The elegant logic of grid-based picture puzzles — how a row of numbers encodes a hidden image, why solving them is a masterclass in constraint propagation, and the fascinating history behind one of puzzle culture's quietest obsessions.
May 16, 2026
38 min listen
Puzzle Formats & Logic
Audio coming soon — read the full episode below
1987First published nonogram, by Non Ishida in Japanese newspapers
NP-CComputational complexity of nonogram solving in the general case
1995Nintendo's Mario's Picross on Game Boy brings the format to millions
What is a nonogram?
A Hidden Picture Encoded in Numbers
Imagine a blank grid — say 10×10 cells. Above each column and beside each row, you're given a sequence of numbers. These numbers are your only guide. They tell you, for each row and column, how many consecutive filled cells appear — in order, separated by at least one empty cell. Using nothing but logic, you deduce which cells to fill and which to leave empty. When you're done, a picture emerges from the filled cells.
That's the entirety of the nonogram format, also called picross, griddlers, hanjie, paint by numbers, or Japanese crossword. The elegance is almost disturbing in its simplicity. There's no vocabulary to know, no geography to recall, no factual knowledge required. Just logic, a grid, and a set of numerical constraints. The picture isn't drawn — it's deduced.
This format has captivated millions of solvers since its independent invention by two people on opposite sides of the world in the late 1980s. It's been featured in Nintendo game cartridges, newspaper puzzle supplements, mobile apps with hundreds of millions of downloads, and world championship competitions. And yet outside of puzzle enthusiast circles, it remains surprisingly little known compared to its grid-based cousins sudoku and crosswords. Today we're fixing that.
A 7×7 Nonogram — Can You See the Picture?
1
3
5
5
5
3
1
1
3
5
7
5
3
1
Filled cell
Empty cell
This symmetrical diamond-like shape is a classic introductory nonogram. Row/column clues together uniquely determine exactly this solution.
How the logic works
Constraint Propagation: The Engine of Nonogram Solving
The mathematical concept underlying every nonogram solve is constraint propagation — the process of using what you know to deduce what you don't know, and then using those deductions to unlock further deductions, cascading information across the grid until every cell is determined.
Let's make this concrete. Suppose you have a 10-cell row with a clue of "7". That means somewhere in this row, there is a run of exactly 7 filled cells (and the remaining 3 cells are empty). The block can start in position 1, 2, 3, or 4 — those are its four possible positions. Now ask: is there any cell that is filled in ALL of those positions? Yes — positions 4, 5, 6, and 7 are covered by all four possible block placements. Therefore those four cells are forced, regardless of where the block actually starts.
This "overlap" technique — sometimes called the "leftmost-rightmost" method — is the fundamental tool. You imagine sliding the block as far left as possible, then as far right as possible, and any cells covered in both the leftmost and rightmost positions are forced filled.
1
Overlap analysis (the leftmost-rightmost method)
For each clue, find its leftmost possible position and its rightmost possible position. Any cells covered by both are forced filled. The larger the clue relative to the row length, the more overlap — and the more forced cells.
Forced = max(0, clue_length - (row_length - clue_length)) cells from center
2
Propagating known cells to neighbors
Every cell you fill or cross out (mark empty) changes the constraints for the row AND column containing it. A filled cell in row 3, column 5 reduces the range of valid block positions in both row 3 and column 5. These reduced ranges may force additional cells — which then propagate into other rows and columns.
Each deduction narrows the constraint space; new forced cells follow
3
Boundary detection — edges and known empties constrain blocks
If you've marked some cells as definitely empty, those cells act as hard boundaries for block placement. A clue "3" in a 10-cell row where cells 1 and 5 are known empty can now only start in positions 6–8 — and if a previously forced cell also sits in column 7, the block's position may be exactly determined.
Known empties divide rows into segments; constraints apply per segment
4
Completing and clearing — exhaustion and edge constraints
Once all cells in a clue's run are filled, the cells immediately on either side must be empty. Similarly, if a run can fit in only one remaining segment, it fills that segment exactly and clears the boundaries. Completed runs cascade empty markers that constrain remaining runs in the same line.
A fully-placed run forces empties at both edges; neighboring runs adjust
A well-designed nonogram is solvable entirely through constraint propagation — no guessing required. You work row by row, column by column, back and forth, each pass revealing new forced cells that constrain the next pass, until the grid is complete. The experience of solving a large, elegant nonogram this way feels remarkably like a proof by contradiction: every deduction is logically necessary, and the picture that emerges wasn't drawn — it was compelled by the constraints to exist.
Advanced solving strategies
Beyond Overlap: The Solver's Toolkit
Experienced nonogram solvers develop a repertoire of pattern-recognition shortcuts that allow them to apply constraint propagation much faster than reasoning from first principles each time. These strategies are essentially named instances of the general propagation logic — pre-compiled inference rules that the trained eye can apply instantly.
The Overlap Rule
When: large clue in long row
Slide the block to its leftmost and rightmost valid positions. Mark the overlap as filled. The larger the clue relative to the row length, the more cells you force immediately.
10-cell row, clue "7" → cells 4–7 forced
Edge Forcing
When: block touches a corner/edge
If a filled cell is at the very edge of a segment (bounded by known empties or the grid boundary), the run it belongs to must start at that boundary — which often forces the entire run to be determined.
Filled cell in position 1 with clue "4" → cells 1–4 all filled
Segment Fitting
When: run equals remaining segment
When the length of a remaining clue run exactly matches the length of the only open segment (bounded by empties), that run fills the segment completely — and the cells immediately outside the segment are empty.
Clue "5" in a 5-cell open segment → fill all 5 + cross both ends
Glue Detection
When: filled isolated cell sits between two blocks
A filled cell with empty neighbors can only belong to one specific run. Knowing which run "owns" the cell determines its contribution to the overlap for that run, often extending the forced region.
Isolated filled cell with multi-run clue → assign to nearest fitting run
Clue Sum Check
When: starting a new row/column
The minimum length required by the clues is: sum of all block lengths + (number of blocks - 1) gaps of at least 1. Subtracting from the row length gives the "slack" — how much room to move. Zero slack means the entire row is forced immediately.
Row 10, clue "3 3 3" → min length = 11 → impossible? No — re-examine clue
Cross-Reference Propagation
When: one axis is heavily constrained
After making deductions along rows, apply them to columns and vice versa. Information propagates across the grid as newly forced cells constrain both their row and column. The most satisfying solves unfold through several complete passes of alternating row-column propagation.
The nonogram format was born twice, independently, in the late 1980s — a testament to how natural and elegant the underlying idea is. Two inventors, on opposite sides of the world, discovered the same puzzle logic within a year of each other.
1987
Non Ishida publishes in Japanese newspapers
Tokyo-based puzzle designer Non Ishida creates "Window Art Puzzles" — small nonogram grids that produce simple pixel-art images. The puzzles appear in Japanese newspaper supplements and quickly attract a following. The format is intuitive, self-contained, and satisfying to solve without any language knowledge.
1988
James Dalgety enters the Sunday Telegraph competition
British surveyor James Dalgety independently invents the same format and enters it in the Sunday Telegraph's puzzle competition. He coins the term "nonogram" — from "non-sequential diagram" — to distinguish it from a misapplication of the term then common in British puzzle circles. Dalgety's coinage sticks in English-language publishing.
1990s
Western newspaper publication begins
The Telegraph and other major newspapers begin regular nonogram (or "hanjie") features. The puzzle gains traction among the same audience that already enjoys crosswords, though it remains a specialty format rather than mainstream. Different outlets coin different names: griddlers, paint by numbers, picross, hanjie — all the same puzzle.
1995
Nintendo releases Mario's Picross on Game Boy
Nintendo's decision to publish Mario's Picross on the Game Boy introduces nonograms to a vast new audience. The game features 256 puzzles across multiple grid sizes. The Game Boy's screen — with its small pixel grid — turns out to be a perfect medium for the format. "Picross" (from "picture crossword") becomes the name most recognized in gaming culture.
2007–2015
Nintendo's Picross DS and Picross e series (3DS eShop)
Nintendo's Picross DS (2007) and the Picross e series for the 3DS eShop (2013–2018) build an enormously dedicated following, particularly in Japan. The series features hundreds of puzzles in increasing size and complexity, with the largest grids reaching 25×20 cells. The e-series becomes one of the most-purchased eShop software lines.
2016–present
Mobile app era and competitive community
Mobile apps bring the format to hundreds of millions of players. Puzzle platforms like Nonograms.org host community-created puzzles in the millions. Competitive solving communities form online, with speed-solving categories for various grid sizes. Nintendo continues releasing Picross titles on Nintendo Switch, including the Picross S series and a collaboration with the Kemono Friends franchise.
Computational theory
The Mathematics: Why Nonograms Are NP-Complete
Nonogram solving sits at a fascinating intersection of recreational mathematics and theoretical computer science. While any specific nonogram puzzle you'd find in a newspaper or app is solvable in seconds, the general question — "can this set of row and column clues be satisfied?" — is NP-complete, meaning it's among the hardest class of problems known in computer science.
Computational Complexity at a Glance
Single row/column
O(n) — Linear
Full nonogram (general)
NP-Complete
Uniqueness verification
NP-Hard
The NP-completeness result (proven by Ueda and Nagao in 1996) applies to the general case — arbitrarily large grids with arbitrary clue patterns. For any specific puzzle you'd actually solve, the constraint propagation algorithms work efficiently. The theoretical hardness shows up in worst-case scenarios, particularly for very large grids or puzzles with highly ambiguous clue patterns that require extensive backtracking.
The uniqueness problem — verifying that a set of clues has exactly one valid solution — is particularly important for puzzle construction. A well-formed nonogram should have exactly one solution, determinable by logic alone without guessing. Constructors of large nonograms (30×30 or larger) typically use software to verify uniqueness, because manual verification is impractical.
The NP-completeness connection also explains why nonograms are such good brain workouts: constraint propagation is exactly the kind of structured logical reasoning that is computationally hard in general, even if specific instances yield to skilled human deduction remarkably quickly. Your brain is, in effect, running an approximate NP solver — and doing a surprisingly good job of it.
Cognitive science
The Cognitive Skills Nonograms Develop
From a cognitive science perspective, nonogram solving is unusually rich as a learning tool. It engages a set of transferable cognitive skills that educational researchers have found difficult to develop through traditional instruction — particularly spatial reasoning, constraint-based deduction, and metacognitive strategy switching.
Cognitive Skill
How Nonograms Develop It
Transfer Evidence
Spatial reasoning
Visualizing block positions across a grid; tracking which cells are forced by which constraints
Strong — spatial reasoning transfers to mathematics, engineering, and STEM tasks broadly
Logical deduction
Applying constraint propagation rules; deriving forced cell states from known partial information
Strong — deductive reasoning transfers to programming, formal proof, and systematic analysis
Working memory
Simultaneously tracking partial row and column solutions; holding multiple tentative block positions in mind
Moderate — working memory capacity has limited transferability but supports learning broadly
Pattern recognition
Rapidly identifying forcing configurations (overlap patterns, edge constraints, segment fits) without re-deriving from first principles
Strong — pattern recognition in structured domains is a core component of expertise across fields
Metacognition
Recognizing when you have insufficient information to proceed; knowing which strategy to apply and when to switch; monitoring your own error rate
Strong — metacognitive skills transfer directly to academic and professional problem-solving
Frustration tolerance
Distinguishing "insufficient information now" from "unsolvable" — learning to suspend judgment and gather more constraints before acting
Moderate — frustration regulation in structured problems supports persistence in open-ended challenges
The metacognitive dimension deserves particular attention. Expert nonogram solvers develop a finely calibrated sense of when they have enough information to commit to a deduction versus when they need to gather more constraints first. This "suspension of judgment under uncertainty" is one of the most valuable and transferable cognitive skills in any knowledge domain — and nonograms provide a uniquely pure training environment for it, because the puzzle's logical structure makes the distinction between "I know" and "I don't yet know" unusually crisp.
The art of construction
How to Create a Well-Formed Nonogram
Solving a nonogram is satisfying. Constructing one that is clean, logically unique, and reveals a recognizable picture at the end — that's an art. The nonogram constructor faces a set of constraints that are, appropriately enough, their own kind of puzzle.
1
Choose or design the image
Start with a pixel-art image on a grid. Simpler images (geometric shapes, silhouettes) produce more solvable puzzles. Overly complex images tend to create ambiguous regions that require guessing.
2
Generate the clues
For each row and column, read off the consecutive runs of filled cells from left to right (rows) or top to bottom (columns). Each run length becomes a number in the clue sequence.
3
Verify unique solvability
This is the hard part. Run a constraint propagation solver on your clues without the image as reference. Does it reach a unique solution using only logic? If not, the puzzle has multiple solutions — redesign.
4
Test solve manually
Solve your own puzzle from scratch without looking at the reference image. This reveals the "feel" of the solve — whether it has interesting constraint cascades, dead spots where progress stalls, or a satisfying reveal arc.
5
Tune the image for solvability
If unique solvability fails, modify the image — add or remove filled cells, simplify complex regions, ensure no 2×2 all-filled blocks appear in ambiguous areas (these often cause multiple-solution problems).
6
Calibrate for the intended audience
A beginner puzzle should have many large single-run rows and columns (maximum overlap forcing). An expert puzzle should have complex multi-run lines where the constraints interact across the grid.
The foundational 1996 paper proving that nonogram solving (and uniqueness verification) is NP-complete. Essential reading for understanding the computational theory.
A mathematical treatment of advanced solving strategies — combining constraint relaxation with propagation to solve difficult nonogram instances efficiently.
Thousands of free nonogram puzzles from the GWN network — beginner to expert, from 5×5 to 30×30 grids.
Listener Q&A
Your Questions Answered
What is a nonogram or picross puzzle?
A nonogram (also called picross, griddlers, or paint-by-numbers) is a logic puzzle where you fill cells in a grid to reveal a hidden picture. Row and column clues — sequences of numbers — tell you how many consecutive filled cells appear in each line, in order. Using logic alone, you deduce which cells to fill and which to leave empty until the complete image emerges.
Who invented nonograms?
Nonograms were independently invented in the late 1980s by two people: Japanese puzzle designer Non Ishida and British surveyor James Dalgety. Ishida published the first nonogram puzzles in Japanese newspapers in 1987–1988. Dalgety independently developed the same concept and entered it in the Sunday Telegraph's puzzle competition in 1988, coining the name "nonogram."
Are nonograms always uniquely solvable?
Well-designed nonograms have exactly one solution determinable by logic alone — no guessing required. Poorly designed puzzles may have multiple solutions or require trial-and-error. The constraint of unique logical solvability is why constructing a good nonogram is an art: the row and column clues must together constrain the grid to exactly one configuration.
What is constraint propagation in the context of nonograms?
Constraint propagation is the process of using known cell states to deduce the state of other cells. The overlap method is the classic example: if a clue says "7" in a 10-cell row, cells 4–7 must be filled regardless of where the block starts. As you fill forced cells, the constraints on neighboring rows and columns tighten, propagating information across the grid until the puzzle is solved.
What cognitive skills does solving nonograms develop?
Nonogram solving develops spatial reasoning, logical deduction, working memory, pattern recognition, and metacognitive awareness. The metacognitive skill — knowing when you have sufficient information to commit to a deduction versus when you need more constraints — is particularly valuable and transferable to professional problem-solving domains.
How hard is it to create a nonogram puzzle?
Creating a well-formed nonogram is significantly harder than solving one. The constructor must design an image, generate the row/column clues, and verify that those clues uniquely determine the image using logic alone. Large nonograms (30×30+) often require software verification. The art lies in choosing images with enough regularity to be solvable, while enough complexity to be interesting.