We can do some pretty remarkable things with our bodies, not all for public retelling. However, even the most mystic yogis among us would probably have trouble immobilising their live food at a distance - and through a highly refractive water-air interface. This, however, is exactly what archer fish (Toxotes jaculatrix) must do successfully from a young age, if they are to survive and grow. They are capable of directing a jet of water directly at perching insects up to a metre away, so dislodging them and causing them to fall back into the water, where they are eaten. All this is done from underwater, so that the insect is unaware of the impending attack. How is this proficiency acquired so successfully? And - perhaps more interestingly still - how can experiments be designed to investigate this behaviour? Stefan Schuster and colleagues provide an answer to both questions in their recent Current Biology article.

They offered single archer fish a choice of different target sizes (8 black discs, from 2-30 mm diameter, printed on white paper) at a range of distances (200-800 mm) above the water surface. Would they try to match some internal `template' (in which case their preferred size of disc would increase as the paper was moved further away); or would they choose the same absolute size of disc, irrespective of range? The latter would have to imply that the fish were actively calculating absolute size and range through a highly distorting air/water interface.

Intriguingly, the fish went for the same absolute size of disc, irrespective of range. Even more impressively, rather than assessing the targets from directly underneath, or from a precise underwater viewing angle, the fish immediately made a choice of which target to aim for from a wide variety of starting positions in the tank, swam directly to a firing position, and fired at one of the discs. As the fish had already made their choice by the time they took up position, their initial calculation of absolute size and position must have been highly complex. Thus, it is puzzling that the fish seemed to make a clear choice of target; how could they tell the absolute size of their target, when the distortion of refraction at the air/water interface could (for example) make a close small disc appear larger than a larger, more distant, one? The authors' clear implication is that the fish must be calculating absolute size and range (despite the vagaries of variable refraction through the air/water interface), rather than matching potential targets to some internal `template'. Surprisingly, naïve fish were nearly as good as seasoned hunters, although their preferred disc size did seem to increase slightly with distance, to the extent that, at the furthest distances, they were firing at sizes that would have been too large to consume. However, it proved relatively easy to train fish to associate a particular absolute size of target with a reward (of a crane-fly). This was achieved by presenting a range of disc sizes at a range of distances from the target, and only rewarding an immediate attack on the correct target when presented. Remarkably, the fish were able to discriminate differences in diameters of only 1-3 mm over a distance of up to a metre - and all from a variable distance, and from underwater.

The authors interpreted the results as implying that the fish could not possibly be comparing their targets with a range of pre-stored, or hard-wired `templates'. Rather they must be able to calculate, over a wide range of conditions, a remarkably accurate absolute size for their potential targets. They are thus necessarily learning the laws of refraction - another remarkable thing we are capable of, although usually only after formal education!