Many fish orient to steady water flow. This behavior, called rheotaxis, is key for river fish like salmon that migrate up- or downstream, or simply have to accommodate the ever-present river flow. How do fish sense the water motion?
It seems likely that the lateral line sense must be involved. The lateral line is an array of flow-sensing hair cells that run in lines along a fish's body. Indeed, if you pharmacologically disable the lateral line, many fish are not able to orient to flow very well.
But that's what's weird. The lateral line doesn't respond well to steady fluid motion. It has two types of sensor: canal neuromasts, which are embedded in pores below the fish's skin, and peripheral neuromasts, which stick out into the flow. Canal neuromasts only respond to differences in the flow speed between each end of the pore, so they can't sense steady flow. Peripheral neuromasts, on the other hand, respond to flow speed, but since they rapidly adapt to different steady flow speeds, it seems likely they wouldn't help much with rheotaxis either.
Boris Chagnaud, working in Horst Bleckmann's laboratory at the University of Bonn, realized that the physiological and behavioral data didn't match up. On the one hand, behavioral studies showed that knocking out the lateral line degrades rheotactic behavior, but on the other hand, the physiology indicates that the lateral line shouldn't be able to sense steady flow in the first place.
The resolution turns out to be simple. Steady flow doesn't exist. Even in highly controlled laboratory conditions, there are always turbulent fluctuations in the water motion, and these fluctuations move downstream at the mean flow speed. The researchers put an anesthetised fish in a laboratory flow tank designed to produce steadily moving water with very low turbulence. To measure the flow speeds, they tracked small particles in the water using a technique called particle image velocimetry. Even though the flow had little turbulence, they found fluctuations in the flow velocity near the fish's body. Points further along the fish's body had the same fluctuations, but slightly later in time, showing that the turbulence was moving downstream at the steady flow speed. The researchers were able to quantify this effect using a mathematical technique called cross-correlation, an estimate of the similarity in the fluctuations in two measurements when one measurement is delayed relative to the other. In this case, the cross-correlation time delay represents the time it takes a turbulent fluctuation to move along the fish's body.
In other words, if fish know the spacing between their lateral line sensors, they ought to be able to estimate the flow speed by looking at the cross-correlations between different neuromasts. To see whether this might be possible, Chagnaud measured the signals from pairs of neuromasts in the lateral line of goldfish. He found many pairs with high cross-correlations,and that the correlation time delay decreased as he increased flow speed.
Fish can't really perform fancy cross-correlation analyses like the researchers did. Instead, they might use what's called a `coincidence detector', known from studies of vision and hearing. Neurons in the brain will only respond when the signals from two different neuromasts reach the brain simultaneously, but the signals take different amounts of time to reach the brain. Such `delay lines' allow sets of neurons to respond to different time delays. With a set of different delay lines, the neurons are, in effect,performing a cross-correlation. These neurons still need to be located, but Chagnaud's work shows they could, in principle, allow fish to extract information on steady flow speed, even with lateral line sensors that don't respond to steady flow.