An HF crossed arm sampling array was used to investigate the effect of the ionosphere on the direction‐of‐arrival (DOA) of 9.3‐MHz transionospheric transmissions from ISIS satellites. The experimental results were compared to theoretical results obtained from a transionospheric raytrace program. The difference in elevation between the transionospheric ray and the line‐of‐sight ray, referred to as the vertical DOA error, varies with elevation angle and maximizes at the iris of the ionosphere. The theoretical results show that if the transionospheric ray passes through a horizontal gradient or blob or trough in the electron density, a refractive error occurs. This error decreases if the ray passes through an electron density enhancement which is greater above the peak of theF2 layer than below it and increases if the ray passes through an enhancement which is greater below theF2 peak than above it. Experimental data exhibit this behavior. TheS4amplitude scintillation index is essentially independent of ray elevation and aspect angle regardless of the presence or absence of magnetic field‐aligned electron density irregularities (FAIs). For times when small‐scale FAIs are clearly present, the DOA of signals from a moving satellite fluctuates both most rapidly and most slowly when the orbit is polar and passes overhead. The very rapid fluctuations occur when the DOA of the transionospheric ray is near the magnetic zenith. This may be a diffractive effect. The fluctuations are slowest, probably indicating a refractive effect, when the satellite is approaching and receeding. This behavior is dependent upon the geometry of the situation but is absent when there are no FAIs. DOA fluctuations occurred at all the invariant latitudes observable (53°–65°), with the largest fluctuations occurring more frequently north of 60° and with increasing magnetic activity. The DOA fluctuations in this area were present in the summer and absent in the winter but were negligible when FAIs were absent. The seasonal variation is believed to be due to a seasonal variation in the chemistry of theF2 layer. When large blobs of FAIs appear to be located at the expected DOA, most connecting ray paths during the satellite transit appear to be refracted from the sides of adjacent blobs. However, one could also positively identify a small number of connecting ray paths per second which arrived at the expected DOA, indicating that a few rays appeared to have woven their way among the irregularit