Distinctive features

1. Ferdinand de Saussure (1916)

Saussure made distinctions between signified and signifier and between form and substance. For example, in his theory the word "dog" (its external spoken and written versions and mental representations of the word's morphology, spelling and pronunciation) is a signifier that relates to our concept of a dog (not to an actual physical dog, but to our idea of what a dog is). In other words, the word "dog" is a signifier and its signified is our concept of what a dog is. The details of this aspect of his theory are better dealt with in a course on semantics (ie. generally about meaning in language) or semiotics (ie. about meaning but from the perspective of the concepts signified and signifier). What interests us here are his ideas about form and substance.

• Form: an abstract formal set of relations.
• Substance: sounds (phonemes) or written symbols (graphemes).
• Word: the union of signified and signifier.

Phonetic segments (speech sounds) are elements that have no meaning in themselves. They have, however, non-semantic and non-grammatical rules of combination etc.

Meaningless elements (phonetic segments) can combine to form meaningful entities. ie. words, which are combinations of phonetic segments (or of phonemes) are meaningful.

There is an arbitrary relationship between the Meaningless and the Meaningful. For example, the relationships between the meaningless elements (phonetic segments or written symbols) and meaningful combinations of those elements (words) are arbitrary.

Each language has its own set of distinct rules for the combination of sounds, or phonotactic rules. The need to find ways of expressing such rules in a simple manner motivated the

2. Nikolai Trubetzkoy (1939)

Nikolai Trubetzkoy was a core member of the Prague school of linguistics which was highly influential in developing some areas of linguistic theory (including phonology), particularly in the 1930s. His most influential work was published posthumously in 1939, shortly after his death. One aspect of Trubetzkoy's work examines the idea of different types of "oppositions" in phonology. These oppositions are based on phonetic (or phonological) features. We won't look at all of the types of oppositions that he described, but only a few that are of particular relevance to this topic. That is, we will examine types of opposition that are relevant to the definition of phonological features.

a) Bilateral oppositions

A bilateral opposition refers to a pair sounds that share a set of features which no other sound shares fully. For example, voiceless labial obstruents = /p,f/. Note that obstruents are defined as having a degree of stricture greater than that of approximants (that is, stops and fricatives).

b) Multilateral oppositions

A group of more than 2 sounds which share common features. For example, labial obstruents, /p,b,f,v/, are both labial and obstruents, so they share two features.

c) Privative (binary) oppositions

One member of a pair of sounds possesses a mark, or feature, which the other lacks. Such features are also known as binary features which a sound either possesses or lacks. Voicing is such a feature. A sound is voiced or NOT voiced. The sound which possesses that feature is said to be marked (eg [+voice]) whilst the sound lacking the feature is unmarked (eg. [-voice]).

d) Gradual oppositions

The members of a class of sounds possess different degrees or gradations of a feature or property. For example, the three short front unrounded vowels in English /ɪ, e, æ/ which are distinguished only by their height. In this system height would be a single feature with two or more degrees of height.

e) Equipollent oppositions

The relationship between two members of an opposition are considered to be logically equivalent. Consonant place of articulation can be seen in this sense. Changes in place involve not just degree of fronting but also involve other articulator changes.

3. Roman Jakobson et al. (1941-1956)

Roman Jakobson was also a member of the Prague school of linguistics and worked closely with Trubetzkoy. Distinctive feature theory, based on his own work and the work of Trubetzkoy, was first formalised by Roman Jakobson in 1941 and remains one of the most significant contributions to phonology. Briefly, Jakobson's original formulation of distinctive feature theory was based on the following ideas:-

1. All features are privative (ie. binary). This means that a phoneme either has the feature eg. [+VOICE] or it doesn't have the feature eg. [-VOICE]
2. There is a difference between PHONETIC and PHONOLOGICAL FEATURES
   • Distinctive Features are Phonological Features.
   • Phonetics Features are surface realisations of underlying Phonological Features.
   • A phonological feature may be realised by more than one phonetic feature, eg. [flat] is realised by labialisation, velarisation and pharyngealisation
3. A small set of features is able to differentiate between the phonemes of any single language
4. Distinctive features may be defined in terms of articulatory or acoustic features, but Jakobson's features are primarily based on acoustic descriptions

Jakobson and colleagues (Jakobson, Fant and Halle, 1952, Jakobson and Halle, 1956) devised a set of 13 distinctive features. They are not reproduced here, but they represent a starting point for the set defined by Chomsky and Halle.

4. Chomsky and Halle (1968-1983)

Distinctive feature theory was first formalised by Roman Jakobson in 1941. There have been numerous refinements to Jakobson's (1941) set of features, most notably with the development of Generative Phonology and the publication of Chomsky & Halle's (1968) Sound Pattern of English but also from phoneticians such as Ladefoged (1971), Fant (1973) and also Stevens (Halle & Stevens, 1971).

In recent years, there have also been proposals that features should themselves be hierarchically structured and arranged on separate levels or tiers (e.g. Clements & Keyser, 1983) which is consistent with the developments in autosegmental phonology since the publication of the Sound Pattern of English.

Regardless of the many differences and controversies, the various kinds of feature systems share the following characteristics.

a) Features establish natural classes

Using distinctive features, phonemes are broken down into smaller components. For example, a nasal phoneme /m/ might be represented as a feature matrix [+ sonorant, -continuant, +voice, +nasal, +labial] (nb. you will often see the features of the matrices arranged in columns in phonology books -- this is exactly equivalent to the 'horizontal' representation used here). By representing /m/ in this way, we are saying both something about its phonetic characteristics (it's a sonorant because, like vowels, its acoustic waveform has low frequency periodic energy; it's a non-continuant because the airflow is totally interrupted in the oral cavity etc.), but also importantly, the aim is to choose distinctive features that establish natural classes of phonemes. For example, since all the other nasal consonants and nasalised vowels (if a language has them) have feature matrices that are defined as [+nasal], we can refer to all these segments in a single simple phonological by making the rule apply to [+nasal] segments. Similarly, if we want our rules to refer to all the approximants and high vowels, we might define this natural class by [+sonorant, +high].

The advantage of this approach is readily apparent in writing phonological rules. For example, we might want a rule which makes approximants voiceless when they follow aspirated stops in English. If we could not define phonemes in terms of distinctive features, we would have to have separate rules, such as [l] becomes voiceless after /k/ ('claim'), /r/ becomes voiceless after /k/ ('cry'), /w/ becomes voiceless after /k/ ('quite'), /j/ becomes voiceless after /k/ ('cute'), and then the same again for all the approximants that can follow /p/ and /t/. If we define phonemes as bundles of features, we can state the rule more succinctly as e.g. [+sonorant, -syllabic, +continuant] sounds (i.e. all approximants) become [+spread glottis] (aspirated) after sounds which are [+spread glottis] (aspirated).

If the features are well chosen, it should be possible to refer to natural classes of phonemes with a small number of features. For example, [p t k] form a natural class of voiceless stops in most languages: we can often refer to these and no others with just two features, [-continuant, -voiced]. On the other hand, [m] and [d] are a much less natural class (ie. few sound changes and few, if any, phonological rules, apply to them both and appropriately it is impossible in most feature systems to refer to these sounds and no others in a single feature matrix).

b) Economy

In phonology, and particularly in Generative Phonology, we are often concerned to eliminate redundancy from the sound pattern of a language or to explain it by rule. Distinctive features allow the possibility of writing rules using a considerably smaller number of units than the phonemes of a language. Consider for example, a hypothetical language that has 12 consonant and 3 vowel phonemes:

We could refer to all these phonemes with perhaps just 6 distinctive features - a reduction of over half the number of phoneme units which also allows natural classes to be established amongst them:

At the same time, each phoneme is uniquely represented, as shown by the distinctive feature matrix:

c) Binarity

We have assumed that features are binary (a segment is either nasal or it is not) following Jakobson's (1941) original formulation of distinctive feature theory and this premise was adopted in Chomsky & Halle's (1968) Sound Pattern of English. There were many reasons why Jakobson (1941) advocated a binary approach. Firstly, as we have seen, this is the most efficient way of reducing the phoneme inventory of a language. Secondly, he argued that most phonological oppositions are binary in nature (e.g. sounds either are or are not produced with a lowered soft-palate and nasalisation) and he even proposed that it has a physiological basis i.e. that nerve fibers have an 'all-or-none' response. But the binary principal is certainly not adopted by all linguists, and many phoneticians in particular have argued that some features should be n-ary (where "n" is any relevant number of degrees or levels - see for example, Ladefoged's 1993 treatment of vowel height which is 4-valued to reflect the distinction between close, half-close, half-open, and open vowels).

d) Phonetic interpretation

According to Jakobson (1941), the distinctive features should have definable articulatory and acoustic correlates. For example, [+nasal] implies a lowering of the soft-palate and also an increase in the ratio of energy in the low to the high part of the spectrum. Chomsky & Halle (1968) abandoned the acoustic definitions of phonological features (inappropriately, as Ladefoged, 1971 and many others have argued: for example [f] and [x] are related acoustically but not articulatorily and they participated in the sound change by which the pronunciation of 'gh' spellings in English changed from a velar to a labiodental fricative e.g. 'laugh', [lɑx]→[lɑf]).

Many of the features are defined loosely in phonetic terms. This is perhaps to be expected. Phonology has established highly abstract representations to explain sound alternations (i.e. to factor out what are considered redundant or predictable aspects of a word's pronunciation) and this abstraction is partly opposed to the principle in phonetics of describing in articulatory and acoustic terms the characteristics of speech sound production that are shared by linguistic communities. Nevertheless, if phonology is to be related to how words are actually pronounced, the features are required to have at least some phonetic basis to them.

e) An overview of commonly used distinctive features

The features described in Halle & Clements (1983) have been commonly used in the phonology literature in their analyses of the sound patterns of various languages. They incorporate many insights of the original features devised by Jakobson (1941) but are mostly based on those of the Sound Pattern of English, taking into account some modifications suggested by Halle & Stevens (1971). Most of these are also discussed below.

i. Major class features

Four features [syll], [cons], [son], [cont] (syllabic, consonantal, sonorant, continuant) are used to divide up speech sounds into major classes, as follows. Note that [syll] means "syllabic" (syllable nucleus), [cons] means "consonantal", [son] means "sonorant" (periodic low frequency energy), [cont] means "continuant" (continuous airflow through oral cavity), and [delrel] means delayed release (release is not "delayed", but there is a longer aspiration phase than oral stops - nb. voice onset is what's actually delayed).

Note that the approximants have been divided into liquids (eg. in English /r, l/) and semi-vowels (eg. in English /w, j/). In this, and most other distinctive feature sets derived from Chomsky and Halle. Semi-vowels (being [-syll, -cons]) form a class of sounds intermediate between vowels ([+syll]) and consonants ([+cons]). The approximants can be defined as a class by the features [-syll, +son, +cont] and can be further sub-divided into liquids and semi-vowels using the [cons] feature. Note that "0" means irrelevant feature for these classes of sounds (there's nothing to release).

We also have a feature [nasal] which, as its name suggests, separates nasal from oral sounds. In the above table, [nasal] would have been redundant as the nasal stops are already defined uniquely as [-syll, +cons, +son, -cont] (ie. as sonorant stops). However. the feature [nasal] is required to define nasal stops, nasalised vowels and nasalised approximants as a single natural class.

ii. Source features

These are related to the source (vocal fold vibration that sustains voiced sounds or a turbulent airstream that sustains many voiceless sounds).

The feature [voice] is self-explanatory (with or without vocal fold vibration). The feature [spread glottis] is used to distinguish aspirated from unaspirated stops (aspirated stops are initially produced with the vocal folds drawn apart). We can therefore make the following distinctions:

The [strident] feature is used by Halle and Clements for those fricatives produced with high-intensity fricative noise: supposedly labiodentals, alveolars, palato-alveolars, and uvulars are [+strident]. There seems to be little acoustic phonetic basis to the claim that labiodentals and alveolars pattern acoustically (as opposed to dentals). In this course, we will use Ladefoged's feature [sibilant] which is defined by Ladefoged (1971) in acoustic terms as including those fricatives with 'large amounts of acoustic energy at high frequencies' i.e. [s ʃ z ʒ]. The English affricates would therefore also be [+sibilant]:

This may be an oversimplification, however, as the alveolar oral stops might also be described as sibilant, so sibilant isn't sufficient to separate oral stops and affricates (and we still need spread glottis).

iii. Vowel features

There are four dimensions to consider: vowel height, backness, rounding, and tensity.

Vowel height is classified using the [high] and [low] features. Palatal and velar consonants are also [+high, -low] and, with the use of the [back] feature, a relationship can be established between high vowels and their corresponding glides:

(NB: [u] is presumed to be a back vowel here as in cardinal vowel 8).

In different languages vowel systems can vary in terms two to four vowel height contrasts.

Vowel systems with only two levels of height simply use [+high] or [-high] to represent high and low vowels respectively.

Vowels systems with three levels of height contrast use the features [high] and [low]. High vowels like [i u y ɯ] are [+high, -low]. Mid vowels like [e o] are neither high, nor low, i.e. [-high, -low]. And open vowels like [a ɑ] are [-high, +low].

Vowel systems with four levels of height contrast require a third feature [mid]. (This feature is additional to the set of features defined in Halle & Clements, but will be used in this course).

The question of vowel tensity is more controversial. In English there is a class of vowels which are more central in the vowel space (i.e. further away from the four vowel space corners) and that cannot end monosyllabic words: in Australian English, these include the short vowels [æ e ɪ ʊ ɔ ɐ] in 'had', 'head', 'hid', 'hood', 'hod', 'hud'. While that is perhaps not so controversial, the issue of whether they can be defined by an articulatory [-ATR] (minus advance tongue root), indicating that the tongue root is not drawn forward to the same degree as [+ATR] vowels (all the 'long' vowels that can end words in English) is a good deal more problematic. It definitely doesn't make sense to apply [+ATR] to lax back vowels. As Halle and Clements note, [ATR] appears to "... never co-occur distinctly in any language ..." with another feature pair "tense/lax" [+- tense]. Whilst the tense/lax distinction is also defined by Halle & Clements in articulatory terms similar to those used for [ATR], the use of the term "tense" versus "lax" to describe respectively the "long" and "short" vowels of English is very familiar to phoneticians. The use of an alternative feature pair "long/short" is avoided because some languages have a distinction between tense and lax vowels which do not appear to be realised acoustically in terms vowel duration (ie. long versus short). In this course the feature [tense] will be used rather than the feature [ATR].

iv. Place of articulation

The [ant] 'anterior' and [cor] 'coronal' features, in combination with [high] and [back] (see 3. above) and [sibilant] (see 2. above) do most of the job of consonantal place classification. These features will be defined in the next section. For example, for fricatives:

([ʍ] is a labial-velar fricative: the fricative equivalent of [w]).

Note that, by having abandoned [strident] (or replaced it with [sibilant]), we leave ourselves the problem of how to differentiate [ɸ] from [f]. Halle & Clements also define a feature [distr] (distributed) that they say can be used, amongst other things, to distinguish bilabial [+distr] and labiodental [-distr] sounds (nb. [θ] is [+distr] if it is lamino-dental, or [-distr] if it is apico-dental). [+distr] sounds have a greater area of contact than similar [-distr] sounds. For example, [+distr] bilabials have two lips in contact so there is a greater area of articulator contact than for [-distr] labiodentals (as the lower lip is in contact with the smaller area of the tips of the upper teeth). Also, apicals use the smaller area of the tongue tip whilst laminals use the greater area of the tongue blade.

5. Distinctive features used in this topic

The following set of distinctive features follows the set defined by Halle and Clements (1983), but with the following exceptions:-

• The feature [ATR] (advanced tongue root) has been omitted, in favour of [tense].
• The feature [strident] has been replaced by Ladefoged's feature [sibilant].
• The feature [rounded] has been omitted as it seems to be mostly redundant given the presence of the feature [labial].
• The feature [mid] has been added to deal with vowel systems with four contrastive levels of height.
• The feature [-cont] does not automatically include all laterals. In this course laterals are [+cont] if approximants or fricatives and [-cont] if lateral clicks or laterally released stops.
• The feature [front] has been added. It is used here exclusively as a vowel feature and is used for languages or dialects, such as Australian English, which exhibit three levels of vowel fronting. The original set only had [back] as American English vowels could be described as [+back] or [-back] (for front) and the central vowel in the word "heard" was disambiguated by a [+rhotic] feature.

When quoted text occurs in the following descriptions, it is taken from Halle and Clements (1983, pp 6-8) and this is indicated by (HC) following the quote.

1. Syllabic / non-syllabic [syll]: Syllabic sounds constitute a syllable peak (sonority peak). [+syll] refers to vowels and to syllabic consonants. [-syll] refers to all non-syllabic consonants (including semi-vowels).
2. Consonantal / non-consonantal [cons]: Consonantal sounds are produced with at least approximant stricture. That is consonantal sounds involve vocal tract constriction significantly greater that that which occurs for vowels. [+cons] refers to all consonants except for semi-vowels (which often have resonant stricture). [-cons] refers to vowels and semi-vowels.
3. Sonorant / obstruent [son]: Sonorant sounds are produced with vocal tract configuration that permits air pressure on both sides of any constriction to be approximately equal to the air pressure outside the mouth. Obstruents possess constriction (stricture) that is sufficient to result in significantly greater air pressure behind the constriction than occurs in front of the constriction and outside the mouth. [+son] refers to vowels and approximants (glides and semi-vowels). [-son] refers to stops, fricatives and affricates.
4. Coronal / non-coronal [cor]: "Coronal sounds are produced by raising the tongue blade toward the teeth or the hard palate; noncoronal sounds are produced without such a gesture." (HC) This feature is intended for use with consonants only. [+cor] refers to dentals (not including labio-dentals) alveolars, post-alveolars, palato-alveolars, palatals. [-cor] refers to labials, velars, uvulars, pharyngeals.
5. Anterior / posterior [ant]: "Anterior sounds are produced with a primary constriction at or in front of the alveolar ridge. Posterior sounds are produced with a primary constriction behind the alveolar ridge."(HC) This feature is intended to be applied to consonants. [+ant] refers to labials, dentals and alveolars. [-ant] refers to post-alveolars, palato-alveolars, retroflex, palatals, velars, uvulars, pharyngeals.
6. Labial / non-labial [lab]: Labial sounds involve rounding or constriction at the lips. [+lab] refers to labial and labialised consonants and to rounded vowels. [-lab] refers to all other sounds.
7. Distributed / non-distributed [distr]: "Distributed sounds are produced with a constriction that extends for a considerable distance along the midsaggital axis of the oral tract; nondistributed sounds are produced with a constriction that extends for only a short distance in this direction." (HC) [+distr] refers to sounds produced with the blade or front of the tongue, or bilabial sounds. [-distr] refers to sounds produced with the tip of the tongue. This feature can distinguish between palatal and retroflex sounds, between bilabial and labiodental sounds, between lamino-dental and apico-dental sounds.
8. High / non-high [high]: "High sounds are produced by raising the body of the tongue toward the palate; nonhigh sounds are produced without such a gesture." (HC) [+high] refers to palatals, velars, palatalised consonants, velarised consonants, high vowels, semi-vowels. [-high] refers to all other sounds. Note, however, the discussion above on how this feature is used in combination with [mid] to describe the distinction between four contrastive vowel heights.
9. Mid / non-mid [mid]: Mid sounds are produced with tongue height approximately half way between the tongue positions appropriate for [+high] and [+low]. This vowel height feature is only required when a language has four levels of height contrast and remains unspecified for languages with fewer vowel height contrasts. [+mid] refers to vowels with intermediate vowel height. [-mid] refers to all other sounds.
10. Low / non-low [low]: "Low sounds are produced by drawing the body of the tongue down away from the roof of the mouth; nonlow sounds are produced without such a gesture." [+low] refers to low vowels, pharyngeal consonants, pharyngealised consonants.
11. Back / non-back [back]: "Back sounds are produced with the tongue body relatively retracted; nonback or front sounds are produced with the tongue body relatively advanced." (HC) [+back] refers to Velars, uvulars, pharyngeals, velarised consonants, pharyngealised consonants, central vowels, central semi-vowels, back vowels, back semi-vowels. [-back] refers to all other sounds.
12. Front / non-front [front]: This is an additional vowel feature added to assist in the description of the vowel systems of languages such as Australian English. To describe the central vowels of Australian English its necessary to define them as [-back, -front].
13. Continuant / stop [cont]: "Continuants are formed with a vocal tract configuration allowing the airstream to flow through the midsaggital region of the oral tract: stops are produced with a sustained occlusion in this region." (HC) For some reason it has been traditional to include lateral consonants as stops in distinctive feature theory. Since laterals can have approximant, fricative or stop (click) stricture there seems to be no justification in including all laterals with the stops, and in this course laterals are not necessarily stops (as is the case for the lateral clicks) but can also be continuants (as is the case for the lateral approximants and fricatives. [+cont] refers to vowels, approximants, fricatives. [-cont] refers to nasal stops, oral stops.
14. Lateral / central [lat]: "Lateral sounds, the most familiar of which is [l], are produced with the tongue placed in such a way as to prevent the airstream from flowing outward through the centre of the mouth, while allowing it to pass over one or both sides of the tongue; central sounds do not invoke such a constriction." (HC) [+lat] refers to lateral approximants, lateral fricatives, lateral clicks. [-lat] refers to all other sounds.
15. Nasal / oral [nas]: "Nasal sounds are produced by lowering the velum and allowing the air to pass outward through the nose; oral sounds are produced with the velum raised to prevent the passage of air through the nose." (HC) [+nas] refers to nasal stops, nasalised consonants, nasalised vowels. [-nas] refers to all other sounds.
16. Tense / lax [tense]: The traditional definition of this feature claims that [+tense] vowels involve a greater degree of constriction then [-tense] (lax) vowels. Tense vowels need not be any different to lax vowels in terms of constriction (e.g. the tense/lax pair /ɐː,ɐ/ in Australian English are produced with the same tongue position but differ in duration). The tense/lax distinction in vowels seems to be related to some kind of strong/weak distinction. In some languages this is realised as a distinction between more peripheral vowels (closer to the four corners of the vowel quadrilateral) and less peripheral vowels (vowels that are either more centred, more mid, or both more centred and more mid). In other languages, a long/short durational distinction is what is often the main acoustic distinction between tense and lax vowels. Note, however, that short vowels are more likely to be produced with under-realised targets (more mid-central) during connected speech than are long vowels because the long vowels have more time to reach their targets. [+tense] refers to tense vowels or long vowels. [-tense] refers to lax vowels or short vowels.
17. Sibilant / non-sibilant [sib]: Sibilants are those fricatives with large amounts of acoustic energy at high frequencies. [+sib] refers to [s ʃ z ʒ]. [-sib] refers to all other sounds.
18. Spread glottis / non-spread glottis [spread]: "Spread or aspirated sounds are produced with the vocal cords drawn apart producing a nonperiodic (noise) component in the acoustic signal; nonspread or unaspirated sounds are produced without this gesture." (HC) [+spread] refers to aspirated consonants, breathy voiced or murmured consonants, voiceless vowels, voiceless approximants. [-spread] refers to all other sounds. It should be stressed that during the occlusion of both voiceless aspirated and voiceless unaspirated (0 VOT) stops the glottis is open. The difference is during the period following release where, for aspirated stops, the glottis stays open much longer than for unaspirated stops.
19. Constricted glottis / non-constricted glottis [constr]: "Constricted or glottalized sounds are produced with the vocal cords drawn together, preventing normal vocal cord vibration; nonconstricted (nonglottalized) sounds are produced without such a gesture." (HC) [+constr] refers to ejectives, implosives, glottalized or laryngealised consonants, glottalized or laryngealised vowels. [-constr] refers to all other sounds.
20. Voiced / voiceless [voice]: "Voiced sounds are produced with a laryngeal configuration permitting periodic vibration of the vocal cords; voiceless sounds lack such periodic vibration." (HC) [+voice] refers to all voiced sounds. [-voice] refers to all voiceless sounds.

6. Australian English vowel features

a) Australian English monophthong vowels

When Chomsky and Halle were using their system of Distinctive Features to analyse American English vowels, they argued that it was necessary to define three levels of height:-

But only two levels of fronting:-

Whilst its not impossible to describe Australian English vowels in terms of [high, low, back, round, tense] so that all monophthongs are provided with a unique featural specification the resulting system doesn't match the phonetics of Australian English vowels. In each category, front, central and back there are several vowels. To call both the central and the back vowels [+back] doesn't capture this pattern satisfactorily. For example, the vowel /ʉː/ has been moving forward over several decades and for most speakers its a high, central, rounded vowel. For some speakers it has moved even further forward (half way between central and front), but isn't yet a front rounded vowel. It makes a lot of sense to regard this as a high rounded vowel that isn't explicitly front or back (i.e. [-front, -back]).

So now we can define three degree of vowel fronting:-

As stated (and justified) in the previous section, we don't use a length feature (e.g. [long]) to indicate the distinction between long and short monophthongs. We can instead refer to long vowels (and diphthongs) as tense and short vowels as lax. Instead we use the [tense] feature to distinguish between tense [+tense] and lax [-tense].

Whilst we can sometimes determine that schwa is in some sense a reduced form of some other vowel (e.g. when vowel quality changes when we add or subtract affixes, such as the first syllable of "phonetics" versus "phonetician") for many words its now impossible to determine what the original vowel might have been (or even if there ever had been an original unreduced vowel). Because of this it seems sensible to treat schwa as a phoneme in English. We treat schwa as a vowel phoneme that has the feature [+syllabic], but is unspecified ("0") for [tense]. All other features are context dependent (ie. optional) so its probably best to also treat these features as unspecified ("0") as well. Whilst schwa could be said to be the least tense (most lax) of all vowels it shares an important property with the [+tense] vowels. It can, in English, occur in open syllables (syllables that end with a vowel). The only other vowels that can occur in open syllables in English are the [+tense] vowels (long monophthong vowels and diphthongs). Traditionally this has been explained as being due to schwa, in these contexts, being the reduced form of a tense vowel (and therefore "underlyingly" tense) but as we have described schwa as an independent phoneme this explanation no longer works. None of the other short vowels can occur in open syllables in normal words (except perhaps in interjections). If we make [tense] unspecified for schwa then we can devise a single rule that determines which vowels can occur in open syllables, i.e. "[-tense] vowels cannot occur in open syllables but all other vowels can". There are other precedents for "0" specifications for a feature. For example the delayed release [delrel] feature can only be applied to phonemes that have a release (i.e. oral stops and affricates). All other phonemes have a "0" specification for this feature. Some specifically vowel features have a "0" specification for consonants and vice versa.

Feature Table for Australian English Monophthongs (all are also [+syll, -cons +son +cont])

b) Centring diphthongs

Mitchell (1946) lists three centring diphthongs for Australian English: /ɪə/,/ɛə/ and /ʊə/.

There was also a possible 4th centring diphthong /oə/ which appears to have been described for Australian English by McBurney (1887), although we have no way of knowing whether words like "poor" [poə] were pronounced as one or two syllables and therefore as a diphthong or as two monophthongs. This fourth centring diphthong is occasionally described for some varieties of British English (eg. Wells, 1982). Further, Bernard (1970, 1986) found that 22 out of 112 /oː/ vowels produced by Australian speakers in an /h_d/ context had an "inglide" (ie. [oə]). Bernard considered this inglide to be no more than an idiosyncratic or allophonic variant of the phoneme /oː/ and no phonetician argues for the existence of /oə/ as a phoneme in modern Australian English.

The continuing existence of /ʊə/ as a phoneme in Australian English is doubtful. It was confidently described by Mitchell (1946) and most phoneticians up until the 1960's to be a phoneme of Australian English realised as a centring diphthong [ʊə]. Its status as a phoneme in Australian English has become quite unclear in recent years. If it still exists, it is a very low frequency phoneme. There is a lot of idiosyncratic variation in the survival of diphthongal realisations of this phoneme (those that still possess an audible or measurable in-glide). Even words such as "tour", which have often been cited as late survivors of words containing /ʊə/ are nearly always now found to contain either [oː] (a monophthong) or [ʉːə] (two monophthongs, ie. two syllables).

Of the remaining two centring diphthongs, only /ɪə/ (a) is common and (b) is produced as an in-gliding diphthong by a majority of Australian English speakers (but a fairly large minority are now pronouncing it as [ɪː]).

The phoneme previously labeled /ɛə/ or /eə/ is now produced as a long mid-high monophthong [eː] by the majority of Australian English speakers in CVC contexts (e.g. "paired") and has been transcribed phonemically by most Australian English phoneticians as /eː/ since the early 1990's. It should be noted, however, that there is evidence that the same people who pronounce it as [eː] in a CVC context pronounce it as [eə] in CV contexts (e.g. "hair").

Why is this discussion of centring diphthongs of relevance to the current topic? The answer lies in the extent to which the centring diphthongs can be accounted for by the features used in tables 1 and 2 if they undergo the process of monophthongisation.

The inclusion of /eː/ in the list of Aus.E. monophthongs causes no problem for this system of distinctive features. /eː/ is simply the tense equivalent of the lax vowel /e/ and the tenseness of /eː/ is realised phonetically as either a long vowel or as an offglided diphthong (varying with speaker and phonetic context). If this system of distinctive features has any cognitive validity, it might help explain the reason why /eː/ has evolved from the earlier /eə/ vowel described by Mitchell in the 1940's whilst /ɪə/ has been resistant to such a change. It might also explain why /ʊə/ and a possible earlier diphthong /oə/ have disappeared (either by merging with an existing tense monophthong, or by splitting into a sequence of a tense monophthong plus schwa). In table 2 there was a vacant slot for /eː/ and, if this distinctive feature model is correct, there was therefore nothing to prevent the diphthong from surviving the process of monophthongisation by becoming a distinct tense vowel. All other centring diphthongs lack such an empty slot to fit into.

So what is happening to /ɪə/ and how should we view its distinctive features?

The phoneme /ɪə/ is realised in one of three major forms:-

1. [ɪə] This is the most common form used by the majority of Australian English speakers, especially when preceding a pause (eg. "ear" uttered by itself or at the end of a sentence). Many of these speakers produce [ɪː] when there is an intrusive /r/ (that is, an [ɹ] inserted between a certain vowels, such as /ɪə/, and a following vowel) or a following alveolar consonant. This pattern of pronunciation appears to be a feature of General and Cultivated Australian English.
2. [ɪː] A growing minority of Australian speakers produce this form even when preceding a pause. As mentioned above, many speakers of Australian English (who normally produce [ɪə] before a pause) also produce this form when there is a following intrusive /r/ or a following alveolar consonant.
3. [iːə] This seems to be a feature of the speech of some speakers of Broad Australian English but this is probably best analysed as two vowel phonemes.

It should be noted that the vowel quality for the [ɪː] allophone is usually that of the lax vowel /ɪ/ rather than that of the tense vowel /iː/ but it should also be noted that the targets of /iː/ and /ɪ/ are becoming more alike (Cox, 1996 - see the page on Australian English Monophthongs).

The vowel /iː/ maintains its distinctiveness due to its onglide and length. Acoustically, and auditorily, the targets of /iː/ and the [ɪː] allophone of /ɪə/ are likely to be distinct for older Australians, but not distinct for younger Australians, but from the point of view of the above system of distinctive features they would be identical. That is they would both be [+high, -low, -back, -round, +tense].

Is this therefore evidence against the cognitive validity of this particular distinctive feature model? Well, possibly! If it can be determined empirically that /ɪə/ is more resistant to this kind of change than /eː/ was, then this might be because there is no available slot for it as was the case for /eː/. But, why should it not simply disappear as is almost the case for /ʊə/? The answer to this question may be lexical rather than phonological or phonetic. There are far more words with this phoneme than there ever was with the phoneme /ʊə/. Phoneme merger would cause numerous word pairs to become phonetically identical. This argument is weakened, however, when we realise that /ɪə/ and /eə/ merged into one phoneme in New Zealand over the past 30 years (Watson et al, 2000). This merger caused many pairs of words to become homophones (e.g. "cheer" and "chair"). The loss of distinctiveness of these word pairs did not prevent the change from occurring.

There may be another explanation for the pattern of changes occurring for /ɪə/. The [ɪː] pronunciation may be more common for those speakers of Australian English who produce the most salient on-glide when pronouncing /iː/. These are particularly speakers of Broad Australian English who produce a strong on-glide (ie. /iː/ = [^əiː]). In these speakers, it could be said that /ɪə/ is becoming a high front unrounded tense monophthong as /iː/ is becoming more like a diphthong. This, however, overlooks the well-established fact that the vast majority of Australian English speakers produce /i:/ with an onglide (regardless of whether they are broad, general or cultivated speakers). This onglide varies from a long onglide ([^əiː]) of broad speakers, through the shorter, but acoustically similar, onglide of general speakers to the short onglide (something like [^ɪiː]) of cultivated speakers.

For those Australian English speakers who move between [ɪə] and [ɪːɹ] sequences in different phonetic contexts, the phoneme /ɪə/ may (according to transformational theory) be described to be "underlyingly" a sequence of the phonemes /ɪ/ + /r/ (or perhaps /iː/ + /r/). If this is so, then the process of monophthongisation of the centring diphthongs may be the result of a more fundamental process whereby the underlying /r/ component of the sequence is deleted. Even in American English, which is a rhotic dialect of English, the surface representations of the underlying post-vocalic /r/ phonemes varies greatly. In American English /ə˞/ the following /r/ is most often entirely represented by the surface rhoticisation of the vowel. In the Eastern New England variety of American English the post-vocalic /r/ in many words is not as rhoticised as is the case of most American English varieties, but the /r/ still exists as an alveolar, but not retroflexed, approximant (Olive et al, 1993, p367).

The next table represents a possible feature representation for the three high front vowels of Australian English speakers. The feature "onglide" is proposed here as a feature that accounts for the virtually universal existence of an on-glide for /i:/ in Australian English.

Feature table for Australian English /iː/,/ɪ/ and /ɪə/

c) Feature analysis of diphthong vowels

Diphthongs are single vowels with two targets. In other words the tongue must attempt to move from one position to another in order for the diphthong to be fully pronounced.

In the following discussion /eː/ is excluded because it as regarded as a tense monophthong (with an optional offglide) and /ʊə/ is excluded because it is either extinct or is almost extinct.

One way of performing a distinctive feature analysis of diphthongs is to analyse each target separately. This type of analysis appears to assume that diphthongs are "underlyingly" a sequence of two monophthongs. In this approach, each target would be allocated a full set of features. This approach has the advantage of not requiring the invention of new features. Its weakness is that implicitly it treats diphthongs as two vowels.

In Australian English all diphthongs are pronounced with distinct first targets and very brief or sometimes absent second targets. They are realised as having a clear first target followed by a long offglide in the direction of some intended second target, which may have a very short or even a zero length. In unaccented contexts (i.e. not affected by sentence stress) diphthongs may be reduced to a monophthong (often similar to the first target) or even to schwa.

Australian English diphthongs can be divided into three classes of diphthongs on the basis of the direction of their offglides. These directions are centring /ɪə/, high-fronting /æɪ, ɑe, ɔɪ/, and a third class of diphthongs /æɔ, əʉ/ whose offglide moves toward a high-central, high-back or a back position. We have already dealt with /ɪə/ above and so we will omit it from this analysis.

These classes could be differentiated by inventing three new features:-

1. [offglide]: true for all diphthongs
2. [centring]: true for centring diphthongs (where the vowel glides to a more central position following the first target)
3. [fronting]: true for the high-fronting diphthongs (where the glides move toward the position of [i]).

The feature [centring] seems redundant, however, as the centring diphthong can be treated as a [-fronting] diphthong. Further, all of these diphthongs can be distinguished by the features for the first target so the fronting feature, whilst descriptive, is redundant.

In the following table, offglide indicates the presence of a second target whilst the other features describe the quality of the stronger first target.

Feature table for Australian English offgliding diphthongs

The main problem with this analysis is its failure to capture the rounding gesture into the second targets of /əʉ/ and /æɔ/. This also results in /æɪ/ and /æɔ/ having identical feature specifications. This failure is a possible argument for providing a featural specification of both diphthong targets. However, if we apply [+round] to any diphthong that contains a round lip gesture in either target (see below in the next table). We continue to apply height and fronting features based on the first target.

We will now bring all of this together into a single table of all Australian English vowel phonemes. Vowels are defined as being syllabic, con-consonantal, sonorant, continuant speech sounds and therefore they all share the following features [+syll, -cons, +son, +cont] . The following table includes only those vowel features necessary for distinguishing between vowel phonemes. In the following table "0" means "inapplicable feature" whilst "+/-" means optional (either are possible as are intermediate values).

Note that schwa explicitly has no onglide or offglide. Its too short to have multiple targets or complex target onsets or offsets. In fact, its generally too short to even have a clear target.

We can also see from this table that we are assuming that Australian English has 19 vowel phonemes. This is based on the following assumptions:-

1. /ə/ is really a phoneme and is not merely a reduced allophone of many vowel phonemes.
2. The occasionally attested examples (only 2 or 3 minimal pairs) of a distinction between /æ/ and /æː/ are not strong enough evidence to support two phonemes. These examples are controversial and there is no agreement on whether they are true minimal pairs or whether they also vary in some other way (grammatical, morphological). For example, the "bad" [bæːd] versus "bade" [bæd] example is weak because the latter word is absent from most people's vocabulary, and the durational distinction claimed to exist between them is less than the normal intrapersonal (same-person) variation in the length of this vowel. It is the most variable Australian English vowel in terms of its duration and can be produced relatively long or short. Further, it has been suggested that the distinction between "bad" and "bade" may merely be evidence of some underlying morphophonetic effect ("bade" being an inflected past tense form of "to bid").
3. The old phoneme /ʊə/ is now extinct and can no longer be considered a phoneme in Australian English.

7. English consonant features (an overview)

The following table brings together (especially in sections 4 and 5) all of the consonants of English into a single table. It is simplified somewhat. For example, dark (velarised or vocalised) /l/ is ignored in this analysis (being an allophonic variation it doesn't really require a phonological feature). The most common features of the most common allophones of each consonant as they exist in Australian English are used as the basis of this table. This table, whilst based on Australian English, is applicable to most dialects of English with only minor changes.

If you carefully examine this feature matrix you will note that no two consonants have the same pattern of features. It is clear that this feature set is sufficient to distinguish between all of the consonants of Australian English.

If you were to consider the distinction between the two main allophones of /l/ (clear /l/ and dark /l/) it would be a simple matter of changing the [-back] to [+back] to indicate the velarising of the dark /l/. That is, its possible to also use distinctive feature to distinguish between allophones of a phoneme. Its not clear how valid it would be to do this as the distinctive features were developed as phonological rather than phonetic features and were in tended to distinguish between phonemes, but distinctive features can be used in rules that explain the selection of competing allophones.

Bibliography

1. Bernard, J., 1970. "On nucleus component durations", Language and Speech, 13(2).
2. Bernard, J., and Mannell, R., 1986, "A study of /h_d/ words in Australian English", Working Papers of the Speech, Hearing and Language Research Centre, Macquarie University
3. Best, C. (1995), "A direct realist view of cross-language speech perception", in Strange, W. (ed.) Speech Perception and Linguistic Experience: Issues in cross-language research, Maryland: York Press.
4. Boersma, P. (1998), Functional Phonology: Formalizing the interactions between articulatory and perceptual drives, PhD dissertation, University of Amsterdam.
5. Browman, C., and Goldstein, L. (1992), "Articulatory phonology: an overview", Phonetica, 49, 155-180.
6. Browman, C., and Goldstein, L. (1993), "Dynamics and articulatory phonology", Haskins Laboratories Status Report on Speech Research, SR-113, 51-62.
7. Chomsky, N., and Halle, M. 1968. The Sound Pattern of English. New York: Harper & Row.
8. Cléments, G., and Keyser, S. 1983. CV Phonology: A Generative Theory of the Syllable. Linguistic Inquiry Monographs, 9. Cambridge, Mass.: MIT Press.
9. Cox, F., An acoustic study of vowel variation in Australian English, Unpublished doctoral dissertation, Macquarie University, 1996.
10. de Saussure, F. 1916, Cours de linguistique générale (publié par C.Bally et A.Sechehaye, avec la collaboration de A.Riedlinger) Paris:Payot. English translation with introduction and notes by W.Baskin 1959, Course in General Linguistics New York: The Philosophical Library. Reprinted 1966, New York: McGraw-Hill.
11. Fant, G. 1973. Speech Sounds and Features. The MIT Press. Cambridge, MA, USA
12. Fowler, C. (1986), "An event approach to the study of speech perception from a direct-realist perspective", Journal of Phonetics, 14:3-28.
13. Goldstein, L., and Fowler, C. (2003), "Articulatory phonology: A phonology for public language use", in Schiller, N. and Meyer, A. (eds.) Phonetics and Phonology in Language Comprehension and Production, New York: Mouton de Gruyter.
14. Halle, M. and Clements, G. 1983. Problem Book in Phonology, MIT Press.
15. Halle, M. and Stevens, K.N. 1971. "A note on laryngeal features", Quarterly Progress Report. MIT Research Laboratory of Electronics, 101, 198-213.
16. Jakobson, R. 1939. "Observations sur le classement phonologique des des consonnes", Proceedings if the Third International Congress of Phonetic Sciences (Ghent) 31-41. Reprinted in Jakobson 1962: 272-9.
17. Jakobson, R., "Child Language, Aphasia and Phonological Universals", 1941 (English translation published by Mouton in 1968)
18. Jakobson, R. 1949. "On the identification of phonemic entities". Travaux du Cercle Linguistique de Copenhague 5: 205-213. Reprinted in Jakobson 1962: 418-425.
19. Jakobson, R. 1962. Selected Writings. The Hague: Mouton.
20. Jakobson, R., Fant, C.G.M. and Halle, M. 1952. Preliminaries to speech analysis: the distinctive features and their correlates. Cambridge, Mass.: MIT Press. (MIT Acoustics Laboratory Technical Report 13.)
21. Jakobson, R. and Halle, M. 1956. Fundamentals of Language. The Hague: Mouton.
22. Ladefoged, P. 1971. Preliminaries to linguistic phonetics. Chicago: University of Chicago Press.
23. Liberman, A.M., Mattingly, I.G., & Turvey, M.T. (1967), "Language codes and memory codes", in Coding processes in Human Memory, eds. Melton, A.W., & Martin, E., Washington: V.H. Winston
24. Liberman, A.M. and Mattingly, I.G. (1985), " The motor theory of speech perception revised", Cognition, 21:1-36.
25. Prince, A. & Smolensky, P. (1993), Optimality Theory: Constraint interaction in generative grammar. Rutgers University Center for Cognitive Science Technical Report 2.
26. Trubetzkoy, N.S. 1939. Grundzüge der Phonologie. Travaux du Cercle Linguistique de Prague 7, Reprinted 1958, Göttingen: Vandenhoek & Ruprecht. Translated into English by C.A.M.Baltaxe 1969 as Principles of Phonology, Berkeley: University of California Press.
27. Watson, C., Maclagan, M. and Harrington, J., 2000, "Acoustic evidence for vowel change in New Zealand English", Language Variation and Change, 51-68.
28. Wikipedia, "Falsifiability", http://en.wikipedia.org/wiki/Falsifiability (accessed 2008/5/1)