The palatal glide /j/ in Persian phonology

Glides have a vague status in feature-based models of phonology. Some classify glides with vowels, and others believe glides have characteristics that differentiate them from vowels and consonants. The present research aims at investigating the acoustic properties of the Persian palatal glide /j/. To this end, a number of words were randomly selected to include /j/ in three different phonological positions, namely word-initial, word-medial and word-final. Acoustic analyses were conducted to identify the spectral properties of the glide. The analyses included measuring F1, F2 and F3 as well as their bandwidths, overall intensity and the intensity between 0to 1000 Hz and 1000 to 5000 Hz. The results suggested that both the formant frequencies and the intensity of the mid and higher formant frequencies are significantly different between the glide /j/ and the neighboring vowels. The results were interpreted to indicate that phonation type is the most important articulatory feature that differentiates between /j/ and the surrounding vowels.

Extended Abstract
The idea that glides occupy a phonological category of their own, especially as distinct from the related high vowels /i/ and /u/, has not been accepted by all researchers.  Some prefer the term “semivowel” to that of “glide”, labeling them as vowel-like segments that only function like consonants (Ladefoged & Maddieson, 1996). Others deny any difference between glides and vowels, other than their relative positions in the syllable. For example, Selkirk (1984) has argued that there is no need for a feature distinction between the glides /j, w/ and the high vowels /i, u/, since they can be differentiated instead by syllable theory alone.
Stevens (1998) defines glides as “a class of consonants produced with a constriction that is not sufficiently narrow to cause a significant average pressure drop across the constriction during normal voicing”. The lack of a vocal tract closure producing a significant pressure drop is a clear distinction between glides and other consonants; its acoustic correlate is a lack of abrupt discontinuity in the acoustic signal (Stevens, 2002).  The dividing line between glides and vowels (especially the closely related high vowels), however, has heretofore been less clearly defined.  Chomsky & Halle (1968) suggest that the [-vocalic] feature that differentiates glides from vowels is defined by a constriction that is greater in degree than that for a high vowel; but the threshold boundary required to create this category distinction along the continuum of constriction degrees has not been established in terms of articulation and acoustics. Previous acoustic analyses have found several potential acoustic correlates of the [-vocalic] feature that differentiates glides from related vowels: (1) Glides, inhabiting the syllable boundaries, should have a weaker intensity than the vowels at the syllable nuclei (Stevens, 1998); RMS amplitude (ARMS) provides a quantitative measure of this intensity relationship (Hon Hunt, 2009); (2) The narrow constriction in the front part of the vocal tract for a glide has the effect of decreasing the frequency of the first formant peak (F1) relative to that of a vowel (Hon Hunt, 2009); (3) The bandwidth of the first formant (B1) is expected to be larger for glides than for vowels, again because of the narrower constriction in the vocal tract for the glide segments (Stevens, 1998); (4)  a less sonorous sound segment (such as a glide) might exhibit more of a noise component in its acoustic signal than a more sonorous sound segment (such as a vowel) (Padgett, 2008); (5) he aerodynamic effects of the oral constriction on the glottal source may also have the effect of decreasing the fundamental frequency of phonation (F0) in a glide relative to that in an adjacent vowel. Acoustic modeling suggests that this effect will be most pronounced when F0 and F1 are close together.  Because F0 is strongly affected by prosodic considerations, this source effect may be variably present in different prosodic environments. (Stevens, 1998).
The present research aims at investigating the acoustic properties of the Persian palatal glide /j/. To this end, a number of words were randomly selected to include /j/ in three different phonological positions, namely word-initial, word-medial and word-final, so we could control for the effect of vocalic as well as prosodic context on the acoustic properties of the glide. The words were selected in such a manner to include the six vowels of the Persian language. The data were elicited by 9 native speakers of Persian, aged 20 to 30 years old. Recordings were made in a sound-attenuating chamber in the laboratory of linguistics at IKIU, with subjects seated with a fixed microphone approximately six inches from the lips. The subjects were prompted by text appearing on a computer monitor to read the phrases that were displayed on the screen.
All acoustic analyses were carried out in Praat. The Acoustic analyses included measuring F1, F2 and F3 as well as their bandwidths, overall intensity and the intensity between 0 to 1000 Hz and 1000 to 5000 Hz. The measurements were made on the steady state of /j/ and the preceding vowel.
The results suggested that both the formant frequencies and the intensity of the mid and higher formant frequencies are significantly different between the glide /j/ and the neighboring vowels. Indeed, the formant frequencies and the intensity of the mid and higher formant frequencies were the only two acoustic properties that systematically distinguished the glide /j/ from the neighboring vowels. The results further suggested that the magnitude of the narrowing of the oral cavity in transition from/i/ to /j/ was too small to excite changes in the first format frequency and bandwidth, but it sufficed to produce changes in phonation type, and thus intensity in the mid and higher formant frequencies. The results were interpreted to indicate that phonation type is the most important articulatory feature that differentiates between /j/ and the surrounding vowels.