The best-fitting two-dimensional plane within the three-dimensional space of spiral galaxy disc observables (rotational velocity vrot, central disc surface brightness μ0=-2.5logI0 and disc scalelength h) has been constructed. Applying the three-dimensional bisector method of regression analysis to a sample of ~100 spiral galaxy discs that span more than 4magarcsec-2 in central disc surface brightness yields vrotproptoI0.50pm 0.050,h0.77pm 0.07 (B band) and vrotproptoI0.43pm 0.040,h0.69pm 0.07 (R band). Contrary to popular belief, these results suggest that in the B band, the dynamical mass-to-light ratio (within four disc scalelengths) is largely independent of the surface brightness, varying as I0.00pm 0.100,h0.54pm 0.14. Consistent results were obtained when the range of the analysis was truncated by excluding the low-surface-brightness galaxies. Previous claims that M/LBvaries withI-1/20,Bare shown to be misleading and/or caused by galaxy selection effects - not all low-surface-brightness disc galaxies are dark matter dominated. The situation is, however, different in the near-infrared where LK'~v4 and M/LK' is shown to vary as I-1/20,Kprime. Theoretical studies of spiral galaxy discs should therefore not assume a constant M/L ratio within any given passband. The B-band dynamical mass-to-light ratio (within four disc scalelengths) has no obvious correlation with (B-R) disc colour, while in the K' band it varies as -1.25+/-0.28(B-R). Combining the present observational data with recent galaxy model predictions implies that the logarithm of the stellar-to-dynamical mass ratio is not a constant value, but increases as discs become redder, varying as 1.70+/-0.28(B-R).