At the recent meeting of the International Neuro-ophthalmology Society in Geneva one of the most interesting presentations was given by two collaborators who disagreed on the interpretation of their joint findings. The atmosphere was more than usually stimulating. The disputed results are published in this issue of BJO (p 30).

It was Josef Zihl from Munich who in 1979, with von Cramon, put forward evidence that it might be possible with practice to extend areas of residual vision in cases with homonymous visual field defects secondary to occipital damage. I have always been sceptical of these findings.

In recent years the work of Erich Kasten and Bernhard Sabel and others in Magdeburg has raised interest again in the prospect of using training methods to bring about a reduction in the extent or density of visual field defects in such patients. The method is referred to as visual restitution training (VRT). However, it is well established that patients with homonymous hemianopia develop eye movement strategies that are adaptive and can potentially improve performance on conventional perimetric tasks unless eye movements are rigidly controlled (see for example Pambakian et al¹). Patients will make an involuntary exploratory saccade into the blind field more frequently than into the sighted field. On the next saccade the fovea is returned to the fixation target. Methods of monitoring fixation, which rely upon the patient reporting a change at the fixation target, may not be fully sensitive to such eye movements and furthermore the authors here suggest that the fixation target used in the previous studies from Kasten’s group may have been detectable eccentrically away from the fovea.

This is a very different situation from the type of eye movement artefacts that are controlled for in conventional perimetry, where the patient’s eye may wander for seconds of time. Rather, we are dealing here with eye movements the duration of which are not much more than two saccadic latencies. These eye movements will also defeat the strategy of presenting targets for less than the latency of saccades. That is fine to prevent patients from shifting gaze towards a target detected in the periphery but if the patient is making frequent exploratory saccades throughout the testing period some targets will be detected.

In the present study the authors have used the “gold standard” for controlled perimetry using the scanning laser ophthalmoscope to monitor the fundus. The conclusion is incontrovertible that using these detection targets there is no expansion of the seeing field as a result of VRT. It remains possible that improvement may have been in the nature of relative defects which would not have been detected by the method employed in this study to detect absolute defects.

This is not to say that there is nothing to be gained from attempts to rehabilitate patients with homonymous hemianopia by encouraging the development of eye and head movement strategies as Zihl himself later reported. These may be tailored for specific tasks—for example, navigation (Christopher Kennard’s group at the Charing Cross Hospital in London) and reading (Richard Wise and Alex Leff at the Royal Free Hospital in London). These strategies do not claim to improve the bare perimetric results but may enable the patient to make better use of his or her residual vision.

Collaborations of this type require a degree of courage and trust and we should suitably applaud these researchers.