We have performed ultrafast absorption bleach recovery and fluorescence upconversion measurements ( approximately 100 fs time resolution) for three CdSe samples, with nanoparticle diameters of 2.7, 2.9, and 4.3 nm. The two types of experiments provide complementary information regarding the contributions of the different processes involved in the fast relaxation of electrons and holes in the CdSe quantum dots. Transient absorption and emission experiments were conducted for the 1S [1Se-1S3/2(h)] transition, 1S(e) and 1S3/2(h) representing the lowest electron (e) and hole (h) levels. The bleach recovery of the 1S transition shows a approximately 400-500 fs initial rise, which is followed by a size-dependent approximately 10-90 ps decay and finally a long-lived (approximately ns) decay. The fluorescence upconversion signal for the 1S transition shows quite different temporal behavior: a two times slower rise time (approximately 700-1000 fs) and, when the fluorescence upconversion signal has risen to about 20% of its maximum intensity, the signal displays a slight leveling off (bend), followed by a continued rise until the maximum intensity is reached. This bend is well reproducible and power and concentration independent. Simulations show that the bend in the rise is caused by a very fast decay component with a typical time of about 230-430 fs. Considering that the 1S quantum dot excitation is comprised of five exciton substates (F=+/-2, +/-1L, 0L, +/-1U, and 0U), we attribute the disparity in the rise of the bleaching and emission transients to the results from the dynamics of the different excitons involved in respectively the bleaching and fluorescence experiments. More specifically, in transient absorption, population changes of the F=+/-1U excitons are probed, in emission population effects for the F=+/-2 ("dark") and the F=+/-1L ("bright") exciton states are monitored. It is discussed that the fast (approximately 400-500 fs) rise of the bleach recovery is representative of the feeding of the F=+/-1U exciton (by filling of the 1S(e) electron level) and that the slower (approximately 700-1000 fs) feeding of the emissive +/-2, +/-1L excitons is determined by the relaxation of the hole levels within the 1S3/2 fine structure. Finally, the approximately 230-430 fs component, typical of the bend in the fluorescence transient, is attributed to the thermalization of the close-lying +/-2 ("dark") and +/-1L ("bright") excitons.