We have previously investigated the effectiveness of a custom built virtual environment in assisting training of a ventriculostomy procedure, which is a commonly performed procedure by a neurosurgeon and a core task for trainee surgeons. The training tool (called VCath) was initially developed as a low fidelity app for a tablet platform to provide easy access and availability to trainees. Subsequently we have developed a high fidelity version of VCath that uses a stereoscopic display to immerse the trainee in the virtual environment. This paper reports on two studies that have been carried out to compare the low and high fidelity versions of VCath, particularly to assess the value of stereoscopy. Study 1 was conducted at the second annual boot camp organized for all year one trainees in neurosurgery in the UK. Study 2 was performed on lay people, with no surgical experience. Our hypothesis was that using stereoscopy in the training task would be beneficial. Results from Study 1 demonstrated that performance improved for both the control group and the group trained with the tablet version of VCath. The group trained on the high fidelity version of VCath with a stereoscopic display showed no performance improvement. The indication is that our hypothesis is false. In Study 2, six different conditions were investigated that covered the use of training with VCath on a tablet, a mono display at two different sizes, a stereo display at two different sizes, and a control group who received no training. Results from this study with lay people show that stereoscopy can make a significant improvement to the accuracy of needle placement. The possible reasons for these results and the apparent contradiction between the two studies are discussed.