A mathematical analysis, called multidimensional scaling, was applied to two sets of data on metastases from head and neck cancers. The analysis places lymph nodes in an imaginary space (here termed a 'nodalgram') such that distances between nodal clusters are proportional to differences in the occurrence of tumor metastases. Two-dimensional scaling explains 93% of the variance in clinical data from 1008 patients and 91% of the variance in pathological data from a different set of 415 patients. These independent results are closely correlated, and both explain vastly more of the data than expected by chance. The dimensions of the nodalgram are not related to anatomical coordinates of the nodes, suggesting that a representation of these structures in an abstract space explains the spread of cancers better than an analysis of spread in real space. Surprisingly, a plot of the extent to which metastases from different primary sites spread along the nodalgram dimensions is closely related to the normal anatomical locations of the primaries. This shows that although the nodalgram may be difficult to interpret, it must have some biological significance. In conclusion, multidimensional scaling successfully quantifies a pattern in the spread of head and neck cancers.