Experimental-neuromodeling framework for understanding auditory object processing: Integrating data across multiple scales

In this article, we review a combined experimental-neuromodeling framework for understanding brain function with a specific application to auditory object processing. Within this framework, a model is constructed using the best available experimental data and is used to make predictions. The predictions are verified by conducting specific or directed experiments and the resulting data are matched with the simulated data. The model is refined or tested on new data and generates new predictions. The predictions in turn lead to better-focused experiments. The auditory object processing model was constructed using available neurophysiological and neuroanatomical data from mammalian studies of auditory object processing in the cortex. Auditory objects are brief sounds such as syllables, words, melodic fragments, etc. The model can simultaneously simulate neuronal activity at a columnar level and neuroimaging activity at a systems level while processing frequency-modulated tones in a delayed-match-to-sample task. The simulated neuroimaging activity was quantitatively matched with neuroimaging data obtained from experiments; both the simulations and the experiments used similar tasks, sounds, and other experimental parameters. We then used the model to investigate the neural bases of the auditory continuity illusion, a type of perceptual grouping phenomenon, without changing any of its parameters. Perceptual grouping enables the auditory system to integrate brief, disparate sounds into cohesive perceptual units. The neural mechanisms underlying auditory continuity illusion have not been studied extensively with conventional neuroimaging or electrophysiological techniques. Our modeling results agree with behavioral studies in humans and an electrophysiological study in cats. The results predict a particular set of bottom-up cortical processing mechanisms that implement perceptual grouping, and also attest to the robustness of our model.