1: \begin{abstract}

2: 

3: We determine how MRI-turbulent stresses depend on gas pressure

4: via a suite of

5: unstratified shearing box simulations. Earlier numerical work reported

6: only a very weak dependence at best, results

7: that call into question the canonical $\alpha$-disk model and the

8: thermal stability results that follow from it. 

9: Our simulations, in contrast, exhibit

10: a stronger relationship, and show that

11: previous work was box-size limited:

12: turbulent `eddies' were artificially restricted by the numerical

13: domain

14: rather than

15: by the scale height.

16: Zero-net-flux runs without physical diffusion coefficients yield a

17: stress proportional to $P^{0.5}$, where $P$

18: is pressure. The stresses are also proportional to the grid length and

19: hence remain numerically unconverged.

20: The same runs with physical

21: diffusivities, however, give a result closer to an

22: $\alpha$-disk: the stress is $\propto P^{0.9}$. 

23: Net-flux simulations without explicit diffusion

24: exhibit stresses $\propto P^{0.5}$, but 

25: stronger imposed fields weaken this correlation.

26: In summary, compressibility is important for the saturation of the

27: MRI, but the exact stress-pressure relationship is difficult to

28: ascertain in local simulations 

29: because of numerical convergence issues and the influence of any imposed flux. 

30: As a consequence, the interpretation of thermal stability behaviour in

31: local simulations is a problematic enterprise.

32: \end{abstract}

33: