This paper analyzes realized volatility in the Japanese stock market by separately computing measures for the morning and afternoon trading sessions, avoiding the non-trading hours issue that complicates daily volatility estimation. Using high-frequency data for five liquid stocks from 2006 to 2009, the authors construct volatility signature plots to quantify microstructure noise bias as a function of sampling frequency. They find that bias is substantial at 5-minute sampling, especially in the afternoon session, and correct for it when examining standardized returns. Under the mixture of distributions hypothesis (MDH), standardized returns should have variance 1 and kurtosis 3. After bias correction, the standardized returns in both sessions exhibit variance close to 1 and kurtosis near 3, and the Anderson-Darling test does not reject normality. Additionally, absolute standardized returns show no significant autocorrelation, further supporting the MDH. The study concludes that price dynamics on the Tokyo Stock Exchange are consistent with the MDH, though slight deviations in variance and kurtosis suggest potential influences from jumps or other factors not captured by the simple diffusion model.