In the present study, 129 of the 172 RFLP alleles detected had frequencies < 0.01. Under the infinite alleles model (KIMURA 1983, p. 208), the expected number of alleles in this frequency class is only 22.9. A similar excess is observed at the 10 polymorphic protein loci scored in Mork et aL’s (1985) study (11 alleles observed with fre- quencies < 0.01 but only 3.7 expected). Excess num- bers of rare alleles, significantly exceeding that ex- pected under the infinite alleles model, have been reported in previous electrophoretic surveys of protein polymorphism (see discussions in OHTA 1976; K”RA 1983). This feature of allele frequency distributions led OHTA( 1975) to propose that many electrophoretic al- leles may not be strictly neutral but subject to weak negative selection. Ohta’s model of slightly deleterious alleles appears unable to account for the large excess of rare RFLP alleles since these are unlikely to experi- ence the same level of negative selection as electropho- retic alleles.