1: \begin{abstract}

2: To better constrain the emission mechanism underlying the hard state

3: of galactic black hole candidates, we use high-time resolution RXTE

4: lightcurves for Cyg X-1 and GX 339-4 to compute two higher order

5: variability statistics for these objects, the skewness and the Fourier

6: bispectrum. Similar analyses, in particular using the skewness

7: measure, have been attempted before, but the photon collection area of

8: RXTE allows us to present results of much greater statistical

9: significance.  The results for the two objects are qualitatively

10: similar, reinforcing the idea that the same basic mechanisms are at

11: work in both. We find a significantly positive skewness for

12: variability timescales less than $\sim 1$ second, and a {\it negative}

13: skewness for timescales from $1- 5$ sec.  Such a skewness pattern

14: cannot be reproduced by the simplest shot variability models where

15: individual shots have a fixed profile and intensity and are

16: uncorrelated in time.  Further evidence against simple shot models

17: comes from the significant detection of a non-zero bicoherence for

18: Fourier periods $\sim 0.1-10$ sec, implying that significant coupling

19: does exist between variations on these timescales.  We discuss how

20: current popular models for variability in black hole systems can be

21: modified to match these observations.  Using simulated light curves,

22: we suggest that the most likely way to reproduce this observed

23: behavior is to have the variability come in groups of many shots, with

24: the number of shots per unit time fitting an envelope function which

25: has a rapid rise and slow decay, while the individual shots have a

26: slow rise and a rapid decay. Invoking a finite energy reservoir that

27: is depleted by each shot is a natural way of producing the required

28: shot correlations.

29: 

30: \end{abstract}

31: