Dispatch 955

Sensory maps: Aligning maps of visual and auditory spaceAdrian Rees

Visual and auditory space are representedtopographically in the superior colliculus in the midbrain.Recent experiments show that N-methyl-D-aspartate(NMDA) receptors play an important role in aligningauditory and visual maps during development.

Address: Department of Physiological Sciences, The Medical School,University of Newcastle upon Tyne, Newcastle upon Tyne NE2 4HH, UK.

Current Biology 1996, Vol 6 No 8:955–958

© Current Biology Ltd ISSN 0960-9822

Survival, for predator or prey, may turn on the fraction of asecond it takes an animal to centre its gaze on the source ofa sound or movement detected in its periphery. Central tothe performance of this task is a nucleus in the midbraincalled, in mammals, the superior colliculus. This structurecombines sensory information with signals from motorareas and computes an output to control the eye, head,body and outer-ear (pinna) movements needed to bringthe animal’s attention to bear on an object of interest (see[1] for review). The superior colliculus includes topo-graphic representations — maps — of visual and auditoryspace, and recent results have shed new light on the mech-anisms that operate to align these two representationsduring an animal’s early life.

Maps of sensory space in the superior colliculusNeurons in the superficial layers in each of the twosuperior colliculi respond to visual stimuli, whereas cells inthe deeper layers respond to auditory, visual and tactileinputs. The receptive field positions for each modality arerepresented topographically in the superior colliculus:stimulus locations in the azimuthal (horizontal) plane arerepresented along the rostrocaudal axis, whereas stimuluselevation (the vertical plane) is represented along themediolateral axis (Fig. 1). Thus, the maps are in register[2–4]. Neurons that receive inputs from motor structuresare also to be found in the deep layers, as are the neuronsthat send control signals to a variety of motor nuclei [5].These topographic representations may have evolvedbecause they are computationally efficient: neighbouringregions of space that will be occupied sequentially by amoving stimulus are linked by the shortest and, therefore,the fastest connections. This arrangement will also makeit easier to relate corresponding regions of space repre-sented by the different modalities, and facilitate theorganization of the motor outputs.

It is important to note, however, that while the concept of amap provides a convenient and helpful analogy for describ-ing the organization of the colliculus, it has limitations. Itis not the case that each auditory neuron in the superior

Figure 1

The responses of neurons in the superiorcolliculus (SC) to visual and auditory stimuliare organized topographically as shown herein the ferret. Neurons in the superficial layersencode the position of visual stimuli, whereasthe maximal responses of auditory neurons inthe underlying deep layers vary with theposition of sounds in the contralateralhemifield. Stimulus azimuth (red arrow) isrepresented along the rostrocaudal axis of theSC, and elevation (blue arrow) across itsmediolateral axis. IC, inferior colliculus (basedon data from [4]).

colliculus responds only to stimuli located at one positioncovering a few degrees of space. Although there is somevariation across species, the receptive fields of many supe-rior colliculus neurons can extend over the whole of thecontralateral hemifield, particularly for sounds that arewell above threshold. A systematic map of auditory spacein the superior colliculus is evident only from theresponses of neurons firing close to their maximal rate[2,4,6]. This has led to the suggestion that informationabout each position in space is encoded in the relativeactivity of an ensemble of neurons [7,8].

The auditory map is dynamic, not fixed. Many animalscan move their eyes independently of their heads, andsome can move their ears too. Unless such movements aretaken into account, the auditory and visual maps will beout of alignment, except when the eyes and ears are point-ing straight ahead. Studies in awake, behaving monkeysand cats show that the coordinates of the auditory map inthe superior colliculus do change when the animal’s eyesmove, and this, at least partially, restores the alignmentbetween the auditory and visual maps [4,9]. An ensemblecode of auditory space, in which some neurons respondover a wide range of spatial locations, might be moreeffective than a discrete point-to-point mapping forkeeping the auditory and visual maps in register underthese dynamic conditions.

Despite the similarity in the way that auditory and visualstimuli are represented in the superior colliculus, informa-tion about spatial position in these two sensory pathwaysis determined according to completely different princi-ples. The position of a stimulus in the visual field is pro-jected directly onto the two-dimensional receptor array inthe retina. By contrast, in the inner ear it is sound fre-quency, not position, that is represented along the array ofsensory hair cells. A sound’s location is computed withinthe auditory system by comparing differences in the timeand intensity of sounds at the two ears, some of which area result of position-dependent spectral filtering propertiesof the pinnae. In mammals, fibres carrying ascending audi-tory input to the superior colliculus originate in the neigh-bouring inferior colliculus, the main midbrain nucleus inthe auditory pathway, with some input from other brainstem sites.

The development of maps in the superior colliculusBecause sound localization depends largely on interauraldifferences determined by the dimensions of the head andears, the cues representing a particular position in spacewill change as the animal’s head and ears grow. Studies inanimals that are born either in an immature state (ferret) orin a mature state (guinea pig) show that, in contrast to thevisual map, which is well organized early in development,the earliest auditory responses measurable in the superiorcolliculus are broadly tuned for sound location [4,10].

The topography of the auditory map and its alignmentwith the visual map only develops gradually over severalweeks. This process depends on the animal receiving theappropriate experience during this period. The mapping isseverely disrupted in guinea pigs that are deprived ofauditory spatial information during this critical time byrearing them in omnidirectional white noise [11]. Duringthis developmental period, the system can compensate fora manipulation of the normal auditory input, caused forexample by plugging one of the ears. After ferrets arereared in this way, not only do the superior colliculusneurons show normal spatial tuning, but the registration ofthe auditory and visual maps develops normally as well[4]. Adult animals subjected to the same procedure neverregain normal spatial responses.

It is the visual input that appears to dictate the alignmentof the auditory and visual maps. When barn owls arereared wearing prisms that displace the position of thevisual world on the retina, the responses of auditoryneurons in the owl’s optic tectum — the avian homologueof the superior colliculus — shift their spatial responses torealign with the visual map [12]. A similar effect is seen inthe ferret when animals are reared following a surgical dis-placement of one eye, a procedure that also displaces thevisual field [4].

NMDA receptors in map alignment during developmentTwo recent studies [13,14] have begun to explore thefascinating mechanisms underlying this alignment andrealignment of the auditory and visual maps in the superiorcolliculus. Both show that neurotransmission mediated byN-methyl-D-aspartate (NMDA) receptors is important inthese processes, albeit in different species and at differentlocations. NMDA-dependent mechanisms are thought tomediate Hebbian-type synaptic modification induced bysensory experience.

In their work, King and colleagues [3,14] implanted smallfragments of a polymer sheet (Elvax) impregnated with anNMDA-receptor antagonist (MK801 or AP-5) on thesurface of the superior colliculus in juvenile ferrets —before the onset of hearing but after the maturation ofvisual connections — and in adult animals. In the adultsand in juvenile controls (implanted with drug-free Elvax),the map of visual fields in the superficial layer and theauditory map in the deeper layers were both normal. Thevisual map was also normal in animals reared with theimplants containing the NMDA-receptor antagonist, butthe spatial tuning of the auditory neurons was significantlyimpaired. Furthermore, the topographical organization ofthe auditory neurons was poorly aligned with the visualmap in these animals [3,14].

From their estimates of how far the antagonists diffused,King and colleagues [3,14] suggested that their effects were

956 Current Biology 1996, Vol 6 No 8

limited to the upper (superficial and intermediate) layers ofthe superior colliculus. Anatomical studies have demon-strated that the dendrites of neurons in the deep layersproject into these upper layers. As King reported at therecent Brain Research Association meeting in Newcastle,UK (25–27 March, 1996), he and his colleagues have alsofound that the number and distribution of NMDA recep-tors within the colliculus varies during development. Thenumber increases up to the age at which the eyes open andthen decreases progressively until adulthood. Over thelatter period, the proportion of NMDA receptors increasesin the superficial layers relative to the deeper layers. Takentogether, these results are consistent with the hypothesisthat NMDA receptors activated by the more preciselyordered visual map strengthen synapses activated byauditory inputs for corresponding positions in space.

A different and novel NMDA-dependent mechanism hasbeen discovered by Feldman et al. [13], who found that, inthe barn owl, developmental plasticity of the auditory mapin the optic tectum is mediated by NMDA-receptor acti-vation in the external nucleus of the owl’s homologue ofthe inferior colliculus. As described above, when owls arereared with prisms over their eyes, the spatial sensitivitiesof auditory neurons recorded in the optic tectum changeto positions corresponding to the optical displacement(Fig. 2). This is a gradual process, taking several weeks,over the course of which the neurons pass through a transi-tional stage. During this transitional period, they respondto the interaural time differences representing both theinitial position and the new visual field position; Feldmanet al. [13] call the latter the learned response. Transitionalresponses are either bimodal or broadly tuned to interauraltime differences (Fig. 2).

In the experiment, neurons exhibiting such transitionalresponses were recorded from in the optic tectum, whilevarious blocking agents were released in the externalnucleus of the inferior colliculus through a fine glassmicropipette. Drugs that blocked the activity of NMDAreceptors caused an overall reduction in the response ofthe auditory neurons in the tectum, but the reduction inthe learned response was significantly greater than thereduction in the response to the initial sound position.This differential effect appears to be specific for NMDAreceptors, because other, non-NMDA antagonists (CNQXand kynurenic acid) reduced both normal and learnedresponses to about the same degree.

In normal birds, NMDA and non-NMDA receptors con-tribute equally to the responses at all stimulus positions.The predominance of NMDA receptors in the learnedresponses suggests that, during prism rearing, a changeoccurs in their expression. It is not known how suchchanges are brought about in a nucleus where visualresponses have not been recorded. The guiding input

does not, of course, need to be a direct visual input, itcould be a feedback from a site where the error betweenthe visual and auditory input is assessed. One intriguingpossibility is that the origin for such an input is the optictectum, where the visual and auditory maps are in closeproximity [12]. But the appropriate connections to theinferior colliculus have not yet been discovered in the owl.

It is necessary to be cautious when comparing data fromspecies as different as the ferret and the owl; the latterhas highly developed sound-localizing abilities, and thisis reflected in the specialized organization of its auditorypathway. Thus, mechanisms operating in one speciesmay not be present in the other. Nevertheless, theseexperiments offer fascinating insights into the Hebbianmechanisms through which the visual input can coercethe reorganization of the auditory representation.

Dispatch 957

Figure 2

When barn owls are reared wearing prisms (top right), the interauraltime difference (ITD, equivalent to sound azimuth) to which a neuron inthe optic tectum responds, shifts from its initial range (red curve) to anew value (blue curve), the learned response, which corresponds withthe spatial displacement of the visual field. In the course of thisprocess, the neuron passes through a transitional state in which itresponds to stimuli representing both positions with a bimodal or flatresponse (green curves).

Neuronal response

Prisms

Neuronal response

Neuronal response

© 1996 Current Biology

Normal LearnedITD

AcknowledgementI thank Alan Palmer for his helpful comments on this article.

References1. Stein BE, Meredith MA: The Merging of the Senses. Cambridge,

Massachusetts: MIT Press; 1993.2. Palmer AR, King AJ: The representation of auditory space in the

mammalian superior colliculus. Nature 1982, 299:248–249.3. King AJ, Schnupp JWH, Carlile S, Smith AL, Thompson ID: The

development of topographically-aligned maps of visual andauditory space in the superior colliculus. Prog Brain Res 1996, inpress

4. King AJ: A map of auditory space in the mammalian brain: neuralcomputation and development. Exp Physiol 1993, 78:559–590.

5. Sparks DL, Hartwich-Young R: The deep layers of the superiorcolliculus. In The Neurobiology of Saccadic Eye Movements.Reviews Of Oculomotor Research 3. Edited by Wurtz RH, GoldbergME. Amsterdam: Elsevier; 1989:213–255.

6. Middlebrooks JC, Knudsen EI: A neural code for auditory space inthe cat’s superior colliculus. J Neurosci 1984, 4:2621–2634.

7. Irvine DRF: The auditory brainstem: a review of the structure andfunction of auditory brainstem processing mechanisms. InProgress in Sensory Physiology 7. Edited by Ottoson D. Berlin:Springer; 1986.

8. Wise LZ, Irvine DRF: Topographic organization of interauralintensity difference sensitivity in deep layers of cat superiorcolliculus: Implications for auditory spatial representation. JNeurophysiol 1985, 54:185–211.

9. Jay M, Sparks DL: Auditory receptive fields in primate superiorcolliculus shift with changes in eye position. Nature 1984,309:345–347.

10. Withington-Wray DJ, Binns KE, Keating MJ: The developmentalemergence of a map of auditory space in the superior colliculusof the guinea pig. Dev Brain Res 1990, 51:225–236.

11. Withington-Wray DJ, Binns KE, Dhanjal Brickley SG, Keating MJ: Thematuration of the auditory space map in the guinea pig isdisrupted by developmental auditory deprivation. Eur J Neurosci1990, 2:693–703.

12. Brainard MS, Knudsen EI: Experience-dependent plasticity in theinferior colliculus: a site for visual calibration of the neuralrepresentation of auditory space in the barn owl. J Neurosci 1993,13:4589–4608

13. Feldman DE, Brainard MS, Knudsen EI: Newly learned auditoryresponses mediated by NMDA receptors in the owl inferiorcolliculus. Science 1996, 271:525–528.

14. Schnupp JWH, King AJ, Smith AL, Thompson ID: NMDA-receptorantagonists disrupt the formation of the auditory space map inthe mammalian inferior colliculus. J Neurosci 1995,15:1516–1531.

958 Current Biology 1996, Vol 6 No 8