Reading it the obvious way I'd have done the job at the time would be to do two or even four prints on each page selecting which chracers fell on that gride each time. The thesis would be written using LISP to do the formatting. This would also only print those characters that fell on the current alignment. It would be a bit fiddly making the 1/24 inch alignments of the pages and I'm sure there would have been plenty of mistakes! But it is straightforwardly doable with the tools he had. Very impressive though. If he did have a LISP program like that I'd have expect him to have developed it further or make some mention of it.