DOI: 10.1126/science.280.5368.1439 , 1439 (1998); 280Science et al.R. Christopher deCharms,Optimizing Sound Features for Cortical Neurons www.sciencemag.org (this information is current as of January 5, 2007 ):The following resources related to this article are available online at http://www.sciencemag.org/cgi/content/full/280/5368/1439version of this article at: including high-resolution figures, can be found in the onlineUpdated information and services, http://www.sciencemag.org/cgi/content/full/280/5368/1439#otherarticles, 30 of which can be accessed for free: cites 62 articlesThis article 135 article(s) on the ISI Web of Science. cited byThis article has been http://www.sciencemag.org/cgi/content/full/280/5368/1439#otherarticles 48 articles hosted by HighWire Press; see: cited byThis article has been http://www.sciencemag.org/cgi/collection/neuroscienceNeuroscience : subject collectionsThis article appears in the following http://www.sciencemag.org/help/about/permissions.dtl in whole or in part can be found at: this articlepermission to reproduce of this article or about obtaining reprintsInformation about obtaining registered trademark of AAAS. c 2006 by the American Association for the Advancement of Science; all rights reserved. The title SCIENCE is a CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. Science (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the on January 5, 2007 www.sciencemag.orgDownloaded fromOptimizing Sound Features for Cortical NeuronsR. Christopher deCharms, David T. Blake,Michael M. Merzenich*The brain’s cerebral cortex decomposes visual images into information about orientededges, direction and velocity information, and color. How does the cortex decomposeperceived sounds? A reverse correlation technique demonstrates that neurons in theprimary auditory cortex of the awake primate have complex patterns of sound-featureselectivity that indicate sensitivity to stimulus edges in frequency or in time, stimulustransitions in frequency or intensity, and feature conjunctions. This allows the creationof classes of stimuli matched to the processing characteristics of auditory corticalneurons. Stimuli designed for a particular neuron’s preferred feature pattern can drivethat neuron with higher sustained firing rates than have typically been recorded withsimple stimuli. These data suggest that the cortex decomposes an auditory scene intocomponent parts using a feature-processing system reminiscent of that used for thecortical decomposition of visual images.Feature processing by neurons in the pri-mary visual cortex has been studied exten-sively since the discovery by Hubel and Wie-sel that the “right” kinds of stimuli for visualcortical neurons are moving, oriented barswithin a spatial receptive field (1)—a find-ing later confirmed and extended in detail byreverse-correlation methods (2). The funda-mental feature-processing characteristicswithin the spectral receptive fields of neu-rons in the awake primary auditory cortexare less completely resolved, except in spe-cies with particularly well understood audi-tory behavior, such as bats (3), owls (4), andsong-birds (5), where stimuli have been se-lected on the basis of ethological principles(6). Our understanding of auditory cortexphysiology in other mammalian species re-sults largely from studies of anesthetized an-imals (7), which has demonstrated that au-ditory neurons are tuned for a number ofindependent feature parameters of simplestimuli including frequency (8), intensity(9), amplitude modulation (10), frequencymodulation (11), and binaural structure(12). However, auditory responses to multi-ple stimuli can also enhance or suppress oneanother in a time-dependent manner (13),and auditory cortical neurons can be highlyselective for species-specific vocalizations(14), indicating complex acoustic processingby these cells. It is not yet known if or howthese many independent selectivities of au-ditory cortical neurons reflect an underlyingpattern of feature decomposition, as has beensuggested (15). Further, because sustainedfiring-rate responses in the auditory cortex totonal stimuli are typically much lower thanvisual responses to drifting bars (16), it hasbeen suggested that the preferred type ofauditory stimulus may still not be known(17). For these reasons, we investigatedwhether a reverse-correlation method simi-lar to that used in the visual cortex woulddiscern the full spectral and temporal fea-ture-response profile of auditory corticalneurons in the awake animal.Reverse correlation stimuli used in visualcortex experiments have consisted of rapid-ly presented two-dimensional checkerboardspatial patterns (Fig. 1A). The auditory stim-uli used here consisted of rapidly presentedrandom chords or asynchronous randomtone progressions that evenly spanned a por-tion of the one-dimensional receptor surfaceof the cochlea (Fig. 1, B and C). Rather thanmeasuring tuning for a preselected parame-ter, this unbiased method constructs the av-erage full auditory stimulus preceding spikesfrom a neuron, whatever form it may take(18). Alternatively, this can be viewed as theaverage response of the neuron driven byeach separate stimulus component, which isnumerically identical (but is reversed intime, and is expressed in units of mean den-sity of stimulus components instead of themean density of spikes). Neuronal reverse-correlation techniques were originally devel-oped for characterizing auditory neurons inthe periphery (19), but while these methodshave been applied to more peripheral struc-tures and visual cortical cells, researchershave only recently begun to succeed in ap-plying this method to auditory cortical neu-rons (20). Data presented here are takenfrom extensive characterizations of 206 iso-Keck Center for Integrative Neuroscience, University ofCalifornia, San Francisco, San Francisco, CA 94143–0732, USA.*To whom correspondence should be addressed. E-mail:[email protected] Cortex: Reverse CorrelationUsing 2D Visual Patterns in TimeAuditory Cortex: Reverse CorrelationUsing 1D Auditory Patterns (Chords) in TimeXSpatiotemporal Receptive FieldXSpike TrainsSpectrotemporal Receptive FieldAuditory Cortex: Reverse CorrelationUsing Asynchronous Poisson Tone TrainsXSpectrotemporal Receptive Fieldtime time time t = 0 mst = 0 ms t = 0 mst = 200 ms t = 200 mst = 400 ms t = 400 mst = 20