Investigations of temperature fluctuation spectra of atmospheric turbulence were performed simultaneously by remote sensing and in situ measurements. A balloon borne instrumented gondola measured temperature and velocity fluctuations up to an altitude of 18 km. Spectral analysis of these data led to one-dimensional spectra. Simultaneously, the nearby stratosphere-troposphere radar “Provence” obtained vertical profiles of reflectivity for both vertical and oblique (15°) lines of sight and for various range resolutions. According to the classical interpretation, the radar echoes result from the three-dimensional spectrum of refractive index fluctuations, and its aspect sensitivity reflects the spectrum anisotropy at the half radar wavelength. We thus separately reconstructed the reflectivity profiles for the two lines of sight. For the oblique one, the in situ observed spectral level of temperature fluctuations (assumed isotropic) was converted into radar reflectivity. For the vertical measurements, a theoretical “universal” level was used, along with an isotropic conversion formula (a tentative account for the known anisotropy is given in the appendix). Comparison of the reconstructed reflectivity profiles with the observed one proves to be very puzzling. For oblique measurements, where the reconstruction process uses only widely accepted hypotheses, the observed profiles are poorly reproduced. By contrast, for vertical measurements, the proposed model leads to a good reconstruction of the shape of the reflectivity profiles. In both cases, the predicted level is always strongly overestimated. The discussion concludes that the discrepancy is real and calls for further studies of this problem.