Copepods may remotely detect predators from the velocity gradients these generate in the ambient water. Each of the different components and characteristics of a velocity gradient (acceleration, vorticity, longitudinal and shear deformation) can cause a velocity difference between the copepod and the ambient water and may, therefore, be perceived by mechanoreceptory setae. We hypothesised that the threshold value for escape response to a particular component depends solely on the magnitude of the velocity difference (= signal strength) it generates. In experiments we isolated the different components and noted the minimum intensities to which the copepod Acartia tonsa responded. As hypothesised, threshold signal strengths due to longitudinal and shear deformation were similar, ~0.015 cm s-1, and were invariant with developmental stage. The latter implies that the threshold deformation rate for response scales inversely with size, i.e. that large stages respond to lower fluid deformation rates than small stages and, hence, may detect predators at longer distances. Signals due to vorticity and acceleration did not elicit escape responses, even though their magnitude exceeded threshold signal strength due to deformation. We suggest that A. tonsa cannot distinguish such signals from those due to their own behaviour (sinking, swimming, passive reorientation due to gravity) because they cause a similar spatial distributions of the signal across the body. Reinterpretation of data from the literature revealed that threshold signal strength due to deformation varies by ca 2 orders of magnitude between copepods and exceeds the neurophysiological response threshold by more than a factor of 10. In contrast, threshold deformation rates vary much less, ~ 0.5 to 5 s-1. Model calculations suggest that such threshold deformation rates are just sufficient to allow efficient predator detection while at the same time just below maximum turbulent deformation rates, thus preventing inordinate escapes.