Mental Models & Information Scent

The full lesson

Every time a user opens your product, they bring an invisible map in their head. That map holds expectations about how things work, where things live, and what labels mean. When your interface matches the map, it feels obvious. When it doesn’t, users stall, misclick, and eventually leave.

Mental models and information scent are two complementary ideas that explain both failure modes — and give you concrete tools to fix them.

What Is a Mental Model?

A mental model is a user’s internal picture of how a system works. Users build it from past experience: other products they’ve used, real-world analogies, marketing copy, and a few minutes of exploration. Mental models are always incomplete, often wrong, and surprisingly hard to change once formed.

Don Norman distinguished three related models:

Good design closes the gap between the system image and the user’s mental model. You rarely change the user’s mental model — you meet it where it already is.

Why Mental Models Are Durable

Jakob’s Law (from Laws of UX) states something important: users spend most of their time on other sites and apps, so they expect your product to work like those products do. That’s why conventions exist — not because they’re perfect in isolation, but because they’re already loaded into millions of users’ mental models at zero cost.

Trying to be novel with navigation, terminology, or interaction patterns forces users to update their mental model before they can accomplish anything. That friction is invisible on a Dribbble mockup and glaring in a usability session.

What Is Information Scent?

Information scent is how well a navigation element — a link label, icon, button, or heading — predicts what a user will find after clicking. The term comes from Peter Pirolli and Stuart Card’s Information Foraging Theory (1999), which models users as information predators following the strongest scent trail toward their goal.

High scent means the label closely matches the words the user already has in their head for that goal. Low scent means the label is ambiguous, full of jargon, or branded in a way that doesn’t map to the task.

The Scent-Killing Habits

Several common design patterns reliably destroy information scent:

• Branded section names — “The Hub,” “My Journey,” “Discover More” tell users nothing about what they contain
• Icon-only navigation without text labels, especially for anything beyond the five most universal icons (home, search, cart, account, menu)
• Hamburger menus on desktop — a well-documented anti-pattern that buries navigation entirely. Research consistently shows hamburger menus on desktop result in roughly 39% slower task completion and halved discoverability compared to persistent visible navigation
• Hover-dependent mega menus where sub-categories stay hidden until the user hovers, giving no upfront scent
• Vague calls-to-action like “Learn More” that give no preview of the destination

Evaluating Scent Without a Full Study

You don’t need an eight-week study to spot low-scent labels. Ask a colleague who hasn’t used the product to say out loud which navigation item they’d click for a specific task. If they hesitate or choose wrong, the scent is broken. This is a five-minute hallway test, not a formal method — but it catches the worst offenders fast.

For quantitative evidence, first-click testing (tools like Optimal Workshop’s Chalkmark) shows where users click first when trying to complete a task. A click map spread across many elements means weak scent everywhere. Clicks concentrated on the wrong element point to the specific misleading label.

How Mental Models and Information Scent Interact

Mental models define what label a user expects to see. Information scent is whether your label matches that expectation enough to trigger a click. Together they describe the full foraging loop:

1. The user forms a goal (“I need to cancel my subscription”)
2. They scan the interface for a cue that matches their mental model of where cancellation lives (“Account” or “Billing,” not “My Profile Settings”)
3. Each label either strengthens or weakens the scent trail
4. They click the strongest signal, scan the new page, and repeat

If your labels use your company’s internal vocabulary instead of the user’s vocabulary, you’ve broken both the mental model match and the scent in a single stroke.

Matching Interface Conventions to Mental Models

Because mental models are built from prior experience, the safest default is to follow established conventions. Innovate only where convention demonstrably fails your users.

Navigation Patterns That Match Mental Models

Tab bars on mobile reflect how users mentally model native apps — the navigation is always present and always shows where you are. Hiding it behind a hamburger menu forces a detour that interrupts the foraging loop.

Label Language

Aligning label language to user vocabulary is the highest-ROI investment in information scent. Practical sources:

• Support ticket analysis — the words users use when they can’t find something are the words your labels should use
• Search query logs — zero-result searches reveal terms users expect the system to understand
• Card sorting — open card sorts reveal how users naturally group and name content
• Tree testing — validates whether an existing IA lets users find things without the visual design providing false cues

Breadcrumbs, Wayfinding, and Persistent Orientation

Information scent doesn’t only govern where users go next — it also tells them where they are. Users who lose their sense of location in an IA become anxious and often restart from the home page, a costly behavior. Persistent wayfinding cues reduce that anxiety:

• Breadcrumbs show the path taken and make parent-level categories visible. They reinforce the site’s IA in the user’s mental model with every page visit.
• Active state styling on nav items — not just color, but shape, weight, or position change — to survive WCAG 2.2 non-color contrast requirements
• Page titles that match the nav label clicked. A mismatch here is a jarring scent break even after a correct click.
• Section headers on long pages that preview sub-content, extending scent below the fold

WCAG 2.2 introduced the “Focus Not Obscured” criterion (2.4.11 at AA, 2.4.12 at AAA). Sticky headers that cover focused elements can fail this. Breadcrumbs and visible nav that comply with WCAG 2.2’s updated target-size minimum (24 x 24 CSS pixels at AA) are the baseline for accessible wayfinding in 2026.

Progressive Disclosure and Scent Layering

Not every piece of information needs to be visible at once. Progressive disclosure reduces cognitive load while maintaining scent — but only when each layer gives users enough of a cue to know the next layer exists and what it contains.

The pattern works when:

• The summary label accurately represents the hidden detail (high scent)
• The expand affordance is obvious — not just a right-pointing chevron in the same color as surrounding text
• The hidden content is genuinely secondary, not primary-path content that most users need

It fails when designers use progressive disclosure to hide content that most users need on the critical path, treating it as a tidying tool rather than a genuine hierarchy signal. If 70% of users need to expand an accordion to complete their task, the accordion is working against you.

Updating Mental Models When You Must

Sometimes your interface genuinely needs to introduce a new interaction pattern — a novel data visualization, a conversational AI layer, a gesture-based workflow. In these cases, you need to actively help users build a new mental model. Don’t assume they’ll figure it out.

Strategies that work:

1. Onboarding that demonstrates, not just describes — show the interaction in a safe sandbox before the user encounters it in a real task context
2. Contextual tooltips triggered by first encounter — not a 12-step product tour at login, but a single tooltip that appears the first time the user is about to need the new pattern
3. Empty state design — a blank dashboard or empty list is a zero-cost onboarding opportunity. Use it to show what the space will contain and how to get there.
4. Analogy anchoring — frame novel patterns in terms of known ones (“like a spreadsheet, but…”) to give the user a starting model to update

What doesn’t work: expecting users to read documentation, burying a help link in a footer, or assuming a single onboarding modal will stick.

Applying This in Practice: A Diagnostic Checklist

Before launch — or as a quick audit of an existing product — run through these checks:

• Every navigation label uses words that appear in user research, support tickets, or search logs
• Primary navigation is visible without interaction on both desktop and mobile
• Page titles exactly match the navigation label that leads to them
• Active/current page state is marked in a way that doesn’t rely on color alone
• Icons in navigation have visible text labels (or are among the five universally recognized icons)
• Accordion/progressive disclosure is only used for genuinely secondary content
• Breadcrumbs are present on content pages more than two levels deep
• A first-click test on the three most common user tasks has been run in the last 12 months