The interpretation of the radar aspect sensitivity observed at VHF frequencies in the lower atmosphere is still a subject of controversy in the radar community. Indeed, scattering from anisotropic turbulence layers and partial reflection from stable thin horizontally stratified layers are generally proposed without leading to a definitive conclusion. A cause of this persistent discussion has been a lack of in situ high-resolution observations which could identify and describe accurately the atmospheric structures at the origin of the radar aspect sensitivity. The Radars, Scidar and Balloons (RASCIBA 90) campaign (February-March 1990, Aire sur l'Adour, France) was performed using simultaneously colocated VHF ST radars and balloon experiments. The objective was to obtain information about the small-scale structures of the lower atmosphere and then to identify the origin of radar echoes. The main result was produced by the high vertical resolution (20 cm) temperature measurements. For the first time, very strong (positive) temperature gradients within thin layers were detected in the lower atmosphere (at least up to 27 km). Such an observation allowed then investigation of the partial reflection interpretation using available bifrequency radar measurements: a 45-MHz radar for which typical 15-dB vertical enhancement is observed just above the tropopause and a 72.5-MHz radar for which an aspect sensitivity of the same order is presented for the first time at this frequency. A simple model considering flat and very extended sheets was assumed to estimate quantitatively the possible contribution of these temperature gradients to the VHF radar vertical power. Radar and reconstructed power reflection coefficient profiles are in good agreement in shape and in level showing that partial reflection from atmospheric sheets is an important and generally dominant process at vertical incidence. Furthermore, the effect of sheets of limited extent is investigated using simple theoretical considerations, and this model shows that the correction to the infinite sheet approximation is probably weak. The sheet distortion effect is also approximately evaluated applying Gaussian rough models. The power loss from the flat sheet case at VHF in the vertical direction is weak for a mean roughness height lower than a few tens of centimeters and for larger heights if the correlation length of the irregularities is larger than a few hundreds of meters. Sheet generation mechanisms are also briefly discussed using two previously published models: the viscosity wave model (Hocking et al., 1991) and “sheet and layer” model (Gossard et al., 1985). Neither of the models in their present forms seems to be able to account for the observations.