Both logit and probit models are used to estimated fractional dependent variables, i.e., Y=0 or 1. Classical model: Y=1 if (Y*=xb+e>0), Y=0 otherwise. Whether the model is logit or probit depends on the error term, "e". If the error follows logistic distribution, then Prob(Y=1)=exp(xb)/(1+exp(xb)) and the model is called logit. If the error follows normal distribution, then Prob(Y=1)=Phi(xb) and the model is probit.