We compared techniques to assess diet at 17 Hen Harrier (Circus cyaneus) nests. Diet was measured directly from observations and compared to estimates from pellets, prey remains and a combination of pellets and remains. For data pooled across nests, pellets over-represented mammalian prey and under-represented avian prey. Prey remains over-represented large prey and under-represented small prey. Combining pellet and remains data did not eliminate these biases. Pellets gave higher diversity values than direct observations and detected more small prey species. For data analyzed on a nest by nest basis, estimates from pellets were significantly related to estimates from direct observations for three prey types by frequency and all types by biomass. These linear relationships were used to predict frequencies and biomass of prey types in the observed diet at five new nests. Our findings suggest that pellets are useful for estimates of prey diversity and as an index of the frequency of certain prey types in the diet, but direct observations are necessary to help quantify the biases inherent in diet estimates.