Language Log

« previous post | next post »

The sound of a cat's purr is a familiar one:

But this familiar sound raises at least two interesting biophysical questions.

In the first place, cats purr both while breathing out and breathing in, while most people can only produce voiced sounds (= laryngeal oscillations) while breathing out. What do cats have or do that we don't have or do?

In the second place, cats' purring is much lower in pitch than we'd expect given their size.

Human children have higher voices than human adults, and adult females generally have high voices than human males, due to differences in the length and mass of the vocal folds.

Wikipedia tells us that

Some traditional Russian religious music calls for A2 (110 Hz) drone singing, which is doubled by A1 (55 Hz) in the rare occasion that a choir includes exceptionally gifted singers who can produce this very low human voice pitch.

But a cat's purr generally has a pitch around 25 Hz — less than half the frequency of those "very low" human voices. Here's an image highlighting one cycle of the purr exhibited above:

So how do they do it? The traditional view was that cats must modulate the tension of their laryngeal muscles with low-frequency control signals from their brains, since their "voice boxes" are way too small to reach those hyper basso-profundo pitches.

A paper published just a few days ago gives a different answer — instead of low-frequency muscle twitches, cats have got some extra "connective tissue masses" in their vocal folds, changing the physical properties of their vocal instrument in relevant ways.

The new paper is Christian Herbst et al., "Domestic cat larynges can produce purring frequencies without neural input", Current Biology 10/3/2023. The abstract:

Most mammals produce vocal sounds according to the myoelastic-aerodynamic (MEAD) principle, through self-sustaining oscillation of laryngeal tissues. In contrast, cats have long been believed to produce their low-frequency purr vocalizations through a radically different mechanism involving active muscle contractions (AMC), where neurally driven electromyographic burst patterns (typically at 20–30 Hz) cause the intrinsic laryngeal muscles to actively modulate the respiratory airflow. Direct empirical evidence for this AMC mechanism is sparse. Here, the fundamental frequency (fo) ranges of eight domestic cats (Felis silvestris catus) were investigated in an excised larynx setup, to test the prediction of the AMC hypothesis that vibration should be impossible without neuromuscular activity, and thus unattainable in excised larynx setups, which are based on MEAD principles. Surprisingly, all eight excised larynges produced self-sustained oscillations at typical cat purring rates. Histological analysis of cat larynges revealed the presence of connective tissue masses, up to 4 mm in diameter, embedded in the vocal fold. This vocal fold specialization appears to allow the unusually low fo values observed in purring. While our data do not fully reject the AMC hypothesis for purring, they show that cat larynges can easily produce sounds in the purr regime with fundamental frequencies of 25 to 30 Hz without neural input or muscular contraction. This strongly suggests that the physical and physiological basis of cat purring involves the same MEAD-based mechanisms as other cat vocalizations (e.g., meows) and most other vertebrate vocalizations but is potentially augmented by AMC.

They don't say whether their set-up exhibited purring in both directions of air flow — perhaps someone will follow up on that point. For more on the context, and on the question of why cats purr, see Jennifer Ouellette, "We now know how cats purr—why they purr is still up for debate", Ars Technica 10/5/2023.

The Herbst et al. paper reminded me of the experiments that I used to watch Tom Baer carrying out 50 years ago, when I was a grad student. Tom's apparatus, set up in a reconfigured closet down the hall from the linguistics department's section of Building 20, involved an incredibly intricate and precise system centered around a freshly-excised dog larynx:

And that diagram doesn't show the multiple fine wires, connected to the attachment points of (the many intrinsic and extrinsic) laryngeal muscles, running over geometrically appropriate pulleys, and linked to small weights providing an array of precisely defined forces.

What Tom found was interestingly complicated, in ways that help us to understand why reducing vocal oscillations to one-parameter "fundamental frequency" is an over-simplification on the production side as well as the perception side.

Here's the abstract from his 1975 thesis, "Investigation of phonation using excised larynxes":

Excised canine larynxes were used to study the mechanism of phonation. Apparatus and procedures were developed for producing laryngeal configuration appropriate for phonation and for supplying airflow through the glottis. Both flow-regulator and pressure-regulator air sources were used. The pseudo-subglottal tract contained a window permitting a direct view of the vocal folds from the inferior as well as the superior aspect. Particles placed on the vocal folds served as discrete visual reference points. A stereoscopic dissecting microscope was used to observe the vibrations from the superior, inferior, or oblique aspect under stroboscopic illumination. A system utilizing the microscope was developed for determining the absolute frontal-plane position of any point that could be observed on the vocal fold surfaces. This system was used with adjustable-phase synchronous stroboscopic illumination to measure the trajectories of particles in the frontal plane when the larvnxes produced steady phonation.

Measurements on three different particles during one run served as quantitative reference points to reconstruct the shapes of the glottis at eighth-cycle phase increments.

In addition to measurements of glottal shape during phonation, observations of gross responses to changes in subglottal pressure and tissue properties were made. Chest register and falsetto could be produced. Fundamental. frequency in chest register varied at rates of 5 to 7 Hz./cm. H20 as subglottal pressure was changed. Substantial changes in fundamental frequency were produced by changes in longitudinal tension. Some aspects of the performance were sensitive to desiccation of surface tissues. The ability to produce falsetto was more dramatically impaired than the ability to produce chest voice when the tissues became desiccated.

Vibration patterns of the vocal folds in chest register were complex. Individual particles were sometimes on the superior vocal fold surfaces and. on the glottal walls during different portions of a glottal cycle. The overall shape of individual particle trajectories was roughly elliptical, but the ellipses contained perturbations which sometimes formed secondary loops. Particles always traversed the ellipses clockwise in the coordinate system with the lateral direction to the right. For particles on the superomedial parts of the folds, trajectories had large vertical and horizontal components. More lateral particles had predominately vertical trajectories and more inferior particles had predominately horizontal trajectories.

Some aspects of the vibration patterns were interpreted as travelling wave phenomena. Apparent wave velocities were determined by direct measurement and were inferred from measurements of particle trajectories at different locations along the vocal-fold surfaces. Glottal closure exhibited wavelike properties.

In addition to travelling wave vibrations on the surface membranes, measured glottal shapes suggested string vibrations of the vocal ligament. A mechanism cf phonation in which aerodynamically-driven waves on the surface membranes drive vibrations of the ligament was suggested. Particle trajectories were measured as the surface tissues of the vocal folds became desiccated. The results of these experiments and analysis of glottal shapes measured during steady phonation suggested that vertical components or forces on the vocal folds were important for phonation. These experiments showed that particle trajectories could be considered as vibrations about a static operating point.

When the vocalis muscles were removed, larynxes could produce nearly normal chest register phonation but not falsetto. Th!s result has implications with regard to the phonatory runction of the vocalis muscle.

Preliminary measurements of the static behavior of preparations for different values of subglottal pressure were made. These results and the results of dynamic measurements were used to directly test a two-mass model of the larynx with near mechanical elements. The aerodynamic aspects of the model were found to be adequate to account for measured pressure-flow relationships. The mechanical aspects were found to be inadequate to account for some experimental results.

For more on the relevant physics, see Ingo Titze, "Comments on the myoelastic-aerodynamic theory of phonation", 1980. Ingo gives this sketch of the history:

Although the myoelastic-aerodynamic theory of phonation, originating with Mueller (1848) and formulated explicitly by van den Berg (1958) in the first volume of this journal, has in principle been widely accepted for more than two decades, there remains a lack of clarity in its application to speech and voice science, specifically with regard to aerodynamic versus myoelastic contributions to stress and intonation. The myoelastic-aerodynamic theory of phonation maintains that vocal fold oscillation is determined by an interaction between aerodynamic stresses applied to the free surfaces of the vocal folds and myoelastic restoring forces generated within the tissues. This biomechanical system is self-oscillating, in other words, the frequency of the mechanical vibration is not determined by periodic neural impulses, or any other periodic input imposed mechanically or aerodynamically upon the system.

 

October 9, 2023 @ 5:22 am · Filed by Mark Liberman under Animal communication, Phonetics and phonology

Permalink