Research ArticleAn experimental study of tongue body loops in V1-V2-V1 sequences
Introduction
In a VCV sequence the shortest movement between the targets of the tongue position would be covered by straight lines. However, as has been observed in many studies, this is not the case when the consonant is dorsal. In VCV sequences with a velar consonant like /k/ or /g/ the trajectory of a fixed point on the tongue dorsum in the midsagittal plane is approximately elliptical in shape, i.e., forming a loop. A loop is mainly occuring following a back vowel and typically shows two characteristics. Namely, a forward sliding of the tongue back along the palate during the velar closure and curved paths during the movements from the vowel towards the consonant and from the consonant towards the vowel. Since the early observations by Houde, 1967 both characteristics have been extensively studied because they differ from the goal-directed, reversible, straight-line movements of extremities like arms or legs (Morasso, 1981).
There have been multiple hypotheses on the causes of the sliding and of the curved paths, but none can consistently explain all observations. This underlines that speech kinematics are quite complex (Fuchs & Perrier, 2005). For the sliding, it was assumed that the air pressure during the velar closure pushes the tongue forwards, so that closure release happens more anterior in the oral cavity ([Houde, 1967], [Kent and Moll, 1972]). However, Hoole, Munhall, and Mooshammer, 1998 compared the loop patterns of VCV sequences when speakers change from an egressive to an ingressive airstream, finding that the intraoral air pressure cannot fully explain the forward sliding of the tongue during the velar closure. Furthermore, it was proposed that the forward sliding is actively controlled to enlarge the volume of the pharyngeal cavity so that phonation can be maintained ([Houde, 1967], [Ohala, 1983], [Coker, 1976]). However, this hypothesis has been ruled out by Mooshammer, Hoole, and Kühnert, 1995 who found that the forward sliding of the tongue is larger for voiceless /k/ than for /g/. The point here is that /k/ does not need cavity enlargement to maintain voicing. For the curved paths, Löfqvist and Gracco, 2002 proposed that they could be the result of an active control process based on cost minimization. However, Perrier, Payan, Zandipour, and Perkell, 2003 and Perrier and Fuchs, 2008 simulated loops with a 2D biomechanical tongue model, which plans only the (virtual) target sequence (e.g., [Flanagan et al., 1993], [Laboissière et al., 1996]) of tongue positions rather than its entire trajectory. They concluded that the active control of the entire trajectory does not seem to be necessary to explain the curved paths. Rather, the curved paths seem to be “due to the biomechanical properties of the tongue model, i.e., the passive tongue elasticity, the muscle arrangements within the tongue, and the force generation mechanism.” (p. 1593 Perrier et al., 2003) In this framework, the virtual target hypothesis ([Löfqvist and Gracco, 1997], [Löfqvist and Gracco, 2002]) suggests that the consonantal target position of the tongue can never be reached because it is located beyond the surface of the palate. However, in trying to approach the virtual target position the tongue generates the consonantal closure and slides along the palate. In line with this hypothesis, data by Brunner, Fuchs, and Perrier, 2011 indicate that both loop characteristics, the curved paths and the sliding, can be interpreted as one joint movement. The kinematic and temporal properties of this movement can be explained by the virtual target hypothesis and the biomechanical properties of the tongue. In a wider sense, the twin study by Weirich, Lancia, and Brunner, 2013 reinforced the general impact of biological and biomechanical properties on loops. Finally, Mooshammer et al., 1995 and Hoole et al., 1998 discussed muscular effects as a potential cause for loops.
The primary goal of the present study was to provide evidence in favor of muscular and biomechanical effects as a major underlying factor. To this end, we studied V1-V2-V1 sequences like, e.g., /i-o-i-o-…/, with no consonants involved. This design was chosen for two reasons: On the one hand, the confounding contact forces between the tongue and the vocal tract walls are typically much weaker than in VCV transitions. Consequently, the sliding is reduced and the curved paths come to light. On the other hand, there is no pressure build-up in the vocal tract. Hence, the aerodynamical influence on the curved paths is reduced, leaving only tongue tissue and muscle properties or active motor control as possible explanations for the observed trajectories. On this basis, we tested the following hypotheses:
1. The tongue body trajectories in V1-V2-V1 sequences form loops, whose sense of rotation and relative width vary systematically with the angle of their main movement line (Section 2, Section 3.1). The reason for this hypothesis is a corresponding observation in a pilot study with two subjects (Birkholz, Hoole, Kröger, & Neuschaefer-Rube, 2011). To examine this hypothesis, a systematic and extensive analysis of this observation for a larger number of subjects was conducted.
2. Muscular and biomechanical effects are a major cause for the observed loop patterns (Section 3.2). On this basis, three sub-hypotheses were formulated. First, the actively controlled speaking rate has no significant effect on the observed patterns. Second, the biological and biomechanical differences across genders have a significant effect on the observed patterns. The reason for this hypothesis is that there are anatomic differences in the oral versus pharyngeal portions of the vocal tract of adult men compared to women (Vorperian et al., 2009). Since the acoustic conduit is formed by its hard and soft tissues like, e.g., the muscles, muscular effects are expected to operate gender-dependently on the tongue body. Third, the limitation of jaw mobility by bite blocks has a significant effect on the observed patterns. The reason for this hypothesis is that vowel production can be attributed to the muscle activities of both the tongue and the jaw (Noiray, Iskarous, Bolanos, & Whalen, 2008). The limitation of jaw mobility is expected to affect the tongue-jaw synergy as well as the effective contractile effects of the extrinsic tongue muscles.
3. Differences in the timing of the extrinsic tongue muscles suffice to explain the observed loop patterns in V1-V2-V1 sequences (Section 4). This hypothesis was tested by means of model simulations. Compared to existing models like the one by Perrier et al., 2003, the proposed model has a higher level of abstraction and models the loop patterns on a macroscopic scale.
In general, an efficient model of loop generation is desirable for two reasons. First, Nam, Mooshammer, Iskarous, and Whalen, 2013 and Iskarous, Nam, and Whalen, 2010 showed that the simulation of loops improves the perceived naturalness of synthetic speech. Therefore, articulatory speech synthesis could benefit from becoming more realistic and natural-sounding. Second, a model of loop generation may provide a speech-related biomechanical characterization of the tongue. Up to now, the possibly confounding effect of biomechanics on the kinematics has rarely been considered in the interpretation of kinematic measurements (e.g., [Kelso et al., 1985], [Kollia et al., 1995], [Recasens and Espinosa, 2010], [Moll and Daniloff, 1971]).
Section snippets
Subjects, stimuli, instrumentation and procedure
The articulatory and acoustic data of eleven male and seven female German adults (19 to 39 years) without any known speech or hearing disorders were recorded in a sound proof and electromagnetically shielded room by means of EMA (AG501, Carstens Medizinelektronik).
Three sensors were attached midsagittally to the tongue. The most anterior sensor was placed at about from the tongue tip, and the most posterior one at the tongue body, about from the tongue tip sensor. The third tongue
Results
In order to quantitatively assess the pattern we carried out statistical analyses of and across all subjects and subsets. Particularly, the effects of gender, speaking rate and jaw mobility on the pattern were examined.
Modeling of results
In order to transfer the mathematical description of the pattern by and into a physical context, a macroscopic tongue movement model similar to the one by Birkholz et al., 2011 was designed. In particular, we derived the model parameters that optimally explain our measurements.
Summary and conclusions
We examined tongue body trajectories in V1-V2-V1 sequences and found that they form loops. This contradicts Löfqvist, 2011, who found that vowel-related movements of a tongue body fleshpoint follow almost straight lines and argued that the whole trajectory may be actively controlled by a minimization of effort. However, in his study fewer vowel pairs flanking a bilabial consonant were investigated in Japanese and Italian words as, e.g., /kami/ and /sa’remo/, respectively. The jaw movement towards
CRediT authorship contribution statement
Christian Thiele: Methodology, Validation, Formal analysis, Investigation, Visualization, Writing - original draft, Writing - review & editing. Christine Mooshammer: Resources, Writing - review & editing, Supervision. Malte Belz: Methodology, Formal analysis, Investigation, Writing - review & editing. Oxana Rasskazova: Investigation, Writing - review & editing. Peter Birkholz: Conceptualization, Methodology, Writing - review & editing, Supervision, Project administration, Funding acquisition.
Acknowledgements
We would like to thank Megumi Terada for assisting in the recording of the EMA data. This study was supported by the Deutsche Forschungsgemeinschaft (GZ: BI 1639/4-1).
References (42)
- et al.
Supralaryngeal control in korean velar stops
Journal of Phonetics
(2011) Organization of tongue articulation for vowels
Journal of Phonetics
(1996)- et al.
On loops
Journal of Phonetics
(1995) - et al.
An MRI analysis of the extrinsic tongue muscles during vowel production
Speech Communication
(2007) - et al.
Dynamics of vowel articulation
Language and Speech
(1982) - et al.
Electromyography of the tongue muscles during vowels in environment
Annual Bulletin RILP
(1988) - Bartoń, K. (2019). MuMIn: Multi-Model Inference.https://CRAN.R-project.org/package=MuMIn r package version...
- et al.
Fitting linear mixed-effects models using lme4
Journal of Statistical Software
(2015) - Birkholz, P., & Hoole, P. (2012). Intrinsic velocity differences of lip and jaw movements: preliminary results. In...
- et al.
Tongue body loops in vowel sequences
Model-based reproduction of articulatory trajectories for consonant-vowel sequences
IEEE Transactions on Audio, Speech, and Language Processing
A model of articulatory dynamics and control
Proceedings of the IEEE
Control of trajectory modifications in target-directed reaching
Journal of Motor Behavior
On the complex nature of speech kinematics
ZAS Papers in Linguistics
How does vowel context influence loops
Do airstream mechanisms influence tongue movement paths?
Phonetica
A study of tongue body motion during selected speech sounds
Perception of articulatory dynamics from acoustic signatures
The Journal of the Acoustical Society of America
A qualitative dynamic analysis of reiterant speech production: Phase portraits, kinematics, and dynamic modeling
The Journal of the Acoustical Society of America
Cinefluorographic analyses of selected lingual consonants
Journal of Speech and Hearing Research
Articulatory organization of mandibular, labial, and velar movements during speech
The Journal of the Acoustical Society of America
Cited by (4)
Velocity differences in laryngeal adduction and abduction gestures
2022, Journal of the Acoustical Society of AmericaVelocity Differences Between Velum Raising and Lowering Movements
2021, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)A framework of motoric complexity: An investigation in children with typical and impaired speech development
2020, Journal of Speech, Language, and Hearing Research