As a legacy of the TOGA programme, it is likely that prediction of interannual climate fluctuations using dynamically-based models will increasingly become an operational activity in many countries. And yet we should remember Tennekes et al's (1987) Popperian assertion that "no forecast is complete without a forecast of forecast skill". The role that ensemble techniques can play in estimating the predictability of seasonal forecasts is addressed here through a series of questions. Within this discussion we will consider the specific role that adjoint techniques can play in seasonal predictability studies