Codes for Sound-Source Location in Nontonotopic Auditory Cortex

Abstract

We evaluated two hypothetical codes for sound-source location in the auditory cortex. The topographical code assumed that single neurons are selective for particular locations and that sound-source locations are coded by the cortical location of small populations of maximally activated neurons. The distributed code assumed that the responses of individual neurons can carry information about locations throughout 360 degrees of azimuth and that accurate sound localization derives from information that is distributed across large populations of such panoramic neurons. We recorded from single units in the anterior ectosylvian sulcus area (area AES) and in area A2 of alpha-chloralose-anesthetized cats. Results obtained in the two areas were essentially equivalent. Noise bursts were presented from loudspeakers spaced in 20 degrees intervals of azimuth throughout 360 degrees of the horizontal plane. Spike counts of the majority of units were modulated >50% by changes in sound-source azimuth. Nevertheless, sound-source locations that produced greater than half-maximal spike counts often spanned >180 degrees of azimuth. The spatial selectivity of units tended to broaden and, often, to shift in azimuth as sound pressure levels (SPLs) were increased to a moderate level. We sometimes saw systematic changes in spatial tuning along segments of electrode tracks as long as 1.5 mm but such progressions were not evident at higher sound levels. Moderate-level sounds presented anywhere in the contralateral hemifield produced greater than half-maximal activation of nearly all units. These results are not consistent with the hypothesis of a topographic code. We used an artificial-neural-network algorithm to recognize spike patterns and, thereby, infer the locations of sound sources. Network input consisted of spike density functions formed by averages of responses to eight stimulus repetitions. Information carried in the responses of single units permitted reasonable estimates of sound-source locations throughout 360 degrees of azimuth. The most accurate units exhibited median errors in localization of <25 degrees, meaning that the network output fell within 25 degrees of the correct location on half of the trials. Spike patterns tended to vary with stimulus SPL, but level-invariant features of patterns permitted estimates of locations of sound sources that varied through 20-dB ranges. Sound localization based on spike patterns that preserved details of spike timing consistently was more accurate than localization based on spike counts alone. These results support the hypothesis that sound-source locations are represented by a distributed code and that individual neurons are, in effect, panoramic localizers.

Cortico‐cortical connectivity within ferret auditory cortex

“…The anterior bank of the PSSC projects to the superior colliculus (Bajo et al, 2010a), and is likely to be homologous to the anterior ectosylvian sulcal field (fAES) of the cat (Jiang et al, 1996; Manger et al, 2005). In accordance with the suggestion that cat fAES may be specialized for spatial processing (Middlebrooks et al, 1998; Malhotra and Lomber, 2007; Las et al, 2008), it has been shown that inputs to the aPSSC from the visual cortex arise predominantly from area SSY (Bizley et al, 2007a), which is thought to be specialized for visual motion processing (Philipp et al, 2006). Previous studies of the PSSC have not demarcated the anterior and posterior banks.…”

Cortical codes for sound localization

“…Spatial profiles of units can show two or more discrete peaks separated by welldefined valleys, as seen in Figs. 1A and B. Middlebrooks and colleagues [18] found that ∼ 20% units in area A2 and AES had two or more peaks.…”

“…Spatial sensitivity tends to be broad. For instance, 58 to 97% of units in areas A2 and AES responded with more than 50% of their maximum spike rates across receptive fields spanning 180 • or greater [18]; the exact percentage of units varied with cortical area and sound level. A similar breadth of sensitivity is reported for the majority of units in area A1 [13,15,16].…”

“…In the cat, the spatial sensitivity of neurons has been studied in the primary auditory cortex (area A1) [13][14][15][16], area A2 [17][18][19], the anterior ectosylvian area (area AES) [17][18][19][20], and to a limited degree in the anterior auditory area (area AAF) [20]. There are good reasons to assume that cortical areas A1, A2, and AES are involved in sound localization.…”

“…Spatial sensitivity varies among units and, to a lesser degree, among cortical areas, but one can make some generalizations. The majority of cortical units shows greater than 50% modulation of their spike counts by the location of the sound source in azimuth (area A1 [14,15]; areas A2 and AES [18]). Spatial sensitivity tends to be broad.…”