This paper extends the concept of dispersion variance to the multivariate c
ase where the change of support affects dispersion covariances and the matr
ix of correlation between attributes. This leads to a concept of correlatio
n between attributes as a function of sample supports and size of the physi
cal domain. Decomposition of dispersion covariances into the spatial scales
of variability provides a tool for computing the contribution to variabili
ty from different spatial components. Coregionalized dispersion covariances
and elementary dispersion variances are defined for each multivariate spat
ial scale of variability This allows the computation of dispersion covarian
ces and correlation between attributes without integrating the cross-variog
rams. A correlation matrix, for a second-order stationary field with point
support and infinite domain, converges toward constant correlation coeffici
ents. The regionalized correlation coefficients for each spatial scale of v
ariability, and the cases where the intrinsic correlation hypothesis holds
are Sound independent of support and size of domain. This approach opens po
ssibilities for multivariate geostatistics with data taken at different sup
port. Two numerical examples from soil textural data demonstrate the change
of correlation matrix with the size of the domain. In general, correlation
between attributes is extended from the classic Pearson correlation coeffi
cient based on independent samples to a most general approach for dependent
samples taken with different support in a limited domain.