Despite a long history of research in motor control, the exact mechanism of how the brain communicates with the the invertebrate ventral nerve cord (VNC) and the vertebrate spinal cord on a single neuron basis remains largely elusive. Drosophila melanogaster is an ideal model organism to address the role of individual neurons, thanks to its stereotyped nervous system. Descending neurons (DNs), interneurons in the brain that project to the invertebrate ventral nerve cord or vertebrate spinal cord, have been shown to elicit specific behaviors upon optogenetic activation in Drosophila. These results suggest that behavior is controlled by sparse sets of command neurons, each eliciting a particular behavior. However, it remains to be seen how many DNs are involved in the control of naturalistic behaviors and what other information they might encode. In order to avoid commands leading to physically impossible or destabilizing actions, the brain has to be aware of the current behavior state. Ascending neurons (ANs), interneurons in the invertebrate ventral nerve cord or vertebrate spinal cord projecting to the brain, are likely conveying this information to the brain. To which degree of fidelity and in what way ANs encode behavior state remains unclear. To address these questions, we developed a dissection approach that allows functional imaging of ANs and DNs in the cervical connective. This dissection gives us optical access to the cervical connective and parts of the VNC in behaving adult Drosophila such that a two-photon fluorescence microscope can be used to image neural activity via calcium indicators. We began by imaging a new library of sparse split GAL4 driver lines targeting ANs. Our recordings from 247 genetically identifiable ANs revealed neurons encoding walking, resting, turning, eye grooming, and foreleg movements. Anatomical characterization of these ANs showed that they predominantly project to two brain regions: the anterior ventrolateral protocerebrum (AVLP) and the gnathal ganglion (GNG). Our results suggest that ANs encoding the motion of the animal with respect to the surrounding environment (self-motion) predominantly project to the AVLP, an integrative sensory hub. ANs projecting to the GNG mostly encode discrete actions. We then proceeded to image populations of DNs. The majority of DNs we recorded encoded walking behavior, with smaller fractions encoding resting and head grooming. Some of these neurons are likely to drive walking, but we also identified neurons encoding walking speed and turning. This suggests that a core set of DNs drives a behavior whose features can be modulated by additional DNs. Besides encoding behaviors, some DNs encode sensory information, including the presence of odors and deflection of the antennae.