What 136 Million Keystrokes Reveal About Human Typing

In 2018, researchers at Aalto University in Finland published what remains the largest study of human typing behavior ever conducted. They recruited 168,000 volunteers through an online typing test and recorded over 136 million individual keystrokes. The resulting dataset—published at the CHI Conference on Human Factors in Computing Systems and awarded an Honorable Mention for Best Paper—upended decades of assumptions about how people type, what makes some typists fast, and why we all make the errors we do.

This article unpacks the key findings of Dhakal et al.’s landmark study and explores what it means for anyone interested in typing behavior, error patterns, and what realistic typing actually looks like.

The Study Design

The researchers built a web-based typing test that asked participants to transcribe English sentences displayed on screen. Each participant typed approximately 15 sentences, and the system recorded every keystroke, including timing, key identity, and whether the key was a correct character, an error, or a correction (backspace). The study captured not just what people typed, but how they typed—key hold durations, intervals between keypresses, rollover timing (pressing the next key before releasing the previous one), and the number of fingers used.

With 168,000 participants, the dataset dwarfs previous typing studies by orders of magnitude. Most earlier research relied on a few dozen participants in lab settings. This study captured the full spectrum of everyday typists—from two-finger hunt-and-peck beginners to experienced touch typists—in their natural environments, on their own keyboards.

How Fast Do People Type?

The headline number: the average typing speed was approximately 52 words per minute (WPM). But that average obscures enormous variation. The distribution ranged from under 20 WPM to over 120 WPM, with a long tail of exceptionally fast typists.

What determines typing speed? Not what most people think. The study found that formal typing training—the kind taught in schools, with prescribed finger assignments for each key—had a measurable but surprisingly modest effect. Many of the fastest typists had never taken a typing course. They had developed their own idiosyncratic methods through years of practice, and those methods worked as well as or better than the textbook approach.

Two factors predicted speed far better than training: the number of fingers used and the degree of hand alternation between consecutive keystrokes.

The Finger Count Discovery

Fast typists use more fingers. This sounds obvious, but the specifics are striking. The study found that fast typists used an average of 8.4 fingers, while slow typists used only 5.3. But the relationship is not linear—it is not simply “more fingers equals more speed.” The data revealed that what matters is not the total finger count but the consistency of finger-to-key assignments.

Fast typists tend to use the same finger for the same key every time. Slow typists are more variable—sometimes using the index finger for a key, sometimes the middle finger, sometimes reaching across hands. This consistency allows fast typists to build stronger motor programs—automated finger sequences that can be executed without conscious planning—which is the key to speed.

Interestingly, many fast typists deviate from the “correct” finger assignments taught in typing courses. They develop personalized mappings that work with their hand size, keyboard, and habits. The departure from textbook technique does not slow them down; what matters is that their personal assignments are consistent.

Eight Distinct Clusters of Typists

By analyzing the keystroke data with clustering algorithms, the researchers identified eight distinct groups of typists, each with characteristic speed ranges, finger usage patterns, and error profiles. These clusters ranged from slow two-finger typists to extremely fast touch typists, with several intermediate strategies in between.

The clusters showed that typing is not a simple spectrum from slow to fast. Different strategies produce different speed-accuracy tradeoffs. Some intermediate-speed typists achieve high accuracy with fewer fingers. Some fast typists sacrifice accuracy for speed, relying on correction (backspace) to fix errors rather than slowing down to avoid them. The relationship between speed and accuracy is mediated by strategy, not just skill.

Rollover Typing: Pressing Before Releasing

One of the study’s most surprising findings was how common rollover typing is. Rollover occurs when a typist presses the next key before fully releasing the previous one, creating a brief overlap in which two keys are simultaneously depressed. On traditional typewriters, this would cause a jam. On modern keyboards, it is a legitimate and common technique.

The researchers found that rollover was prevalent across all speed groups, not just fast typists. Even relatively slow typists exhibited rollover behavior on certain key combinations. The amount of rollover increased with speed, but its presence at all speed levels suggests it is a fundamental aspect of how humans interact with keyboards, not an advanced technique reserved for experts.

Rollover has direct implications for error patterns. When two keys are pressed in rapid succession with partial overlap, the likelihood of transposition errors increases—the intended sequence “ab” can become “ba” if the second key registers before the first is fully released. The study found that rollover-related transpositions account for a significant portion of all typing errors, particularly for fast typists.

Hand Alternation and Speed

The strongest predictor of typing speed, beyond finger count, was the degree to which typists alternated hands between consecutive keystrokes. When consecutive characters are typed by different hands, the typing is faster because one hand can pre-position for its key while the other hand is still completing its keystroke. The two hands work in parallel rather than in series.

The data showed that letter pairs typed by different hands were typed significantly faster than pairs typed by the same hand. This bimanual advantage was consistent across all speed groups and all typing strategies. It explains why certain words and character sequences are typed faster than others—the determining factor is not letter frequency or familiarity but the hand alternation pattern.

This finding also explains one of the most distinctive features of transposition errors: the majority of character transpositions occur across hands. When two characters are typed by different hands in rapid succession, the coordination between the two independent motor systems can break down, causing the second hand to fire slightly before the first. The result is a transposition error that is fundamentally bimanual—a coordination failure between two limbs executing independent movements in parallel.

What This Means for Error Patterns

The 136-million-keystroke dataset provides the most comprehensive picture ever assembled of how typing errors actually occur in the wild. Several patterns emerge:

• Speed amplifies motor noise. Faster typists make more errors per keystroke, but they also correct more aggressively, resulting in similar final error rates. The errors themselves differ: fast typists produce more transposition and rollover errors, while slow typists produce more substitution errors.
• Finger count changes error distribution. Typists who use more fingers distribute errors more evenly across the keyboard, while typists using fewer fingers concentrate errors on the keys that their limited finger set must reach for.
• Bimanual coordination dominates transpositions. When characters get swapped, it most often involves characters typed by different hands. This is not a random phenomenon but a systematic consequence of the brain coordinating two independent motor systems.
• Key hold duration signals confidence. Correct keystrokes have shorter hold durations than incorrect ones. The motor system appears to “hesitate” slightly when a finger lands on a key that might be wrong, even before the conscious mind recognizes the error.
• Error correction is strategic. Fast typists tend to detect and correct errors more quickly, often within one or two subsequent keystrokes. Slow typists sometimes type several additional characters before noticing and correcting an error. The correction strategy—not just the error rate—differs across typing profiles.

Beyond Speed: What the Data Reveals About Practice

One of the study’s most practical findings is about how typing skill develops. The researchers found that self-reported daily typing duration correlated with speed, but the relationship plateaued. Beyond a certain amount of daily practice (roughly 4-6 hours), additional typing time did not produce additional speed gains. This suggests that typing speed has a ceiling determined more by motor system constraints and strategy than by raw practice volume.

The data also showed that typing speed continues to develop well into adulthood. Unlike many motor skills that plateau in adolescence, typing speed increases through the 20s and 30s, likely because typing practice accumulates over years of professional and personal keyboard use. The oldest age groups in the study showed only modest speed declines, suggesting that typing is a remarkably durable skill once acquired.

How LikelyTypo Uses This Research

LikelyTypo translates these research findings into a practical tool. Its typing profiles model the different strategies identified in the study: the careful, accurate typist; the fast typist who relies on correction; the aggressive typist who pushes past the speed-accuracy tradeoff. Each profile produces a different distribution of error types, matching the patterns the research documented.

The device models in LikelyTypo reflect the physical constraints that the 136-million-keystroke study quantified. Keyboard errors follow the proximity-driven patterns of physical key pressing. Phone errors reflect the wider touch radius and absence of tactile feedback that mobile studies have documented. The result is generated errors that match what the largest typing dataset in history says real errors look like.

The next time you sit down to type, consider what the 136-million-keystroke study reveals about your own fingers. You are one of billions of typists, each with a unique strategy shaped by practice, habit, hand size, and keyboard. Your errors are not random failures. They are the predictable output of a motor system executing a complex bimanual coordination task at remarkable speed, under constraints that researchers are only beginning to fully understand.