The annual mean temperature on the Tibetan Plateau (TP) has strongly increased over the past few decades, with larger warming in winter than in summer. Whether this different amplitude of change between seasons has persisted over longer time-scales in the past remains poorly understood, limiting our understanding of the mechanisms responsible. Here, we apply multivariate regression analysis and ensemble empirical mode decomposition (EEMD) to decompose winter (T-DJF) and summer (T-JJA) temperature reconstructions over the 1718-2005 CE period for the southeastern TP to investigate similarities and differences between winter and summer temperature changes, over multiple time-scales, as well as the driving factors behind the seasonal differences. The results reveal that the T-DJF and T-JJA reconstructions were significantly correlated throughout the study period, with the magnitude of the T-DJF variations approximately six times greater than the T-JJA variations. When the two reconstructions were decomposed over multiple time-scales, it was found that the consistency between winter and summer temperature reconstructions only existed at inter-annual scale. Assessing the driving factors, the main contributions to the T-JJA and T-DJF changes at the inter-annual and inter-decadal scales appear to be mainly the El Nino-Southern Oscillation (ENSO). The Pacific Decadal Oscillation (PDO) contribution was important to T-JJA and T-DJF changes at multi-decadal scales. Furthermore, we found that orbital parameters and the Atlantic Multidecadal Oscillation (AMO) was a major contributor to the changes in T-JJA and T-DJF at centennial scales, respectively. Both the T-JJA and T-DJF have a significant long-term increasing trend since c. 1850, mainly attributed to anthropogenic forcing. The detected similarities and differences between T-DJF and T-JJA at multiple time-scales provide new perspectives on the understanding the mechanisms behind climate change on the Tibetan Plateau and even entire East Asia.