To simplify testing of copy/paste and screen readers, the text in the
markup of each heading is a brief descriptive of the method used. In
normal use of IR methods, just as with alt text, the replaced text
should be the same as provided in the image. I also created linked
versions of each of the methods..this was a PITA, but most of the methods
work back to IE5.01/win (validity note: cursor:hand was needed. If you
don't need it, drop it). I included the methods of Mike
Rundle and Seamus
Leahy as they require no SPAN elements, but, just like
the original Fahrner IR, these also hide text in different ways.
My initial testing  seems to conclude that Levin's
IR method is still the best choice for three reasons: 1) it
allows sighted users to browse with images off. 2) it works in most modern
browsers. 3) it's fairly straightforward to make the heading also a
CSS 2.1 IR technique essentially creates the SPAN in Levin's
method using generated content, but this is only currently reliable in
Opera7 and Safari 1.2.
Although the text in Levin's simple test
case of Andrew's IR method appears to be visible with images off,
in a more complex page with nested elements, the
use of z-index either places
the element too deep to be visible, or, upon adding position:relative;
z-index:2; to the heading as Levin suggested, puts the text on
all browsers I tested besides Gecko.