To clarify the mechanism of phenol toxicity, the radical-scavenging activity of 2- or 2,6-di-tert-butyl- and 2-methoxy-substituted phenols was investigated by combining two distinct approaches: first, the induction period method for methacrylate polymerization initiated by benzoyl peroxide or 2,2'-azobisisobutyronitrile; and secondly, 1,1'-diphenyl-2-picrylhydrazyl (DPPH) radical scavenging test. The homolytic bond dissociation enthalpy (BDE) and ionization potential (IP(koopman)) were calculated by the DFT/B3LYP method. The cytotoxicity was investigated using tumor cells (human submandibular gland carcinoma cells, HSG; human promyelocytic leukemia cells, HL-60) and primary cells (human gingival fibroblasts, HGF; human periodontal ligament fibroblasts, HPLF; human pulp fibroblasts, HPF) derived from oral tissues. The cytotoxicity between tumor and primary cells was similar, except for eugenol dimer showing less toxicity for primary cells. The DPPH assay was not useful for tert-butylphenols due to their steric hindrance. For both HSG and HGF cells, a linear relationship was found between 50% cytotoxic concentration (CC(50)) and inhibition rate constant (k(inh)), but not BDE, IP, or logP. The acceptable quantitative structure-activity relationships (QSAR) obtained for cytotoxicity vs. k(inh) suggested that the cytotoxicity of phenols may be dependent on radical reactions. The cytotoxicity of vanillin and 3,5-di-tert-butyl-4-hydroxy-benzaldehyde with large k(inh) values, weak antioxidants was markedly less than that of 2,6-di-tert-butyl-4-methylphenol, 2,4,6-tri-tert-butylphenol and curcumin with small k(inh) values, potent antioxidants.