An owl knows where to aim its talons even in the dark, thanks to neurons that can pinpoint the sound of a rustling mouse. These space-specific neurons perform more sophisticated computations than expected, researchers say. While most neurons simply add incoming signals to come up with an answer, these neurons can multiply.
Space-specific neurons receive two kinds of inputs. If a mouse squeaks to an owl's right side, that ear registers a slightly louder signal, and slightly sooner, than the left ear. Earlier research by Masakazu Konishi and colleagues showed that a set of auditory neurons calculates the difference in loudness and time and sends the results to neurons that are precisely tuned to particular locations. To learn how these neurons process the signals, neuroscientists José Luis Peña and Konishi of the California Institute of Technology (Caltech) in Pasadena outfitted 14 barn owls with headphones and monitored space-specific neurons' responses to pairs of sounds.
Two properties convinced the researchers that these neurons multiply signals. First, when very faint timing and loudness signals correspond to the same spot, the neurons in the auditory map fire robustly. This happens even if the two signals, when simply added, are too faint to rile up the neurons enough to fire. Second, the lack of either a loudness or timing signal can veto a space-specific neuron's firing--just as in multiplication, 2 x 0 = 0. As the researchers report in the 13 April issue of Science, a multiplicative model predicts how the neurons respond to different stimuli with about 98% accuracy.
Archetypal neurons don't compute this way. Normally, a neuron receives a host of excitatory and inhibitory signals of various volumes along its dendrites. When the signals add up to surpass some threshold, the neuron fires. Such a neuron acts like a transistor in an electronic circuit, says neuroscientist Christof Koch, also of Caltech but not involved in this project. But a neuron with the power to multiply, he says, "is more like a little processor; computationally it's much more powerful."