- Email: [email protected]
Sound analysis in auditory cortex
Update
229
TRENDS in Neurosciences Vol.26 No.5 May 2003
| Research News
Sound analysis in auditory cortex Robert J. Zatorre Montreal Neurological Institute, McGill University, 3801 University Street, Montreal, Quebec, Canada H3A 2B4
A recent paper by Seifritz and collaborators [1] provides an excellent example of the sophistication that analysis of the auditory cortex is now experiencing. These scientists provide evidence that distinct subareas of human auditory cortex have different temporal response properties – a feature that could prove crucial in understanding how the auditory nervous system temporally decomposes and represents complex incoming sound waves. It is important to recall that auditory cognition must solve a tricky computational problem: all incoming input is combined into a single dimension. That is, as far as the auditory system is concerned, the entire universe of sound must be decoded from the inward and outward movements of two small membranes (the ear drums). Moreover, all sound unfolds over time. These constraints mean that temporal properties assume a tremendous importance. Hence, a great deal of the hardware, both in the ascending auditory neuraxis and at the cortical level, is concerned with extracting and recoding information that is contained within the temporal waveform. Seifritz and colleagues made use of functional magneticresonance imaging (fMRI) to measure blood oxygenation responses (an indirect index of brain activity) and independent-component analysis to identify the temporal and spatial properties of the signal. The latter technique can be characterized as a data-driven multivariate statistical approach that allows one to infer the presence of multiple independent components in a given data set. This approach is an interesting one that is increasingly being applied to neuroimaging data; its advantage is that it reveals patterns of activity that can not be predicted and, hence, serves an excellent role as an exploratory heuristic tool. The end result of this analytical process was the Corresponding author: Robert J. Zatorre ([email protected]). http://tins.trends.com
identification of two distinct response types within the auditory cortex: one showing a sustained response that continues for the duration of sound stimulation, the other showing a transient response peaking at about five seconds after stimulus onset and then returning to zero. Further, these two response types appear to be spatially segregated within the auditory cortex: the sustained response localizes primarily to the area of Heschl’s gyrus, which contains core auditory cortex, whereas the transient responses are distributed surrounding the core areas (Fig. 1). This finding represents the beginning of new avenues of research. Having described this pattern of brain activity, many new questions arise. In particular, it will be crucial that work in the near future build on these findings, by trying to understand how the sustained and transient fields relate on the
Right temporal lobe
Left temporal lobe
Heschl's gyri Anterior poles
Transient response 100 60 20
Sustained response 60 100 (%)
100 BOLD (%)
Not so long ago, the auditory cortex took a back seat to the visual system in neuroscience research. With some notable exceptions outside the primate order, such as the classic work on echo-locating bats, owls and birdsong, the auditory cortex has been overlooked: only a few investigators were involved in understanding the structure and function of the monkey auditory cortex, and even fewer had the means to study its human counterpart. This situation has undergone a dramatic change in the past decade or so. Spurred on by advances in primate neurophysiology and neuroanatomy, and especially by developments in functional neuroimaging, substantial progress is now being made into understanding how the human auditory cortical system works.
Transient predictors
50
Sustained predictors
0 0
20 40 Time (s)
60
0
20 40 Time (s)
60
TRENDS in Neurosciences
Fig. 1. Sustained and transient activity patterns in auditory cortex. Top: the relative-contribution map of transient and sustained blood-oxygen-level-dependent (BOLD) signal sources in the auditory cortex, projected on the reconstructed cortical surface of the temporal lobes of a standard brain template. Color-coding indicates the relative contribution of the two predictor classes and suggests a spatial continuum between the temporal response patterns. The contribution of the sustained response type becomes less predominant as one moves from the core to the belt areas. Bottom: signals represent intra-individual averages of the trials used as predictors within a group multiple-regression analysis. Reproduced, with permission, from Ref. [1], q 2002 American Association for the Advancement of Science (http://www.sciencemag.org).
230
Update
TRENDS in Neurosciences Vol.26 No.5 May 2003
one hand to neurophysiological and neuroanatomical phenomena, and on the other hand to cognitive neuroscience phenomena. Neuroanatomy and neurophysiology With respect to neuroanatomy, recent attention has focused on the parcellation of the auditory cortex into various subfields, based on their cytoarchitecture and connectivity, leading to the identification of core, belt and parabelt regions that are arranged in a concentric fashion on the superior temporal plane [2,3]. The correspondence between monkey and human auditory cortices has also begun to be worked out, and indications are that a similar organization underlies both [4]. Hence, one important question is how the sustained-versus-transient responses seen by Seifritz et al. fit in to the core-versus-belt anatomical distinctions. Although one might be tempted to assume a close fit between functional and structural properties (e.g. such that sustained responses are confined to the core area), the situation is likely to be more complicated. Opportunities abound for clarification of this important issue. With respect to neurophysiology, the issues are, if anything, even more complex. A good deal of progress has been made recently in characterizing the response properties of auditory neurons in the monkey brain. Many investigators have begun to explore these issues using awake preparations and looking at correlations with behavior [5]; others have used sophisticated data analysis tools to extract information contained in interspike interval values [6]; still others have used ecologically meaningful stimuli to elucidate complex response properties that might not be obvious with artificial stimuli [7]. It is an exciting time in the field, with new information appearing regularly and generating both interest and debate. It is still too early to know just how the discovery of the different fMRI signals observed by Seifretz et al. fits into the neurophysiological picture. What is clear is that there is not a one-to-one mapping of transient-versussustained blood oxygenation signals to transient or sustained single-unit responses. For one thing, the time scales are at least an order of magnitude different, because of the hemodynamic delay involved. Moreover, complex interactions between neurons, including habituation and inhibitory processes, contribute to the global changes in hemodynamic signal to which fMRI is sensitive. One intriguing question is how (or whether) the different classes of fMRI response might be related to different classes of neurons that either are synchronized to the fine structure of the stimulus or exhibit nonsynchronous rate-coding properties [8]. Cognitive neuroimaging Finally, the data concerning distinct temporal processing characteristics must also be tied in to results from cognitive neuroscience and neuroimaging. Several prior investigations, notably those using magnetoencephalography, have identified the existence of steady-state responses that are distinct from transient ones [9] but, http://tins.trends.com
once again, it is not known how these physiological signatures relate to the hemodynamic changes seen by Seifritz and colleagues. To make advances that would cut across these methodologies, it would be useful for scientists working on these problems to get together and attempt to relate the findings from one technique directly to those from the other. Using fMRI or positron emission tomography, several investigators have also identified cortical areas that respond to stimuli that vary systematically in their temporal properties [10 – 12] – findings that, generally speaking, support the idea of hierarchical processing. The sustainedversus-transient dichotomy has yet to be integrated with these results, but the topographical distribution distinguishing between the core and surrounding belt regions certainly fits with the idea that these regions have distinct functional characteristics. Another aspect of human auditory cortex is that it exhibits hemispheric specialization [13]. Functional asymmetries were not observed by Seifritz et al., however, raising the additional question of whether the sustained-versustransient effects are truly bilaterally symmetric, or whether there are more subtle differences in temporal response that await discovery. All of these phenomena must ultimately be understood in a broader context. They will hopefully be integrated into computational models that help to explain how elementary features of neural signals, such as those discussed here, shape the overall neural responses that underlie perception and behavior. References 1 Seifritz, E. et al. (2002) Spatiotemporal pattern of neural processing in the human auditory cortex. Science 297, 1706 – 1708 2 Kaas, J.H. et al. (1999) Auditory processing in primate cerebral cortex. Curr. Opin. Neurobiol. 9, 164– 170 3 Romanski, L. et al. (1999) Dual streams of auditory afferents target multiple domains in the primate prefrontal cortex. Nat. Neurosci. 2, 1131– 1136 4 Rauschecker, J. (1998) Cortical processing of complex sounds. Curr. Opin. Neurobiol. 8, 516 – 521 5 Recanzone, G.H. et al. (2000) Correlation between the activity of single auditory cortical neurons and sound-localization behavior in the macaque monkey. J. Neurophysiol. 83, 2723 – 2739 6 Furukawa, S. et al. (2000) Coding of sound-source location by ensembles of cortical neurons. J. Neurosci. 20, 1216 – 1228 7 Tian, B. et al. (2001) Functional specialization in rhesus monkey auditory cortex. Science 292, 290– 293 8 Lu, T. et al. (2001) Temporal and rate representations of time-varying signals in the auditory cortex of awake primates. Nat. Neurosci. 4, 1131– 1138 9 Ross, B. et al. (2002) Temporal integration in the human auditory cortex as represented by the development of the steady-state magnetic field. Hearing Res. 165, 68– 84 10 Giraud, A-L. et al. (2000) Representation of the temporal envelope of sounds in the human brain. J. Neurophysiol. 84, 1588 – 1598 11 Patterson, R.D. et al. (2002) The processing of temporal pitch and melody information in auditory cortex. Neuron 36, 767 – 776 12 Zatorre, R.J. and Belin, P. (2001) Spectral and temporal processing in human auditory cortex. Cereb. Cortex 11, 946 – 953 13 Zatorre, R.J. et al. (2002) Structure and function of auditory cortex: music and speech. Trends Cogn. Sci. 6, 37– 46 0166-2236/03/$ - see front matter q 2003 Elsevier Science Ltd. All rights reserved. doi:10.1016/S0166-2236(03)00074-2
229
TRENDS in Neurosciences Vol.26 No.5 May 2003
| Research News
Sound analysis in auditory cortex Robert J. Zatorre Montreal Neurological Institute, McGill University, 3801 University Street, Montreal, Quebec, Canada H3A 2B4
A recent paper by Seifritz and collaborators [1] provides an excellent example of the sophistication that analysis of the auditory cortex is now experiencing. These scientists provide evidence that distinct subareas of human auditory cortex have different temporal response properties – a feature that could prove crucial in understanding how the auditory nervous system temporally decomposes and represents complex incoming sound waves. It is important to recall that auditory cognition must solve a tricky computational problem: all incoming input is combined into a single dimension. That is, as far as the auditory system is concerned, the entire universe of sound must be decoded from the inward and outward movements of two small membranes (the ear drums). Moreover, all sound unfolds over time. These constraints mean that temporal properties assume a tremendous importance. Hence, a great deal of the hardware, both in the ascending auditory neuraxis and at the cortical level, is concerned with extracting and recoding information that is contained within the temporal waveform. Seifritz and colleagues made use of functional magneticresonance imaging (fMRI) to measure blood oxygenation responses (an indirect index of brain activity) and independent-component analysis to identify the temporal and spatial properties of the signal. The latter technique can be characterized as a data-driven multivariate statistical approach that allows one to infer the presence of multiple independent components in a given data set. This approach is an interesting one that is increasingly being applied to neuroimaging data; its advantage is that it reveals patterns of activity that can not be predicted and, hence, serves an excellent role as an exploratory heuristic tool. The end result of this analytical process was the Corresponding author: Robert J. Zatorre ([email protected]). http://tins.trends.com
identification of two distinct response types within the auditory cortex: one showing a sustained response that continues for the duration of sound stimulation, the other showing a transient response peaking at about five seconds after stimulus onset and then returning to zero. Further, these two response types appear to be spatially segregated within the auditory cortex: the sustained response localizes primarily to the area of Heschl’s gyrus, which contains core auditory cortex, whereas the transient responses are distributed surrounding the core areas (Fig. 1). This finding represents the beginning of new avenues of research. Having described this pattern of brain activity, many new questions arise. In particular, it will be crucial that work in the near future build on these findings, by trying to understand how the sustained and transient fields relate on the
Right temporal lobe
Left temporal lobe
Heschl's gyri Anterior poles
Transient response 100 60 20
Sustained response 60 100 (%)
100 BOLD (%)
Not so long ago, the auditory cortex took a back seat to the visual system in neuroscience research. With some notable exceptions outside the primate order, such as the classic work on echo-locating bats, owls and birdsong, the auditory cortex has been overlooked: only a few investigators were involved in understanding the structure and function of the monkey auditory cortex, and even fewer had the means to study its human counterpart. This situation has undergone a dramatic change in the past decade or so. Spurred on by advances in primate neurophysiology and neuroanatomy, and especially by developments in functional neuroimaging, substantial progress is now being made into understanding how the human auditory cortical system works.
Transient predictors
50
Sustained predictors
0 0
20 40 Time (s)
60
0
20 40 Time (s)
60
TRENDS in Neurosciences
Fig. 1. Sustained and transient activity patterns in auditory cortex. Top: the relative-contribution map of transient and sustained blood-oxygen-level-dependent (BOLD) signal sources in the auditory cortex, projected on the reconstructed cortical surface of the temporal lobes of a standard brain template. Color-coding indicates the relative contribution of the two predictor classes and suggests a spatial continuum between the temporal response patterns. The contribution of the sustained response type becomes less predominant as one moves from the core to the belt areas. Bottom: signals represent intra-individual averages of the trials used as predictors within a group multiple-regression analysis. Reproduced, with permission, from Ref. [1], q 2002 American Association for the Advancement of Science (http://www.sciencemag.org).
230
Update
TRENDS in Neurosciences Vol.26 No.5 May 2003
one hand to neurophysiological and neuroanatomical phenomena, and on the other hand to cognitive neuroscience phenomena. Neuroanatomy and neurophysiology With respect to neuroanatomy, recent attention has focused on the parcellation of the auditory cortex into various subfields, based on their cytoarchitecture and connectivity, leading to the identification of core, belt and parabelt regions that are arranged in a concentric fashion on the superior temporal plane [2,3]. The correspondence between monkey and human auditory cortices has also begun to be worked out, and indications are that a similar organization underlies both [4]. Hence, one important question is how the sustained-versus-transient responses seen by Seifritz et al. fit in to the core-versus-belt anatomical distinctions. Although one might be tempted to assume a close fit between functional and structural properties (e.g. such that sustained responses are confined to the core area), the situation is likely to be more complicated. Opportunities abound for clarification of this important issue. With respect to neurophysiology, the issues are, if anything, even more complex. A good deal of progress has been made recently in characterizing the response properties of auditory neurons in the monkey brain. Many investigators have begun to explore these issues using awake preparations and looking at correlations with behavior [5]; others have used sophisticated data analysis tools to extract information contained in interspike interval values [6]; still others have used ecologically meaningful stimuli to elucidate complex response properties that might not be obvious with artificial stimuli [7]. It is an exciting time in the field, with new information appearing regularly and generating both interest and debate. It is still too early to know just how the discovery of the different fMRI signals observed by Seifretz et al. fits into the neurophysiological picture. What is clear is that there is not a one-to-one mapping of transient-versussustained blood oxygenation signals to transient or sustained single-unit responses. For one thing, the time scales are at least an order of magnitude different, because of the hemodynamic delay involved. Moreover, complex interactions between neurons, including habituation and inhibitory processes, contribute to the global changes in hemodynamic signal to which fMRI is sensitive. One intriguing question is how (or whether) the different classes of fMRI response might be related to different classes of neurons that either are synchronized to the fine structure of the stimulus or exhibit nonsynchronous rate-coding properties [8]. Cognitive neuroimaging Finally, the data concerning distinct temporal processing characteristics must also be tied in to results from cognitive neuroscience and neuroimaging. Several prior investigations, notably those using magnetoencephalography, have identified the existence of steady-state responses that are distinct from transient ones [9] but, http://tins.trends.com
once again, it is not known how these physiological signatures relate to the hemodynamic changes seen by Seifritz and colleagues. To make advances that would cut across these methodologies, it would be useful for scientists working on these problems to get together and attempt to relate the findings from one technique directly to those from the other. Using fMRI or positron emission tomography, several investigators have also identified cortical areas that respond to stimuli that vary systematically in their temporal properties [10 – 12] – findings that, generally speaking, support the idea of hierarchical processing. The sustainedversus-transient dichotomy has yet to be integrated with these results, but the topographical distribution distinguishing between the core and surrounding belt regions certainly fits with the idea that these regions have distinct functional characteristics. Another aspect of human auditory cortex is that it exhibits hemispheric specialization [13]. Functional asymmetries were not observed by Seifritz et al., however, raising the additional question of whether the sustained-versustransient effects are truly bilaterally symmetric, or whether there are more subtle differences in temporal response that await discovery. All of these phenomena must ultimately be understood in a broader context. They will hopefully be integrated into computational models that help to explain how elementary features of neural signals, such as those discussed here, shape the overall neural responses that underlie perception and behavior. References 1 Seifritz, E. et al. (2002) Spatiotemporal pattern of neural processing in the human auditory cortex. Science 297, 1706 – 1708 2 Kaas, J.H. et al. (1999) Auditory processing in primate cerebral cortex. Curr. Opin. Neurobiol. 9, 164– 170 3 Romanski, L. et al. (1999) Dual streams of auditory afferents target multiple domains in the primate prefrontal cortex. Nat. Neurosci. 2, 1131– 1136 4 Rauschecker, J. (1998) Cortical processing of complex sounds. Curr. Opin. Neurobiol. 8, 516 – 521 5 Recanzone, G.H. et al. (2000) Correlation between the activity of single auditory cortical neurons and sound-localization behavior in the macaque monkey. J. Neurophysiol. 83, 2723 – 2739 6 Furukawa, S. et al. (2000) Coding of sound-source location by ensembles of cortical neurons. J. Neurosci. 20, 1216 – 1228 7 Tian, B. et al. (2001) Functional specialization in rhesus monkey auditory cortex. Science 292, 290– 293 8 Lu, T. et al. (2001) Temporal and rate representations of time-varying signals in the auditory cortex of awake primates. Nat. Neurosci. 4, 1131– 1138 9 Ross, B. et al. (2002) Temporal integration in the human auditory cortex as represented by the development of the steady-state magnetic field. Hearing Res. 165, 68– 84 10 Giraud, A-L. et al. (2000) Representation of the temporal envelope of sounds in the human brain. J. Neurophysiol. 84, 1588 – 1598 11 Patterson, R.D. et al. (2002) The processing of temporal pitch and melody information in auditory cortex. Neuron 36, 767 – 776 12 Zatorre, R.J. and Belin, P. (2001) Spectral and temporal processing in human auditory cortex. Cereb. Cortex 11, 946 – 953 13 Zatorre, R.J. et al. (2002) Structure and function of auditory cortex: music and speech. Trends Cogn. Sci. 6, 37– 46 0166-2236/03/$ - see front matter q 2003 Elsevier Science Ltd. All rights reserved. doi:10.1016/S0166-2236(03)00074-2