Why a Violin Bow Does Not Simply Drag the String
The Bow Lets Go
A bow travels one way. A violin string has to come back.
That sounds too obvious to deserve a question, until you try to put the two motions together. If the bow simply grabs the string and drags it, where does the return fit? And if it lets go, how can it keep a note going?
The familiar answer is friction. The more interesting answer is that friction has to cooperate with a traveling disturbance—and with a string that can twist as well as move sideways. A sustained note is not the result of holding on continuously.
The corner makes the trip
Jim Woodhouse's account of Helmholtz motion begins by distinguishing the shape we might imagine from the motion the ideal model describes. The blurred, lens-shaped envelope around a vibrating string can suggest a smooth curve swinging back and forth. But in the ideal bowed-string motion, the string at any instant consists of two straight portions joined at a corner. That corner travels along the string and reflects at the ends.
The corner is a moving change in slope, not a particular bit of string making the whole journey.
At the bow, its arrival separates two parts of the cycle. While the corner makes the longer trip toward the finger-side end and back, the string sticks to the bow and moves with it. During the shorter trip toward the bridge and back, the string slips rapidly in the other direction. The bow keeps traveling forward through both phases. Wave travel sets the approximate repetition time. [Woodhouse, §9.1, discussion of Fig. 2.]
A little arithmetic makes the return less mysterious. For a point-contact example that ignores twisting, imagine an ideal 200 Hz note, with the bow one tenth of the vibrating length from the bridge. Each cycle lasts five milliseconds. In this model, the string goes with the bow for 4.5 milliseconds and slips back for the remaining half-millisecond.
Give the bow an assumed speed of 0.1 meters per second. During the sticking part, the string moves 0.45 millimeters in that direction. To return through the same distance in one ninth of the time, it must move at 0.9 meters per second in the opposite direction during slipping.
Those are calculated numbers for an ideal example, not measurements of somebody's playing. But they expose the arrangement: a long, slower movement one way and a short, faster movement back. The string can return without requiring the bow to reverse every cycle.
The bow's job is not to carry the string around its whole path. It supplies energy through a contact that repeatedly changes state.
More pressure is not the whole instruction
This makes bowing sound as if the right material should do everything automatically. Rosin helps make the contact grip and release, so why should drawing a bow be difficult?
Because a possible repeating motion is not the only motion available.
In Woodhouse's explanation of Schelleng's diagram, bow force has both a lower and an upper limit for maintaining Helmholtz motion at a given speed and position. Too little can produce extra slips within a cycle. Too much can take the string into nonperiodic motion. An ideal lossless string would not need a minimum sustaining force, but real motion loses energy; the bow must replace it without upsetting the intended sequence of sticking and slipping.
The limits also move when the bow moves along the string. In Schelleng's approximate treatment, taking the bow closer to the bridge requires greater force and makes the relative allowable range narrower. It is therefore possible to cross a boundary without changing force at all—just by letting the contact position wander. [Woodhouse, §9.3, discussion of Figs. 1–3.]
There is a useful catch. Being inside the diagram's permitted region does not guarantee the intended motion. Woodhouse describes measurements in which other repeating regimes interrupt that region. The experiment he discusses used a cello string and a rosin-coated Perspex rod rather than ordinary bowhair, and deliberately established Helmholtz motion before testing whether it survived.
The map is useful. It is not a universal pressure recipe, and maintaining a note is not the same problem as starting it.
Nor is a sharp corner and a simple grip-versus-slide rule the complete physics. Real corners round off; a ribbon of hair is not a point; rosin's behavior is more complicated than a force determined only by current sliding speed. Woodhouse explicitly warns about these simplifications. The sketch earns its place by explaining an organization, not by including everything.
A real string has somewhere to twist
The complication that most changed my picture was not a correction to the corner's sharpness. It was the location of the bow.
The bow touches the outside of the string, not its center. Its force can therefore twist the string as well as pull it sideways.
In a 2005 experiment, Eric Bavu, John Smith and Joe Wolfe measured both kinds of motion together. They did not measure a violin in performance. They stretched a wound bass-guitar string between rigid supports on a laboratory rig and bowed it with a violin bow. The displayed measurements were made by Bavu, an experienced violinist. Small attached wires and a coil, moving in a magnetic field, provided separate signals for transverse and twisting motion. [Bavu et al., pp. 242–243.]
That distinction between the outside and the center changes what “sticking” means. The sideways velocity of the surface point under the bow combines the motion of the centerline with the contribution from rotation. The surface point is what must move with the bow during sticking. The centerline alone need not tell the whole story.
Twisting was not simply negligible. In one reported condition, the peak surface-speed contribution from torsion was about a third of the string's transverse speed during the sticking phase. The sensors were near the bow rather than precisely at the contact; the authors relate the timing of those nearby motions. It is a result for that condition—not a claim that one third of a violin's sound, energy or movement always comes from twisting. [Bavu et al., Fig. 3 and p. 244.]
Yet the twisting motion did not march to an unrelated clock. Both measured motions repeated with the same overall period. Faster torsional oscillations occurred within that repeating pattern; the torsional spectrum contained harmonics of the transverse fundamental, with strong regions shaped by the string's torsional resonances. In their 40 Hz example, the torsional signal lacked the 40 Hz component itself, but its collection of higher harmonics still made a pattern that repeated every 25 milliseconds. [Bavu et al., p. 244.]
That is different from saying the string's free twisting resonance had become identical to its transverse fundamental. A preferred resonance and the full repetition period of a driven motion are not the same quantity.
The result suggests a better picture than “a clean note means the complications are too small to matter.” Here, a complication could be appreciable and still participate in a shared cycle.
Where the measurement stops
It would be satisfying to finish by saying the experiment showed the sideways and twisting fluctuations exactly cancel, leaving the surface perfectly attached to the bow during sticking.
It did not.
The authors reported smaller fluctuations when they combined the measured contributions, but their transverse channel attenuated high-frequency motion. That prevented a quantitative answer to the compensation question. They also could not measure cycle-to-cycle timing jitter at their 0.1-millisecond resolution. Not measurable at that resolution is not perfectly absent. [Bavu et al., pp. 243–244.]
They attributed the coordination to the much lower torsional Q—about 17, compared with about 900 for transverse motion in their rig. Q describes how lightly damped a resonance is; lower Q means it loses a larger share of its stored energy per oscillation. That is their physical explanation, not an experiment that independently changed damping to prove it was the sole cause. The rigidly supported bass string also cannot stand in for every string on a complete violin.
These limits do not undo the finding. They tell us what the finding is: appreciable twisting motion, coordinated with the transverse cycle, observed through an apparatus with real constraints.
I started with a bow that dragged a string. I ended with a contact that supplies energy, releases, catches again, and acts on a surface whose movement is richer than its centerline alone suggests.
The return is not an interruption of how the note works. It is part of how the note works.
And the next question sits right at its beginning: how does a player find that coordinated cycle quickly enough that we hear an intentional attack rather than the search for one?
Sources and limits
Jim Woodhouse, Euphonics, §9.1 and §9.3: complete main texts, references and captions read. Two sections are one specialist exposition, not two independent studies. Original historical works and linked technical supplements were not read.
Eric Bavu, John Smith and Joe Wolfe, “Torsional Waves in a Bowed String,” Acta Acustica united with Acustica 91 (2005), 241–246: complete main text, methods, captions and references read from the author PDF, with OCR used to recover missing symbols in text extraction. This separate team's measurements share an older theoretical lineage with Woodhouse's account; they are not an independent replication of every claim in it. The ideal-cycle arithmetic above was checked computationally. No figure-image inspection, listening, playing, movie viewing, raw-data reanalysis or new physical experiment is claimed.