, the term non-rhotic is in
fact misleading. In RP, the phoneme /ɹ/ is not absent word-finally, it is simply not pronounced.

- A test of changing the word forms is conclusive in showing that the /ɹ/ is indeed present. It emerges
in the word-medial position in all samples. Furthermore, we can see that it is obligatory. But where is
the /ɹ/ in the base forms?
- The words in the image above prove that the allophone of a phoneme can be ‘zero’. In other words,
/ɹ/ is present, and we can prove it with the -ing form. But it does not have a phonetic realisation.

- However, doesn’t this mean that we can’t know when or where a ‘zero’ allophone is present? Must
we always be suspicious? The answer of how to find if it is present lies in the previous graphics. One
can only assume a zero allophone if we see that it alternates with a non-zero allophone.

- Non-rhotic varieties of English have the zero-allophone realisation of /ɹ/ in the word-final position.
Rhotic varieties have a non-zero-allophone realisation instead.

Phonological rule predicting allophonic realisations of /ɹ/ in RP

/ɹ/ is realised:

- as [ɹ̥] after voiceless consonants

- as zero word-finally after long vowels and diphthongs

- as /ɹ/ elsewhere

2.5 The syllable

2.5.1 The structure of the syllable

- The syllable is the intermediate category of sound sequencing that lies between the sound and the
word

- Syllables consist of one vowel, and a consonant/-s, either in the front, in the back, or both. C
(optional)_ V (obligatory) _ C (optional

- little [lɪtl ̩], button [bʌtn̩ ], cupboard [kʌbɹ̩d]. Consonants which occupy the central part of the syllable
are termed syllabic consonants

- The slot for the vowel is termed the nucleus of the syllable. In English this slot can be filled by a
vowel, a diphthong or a syllabic consonant. The prevocalic slot is termed onset and the postvocalic
coda, both can be filled by one or more consonants, furthermore, neither is obligatory.
2.5.2 Syllabification

- Given a sequence of consonants and vowels, syllabification proceeds in such a way that as many
consonants as possible end up in an onset, even if the language allows codas. This generalisation is
termed the Maximal Onset Principle. Al-fred, instead of Alf-red. Su-san instead of Sus-an. Basically,
we want a syllable to start with a consonant(onset) if possible. However, why is it then not A-lfred? Is
‘lfr’ not a possible onset? Why?

- The answer to that is a bit complicated. A syllable is a set of sounds which form a group around its
most sonorous (sonority is a relative term that describes something as being more or less clearly
audible) member. In almost all cases, the nucleus of a syllable is its most sonorous part. E.g. graph.

How does this help us determine the rule which states that a-lfred is not a valid syllabification? Well,
firstly, seeing as the sonority exists on a curve, we can determine that another rule in regards to
syllables is that while onsets and coda are lower in sonority then the nuclei, an onset must rise in
sonority as it approaches the nucleus, and the nucleus must fall in sonority as it approaches the coda.
Sonority sequencing principle.