Learning by doing is strongly supported for practical skills when an expert models the task, coaches the learner through it, and withdraws support step by step, but it is poorly supported when novices are left to discover concepts alone with minimal guidance. The research distinguishes guided practice from unguided discovery, and the five-step apprenticeship sequence of watching, assisting, performing under supervision, performing alone and teaching belongs in the first group.
What the research on skill learning supports about practice-based teaching, and where the popular claim goes too far. Every study named here is cited from recall.
Two Claims Hide Inside One Slogan
“Learning by doing” can mean (1) a learner acquires a craft by taking part in the work under an expert, or (2) a learner works out concepts and rules unaided by exploring. These get different support.
Claim 1, craft by participation. Anthropologists and psychologists describe children in many communities learning by observing and joining in real work, with adults adjusting what the child does as skill grows. Rogoff and colleagues (2003) call this “intent participation”. The educational version is the cognitive apprenticeship model of Collins, Brown and Newman (1989): the expert models the task and thinks aloud, coaches while the learner tries, provides support that is gradually faded, and asks the learner to explain and then try new variants. The parent article’s five-step sequence of watch, assist, perform supervised, perform independently and teach is a close match.
Claim 2, unguided discovery. Kirschner, Sweller and Clark (2006) argued that minimal-guidance instruction ignores the limits of working memory, which holds only about four chunks at once (Cowan, 2001). A beginner juggling a new task and a search for the rule can be overloaded. A meta-analysis by Alfieri and colleagues (2011, recalled) found unassisted discovery did worse than explicit instruction, while discovery with guidance, feedback or worked examples did better than both. This sits among arguments that are still debated; the safe reading is that guidance matters most for novices.
The Worked-Example Effect And Its Reversal
Sweller and Cooper (1985) found that beginners in algebra learned more from studying solved examples than from solving the same number of problems. The advantage reverses as skill grows (the “expertise reversal effect”, Kalyuga and colleagues, 2003): guidance that helps a novice slows an experienced learner. Practical rule: show complete examples early, then pose problems, then give open jobs.
Method: A Hand-Over Plan
Step 1: Demonstrate the whole task, then the parts. The learner sees the finished goal first.
Step 2: Name what you are doing and why. Spoken reasoning exposes steps the expert no longer notices.
Step 3: Let the learner do one sub-step with you. The expert keeps the hardest part.
Step 4: Give feedback while the action is fresh. For motor skills, correction soon after the try is more useful than a later summary.
Step 5: Fade. Reduce help each repetition. Introduce variations (a different tool, a different material) so the skill is not tied to one setting.
Step 6: Have the learner explain, then teach. Explaining shows whether the reasons were understood.
Practice Alone Does Not Make Expertise
The popular “ten thousand hours” claim derives from work on deliberate practice by Ericsson and colleagues (1993). A later meta-analysis (Macnamara and colleagues, 2014, recalled) found that deliberate practice accounted for a fraction of performance differences, varying by field. Hours matter; their quality, and the learner’s starting point, also matter. Do not tell a child that skill is simply a matter of time.
Limits Of This Evidence
Most controlled studies use school subjects or laboratory tasks, with few on traditional crafts. The reasoning for crafts rests on the same mechanism (limited working memory, benefit of feedback) plus ethnographic description. Percentage tables claiming “we remember 10% of what we read and 90% of what we do” have no experimental source: the Dale “cone of experience” of 1946 had no such numbers. A page elsewhere in this guide (Hands-On Learning) repeats such percentages; disregard them. The page’s practical advice to teach by doing is consistent with the above; its numbers are not.
Common Mistakes
- Throwing novices in “to learn from mistakes”. Where mistakes are costly or dangerous, the first tries must be protected.
- Never fading help. The learner stays dependent.
- Skipping the reasons. A child who can repeat a motion but cannot say why will fail when conditions change.
Practising on hazardous tasks
Axes, knives, fire, boiling liquid, heavy loads and moving machinery are learned in stages, starting with a safe model or a low-hazard version, and the expert stays within arm’s reach until the learner has done the task correctly many times in different conditions. Hazard limits come before the hand-over plan; see Task Assignment by Age.
What’s Next
- Learn next: Apprenticeship System Design & Knowledge Transfer
- Combine with: Skill Demonstrations and Building Competence & Pride