| 50. When we speak of a sentence consisting of words and words consisting of sounds and sounds consisting of attributes, we are concerned with how the various wholes are seen and analyzed as composed of parts. With objects of natural language most of these parts are viewed as occurring in a specific order: words in sentences occur in sequence; their sounds are uttered in sequence, but the attributes of a sound all come together at once. It is possible for the linguist to describe the expressions of language by applying the generative rules of a formal Generative-Transformational Grammar. In describing the syntax of sentences these are called the phrase structure rules. PS-rules describe the structure of a language by giving the constituency of its parts explicitly. The computer scientist calls this the aggregate model and the descriptive rules Backus-Naur form. The same kinds of rules that are helpful in describing the arrangement of words in phrases are useful to describe the arrangement of morphemes within the word and phonemes within the morpheme. |
| 51. Ordering rules. The linguist orders the PS-rules strictly, each one (except the first) describing the constituency of some element introduced in a previous rule. It is sometimes useful to loosen this restriction and open up the possibility of recursion, i.e., defining a element in terms of itself. Some elements, at least the initial one, may describe a constituent part of itself or of one of its parts. |
| 52. The segments of language are the elements that the speaker orders sequentially. The characterization of such elements in terms of features or class membership may make use of segment structure rules. These rules do not say anything about composition, but instead, classification. I will give these rules in figures, but to help in the interpretation of the rules I have provided meta-rules in boxes. These meta-rules delineate conventions for describing the multiple class membership of the various phonological segments. Each distinctive feature within a segment tells of its memebership in a different class. In some rules these features are joined with AND so as to abut its neighbor. This convention will be seen on the right side of the subcategorization rule in box 1. Some rules will require that features be joined with OR. In this situation the segment is really a more general specification of several possible segments. In this case the same abbreviatory convention used for BNF rules in the syntax is made to work in the segment structure rule described in box 2. |
| 53. Figure 7 shows a fragment of the SS-rules that characterize the phones of the IPA system. The first rule contains two green boxes each meant to represent its own kind of phone as a segment. The intent is to state that a phone that is either a consonant [+Consonantal] or a vowel [–Consonantal] is characterized as a segment having features. The orange boxes represent sets of features to specify the place of articulation, the manner of articulation, and the presence (or absence) of voice. The next two rules state that any consonantal segment has a specific place of articulation [kPlace] and a specific manner [lManner]. There is then a rule that states that for non-consonantal segments the place of articulation is along two dimensions: horizontal (front-to-back), and vertical (close-to-open). There follow rules that express the fact that consonants are either voiced or unvoiced and vowels have a feature of being rounded or not. The last two rules tell us that the dimensions of place for vowels are divided up into five, viz. six grades or steps. Thus, these rules divide up the IPA features of phones into 9 x 10 x 2 or 180 consonantal and 6 x 5 x 2 or 60 vowel distinctions. IPA has but 100 basic symbols for phones and many more modifications that can be added to them. Later rules reduce these many distinctions by eliminating those impossible to articulate and many made redundant by their dependence on each other. |
| 54. Figure 8 gives the usual names for each of the features of the IPA system for which I have assigned numbers by rule. Two of them (in turquoise boxes) are generalized features created to group some of the features together so as to be able to refer to any of the group with a single label. This sort of abbreviation is provided by rules formatted identically to those used to combine segments, cf. boxes 3, 5, 6. |
| 55. Redundancy Rules. There are a number of features whose presence on a segment can be predicted based on the presence of certain other features. The need to state these dependencies gives rise to redundancy rules.. The first rule in figure 9 states that there are voiced and unvoiced reflexes for the fricative, plosive, and implosive sounds. The next rule gives the feature of voiced [+Voice] to the nasals, trills, taps/flaps, approximants, and lateral approximants, whereas the one after that characterizes the ejective stops and glottal sounds as unvoiced [–Voice]. Similarly the other rules characterize the other IPA phones according as their features may be predicted on the basis of those already specified. The first segment structure rule gave voice to vowels and there are no redundancy rules to take it away. Voiceless vowels, as occur in Piute or Japanese, are allophonic variations in a particular language, not of the IPA phones, which are used to describe what must lie at the heart of any spoken language. |
| 56. The rules of figures 11 and 12 read like redundancy rules, but relate specifically to the way sounds are produced and perceived in English. For this reason we place the features characterizing the phonemes in bright yellow boxes as explained in box 4. When the rules are observed, the language distinguishes vowels along the dimensions of horizontal (front-to-back), vertical (high-to-low), length (short vs. long), and manner (rounded vs. unrounded). Length was introduced as a variation or interpretation of the closeness, whereas being rounded or unrounded is predictable in English on the basis, generally, of being front or back (when not low). The subclasses of consonants do not correspond fully with the IPA phones either. This is most obvious with the glides and affricates which in IPA are sound complexes. I have placed these classes within turquoise boxes to indicate that they are defined by special rules to be explained later. |
|
57.
At this point the sound rules of English allow us to distinguish all the phonemes of the language from each other on the basis of features that relate directly to the features of the IPA system.
Figure 13 lists the symbols I have chosen to represent these distinct sounds as we conceptualize them alongside the features that make them different from all the others.
These distinctive features are used to model each sound as a segment of speech.
A single feature specifies a whole class or sub-class of sounds, whereas it usually takes more than one feature to specify a phoneme as perceived acoustically.
(It only takes one feature to specify the sound of “h,” classed here as the semivowel.) There is a special challenge in our systemization of the vowel phonemes. It seems clear that the schwa is unrounded. We should like to maintain the symmetry between the long and the short sounds and so /oy/, which is rounded, needs to be taken as unrounded. It appears that the unroundedness in the /y/ must be taken to characterize the long vowel as a whole. |
| 58. Complex Sounds of English. There are two origins of what we might call compound phonemes in English. The first is an artifact of the definitions of the IPA phones. We see this in the necessity of defining tha affricates as combinations of a plosive followed by a fricative. The temporal sequencing of the two articulations is a consequence of the structure of the speech apparatus. Their combined sound or co-articulation is perceived as a single sound distinct from both. |
|
59. Multiple Definitions.
We also see uncertainty in the multiple ways in which the back glide might be defined.
The second rule in figure 14 sees the back glide as the result of pronouncing the voiced bilabial fricative at the same time as the velar approximant.
Neither one of these sounds occurs in the language apart from the back glide, so perhaps this complex definition is not appropriate.
Yet the other alternative is to make it a furtive allophone of the vowel sound of “oo.”
In contrast to the “w” sound is the sound of “y,” whose consonantal character seems not to bother phoneticians.
This too could be seen as a furtive allophone of the vowel sound of “ee.”
In both cases the glide structures as a consonant, but sounds very much like a vowel. The last rule shows how the IPA sound of an unvoiced palatal fricative is realized as one of the allophones of /h/. The environment of this allophone is preceding an /uw/ sound. The question here arises about whether the “y” should be taken as a separate phoneme /y/ or simply as palatalization before the /uw/ sound. However, since the /uw/ does not normally cause palatalization by itself, it appears that it is the front glide combining with the /uw/ that is causing the formation of a distinct vowel allophone. Hence, the difference between “beauty” /byuwtiy/ and “booty” /buwtiy/ is in the vowel, not the consonant. It appears that it is this vowel’s pronunciation as palatalized that is being preserved after the /h/: “hue” /hyuw/ differs from “who” /huw/ in the vowel. |
