Reconclining phi radiative decays with other data for a0(980), fo(980), pi-pi -> KK and pi-pi -> eta-eta

Data for phi -> gamma (eta-pizero) are analysed using the KK loop model and compared with parameters of a0(980) derived from Crystal Barrel data. The eta-pi mass spectrum agrees closely and the absolute normalisation lies just within errors. However, BES parameters for fo(980) predict a normalisation for phi -> gamma (pizero-pizero) at least a factor 2 lower than is observed. This discrepancy may be eliminated by including constructive interference between fo(980) and sigma. The magnitude required for sigma -> KK is consistent with data on pi-pi -> KK. A dispersion relation analysis by Buttiker, Descotes-Genon and Moussallam of pi-pi -> KK leads to a similar conclusion. Data on pi-pi -> eta-eta also require decays of sigma to eta-eta. Four sets of pi-pi -> KK data all require a small but definite fo(1370) signal.