The NUMRM structure is used to track revision marking data for paragraph numbers, and is stored in the PAP for each numbered paragraph. When revision marking tracking is turned on, we fill out the NUMRM for each number with the data required to recreate the number's text. Then at display time, that string is compared with the current paragraph number string, and displayed as changed (old deleted, current inserted) if the strings differ. The string construction algorithm is the same as for an LVL structure.
| b10 | b16 | field | type | size | bitfield | comment |
| 0 | 0 | fNumRM | uns char | 1 | True if this paragraph was numbered when revision mark tracking was turned on | |
| 1 | 1 | Spare | uns char | 1 | ||
| 2 | 2 | ibstNumRM | short | 2 | index to author IDs stored in hsttbfRMark for the paragraph number change | |
| 4 | 4 | dttmNumRM | DTTM | 4 | Date of the paragraph number change | |
| 8 | 8 | rgbxchNums[9] | uns char[9] | 9 | Index into NUMRM.xst of the locations of paragraph number place holders for each level (see LVL.rgxchNums) | |
| 17 | 11 | rgnfc[9] | uns char[9] | 9 | Number Format Code for the paragraph number place holders for each level (see LVL.nfc) | |
| 26 | 1A | Spare | short | 2 | ||
| 28 | 1C | PNBR | int [9] | 36 | Numerical value for each level place holder in NUMRM.xst. | |
| 64 | 40 | xst | XCHAR[32] | 64 | The text string for the paragraph number, containing level place holders |
cbNUMRM (count of bytes of NUMRM) is 128 (decimal), 80 (hex).