The issue of whether pain is represented by specific neural elements or by patterned activity within a convergent somatosensory subsystem has been debated for over a century. The gate control theory introduced in 1965 denied central specificity, and since then most authors have endorsed convergent wide-dynamic-range neurons. Recent functional and anatomical findings provide compelling support for a new perspective that views pain in humans as a homeostatic emotion that integrates both specific labeled lines and convergent somatic activity.

This article addresses the neuroanatomical evidence for a progression of integrative representations of affective feelings from the body that lead to an ultimate representation of all feelings in the bilateral anterior insulae, or "the sentient self." Evidence for somatotopy in the primary interoceptive sensory cortex is presented, and the organization of the mid-insula and the anterior insula is discussed. Issues that need to be addressed are highlighted. A possible basis for subjectivity in a cinemascopic model of awareness is presented.

The distribution of terminal projections in the brainstem from lamina I neurons in the spinal dorsal horn was investigated with the anterograde tracer Phaseolus vulgaris-leucoagglutinin in the cat and the cynomolgus monkey. Iontophoretic injections made with physiological guidance were restricted to lamina I or to laminae I-III in the cervical (C6-8) or lumbar (L6-7) enlargement. The distribution of terminal labeling was essentially identical in the cat and the monkey, although consistently of greater intensity in the monkey. Terminations were observed in the solitary nucleus, the dorsomedial medullary reticular formation, the entire rostrocaudal extent of the ventrolateral medulla, the locus coeruleus, the subcoerulear region and the Kölliker-Fuse nucleus, the lateral and medial portions of the parabrachial nucleus, the cuneiform nucleus, the ventrolateral and lateral portions of the periaqueductal gray, and the intercollicular nucleus. Lamina I terminations were generally bilateral in the medulla but more dense contralaterally in the pons and mesencephalon. The density and laterality of labeling in the medulla varied between cases independently from that in the pons and mesencephalon, suggesting that the lamina I projections to these regions may originate from different subsets of neurons. A clear topographic organization was observed only in the lateral column of the periaqueductal gray, where lumbar lamina I terminations were found caudal to cervical terminations. These observations indicate that spinal lamina I neurons project to a variety of brainstem sites involved in autonomic (cardiovascular, respiratory) and homeostatic processing and the control of behavioral state. These projections provide an afferent substrate for spino-bulbo-spinal somatoautonomic reflex arcs activated by nociceptive, thermoreceptive activity and for a spino-bulbo-hypothalamic relay of such activity by cells in the caudal ventrolateral medulla. These observations support the general concept that lamina I projections distribute modality-selective sensory information relevant to the physiological status and maintenance of the tissues and organs of the entire organism.