\documentclass{article} \input mtex \input diagxy \input xymatrix \begin{document} When I created diagxy I did not set out to make diagrams that were better than those of xy-pic, but only to make a simpler and more familiar (to users of my diagram macros) interface. Since it is based entirely on xy-pic, it is not possible in any case. But then there are defaults\ldots. A colleague asked me to look at a paper to help him choose a referee and I came on the following display: $$\xymatrix { X_{rs}\ar[rr]^{\Upsilon_{(rs,\tau)}}&& X_{^\tau(rs)}\ar[rr]^{\Upsilon_{(^\tau(rs), \sigma)}}&&X_{^{\sigma\tau}(rs)} } $$ I don't know that the above code is the exact code the authors used, but the results are identical. I call your attention to the varying heights of the three $X$'s. For an extreme example of this, consider $$\bfig \xymatrix{X^\Gamma\ar[r]&x_\gamma} \efig$$ This happens because xymatrix centers the arrow on the middle of the node. I do not know if xymatrix can be told to do otherwise, but the underlying xy-pic code certainly can be and that was the default I chose in designing my morphism macro. Look at $$\bfig \morphism<900,0>[X_{rs}`X_{^\tau(rs)};{\Upsilon_{(rs,\tau)}}] \morphism(900,0)<900,0>[X_{^\tau(rs)}`X_{^{\sigma\tau}(rs)}; {\Upsilon_{(^\tau(rs),\sigma)}}] \efig$$ by contrast. The morphism code actually raises the arrow by a fixed amount (some fraction of the xheight, a font-dependent quantity) above the baseline, which is independent of descenders, ascenders, and sub- and superscripts. Actually, I would have done this with online to, which uses xy-pic only for the arrows, not the nodes and adjusts the lengths of the arrows to fit the labels: $$X_{rs}\to^{\Upsilon_{(rs,\tau)}}X_{^\tau(rs)}\to^{\Upsilon_{(^\tau(rs), \sigma)}}X_{^{\sigma\tau}(rs)}$$ This is not intended as a criticism of xy-pic. It is very general and it would be easy to find examples in which their default was what is wanted. It is just not likely to be in commutative diagrams. They do give the option of placing the arrow with respect to the baseline, which is more than Latex' picture mode does. My old macros were based on that. I now think it may be possible to coerce the picture mode into using the baseline (by putting things into boxes and defining them to have depth 0, if you want the details), but I knew much less 15 years ago when I started using Latex. So I used a different, much less satisfactory, approach that sort of worked but caused other problems and required extensive use of phantom arguments. \kern5pt\noindent Michael \end{document}