Background The MRC 0–5 grading system is ubiquitously used to measure muscle strength. However, its criterion status is undermined by insensitivity of grades 4 and 5, which reduce its ability to detect change.

Aim To test the measurement properties of an expanded MRC strength scale.

Method We divided MRC grade 4 into three distinct grades, obtaining an 8-grade scale. We scored 53 muscles (25 pairs of limb muscles; three trunk muscles) in 250 people with multiple sclerosis (EDSS 0–9.0; median 6.0) on both original and expanded MRC grades. Data were analysed using the Rasch measurement model.

Results Rasch analysis revealed that in 52 (original 5-grade version) and 53 (expanded 8-grade version) muscles, the grading system did not work as intended. However, grades 0–3 were largely unendorsed, despite high median disability. Uniformly rescoring these four grades as one grade rectified the problem in 52 muscles (original version), but only in 15 on the expanded version. Individualised muscle grade scoring suggested different muscles operate best with different numbers of grades.

Conclusion In principle, a uniform muscle grading system is attractive. In practice, bespoke grading works better, possibly because of muscle diversity. Furthermore, MRC grades 1–3, while useful in some disorders, appear less relevant in MS.