Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Nat Neurosci. Author manuscript; available in PMC 2010 November 22.
Published in final edited form as:
PMCID: PMC2989898

Changing tune in auditory cortex


Investigating the organization of tone representation in the rodent auditory cortex at high resolution, two new studies in this issue find that the arrangement of relative frequency responsiveness is not preserved at a fine-scale cortical level.

Staying organized is tough, but usually worth the effort. Even if you don’t follow that maxim, your brain certainly seems to. Neural circuits are wired with marked accuracy, with neurons being capable of contacting select targets with single-cell precision. Some of the most conspicuous and well-studied examples of neuronal organization occur in sensory systems, many of which have highly ordered representations, or maps, of the sensory periphery in respective brain regions. In the visual and somatosensory systems, for example, the relative spatial arrangements of receptors at the periphery, the retina and the body surface, are maintained in cortical representations as retinotopic and somatotopic maps. The fact that smooth and orderly maps of the periphery are preserved throughout the chain of successive nuclei up to primary cortical areas seems to argue for their general usefulness. Indeed, the ‘neighbor preserving’ organization of sensory maps is thought to facilitate circuit operations, such as lateral inhibition, that benefit from relative proximity.

Are such high-fidelity maps of the sensory periphery also the basic organizing principle of auditory cortex? Two studies in this issue say no. Rothschild et al.1 and Bandyopadhyay et al.2 report that the fine-grained topographic representation of frequency or tonotopy, which is established in the cochlea and faithfully maintained throughout all subcortical nuclei, is lost in the primary auditory cortex (A1) (Fig. 1a,b). Although previous work has pointed to a similar conclusion, technical limitations of these earlier studies left room for conflicting interpretations. Experiments using techniques with wide coverage and low spatial resolution, such as microelectrode mapping3 or intrinsic signaling imaging4, generally found evidence of cortical tonotopy. However, recordings from neighboring neurons provided several counter-examples in which even adjacent cells had markedly different tuning properties5,6. Reconciling these discrepant findings is further complicated by the fact that tone-evoked spiking in A1 is sparse7, which exaggerates tonotopy assayed with low-resolution methods and hinders a fine-grained mapping of tonotopy using microelectrode recordings. Thus, settling the question of tonotopy in A1 decisively is a tall order; it requires monitoring activity of a wide area of A1 at single-cell resolution.

Figure 1
Tonotopy and network architecture in primary auditory cortex (A1). (a) Schematic of the classical tonotopic map. There is a smooth increase in the preferred frequency (best frequency, indicated by color) of neurons along the rostrocaudal axis. (b) Tonotopic ...

To do this, Rothschild et al.1 and Bandyopadhyay et al.2 applied in vivo two-photon microscopy, which has been successfully used to map the fine-scale organization of circuits in the visual8, 9 and barrel10,11 cortex, to study tone-evoked activity in A1 of mice. Both groups used similar methods, exposing the surface of auditory cortex in lightly anesthetized mice and bulk-loading membrane-permeable calcium indicators across a relatively wide area. After neurons had taken up the dye, both groups could successfully track sound-evoked calcium transients of up to several dozen neurons simultaneously, providing a live, detailed and panoramic activity map of A1.

Both studies revealed that, unlike the visual cortex, which shows a dense and graded tiling of receptive-field properties9, auditory cortex is more of a mixed bag. First, less than half of all cortical neurons were even tone responsive. Even among the neurons that did respond, responses were unreliable and only a minority had the classical V-shaped tuning curves characteristic of subcortical neurons. Although a loose tonotopic trend was evident on the scale of several tens to hundreds of microns, this tonotopy lacked the smooth, single-cell gradations of tuning observed in visual cortex. In A1, neurons with similar tuning properties were only slightly more likely to be neighbors than neurons with very different tuning properties. Therefore, on a local scale, the tonotopic map in A1 is fractured (Fig. 1b).

Does a lack of smooth global tonotopy imply that frequency is discarded as an organizing principle? Perhaps not. Rothschild et al.1 suggest that, although a cell’s preferred frequency doesn’t tell us where it’s likely to be, it may tell us who it’s connected to. Using an analysis of trial-to-trial variations in sound-evoked activity, Rothschild et al.1 found that nearby neurons with similar frequency tuning also had similar fluctuations in their activity levels. In addition, neurons with similar patterns of driven, sound-evoked discharge also covaried in their spontaneous activity. Although several models of afferent and intracortical connectivity could potentially accommodate these data, the authors’ simulations suggest that they are most parsimoniously explained by the presence of partially overlapping subnetworks in A1.

According to this view, similarly frequency-tuned neurons (which probably receive common thalamic input) are distributed semi-randomly in local patches of A1, and interdigitated with other such populations. In addition to sharing common input, members of each such network may also be strongly and selectively interconnected (Fig. 1c). Although this interpretation awaits verification in the form of paired recordings from putative intra- and internetwork neurons, it is certainly consistent with in vitro mapping studies of visual cortex that show selective synaptic connections between L2/3 pyramidal cells with common L4 inputs12. Investigating the fine-scaled connectivity of subnetworks in A1 will be a fascinating area of future research and one greatly facilitated by Rothschild et al.1’s study.

Even though tonotopy is not the major organizing principle of A1, it is still interesting to ask why it is absent. In fact, knowing exactly how the expected tonotopic map in A1 becomes scrambled may tell us the type of sensory transformation occurring there. One possibility is that, despite the orderly, tonotopic arrangement of neurons in the auditory thalamus, thalamocortical axons become scattered en-route to the cortex, establishing patchy and diffuse isofrequency domains in A1 such that neighboring cortical neurons receive mixed frequency inputs. Alternatively, thalamocortical projections may be arranged tonotopically, with tonotopy becoming fractured by intracortical processing.

To differentiate between these two possibilities, Bandyopadhyay et al.2 mapped A1 using two calcium indicators with different affinities, which allowed them to differentiate between subthreshold and suprathreshold contributions to postsynaptic calcium responses. The low-affinity dye (Fluo-4) detected only large calcium fluxes associated with spikes, whereas the high-affinity dye (OGB-1) could detect spikes and smaller fluctuations in calcium that result from barrages of subthreshold synaptic input. Thus, Bandyopadhyay et al.2 had a tool for determining whether the fragmented organization of Al is the result of heterogeneity of input organization or postsynaptic processing.

Using a clustering analysis that assessed the similarity of tone or noise-evoked neuronal responses in small neighborhoods of A1, Bandyopadhyay et al.2 found that subthreshold input maps to A1 were considerably more organized than output maps that were based on spiking alone. Input maps were tiled with cleanly separable clusters of similarly responsive neurons, whereas output maps lacked any obvious organization (Fig. 1d). This result suggests that relative disorder in A1 in the form of its spike output exists despite the presence of orderly input.

So how does this disorder arise? One possibility, suggested by Bandyopadhyay et al.2, is that neighboring neurons in A1 carry out different operations in parallel; although adjacent cells receive similar, correlated inputs, they may be members of separate fine-scaled assemblies that process these inputs differently. Such parallel processing streams, which are a common theme of the ascending auditory pathway, may be selective for different input features or specific stimulus attributes. This parcellation into parallel streams in A1 could arise from differences in intrinsic properties of cortical neurons, such as morphology or ion channel expression. In addition, different parallel streams may recruit different neuronal populations with different interconnectivity and possibly different synaptic strengths and dynamics. Notably, this latter view is cofnsistent with the subnetwork model supported by Rothschild et al.1.

Taken together, these studies clearly show that frequency is not the most prominent motif that defines the functional organization of A1. Furthermore, Bandyopadhyay et al.2 found a lack of organization on the basis of bandwidth or intensity tuning, making it unlikely that any set of simple sound parameters is topographically mapped onto the surface of A1. More likely, A1 neurons encode complex and ethologically meaningful stimuli. This view is supported by physiological studies showing that A1 neurons are best driven by spectrally and temporally rich stimuli13, including conspecific voice calls14, and can discern between complex sound sequences. In fact, it has been proposed that A1 may represent higher-order auditory objects, serving a role analogous to that of the inferior temporal cortex in the visual system15.

Although we should proceed carefully before generalizing these results to all mammalian species and unanesthetized preparations, Rothschild et al.1 and Bandyopadhyay et al.2 have pointed the way toward a more refined understanding of A1’s functional architecture. Perhaps more importantly, they have also opened a new avenue for addressing questions of auditory coding that move beyond cortical cartography. The time is now ripe to answer the questions of what exactly auditory cortex represents and how such representations are built out of spatiotemporal assemblies of A1 neurons. The future will be exciting. Keep your ears to the ground and stay tuned.



The authors declare no competing financial interests.


1. Rothschild G, Nelken I, Mizrahi A. Nat Neurosci. 2010;13:353–360. [PubMed]
2. Bandyopadhyay S, Shamma SA, Kanold PO. Nat Neurosci. 2010;13:361–368. [PMC free article] [PubMed]
3. Merzenich MM, Knight PL, Roth GL. J Neurophysiol. 1975;38:231–249. [PubMed]
4. Nelken I, et al. J Neurophysiol. 2004;92:2574–2588. [PubMed]
5. Evans EF, Ross HF, Whitfield IC. J Physiol (Lond) 1965;179:238–247. [PubMed]
6. Goldstein MH, Jr, Abeles M, Daly RL, Mcintosh J. J Neurophysiol. 1970;33:188–197. [PubMed]
7. Hromádka T, Deweese MR, Zador AM. PLoS Biol. 2008;6:e16. [PMC free article] [PubMed]
8. Ohki K, Chung S, Ch’ng YH, Kara P, Reid RC. Nature. 2005;433:597–603. [PubMed]
9. Ohki K, et al. Nature. 2006;442:925–928. [PubMed]
10. Kerr JN, et al. J Neurosci. 2007;27:13316–13328. [PubMed]
11. Sato TR, Gray NW, Mainen ZF, Svoboda K. PLoS Biol. 2007;5:189. [PMC free article] [PubMed]
12. Yoshimura Y, Dantzker JL, Callaway EM. Nature. 2005;433:868–873. [PubMed]
13. Sadagopan S, Wang X. J Neurosci. 2009;29:11192–11202. [PMC free article] [PubMed]
14. Wang X, Kadia SC. J Neurophysiol. 2001;86:2616–2620. [PubMed]
15. King AJ, Nelken I. Nat Neurosci. 2009;12:698–701. [PubMed]