BACKGROUND: The purpose of this study was to examine the association between habitual walking and multiple or injurious falls among community-dwelling older adults. METHODS: Cross-sectional analysis was conducted on the data from 708 community-dwelling older adults, aged 60-91 years (72.3 +/- 6.6 yr, 233 men and 475 women). Prevalence of falls between walkers and non-walkers was compared separately by the number of risk factors (Groups R0, R1, R2, R3 and R4+). Logistic regression analysis was used to assess the association between habitual walking and falls separately by lower (R<3) and higher (R3+) risk groups. An interaction between habitual walking and risk of falling was examined in logistic regression analysis among all participants. RESULTS: In Groups R0, R1 and R2, prevalence of falls was lower in walkers than non-walkers; however, in Groups R3 and R4+, prevalence of falls was higher in walkers. Logistic regression analysis showed that habitual walking was significantly associated with fewer falls (Odds ratio (OR): 0.44, 95% confidence interval (CI): 0.20-0.97) among the lower risk group, but significantly associated with greater falls (OR: 4.61, 95% CI: 1.32-16.09) among the higher risk group. The interaction between habitual walking and higher risk of falling was significant (P < 0.05). CONCLUSION: Habitual walking seems to positively affect the prevention of multiple or injurious falls but only in community-dwelling older adults who have less than three risk factors.