Most studies on icon search have assumed a normal distribution for search times, using the arithmetic mean as a representative value of the measured target icon search times. In this study, various types of icon search experiments were performed and search times were measured. The research question was, “What probability distribution does the measured search time correspond to?” In all experimental conditions, the search time was found to best fit the ex-Gaussian distribution rather than the normal distribution, exponential distribution, and ex-Wald distribution. Therefore, this suggests that it is more appropriate to use the mean of the ex-Gaussian distribution to express the central tendency of icon search times. In all experimental conditions, the arithmetic mean value was greater than the mean of the ex-Gaussian distribution. In addition, the higher the difficulty of the search experiment, the larger the difference between the arithmetic mean and the mean of the ex-Gaussian distribution. In particular, if the difficulty of icon search is high, it would be appropriate to use the mean of the ex-Gaussian distribution.