One cause for caution is the fact that the Society has given two totally different figures for memorial attendance in 1935:
*** w88 12/15 p. 12 Angelic Messages for Our Day ***
In 1935 attendance worldwide at the Memorial observance of Jesus’ death was 32,795. Of these, 27,006 partook of the emblems as being the remaining ones on earth of the 144,000, whose hope is heavenly.
*** w96 8/15 p. 31 Questions From Readers ***
For example, at the Memorial celebration in 1935, attended by 63,146, those partaking of the emblems in evidence of their profession to be anointed numbered 52,465. Thirty years later, or in 1965, the attendance was 1,933,089, while the partakers decreased to 11,550.
The second figure also appears in the 1993 Proclaimers book (p. 717) and the 15 February 2003 Watchtower (p. 18). And yet, when you look at the original 1936 Yearbook, the actual published figures are the ones that appeared in the original 1988 Watchtower article:
So clearly there has been some revisionism....either the originally published figures were in error (and corrected in the research that went into the Proclaimers book) or they have been inflated to almost twice their original size. It is also worth nothing that the 1937 Yearbook gives the memorial figures for 1936 as 35,172 attending and 25,435 partaking (p. 69), which would seemingly support the older figures. However, it is quite striking that the grand totals for "Other Outlying Countries" for both years is so pitifully small. For instance, the 1937 Yearbook notes that there were 1,060 memorial partakers in Hungary (p. 153), a figure that is clearly not added to the overall total which has only 243 partakers outside the U.S. Similarly, the 1936 Yearbook noted that 10,537 attended the memorial in central Europe -- a number clearly exceeding the 300 for "Other Outlying Countries". This suggests to me that the newer Proclaimers numbers are indeed more accurate, since the older totals included mostly only the U.S. figures. But this raises the question of whether the 1928 figure is comparable, whether it might omit non-U.S. figures as were done in 1936 and 1937. However, I find it hard to believe that such a basic computational error was committed independently so early.