Jason Kiss has posted some results from his tests done on various screen readers on pages with the so-called "responsive data tables" (I guess "responsive" is the new "also works on mobiles and stuff" hype term... bleh).
Anyway, this is pretty interesting!
Apparently, some browsers are telling some AT (accessibility technology) that a table isn't a table when your data table is set to display: block (or any of its parts), as seen in Chris Coyer's "Responsive Data Tables" experiment (link to that is on Jason's page). Seems to depend on the browser what the table is reported back as.
For Saffy-VO and NVDA-Firefox (Win), this means table navigation is gone, and the CSS-created content is read out. This would also mean then that the semantic relationship between your rows and columns are gone, but then while I've used this technique (and I'd only tested it in JAWS where I just got a regular table) on a long silly table, it's probably better to keep this change to simple, small tables. A complex table relying on stuff like axis or headers probably loses a lot more data than you want to bother faking back in with CSS.
JAWS is still treating them as tables, but since it's sticking to the old policy of not announcing CSS-created content, and giving users regular table feedback and navigation... really, no change for those users, except of course their browsers shouldn't be showing a "responsive" table unless they've shrunk their browser to the width of a mobile phone.
Assuming tables are set to display: block specifically for this "responsive" effect (making them fit on a screen the size of a mouse turd), could we ignore JAWS (and Window-Eyes) at this point anyway... until more Microsoft phones are out anyway? While I know VO has stepped over to iWhatevers pretty well, I dunno if there are any mobile versions of the Big Two for Windows.