The mechanism of polar amplification in the absence of surface albedo feedback is investigated using an atmospheric model coupled to an aquaplanet slab ocean forced by a CO2 doubling. In particular, we examine the sensitivity of polar surface warming response under different insolation conditions from equinox (EQN) to annual mean (ANN) to seasonally varying (SEA). Varying insolation greatly affects the climatological static stability. The equinox condition, with the largest polar static stability, exhibits a bottom-heavy vertical profile of polar warming response that leads to the strongest polar amplification. In contrast, the polar warming response in ANN and SEA exhibits a maximum in the midtroposphere, which leads to only weak polar amplification. The midtropospheric warming maximum, which results from an increased poleward atmospheric energy transport in response to the tropics-to-pole energy imbalance, contributes to polar surface warming via downward clear-sky longwave radiation. However, it is cancelled by negative cloud radiative feedbacks locally. Furthermore, the polar lapse rate feedback, calculated from radiative kernels, is negative due to the midtropospheric warming maximum, and hence is not able to promote the polar surface warming. On the other hand, the polar lapse rate feedback in EQN is positive due to the bottom-heavy warming response, contributing to the strong polar surface warming. This contrast suggests that locally induced positive radiative feedbacks are necessary for strong polar amplification. Our results demonstrate how interactions among climate feedbacks determine the strength of polar amplification.