In a study to evaluate the reproducibility and accuracy of the sonar technique of measurement of the in vivo fetal crown-rump length (Robinson, 1973), a series of in vivo and in vitro experiments was performed in which the random and systematic errors inherent in the technique were assessed. The potential sources of random error were those of operator judgement, movement of the fetus and mother, machine sensitivity settings and measurement from the photograph; while the sources of systematic error were those of oscilloscope scale factor, and velocity calibration inaccuracies, and the effect of beam width. The overall effect of the random errors, that is, the reproducibility of the technique, was assessed in an in vivo blind trial in which three independent measurements were made of the fetus. In a series of 30 experiments the average standard deviation of the three readings was found to be 1.2 mm. Evaluation of the systematic errors by in vivo experimentation, on the other hand, showed that the basic sonar measurements were in error by an overestimate of 1 mm for the beam width effect and 3.7 per cent for the scale factor and velocity calibration errors. A weighted non-linear regression analysis of 334 measurements was performed in order to obtain a "curve of best fit" for the period covering 6 to 14 weeks of menstrual age. The values obtained were corrected for the systematic errors and compared with widely quoted anatomical figures. In the second part of this investigation the original data was further analyzed to determine on a statistical basis the accuracy of the technique as a method of estimating maturity. It was shown that such an estimate could be made to within 4.7 days with a 95 per cent probability on the basic of a single measurement, and to within 2.7 days if three independent measurements were made.