We review evidence concerning the basis for tissue segregation during embryonic development. This compartmentalization is shown to be an immiscibility phenomenon caused by changes in the strengths of adhesions between mobile cells which accompany their differentiation and generate interfacial tensions at cell population boundaries. The mobile cells exchange neighbors in response to these adhesion-generated forces which impel the system toward the configuration of maximal binding. Cadherins dominate these intercellular adhesions, but integrin-fibronectin-based adhesions also contribute to them as well as to cell-matrix adhesions. At the interface between two segregating cell populations are three kinds of cell-cell interfaces: a-a, b-b and a-b. Tissue immiscibility (segregation) results when the cross-adhesion is weaker than the mean value of the two kinds of self-adhesions, does not require (although it permits) qualitative changes in cell adhesion molecules and is easily generated even by moderate changes in the quantities of adhesion molecules on the cell surfaces. All type I and II cadherins tested cross-adhere, in most cases with strengths close to those of their self-adhesions. Is malignant invasion a process of cell segregation in reverse, in which the cross-adhesion between cancer cells and host tissue components is strong relative to their self-adhesions? We review evidence for cadherin involvement in breast, prostate and brain cancers. Despite evidence that N-cadherin enhances the invasiveness of certain cancer cells, we have found that increasing the expression not only of functional E-cadherin but also of P- or N-cadherin restrains the spreading of other malignant cell lines over (and through) a reconstituted extracellular matrix.