In psychology, a learning curve is a graphical representation of how performance improves with practice or experience over time. It plots some measure of skill, accuracy, or speed on one axis against the number of practice trials or time spent learning on the other. The concept dates back to the 1880s, when Hermann Ebbinghaus conducted some of the earliest quantitative experiments on memory, and it remains a core tool for understanding how humans acquire skills and knowledge.
Origins of the Learning Curve
The learning curve concept traces to Ebbinghaus, who spent seven months testing only himself as a subject, often running up to three sessions per day. He wanted to find the lawful relationship between how well we retain information and how much time has passed since we learned it. To control for the effects of meaning and familiarity, he invented nonsense syllables as his test material. He had rejected tones (too cumbersome), digits (not enough variety for long experiments), and poetry (too variable in the associations it triggered).
Ebbinghaus would study rows of thirteen nonsense syllables until he could recall them correctly twice in succession, then measure how much effort it took to relearn them after various delays. His data revealed a characteristic pattern: forgetting was steepest right after learning and then gradually leveled off. He fit this pattern first to a power function in 1880, then to a logarithmic function in 1885, producing what became known as the Ebbinghaus Forgetting Curve. Though technically a forgetting curve rather than a learning curve, this work established the idea that changes in performance follow predictable mathematical shapes.
What the Curve Actually Looks Like
Most learning curves share a common shape: rapid improvement early on, followed by slower gains as you approach your maximum ability. This “hollow” or concave shape is the most frequently observed pattern in psychology experiments. You make big jumps in the first few practice sessions, then each additional hour of practice yields smaller and smaller returns.
Not all learning curves look the same, though. Researchers have documented at least three distinct shapes:
- Negatively accelerated (concave) curves: The most common type. Performance gains are largest at the start and diminish over time. This is the shape you see when someone picks up the basics of a skill quickly but takes much longer to refine it.
- S-shaped curves: Performance starts slowly, then accelerates through a middle phase, and finally levels off. These tend to appear when the learner is unfamiliar with the material or is a slower learner. The initial flat portion reflects a phase where the learner is building foundational knowledge before visible improvement kicks in. With prolonged learning, the S shape often disappears.
- Positively accelerated curves: Performance improves slowly at first, then speeds up. These are less common and typically appear in tasks where a certain threshold of understanding must be reached before progress can accelerate.
The “Steep Learning Curve” Misconception
In everyday speech, people say something has a “steep learning curve” to mean it’s difficult and will take a long time to master. In psychology, a steep learning curve means the opposite. If you plot performance (vertical axis) against time (horizontal axis), a steep curve rises quickly, meaning the learner is gaining skill at a fast rate. A slow, painful learning process would actually produce a shallow, drawn-out curve. The colloquial usage likely comes from the intuition that climbing a steep hill is harder, but on the graph itself, steepness represents speed of improvement.
Three Stages of Skill Acquisition
One of the most influential frameworks for understanding what happens at different points along a learning curve comes from Fitts and Posner, who proposed three sequential stages of motor skill acquisition. Learning curves across a wide range of tasks conform to this basic model.
In the cognitive stage, you figure out what you’re trying to do. You establish goals, determine the right sequence of actions, and rely heavily on conscious thought. Performance is inconsistent, and errors are frequent, but improvement is rapid because even small insights lead to big gains.
In the associative stage, you’ve worked out the basic approach and begin refining the details. Attention shifts to specific parts of the sequence, transitions between steps, and error correction. Improvement continues but at a slower pace than before.
In the autonomous stage, the skill becomes automatic. You can perform it with minimal conscious attention, the way an experienced driver navigates familiar roads. Performance gains in this phase are very small and come only with sustained practice. This is the long, flat tail of the learning curve where experts are making marginal refinements.
Learning Plateaus
A plateau is a flat stretch in the learning curve where performance seems to stall despite continued practice. Plateaus are a normal part of skill acquisition and can happen for several reasons. One explanation from information-processing psychology is that the brain needs to accumulate enough experience before it can cross a decision threshold. Until that threshold is crossed, the underlying learning doesn’t show up in measurable behavior, creating the appearance of stagnation.
Another common cause is that your current strategy has reached its limit. Early strategies get you through the initial phase, but hitting a plateau often signals that you need to restructure your approach, such as switching from a hunt-and-peck typing method to touch typing. The plateau reflects the transition cost of adopting a new, ultimately more effective strategy. Cognitive interference also plays a role: as tasks become more complex, the demands on working memory increase, and overloading memory with new information can temporarily slow down performance gains.
The Math Behind the Curve
Psychologists have long debated which mathematical function best describes learning. Two models have dominated the discussion. The power law of learning, formalized by Newell and Rosenbloom in 1981, states that performance improves as a power function of practice. In practical terms, this means early practice produces large gains that shrink proportionally as you continue. The relative rate of improvement slows steadily with each additional session.
The competing model is the exponential function, which predicts that the proportional progress toward your maximum performance stays constant with each trial. In other words, you always close the same percentage of the remaining gap, regardless of how far along you are.
For decades, the power law was treated as a near-universal principle. More recent work has challenged this, showing that the power law pattern can emerge as a statistical artifact when you average together the data of many individuals who each learn at different rates following exponential curves. When individual learning curves with varying rates are averaged, the resulting group curve naturally takes on a power function shape, even if no single person’s learning follows that pattern. Despite extensive research, psychologists have not been able to definitively rule in or out either model. Both continue to provide useful approximations depending on the context.
How Learning Curves Are Used in Practice
Outside the lab, learning curves have significant practical value. In organizational psychology and workforce planning, the principle that average productivity increases as a function of cumulative experience is well established. Research from MIT Sloan found that paramedics with lower experience improved their productivity by 1.7% to 2.8% after accumulating 500 patient calls. More notably, their performance consistency improved dramatically: the variability in how long tasks took was reduced by 3.7% to 8.7% over the same period.
This matters for planning. If you’re scheduling surgeries or staffing an emergency service, knowing the average learning curve lets you predict how long a newer worker will take. But the reduction in variability is equally important, because it means experienced workers are not just faster on average but more predictable. Ignoring this consistency effect can cause you to underestimate the benefits of experience by as much as 4%, leading to scheduling and staffing errors. Organizations use learning curve data to set realistic training timelines, determine when new employees will reach full productivity, and decide how much supervision different experience levels require.
What Shapes Your Personal Learning Curve
Several factors determine whether your learning curve will be steep (fast improvement) or shallow (slow going). Task complexity is one of the most obvious: simple, repetitive tasks produce steep curves, while skills with many interacting components tend to generate S-shaped curves with a slow initial phase. Prior knowledge matters too. Familiarity with related material effectively lowers the threshold that must be crossed before visible improvement begins, which is why the S-shaped pattern is most prominent in learners who are completely new to a domain.
The quality and timing of feedback accelerates or slows learning considerably. Immediate, specific feedback helps you correct errors before they become habits, steepening the curve during the cognitive and associative stages. The learner’s own strategies also matter. Research on motor learning shows that the cognitive stage is defined by active strategy selection, and choosing effective strategies early can compress the time spent in that initial phase. Finally, the spacing of practice sessions influences the curve’s shape: distributed practice (spread out over time) generally produces more durable learning than massed practice (cramming), even if the curve appears to rise more slowly in the short term.