PHONOLOGY

Phonology I: Features and Phonemes

Distinctiveness
Although there is a continuum of sounds but there is only a very limited set of distinctions that we
look out for. Just like letters: can write them in many different ways most differences do not matter at
all. Similarly, some phonetic contrasts are relevant others are not. The question is: what do we mean
by relevance? The answer is: if the contrast makes a difference in meaning it is relevant. The easiest
test is to find two words that mean different things but differ only in one sound. These are called
minimal pairs. HANDOUT PIC1

spell out a few assumptions:

1. the sound stream is segmentable into unique and identifiable units.
The sound stream of an utterance of /hat/ will thus consist of three sounds, which I write as [h], [æ]
and [t]. Similarly, the sound stream of an utterance of /cat/ consists of three sounds, [k], [æ] and [t].
2. two sequences are of equal length to align the particular sounds with each other
3. ‘exchange’ particular occurrences of sounds in the stream.

Definition 4 Two sound streams form a minimal pair, if their segmentations are of the same
length, and one can be obtained from the other by exchanging just one sound for another, and
that the change results in a change of meaning.

consist of two entire words

for two words to be minimal pair they must be of equal length in terms of how many basic sounds
constitute them, not in terms of how many alphabetic characters are needed to write them.
The same length is important - formal reason – to associate the sounds with each other

Many of the contrasts between the 40 or so basic sounds of English can be demonstrated to be
relevant by just choosing two words that are minimally different in that one has one sound and the
other has the other sound.
- note : in English certain sounds just do not exist. E.g. retroflex consonants, lateral fricatives
- English uses only some of the available sounds, and other languages use others
- the set of English sounds is divided into 40 groups, each corresponding to one letter
- phoneme: groups of realizations of various sounds. /l/ is different in file, slight, wealth, listen –
but it is represented as one sound /l/. So one phoneme contains many ‘sounds
the different sounds of a phoneme do not always occur in the same environment.
of the same phoneme, then σ and σ′ are said to be
in free variation. If however σ and σ′ are not in free variation, we say that the realization of the
phoneme as either σ or σ′ is conditioned by the context.
Table 9 gives a list of the phonemes of American English. The slanted brackets denote phonemes not
sounds. On the whole, the classification of phonemes looks very similar to that of sounds. But there
are mismatches. There is a series of sounds, called affricates, which are written as a combination of a
stop followed by a fricative: English has two such phonemes, [t ] and [d ]. Similarly, diphthongs, which
S Z

are written like sequences of vowels or of vowel and glide, are considered just one phoneme. Notice
also that the broad transcription is also hiding some diphthongs, like [e] as in /able/. This is a
sequence of the vowel [e] and the glide [j]. The reason is that the vowel [e] is obligatorily followed by
[j] and therefore mentioning of [j] is needless.

Some Concerns in Defining a Phoneme

a phone is just a sound, a concrete, linearly indecomposable sound (with the exception of certain
diphthongs and affricates), a phoneme on the other hand is a set of sounds.
•
Definition 5 A phoneme is a set of phones (= speech sounds). In a language L, two sounds a
and b belong to the same phoneme if and only if for all strings of sounds x and y: if both x a y
and x b y belong to L, they have the same meaning. a and b are allophones if and only if they
belong to the same phoneme.

We also say the following. If ~xaaa~y ∈ L then the pair h~x,~yi, which we write ~x ~y is an
environment for a in L. Another word for environment is context.
So, if a and b belong to the same phoneme, then either in a given word (or text) containing a one
cannot substitute b for a, or one can but the result has the same meaning; and in a text containing b
somewhere either one cannot substitute a for b or one can and the result has the same meaning.

Definition 6 (Phoneme) If L is a language, and p a specific sound then /p/L denotes the
phoneme containing p in L.

Features
phonemes are constituted by classes of sounds that have certain properties in common. These are
defined by features. Features are phonetic, and supposed to be not subject to cross-language
variation. take a (more or less) articulatory standpoint.
Take any sound realizing English /b/. It is produced by closing the lips, thereby obstructing the air
flow (‘bilabial’) and then releasing it, and at the same time letting the vocal cords vibrate (‘voiced’). If
the vocal cords do not vibrate we get the sound corresponding to /p/. We can analyse the sound as a
motor program that is executed on demand. Its execution is not totally fixed, so variation is possible
(as it occurs with all kinds of movements that we perform). Second, the motor program directs various
parts of the vocal tracts, some of which are independent from each other. The value ‘+’ tells us that
the cords have to vibrate during the production of the corresponding sound, while ‘− ’ tells us that they
should not.
while phonetically we are dealing with a continuous scale (onset time measured in milliseconds), at
the phonemic level we are just looking at a binary contrast.
following notation. There is a set of so-called attributes and a set of so-called values. A pair [ATT :
val] consisting of an attribute and a value is called a feature. We treat +voiced as a notational
alternative of [VOICED : +]. An attribute is associated with a value range. For phonology, we may
assume the following set of attributes:
PLACE, MANNER, VOICED, CONSONANTAL, ASPIRATED, APERTURE
and we may assume the following set of values:
bilabial, labiodental, plosive, approximant, high,mid, +, − , . . .
The range of PLACE is obviously different from that of MANNER, since ‘dental’ is a value of the
former and not of the latter. A set of features is called an attribute value structure (AVS). You have
seen AVSs already in the first lecture. The notation is as follows. The attributes and values are
arranged vertically, the rows just having the attribute paired with its value, separated by a colon:

vowels in English are voiced. Not in all. think about whispering. Whispering is speaking without the
vocal chords vibrating. In effect, whispering is systematically devoicing every sound. That this does
not remove the distinction between [p] and [b] shows you that the distinction is not exclusively a
voicing contrast! One additional difference is that the lip tension is higher in [p].

arguments to show that features exist:

- features encode a certain linguistic reality
- the features that we have spoken about so far have phonetic content
- articulatory properties
- many rules can be motivated from the fact that the vocal tract has certain properties

This says that a consonant becomes devoiced at the end of a word. The part before the arrow
specifies the situation before the rule applies; the part to the right and before the slash show us how it
looks after the application of the rule. The part after the slash shows in what context the rule may be
applied. The underscore shows where the left part of the rule must be situated (and where the right
part will be substituted in its place). Here, it says: it must occur right before #, which signals the end of
a word.
Features have an independent motivation the correspondence is specified uniformly for all sounds
e.g. (‘voice’ refers to the fact whether or not the vocal cords vibrate. some rules are not really specific
to one language but a whole group of them (final devoicing is a case in point.
For example, Hungarian, Turkish and Finnish both have a rule called vowel harmony. Modulo some
difficulties all rules agree that words cannot both contain a back vowel and a front vowel.
Systematically oppose /o/ with /o/, which differs in the position of the tongue body (close-mid back
rounded vowel). The situation is complicated through the fact that Hungarian long and short vowels
do not only contrast in length but also in a feature that is called tension. /ö/ in Hungarian it is lax
(which means less rounded and less close). - the contrast is between ‘front’ and ‘back’, no matter
what else is involved. Thus, among the many parameters that define the actual sounds languages
decide to systematically encode only a limited set (which is phonologically relevant and on which the
rules operate) even though one still needs to fill in details as for the exact nature of the sounds.
Precisely this is the task of realization rules. These are the rules that make the transition from
phonemes to sounds.

Natural Classes
Definition 7 (Natural Class; Provisional) A natural class of sounds is a set of sounds that can be
specified by a single AVS.
Definition 8 (Natural Class) Let S be a classification system for P. A subset U of P is natural in S if
and only if it is an intersection of sets of the form [A : v] for some attribute and some legitimate value.

The Classification System of English Consonants

three attributes: PLACE, MANNER, and voice.