We perform comprehensive analyses of trapped waves and signals of damaged fault zone rocks associated with time delays of body waves along the Parkfield section of the San Andreas Fault (SAF). Waveforms generated by thousands of earthquakes and recorded by near fault stations in several permanent and temporary deployments are examined, with attention to the possible influence of a lithology contrast across the fault on signals of the low velocity damage zone. Clear candidate trapped waves are identified at only three stations, MMNB and two of the other near fault stations (FLIP and PIES) further to the NW. Clear candidate trapped waves are not seen at any of the near fault stations to the SE of MMNB. The locations of the events generating good candidate trapped waves at MMNB and the other two stations are not distributed broadly in space, but clustered in a small number of locations. Moreover, events that generate clear candidate trapped waves at one station do not typically generate trapped waves at the other two stations. These observations imply that the damage zone is highly variable along strike and that a coherent connected waveguide does not exist for distances along strike larger than at most 3–5 km (the distance between stations). Synthetic waveform fits for observed trapped waves at stations MMNB and FLIP indicate that the most likely parameters of the trapping structures at these locations are widths of about 150 m, depths of about 3 km, velocity reductions of 30–40 per cent, and Q values of 10–40. Synthetic calculations of trapped waves demonstrate that if there is a contrast of seismic velocity across the fault, the trapped waves are delayed relative to the S wave arrival. Trapped waves at station MMNB, and to a lesser extent also at stations FLIP and PIES in the NW section, show this characteristic. This suggests a lithology contrast in the top few km at these locations, in agreement with results from tomography and studies of head waves in the Parkfield area. At several mini across-fault arrays, where trapped waves are not observed, a low velocity damage zone is detected from the delay in the arrival time of body wave phases relative to a nearby off-fault station. The observed delay of the S wave is greater than the P wave delay, consistent with the existence of a damage zone with Poisson ratio of about 0.33. The observations of time delays without trapped waves indicate that parts of the damage zone are insufficiently coherent to generate trapped waves. A broader damage zone may exist in the region between the SAF and the South West Fracture Zone. The results highlight the diversity of damage structures along the ∼40 km of the SAF examined in this study, and imply that fault imaging based on data at single sites does not necessarily apply to a larger section.