How the nematode C. elegans moves in a serpentine fashion is still unknown despite a detailed anatomical knowledge, connectome and genetic access to each of its 302 neurons. Two main models exist for C. elegans locomotion: a body cascade model where the back and forth undulations of the head set up an oscillatory pattern that propagates down the body via lateral connections between neuromuscular units and biomechanical linkage, and an alternative active posture model where the sinusoidal body posture along the entire body is effected by active neuromuscular control not solely deriving from lateral neuromuscular signaling from the head to tail. We recorded high resolution videos of worms crawling on an agar surface and extracted time series of the centerline postural angle for each of 13 body segments. As expected, the time series appeared to be a series of phase-lagged noisy sinusoids. We computed the peak cross-correlation between each body segment angle to the anterior-most segment angle-for a simulated worm under the body cascade model, this function monotonically decreases as function of segment number. However, we found a strong breaking of monotonicity in our experimental data, arguing against a strict body cascade model. We then performed mutual information analysis between segment pairs, and found a strong deviation from monotonic information loss. This deviation from monotonicity was further evidence against the strict body cascade model, as a consequence of the data processing inequality theorem. We then performed the same analysis for backwards locomotion and swimming versus crawling behaviors and again found a breaking of monotonicity, although at a different body segments, suggesting a different pattern of neural control. Finally, we performed our analysis on various mutant worms tracked through the Open Worm Movement Database. Overall, our analysis results resembled N2 worms for all analyzed genotypes, though
unc-37 and
egl-8 displayed reduced strength of the mid-body increases in peak cross-correlation and mutual information, suggesting a loss of central coordination. We conclude that worms employ centrally driven active postural control to locomote in addition to lateral neuromuscular signaling. Thus, basic information theory applied to animal behavior can yield insights into models of neurobehavioral control.