Coincidence Detection in the Auditory System 50 Years after Jeffress

for cell 1 and vice versa for the branch from the left The idea that cortical neurons code information not only side. Due to the finite axonal conduction speed, these in their average rate of discharge but also in the temporal branching patterns set up delay lines that run in opposite pattern of spike discharge has generated much recent fashion for the two monaural input channels. We refer to controversy (Shadlen and Newsome, 1998; Stevens and the differences in delay between ipsiand contralateral Zador, 1998). Proponents of time codes often use examsignals at their points of convergence as internal delays. ples in organisms portrayed as being “specialized” for Imagine a sound originating from straight ahead. It temporal processing (e.g., electric fish, barn owl, echoreaches the cochleae without interaural delay (ITD 5 0 locating bats; see Carr, 1993). It is less widely known that ms). Only cell 4 receives coincident input signals and even in “nonspecialized” mammals, including humans, a will be activated, since the total pathlengths traversed well-documented case of information encoding based by its ipsiand contralateral inputs are equal with an on submillisecond spike timing is found in the auditory internal delay of 0 ms. If the tone originates from the left brainstem. hemifield, it reaches the left ear earlier than the right. The role of the time dimension in hearing has been Input signals coincide if the signals from the left have debated for over a century, and the extent to which to travel a longer path length than those from the right perception of certain sound attributes (e.g., pitch; Cariby an amount that exactly offsets the acoustic delay, ani and Delgutte, 1996) depends on fine timing of firing e.g., at cell 1. Thus, the overall result of this scheme is patterns in the auditory nerve is still unsettled. However, the creation of a spatial array of cells, each tuned to a clear example of the importance of the temporal dia specific ITD and arranged orderly according to the mension comes from binaural hearing. The interaural azimuth to which they are tuned. time difference (ITD) between the arrival time of acoustic Modern anatomical and physiological evidence shows energy at the two ears provides an important cue for that the MSO is the likely site of Jeffress’ model. It the spatial location of sound and can be discriminated consists of a long sheet of cells oriented in a roughly at very small values (z10 ms). Physiologically, neurons parasagittal plane. Afferents from bushy cells in the in the medial superior olive (MSO) are sensitive to microanteroventral cochlear nucleus (AVCN) are segregated second differences in their afferent signals. This is one through the cells’ bipolar dendritic morphology: ipsilatof the few sensory circuits in the mammalian CNS for eral afferents terminate on the lateral, contralateral afferwhich a strong functional hypothesis can be formulated ents on the medial dendrites. What, then, is the current and the mechanisms underlying its physiological propevidence in the MSO circuit for the three key assumperties are relatively well understood. tions of the Jeffress model and its main outcome, a In a seminal short paper published exactly 50 years spatial map of ITD? ago, Jeffress (1948) outlined a model to transform acousSynchronization of Discharge Pattern tic timing differences into a neural spatial gradient, in to the Acoustic Stimulus effect transforming a time code into a space code. To Clearly, to encode ITDs, it is necessary that the temporal a remarkable extent, the key assumptions of his model information in the acoustic signal be encoded by the have been borne out, even though they were formulated cochlea, as evidenced by phase locking to low-freat a time when single cell physiology of the brainstem quency tones in auditory nerve fibers (Figure 2A). Recent was in its infancy. The model has guided much subseevidence (Joris et al., 1994) shows that phase locking quent psychophysical and physiological research into in bushy cells is much more precise than in auditory the mechanisms of interaural time comparisons and is nerve; at low frequencies (,1 kHz), spikes lock to a currently the backbone of virtually all computational narrower range of phases and discharge on every stimumodels of sound localization (Colburn, 1996). lus cycle (Figure 2B). This higher phase locking reduces Jeffress made three assumptions (Figure 1A). First, the temporal window over which binaural coincidences monaural channels (“secondary fibers”), which originate can occur and thereby sharpens ITD tuning in the MSO.