The majority of our understanding of the ability of the human nervous system to merge multiple sensory modalities originates from single-unit studies of non-human animals. These groundbreaking neurophysiological studies established many principles for understanding multisensory processing at the level of single neurons, including the criterion with which experimenters assess integration of multiple senses. It is tempting to consider that neuroimaging measurements, like blood oxygenation-level dependent (BOLD) activation measured with fMRI, are directly comparable with findings from single-unit recordings. Although several studies have established clear links between BOLD activation and neural activity there remains a fundamental difference between BOLD activation and single-unit activity: BOLD activation is measured from the vasculature supplying a heterogeneous population of neurons, whereas singe-unit measures are taken from individual neurons. The ramifications of this difference are not inconsequential, because the principles of multisensory phenomena established using single-unit recording may not apply to population-based neuroimaging data The established principles must be tested theoretically and empirically, and where they fail they must be replaced with new principles that are specific to the new technique.
Using a modeled BOLD response based upon known populations of neurons within the superior colliculus, including both unisensory and multisensory neurons, we have assessed three criteria commonly used in BOLD fMRI to identify purported multisensory brain regions: the maximum criterion (AV > Max(A,V)), the additive criterion (AV > Sum(A,V)), and the mean criterion (AV > Mean(A,V)). We show that the linear, time-variant properties of the BOLD response suggest that a single region consisting of only unisensory neurons would produce a BOLD response that exceeds both the maximum and mean criterion, and as such, the presence of unisensory neurons within a multisensory brain region invalidates the usage of both the criteria. Our models show the additive criterion to fall short of the ability to assess neuronal convergence on two accounts. Single-unit studies have shown that only a minority of multisensory-enhancing cells respond in a superadditive fashion, but to exceed the additive criterion the mean response of all multisensory neurons within a given region must be superadditive. Also, the additive criterion is influenced by the experimenter’s choice of baseline. The lower the unisensory response is relative to baseline, the more liberal the additive criterion becomes.
Here, we propose a criterion to assess neuronal integration of senses specifically designed to be both theoretically and empirically viable when used with the BOLD signal. By varying an added stimulus factor such as stimulus saliency, one can measure the change in BOLD activation in unisensory and multisensory conditions. Inequalities between the relative differences in the multisensory condition and the sum of the relative differences in the unisensory conditions (∆AV ≠ Sum (∆A,∆V)) indicates an interaction between the two sensory streams in a way that is not sensitive to experimenter-chosen baseline and accounts for the population-based aspect of the BOLD signal. Additionally, such a measure provides insights into what stimulus aspects are being integrated, as opposed to merely labeling a particular brain region as a site of integration.