1: \begin{abstract}

2: We propose to use multiple-imaged gravitational lenses to set limits on 

3: gravity theories without dark matter, specificly TeVeS (Bekenstein 2004),  

4: a theory which is consistent with fundamental relativistic principles 

5: and the phenomenology of MOdified Newtonian Dynamics (MOND) theory.

6: After setting the framework for lensing and cosmology, 

7: we derive analytically the deflection angle for the point lens and the Hernquist

8: galaxy profile, and study their patterns in convergence, shear and

9: amplification.  Applying our analytical lensing models we fit galaxy-quasar

10: lenses in the CASTLES sample.  We do this with three methods, fitting

11: the observed Einstein ring sizes, the image positions, or the flux

12: ratios.  In all cases we consistently find that stars in galaxies in  

13: MOND/TeVeS provide adequate lensing.  Bekenstein's toy

14: $\mu$ function provides more efficient lensing than 

15: the standard MOND $\mu$ function.

16: But for a handful of lenses a good fit would require

17: a lens mass orders of magnitude larger/smaller than the stellar mass

18: derived from luminosity unless the modification function $\mu$ and

19: modification scale $a_0$ for the universal gravity were allowed to be

20: very different from what spiral galaxy rotation curves normally imply.

21: We discuss the limitation of present data and summarize 

22: constraints on the MOND $\mu$ function.  

23: We also show that the simplest TeVeS "minimal-matter" cosmology,

24: a baryonic universe with a cosmological constant,  

25: can fit the distance-redshift relation from the supernova data,

26: but underpredicts the sound horizon size at the last scattering.  

27: We conclude that lensing is

28: a promising approach to differentiate laws of gravity 

29: (see also astro-ph/0512425).  

30: \end{abstract}

31: