Volume 44, Issue 4

We present two magnetic methods for measuring articulatory activities. Unlike previous methods, the magnetic fields are uniform. Three alternating magnetic fields induce voltages in two detectors, a magnetic potentiometer and a flat coil. Thus we are able to measure jaw and tongue movements during speech production. These methods are biologically safe. The miniaturized detectors do not disturb the speech movements. Some preliminary results are presented. These methods can also be applied for other kinds of vectorial distance measurement.

Prior electromyographic (EMG) research reveals a lack of agreement as to the role of the palatoglossus muscle in speech. Some reports have concluded that it bears primary responsibility for lowering the velum and that it actively controls velar lowering on nasal sounds in speech, whereas others have concluded that it acts to assist in the tongue-body movements associated with the production of back vowels and linguavelar articulations. To clarify these conflicting claims, EMG recordings were obtained from the palatoglossus (as well as the levator palatini) muscle of a native speaker of Hindi who produced CVC nonsense and meaningful syllables containing a nasal or nonnasal vowel in a symmetric consonantal environment. The results showed the following: Palatoglossus activity for the central and back nasal and nonnasal vowels was moderately to considerably higher than for the front nasal and nonnasal vowels. The levels of palatoglossus activity for the back rounded nasal vowels and for the front nasal vowel /ε/ were comparable to those for the corresponding nonnasal vowels, while in all other vowels, the nasal vowels exhibited relatively higher levels of palatoglossus activity than their nonnasal counterparts. In all nonnasal vowels, the increase in palatoglossus activity occurred simultaneously with strong levator palatini activity. On the other hand, in nasal vowels, the increase in palatoglossus activity and the decrease in levator palatini activity were virtually synchronous for the front nasal vowels, while the increase in palatoglossus activity began much earlier than the decline in levator palatini activity for the central and back nasal vowels. This difference in temporal relationship between palatoglossus activation and levator suppression for different vowel types is important since it unambiguously supports the ‘gate-pull’ model (that is, active velar lowering) for the production of front nasal vowels whereas in the case of central and back vowels, nasal and nonnasal, the palatoglossus appears to be primarily involved in moving the tongue-body.

It has been commonly observed in the speech of English-speaking adults and children that vowels are longer when they precede voiced versus voiceless final obstruents. However, less is known about articulatory characteristics of such productions, particularly in the case of children. The present study investigated kinematic properties of children’s and adults’ productions of labial stop consonants in postvocalic position. Acoustic measures of the same segments were also obtained. It was observed that for all three groups of children and the adults, closing gestures into [p] occurred more rapidly than did closing gestures into [b]; however, there was a tendency for the difference between [p] and [b] closing velocities to increase with age. Possible relationships between acoustic and kinematic factors were also explored.

Past studies of declination make contradictory claims about the relationship between the length of an utterance and the height of its first accent peak (P1): Some find a large increase in P1 with increased length, others find little or no increase. A possible explanation for this contradiction is that large increases in P1 occur when the first major constituent of the sentence increases from one accented item to more than one, but not as a function of overall length alone. In an experiment, 1 of 2 speakers behaved in the way just outlined; data from the 2nd speaker supported the more general hypothesis that ‘metrical’ constituent structure is relevant to P1.

An attempt was made to see whether realizations of eight different pitch patterns can be classified automatically by subjecting fundamental frequency curves to statistical treatment using two kinds of discriminant functions. The materials consisted of 150 imitations of each of the eight prototypes (eight intonations of a short Polish phrase). Approximately 80% of the time the statistical procedures assigned the individual tokens to the appropriate classes. Most of the misassignments were due to wrong imitations. The discriminant functions also indicated similarities and dissimilarities among the pitch patterns. It is suggested that perceptually and linguistically, some differences between pitch patterns may not be quite categorical.