We analyze the Crab pulsar at ten frequencies from 0.43 to 8.8 GHz using data obtained at the Arecibo Observatory and report the spectral dependence of all pulse components and the rate of occurrence of large-amplitude 'giant' pulses. Giant pulses occur only in the main-and-interpulse components that are manifest from radio frequencies to gamma-ray energies (known as the 'P1' and 'P2' components in the high-energy literature). Individual giant pulses reach brightness temperatures of at least 10 32 K in our data, which do not resolve the narrowest pulses, and are known to reach 10 37 K in nanosecond-resolution observations (Hankins et al. 2003). The Crab pulsar's pulses are therefore the brightest known in the observable universe. As such, they represent an important milestone for theories of the pulsar emission mechanism to explain. In addition, their short durations allow them to serve as especially sensitive probes of the Crab Nebula and the interstellar medium. We identify and quantify frequency structure in individual giant pulses using a scintillating, amplitude-modulated, polarized shot-noise model (SAMPSN). The frequency structure associated with multipath propagation decorrelates on a time scale ∼ 25 sec at 1.5 GHz. To produce this time scale requires multipath propagation to be strongly influenced by material within the Crab Nebula. We also show that some frequency structure decorrelates rapidly, on time scales less than one spin period, as would be expected from the shot-noise pattern of nanosecond duration pulses emitted by the pulsar. We discuss the detectability of individual giant pulses as a function of frequency and provenance. Taking into account the Crab pulsar's locality inside a bright supernova remnant, we conclude that the brightest pulse in a typical 1-hour observation would be most easily detectable in our lowest frequency band (0.43 GHz) to a distance ∼ 1.6 Mpc at 5σ. We also discuss the detection of such pulses using future instruments such as LOFAR and the SKA.