A note on Thomson's sampling theory for correlations among mental tests