Use a small discrete-outcome encoding to support uncertainty recall

Use a small discrete-outcome encoding instead of a smooth continuous probability curve when readers need to remember a shown uncertainty distribution. For example, present about 20 discrete replication outcomes rather than one continuous density when displaying the sampling distribution from a single experiment.

A small set of discrete outcomes can be remembered as a shape. That made the shown uncertainty distribution easier to reproduce later in a graphical recall task, even after an intervening task.

Mechanism: Discrete outcomes compress the distribution into a limited visual pattern that readers can store and reproduce more easily than a continuous curve.

Evidence: Participants who viewed uncertainty as a small discrete-outcome display recalled the shown sampling distribution more accurately in the graphical recall task than participants who viewed a continuous display; the paper notes that many readers appeared to remember the discrete shape directly (Hullman et al., 2018).

Notes: The same clear advantage did not appear on the text probability recall task, so the benefit may be tied to remembering graphical shape rather than fully translating its meaning.

• User Goal: Remember the uncertainty shown for one reported experiment.
• Task: Graphically reproduce a previously shown uncertainty distribution.
• Data: One sampling distribution associated with a single experiment.
• Chart Setting: A display where uncertainty can be shown as a limited set of discrete outcomes.
• Audience: Readers with limited statistics training.
• Success Criterion: More accurate redraws of the shown distribution.

Break it when: Readers must estimate the uncertainty of a new study rather than recall one already shown. Why: Discrete displays reduced accuracy on the transfer task.

Sacrifice: You give up some precision and expressiveness compared with a continuous distribution. Risk: Some readers may remember the shape without understanding what the individual outcomes mean. Mitigation: Use this encoding when graphical recall of the shown distribution matters more than transfer to a new estimate.

Mistake: Use a smooth continuous uncertainty curve in a recall-focused view. Why it fails: Continuous conditions produced less accurate graphical recall than small discrete-outcome displays.

Failure Sign: Readers cannot redraw the shown uncertainty distribution after a short delay or intervening task. Quick Check: Compare a small discrete-outcome version against a continuous version and ask readers to redraw the distribution later. Stronger Test: Score the redraws against the shown distribution and compare error across the two encodings.

• Replace the continuous uncertainty curve with a discrete-outcome display.
• Keep the discrete-outcome count small rather than increasing the number of outcomes.
• Use the discrete version only on views where recall of the shown graphical distribution is the priority.
• Switch back to a continuous encoding if the same view must support new-study estimation.