The autosegmental formalism deals with several separate linear sequences in such a way that a phonological representation is depicted on several distinct tiers, e.g. the segmental tier, the stress tier, or the tonal tier. Each of these tiers consists of different features.
2.4.1 Feature geometry for consonants
Fig. 2 displays how the relevant features for consonants are organized in the form of a feature-geometric tree.
Fig. 2: Feature geometry for consonants
←41 | 42→
Fig. 2 shows the different features used for the consonants that were listed in Section 2.3.1. McCarthy (1988) assumes that the grouping of features is based not so much on any similarity of articulatory or acoustic correlates as on the functional coherence of the feature groupings in phonological rule systems. Feature geometry gathers groups of features under a shared parent node or class node in a tree, and operations on the entire set can be encoded as operations on the class node. The tree in Fig. 2 contains a root node—the topmost node—which includes the binary major class features [±consonantal], [±vocalic], and [±sonorant]. The reason why these features constitute the root node is that none of them can spread separately, as is the case for most other features, like the feature [nasal], which can spread without spreading any other part of the tree; see Clements and Hume (1995). The distinctive features are grouped into categories with regard to the different natural classes of segments that they describe. These groups of categories are the major class features, laryngeal features, manner features, and place features.
From the root node in Fig. 2, there is a main division between the Laryngeal (L) and Supralaryngeal (SL) nodes. The L node comprises the features of the larynx. Features such as [nasal] and [continuant] are dependent on the SL node, which comprises the articulators [labial], [coronal], and [dorsal], which are grouped under the class node Articulators (Artic).
Segments are only specified by the relevant features that distinguish them from the other segments. Let us start with the L consonants. These consonants use the glottis as their first articulator. They have three distinctive features: [±voiced], indicating whether or not there is vibration of the vocal cords; [spread glottis] or [s.g.], indicating the aspiration of a segment with regard to the openness of ←42 | 43→the glottis; and [constricted glottis] or [c.g.], which denotes the degree of closure of the glottis. The feature [c.g.] is used to specify the glottal stop [ʔ] and the feature [s.g.] is used for the glottal fricative [h];. The feature tree for the glottal stop appears in Fig. 3. Only the L feature [c.g.] needs to be specified, but no SL feature. The same is true for [h]: only the feature [s.g.] is needed for this segment.
Fig. 3: Feature-geometric tree for the glottal stop [ʔ]
Turning next to the obstruents, it was shown above that the obstruents consist of stops and fricatives, distinguished by the feature [±continuant]: the stops are [-continuant] and the fricatives [+continuant]. They are all [+consonantal], [-vocalic], and [-sonorant]; and their articulators are [labial], [coronal], and [dorsal]. They can be voiced or voiceless.
Fig. 4 shows the feature-geometric representation for a voiceless labial stop [p];.
Fig. 4: Feature-geometric tree for [p];
Fig. 5 shows the features of the stops [t]; and [g].
Fig. 5: Feature-geometric tree for coronal [t]; and dorsal [g]
The labiovelar segments [kp] and [gb] combine two articulators, [labial] and [dorsal]; see Fig. 6.
Fig. 6: Feature-geometric tree for the labiovelar consonants
←43 | 44→
Fricatives are [f]; and [s]; see Fig. 7 for a tree-geometric representation of [f]. Recall that there is no dorsal fricative in Fròʔò.
Fig. 7: Fricative [f]; in a feature-geometric tree representation
Sonorant consonants are distinguished from obstruents with the feature [±sonorant] bearing the positive value [+sonorant].
Depending on whether the velum is lowered or raised, a nasal or oral segment is produced, respectively. The trees in Figs. 8 and 9 show the features for [m]; and [j]. Recall that nasals are [-continuant].
Fig. 8: Feature-geometric tree for [m];
Fig. 9: Feature-geometric tree for [j];
←44 | 45→
Both glides, [j]; and [w], are in some instances considered as [+cons], [+voc], thus entailed [+son] as it is shown in Fig. 10. But note that in my data the letter segments are performing consonantal function instead of vocalic function.
Fig. 10: Features for the glides [w]; and [j]
2.4.2 Feature geometry for vowels
Fig. 11 shows all vocalic features needed for Fròʔò. The root node contains the features [-consonantal], [+vocalic], and [+sonorant]. There are no L features for vowels, only SL ones. Vowel Specification (VS) is a class node dependent on the SL class node that regroups all vocalic features.
Fig. 11: Feature tree for the vowels of Fròʔò
←45 | 46→
The vocalic features [±high], [low], [±back], [±ATR], and [nasal] are the features needed to distinguish all vowels and make the necessary natural classes among them. Some examples appear in Fig. 12.
Fig. 12: Feature-geometric tree for [i];, [o], [ɛ], and [a]
←46 | 47→