Nowadays, the majority of stock market trading is performed electronically, based on pre-programmed computer algorithms. We obtain five-minute high-frequency data from the Turkish Stock Exchange to investigate the data-generating process of emerging market returns during the global financial crisis of 2007-2009. We test tail behaviour and how data-generating processes changed during the intraday trading period in both crisis and non-crisis periods. We also examine whether price asymmetry has a significant effect on the diffusion and jump characteristics of emerging market returns. The results identify a clear increase in jumps with infinite activity in crisis periods and a decreased identification of jumps with finite activity in non-crisis periods. In crisis periods, the proportion of large and small jumps increased and the proportion of Brownian motion decreased. We show that data-generating processes are not stable during the intraday trading period, which fluctuates slightly, particularly right after the market opening times in the morning and in the afternoon. Finally, we conclude that there are many more stressful days in crisis periods than in non-crisis periods in emerging markets returns.