We show that the coherent response of semiconductor band-to-band continuum transitions at room temperature, excited by 11 fs pulses at elevated carrier densities, deviates substantially from that expected for an ideal photon echo. A microscopic non-Markovian quantum kinetic theory of both electron-electron and electron-LO-phonon scattering reproduces these unusual experimental observations qualitatively and, furthermore, the famous scaling of the decay time t with carrier density n eh according to Consider a dense gas of electrons and holes generated at random places in pairs with opposite momentum in coherent quantum states in a semiconductor. It is clear that the particles will move in such a way that they screen their mutual Coulomb interaction. How long does it take that screening is built up and that scattering has destroyed the initially quantum mechanically coherent state? About one decade ago, Shank et al. addressed this fundamental question when they studied the variation of the Coulomb scattering in a dense gas of optically excited crystal electrons and holes [1,2] with density n eh using ഠ10 fs pulses and a photon echo technique. They reported that the decay time t of the photon echo signal at room temperature scales approximately according to t~n 21͞D eh , where D is the dimension of either bulk GaAs ͑D 3͒ or GaAs quantum wells ͑D 2͒. At that time, it was tempting and appealing to interpret this seemingly simple result in terms of the nearest-neighbor scattering of D-dimensional hard spheres in which case it is simple to show that the average time between different collisions scales~n 21͞D eh indeed. Today, a number of important questions are still unanswered. (i) Is it a photon echo? In analogy with a spin echo this would mean that a first short pulse excites the system, a second pulse arrives after time delay t 21 , and the coherent collective response of the system, the photon echo, peaks at yet another time delay t 21 later. Without real-time resolution [1,2], however, one only measures the energy of the echo signal versus t 21 and its decay time t. For a phenomenological dephasing time T 2 it is known that T 2 4t holds for a four-wave mixing (FWM) geometry under these conditions. Not a single experiment, however, has shown directly that the coherent real-time response of semiconductor band-to-band continuum transitions under these conditions is a photon echo indeed. Under quite different conditions (T 4 K, 100 fs pulses, and much lower carrier densities), Ref. [3] showed that the total response is not a photon echo. Only after spectral filtering (behind the sample) of those components of the FWM-signal resonant with the band-to-band continuum, were they able to isolate a contribution which corresponded to an echo. The position of this signal shifted with time delay as expected [4], its width was constant (150 fs) and given by the time resolution of their setup for all conditions. The picosecond response of inhomogeneously broadened excitonic transitions [5] as well as the response of modulation-...