[Objective] This study investigated the applicability of a 3-m zigzag walk test for the prediction of falls and examined the relationships among fall history, the 3-m zigzag walk test, 10-m walk, and age. [Subjects] A total of 50 elderly individuals (23 males and 27 females) aged 65 and over, who were able to walk independently, were studied. [Methods] Four poles made of PET bottles were placed on a 3-m walkway in a straight line to create a zigzag course, and the time needed to walk it was measured. The best results on days 1 and 2 were adopted for the fall and no-fall groups, and intra-rater reproducibility was evaluated by calculating the intra-class correlation coefficient and performing the paired t-test. For comparison of the time needed to walk the zigzag between the 2 groups, the unpaired t-test was performed. The relationships between the times needed to walk the 3-m zigzag and 10 m and age were analyzed by calculating the correlation coefficient with fall history as the dependent variable, in multiple logistic regression analysis with the times needed to walk the 3-m zigzag and 10 m and age as independent variables. For the optimal classification of the fall and no-fall groups, cutoffs were calculated based on the ROC curve. [Results] The paired t-test results did not show differences between measurements, and the ICC was 0.97 in the fall, and 0.94 in the no-fall groups. The fall group needed significantly more time than the no-fall group to walk the 3-m zigzag. Further, the Pearson product-moment correlation coefficient revealed a significant correlation between the times needed to walk the 3-m zigzag and 10 m, while no correlation was observed between the time needed to walk the 3-m zigzag and age (r=0.225). The time needed to walk the 3-m zigzag was extracted as a factor associated with fall history in multiple logistic regression analysis, with an odds ratio of 0.377. Its significance as a variable was p<0.01. In the Hosmer-Lemeshow test of the study model, the rate of discrimination between the predicted and actual values was 82.0%. [Conclusion] The cutoff time to walk the 3-m zigzag was estimated to be 10.5 seconds, suggesting that this model may be a valid index for the prediction of falls.