Hick's Law
Sliders and toggles multiply faster than anyone plans for. When a screen turns into a wall of peer options, people slow down. Hick’s law is the clean statement of that curve.
Hick’s law describes the time it takes for a person to make a decision as a result of the possible choices. Increasing the number of choices increases the decision time logarithmically, not in lockstep with each new row on the page.
That summary tracks how the experimental literature states the finding. Reaction time grows with the depth of a binary decision tree you would need to distinguish among equally likely alternatives, which is why doubling the count does not double the time. The same framing ties the law to information rate in choice-reaction tasks rather than to raw list length in every real interface.
Product teams cite the law when they argue for shorter menus and fewer primary actions. The practical read is conservative. Cutting options helps when each option truly competes for the same decision in the same moment. It does not help when you hide a task someone still has to finish later under an extra tap they cannot predict.
Long flat lists are a special case. If the user scans a shuffled list of commands to find a word they already know, each item gets a turn in working memory and time can grow about linearly with length. If the list is ordered predictably (alphabetical labels the user knows), people can subdivide and the logarithmic story fits better. Nielsen Norman Group’s Hick’s law clip on long menus states the user-facing version plainly. More choices mean longer time to a decision, though you can pair that constraint with structure so long menus stay workable.
Treat it as a design pairing problem. Grouping, hierarchy, defaults, recency, and search change the effective branching factor the user experiences in one glance. Deleting options alone rarely fixes a task people still have to finish.
On-call consoles stack choices when people are already overloaded. Playbooks, remediation buttons, and cross-service links all read as “just one more control” until the primary incident path competes with dozens of peers. You do not always need fewer buttons. You need a stable default path (ack, silence, escalate, open runbook) with everything else reachable but not co-equal on the first row.
When every action is red and loud, none of them are. When the same panel serves triage and deep diagnosis, split the modes or you pay twice. First you pay in scan time while someone picks among equals. Second you pay in mistakes when someone picks fast because the clock is loud.
Laboratory exceptions are worth humility. They are not a reason to ignore the main slope. Familiar stimuli, highly practiced responses, and some eye-movement tasks break the neat log relationship in published studies. “Never add a twelfth button” is too blunt a slogan. Each additional peer choice has a real time cost unless you change how people search or recall, and that cost is sublinear in the lab but still painful when latency is money or safety.
Use the law as a pressure gauge on density and on sameness of rank. Fix the information architecture before you polish the hover states. For motor pointing, Fitts’s law is the paired constraint on distance and target size.