This paper uses fractional integration methods to obtain new evidence on polar amplification. The adopted modelling framework is very general since it allows the differencing parameter to take any real value, including fractional ones, and provides useful information on both the short and the long run. The analysis is carried out using monthly temperature anomaly data for both the Arctic and the Antarctic, as well as the Northern and Southern Hemisphere, which have been obtained from the NOAA (National Center for Environmental Information) archive. The main findings can be summarised as follows. There is evidence of Arctic amplification, since the upward trend in the Arctic data is more pronounced compared to that in the Northern hemisphere series, but not of Antarctic amplification, where the opposite holds. Also, the effects of shocks are more long-lived in the Arctic/Northern hemisphere than in the other pole/hemisphere. These results are robust to whether or not seasonality is explicitly modelled. In addition, temperature changes in the poles have bigger effects on those in the corresponding hemisphere if they occur in the Antarctic rather than in the Arctic.