← All Guides

Heuristic Evaluation, Fitts's Law, Hick's Law & F-Pattern Scanning: A Learning Guide

What You're About to Understand

After working through this guide, you'll be able to audit any digital interface and explain exactly why a button placement is costing conversions, why a navigation menu feels sluggish, or why users are missing your best content. You'll spot Fitts's Law violations on a checkout page, recognise when someone is misapplying Hick's Law to justify removing useful options, and know why "design for the F-pattern" is the opposite of good advice. You'll also be able to run a credible heuristic evaluation — and, critically, know what it can't tell you.

The One Idea That Unlocks Everything

Humans are satisficers with finite bandwidth.

Think of every user as a forager in a vast forest. They don't examine every tree. They move fast enough (not perfectly), decide quickly enough (not thoroughly), and scan just enough (not completely). Three biological bottlenecks shape every interaction:

Every UX law in this guide is a consequence of one of these three bottlenecks. Every design mistake is a failure to respect one of them. And heuristic evaluation is a structured method for checking all three at once.

If you remember nothing else: users perceive, then decide, then act — and each phase has a measurable speed limit.

Learning Path

Step 1: The Foundation [Level 1]

Forget theory for a moment. Open your phone. Try to tap the tiny "x" that closes a banner ad. Now tap a big "Add to Cart" button. The first tap was slow, hesitant, maybe you missed. The second was instant and confident. You just experienced Fitts's Law.

Fitts's Law says movement time depends on two things: how far away the target is (distance, D) and how big it is (width, W). The formula: MT = a + b × log₂(1 + D/W). Closer and bigger = faster. But the relationship is logarithmic — doubling button size doesn't halve movement time, it shaves off a fixed chunk. After a point, making things bigger has diminishing returns.

Now picture yourself at a restaurant. A menu with 3 entrées — you decide in seconds. A menu with 64 entrées — you freeze. That's Hick's Law: RT = a + b × log₂(N), where N is the number of equally likely choices. Each doubling of options adds about 150 milliseconds. Two choices: ~350ms. Four: ~500ms. Eight: ~650ms. Again logarithmic — going from 100 to 200 options barely matters, but going from 2 to 4 matters a lot.

Now open any news website. Watch where your eyes go. If the page is a wall of text, you'll sweep across the top, do a shorter sweep partway down, then skim down the left side. That's the F-pattern — discovered by Jakob Nielsen in 2006 via eye-tracking with 232 users. It's shaped like the letter F because users invest the most reading at the top (establishing context), less below (confirming relevance), and then just spot-check the left margin.

Finally, heuristic evaluation is the method that ties these together. Developed by Nielsen and Molich in 1990, it's a structured expert review of an interface against 10 principles (like "visibility of system status" and "error prevention"). Three to five evaluators independently inspect a design, record problems with severity ratings (0–4), then aggregate findings. It's fast, cheap, and catches about 75% of usability problems with five evaluators.

Check your understanding:
1. If a designer doubles the size of a button, does Fitts's Law predict that click time will be cut in half? Why or why not?
2. A user is browsing a dropdown list of 200 countries to find "Australia." Is Hick's Law the right framework to predict how long this takes? Why or why not?

Step 2: The Mechanism [Level 2]

Here's where it gets interesting: both Fitts's Law and Hick's Law are logarithmic for the same deep reason. Both emerge from the brain processing information in bits — and reducing uncertainty by roughly half at each step.

Why Fitts's Law is logarithmic: Your brain uses a feedback loop. It fires a fast ballistic movement toward the target, then corrects. Each correction cycle halves the remaining error. The motor cortex computes a "difference vector" — the gap between where your hand is and where the target is — and this gap shrinks exponentially. Since each correction takes constant time and halves the error, the total time scales as log₂ of the initial uncertainty (D/W). The motor system is literally transmitting information at a constant rate of about 10–12 bits per second.

Worked example: A button is 300px away and 50px wide. ID = log₂(1 + 300/50) = log₂(7) ≈ 2.8 bits. Now move it to 150px away: ID = log₂(1 + 150/50) = log₂(4) = 2.0 bits. Halving the distance saved 0.8 bits — roughly 0.8 × b milliseconds. If you instead doubled the width to 100px: ID = log₂(1 + 300/100) = log₂(4) = 2.0 bits — identical improvement. Key Insight: Reducing distance and increasing size are mathematically interchangeable in Fitts's Law. Designers who only think about button size are missing half the toolkit.

Why Hick's Law is logarithmic: Humans don't scan options one by one. They subdivide — like a binary search. "Is it in this half? No? Then the other half. Now which half of that?" With N options, you need about log₂(N) such halvings, each taking a constant ~150ms. This maps directly to Shannon's information entropy: N equally probable options contain log₂(N) bits of information.

Why the F-pattern has its specific shape: The top bar exists because users front-load their effort — "Is this page even relevant?" The first few lines get near-complete reading. The second, shorter bar is a confirmation check — "Is there more good stuff below?" — but motivation is already declining. The left-side stem exploits a habit from left-to-right reading: the first few words of each line act as a table of contents. The whole pattern is an optimisation strategy: minimise reading effort while maximising information yield. It's information foraging — identical in structure to how animals forage for food.

Why heuristic evaluation works (and where it doesn't): Nielsen's 10 heuristics were derived from factor analysis of 249 real usability problems. They capture fundamental mismatches between human cognition and interface design — memory limits, attention limits, mental model conflicts. These mismatches haven't changed since 1994 because human cognitive architecture evolves on millennia-scale timelines. But the method has a fundamental ceiling: evaluators imagine user problems rather than observing them, and the "curse of knowledge" means experts can't un-know their expertise. This is why only ~36% of heuristic evaluation findings are confirmed as real problems in usability testing.

Check your understanding:
1. A colleague proposes putting the primary CTA at the exact centre of the page to "minimise average distance from all starting positions." Using Fitts's Law, what's the flaw in this reasoning for desktop mouse users? (Hint: think about screen edges.)
2. Why does the F-pattern break when content uses strong headings, bullet points, and bold text?

Step 3: The Hard Parts [Level 3]

Everything above works cleanly in a textbook. The real world is messier. Here's where the simple models crack.

Fitts's Law on touch: a different universe. The most common misapplication of Fitts's Law is carrying desktop assumptions to mobile. On a mouse, screen edges are "infinitely wide" targets — the cursor stops at the boundary, so you can overshoot and still hit. This is why macOS's top menu bar is ~5× faster than Windows menus inside windows. On touch, this advantage reverses. The finger IS the pointer — overshoot the edge and you're off the device. Edges and corners become the hardest spots to tap, not the easiest. There's no cursor, so the starting position is unknown. The finger occludes the target ("fat finger problem"). The standard formula needs modification — Bi & Zhai's FFitts model (2013) uses dual Gaussian distributions for intended touch point vs. perceived target centre, but there's no consensus model yet.

Hick's Law doesn't apply to experts — or to recognition tasks. An expert typist uses 26+ keys. Hick's Law predicts ~700ms per keystroke (4.7 bits × 150ms). Actual expert typing: ~200ms per keystroke. Practice converts conscious decisions into automatic motor programs, shifting neural processing from the prefrontal cortex (deliberate, serial) to the basal ganglia (automatic, parallel). The law also fails for recognition tasks — browsing a list of search results isn't "deciding between alternatives," it's pattern-matching. Kveraga et al. (2002) found saccadic eye movements are completely unaffected by stimulus-response uncertainty, directly violating Hick's Law. The takeaway: Hick's Law governs novel decisions between alternatives, not all interaction.

The evaluator effect is unsolvable. Agreement between any two heuristic evaluators ranges from 5% to 65%. This isn't a training problem — it persists even among experienced evaluators. The root cause: usability isn't a single objective property. It's an emergent property of the interface-user-context system. Two equally expert evaluators literally attend to different things because attention is selective and what "counts" as a problem depends on implicit, personal models of user competence. Training improves agreement modestly but can't eliminate it. The practical implication: never rely on a single evaluator, and never use heuristic evaluation as your sole method.

The F-pattern may be partly a statistical artifact. The famous heat maps average across hundreds of users, producing a clean F shape. But individual gaze plots show far more variable behaviour. Critics (EyeQuant and others) argue the "F" is partly an artefact of aggregation methodology — the population-level tendency is real, but it poorly represents any individual user's actual behaviour. The F-pattern is best understood as a gravitational tendency in a population, not a rule each user follows.

Check your understanding:
1. Your team is designing a mobile checkout and someone says "put the Buy button in the bottom-right corner — corners are Fitts's Law sweet spots." What's wrong with this reasoning?
2. A heuristic evaluation of an e-commerce site found 41 usability problems; a subsequent usability test found only 10. Does this mean the heuristic evaluation was more thorough? What's the catch?

The Mental Models Worth Keeping

1. Perceive → Decide → Act (the interaction triad)
Every user interaction passes through three phases, each governed by a different law. Scanning/seeing = F-pattern (perceptual). Choosing what to do = Hick's Law (cognitive). Clicking/tapping = Fitts's Law (motor). Diagnosing a conversion problem means identifying which phase is the bottleneck.
Example: Users aren't clicking your CTA. Is it because they can't see it (perception — bad placement relative to scan pattern)? Can't decide what to do (cognition — too many competing options)? Or can't reach it easily (motor — too small or too far away)?

2. The Logarithmic Uncertainty Principle
Both Fitts's and Hick's Laws are logarithmic because the brain reduces uncertainty by halving at each step. This means the first few additions (of distance, of options) hurt the most. Going from 1 to 2 choices is devastating. Going from 50 to 51 is invisible.
Example: Adding a second CTA to a hero section doubles decision time. Adding one more item to a 30-item menu changes nothing measurable.

3. Information Foraging
Users behave like animals foraging for food — they follow "information scent" (cues that suggest relevance) and invest effort proportional to expected payoff. The F-pattern is a foraging strategy. Good headings and front-loaded keywords are high-scent markers.
Example: A product page buries the price in paragraph 4. The information scent is weak, so foragers abandon before finding it.

4. Expert Simulation ≠ User Observation
Heuristic evaluation predicts problems; usability testing observes them. These find different things. HE catches more issues (especially cosmetic ones) but confirms only ~36% as real. Testing catches fewer but more severe problems (70% major vs. 12% for HE). Neither alone is complete.
Example: An evaluator flags a label as "unclear." In testing, every user ignores the label and taps the icon instead. The "problem" doesn't exist for real users.

5. Design for Satisficers, Not Optimisers
All three laws describe satisficing behaviour — users move fast enough, decide quickly enough, scan just enough. They never optimise. Designing for idealised thorough-reading, careful-comparing users is designing for people who don't exist.
Example: A long-form comparison table assumes users will read all columns. Real users scan the first two rows, check the price column, and decide.

What Most People Get Wrong

1. "Bigger buttons are always better"
Why people believe it: Fitts's Law says bigger targets are faster. True, but logarithmic — the benefit diminishes rapidly. What's actually true: Reducing distance is often more impactful than increasing size. Moving a CTA closer to where the cursor rests can matter more than enlarging it. Tell the difference: If the button is already 44pt+ and near the cursor's resting zone, making it bigger won't help. Look at distance first.

2. "Fewer options are always better"
Why people believe it: The famous jam study (24 jams → 3% purchase; 6 jams → 30%) plus a simplified reading of Hick's Law. What's actually true: The jam study has failed to replicate consistently — a 2010 meta-analysis found the average effect of assortment on purchase was essentially zero. Hick's Law applies to decisions, not recognition or browsing tasks. Aggressively reducing navigation options can make things worse. Tell the difference: Ask whether the user is choosing between alternatives (Hick's applies) or searching for a known item (it doesn't).

3. "The F-pattern is how you should design pages"
Why people believe it: Designers hear "users scan in an F" and place all key content along the F shape. What's actually true: The F-pattern is what happens when design fails to guide the eye. It's a "laziness" pattern triggered by dense, unformatted text. Good formatting (headings, bullets, bold text) breaks the F-pattern and produces better scanning patterns like "layer-cake" or "spotted." Tell the difference: If your page triggers the F-pattern, that's a signal to improve formatting, not a validation of your layout.

4. "Heuristic evaluation replaces usability testing"
Why people believe it: HE finds 4× more problems and is cheaper. What's actually true: HE skews heavily toward cosmetic issues (only 12% of identified problems are major). Usability testing finds fewer total issues but 70% are major. And evaluators only agree 5–65% of the time. Tell the difference: If your most recent CRO insight came only from expert review and was never validated with real user data, treat it as a hypothesis, not a finding.

5. "Fitts's Law works the same on touch as on mouse"
Why people believe it: The law is presented as universal. What's actually true: On touch, screen edges become the hardest targets (opposite of mouse), there's no cursor starting position, and the finger occludes the target. The "infinite edge" effect that makes macOS menus 5× faster simply doesn't exist on touch. Tell the difference: When someone cites Fitts's Law for a mobile design, check whether they're applying mouse-era assumptions.

The 5 Whys — Root Causes Worth Knowing

Chain 1: "Fitts's Law is logarithmic"
Claim: Movement time scales with log₂(D/W) → Why? The motor system works as an information channel, transmitting bits of positional information → Why? Neural feedback loops reduce error exponentially — each correction halves the gap → Why? The motor cortex uses a difference-vector computation where correction magnitude is proportional to remaining error → Why? Neural population coding uses rate-coded signals where output magnitude tracks input magnitude → Why? This is a fundamental property of how neuron populations integrate information — not designed, but a consequence of circuit dynamics.
Root insight: The logarithm isn't a mathematical convenience. It's a physical property of how neurons process uncertainty.

Chain 2: "Users only read 20–28% of words on a page"
Claim: Most web content goes unread → Why? Users scan to decide whether to read, not because content is bad → Why? Cost-benefit: thorough reading has poor ROI for most pages → Why? Information foraging theory: users invest effort proportional to expected value → Why? Every page competes with millions of alternatives one click away → Why? Digital media shifted the scarce resource from information to attention.
Root insight: Scanning is a pre-reading behaviour. You must win the scanning battle before you can win the reading battle. Formatting serves scanning; quality serves reading. They're sequential, not substitutable.

Chain 3: "Heuristic evaluation has ~34% false positive rate"
Claim: A third of identified problems aren't real → Why? Evaluators imagine difficulties real users don't experience → Why? They lack context about actual user goals and workarounds → Why? HE is a simulation of user behaviour, not observation of it → Why? The curse of knowledge prevents experts from un-knowing their expertise → Why? There's a fundamental epistemological gap between prediction and observation — all user models are wrong.
Root insight: Expert evaluation and user observation find different problem sets. Neither subsumes the other.

The Numbers That Matter

10–12 bits/second — The throughput of the human motor system (Fitts, 1954). Remarkably constant across tasks. To put it in perspective: that's about the data rate of tapping out Morse code. Your hand is a slow, precise modem.

~150ms per bit — The slope of Hick's Law. Each doubling of equally-likely options adds 150ms of decision time. Going from 2 to 4 choices costs the same as going from 50 to 100.

20–28% — The percentage of words users actually read on a web page (NNGroup). That means your product page's three paragraphs of lovingly-crafted copy? Users read maybe the first sentence of each.

80% — The portion of viewing time spent on the left half of pages in left-to-right reading contexts. The right side of your page is a ghost town for casual scanners.

75% / 35% — Five evaluators catch ~75% of usability problems. One evaluator catches ~35%. The jump from 1 to 3 evaluators (35% → 60%) is where the real value is. After 5, diminishing returns hit hard.

5–65% — The range of agreement between any two heuristic evaluators. That's not a typo. Two experts can look at the same interface and agree on as few as 5% of the problems. This is the "evaluator effect" — and it has no known fix.

~36% — The "hit rate" of heuristic evaluation. Only about a third of problems identified by evaluators are confirmed as real issues in usability testing. The rest are false positives — problems that exist in theory but not in practice.

5× faster — How much faster macOS menu acquisition is compared to Windows, because Mac places the menu bar at the screen top (infinite depth) while Windows puts it inside application windows (finite depth). One of the most dramatic real-world Fitts's Law demonstrations.

44×44 pt / 48×48 dp — Apple's and Google's minimum touch target sizes. Below these thresholds, Fitts's Law predicts (and testing confirms) that error rates spike.

Where Smart People Disagree

Does Fitts's Law apply to touch at all?
Steven Hoober (2022) argues it fundamentally doesn't — no cursor, no infinite edges, unknown hand position, finger occlusion. The basic assumptions break. Bi & Zhai (2013) counter that their FFitts model, with modified parameters and dual Gaussian distributions, fits finger-pointing data well. Both sides agree the standard formulation needs modification; they disagree on whether the result is still meaningfully "Fitts's Law." Unresolved. The field lacks a consensus touch model.

Is choice overload real?
Iyengar & Lepper's jam study (2000) launched an industry of "fewer = better" design advice. Scheibehenne et al.'s 2010 meta-analysis found the average effect was essentially zero — possibly publication bias. Chernev et al. (2015) offered a reconciliation: choice overload is real, but only under specific conditions (high complexity, low expertise, unclear preferences). The oversimplified version persists in CRO folklore long after the science became nuanced.

Is the F-pattern real or a statistical artefact?
NNGroup's position: robust, replicated across studies. Critics (EyeQuant and others): heat maps average out individual variation, and individual gaze plots show diverse patterns. The "F" may emerge from aggregation methodology more than from individual behaviour. Both sides have valid points — it's a real population tendency that poorly represents any individual user.

Is heuristic evaluation still worth doing?
Proponents: it's cost-effective, fast, and catches most problems early. Critics: 34% false positive rate, evaluator effect, skews toward cosmetic issues, gives false confidence. Current consensus: a useful complement to user testing, never a replacement. But "complement" is doing heavy lifting in that sentence.

Is the information-theoretic basis of Fitts's Law real or metaphorical?
MacKenzie's Shannon formulation is enshrined in the ISO 9241-411 standard. Heiko Drewes (LMU Munich) has published ongoing critiques: the connection to Shannon's theorem is superficial, with no rigorous derivation from information theory to motor behaviour. The formula works empirically but the "why" remains debated. This matters because it determines whether we can extend the law to new domains (VR, BCIs) or whether it's just a good curve fit for mice and keyboards.

What You Don't Know Yet (And That's OK)

After absorbing this guide, you can confidently apply these laws to desktop and mobile web interfaces and run a competent heuristic evaluation. Here's where your knowledge runs out:

Subtopics to Explore Next

1. Information Foraging Theory (Pirolli & Card)
Why it's worth it: It's the theoretical backbone of the F-pattern — once you understand foraging, you can predict scanning behaviour in any context, not just web pages.
Start with: Search "information scent" and read Pirolli & Card's 1999 paper abstract, then NNGroup's articles on information scent.
Estimated depth: Medium (half day)

2. Choice Architecture & Nudge Theory (Thaler & Sunstein)
Why it's worth it: Extends Hick's Law into behavioural economics — how defaults, framing, and salience modify decision time and quality beyond just "number of options."
Start with: Read chapters 1–3 of "Nudge" (2008), then look for "choice architecture" in CRO contexts.
Estimated depth: Medium (half day)

3. Nielsen's 10 Heuristics — Deep Dive with Examples
Why it's worth it: The heuristics are only useful when you can recognise violations instinctively. This requires studying dozens of examples per heuristic.
Start with: NNGroup's updated 2020 article on the 10 heuristics, then their video series walking through examples.
Estimated depth: Medium (half day)

4. Eye-Tracking Methodology & Alternative Scan Patterns
Why it's worth it: Unlocks layer-cake, spotted, commitment, Gutenberg, and Z-patterns — the full vocabulary of how users actually scan, beyond the oversimplified F-pattern.
Start with: NNGroup's "Text Scanning Patterns: Eyetracking Evidence" and "Eyetracking Web Usability" (2010 book).
Estimated depth: Medium (half day)

5. Touch Interaction Design (Beyond Fitts's Law)
Why it's worth it: Mobile is the majority of web traffic; applying mouse-era Fitts's Law to touch is a common and costly error.
Start with: Steven Hoober's 2022 Smashing Magazine article "Fitts' Law In The Touch Era."
Estimated depth: Surface (1–2 hours)

6. The Power Law of Practice & Expertise
Why it's worth it: Explains exactly when and why Hick's Law stops applying — critical for designing interfaces that serve both novices and experts.
Start with: Search "power law of practice Newell Rosenbloom" and read about the novice-to-expert transition.
Estimated depth: Surface (1–2 hours)

7. Cognitive Walkthrough vs. Heuristic Evaluation
Why it's worth it: Cognitive walkthrough focuses on learnability for new users — a different lens than HE's broad heuristic sweep. Knowing when to use which saves time and catches different problems.
Start with: Search "cognitive walkthrough method" and compare with heuristic evaluation on NNGroup.
Estimated depth: Surface (1–2 hours)

8. Fitts's Law in Brain-Computer Interfaces & Motor Rehabilitation
Why it's worth it: Reveals the law's limits — signal-independent noise in BCIs violates Fitts's assumptions, and throughput (bits/second) serves as a clinical measure of motor recovery after stroke.
Start with: Search "Fitts's Law BCI throughput" in PubMed/PMC.
Estimated depth: Deep (multi-day)

Key Takeaways

Sources Used in This Research

Primary Research:
- Fitts, P.M. (1954). "The information capacity of the human motor system in controlling the amplitude of movement." Journal of Experimental Psychology.
- Hick, W.E. (1952). "On the rate of gain of information." Quarterly Journal of Experimental Psychology.
- MacKenzie, I.S. (1989). "A Note on the Information-Theoretic Basis for Fitts' Law." Journal of Motor Behavior.
- Bullock & Grossberg (2006). "Fifty years later: a neurodynamic explanation of Fitts' law." PMC.
- Nielsen, J. (1994). "Usability Inspection Methods." Wiley. / 10 Usability Heuristics (NNGroup, updated 2020).
- Nielsen, J. (2006/2017). "F-Shaped Pattern for Reading Web Content." NNGroup.
- Callahan et al. (1988). "An empirical comparison of pie vs. linear menus." ACM.
- Kveraga, Boucher, & Hughes (2002). "Saccades operate in violation of Hick's law." ResearchGate.
- Hertzum & Jacobsen (2001). "The Evaluator Effect: A Chilling Fact About Usability Evaluation Methods."
- Proctor & Schneider (2018). "Hick's law for choice reaction time: A review." Purdue.
- Bi & Zhai (2013). FFitts model for touch interaction (referenced in multiple sources).
- Scheibehenne et al. (2010). Meta-analysis on choice overload.
- Chernev et al. (2015). Choice overload moderators.

Expert Commentary:
- Hoober, S. (2022). "Fitts' Law In The Touch Era." Smashing Magazine.
- Tognazzini, B. "A Quiz Designed to Give You Fitts." AskTog.
- Atwood, J. "Fitts' Law and Infinite Width." Coding Horror.
- Sauro, J. "How Effective are Heuristic Evaluations?" MeasuringU.
- EyeQuant. "Eye Tracking Studies: Does the Famous F-shape Pattern Really Exist?"

Good Journalism:
- Smashing Magazine (2012). "Redefining Hick's Law."
- Scandiweb. "How Heuristic Evaluation Can Increase Your Revenue."

Reference:
- Wikipedia: Fitts's law, Hick's law.
- Laws of UX (Jon Yablonski): Fitts's Law, Hick's Law.
- Interaction Design Foundation: Fitts's Law, Hick's Law.
- Nielsen Norman Group: Severity ratings, scrolling and attention, text scanning patterns.