Current methods of interpreting critical supercooling are discussed. In the method of Turnbull it is assumed that the experimental observation time matches the reciprocal of the steady-state nucleation frequency. This assumption is shown to be incompatible, for certain types of experiment, with the nucleation time lag predicted by the theory of part II. An alternative criterion of freezing which avoids this difficulty is that under the conditions of a threshold experiment the observation time is comparable with the nucleation time lag. This criterion may be used in conjunction with the theory of part II to interpret critical supercoolings without making separate assumptions about nucleation frequencies. Threshold nucleation rates, sizes of crystal nuclei, and liquid-solid interfacial free energies for alkali halides are calculated from the supercooling data of part I. Data for nuclei are examined in relation to the properties of the liquid and solid compounds in bulk form. Certain outstanding problems in the interpretation of results obtained in this and other studies are discussed with respect to present nucleation theory.