Although for a fully ionised plasma the main dimensionless parameters which control the plasma confinement are the normalised Larmor radius ρ*, plasma β, and normalised collisionality ν*, some experimental results have indicated that the density normalised to the Greenwald density limit Fgr (≡ nπa2/I ) may be a relevant parameter. To resolve this question identity experiments have been completed on tokamaks of different size in which first ρ*, β, ν* and then ρ*, β, Fgr are matched. The dimensionless confinement times ωcτE on each tokamak are then compared. Initial experiments on JET and DIII-D indicated that ρ*, β, and ν* were the key parameters, however one could not exclude ρ*, β, and Fgr within the errors of the measurement. The reason for this is that the difference in n* at fixed r*, b, Fgr and Fgr at fixed ρ*, β, ν* scales with minor radius a and since the size ratio of JET to DIII-D is only a factor of 1.6 the differences in the unmatched parameter and hence any effect on τE is small. This paper describes similarity experiments performed on JET and Alcator-CMOD, which have a size ratio of factor 4. The ρ*, β, ν* and ρ*, β, Fgr matches were achieved by performing an ELMy H-mode ν* scan on JET in the Alcator-CMOD geometry with ρ*, β, and q at the same value as an Alcator-CMOD pulse. The scan included separate pulses where both ν* and Fgr were matched in both machines. The ratio of the normalised global confinement for the two machines (ωcτE (JET)/ωcτE (CMOD)) for the ρ*, β, ν* match was 1.08, for the ρ*, β, Fgr match the ratio was 2.65. This indicates that ν* is the more relevant parameter for confinement scaling.