Definition
Deduction is a mechanic in which the game hands over a body of material and then only judges whether the conclusion the player assembled is right. The player's verbs are look, read and cross-reference; there is almost nothing on a board to move. What changes each turn is not the state of the world but the player's own understanding. The puzzle is solved when that understanding is written into the required fields and matches the answer.
The core of the mechanic is that the reasoning happens outside the game. Forming a hypothesis, testing it against evidence and discarding contradictions all take place in the player's head. The game's job is to lay out the material and, at the end, to mark it. Two players given the same material therefore never take the same route.
Origins and lineage
The best known ancestor that put reasoning at the centre of a game is the board game Cluedo, designed by Anthony E. Pratt. He devised it in 1943, applied for a patent in 1944 and was granted it in 1947; postwar shortages held the release back to 1949 (Waddingtons in the UK, Clue in the United States). Cluedo's reasoning, however, is elimination from a hand of cards, not the assembly of meaning out of documents.
A skeleton much closer to the modern mechanic appears in the early 1980s with Sherlock Holmes: Consulting Detective, a casebook-format game from Sleuth Publications designed by Raymond Edwards, Suzanne Goldberg and Gary Grady. Players read the case in any order, lose points for every location visited, then answer a quiz and compare their score with Holmes's. Sources disagree on whether the first edition is 1981 or 1982, so the year is not settled. Three traits are already present: free browsing, stopping when you judge you are ready, and a single verdict at the end.
The digital form settled in the 2010s. Her Story (Sam Barlow, 2015) shrank access to the material down to a single search box. In his GDC 2016 talk Barlow took the position that the player's brain is the most powerful game engine there is (the philosophy of Sam Barlow). Return of the Obra Dinn (Lucas Pope, 2018) worked out the answer fields and the rule for checking them, and became the reference point for what followed (the philosophy of Lucas Pope). The Case of the Golden Idol (2022) and The Roottrees are Dead (2025) generalised that shape into fill-in-the-blank word banks.
Variations
How the material is handed over changes the feel completely. In the search form, everything is present from the start and only what the player types is retrieved. Her Story and Telling Lies are the models here, and the vocabulary you can think of to search becomes the engine of progress. The unease of never knowing what you have not seen is the character of this form (searching the archive).
In the fill-in form, the material arrives as scenes and the answer goes into fixed fields. Return of the Obra Dinn assigns a name and a cause of death to each crew member. The Case of the Golden Idol and The Rise of the Golden Idol drop collected words into gaps in sentences. The Roottrees are Dead offers a family tree as one very large answer sheet. In each case the shape of the fields is itself the statement of what must be pinned down.
In the exploratory form, reasoning and movement dissolve into the same act. Strange Horticulture turns matching a specimen against a field guide into the daily work of minding a shop, and Botany Manor lays the inference of flowering conditions over a walk through a house. Paradise Killer and Shadows of Doubt make gathering evidence identical with surveying a large world. Blue Prince shows another combination, where generating the floor plan and reasoning about it sit together.
Design pitfalls
The most common failure is the granularity of checking. Return a verdict on each item the moment it is entered and brute force becomes the optimal tactic, which removes the need to reason at all. Withhold every verdict until the end and the player never gets confirmation, so the satisfaction disappears. Locking in only when three fates are correct together, as Return of the Obra Dinn does, is a settlement between those two extremes (designing the in-game notebook, designing hint systems).
The second is assistance that goes too far. Linking clues automatically, flagging contradictions, highlighting empty fields — each is helpful on its own, but together the notebook starts doing the reasoning. Lorelei and the Laser Eyes takes on the storage of documents and draws not one line between them. Storage to the machine, connection of meaning to the human: that division of labour is the lifeline of this mechanic.
The third is a shortage of information channels, which surfaces in two ways. Too few clues and the player is forced into guessing, unable to attribute a failure to their own reasoning (designing luck out of the puzzle, legible failure). Show the word bank too early and the fields can be filled with plausible-sounding words before the evidence is read at all (filling in the Golden Idol by vibes). The harder the work, the more readily these two complaints show up in its reception (the case against Obra Dinn, the case against Lorelei).
How to tell it apart
What separates this from logic grid deduction is whether the constraints are formalised. In a logic grid the conditions are written on the board as numbers or symbols, and solving is a search for the next forced move from the outset. In deduction the conditions are buried inside prose, photographs and audio. Work therefore begins with deciding what even counts as a clue, and that sorting is itself the hard part.
Neighbouring mechanics are told apart by where the verdict lives. Searching the archive and environmental observation name ways of gathering material, and point-and-click names an input scheme; none of them necessarily has a field in which a conclusion is submitted. Document inspection turns on matching against a rulebook, so the answer lives in the rules. Language decoding is the building of a correspondence table, which can be reused once fixed. If there is a field that asks, all at once at the end, whether the conclusion is right, the mechanic is deduction.
Further reading
On this site: designing the in-game notebook (the central essay on this mechanic) / designing hint systems / designing luck out of the puzzle / legible failure / the philosophy of Lucas Pope / the philosophy of Sam Barlow / reading an unreadable script / ten thinking puzzles on Steam, 2025 / getting it wrong is not a punishment: twenty developers on redesigning failure / Hanabi (2010)
Primary sources off-site: Wikipedia (English): Cluedo (the citations for devised 1943, filed 1944, granted 1947, released 1949), Wikipedia (English): Sherlock Holmes: Consulting Detective (how play and scoring work; its first-edition year disagrees with other sources), and GDC 2016: Making Her Story (Barlow's own talk). The naming of this mechanic is not settled, and sources differ on which work to call its first appearance. This lexicon records only years that could be confirmed and marks the rest as unconfirmed.
On the table
- Sherlock Holmes: Consulting Detective — read the casebook freely, answer the quiz at the end, compare your score with Holmes's. The skeleton of this mechanic is already in place.
- Hanabi — narrows the object of deduction to a single thing, your own hand, with other players' actions as the only clues.
See also
Logic Grid Deduction / Searching the Archive / Environmental Observation / Document Inspection / Language Decoding
