This study compared parameter recov eries of the following IRT calibration procedures in a two-parameter logistic model: (1) heuristic method which transforms biserial correlation coefficients and proportion passing the items into item parameters; (2) heuristic method which transforms factor loadings and proportion passing the items into item parameters; (3) MMAP method which applies Bayesian priors to traits and item discriminating power and/or difficulty; (4) marginal maximum likelihood method; and (5) pseudo Bayesian method. Numerical simulation studies were used to evaluate these methods for varying distributions of traits and item parameters, test lengths, and sample sizes. Fifty datasets were generated for each of the combinations of factors. All these methods produced estimates of essentially equivalent accuracy when the distribution of taits was normal, although pseudo Bayesian and MMAP method produced more precise estimates than other methods when the distribution of traits was skew, particularly for longer tests and larger samples.