Because of the pathogenicity and low incidence of avian influenza virus infections in humans, the immune correlates of protection for avian influenza vaccines cannot be determined from clinical studies. Here, we used the ferret model to address this for an avian influenza H5N1 vaccine. Using oil-in-water adjuvants, we generated groups of ferrets with undetectable (geometric mean titer [GMT] < 10), low (GMT = 28.3), or high (GMT > 761.1) hemagglutination-inhibition (HAI) titers to the A/Viet Nam/1203/2004 (H5N1) virus. Ferrets were then challenged with the wild-type virus and disease severity and immunologic parameters were studied. The severity of infection and symptom profile were inversely associated with pre-challenge HAI titers in a dose-dependent manner. A vaccinated ferret with no detectable HAI-antibodies but high flu-specific IgG-antibody titers mounted rapid functional antibodies after infection and experienced milder disease compared to other ferrets in the group. Compared to naïve ferrets, all vaccinated ferrets showed improved cellular immunity in the lungs and peripheral blood. High number of IFNγ+ CD8- T cells in the airways was associated with early viral clearance. Thus, while neutralizing antibodies are the best correlate of protection, non-neutralizing antibodies can also be protective. This should be taken into consideration in future avian influenza vaccine trials.