Navigation and link labels: what NN/g research shows

Navigation decides whether the rest of a page ever gets seen, and a link label decides whether the click lands where the reader expects. NN/g measured part of that cost across two decades of articles, and the measurements are more useful than the slogans built on top of them. The navigation topic index collects the pieces behind this post.

Two questions show up in the archive. Did people find the navigation, and how much effort did it take once they started looking? The hidden navigation study tackled the first with a quantitative test in 2016: 179 participants, 6 sites, three conditions.

Every figure in the next section comes from that one study, so they share the same limits. One round, six sites, one style of interface.

Hidden menus: when they work and what they charge

On desktop, use of the navigation fell from 48% when it was visible to 27% when it was hidden. On mobile the gap was wider: 86% with the combined menu against 57% with the hidden one. Content discovery dropped by more than 20%. Perceived difficulty climbed 21% against the visible menu and 11% against the combined one. Tasks ran at least 39% longer on desktop and 15% longer on mobile. Reaching the navigation cost an extra 5 to 7 seconds on desktop and about 2 on mobile (NN/g).

A drop like that has a cost, but it isn’t a verdict. The 2025 study of icon recognizability confirms the three-line icon is now a familiar convention that most people read as the main menu. It also states what it didn’t measure: recognition, not task time or success. Recognizing the icon doesn’t remove the price of hiding the menu.

On placement, the articles agree. Don’t hide the navigation on desktop. On mobile, show up to four top-level links, and hiding the rest is acceptable past that count. Use the standard icon, put it in the corner people expect, label it “Menu”, and give it contrast and padding. Then keep other routes open, with links in the body, a visible search field and a footer.

How many items before the list becomes a scroll

An expanding menu carries two costs. First the reader has to notice the trigger, then pick an item. The review of expandable menus splits the second cost into visual search plus pointer movement. Fitts’s law governs the movement, so a small target far from the cursor costs more time. A linear menu therefore punishes its own last items, which is one reason it should stay short.

Layout can bring that cost down. Aligning the trigger with the middle item cuts the average travel. A rectangular megamenu keeps its items closer on balance. A radial menu makes every item equidistant and pays for being unfamiliar. A marking menu turns direction into a shortcut and, for the expert, skips the menu.

With touch, the question changes. The finger may not rest on the trigger when the menu opens, so what counts is where the panel appears relative to where the hands probably are.

Dropdowns and the corridor

The guidance on dropdowns treats the widget as overused and names four legitimate jobs: issuing a command, navigating, filling a form and choosing an attribute. Keep the label visible while the menu is open. Disable an unavailable option in gray instead of removing it, so the position people learned stays put. A long dropdown with scrolling forces the pointer down a corridor, a cost the steering law predicts. Where typing is faster, as with a state or a country, take the text field instead. The same goes for a value people know by heart, like a birth month. On desktop, keep global navigation out of a dropdown entirely, and support the keyboard.

Avoid one more habit: a menu that changes because of another menu on the same page.

Breadcrumbs: a thin line with a real job

The 2007 case for breadcrumbs frames them as secondary one-line navigation. They show where the page sits, give one-click access to the levels above, help someone who arrived from search or an external link, and take almost no room. The author states he has no exact figure for how many people use them and didn’t think the question justified its own study.

The 2018 guidance, revised in 2026, tightens the rules. A breadcrumb complements navigation rather than replacing it, and it shows site hierarchy rather than session history. On a site with several paths it shows one canonical path. Include the current page as the last item without a link. Link only real pages. Skip the trail when the hierarchy is one or two levels deep, and use a one-character separator.

Link labels and the four S

A label has one job before the click: say what is on the far side. The guidance on link labels lists four qualities, all starting with S. Specific means the label names the destination, and “learn more” is the new “click here”. Sincere means the link is a promise, and a broken one shrinks the appetite to click again. Substantial means the label must survive on its own, because people scan links while skipping the sentence around them. Succinct means no word too many, with the useful part first.

The four pull against each other, and the article admits it. Specific pushes a label longer, and longer is harder to scan. The common fix is to lead with the noun the reader is hunting for. That is the same problem as matching search intent, where the label and the query have to meet.

Visited links and what the visual state decides

Link appearance carries information as well as style. The guidelines for visualizing links once asked for color plus underline. A 2026 editor’s note withdraws that visual rule and points instead to WCAG 2.1 success criterion 1.4.1, which requires that color alone doesn’t separate a link. Most readers now combine color, position and context.

What survived the revision is a warning. Don’t use the link color on text that isn’t a link, and don’t let a link read as plain text.

The visited state is quieter and more useful. The article on visited links argues that a changed color tells people where they have been, where they are, and where they can still go. Ignore the convention and people reopen pages by accident, lose their place, mix up two similar links and quit sooner.

This problem is hard to catch in testing. Almost no one complains about link color. People just keep going in circles.

Two numbers from that article have no declared source, so leave them out of any summary.

Pagination, infinite scroll and long lists

A long list needs a strategy, and the size of the collection picks it. The comparison of listing page options sorts the choices. Numbered pagination suits a large collection and offers precise control. Loading everything at once suits a small site. Infinite scroll works with few items and strong filters, and fails on a large collection, because the footer becomes unreachable and the reader loses their position in the list. A “load more” button reaches the footer and suits mobile, provided it states the total, the amount loaded and the amount left.

One rule carries across all of them. When someone opens an item and returns, put them back where they left the list. The same listing problem shows up in data-dense product pages, where the collection is the product.

Where the other side wins

The articles don’t always agree, and the five disagreements below are worth keeping in view.

Breadcrumbs on mobile: the 2007 text calls them problem-free and cheap. The 2018 revision says the cost outweighs the benefit on a phone, because the trail wraps onto several lines and the targets get small.
Full trail against short trail: 2007 wants the whole path for consistency. 2018 says to shorten it on mobile, and then admits the advice clashes with its own rule about showing the full path.
Underline: the 2004 text wanted color plus underline. The 2026 editor’s note revokes the visual rule and sends readers to WCAG 2.1 criterion 1.4.1.
Icon familiarity: the 2016 study measures a high cost for hiding the navigation. The 2025 study finds the icon widely recognized but says it tested recognizability alone and restates the older advice. That is a limit of scope, not a reversal, and it should be described that way.
Mechanism: the 2016 study explains the cost through perception and cognition. The menu and dropdown articles explain the same defect through movement, with Fitts’s law and the steering law. The accounts complement each other rather than compete.

A label also has to survive translation, which pushes the same discipline into multilingual content decisions.