deletions | additions       diff --git a/Deterministic phase portrait.tex b/Deterministic phase portrait.tex  index 25beb47..8096c67 100644  --- a/Deterministic phase portrait.tex  +++ b/Deterministic phase portrait.tex 
...
The second region, close to the origin of the plane, correspond to divisions occurring early after a circadian peak. As the deterministic attractor do not go through this region, traces rarely explore that interaction.  Two trajectories showing the effect of the upper deceleration region are shown in Figure \ref{fig:someTraces}. To verify that this slow down wasn't an artifact of the phase inference we selected all traces passing through the deceleration region and compared the length of the encompassing circadian intervals (peak to peak times) to the ones of traces dividing earlier in the circadian cycle. We defined the first group by taking the circadian phase at cell cycle phase equal to $0.75 \times 2\pi$ (that is, $\theta(t_d) : \phi(t_d) = 0.75 \times 2\pi$) included between $0.5$ and $0.75 \times 2\pi$ (gray rectangle in Figure \ref{fig:someTraces} A) while the second group was defined by a circadian phase included between $0.25$ and $0.5 \times 2\pi$ (gray rectangle in Figure \ref{fig:someTraces} C). The mean circadian interval was indeed longer ($24.7$h) in the first group than in the second one ($21.8$h, $p<10^{-16}$ t-test). This suggest that the circadian cycle is only slowed down by cell divisions in a specific window, and that our inference procedure is able to capture this effect, even though we measure only a proxy of one state variable of the circadian  oscillator.