Pulling a Spool:
Reproducing and Testing the Static-Friction Model
of a Rolling Spool
Group Project — General Physics I (BS103A)
DGIST, 2026 Spring
June 13, 2026
Abstract
We study the classic introductory-mechanics problem of a spool pulled by a ribbon at an
angle θ to a horizontal table, following Carl E. Mungan, “Pulling a Spool,” Phys. Teach. 61,
178–181 (2023). The paper argues, against the common student intuition, that the static
friction on a rolling spool need not oppose either the direction of motion or the horizontal
pull, and that there is a critical pulling angle θc = cos−1 (Ri /Ro ) at which the spool does
not roll at all. We reproduce the two equations of motion (force and torque), eliminate the
friction force to obtain the acceleration of the center of mass, and recover the three behavioral
regimes. We then verify the model with two experiments. In the first, we pulled an ordinary
spool at a range of angles and observed it wind in, stay put, or unwind in agreement with
the predicted regimes. In the second, we varied the radius at which the string effectively
acts (larger than, equal to, and smaller than the contact radius) and observed the spool
move away, spin in place, and roll toward us, respectively. Both experiments confirm that
the sense of rolling is set by the torque of the pull about the contact point.
Contents
1 Introduction and Motivation 2
2 Theoretical Model 2
2.1 Setup and the critical angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Equations of motion for θ < 90◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.3 Equations of motion for θ > 90◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.4 Direction of the friction force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.5 Answers to the five questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3 Experiment 1: Rolling Direction vs. Pulling Angle 4
3.1 Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.2 Results and comparison with the model . . . . . . . . . . . . . . . . . . . . . . . 4
4 Experiment 2: Effective Acting Radius vs. Contact Radius 5
4.1 Motivation and setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
4.2 Why the contact radius is the deciding length . . . . . . . . . . . . . . . . . . . . 5
4.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
5 Discussion 6
6 Conclusion 6
1
A Reproducibility 6
1 Introduction and Motivation
A simple model of kinetic friction is taught early in introductory physics: its magnitude is the
coefficient of kinetic friction times the normal force, and it points opposite to the sliding velocity.
Static friction is more subtle. It has no fixed magnitude; it adjusts itself, up to a maximum
value, so that the contact point does not slip, and its direction is whatever is needed to prevent
incipient slipping. Because of this, students often carry over a wrong intuition from kinetic
friction, assuming that static friction must oppose the motion or oppose the applied force.
A pulled spool is an ideal counterexample. Because the contact point between the spool and
the table does not slip, any friction acting there is necessarily static. The surprising behavior of
the spool—rolling toward the puller for some angles and away for others, with friction pointing
in directions that defy the naive guess—makes the limitations of the simple friction picture
vivid.
The original paper poses five questions to students about a gently pulled spool: must the
friction be nonzero; must it equal µs N ; must the velocity point forward; must the friction point
backward; and must the friction oppose the velocity? The central result is that the answer to all
five is no. Our project reproduces the reasoning that establishes this and tests it experimentally.
2 Theoretical Model
2.1 Setup and the critical angle
Consider a cylindrically symmetric spool of mass M and moment of inertia I about its central
axis O. The ribbon is wound on the inner axle of radius Ri , and the spool contacts the table
on its outer rim of radius Ro > Ri . The pulling angle θ is measured from the table: θ = 0◦
means the ribbon unwinds from the bottom of the axle (pulled horizontally), θ = 90◦ means it
is pulled straight up, and θ = 180◦ means it unwinds from the top of the axle. We define the
forward (+x) direction as the direction of the horizontal component of the pull, so the analysis
is split at θ = 90◦ , where the horizontal component vanishes.
We omit the vertical normal force N and gravity M g from the horizontal and torque analysis
(they balance vertically while the spool stays on the table), and we assume rolling without
slipping, so the friction f at the contact point P is static.
F⃗
Ri O
θ
S Ro
Figure 1: Geometry of the pulled spool. O is the center, P the contact point with the table,
and S the point where the ribbon leaves the inner axle. At the critical angle θc the line of action
of F⃗ passes through P , so F⃗ exerts no torque about P .
The geometry shows that there is a critical pulling angle
−1 Ri
θc = cos , (1)
Ro
2
at which the line of action of the pull passes through the contact point P . At that angle F⃗
produces no torque about P , and since the friction f also has no torque about P (it acts there),
there is no torque at all about P and the spool cannot begin to roll.
2.2 Equations of motion for θ < 90◦
For pulling angles 0◦ ≤ θ < 90◦ (Figs. 1 and the three small-, intermediate-, and large-angle
cases), Newton’s second law for the horizontal translation of the center of mass is
F cos θ − f = M ax , (2)
and the rotational form of Newton’s second law for clockwise rotation about the center O, using
the rolling constraint α = ax /Ro , is
Iax
Ro f − Ri F = . (3)
Ro
The torque term −Ri F appears because the ribbon acts on the inner axle at radius Ri , while
the friction acts at the outer radius Ro .
Eliminating the friction. Solving Eq. (2) for f = F cos θ − M ax and substituting into
Eq. (3):
Iax
Ro (F cos θ − M ax ) − Ri F = , (4)
Ro
Iax
Ro F cos θ − Ro M ax − Ri F = , (5)
Ro
I
F (Ro cos θ − Ri ) = ax + M Ro . (6)
Ro
Therefore
F Ro (Ro cos θ − Ri ) F Ro2 (cos θ − cos θc )
ax = = , (7)
I + M Ro2 I + M Ro2
where the second form uses cos θc = Ri /Ro from Eq. (1). The denominator I + M Ro2 is positive,
so the sign of ax is the sign of cos θ − cos θc :
ax > 0 (rolls forward, toward the puller) when 0◦ ≤ θ < θc ;
ax = 0 (does not roll) when θ = θc ;
ax < 0 (rolls backward, away from the puller) when θc < θ < 90◦ .
This is the key counterintuitive prediction: for intermediate angles a forward pull produces
backward acceleration.
2.3 Equations of motion for θ > 90◦
For 90◦ < θ ≤ 180◦ the spool is turned around so that S is above O and cos θ < 0. The signs
in the two equations flip accordingly:
−F cos θ + f = M ax , (8)
Iax
−Ro f + Ri F = . (9)
Ro
Eliminating f between Eqs. (8) and (9) gives the same expression as Eq. (7) but with overall
sign reversed,
F Ro2 (cos θc − cos θ)
ax = > 0, (10)
I + M Ro2
since cos θ < 0 < cos θc here. Hence the spool always rolls forward for 90◦ < θ ≤ 180◦ .
3
2.4 Direction of the friction force
Eliminating ax instead of f from either pair of equations yields a single expression valid for all
pulling angles,
F (M Ri Ro cos θ + (I − M Ri Ro ) cos θ − . . . ) F (M Ri Ro − I cos θ)
f= 2
−→ f = . (11)
I + M Ro I + M Ro2
Two consequences follow. First, because M Ri Ro /I < M Ro2 /I and | cos θ| ≤ 1, the magnitude
of the friction is always smaller than the applied force. Second, for θ < 90◦ we have cos θ > 0
and the numerator is positive, so f points backward throughout Figs. 2–4 of the original paper.
But for θ > 90◦ , cos θ < 0 and the sign of f depends on the relative size of the two numerator
terms. The friction reverses (points forward) once I > M Ri Ro , defining a second special angle
θm with
M Ri Ro
cos θm = − , (12)
I
at which the friction is exactly zero. For an ordinary spool I ≈ M Ri2 < M Ri Ro , so θm does
not exist and the friction simply points forward for all θ > 90◦ —meaning, remarkably, that the
friction does not oppose the horizontal pull.
2.5 Answers to the five questions
The derivation answers all five framing questions in the negative:
1. Friction need not be nonzero — it is exactly zero at θm for a suitably mass-loaded spool.
2. Friction need not equal µs N — for gentle pulling it is below the slipping limit (Eq. (11)).
3. Velocity need not point forward — it points backward for θc < θ < 90◦ .
4. Friction need not point backward — it points forward for θ > 90◦ .
5. Friction need not oppose the velocity — in Fig. 4 they point the same way.
3 Experiment 1: Rolling Direction vs. Pulling Angle
3.1 Procedure
We took an ordinary spool with a ribbon wound on its inner axle and placed it at rest on a flat
horizontal table. We pulled the free end of the ribbon gently (to keep the spool rolling without
slipping) at a range of angles θ measured from the tabletop, starting from a shallow, nearly
horizontal pull and increasing the angle progressively. For each angle we recorded whether the
spool wound the ribbon in (rolled toward us, “winds up”), stayed in place, or unwound the
ribbon (rolled away, “unwinds”).
3.2 Results and comparison with the model
Table 1: Observed rolling behavior as a function of pulling angle, compared with the prediction
of Eq. (7).
Pulling angle range Predicted (ax ) Observed behavior
0◦ ≤ θ < θc forward (ax > 0) spool rolls toward us, ribbon winds in
θ = θc none (ax = 0) spool stays in place (cannot roll)
θc < θ < 90◦ backward (ax < 0) spool rolls away, ribbon unwinds
The observations matched the predicted regimes. At shallow angles the spool rolled toward
us and wound the ribbon up; as we steepened the pull there was a narrow band of angles near
4
the critical value where the spool refused to roll and the ribbon merely slid against the axle
without translating the spool; and at still steeper (but sub-vertical) angles the spool reversed
and rolled away from us even though we were pulling it “forward.” This reversal is the most
striking confirmation of the model: a forward pull producing backward motion. The transition
occurs near θc = cos−1 (Ri /Ro ), so a spool with a thin axle (small Ri /Ro ) has a critical angle
close to 90◦ , while a spool with a thick axle reverses at a shallower angle.
4 Experiment 2: Effective Acting Radius vs. Contact Radius
4.1 Motivation and setup
Our second experiment was inspired by the demonstration in the referenced video (https://
[Link]/wWKfDffm3jU). Instead of varying the pulling angle, we fixed a low, nearly horizontal
pull and varied the radius at which the string effectively acts on the spool. We attached the
connection point of the string to the spool at three different radii relative to the contact radius
Ro (the outer rim that touches the table):
1. the acting radius larger than the contact radius,
2. the acting radius equal to the contact radius, and
3. the acting radius smaller than the contact radius.
This isolates the geometric quantity that governs the direction of rolling: the torque of the pull
about the contact point P .
4.2 Why the contact radius is the deciding length
The cleanest way to predict the rolling direction is to take torques about the contact point P ,
because the friction f acts at P and contributes no torque there. For a (nearly) horizontal pull,
let r be the perpendicular distance from P to the line of action of the string—equivalently, the
height above the table at which the string leaves the spool. The torque of the pull about P is
τP = F (r − Ro ) in the sense that determines forward/backward rolling:
if the string acts below the contact radius (r < Ro ): the torque rolls the spool toward the
puller;
if the string acts at the contact radius (r = Ro ): the pull’s line passes through P , the
torque about P vanishes, and the spool cannot roll—it can only spin in place if pulled
hard;
if the string acts above the contact radius (r > Ro ): the torque rolls the spool away from
the puller.
This is the same logic that produced the critical angle in Eq. (1): there, varying θ moved the
line of action across P ; here, varying the acting radius does the same thing directly.
4.3 Results
Table 2: Observed behavior as a function of the acting radius relative to the contact radius Ro .
Acting radius Torque about P Observed behavior
larger than Ro rolls away spool moved away from us (rolled backward)
equal to Ro zero torque spool did not roll; it rotated/spun in place
smaller than Ro rolls toward spool was dragged toward us (rolled forward)
The results in Table 2 match the torque prediction exactly. When the device was attached
at a radius larger than Ro , the spool rolled away from us; when it was attached exactly at
Ro , the line of action of the string passed through the contact point, no torque acted about P ,
5
and the spool merely spun in place without translating; and when it was attached at a radius
smaller than Ro , the spool was pulled toward us as expected. This experiment is the “radius”
analogue of the “angle” experiment, and the two together show that both knobs—the pulling
angle and the acting radius—work through the same underlying quantity, the torque about the
contact point.
5 Discussion
The two experiments probe the same physics from complementary directions. The governing
principle is that the friction at the contact point has no torque about that point, so the rolling
direction is set entirely by the torque of the pull about P . Whether one varies the pulling angle
(Experiment 1) or the radius at which the string acts (Experiment 2), the spool rolls toward
the puller when the pull’s line of action passes below P , refuses to roll when the line passes
through P , and rolls away when the line passes above P .
A few practical sources of imperfection are worth noting. Near the critical configuration the
behavior is very sensitive: a small change in angle (or a slight slipping of the string against the
axle) flips the direction, which makes the “stays in place” case hard to hold steady. Pulling too
hard breaks the rolling-without-slipping assumption, after which the spool slides and the simple
analysis no longer applies (the original paper’s appendix treats this regime, where increasing
the pull first causes slipping and eventually lifts the spool off the table). We also neglected
rolling friction and air drag, which is reasonable for a rigid spool on a hard table.
6 Conclusion
We reproduced the derivation of the acceleration of a pulled spool, recovering the critical angle
θc = cos−1 (Ri /Ro ) and the three rolling regimes, and we verified the central, counterintuitive
claim that static friction on a rolling spool need not oppose either the motion or the applied
pull. Our first experiment confirmed the forward / stationary / backward regimes as the pulling
angle was increased through θc . Our second experiment, varying the radius at which the string
acts relative to the contact radius, produced “away,” “spin in place,” and “toward” behaviors
in exact agreement with a torque analysis about the contact point. Together the experiments
demonstrate that the sense of rolling of a pulled spool is governed by the torque of the pull
about its point of contact with the table.
A Reproducibility
All quantities used in the analysis (Ri , Ro , M , I, θ) are directly measurable. The predictions
are evaluated from Eqs. (1), (7), and (11); no fitting parameters are involved. The experimental
procedures in Sections 3 and 4 can be repeated with any spool that has a clearly distinct inner
axle and outer rim and a non-slip ribbon.
References
[1] C. E. Mungan, “Pulling a Spool,” The Physics Teacher 61, 178–181 (2023). [Link]
org/10.1119/5.0042450
[2] P. G. Hewitt, “December Figuring Physics answer,” Phys. Teach. 59, 67 (2021).
[3] R. Cross, “Comment on ‘Pulling a spool’,” Phys. Teach. 61, 425–426 (2023). [Link]
org/10.1119/5.0148218
6
[4] Demonstration video of a spool pulled at different effective radii. [Link]
wWKfDffm3jU