We present a new gap-free version of the seasonal and longitudinal (s/1) variations of PEFI, the equatorial F region irregularity (EFI) occurrence probability, based on data from the AE-E spacecraft. The agreement of this and three earlier partial PEFI patterns verifies all four. We reinterpret another earlier gap-ridden pattern, that of D¿RSF, a topside ionogram index of average darkening by range spread F. We compare it with PEFI and, using ionosonde radio science considerations, we conclude that D¿RSF=PEFI times a factor depending on the average number of topside plasma bubbles visible to the ionosonde. The s/1 variations of D¿RSF thus imply s/1 variations in the average spacing of bubbles, whose seeds have an occurrence probability pattern Pseed. For discussion we assume PEFI=PinstPseed, where Pinst is the pattern of F region instability. The PEFI pattern, which is by definition independent of seed and/or bubble spacing, is far too complex to be explained by the dominant paradigm, that of changes in Pinst by simple changes in the F region altitude and/or north-south asymmetry. We examine evidence behind this dominance, and find it unconvincing. Both the asymmetry and sunset-node/altitude hypotheses of 1984 and 1985, respectively, seem to be partly based on misunderstood data, and their features appear displaced in time and space from those of our repeatable PEFI pattern. In contrast, if Pseed variations influence the PEFI pattern and depend on thermospheric gravity waves from tropospheric convection near the dip equator, then the seasonal maxima (minima) of PEFI could be explained, since they all occur above relatively warm (cold) surface features, where convection is maximal (minimal). Also, thehypothesis of the dominance of the Pseed term could explain an unusual December/January PEFI maximum in the deep, wide, normal Pacific minimum in the one data set obtained in El Ni¿o years. Based on the experiments we consider, we predict that the s/l variations of Pseed will be found to be similar to those of PEFI, and largely to explain them. Finally, we find reasons, based on the similarity of the D¿RSF variations to s/l patterns of the average scintillation index, for not using, as is commonly done, such scintillation patterns as substitutes for PEFI or Pinst patterns. ¿ 1998 American Geophysical Union