Course: Phonology

Vincent Liu

June 14, 2002


Vocoid Approximants

6.1 Introduction

a.            Approximantsè (1) sounds which have a lateral escape, or (2) belong to the family of r-sounds, and (3) those with vocoid characteristics.  They are often known as ¡§semi-vowels¡¨. 

b.           In the UPSID file, vocoid approximants are treated as consonants if (1) they don¡¦t alternate with syllabic vocoid pronunciation and (2) share distributional properties with other consonants.


6.2 Frequency of vocoid approximants

a. Two most frequently-occurring approximants---/j/ and /w/    

The frequency of /j/ and /w/ and their co-occurrence is shown in table 6.1.


Table 6.1 Distribution of /j/ and /w/ in UPSID languages


With /j/

No /j/

With /w/

226  (71.3%)

  14 (4.4%)

No /w/

47  (14.8%)

30 (9.5%)

Total number of languages in UPSID: 317


Analysis of table 6.1:

(1)   The occurrence of /w/ is associated with the occurrence of /j/.


(2)   In the UPSID data, therefore, the occurrence of /w/ usually implies the occurrence of /j/ in the same language. 

(3)   The analysis is different from the results found in Stanford Phonology Achieve, because

(i)                The different selection of languages in the two surveys,

(ii)              The application of different criteria for phonemic status of approximants.

b.      Other vocoid approximants are divided into two groups---

(1)   Those which are modified variants of /j/ and /w/.

(i)                Modified forms of /j/ èvoiceless /j¥/, laryngealized /j¼/, and nasalized /j/.

(ii)              Modified forms of /w/èvoiceless / ã / and laryngealized /w/.  


(2)   Those which have different place of articulation include the labial-palatal approximant / / (4 instances) and the velar approximant / / (5 instances).  


The frequency of these modified segments is given in table 6.2.


Number of languages

Percent of languages










/ ã /







Analysis of table 6.2:

(1)         There is an association between the occurrence of /j/ and /w/.  A diachronic source of this kind for /j/ would predict that it would be more frequent than /w/.  In 12 cases /w/ and /j/ occur together. (Only one exception)

(2)    The voiceless approximants /j¥/ and /ã/ fairly markedly in frequency, / ã / is 1.6 times more frequent than /j¥/çèopposed to their voiced counterparts.

(3)   Diachronic source of these voiceless segments is likely to be similar in both cases.

(i)                  From a cluster of a voiceless obstruent and the voiced approximant.

(ii)                From labialised or palatalized voiceless obstruents.

E.g., (a) */ SW / in Proto-Athabaskan è / ã / in Hupa

    (b) / xW / in Old English è / ã / in Middle English

    (c) */ xW / in Proto-tai è */ ã / in early Northern Tai

(4)   Ways to exit from the inventories:

(i)                  Vocalization (as in the widespread merger of */ã / with /w/ in many varieties of Modern English)

(ii)                Collapse into an undifferentiated voiceless vowel phoneme /h/ (as in the special development of */ã/ in English before /u/ and /o/ in words such as ¡§who¡¨, ¡§whole¡¨)

(iii)               Fricativization (as in the idiolectal /C / for the segments /hj/ = [jj] in English words such as ¡§huge¡¨, ¡§human¡¨)

(5)   The relative frequency of /j¥/ and / ã / suggests that there may be some factor which favors the development of / ã / over /j¥/ or favors the loss of /j¥/ more than /ã/. (Reason: A true voiceless palatal approximant is poorly distinguishable from /h/, which occurs in most languages, and hence is likely to collapse together with it.  If, on the other hand, /j¥/ is articulated more forcefully to preserve the distinction it would become a palatal fricative.  A voiceless labial-velar approximant may survive better because its two strictures produce two cavities with sonorants which are rather close to each other in frequency and reinforce each other.)


6.3 Approximants and related vowels

/j/ è /i/

/w/ è /u/

The greater frequency of /i/ is undoubtedly a predictor of the greater frequency of /j/.  However, for both /j/ and /w/ there are a few languages which have the approximant but lack the corresponding vowel.è Table 6.3


Table 6.3 Common approximants occurring without cognate vowels


No. of languages

% of sample

/j/ but no /i/



/w/ but no /u/




Analysis of table 6.3:

(1)       The systems without /u/ may be regarded as falling into two principal classes:

(i)                  Those with a ¡§compensating¡¨ vowel which is high or back or rounded but not all three.

(ii)      Those which simply have a gap (and whose highest back vowel is usually /o/).

(2)       This is suggestive of a variety of possible sources for /w/ and may predict that the class of /w/ segments in languages may vary phonetically through a greater range than /j/.


The less frequently occurring approximants / ç / and / Ä / are also investigated in relation to the corresponding vowels, in the case /y/ and / µ / respectively. èTable 6.4


Table 6.4 Other approximants and vowels


No. of languages

/ ç / and /y/


/ ç / but no /y/


/ Ä / and /µ /


/ Ä / but no /µ /



Analysis of table 6.4:

Since the numbers are so small no great reliance should be placed on the indications.


6.4 Approximants and related consonants

/j/ and palatalized consonants---

Since disyllabification of high vowels is a major process creating both /j/ and palatalized consonants, it might be expected that palatalized consonants would occur only in languages with /j/. [Three exceptions (Ocaina 805, Muinane 806, Ket 906) but only one (Ocaina) is straightforward]


/w/, labial-velar tops and labialised velars

The labial-velar stops (/k?p/ and /g?b/ are the most common ones) may vary a good deal in their initiation but belong together by virtue of their shared place of articulation.è Table 6.5


Table 6.5 Co-occurrence of /w/ and labial- velar stops


No. of languages

% of sample

/w/, /k?p/ and /g?b/



/w/ and /k?p/



/w/ and /g?b/



No /w/ but /k?p/ and /g?b/




Analysis of table 6.5:

(1)         There is a tendency for /k?p/, /g?b/ to occur in systems with /w/ in preference to those lacking /w/. [One exception (Keplle 103)]

(2)         There is an obvious similarity between labialised consonants and /w/, and there is a historically similar source for both types of sounds in disyllabification of /u/ in many instances.

(3)          Use the most frequent /kW / as the archetypes pf labialised consonants, and the co-occurrences of /w/ with /kW / are shown in table 6.6.


Table 6.6 /w/ and labialised velar stops


No. of languages

% of sample

/w/ and /kW /



No /w/ and /kW /




Analysis of table 6.6:

There is a weak tendency for /kW / to be more likely to occur in languages which have /w/. (Five exceptions: Mixtec 728, Guarani 828, Wantoat 615, Chipewyan 703 and Kpelle 103)


6.5 Other approximants

Apart from the 4 most common approximants /j, w, ç, Ä9/, there are two more found in the UPSID.

(1)   Bilabial approximant / B9 / (6 languages, 1.9%)

(2)   Labio-dental approximant / / (6 languages, 1.9%)