3 Phonetic and phonological parameters of approximants
3.1 Phonetic parameters
Three kinds of phonetic parameters can be distinguished: articulatory, acoustic, and auditory.
3.1.1 Articulatory parameters
In articulatory phonetics, Bicker & Floyd (2006: 99) mention two parameters permitting the delimitation of approximants: length and degree of constriction.
Length
Gendrot et al. (2015) observe that the French voiceless fricative rhotic [χ] has longer durations than the corresponding reduced form realized as a voiced approximant.
A discriminant analysis of French rhotics in Ramasse (2017: 29) shows that approximants are shorter than fricatives. A short study[1] on vowels to complete this description was carried out: the length of 120 back vowels of 12 speakers (6 males, 6 females) in the corpus described by Ramasse (2017) was measured. Welch’s ANOVA showed a significant difference in length between voiceless fricatives, approximants, and vowels. However, there was no significant difference in length between approximants and voiced fricatives.
The length values of the compared categories of sounds are shown in Table 4. This comparison permitted us to determine that the two limits between the three distributions are significantly different: 66 ms between approximants (+ voiced fricatives) and voiceless fricatives and 84.5 ms between voiceless fricatives and vowels. These limits are illustrated in Figure 1.
Table 4. Length of French rhotics and back vowels in ms.
| N | Mean | Std. deviation | ||
| 1 | Approximants | 452 | 55.50 | 23.051 |
| 2 | Voiced fricatives | 249 | 56.49 | 23.383 |
| 3 | Voiceless fricatives | 299 | 68.28 | 24.975 |
| 4 | Vowels | 120 | 84.47 | 44.903 |
Figure 1. Limits for the parameter of length (in ms) between approximants, voiceless fricatives, and vowels
As shown in Figure 1, the approximants studied here have a maximal length of 66 ms and are shorter than voiceless fricatives and vowels.
Degree of constriction
Bicker & Floyd (2006) consider that semivowels present a greater constriction than vowels. However, Bothorel et al.’s (1986) cineradiographic study contradicts this assertion, at least for French: for the same subjects, there is no difference in the degree of constriction between a semivowel and the corresponding close vowel (between [j] and [i], [ɥ] and [y], and [w] and [u]).
Nevertheless, other American English approximants, liquids, present a different degree of constriction than fricatives and vowels. The degree of constriction is estimated by Catford (1977) in terms of “cross-sectional area of the oral articulatory channel”. Studies using resonance magnetic imaging permit the cross-sectional area of the oral articulatory channel for the classes of sounds in question to be precisely determined.
Narayaman et al. (1995) study cross-sectional areas of American English fricatives produced by four speakers. There is no significant difference between voiceless (with a mean of 17.9 mm2) and voiced (with a mean of 19.1 mm2) fricatives for this parameter. These results contradict Catford (1977: 123-124), who estimated two different cross-sectional area values, 10 mm2 for voiceless fricatives and 10.6 mm2 for voiced fricatives.
As shown in Table 5, the mean and standard deviation of all (voiceless and voiced) fricatives are 18.5 and 5.8, respectively.
Table 5. Cross-sectional area values (in mm2) of 3 categories of American English sounds.
| N | Mean (mm2) | Std. deviation | ||
| 1 | Fricatives | 32 | 18.5 | 5.8 |
| 2 | Approximants (laterals) | 25 | 37.9 | 19.2 |
| 3 | Front vowels | 32 | 64.3 | 19.9 |
Figure 2. Limits, for the parameter of cross-sectional area (in mm2), between three significantly distinct distributions of American English sounds: fricatives, approximants (laterals) and front vowels.
Values of cross-sectional areas of approximant liquids are given in two studies: Alwan et al. (1997) for rhotics and Narayanan et al. (1997) for laterals. However, because of their instability, rhotics cannot be used here in the comparison with sounds of other classes [2].
For laterals, Narayanan et al. (1997) showed that there are two flow channels along the side of the tongue. To compare areas of the central channel of fricatives and vowels, the two lateral areas were added, and the obtained mean and standard deviation were 37.9 and 19.2, respectively.
Only the buccal cross-sectional area values of front vowels given by Story (2008) are considered here. The mean and standard deviation are 64.3 and 19.9, respectively. Figure 2 shows that the cross-sectional area values of approximants stand between 28 and 52. The value used by Catford (1977: 124) is 40; this estimated value agrees with the observed values.
3.1.2 Acoustic parameters
Intensity
In Ramasse’s (2017: 28) study of French rhotics, intensity, more precisely the sound pressure level (SPL), appears to be the most important parameter allowing discrimination between fricatives and approximants. The SPL fricative values are significantly lower than the approximant values.
A short supplementary study, analogous to the study of vowel length, was carried out: the intensity of 120 back vowels of 12 speakers (6 male, 6 female) in the corpus described by Ramasse (2017) was measured. The SPL values of the categories of the compared sounds are shown in Table 6. The limits between the categories are shown and illustrated in Figure 3. Welch’s ANOVA and post hoc Tamhane’s T2 tests show that there is a significant difference between the distributions. It appears that, according to intensity, approximants are intermediate between fricatives and vowels, with limits of 65.5 dB and 74.3 dB.
Table 6. Intensity values (in dB SPL) of French fricatives, approximants, and vowels.
| N | Mean | Std. Deviation | |
| 1 voiceless fricatives | 299 | 59.49 | 8.04 |
| 2 voiced fricatives | 249 | 63.66 | 6.51 |
| 3 voiceless approximants | 52 | 67.19 | 6.70 |
| 4 voiced approximants | 391 | 70.60 | 5.74 |
| 5 vowels | 120 | 78.45 | 3.60 |
Figure 3. Limits for the parameter of intensity (in dB SPL) between fricatives, approximants (French rhotics) and vowels (back vowels).
Formant structure
Frequency
According to Stevens (1998: 532), semivowels present “a relatively low first-formant frequency in the range of 250 to 300 Hz”.
In a comparison between the vowels and semivowels of Amharic, Yoruba, and Zuni, Maddieson & Emmorey (1985:171) conclude that semivowel F1 values are lower than those of the corresponding vowels. This proves a narrower constriction of semivowels. The semivowel narrower constriction not noticeable in French on Bothorel et al.’s (1986) cineradiographic study drawings is revealed by a lower F1, because F1 values are correlated with the degree of constriction, for languages other than French.
Bandwidth
Compared to vowels, there is an increase in the bandwidth of formants, especially the first formant of glides and liquids, according to Stevens (1998: 534).
According to Ramasse (2017: 26), the formant bandwidth is narrower for approximants than for voiced fricatives. There seems to be an inverse correlation between the vocalic quality of a sound and the value of its bandwidth.
3.1.3 Auditory criterion: relative sonority
Relative sonority, according to Jones (1922), is related to two factors: intensity (force of the breath) and what he called the quality of the sound, which is directly linked to the degree of constriction of the produced sound.
According to Jones (1922: 12), approximants, especially semivowels, are distinct from vowels because of their lesser degree of relative sonority and their shorter length.
Approximants are thus between fricatives (less sonorous) and vowels (more sonorous) on a continuum of sonority.
3.2 Phonological parameters
As described by Roach (2011), English approximants are composed of glides and liquids.
Only liquids are generally considered in phonological descriptions except in Williamson (1977), where glides are considered.
3.2.1 Binary features
Jakobson et al.’s (1951) two fundamental source features, vocalic and consonantal, permit the characterization of liquids and oppose them to fricatives and vowels. These two features are borrowed by Chomsky and Halle (1968), but vocalic is then replaced by the feature syllabic “which would characterize all segments constituting a syllable peak” (Chomsky & Halle,1968: 354).
Introducing this feature, they distinguish nonsyllabic liquids and syllabic liquids and do the same with nonsyllabic nasals and syllabic nasals. A third feature, sonorant (production of the sound with a vocal tract cavity configuration in which spontaneous voicing is possible), is used.
Clements (1990) replaces consonantal with vocoid, a term borrowed from Pike, which he defines as the converse of consonantal. He puts forward a fourth feature, approximant: “I will consider an approximant to be any sound produced with an oral tract stricture open enough so that airflow through it is turbulent only if it is voiceless. . . . We will treat approximant as a binary feature, like the other major class features” (Clements, 1990: 293). He uses a new feature, vocoid, which he defines as the converse of consonantal. “‘Vocoid,’ a term introduced by Pike (1943), is simply the converse of the traditional feature ‘consonantal’ and is defined accordingly”. Clements (1990: 293)
The last feature is sonorant.
As shown in Table 7, these three features characterize the four degrees of sonority of the sound classes: obstruents, nasals, liquids and vowels. [approximant] is necessary to oppose approximants (liquid approximants) to nasals on the sonority scale.
Table 7. Opposition between obstruents, nasals, liquids, and vowels according to Clements (1990)
3.2.2 Multivalued feature: Williamson (1977)
Ladefoged (1967: 81) puts forward a phonetic parameter, articulatory stricture, with four principal values: 1. stops, 2. fricatives, 3. approximants – near vowels, and 4. approximants – far vowels.
Lindau (1975), applying this analysis to phonology and to vowels, proposes a multivalued feature [F1] for vowels. Williamson (1977: 845) proposes a multivalued feature [stricture] for consonants, at first with four degrees:
2 Stop,1 Fricative, 0 Approximant, X Absence (of any stricture)
Basing her description on Chomsky & Halle (1968) and Ladefoged (1975), she considers that “nasals and laterals behave in some languages like obstruents and in others like sonorant.” They have the values 2/0, breaking the continuity on the [stricture] continuum, and two other features are necessary to depict the distinctive characteristics of these two classes. As shown in Table 8, these two classes are not kept afterwards. And she keeps only two degrees for vowels (instead of Lindau’s four degrees) because “in the vowel domain, the Ijo evidence suggests that no more than two values are required: high and non-high.” It appears that approximants are at degree 0, intermediate between high vowels (-1) and fricatives (+1).
Table 8. Williamson’s (1977) multivalued [stricture] feature
3.4 Summary of the second approach
Approximants constitute a class of sounds that are shorter than fricatives and vowels. Their cross-sectional area and intensity are intermediate between those of vowels and fricatives.
Phonologically, they are defined as [+ consonantal] or [-vocoid], [-syllabic], [+sonorant] and [+ approximant]. They are opposed to vowels by the feature [vocoid].
It appears that the feature [approximant] is necessary to oppose approximants to nasals.
According to Williamson (1977), approximants correspond to degree 0 of her multivalued feature [stricture] between high vowels (-1) and fricatives (+1). Clements and Williamson treat two different notions of approximants; these notions will be presented in the next section.
[1] As described in Ramasse (2017: 26), length, in the short supplementary study described here, was measured by Speech Analyzer and intensity, for the one described below, by Praat.
[2] Alwan et al. (1997: 1084) sum up their results for [ɹ] in these words: “The areas vary between 0.2–2.0 cm2 depending on the subject’s oral morphology and how the oral constriction was formed”. Such a distribution (mean 73 mm2, standard deviation 68,32) could not be used here for a comparison.