The main concern for production-oriented reticles has so far been the control of the critical dimension (CD), i.e. the CD uniformity and the CD mean-to-target. Feature linearity and overall feature size control is as much a strong requirement for a good reticle quality, in view of the Mask-Error-Enhancement-Factor (MEEF), which describes how the deviation of an actual feature from its nominal size is magnified onto the wafer. The present paper will discuss the results obtained from a mask linearity study, in which the simultaneous feature size control of different feature types and tone (lines, spaces, dots, contacts) is assessed for varying nominal feature size and proximity (isolated vs. equal lines and spaces), using a reticle CD-SEM. Today mask shops have a range of mask writers available, i.e., 10kV raster scan gaussian beam e-beam tools, laser based optical tools and 50kV vector scan variable shaped e-beam systems. The tool used leaves a so-called "finger print" on the reticle, as a varying linearity behavior. It will be shown that the 50kV writers are superior in terms of mask linearity, as can be expected. Yet, the patteming process used (especially the resist and etching of the patteming layer) is as important. In addition, the mask shop may have compensated for etch bias and for known proximity effects, by tweaking the mask data. The paper will show and discuss the results of masks obtained from over 15 different sources (varying mask shop, tool used and process used), in an anonymous way, as an impression of what is presently available on the market. The results are discussed against the mask requirements table in the ITRS2001.