Disfluency Penalty

What it is

The disfluency penalty is the specific, unflattering half of processing fluency. Fluency says that ease of processing is read as a property of the thing processed. The penalty is what happens when the ease runs out. Effort of comprehension does not get filed where it belongs, as a fact about the encounter between a reader and a page. It gets filed as a fact about the page, and then as a fact about whoever made it.

Oppenheimer's 2006 paper in Applied Cognitive Psychology is the cleanest statement of it, and its title is a joke at the expense of its own subject: "Consequences of erudite vernacular utilized irrespective of necessity: problems with using long words needlessly". Experiments 1 to 3 manipulated the complexity of a text. Longer words and more elaborate constructions, added without adding meaning, lowered judged intelligence of the author, and the mediation analysis put processing fluency in the causal path. That much could still be explained away. A reader might reasonably think that someone reaching for unnecessary vocabulary is covering for something, in which case the judgement is inferential rather than perceptual, and arguably correct.

Experiment 4 closes that escape route. The text was held constant and only its legibility was varied. Nothing about the writing changed, nothing about the vocabulary changed, nothing about the argument changed. A harder-to-read font was enough to lower the estimate of the author's intelligence. There is no inference available there about the writer's judgement, because the writer made no choice that the reader could be responding to. What moved was the reader's own experience of decoding, misattributed to the person on the other end.

That is why this concept earns a page separate from fluency in general. Fluency is a broad claim about judgement under low effort. The penalty is a narrow and testable consequence with an identifiable victim. Someone is being judged, and the variable doing the judging is not in their work.

The practical extension, which is the reason the concept appears on this site, is that unfinished work is disfluent by definition. Missing sections, placeholder text, an argument that has not yet found its order, a rough edit that still has its seams showing: every one of those raises the cost of understanding. Oppenheimer's font manipulation establishes how little of that cost it takes before the penalty attaches. If a typeface is enough, a half-built thing is more than enough.

There is a companion finding in the same literature that pulls the other way and should be held alongside this one. Disfluency also slows people down and makes them more careful, which is why harder-to-read type has been used deliberately to produce more sceptical reasoning. Ease is not simply good and difficulty is not simply bad. The direction of the effect depends on what you are asking the reader to do. What the penalty says is that if you are asking them to judge, difficulty costs you.

In effect

Research: the fourth experiment, where the content stops mattering

The load-bearing result is Experiment 4, and it is worth being precise about why. Experiments 1 to 3 are a study of writing: they show that a stylistic choice can be read as evidence about the person who made it. Experiment 4 is not a study of writing at all. The author made no choice. A property of the printed page, invisible as a variable to the person reading, moved the judgement of a person they had never met. That is the strongest version of the claim available in this literature, and it is the version the site relies on. The theoretical framework it sits inside was laid out two years earlier by Rolf Reber, Norbert Schwarz and Piotr Winkielman in Personality and Social Psychology Review: the more fluently something is processed, the more positive the response to it, and people draw on that experience for judgements that have nothing to do with aesthetics, including judgements of truth.

Everyday life: the draft that is read as bad thinking

The everyday case is the one this concept exists to name. Hand someone a piece of work that is genuinely rough and two different things are available to be judged: the quality of the thinking, and the difficulty of following it. The reader has no reliable way to keep those separate, and no particular motive to try, because the difficulty does not present itself as difficulty. It presents itself as an impression of the work. A draft with its argument in the right order is processed as a draft. A draft with its argument out of order is processed as bad thinking, and the person who wrote it is downgraded accordingly. The useful distinction for anybody showing work early is therefore not finished against unfinished. It is legible against illegible, and legibility is cheap.

Commerce: names, tickers and the price of being hard to say

Our notes record a set of archival findings in the same fluency family, reported in Adam Alter's Drunk Tank Pink: newly listed stocks with easier-to-pronounce ticker symbols and company names performing better in early trading, and lawyers with more pronounceable surnames reaching partnership faster than comparable peers. Those are the disfluency penalty operating on people and firms rather than on drafts, and they are also the weakest evidence on this page, because they are archival rather than experimental and our notes attach no journals, years or sample sizes to them. See Processing Fluency for the caveats in full. They are worth naming because they suggest the penalty is not confined to a laboratory rating scale, and worth holding loosely for exactly the reason stated.

What it does not say

It does not say that expertise fails to protect against the penalty, and it does not say that expertise protects against it either. Oppenheimer's participants were ordinary readers, not domain experts assessing work in their own field. It is reasonable to expect that a practised editor reading a rough draft has learned to correct for roughness. It is also possible that the correction is incomplete, and that the expert feels the drag without ever attributing it to the format. Nothing cited here tests it. The question is open, and anyone asserting that the expert is immune is guessing.

It does not say that showing work early is a mistake. It prices one cost of doing so. Anything iterative gets cheaper to correct the earlier it is examined, and a rule of withholding until complete guarantees that every error is found at the point of maximum expense. Both costs are real, and this page only measures one of them.

It does not say that clarity is always the goal. The same literature reports disfluency producing more careful and more sceptical reading. If what you want from a reader is scrutiny rather than a verdict, ease is working against you.

It does not say the penalty is about the audience. Nothing in this literature sorts viewers by discernment, taste or intelligence. The variable that moved in Experiment 4 was the format, and it moved for everyone.

It carries no effect sizes on this page, because our notes carry none. The 2006 paper's numbers should be read from the paper by anyone who needs them.

---