The paper, Threshold models in time series analysis - 30 years on, of Howell Tong is discussed and some remarks on the impact and possible future development of threshold models and their continuous-time analogues are made.