Research Article
An experimental study of tongue body loops in V1-V2-V1 sequences

https://doi.org/10.1016/j.wocn.2020.100965Get rights and content

Highlights

  • Tongue body trajectories in V1-V2-V1 sequences form loops.

  • Loop shape varies systematically with angle of main movement line between V1 and V2.

  • Biomechanics suffice to generally explain loops and angle-dependence of their shape.

  • Macroscopic model of loop generation is proposed.

Abstract

Loops are well known as the elliptical tongue fleshpoint paths in vowel-consonant-vowel (VCV) sequences with a velar consonant. However, the discussion about the relative influences of aerodynamics, active motor control and biomechanics on loops is controversial. In order to characterize the biomechanical influence, the design of the current study was specifically based on V1-V2-V1 sequences with no consonants involved, recorded by means of electromagnetic articulography (EMA). In this context, we found evidence that the loop shape varies systematically with the angle of the main movement line between the vowels V1 and V2. Moreover, the regular angle-dependence of the loop shape is examined for eleven male and seven female German adults under different conditions for speaking rate (normal and slow) and for jaw mobility (large, small and no bite block). Model simulations suggest that both the loops and the angle-dependence of their shape can be attributed to biomechanical properties of the extrinsic tongue muscles.

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 0.5cm from the tongue tip, and the most posterior one at the tongue body, about 3.5cm from the tongue tip sensor. The third tongue

Results

In order to quantitatively assess the wrel(φ) pattern we carried out statistical analyses of |am| and arg(am) across all subjects and subsets. Particularly, the effects of gender, speaking rate and jaw mobility on the wrel(φ) pattern were examined.

Modeling of results

In order to transfer the mathematical description of the wrel(φ) pattern by |a1| and arg(a1) 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)

  • J. Brunner et al.

    Supralaryngeal control in korean velar stops

    Journal of Phonetics

    (2011)
  • K. Honda

    Organization of tongue articulation for vowels

    Journal of Phonetics

    (1996)
  • C. Mooshammer et al.

    On loops

    Journal of Phonetics

    (1995)
  • S. Takano et al.

    An MRI analysis of the extrinsic tongue muscles during vowel production

    Speech Communication

    (2007)
  • P.J. Alfonso et al.

    Dynamics of vowel articulation

    Language and Speech

    (1982)
  • T. Baer 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...
  • D. Bates 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...
  • P. Birkholz et al.

    Tongue body loops in vowel sequences

  • P. Birkholz et al.

    Model-based reproduction of articulatory trajectories for consonant-vowel sequences

    IEEE Transactions on Audio, Speech, and Language Processing

    (2011)
  • C.H. Coker

    A model of articulatory dynamics and control

    Proceedings of the IEEE

    (1976)
  • J.R. Flanagan et al.

    Control of trajectory modifications in target-directed reaching

    Journal of Motor Behavior

    (1993)
  • S. Fuchs et al.

    On the complex nature of speech kinematics

    ZAS Papers in Linguistics

    (2005)
  • C. Geng et al.

    How does vowel context influence loops

  • P. Hoole et al.

    Do airstream mechanisms influence tongue movement paths?

    Phonetica

    (1998)
  • R.A. Houde

    A study of tongue body motion during selected speech sounds

    (1967)
  • K. Iskarous et al.

    Perception of articulatory dynamics from acoustic signatures

    The Journal of the Acoustical Society of America

    (2010)
  • J.S. Kelso et al.

    A qualitative dynamic analysis of reiterant speech production: Phase portraits, kinematics, and dynamic modeling

    The Journal of the Acoustical Society of America

    (1985)
  • R.D. Kent et al.

    Cinefluorographic analyses of selected lingual consonants

    Journal of Speech and Hearing Research

    (1972)
  • H.B. Kollia et al.

    Articulatory organization of mandibular, labial, and velar movements during speech

    The Journal of the Acoustical Society of America

    (1995)
  • Cited by (4)

    View full text