The Microsecond Map

Masakazu Konishi, Barn Owls, and Coincidence-Detector Delay Lines

Sound travels through dry air at 343 meters per second. When a meadow vole rustles a dry leaf on a forest floor, the acoustic wavefront striking the head of a perching barn owl (Tyto alba) from ten degrees off the midline reaches the nearer eardrum just 30 microseconds before it reaches the farther eardrum. To a neurobiologist, that number poses an immediate physiological paradox. A single action potential in a vertebrate auditory nerve fiber lasts roughly 1,000 microseconds (1 millisecond), followed by an absolute refractory period of another 1,000 to 2,000 microseconds, and synaptic transmission across a chemical cleft introduces jitter measured in hundreds of microseconds. How can a brain built out of millisecond-scale neurons resolve interaural time differences (ITDs) down to ten microseconds—corresponding to a spatial angle of just one degree?

The behavioral reality of that precision was established in 1962 by graduate student Roger Payne in a light-sealed barn at Cornell University. Payne covered the floor with dry leaves, sealed every crack against external light, and verified total darkness by exposing ultra-fast infrared and panchromatic photographic film on the floor during trials. When a mouse walked across the leaves, the barn owl launched from its perch and struck the rodent with its talons within one degree of accuracy in both horizontal azimuth and vertical elevation. When Payne towed a crumpled ball of paper across the sand with a string, the owl struck the paper ball with identical precision; when he tied a rustling leaf to the tail of a silent mouse, the owl struck the leaf and ignored the mouse.

In 1977 and 1978 at the California Institute of Technology, Masakazu Konishi and Eric Knudsen inserted tungsten microelectrodes into the midbrain of anesthetized barn owls while moving a miniature loudspeaker along a semicircular mechanized hoop around the bird's head. In the external nucleus of the inferior colliculus (ICx), they discovered something never before seen in an auditory system—a two-dimensional topographic map of auditory space. Unlike the cochlea or primary auditory cortex, where neurons are arranged like piano keys by sound frequency (tonotopy), every neuron in the owl's ICx fired only when a sound originated from a specific, palm-sized receptive field in three-dimensional space, arranged sequentially from left to right along the anterior-posterior axis and from up to down along the dorsoventral axis.

Ten years later, in 1988 and 1990, Catherine Carr and Masakazu Konishi traced the upstream brainstem pathways from the auditory nerve to uncover how the owl computes that map. Immediately after entering the brainstem, each eighth-nerve auditory fiber bifurcates into two strictly segregated anatomical branches. One branch terminates in nucleus angularis, where neurons discard timing information and encode purely the decibel amplitude (intensity) of each frequency band. The second branch terminates via giant calyceal endbulbs of Held in nucleus magnocellularis, whose low-threshold potassium channels lock action potentials to the exact zero-crossing phase of acoustic carrier frequencies up to 9,000 hertz—three times higher than any mammalian auditory nerve can phase-lock.

For a spherical head of radius $r$ and sound incidence angle $\theta$, the Woodworth acoustic path-difference equation gives the interaural time difference $\Delta t_{\text{ITD}}$ that these phase-locked spikes must measure:

In 1948, the psychologist Lloyd Jeffress had proposed a theoretical circuit that could convert microsecond time differences into a spatial place code without requiring fast clocks: two opposing bundles of axons entering a linear array of binaural coincidence-detector neurons from opposite ends. When Carr and Konishi injected horseradish peroxidase into axons projecting from the left and right nucleus magnocellularis into the third-order brainstem station, nucleus laminaris, they saw Jeffress's 1948 theoretical diagram realized in living tissue. Axons from the ipsilateral and contralateral ears enter nucleus laminaris from opposite dorsal and ventral surfaces, branching into interdigitating comb-like delay lines with shortened internodal myelin segments that slow conduction velocity down to 3 to 5 meters per second. Each target neuron in nucleus laminaris acts as a sub-millisecond coincidence detector that fires maximally only when spikes from the left and right ears arrive within the same 20-microsecond window—occurring at the exact depth where internal axonal propagation delay ($\tau_{\text{left}} - \tau_{\text{right}}$) cancels external acoustic travel time $\Delta t_{\text{ITD}}$.

Yet on a symmetrical head, both time difference and intensity difference vary along the horizontal plane. The barn owl solves two-dimensional localization through a radical evolutionary distortion of its skull. Behind the heart-shaped acoustic ruff of stiff facial feathers, the owl's external ear openings and bony squamosal flaps are vertically asymmetrical: the left ear aperture sits higher on the skull and points downward at a 15-degree angle below the horizontal plane, while the right ear aperture sits 6 millimeters lower and points upward at 15 degrees. For high frequencies above 4 kilohertz, a mouse directly below the owl sounds up to 20 decibels louder in the left ear than the right; a sound above the horizon is louder in the right ear. By intersecting the microsecond horizontal phase map from nucleus laminaris with the vertical decibel intensity map from nucleus angularis, the midbrain constructs a stereoscopic acoustic image in complete darkness.