Left–right alternation is a defining feature of spinal locomotor circuits, yet the level of neuronal detail required to generate and maintain this pattern remains unclear. This thesis investigates how models spanning multiple levels of abstraction—from biophysically detailed Hodgkin–Huxley (HH) neurons to adaptive integrate–and–fire (I&F) formulations and synfire-chain modules—can account for the generation of fictive swimming in the spinal cord of the Xenopus laevis tadpole. The guiding hypothesis is that a small set of neuronal mechanisms is sufficient to reproduce the essential features of rhythmic alternation, and that moving between modeling scales helps distinguish core principles from biological detail. A minimal bilateral HH network comprising only four canonical neuron classes—excitatory descending interneurons (dINs), inhibitory commissural interneurons (cINs), ipsilateral inhibitory interneurons (aINs) and motoneurons—served as a biophysical proof of concept. Tuned to reproduce experimentally observed firing modes, the model demonstrated that rebound-prone dIN excitability, contralateral inhibition and modest electrical coupling are sufficient to generate stable alternating activity, even in very small networks. These results motivated the transition to simpler models capable of efficient analysis and scaling. Adaptive exponential I&F (AdEx) neurons were calibrated to physiological recordings using simulation-based inference, yielding tonic and phasic/rebound templates that preserved the key dynamical signatures of the HH model. Phase-plane analysis clarified the mechanisms underlying single-spike responses and rebound firing in dINs. At network level, the I&F models robustly reproduced left–right alternation, while highlighting constraints on synaptic kinetics and adaptation needed to avoid multi-spike responses. Finally, a synfire-chain framework provided a complementary, timing-centric perspective, demonstrating how precise spike synchrony, synaptic delays and minimal inhibitory coupling can generate alternating left–right sequences in a feedforward setting. Together, these approaches converge on a common conclusion: rebound-prone ipsilateral excitation combined with precisely timed contralateral inhibition constitutes a sufficient substrate for alternating spinal rhythms. By integrating bottom-up and top-down modeling strategies, this thesis provides a unified, extensible framework for studying spinal pattern generation. The results show that essential locomotor dynamics can be captured across multiple abstraction levels, offering both mechanistic insight and practical tools for future data-driven investigations of spinal circuit development, robustness and modulation.
Alexia C. Wilson (Wed,) studied this question.