[en] A continuous reduction model implying the dynamical suppression of linear superpositions of macroscopically distinguishable states, presented recently by L. Diosi, is investigated. The model exhibits appealing features, in particular it relates reduction to gravity and contains no constants besides Newton's gravitational constant G. It turns out, however, that the model is not fully consistent. A modification of it is proposed, which overcomes the difficulties and retains partially its nice features. The resulting model deals with systems containing distinguishable or identical constituents, allows the derivation from microdynamics of wave packet reduction and accounts for the emergency of definite macroscopic properties for macro-objects. Reduction is related to gravity in the same way as in Diosi's model, but a fundamental length must be introduced to avoid inconsistencies. (author). 12 refs