Antimicrobial resistance in Staphylococcus aureus and coagulase-negative staphylococci (CNS) has become an increasing problem in hospital settings. These strains often reveal resistance to various drug classes in addition to beta-lactam resistance. Clindamycin, as a well-tolerated and cost-effective antimicrobial agent, is used widely in the treatment of intra-abdominal and skin-soft tissue infections. However, a major concern with regard to the use of clindamycin for staphylococcal infections is the possible presence of inducible resistance to clindamycin. The aim of this study was to determine the prevalence of constitutive and inducible clindamycin resistance among nosocomially acquired S. aureus and CNS strains. A total of 375 staphylococcal isolates were tested for clindamycin and erythromycin by the disk diffusion induction test (D-test) according to CLSI criteria. Oxacillin disks were used for the detection of methicillin resistance. The isolates resistant to erythromycin (ER-R) and susceptible to clindamycin (CL-S) with a D-shaped zone around clindamycin disk were considered positive for inducible resistance (D-test positive). Constitutive clindamycin resistance was found in 64.6% of methicillin-resistant S. aureus, 11.8% of methicillin-susceptible S.aureus, 53.8% of methicillin-resistant CNS and 4.9% of methicillin-susceptible CNS. Erythromycin-resistant clindamycin-susceptible (ER-R/CL-S) phenotype was found more frequent in CNS than in S. aureus strains. Among these strains, inducible clindamycin resistance was detected in 90% of S. aureus and 52.2% of CNS. In conclusion, to avoid using clindamycin when the antibiotic susceptibility test result showed an ER-R/CL-S phenotype may prevent a possible clindamycin treatment failure since inducible clindamycin resistance is frequent among such isolates, however, it may also deter the use of clindamycin in the treatment of infections that would likely respond to clindamycin therapy. Decision about clindamycin use for staphylococci with the ER-R/CL-S phenotype should be made according to the local prevalence data.