Hick's law
Choice reaction time grows with the logarithm of the number of equally likely alternatives.
For a practised person responding to one of several equally likely signals, decision time rises roughly with the logarithm of the number of alternatives. The scope is narrow: simple, well-learned stimulus-response mappings with no meaning to read and no visual search. It says nothing about how long it takes to choose between options that carry semantics.
Reaction time against number of alternatives
How it shows up in software
The law is quoted constantly as a reason to cut options, and the citation is usually wrong. A dropdown of 30 country names is dominated by visual search and reading, which scale differently. Where it does apply is genuinely simple choices from practised, equally weighted signals: a segmented control, a small toolbar, a set of identical keys.
Using it well
- Use the law only for simple, practised choices between equivalent signals. For labelled options, optimise for scanning: sorting, grouping, search and good labels.
- Prefer breadth over depth when the target is findable by a rule, since each extra level adds a whole new decision.
- Set a default and a recommended option so most people make no decision at all, which beats shortening the list.
- Make the list searchable once it exceeds what a person can scan in a glance, rather than deleting options people need.
Where it turns manipulative
- Citing Hick to remove features or settings a person depends on. Cutting real function to shave milliseconds off a decision that was never the bottleneck is a loss dressed as a law.
- Reducing a choice to one option and a hidden link so there is technically no decision, which is a default imposed rather than a decision simplified.
- Collapsing consent choices into a single accept button while the decline path stays multi-level, then calling the asymmetry simplicity.
Where you have seen it
Google Search
The home page offers one input and a small number of practised actions rather than a directory of services.
iOS Control Centre
A small grid of toggles with fixed positions, learned by location, so choosing is closer to a practised response than to reading a list.
Stripe Dashboard
Long resource lists get filter and search rather than truncation, because the choices carry meaning that has to be read.
What the research says
- Hick, 1952Well evidenced
Reaction time to one of n lit lamps increased approximately with the logarithm of n, across practised participants.
The task was pressing keys in response to lights. There were no labels to read and no meaning to weigh.
- Hyman, 1953Well evidenced
Extended the result by varying alternative probabilities and sequence, showing reaction time tracked information content rather than raw count.
- Landauer and Nachbar, 1985Mixed evidence
Applied a logarithmic model to menu depth and breadth, finding broad shallow menus outperformed deep narrow ones for ordered numeric and alphabetic targets.
The menus were ordered numbers and words with an obvious search rule. Generalising to a labelled dropdown of unfamiliar options, where reading and visual search dominate, goes beyond what the data supports.
Grades are a judgement about the evidence, not about how useful the idea is. Plenty of contested effects are still worth knowing, as long as you do not cite them as settled.