As was the case with simple spike suppression, neither average CS-complex spike occurrence nor its average latency showed between-cell correlations to average CR properties (each p > 0.25), except for a small correlation between mean CS-complex spike latency and mean CR onset (n = 14, r = 0.567, p = 0.0346). However, on a trial-by-trial basis, mixed-effects regression on the 14 cells with a clear CS-complex spike did show that CRs in trials with a CS-complex spike were on average 13.7% higher in amplitude than those without (p = 0.0194). Also, in a similar mixed model, CR amplitude related inversely to CS-complex spike latency, with an average decrease in percentage eyelid closure of 0.26 percentage points per millisecond increase of CS-complex spike latency (p = 0.0034). Conversely, CS-complex spike latency was 4.9 ms earlier in trials with a CR compared to those without (p < 0.0001). Together, these mild effects imply a modest contribution of CS-complex spikes to conditioned behavior on a trial-bytrial basis. The significance of the link between CS-complex spike occurrence and CR amplitude disappears when including simple spike suppression in the regression model (p = 0.16). This raises the possibility that CS-complex spikes affect conditioned behavior in part through their effect on simple spike suppression. Still, CS-complex spike latency retains some significance as a predictor of CR amplitude, even when including simple spikes in the model (p = 0.0083). This raises the possibility that climbing fiber signals and/or that of their collaterals in the nuclei directly contribute to the conditioned motor response.