We present a quantitative study of the evolution of point defects into clusters and extended defects in ion-implanted Si. Deep level transient spectroscopy (DLTS) measurements are used to identify and count the electrically active defects in the damaged region produced by Si ion implantation at energies of 145 keV–2 MeV, and fluences from 1×108 to 5×1013 Si/cm2. Analyses of silicon annealed in the temperature range 100–680 °C allow us to monitor the transition from simple point defects to defect clusters and extended defects that occur upon increasing the ion fluence and the annealing temperature. At low doses, <1010 Si/cm2, only about 2% of the Frenkel pairs generated by the ion beam escape recombination and are stored into an equal number of interstitial and vacancy-type point defects. Thermal treatments produce a concomitant annealing of interstitial and vacancy-type defects until, at temperatures above 350 °C, only two to three interstitial-type defects per ion are left, and the DLTS spectra contain signatures of second-order point defects. Interstitial clusters at Ev+0.29 and Ev+0.48 eV are found to dominate the residual damage of silicon implanted at higher fluences, 1×1012–7×1013 Si/cm2, and at annealing temperatures, T⩾600 °C. These interstitial clusters have point defect capture kinetics and are not observable in transmission electron microscopy (TEM), suggesting that they are smaller than ≈50 Å. Finally, for silicon implanted at higher Si doses, ⩾5×1013 Si/cm2, thermal treatments at 680 °C result in a strong decrease in the concentration of the interstitial cluster signatures and in the introduction of a different DLTS signal, Ev+0.50 eV, which exhibits logarithmic rather than exponential carrier capture kinetics, a feature typical of an extended defect. Comparison of the formation and dissolution of this extended defect signature with TEM analyses indicates that this level is a signature of the rodlike {311} defects that are known to store the interstitials responsible for transient enhanced diffusion. These results suggest that the small interstitial clusters are either the precursors of the {311} defects or that they compete with {311} defects as sinks for self-interstitials.