This paper considers the dynamic lot sizing problem when there is learning and forgetting in setups. Learning in setups takes place with repetition when additional setups are made and forgetting takes place when there is a break between two setups. We allow the amount forgotten over a break to depend both on the length of the break and the amount of learning at the beginning of the break. The amount of forgetting could be small over the short initial interval of interruption. We present some decision/forecast horizon results and managerial insights based on computational results. We found that the characteristics of worker forgetting influence the setup frequency. For instance, the presence of forgetting characterized by a slow-forgetting S-shaped function leads to more frequent setups. Based on our computational experiments and analytical insights, the problems with a slow-forgetting S-shaped function are harder to solve and consequently provide longer forecast horizons.