Chapter V

Segmental Strength, Hierarchies,
and Phonological Primes¹

5.1.0 Linguists have felt that the segments that occur in languages have a relationship of relative strength. Hooper (1975) offers a strength hierarchy similar to the following one as a strategem for explaining the tautosyllabic distribution of segments (cf. 206):

She considers the optimal distribution of segments in the syllable to be:

Suggestions such as this abound in the literature in which linguists have attempted to define the syllable on phonetic grounds. This pattern and the hierarchy she proposes as well as their congeners seem correct from a phonetic point of view, yet Hooper takes no position concerning physical correlates to the hierarchy. She views strength as (198) a .. cover feature ... [a] theoretical construct, not entirely divorced from physical reality, . . . but . . . [whose] importance is seen only in . . . [its] function in a linguistic system.” For this reason she determines her hierarchy on “phonological” grounds which are functional and dynamic.

5.1.1 Strong position. The number of contrasts in a particular position determines the relative strength of such a position. Hooper uses the Spanish syllable as an example showing that all phonemes in that language contrast syllable initially which suggests to her that syllable onsets are the strongest position from which it follows that initial position is the natural position for strong consonants.

5.1.2 Weak position. Positions within the syllable in which relatively few phonemes contrast are weak. In most languages fewer consonants contrast in syllable final position, so, she reasons, syllable final position, a weak position, is the ideal place for weak consonants.

5.1.3 Reduction of contrasts. Hooper uses the Spanish syllable as an example of this phenomenon. Different registers of that language allow a greater or smaller number of consonants to contrast syllable finally. Most “formal” or “academic” varieties permit up to 15 consonants syllable finally: /p t k, b d g, f,θ, s, x, r, 1, n/ as well as the clusters /ns, rs, ks/. Some “folk” varieties only permit 6: /(θ), (d), r, 1, s, n/ while others delete all syllable final contoids. While Hooper fails to mention it, the stronger segments are most readily deleted in this position.

5.1.4 Neutralization of contrasts. Positions in which contrasts are neutralized are weak positions. Contrasts such as [±surd], [±high], [±occlusion], and [±delayed transition] are often neutralized syllable finally.

5.1.5 Susceptability to deletion. The greater likelihood that a segment may be deleted is an indication of its relative strength. Hooper claims that historical evolution of the type verified in Iberian Romance: tt>t; t>d; d>; >ϕ, suggests that voiceless stops are stronger than voiced and so on.

5.2.0 Foley (1977) postulates a set of primes of phonological strength which he derives from (25) “. . . thoughtful considerations of natural language.” and from the examination of phonological processes. These primes are:

5.2.1 Propensity towards spirantization. The weaker segments have a greater tendency to spirantize. He assigns number values to the segments as follows: strong (labials) - 3; medium (dentals) - 2; weak (velars²) - 1.

5.2.2 Segmental strength. Geminate voiceless stops are the strongest, then voiced stops, then voiced fricatives:³ kk - 4, k - 3, g - 2, g - 1.

5.2.3 Degree of resonance. This is the propensity that a segment has for vocalization: stops are least likely, non-occlusive, non-fricative segments are most likely: r - 5, 1 - 4, n - 3, s - 2, t - l.⁴

5.2.4 Binding strength. Sequences of segments, e.g. kw, are weakest, doubly articulated segments are stronger, e.g. k^w, and units, e.g./p, are strongest:⁵ sequence - 1, complex unit - 2, unit - 3.

5.2.5 Vowel strength. 1. High vowels are weakest and low are strongest: high - 1, mid - 2, low - 3.

5.2.6 Vowel strength 2. Front vowels are weak and back are strong: front - 1, back - 2.

5.2.7 Foley seems to regard the total strength of segments to be the sum of the number values of his theoretical primes (cf. 45, 90), yet his primes do not combine into a coherent system which establishes strength relationships among classes of segments.

5.3 There are many shortcomings to both approaches to segmental strength. Despite his admonitions against “phonetic reductionism” (25) Foley’s parameters may be directly correlated with articulatory gestures. Even though Hooper’s attempt seems to be the better of the two and her hierarchy of segmental strength seems intuitively correct, it is disturbing that diachronic evidence from only 4 languages (Spanish, Icelandic, Pali-Sanskrit, and Ijo) suffices for her to claim that her hierarchy is universal. It is even more disturbing that she suggests no articulatory correlates for a hierarchy whose members (including the glides) are articulatorily defined. Both researchers’ avoidance of articulatory considerations stems from the parameters with which they view phonology. This avoidance is inappropriate because all phonological systems have an articulatory origin, they are products of the (universal) human vocal tract which is regulated by the brain. Neither approach defines segmental strength; each only amasses sets of disparate behavioral characteristics which the authors take to reflect the strength of segments.

5.4.1 Segments themselves are only constructs or artifacts that linguists use to transcribe speech. They are signs which denote complex articulatory gestures imperfectly, but, in phonemic transcription, in a manner sufficient to represent distinctive phonic elements used in transmitting messages. Indeed, segments are as theoretical as Hooper and Foley feel that strength is, but strength is not difficult to define. It is a combination of the amount of impedence applied to the supraglottal egressive air stream, the number of facultative resonance chambers used, and deviation from the voiced state of the glottis.

5.4.2 I assume egressive air and laryngeal resonance to be basic to the act of speaking. All languages employ egressive pulmonic air; all languages have voiced segments; all have suprasegmentals requiring laryngeal resonance; and primative communication, crying, involves laryngeal resonance. The ubiquity of voice in human communication suggests that voicelessness is the marked state. The ubiquity of egressive pulmonic air flow in human communication suggests that impediments or deviations from this state represent markedness—the greater the deviation and the greater the impediment(s), the greater the strength or the markedness of the segments involved. By way of illustration, compare the amount of force involved in articulating a voiceless as opposed to a voiced stop. In the production of, say, p, the lips need to be tightly closed since the egressive pulmonic air comes to the oral cavity unimpeded by the vocal bands (they are open), therefore great stricture is required to occlude the air stream. In the articulation of a b the air is already partially impeded at the vocal bands thus the occluding force at the lips need not be so strong. A click, , for example, involves voicelessness, two points of articulation (both occlusive), and suction, and is stronger and more marked than a segment such a p.

5.5.1 The ideal distinctive feature inventory, as a phonetic theory of phonological systems, should reflect greater and greater strength with more and more positive specifications of segments thus providing more marks for marked segments and a positive correlation for segmental strength hierarchies such as Hooper’s. To test both the strength hierarchy and distinctive feature hypotheses one can compare the average numbers of positive distinctive feature specifications of segments included in the hierarchy with their intuitive arrangements in the hierarchy. To this end I have made an inventory of segments which purports to represent the distinctive human articulatory possibilities. While it does not exhaust the anthropophonic potential (probably a half million sounds), it represents a reasonably thorough sampling from which reliable averages of positive specifications can be acquired to represent the relative strength of the segment types in the hierarchy. The segments are:

5.5.2.1 In order to test the way in which distinctive feature inventories reflect the strength of segments, I devised matrices (see Appendix B) in which I specified the segments in the inventory above first using the SPE feature inventory which consists of the following primes:

Counting and averaging the positive specifications in the matrices yielded the following results, in the order of descending positive specifications:

5.5.2.2 The hierarchy established by counting the average number of positive specifications of all segments is unsatisfactory, but it becomes even more so if one counts only the natural, i.e., voiced, liquids and nasals. It is small wonder that linguists trying to establish strength hierarchies, but committed to using SPE features, had to claim that strength was not a phonetic concept. One can, of course, argue that the SPE features were not designed to reflect the strength of segments, but, then, one must ask, what was their purpose? If an alternate set, with fewer primes, can distinguish all the segments the SPE set distinguishes, and can be used to write rules in an economic fashion, then it is a better theory. If it also reflects the strength of segments in a manner consistent with that which linguists have felt to be correct, then it is even better. The revised set of articulatory primes accomplishes both missions.

The following table shows the average number of positive specifications in the matrices using the revised inventory of distinctive features.

5.6 In Chapter IV, I compared and contrasted the distinctive features proposed by JFH with those proposed here. I found that seven of the features are the rough equivalents of one another, that [lateral] is an apt substitute for [vocalic] and that one more articulation parameter [nondiffuse] must be added as well as a friction parameter. Adding these parameters, replacing [voiced] with [surd], [tense] with [widened glottis] (see 3.2.3.1-4) and adding the prosodic features make the inventory proposed here and the JFH inventory rough equivalents—the only difference is in the point of articulation features [labial, high, raised, dorsal] which contrast with [compact, nondiffuse, grave, flat]. The revised JFH inventory contains the following primes:

The revised JFH specifications reconfirm the strength hierarchy as set forth by Hooper. Had [voiced] and [tense] been maintained as primes, along with [vocalic] and [strident], this would not have been the case. The classification that the parameters [compact, nondiffuse, grave, and flat] provide for vocoids: high - 1.7, mid - 2.6, and low - 2 run against the belief that the higher the vocoid the higher the strength. For the time being then, I submit that the feature system that I propose is the best of the three in existence since it distinguishes all that the SPE set does but uses fewer primes; it is of higher observational adequacy than the PSA primes; and it reflects the strength of segments better than the official version of either of the opposing theories.

5.7 The distinctive feature inventory that I propose is motivated on phonetic, phonological, and semiotic grounds rather than as an attempt to reflect the strength hierarchy that linguists have believed in. Its reflection of this hierarchy gives the inventory added credibility and buttresses the hierarchy which up to now has been based on linguists’ intuitions rather than on theoretical primes. Future research in syllable structure, dynamic phonology (metasteses), phonological evolution, segmental relative frequency in texts, and markedness will serve as tests of descriptive and explanatory adequacy of these primes as a theory of phonological systems. I hope that the five chapters of this monograph plus the information in Appendices A and B will motivate phonologists to take these primes seriously and to reëxamine problematic areas of segmental description using them as analytic primes. I have used them in studies I have already published (1979, 1980a) and have found that they work quite well. I must admit, however, that much more rigorous testing is necessary before they can be accepted as the distinctive phonetic primes of phonological analysis.

Footnotes

¹An earlier version of this chapter is Brakel (1979). I am grateful to the Chicago Linguistic Society for allowing me to publish this version.

²Foley takes no stand on the relative strength of palatal and pharyngeal segments.

³Foley does not include voiceless fricatives in this sub-hierarchy.

⁴Foley considers the propensity to vocalize a strength. Vocalization, to be in line with primes 1 and 2, should be seen as a weakening and r should, perhaps, be given the lowest number and t, the highest.

⁵Foley considers p a tightly bound kw.

⁶Pitch and stress are not considered in segmental specification. They are included here because they are part of the SPE inventory.

⁷The matrices for all these tables can be found in Appendix B.