On the range of validity of the QCD-improved parton model

Based on the world DIS data we extract the experimental F_2^p-F_2^n as a function of Bjorken-x and photon virtuality Q^2, using two different methods. Both methods lead to identical results. We find that the standard PDFs fail to describe the experimental data below Q^2 < 7 GeV^2, which is much higher than for F_2^p and F_2^d separately. The difference between PDFs and the experimental data cannot be understood as due to nuclear effects in the deuteron, and evidently suggests substantial nonsinglet higher-twist effects. The trend of the experimental data is approximately explained by a recent two-component model of the nucleon structure functions and suggests strong Q^2-dependence of the Gottfried Sum, in disagreement with the parton model interpretation. The large negative higher-twist effects can explain the difference between the value of the Gottfied integral obtained recently by the E866 Drell-Yan experiment at Fermilab and an older NMC result.