Specificity in the self-efficacy construct requires that instruments assess domain-specific behaviors. The objectives of the study were to (1) reexamine and shorten the original 57-item Cholesterol-Lowering Diet Self-Efficacy Scale (CLDSES), (2) estimate reliability and validity of the short form CLDSES (CLDSES-SF), (3) examine the dimensionality of the CLDSES-SF, (4) examine discriminant validity of the scale by its ability to differentiate between adherence and nonadherence to saturated fat intake guidelines, and (5) examine the CLDSES-SF's sensitivity and specificity. A sample of 238 patients being treated for hypercholesterolemia completed three instruments: the CLDSES, the Connor Diet Habit Survey (DHS) and a Three-Day Food Record. Internal consistency for the CLDSES-SF was .95. The inter-correlation coefficients among the seven sub-scales ranged from .28 to .62, and for the total to sub-scales .58 - .87, significant at p < .001. Evidence for criterion validity was shown in the correlations between the CLDSES-SF and the Connor DHS scores and between the CLDSES-SF and the Three-Day Food Record. Factor analysis showed a one-factor model that fits the data with a sufficient amount of total variance explained among the seven subscale items in each subscale, providing evidence of the uni-dimensionality of each subscale. Sensitivity (76%) and specificity (63%) for the CLDSES-SF were good with 88% positive predictive value for adherence. The shortened CLDSES retained its good psychometric properties and remains comprehensive in its assessment of self-efficacy for adherence to a cholesterol-lowering diet.