Based on a simplified model of the all-α class of protein, all packing arrangements of α-helices were generated and assessed by both general and specific structural rules. The method was applied to myoglobin and parvalbumin, which were both ranked in the top 4% of folds under the general packing constraints. Incorporation of the restrictions implied by the EF-hand motifs of parvalbumin were sufficient to select the correct fold as one of two (equal scoring) possibilities. Myoglobin scored well under the general packing constraints and the addition of a single distance constraint, implied by haem binding, was sufficient to select the correct fold as one of several candidates. Incorporation of a score for complementary hydrophobic packing between helices further selected myoglobin as a unique fold but did not improve the ranking of parvalbumin. For both proteins, the α-helices were predicted from multiply aligned sequences using pattern-matching methods and no specific aspect of the known X-ray structures influenced this or the prediction of the correct folds. Although the method is currently of limited generality, its further applications and extension to a more detailed structural level are discussed.