Loading...

When one arm stops swinging

An ordinary walk, carrying a phone
MOTION

Accuracy

89.7%

Step Count Accuracy, Hands Full

walking while carrying something in the hand, against an observer’s hand tally count

Reference standard

Hand tally counter

Sample size

10 participants

Carrying a phone takes an arm out of the walk. The Ultrahuman Ring’s typical error widens from 8.2% on an ordinary outdoor walk to 10.3% with something in the hand — a little over two percentage points, for an arm that has stopped swinging.

80%85%90%95%100%91.8%Hands freeordinary outdoor walks89.7%Carrying somethingwalks with a phone in handstep count accuracy on a single walk (%)
89.7%accuracy carrying something in the hand
91.8%outdoor walks, hands free
2.1 ptsthe gap between the two conditions
1,010steps recorded per 1,000 taken — the one condition where it counts a few too many
Figure 1 — Two conditions from the same study, each measured against its own hand count. The dashed line is the wider end of the range the expert consensus on step-count validation treats as acceptable for general use: an error of 10% to 15%, drawn here at the 15% end.

Everything a step counter on the hand knows about walking, it learns from the arm. A stride sends a swing through the shoulder, the elbow and the wrist, and that swing is the rhythm the sensor locks onto. Put a phone in that hand and the swing stops. The hand is now held forward and steady, and the only motion left reaching it is what travels up the body from the feet — a smaller, muddier version of the same signal.

This is the condition the published literature treats as the hard case for anything worn on the arm, and the striking thing is that it does not resolve to one number. In the study cited below, trackers worn at the wrist, chest and pocket were significantly less accurate walking with a rolling walker — but pushing a shopping trolley produced no significant difference at all against a video count. What decides the damage is how rigidly the arm is held, not whether the hands happen to be full — and a phone in one hand is the mildest version there is.

That is the version this card measures: one hand occupied, intermittently, on an otherwise ordinary walk. Set against a comparable set of ordinary walks, it costs 2.1 percentage points, and two in every three of those walks still land within a tenth of the hand count, against six in every seven of the hands-free ones.