We show that the hedgehog soliton solution describing the nucleon in the SU(3)xSU(3) linear sigma model breaks down when the pion mass becomes too large. PACS: 11.30Rd; 12.50Lr The experimental value of the nucleon sigma-term aN ~60 MeV [1] is roughly twice the theoretical quark model prediction obtained by relating aN to the octetbaryon mass-splittings. It has been suggested [2] that this discrepancy points to a large strange-quark content (gs)N in the nucleon, in violation of the OZI rule: Another solution to this puzzle, advocated by Jaffe [3], is that the variation of baryonic masses with (rh-m~) (where rh and m~ are the non-strange and strange currentquark masses) is strongly non-linear, so that first-order perturbation theory does not give a reliable estimate of the octet-baryon mass-splittings. Some recent model investigations [4,5] have indeed shown evidence that the nucleon mass may depend non-linearly on the strangequark mass. A more natural possibility, which is suggested by chiral perturbation theory [2,6], is that the nucleon mass depends non-linearly on the non-strange quark mass.We showed recently [7] that, in the framework of the SU(3)x SU(3) linear sigma-model I-8], the nucleon mass does not vary significantly with the strange currentquark mass, but that its dependence on the non-strange current-quark mass rfi (proportional to. the SU(2) x SU(2) dynamical symmetry breaking parameter -cl=F,~rn~) is strongly non-linear. Because this nonlinearity has nothing to do with strangeness, it also occurs in the SU(2) x SU(2) linear sigma model [9]. Actually, we shall show that, when Cl is varied sufficiently far away from the chiral limit ca =0, something more dramatic happens in the (two-or three-flavor) linear sig-* On leave of absence from the Laboratoire de Physique Th6or-ique, Universit6 de Nice, parc Valrose, F-06034 Nice Cedex, France ma model. The nucleon is described in this model as a non-topological soliton 1-10-12]. It is well known that classically stable non-topological solitonic solutions exist only for a limited range of the model parameter values [13][14][15]. This general property, which has been well documented by numerical investigations in various field-theoretical models [16][17][18][19], has the consequence that, when -ca (or mR) is varied, all the other model parameters being fixed, the solitonic solution describing the nucleon ceases to exist above a critical value -c] (or m~).To show this, we consider the SU ( (a:= 0, ..., 8) combines a scalar nonet ~ and a pseudoscalar nonet ~b~ coupled to the quark triplet field 0, and