Instances of EGG-based description of linguistic phenomena are rare. However, even in the early stages of electroglottography some experiments were conducted. The experiments take into account segmental but also suprasegmental settings related to voice quality in the sense of phonation type. Despite being easy to use and despite its non-invasive nature EGG has seldom been employed in studies of intonation.
Abberton | Esling | Dromey |
It is usually assumed that an EGG waveform does not change at all or changes only little across a voiced segment and that in particular it also does not indicate differences among vowels, because the signal is not influenced by articulation. However, the vocal tract coupling changes the aerodynamic conditions of the glottal flow, which induces changes in the vibration mode of the vocal folds (compare section 27). This can be related to a progression of the changes in the contact area. Nevertheless, no evidence of such effects has been published yet. Despite this fact, it is to be expected that the laryngeal settings will be more conspicuously present in the EGG signal than the phenomena related to articulation.
As stated in section 5.1 contrasts in phonatory settings are used rather seldom in the major European languages to distinguish phonemes. If at all, they are used rather on the paralinguistic layer of communication. Thus, it is to be expected that the linguistic use of phonation in German, to which this study is confined, will be limited to word stress and intonation11.
In the thesis of Abberton (1972) the movement of the larynx structure was investigated by means of the Gx component of the EGG signal during the production of the contrastive initial plosive triplets (plain, aspirated and tense) in Korean /p, ph, p*/. Evidence for a creaky voice vowel onset and and larynx movement was found for tense stops.
Esling (1991) tests the hypothesis that in Korean a distinctive breathy voice quality is associated with the lenis (versus aspirated or fortis) consonant series as a principal cue in identifying certain consonant-vowel (CV) sequences.
Dromey et al. (1992) contribute a study of the influence of growing speech intensity on glottal airflow and EGG signals. The study is not directly related to the linguistic research but intensity is often related to overall prominence (see section 15), so it may be regarded as a suprasegmental setting of the phonatory apparatus. Ten female speakers with no history of speech disorders produced a syllable chain conisting of multiple repetitions of /pa/ with intensities varied from 70 to 95 dB in 5 dB increments. The intensity was held within ±1 dB of the target. Measurements of the airflow Open Quotient and Speed Quotient (using two estimation methods for both quotients) were compared with the Open and Speed Quotients measured in the EGG domain. The Open Quotient was measured using the extremes (minimum and maximum) of the differentiated EGG (as described in section 7.2). These markers are also used to compute the skewness of the EGG pulse. In conclusion, the authors state that the EGG measures of the Open Quotient do not corroborate the glottal flow coefficients. As the intensity grows, the glottal airflow's OQ continuously decreases, whereas the EGG measure remains almost unchanged. The Speed Quotient (defined as the ratio of rise time and fall time) exhibits the general tendency to decrease (for low to moderate intensities), although the EGG values are considerably higher than those of the aerodynamic measures. This tendency to decrease is more exaggerated for EGG than for the airflow. They also claim that the speeds of abduction and adduction which are measured as peak values of the differentiated EGG waveform depend on voice intensity. They observed that the opening speed increases continuously across intensities, whereas the closing speed saturates at the high intensity levels. Unfortunately, this important interdependence was not statistically examined. At the highest intensity levels, the ratio of closing to opening speed decreases. The unsatisfactory results of the EGG measure of the OQ are caused by the problems already mentioned (the estimate of the true instant of the glottal closure). Nevertheless, the results of the experiment show that the variation in speech intensity influences the EGG waveform.
In the following chapter, we present data which shows that important linguistic and paralinguistic categories can indeed be investigated objectively on the basis of an appropiate parametrization of the EGG signal. The parametrization presented in the subsequent chapters will be applied to the characterization of linguistic stress, pitch accent (chapter III) and the complex voice quality distortion characteristics of neurological and organic speech impairments (chapter IV).
11the tenseness of German vowels was also taken into account in the conducted experiments.