We made a parameter fit to the Haleakala neutron monitor counting rate during the 1991 March 22 solar flare (Pyle and Simpson, 1991) using the time profiles of γ-rays at 0.42 80 MeV obtained with the GRANAT satellite (Vilmeret al., 1994) and the microwave data from Owens Valley Radio Observatory. We use a two-component neutron injection function to find that either an impulsive injection or the ‘impulsive-plus-prolonged’ neutron injection is possible. In both cases, the number of > 300 MeV neutrons emitted towards the Earth is estimated as ≈ 2 × 1027 sr-1, which is less than that of the 1990 May 24 flare by an order of magnitude. We tested if such a big difference in neutron number detected on the Earth can be accounted for solely by their different positions on the solar disk. For the estimation of the degree of anisotropy of high-energy secondary emission, we made use of macroscopic parameters of the flare active region, in particular, the vector magnetogram data from the Big Bear Solar Observatory. In our result, the anisotropy factor for the neutral emissions of the 1991 March 22 flare is only ≈ 1 10, which is rather small compared with previous theoretical predictions for a disk flare. Such a moderate anisotropy is due to the relatively large inclination angles of the magnetic fields at the footpoints of the flaring loop where accelerated particles are trapped. We thus concluded that the smaller number of neutrons of the 1991 March 22 flare would be not only due to its location on the disk, but also due to fewer protons accelerated during this event as compared with the 1990 May 24 limb event. For a more precise determination of the anisotropy factor in a flare, we need a detailed spectrum of electron bremsstrahlung in 0.1 10 MeV and the fluence of γ-ray emission from the π0-decay.