The back-of-book index has long attracted a peculiar suspicion. Early critics complained that it allowed readers to raid a learned volume for isolated facts without submitting to the sequence of its argument. An index, on this view, converts disciplined reading into extraction: instead of following a demonstration, the impatient reader jumps to a name, harvests a page number, and departs. Modern search engines seem to have perfected the offense by locating any repeated string instantly.

This complaint mistakes an index for a merely mechanical doorway. A competent index does not list every word. It groups variant expressions under a chosen heading, separates different senses of the same term, and creates cross-references among concepts that may never appear together on a page. To place “rent” beneath “political economy” rather than merely under the letter R is to make a claim about the book's conceptual organization. The indexer reads the work twice: first as a sequence and then as a field of possible routes.

Those routes can expose features obscured by sequence. An argument presented gradually may rely on a concept whose scattered appearances become visible only when gathered under one entry. Cross-references can reveal that two vocabularies the text keeps apart perform similar work. The index may even register hesitation: repeated subentries and circular references can show that a book has no stable term for a problem it repeatedly encounters. Far from releasing readers from interpretation, a strong index offers an interpretation concise enough to be tested against the text.

The danger is not that indexes interpret, but that their interpretations masquerade as neutral access. A heading can make one category seem primary, bury a dissident term beneath “see also,” or omit an episode that complicates the indexer's scheme. Readers often treat absence from the index as absence from the book, although omission may reflect judgment, oversight, or limited space. The proper response is not to abandon indexes but to read them critically, as arguments about relevance.

Comparison among editions can make this argument visible. A later index may replace personal names with institutional headings, or split one broad entry into competing concepts. Such changes need not track revisions to the main text. They can record a new reader's theory of what the unchanged book is about. The index thus belongs to a history of reception as well as to the physical volume it serves.

Digital search does not eliminate this need. It excels at finding exact strings, including occurrences an indexer missed, but it cannot by itself decide that “wages,” “compensation,” and “the price of labor” belong to one conceptual family—or that in a particular book they emphatically do not. Search supplies locations; an index proposes relations. Used together, they can correct one another. The index remains valuable not as a shortcut around thought, but as a compact, contestable map of someone else's reading.