The distribution of thalamocortical lags (broken down by ripple co-occurrence) is shown on Figure 7. In short, thalamocortical lags were generally either negative or not significantly different from zero, indicating that either sleep spindles preferentially occurred on the scalp first or occurred simultaneously on scalp and thalamic channels. Very little variance in spindle lag was accounted for by this model (R2 = 0.014), suggesting that lag mainly depends on extraneous factors. Still, fast spindles (β = 0.015), spindles in the MD nucleus (β = 0.017) and spindles occurring on F4 (β = 0.010) were associated with more positive lags, while spindles occurring on P3 (β = -0.007) and spindles with ripples (β = -0.017) were associated with more negative lags. That is, in case of fast spindles, in the MD and on F4, there was a diminished tendency for scalp spindles to occur first, but this tendency increased on P3 and especially when ripple activity was present in the thalamus. All effect sizes are unstandardized and given in seconds. All p<0.001, except for the effect associated with P3 (p = 0.010).