1947 yearbook of Kansas State College of Agriculture and Applied Science, Manhattan, Kansas — a group photograph whose printed caption names Joe I. Cooper ninth of ten in the third row, together with four face crops that Ancestry's index attaches to the same name on other pages of the same volume.

What this document records

What the caption actually says

A group photograph of a uniformed formation, captioned by rows. The portion held reads: "… W. Walters. Third row: Kenneth L. Myer, Donald D. Wilson, Herbert A. Traulsen, George D. Allan, Arthur L. Miller, Thomas K. Willard, Lewis D. Stratton, Robert A. Fulmer, Joe I. Cooper, Willis C. Patten. Second row: Thomas E. Levick, John O. Fuller, James M. Brown, Phillip J. Hampton, Jessie C. Mayfield, Leland W. Leonard, Robert L. Pettit, Robert J. Lehnen, Robert L. Daly, Carmen K. Wilcox, Forest D. Campbell. Front row: Walter J. Throckmorton, Robert L. Yates, John F. Meisner, Lee M. Stratton, Fred C. Kramer, Don R. Borthwick, William H. Houghton, Robert L. Mingle, Robert Kern, Kenneth Hunt, Robert E. Crackel." This is the yearbook's own text naming him — not an index, not a crop. He is ninth of ten in the third row, between Robert A. Fulmer and Willis C. Patten.

What the caption does not say here

The crop begins mid-caption. It opens on "… W. Walters." — the tail of an earlier row, with the lines above it (which would carry the fourth row and the unit's name) outside the member's crop. So the unit is not named in anything the catalog holds. Ancestry's index gives a PHOTO DESCRIPTION of "R O T G COMPANY G" — evidently R.O.T.C. Company G — for one of the seven records, but that is Ancestry's reading of some page of this volume and not necessarily this page. Recorded as the index's word, not the document's.

The face crops and why none of them can be asserted

Ancestry's face-crop-to-name mapping in this collection is demonstrably wrong. The seven "Joe I Cooper" records for this school and year carry face thumbnails in the results table, and: - three of them are the identical crop of one portrait (same neighbours visible in the grid); - one is a visibly different young man, cropped at the edge of a different page; - one is a tight enlargement of a third face; - and one is a woman. ⇒ The index attaches at least three different male faces and one female face to a single name. These records index a page; the face crop is a guess at which portrait in the grid the OCR'd name belongs to, and on a dense portrait grid an off-by-one is routine. Therefore no crop here is recorded as a likeness of Joe Irvin Cooper. All four are filed as candidates, and the woman is filed with them deliberately, as the evidence that the mapping fails.

What would settle it

The yearbook page itself. A college portrait grid prints the names in a list keyed to the grid; reading that list against the faces settles the mapping in one look. Ancestry's image viewer would not render the page in this session — the FamilySearch or a library copy of the 1947 volume, or an Internet Archive scan, would.