Antifreeze glycopeptide and peptides from the blood of polar fishes prevent the growth of ice crystals in water at temperatures down to approximately 1 degree C below freezing point, but do not appreciably influence the equilibrium freezing point. This freezing point hysteresis must be a disequilibrium effect, or it would violate Gibbs' phase rule, but the separate freezing and melting points are experimentally very definite: ice neither melts nor freezes perceptibly within the 'hysteresis gap', for periods of hours or days. We report here unusual crystal faces on ice crystals grown from solutions of very low concentrations of the anti-freeze glycopeptides and peptides. This is a clue to the mechanism of freezing inhibition, and it may be the basis of a simple, very sensitive test for antifreeze material. Very low concentrations of the antifreeze protein are also remarkably effective in preventing the recrystallization of ice.