It is still a popular view that primary sensory cortices are unimodal, but recent physiological studies have shown that under certain behavioral conditions primary sensory cortices can also be activated by multiple other modalities. Here, we investigate the anatomical substrate, which may underlie multisensory processes at the level of the primary auditory cortex (field AI), and which may, in turn, enable AI to influence other sensory systems. We approached this issue by means of the axonal transport of the sensitive bidirectional neuronal tracer fluorescein-labeled dextran which was injected into AI of Mongolian gerbils (Meriones unguiculatus). Of the total number of retrogradely labeled cell bodies (i.e. cells of origin of direct projections to AI) found in non-auditory sensory and multisensory brain areas, approximately 40% were in cortical areas and 60% in subcortical structures. Of the cell bodies in the cortical areas about 82% were located in multisensory cortex, viz., the dorsoposterior and ventroposterior, posterior parietal cortex, the claustrum, and the endopiriform nucleus, 10% were located in the primary somatosensory cortex (hindlimb and trunk region), and 8% in secondary visual cortex. The cortical regions with retrogradely labeled cells also contained anterogradely labeled axons and their terminations, i.e. they are also target areas of direct projections from AI. In addition, the primary olfactory cortex was identified as a target area of projections from AI. The laminar pattern of corticocortical connections suggests that AI receives primarily cortical feedback-type inputs and projects in a feedforward manner to its target areas. Of the labeled cell bodies in the subcortical structures, approximately 90% were located in multisensory thalamic, 4% in visual thalamic, and 6% in multisensory lower brainstem structures. At subcortical levels, we observed a similar correspondence of retrogradely labeled cells and anterogradely labeled axons and terminals in visual (posterior limitans thalamic nucleus) and multisensory thalamic nuclei (dorsal and medial division of the medial geniculate body, suprageniculate nucleus, posterior thalamic cell group, zona incerta), and in the multisensory nucleus of the brachium of the inferior colliculus. Retrograde, but not anterograde, labeling was found in the multisensory pontine reticular formation, particularly in the reticulotegmental nucleus of the pons. Conversely, anterograde, but no retrograde, labeling was found in the visual laterodorsal and lateroposterior thalamic nuclei, in the multisensory peripeduncular, posterior intralaminar, and reticular thalamic nuclei, as well as in the multisensory superior and pericentral inferior colliculi (including cuneiform and sagulum nucleus), pontine nuclei, and periaqueductal gray. Our study supports the notion that AI is not merely involved in the analysis of auditory stimulus properties but also in processing of other sensory and multisensory information. Since AI is directly connected to other primary sensory cortices (viz. the somatosensory and olfactory ones) multisensory information is probably also processed in these cortices. This suggests more generally, that primary sensory cortices may not be unimodal.