Should Doe, John V report that the V might be a middle initial? #877

The shape

After a family comma, a bare I or V ending the given part reads as a generational suffix, and nothing reports that this was a guess. The same letter without a comma reads the same way, but there it reports suffix-or-name:

"Doe, John V"       →  given="John"  suffix="V"                 (no report)
"DOE, JOHN I"       →  given="JOHN"  suffix="I"                 (no report)
"Doe, John Q I"     →  given="John"  middle="Q"  suffix="I"     (no report)
"Doe, John V."      →  given="John"  middle="V."                (C9: a period marks an initial)
"Doe, John V, Jr"   →  given="John"  middle="V"  suffix="Jr"    (C9: the suffix slot is taken)
"John Doe V"        →  given="John"  family="Doe"  suffix="V"   reports suffix-or-name

This is rules/C9 (a one-character suffix word inside a comma part), and it's 1.4.0's reading: Doe, John V gave suffix V there too. It's been the reading since #144, whose reporter's data wrote I as a suffix (DOE, JANE MARIE I). decisions/C9 already records the missing report as "not done": the comma path "has never reported it", and adding an emitter was more than #430 needed.

Why it comes up now

#835 (PR #876) reads a bare salutation at the end of the given part as the title. That puts more names into this shape: DOE, JANE I MRS now reads title MRS, suffix I, where it read middle I MRS. In the FEC donor file that's 59 names. 24 of them used to report conjunction-or-initial on the I, so they lost the only signal the letter was in doubt.

How the data writes it

In the FEC's itemized contributions (fec-individuals, 1,890,655 names with exactly one comma, measured 2026-10-24):

• The letter matches ordinary middle-initial frequency. 214,856 names end the given part in a lone capital, a salutation set aside. V (1,956) ranks 19th of 26 letters and I (1,305) 21st, close to where ordinary initials like O (1,324) fall. Nearly all stand behind a full given name (V 1,922 of 1,956), not behind another initial.
• The period-marked spelling, which C9 reads as an initial, is just as common. I. and V. appear about as often as I and V, so it looks like one population written two ways.
• The real generational spellings are rarer. IV appears 765 times in that slot, so a bare V read as "the fifth" would have to outnumber "the fourth" by more than 2 to 1. III appears 6,254 times and II 2,159.
• Today's reading: of the 3,261 bare I/V names, 3,180 read the letter as a suffix.

So in this file the letter is mostly a middle initial. #144's file wrote the suffix. Both are real data, and nothing in the writing tells them apart.

Proposed rule (a C9 amendment)

When: a bare one-character suffix word ends the given part after a family comma, behind a name word, with no further comma, as in Doe, John V.

Then: it still reads as the generational suffix (C9, 1.4.0), and the parse reports suffix-or-name on it, naming the field it lands in.

Because: one letter is the shape a middle initial is written in, and at this slot the data goes both ways. One large file writes the letter as the initial about as often as letter frequency predicts, and another writes it as the generation. When a reader can't tell either, the fork should be reported, as it already is for the comma-less John Doe V.

Except:

• A period-marked letter stays an initial with no report (Doe, John V.). C9's period is a writing signal, and nothing is in doubt.
• A letter that follows another suffix gets no report (Doe, PSM I). The run settles it.
• A further comma (Doe, John V, Jr) keeps the middle initial, and gets no report.

Accepted: callers with FEC-style data will see a report on every such name. That's the point, but it's a new report on roughly 3,200 names in that file.

Open questions

• Report only, or also flip the default? The FEC counts argue the default should be the middle initial. Issues with names in the format Last, First Middle Suffix where Suffix is I #144 and 1.4.0 argue for the suffix. Reporting changes no field, so it can land first and the flip can be decided separately. A flip would need its own measurement of how often Issues with names in the format Last, First Middle Suffix where Suffix is I #144's shape occurs.
• Does the AGENTS.md example need correcting? The "a kind is worth adding only if a reader would hesitate" bullet uses Smith, John V as its example of a fork nobody hesitates over: "reads as a middle initial to anyone -- the comma settles it". But the parser reads it as the suffix, so the example contradicts both the code and this data. Whichever way this issue is decided, that bullet needs a different example.
• Is the strict switch involved? decisions/C9 notes the given-part walk is "strict-blind": Policy(lenient_comma_suffixes=False) still reads suffix V. Should the switch also turn this reading off?
• Comma shapes still to sweep: a nickname head, a particle surname (van Doe, John V), a trailing title (Doe, John V Prof.), and all three name_order values.

Where it was found

The read of fec-individuals for #835 (PR #876), recorded in decisions/H8 as "a consequence not decided here".