It often occurs in J-R testing that some initial crack extension (.DELTA.a) data points have anomalous negative values. The reason for the apparent negative crack growth is due to the analysis method. The phenomenon as a possible source of error in determining $J_{IC}$ or J-R curve from partial unloading compliance experiments may be eliminated by the compliance correction equation or the offset technique. In this study, the correction methods are suggested and examined by the measurement of the actual crack length and $J_{JC}$ analysis.