Description: The Milankovitch theory of climate change predicts that variations of the climate system should match the dominant frequencies of the orbital forcing in the 41 and 23 kyr**-1 frequency bands. Such a linear theory would predict that the amplitude variations of the climate response in these bands should match amplitude variations in orbital forcing. Here we compare amplitude variations of the marine oxygen isotope record with orbital forcing in these bands over the last 700,000 years and find systematic changes through time. We express these amplitude mismatches as variations in the glacial response time, a measure of the climate system's sensitivity to orbitally induced insolation changes. Variations in the glacial response time occur in all frequencies bands without strong concentration of variance in any given band, and have a 'red' spectrum with larger variations at the longer periods. The response time is coherent with delta18O at periods of 100 and 41 kyr, which suggests that the variations in glacial response time in part reflect internal feedback mechanisms of the global climate system. The phase relationship between the estimated glacial response time and the delta18O (ice volume) record is very different at these two frequencies, which suggests at least two separate feedback mechanisms. The first mechanism enhances the 100,000-year climate cycle by increasing rates of change during major glacial terminations. Candidates for this feedback include lithospheric depression and rebound, enhanced ice calving from large marine based ice sheets, and possibly others. A second set of mechanisms, which is detected in the response to the 41,000-year orbital cycle of Earth's obliquity, accelerates ice growth events and slows glacial melting. Some models which include feedbacks between ice sheets, sea ice, and deep ocean temperatures predict early rapid ice growth, followed by slower growth, and this general feature is consistent with our analysis. While we can not at present identify the specific feedbacks leading to asymmetry of growth and decay rates at different frequency bands, the finding of this ice-growth acceleration mechanism in the 41,000-year frequency band suggests that high-latitude processes, where insolation varies most strongly at this rhythm, may be involved. Our finding of systematic changes in climate sensitivity has implications for orbitally tuned chronologies in Pleistocene sediments. Instead of a constant phase shift within a frequency band between orbital forcing and glacial response, as has been assumed in the past, we suggest a variable phase. The largest changes in age estimates for isotopic events are at the glacial terminations, which in our chronology are as much as 3500 years older that estimated previously.