Phonemes, Allophones and Suprasegmentals

Sonority hierarchy and syllable structure

Sonority is the relative loudness or resonance of a speech sound, and it provides the basis for syllable structure. Segments are ranked on a scale in which stops are the least sonorous, followed by fricatives, nasals, liquids, glides, and vowels. Among vowels, low vowels are more sonorous than mid vowels, and mid vowels are more sonorous than high vowels. Within a manner class, voiced segments are more sonorous than voiceless ones, so voiced stops rank above voiceless stops. A sequence that rises in sonority from left to right follows the scale at every step, whereas a sequence that falls in sonority at some point does not. For example, a voiceless stop followed by a voiced stop, then a lateral, then a low vowel is a rising sequence. A syllable has an obligatory nucleus, usually a vowel, which is the peak of sonority. Onset and coda consonants are optional, so structures such as nucleus-coda, onset-nucleus, and onset-nucleus-coda are all well formed. A string consisting only of an onset and a coda, with no nucleus, cannot form a syllable. Languages differ in which onsets and codas they allow. In a language that forbids onset clusters, each syllable begins with at most one consonant. Any form with a cluster in onset position must therefore be repaired, typically by vowel epenthesis or by consonant deletion, and a data problem of this type requires applying the constraint to every form before counting the repaired outputs.

Consonant articulatory features

Phonological features describe a segment as a bundle of properties, and a natural class is a set of segments that share one or more features. For place of articulation, the alveolar class includes /s/, /z/, /n/, /t/, /d/, and /l/. Dental, postalveolar, and palatal consonants belong to different place classes, so a set that mixes these places with alveolars is not a natural class of alveolars. A set such as /t/, /d/, /l/ shares a single place and is therefore a natural class. The feature [continuant] distinguishes sounds in which air flows through the oral tract without complete closure, which are [+continuant], from sounds with complete oral closure, which are [-continuant]. Fricatives, approximants, liquids, and glides are [+continuant], including the glottal fricative /h/. Stops and affricates are [-continuant], and nasals are also [-continuant] in standard feature systems because the oral tract is closed even though air escapes through the nose. A set containing several fricatives and glides therefore contains at least two [+continuant] sounds, whereas a set made only of stops contains none.

Lexical tone and autosegmental representation

Lexical tone is pitch used to distinguish word meaning, and it is not a feature of all languages in India. Dravidian languages such as Tamil, Telugu, Kannada, and Malayalam are non-tonal, so a statement that all of them lack lexical tone is true. Munda languages, which belong to the Austroasiatic family, such as Mundari and Santali, are also non-tonal, so the claim that all Austroasiatic languages in India are tonal is false. Indo-Aryan languages are largely non-tonal, while Tibeto-Burman languages show internal variation, with some members tonal and others not, so no universal claim can be made for the whole family. Autosegmental representation separates the tone tier from the segmental tier. Tone-bearing units (TBUs), usually vowels or syllables, are linked to tones by association lines. Three well-formedness conditions govern these links: every TBU is associated with at least one tone, every tone is associated with at least one TBU, and association lines do not cross. A TBU left unlinked violates the first condition, and a tone left unlinked violates the second. Crossing association lines violate the third. A single tone may link to several adjacent TBUs, as in tone spreading, without violating any condition.

Phonological rule derivation

A phonological derivation starts from an underlying form, applies rules in a fixed order, and yields a surface form. To test whether a candidate output is possible under a rule, first check whether the rule's structural description is met in the candidate, and then check whether the candidate satisfies the structural change. For example, if a nasal assimilates in place to a following consonant, any candidate in which a nasal precedes a consonant of a different place is ruled out, because the output must show the required place agreement. Rule ordering describes how one rule affects the application of another. In a feeding relation, an earlier rule creates the environment for a later rule. In a bleeding relation, an earlier rule removes the environment of a later rule, so the later rule cannot apply. Counter-feeding and counter-bleeding are the corresponding patterns in which the later rule, rather than the earlier one, creates or destroys the environment. In English plurals, the plural /z/ devoices after voiceless consonants, as in cats, and an epenthetic vowel placed between the stem and the affix removes the voiceless context for that devoicing. Applying epenthesis before devoicing therefore produces a bleeding relation.