Wordless instructions
A Socratic walk-through of wordless instructions — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why do assembly instructions drop words entirely when a written sentence could explain the step faster?
Open a flat-pack instruction booklet and there is not a word in it. A drawing shows a dowel, an arrow, a hand, a face that is either pleased or alarmed. Some steps are instantly clear. Others are ambiguous in a way that one short sentence — tighten this only after fitting the back — would fix completely.
So the wordless booklet is not obviously better at instructing. It is plainly worse at some steps. Which means it is being chosen for a reason other than instructional quality, and the interesting question is what that reason is and what it costs.
Reasoning it through
REASONING #Start with the economics, because that is where the pressure comes from.
A product sold into forty markets with a text manual needs forty translations. Each is a cost, but the cost is not only the translation fee. It is the version control — forty files that must all be updated when a fitting changes. It is the review, since a mistranslated safety warning is a liability. It is the physical bulk, because forty languages in one box means a thick booklet, more paper, more weight, and a printing plate change whenever a market is added. And it is the risk of a market being shipped the wrong booklet.
A wordless booklet collapses all of that to one artefact. Print one, ship everywhere, update once. The saving is not the translation fee; it is the elimination of an entire dimension from the production system.
Now the second driver, which is separable and often stronger: the drawing has to exist anyway. You cannot describe "align the cam bolt with the pre-drilled hole on the underside of panel C" in a way anyone can act on without showing them. So the illustration is not an alternative to the text — it is the load-bearing element, and the text was always the annotation. Once you accept that the picture carries the information, dropping the annotation looks like removing a cost rather than removing content.
So we have a real saving and a real reason. What is being given up?
Ask what language does that a picture cannot. Three things stand out. Negation — a picture shows a state; showing "do not do this" requires a convention, usually a red cross or a slash, which must be learned. Conditionals and sequence dependencies — "only if you fitted the long screws in step 4" has no natural pictorial form. And abstraction over time — "hand-tighten now, fully tighten after step 9" needs a way to point at a future step, which pictograms handle poorly.
Those are exactly the places where wordless instructions fail, and the failures are predictable from the list. That is the useful test of the account: the confusion is not spread evenly across the booklet but concentrated where the missing capabilities were needed.
There is one more piece. A pictogram is not self-explanatory — it is conventional, and its conventions are learned. The exploded view, the ghosted outline showing a previous position, the curved arrow for rotation, the magnified circle for a detail: all of these must be read, and readers who have assembled furniture before read them fluently while first-timers do not. The wordless booklet is not language-free. It is written in a visual language that happens not to need translating, which is a different and much narrower claim.
The analogy
THE ANALOGY #Think of road signs at an international border. A sign showing a stylised deer needs no translation and works for a driver from anywhere — provided they have learned that a red triangle means warning and that the picture inside names the hazard.
Now try to sign something a picture handles badly: no overtaking between 7am and 9am on schooldays. Every country ends up bolting a plate of text under the symbol, because the conditional and the time-window have no pictorial form. The symbol did not fail; it was asked to carry something outside its range.
A driver meets road signs constantly and learns the vocabulary through repetition, whereas most people assemble furniture rarely, so the pictographic conventions never become fluent — which means the instruction booklet is relying on a shared visual language its readers have far less practice in than the analogy suggests.
Clarifying the model
THE MODEL #"Pictures are universal" is the false premise underneath the practice, and it should be named. Pictograms rest on learned conventions, and some of those conventions are culturally variable: reading order, whether a hand is shown from the user's viewpoint, how a "wrong" state is marked, whether a face conveys the intended feeling. The claim that survives is much weaker and still useful — pictograms are cheaper to deploy across languages, not understood equally by everyone. Firms adopting them for the first reason often advertise the second.
The trade is being made by someone who does not bear its cost. The saving accrues to the manufacturer; the confusion is paid by the customer, in time and swearing, and only reaches the firm if it produces returns or support calls. That asymmetry predicts the observed equilibrium fairly well: booklets that are wordless throughout, including at the steps where a sentence would obviously help, because the marginal cost of the bad steps is borne elsewhere. It also predicts the exception — warnings with legal consequences usually do get text, in all forty languages, because there the liability lands back on the manufacturer.
Wordless is not the only way to avoid translation. Numbers, part codes and letters survive across languages without translation, which is why they are everywhere in these booklets. So the constraint is not really "no language" but "nothing needing translation", and the design space includes more than pictures — a point that is easy to miss when a booklet looks purely graphical.
Testing is the part most often skipped. The failures above are discoverable cheaply by watching a handful of first-time assemblers, and the firms whose instructions are conspicuously good are generally the ones that do this. That the technique is well known and unevenly applied says the constraint is not knowledge.
The falsification test. If the driver is localisation cost rather than instructional quality, then products sold in a single language market should carry more text in their instructions than otherwise-similar products sold internationally. If single-market products were just as wordless, the explanation would have to be about drawings being intrinsically better, or about design convention, rather than about translation.
A picture of it
THE PICTURE #How to readThis is a requirements chart repurposed as a capability comparison, so read the boxes as things an instruction has to achieve and the two elements as the notations available. A satisfies link means the notation does the job; the single traces link means it addresses the requirement only partially, by learned convention. The point is the pattern of missing links: the pictogram carries the two requirements at the top and cannot reach the two marked high risk, which are exactly the steps where real booklets confuse people. The risk fields are argued, not measured.
What became clearer
WHAT CLEARED #Wordless instructions are not a claim that pictures teach better than sentences. They are a production decision: one artefact instead of forty, updated once, with no translation risk — made easier by the fact that the drawing was always doing the real work. What is surrendered is a specific and predictable set of capabilities, negation, conditionals and forward references, and the confusion in any booklet clusters exactly there. The cost of that surrender falls on the reader rather than the firm, which is why the practice is applied even at the steps where one short sentence would obviously help.
Where to go next
ONWARD #- Why safety warnings still get translated text when the rest of the booklet does not.
- How pictogram conventions are learned, and which of them turn out to be culturally variable.
- Why part codes and numbers survive in wordless booklets, and what that says about the real constraint.