The present study investigated the difference in voice perturbation measures and parameters obtained from nonlinear dynamic analysis between normal laryngeal phonation and standard esophageal (SE) phonation. Jitter, shimmer, correlation dimension and Kolmogorov entropy were measured from 10 SE and 10 normal male speakers of Cantonese. Jitter and shimmer values were significantly higher for SE than laryngeal voice. But jitter values were found to be significantly different when length of sound samples was altered. In addition, both correlation dimension and Kolmogorov entropy values were significantly higher for SE than laryngeal voice and sample length did not appear to affect the result. These results suggest that SE voices are more chaotic than laryngeal voice. It follows that the technique of nonlinear dynamic analysis may be more reliable and stable for evaluating the acoustic characteristics of SE voices.