Around 1979, two realistic Nijmegen models of the baryon--baryon interaction were available: the elder model D and the improved new model F. Only model F led to the semiempirical value of the {\Lambda } binding in nuclear matter. When the first CERN observation of \Sigma hypernuclei was announced, model F was used to calculate V_\Sigma , the potential felt by \Sigma in nuclear matter. The result, a repulsive V_\Sigma , was unreconcilable with the CERN observation, and prompted theoreticians to use model D which led to an attractive V_\Sigma . To explain the existence of narrow \Sigma hypernuclear states at a relatively high energy, the theoreticians came forward with such ideas as the `bound states embedded in continuum', or V_\Sigma (r) with a repulsive barrier at the hypernuclear surface. A possible inaccuracy in the CERN experiments was not considered. The first empirical indication that V_\Sigma may be repulsive inside the nuclear core came from the analysis of strong interaction shifts and widths of \Sigma^- atoms, which could be explained with the help of model F of the baryon--baryon interaction. Final evidence of the repulsiveness of V_\Sigma was supplied by the new (K^-,\pi ) experiments performed at Brookhaven with an order of magnitude better statistics than the old CERN experiments. In the Brookhaven experiments the narrow states of \Sigma hypernuclei observed at CERN disappeared. The pion spectra measured in these new experiments are consistent with V_\Sigma repulsive inside nuclei and with model F of the baryon--baryon interaction.

PACS numbers: 13.75.Ev, 36.10.Gv