Zatorre, R.J., Bouffard, M., Ahad, P. and Belin, P. (2002) Where is 'where' in the human auditory cortex?. Nature Neuroscience, 5, 905-909.
We examine the functional characteristics of auditory cortical areas sensitive to spatial cues in the human brain, and whether they can be dissociated from parietal-lobe mechanisms. Three PET experiments were conducted using a speaker array permitting quasi free-field sound presentation within the scanner. In the first two experiments posterior auditory cortex was found to respond to sounds that varied in their spatial distribution, but only when multiple complex stimuli were presented simultaneously, implicating this cortical system in disambiguating overlapping auditory sources. The third experiment demonstrated that the right inferior parietal cortex is specifically recruited in localization tasks, and that its activity predicts behavioral performance, consistent with its involvement in sensory-motor integration and spatial transformation. The findings clarify the functional roles of posterior auditory and parietal cortices, and help to reconcile competing models of auditory cortical organization.
Performance on 5-key (L5) and joystick (LJ) localization tasks. Red bars indicate mean absolute error as a function of azimuth (with standard error bars indicated). Filled points indicate mean angle at which the joystick was placed for each of the five positions sampled (standard error bars). Solid line represents veridical localization performance.
Performance on right-left (RL) and same-different (SD) discrimination tasks. Blue bars (referenced to left ordinate) indicate mean percent error as a function of position of the stimuli. Solid line (referenced to right ordinate) indicates mean latency to respond.
A series of measures was taken to validate the small (radius 24 cm) stimulus array used inside the PET scanner. Three considerations make the behavioral testing potentially different within the scanning environment than under more conventional situations: first, the proximal location of the sound sources may introduce different acoustical cues than are found with more distal sound sources; second, the subject is lying supine on the scanner bed, whereas sound localization tests are typically conducted with the subject in an upright sitting position; third, the PET scanner gantry contains fans which produce continuous background noise (approx. 56 dBA SPL) which might affect performance. To address these points, we carried out acoustical measures of the physical sound cues available to the listener, and conducted a validation study to compare performance within the scanner and in a more conventional auditory localization setup.
Previous work (Brungart et al., J. Acoust. Soc. Am., 1999, 106, 1465-1479; J. Acoust. Soc. Am., 1999, 106, 1956-1968) indicates that for sound sources fairly close to the head, the binaural cues that are important are similar to those for sources further away, except that interaural level differences (ILDs) increase substantially for close sources, whereas interaural time delays (ITDs) are independent of distance.To document interaural ILDs we measured the dB values (A scale, slow speed) with a sound pressure meter (Larson-Davis DSP83) positioned next to the ear for the noise stimulus used in Exp 1. To measure ITDs, we used a square-wave click (the onset of the noise stimulus was not sufficiently crisp against the background noise to be able to discern the small time differences), and recorded the onsets in the two ears using insert earphones placed within the ear canals; we then analyzed the time differences between the two channels with sound-editing software. The mean maximum ITD for peripheral positions in the scanner device was 0.78 ms (range .75-.81 across different listerners), and was zero at the center; values measured with the larger array were essentially identical. The mean maximum ILD in the scanner was 9.5 dB (range 8-11) for the most peripheral locations, in contrast to those measured in the lab with more distal sources which were about 4.5dB. These measures are in keeping with the data reported by Brungart et al., and confirm that ILDs are more prominent for proximal sources, while ITDs are constant.
To ascertain empirically whether behavioral localization performance was any different using the scanner array, as compared to a more standard setup, we carried out a series of behavioral measures of localization using the joystick in our PET scanner apparatus, and in a more conventional stimulus presentation array of 1 m radius used in our prior studies (Zatorre & Penhune, J Neurosci, 2001, 21, 6321-6328; q.v. for additional details). We tested 6 subjects (3 male, 3 female, age range 22-50) using identical stimuli (broad-band noise) and task (joystick as used in our Exp 3) in the two situations: lying within the scanner, and seated in a sound-treated lab with the larger array. Twenty trials per position (range 90º to +90º) were administered in a randomized order, and subjects were simply instructed to place the joystick in the position corresponding to the perceived angular position.
The figure to the right illustrates the results obtained in the two test situations. The dependent variable is the mean joystick position. Subjects in both the scanner and standard array had very similar functions; the correlation between positions perceived across the two arrays was 0.995. Inaccuracy at the peripheral positions was somewhat more exaggerated in the scanner device than in the larger array at the far peripheral locations. However, signed error scores tended to be smaller (i.e. more accurate) in the scanner array than in the 1 m array over the mid range.
An additional way to evaluate performance in the two devices is to compute the absolute error score (absolute value of difference between obtained response and veridical stimulus position). This index has the advantage that it is independent of directional bias, and reflects overall accuracy of localization. Mean absolute error as a function of position is shown in the figure to the right. There was no significant difference in this variable across the two arrays [F(1,5)=2.19, p=.19]; measures in both situations showed the characteristic better performance near the center and greater error towards the periphery, as expected (e.g. Perrott et al, J. Acoust. Soc. Am., 1993, 93, 2134-2138).
These findings indicate that, generally speaking, behavioral localization performance is not substantially different when tested in the scanner, as compared to what might be expected under conventional testing conditions.
One additional source of noise in the behavioral data needs to be considered: the motor control of the joystick. It is possible that listeners might give responses which are somehow distorted because of the somewhat unusual position of the joystick (by the persons side as they are lying down), and given that it was out of view. To verify that motor aspects of manipulating the joystick were not an issue, we had asked each of our original 12 PET volunteers to position the joystick at each of 5 positions (-90, -45, 0, 45, and 90) via verbal instructions prior to beginning the tasks. The data from this calibration series are shown in the figure. As can be seen, the responses were very close to the values expected assuming perfect accuracy in joystick orientation. Thus, it appears that controlling the joystick does not introduce any confounds.