This study examines the existence, tendency and determinants of herding in the Malaysian stock market under market stress from 2016 to 2020. This study adopts ordinary least square and quantile regression models to estimate herding. Three types of measurements are used to capture volatility, which are realized volatility, Parkinson volatility and Garman and Klass volatility. The result shows that herding exists in the Malaysian stock market. Investors are observed to herd stronger in the bearish (down) market condition compared to bullish (up) market condition, especially in the upper quantile (τ > 50%). Realized volatility is found to be significant in every quantile except for the median quantile (τ = 50%) and Garman and Klass’s volatility is significant in the upper quantiles of 0.75 and 0.90. This study assists analysts and investors to formulate better investment strategies. Regulators and policymakers shall also control and regulate the herding behaviour of investors, which can deviate the stocks from their fundamentals. The existence of herding also violates the assumptions of EMH in assuming that investors are rational.