Stimulus 1 0.35 sec
Delay 1.0 sec
Stimulus 2 0.35 sec
Response 2.0 sec
Fig. 3. Timeline of a single delayed match-to-sample (DMS) trial for the auditory model (and the corresponding auditory fMRI experiment). Shown are frequency vs. time representations of the tonal contours (TCs)
The experimental evidence for the presence of neurons with these properties was reviewed in Husain et al. (2004). In brief, there were reports showing that there are neurons in auditory cortex that respond to the direction of frequency sweeps (e.g., Mendelson & Cynader 1985), and that there are neurons in prefrontal cortex that are active during the delay interval in a delayed response task for auditory stimuli (Kikuchi-Yorioka & Sawaguchi 2000). However, unlike the situation in visual cortex, there have been relatively few studies in awake monkeys and other mammalian preparations in which the response properties of neurons in various parts of the STG were evaluated. So, for example, there is only a small amount of evidence showing that neurons in the anterior part of the STG respond to complex auditory patterns (e.g., Kikuchi et al. 2004). Similarly, a number of other crucial assumptions that Husain et al. (2004) made in constructing the auditory model also rested on either weak experimental data (i.e., just a few studies), or else were made in analogy to the visual case. To illustrate, a key assumption we used was that in going from primary to secondary auditory cortex and thence to ST, the spectrotemporal window of integration increased (analogous to the increase in the size of the spatial receptive field in the visual model), so that, for example, neurons in the secondary auditory area respond best to longer frequency sweeps that do the neurons in A1/A2. Experimental support for this assumption could only be found in Harrison et al. (2000). Another key assumption was that in the secondary auditory area there was a neuronal population that responded best to a change in the direction of frequency sweeps. There was no published evidence for such 'second-derivative' neurons in auditory cortex.
In essence, each of these assumptions can now be considered predictions of the model. To focus on a case that we shall come back to, the assumption concerning the increase in the spectrotemporal window of integration is instantiated in the model by increasing both the divergence and sparseness of connections in the feedforward direction, with neighboring A1/A2 units sending connections to neighboring units in secondary auditory cortex. Likewise, these latter units project in a similar divergent fashion to ST. Consequently, the assumption of increasing spectrotemporal receptive field size ultimately rests on assumptions concerning the pattern of interregional neuroanatomical connectivity in STG.
Identical stimuli were used for both modeling and the corresponding fMRI experiment (Husain et al. 2004). The bar graph on the right of Fig. 2 shows the percent signal changes (comparing the fMRI activity for tonal contours to that for pure tones) in each brain region for both the simulated and the experimental data. Two important points arise from this part of Fig. 2. First, our simulated results in primary auditory cortex (A1/A2) did not match the experimental value (in the simulation, the percent change between tonal contours and tones was near zero). A likely reason for this is that we included in our model only one type of neuron found in primary auditory cortex (selective for frequency sweeps), but there are many neuronal types in the brain selective for other features in the auditory input (e.g., loudness, on and off properties of the input) that we were not modeling. Moreover, there was a large amount of scanner noise during the experiment that could have had some effect on the experimental data that was not taken into account in the simulation. The second important point is that we were able to get close quantitative agreement between simulated and experimental data in all the right hemisphere regions that corresponded to those in the model (except, as noted above, A1/A2). As far as we know, this was the first study in which a biologically realistic neural model generated stimulated fMRI data that generally agreed quantitatively with experimental fMRI values in which task design and stimuli were identical to those used in the modeling.
To test the robustness of the auditory model, we (Husain et al. 2005) used it to investigate the auditory continuity illusion, which is an example of one type of auditory perceptual grouping phenomenon. Perceptual grouping permits the auditory system to integrate brief, disparate sounds into cohesive perceptual units, which is important for perception because it enables, for example, one to separate attended sounds from environmental noise. The auditory continuity illusion emerges when a sound object (e.g., pure tone, frequency sweep, word) is perceived to continue through occluding noise even though no such signal need be physically present in the noise. Although it serves the important purpose of making communication sounds intelligible in a noisy environment and although it been extensively studied by means of psychophysical experiments, little is known concerning neural basis of this illusion.
In our simulations, intact stimuli (tonal contours) were matched with fragmented versions (i.e. with inserted silent gaps) of the stimuli (Fig. 4). The ability of the model to match fragmented stimuli declined as the duration of the gaps increased (Fig. 5, top). However, when simulated broadband noise was inserted into these gaps, the matching response was restored indicating that a continuous stimulus was perceived (Fig. 5, bottom). The electrical activities of the neuronal units of the model agreed with electrophysiological data obtained by Sugita (1997), and the behavioral activity of the model matched human behavioral data (Ciocca & Bregman 1987; Dannenbring 1976). The most important aspect of this study relevant to this chapter concerns how our model implements the illusion. The predominant mechanism is the divergence of the feedforward anatomical connections along the auditory processing pathway in the temporal cortex. Not only do our results attest to the robustness of the model, but further, they predict the primary role of the anatomical connectivity of the auditory processing areas in mediating the continuity illusion. Note that these results were obtained without changing any of the parameters of the auditory model.
In summary, our simulation results for the auditory model demonstrate that our assumptions concerning the neural mechanisms by which auditory objects are processed in the cerebral cortex are such that they enabled us (1) to simulate neuronal data from multiple brain regions that agreed with available experimental data, (2) to simulate fMRI data that quantitatively
C Fragmented with inserted noise
Was this article helpful?