The response of a neuron to a linear combination of the first two harmonics of a fundamental frequency is studied by means of a leaky integrate and fire model. A suitable modification of the classical stochastic model is introduced to consider such input. The resulting interspike interval distribution exhibits maxima in correspondence with the fundamental frequency that was absent in the input signal. This fact shows the ability of the system to recognize the ''ghost'' frequency. Resonance-like behavior is also showed by the model neuron in a set of instances. The simplicity of the considered model makes also easy to understand the features involved in the ghost resonance phenomenon and the recognition of the parameters ranges compatible with such behavior.