This section turns to the repeated acts by which the studio favors schema formation.

One common strategy is to ask the student to actively recall previously encountered material. Even an unsuccessful attempt at recall, when followed soon afterward by the correct answer, can support later learning (Kornell, Hays, & Bjork, 2009). Cognitive scientists have a name for deliberate recall: retrieval practice, which has been associated with stronger long-term retention than passive rereading (Roediger & Karpicke, 2006). In the studio, retrieval works at every scale: a single axiom, a grammatical pattern, a chain of reasoning, or a long poem.

Another move is to present a fully worked-out model—a solved problem, a demonstrated proof, or a model passage—before the student attempts one independently. In language arts, there is the classical practice of imitation: a model passage is read carefully, copied by hand, paraphrased, and analyzed for structure until the student is able to compose a new piece in its form (Foster, 1989; Murphy, 1990). In math, there is the worked example: a problem solved in full on the page. Because the steps are visible, the student is spared the burden of searching for a solution path and can devote attention to the problem's structure: what principles are at work and what makes this type of problem recognizable (Sweller, van Merriënboer, & Paas, 1998; Chen et al., 2023). The completely solved example is slowly replaced with partially solved versions, with the solution faded step by step and the student filling in the missing portions on each try until the entire problem can be worked alone.

When students are asked to explain the reasoning behind each step of such a model, new material binds to schemas already held (Ausubel, 1968; Chi et al., 1989). When studying why slope is constant along a line, the student connects the algebraic formula to its geometric reason: between any two points on the line, the vertical distance is the "rise" and the horizontal distance is the "run." These distances vary depending on which two points you pick, but the right triangles they form (rise, run, and corresponding stretch of the line as hypotenuse) are all similar because the line crosses every horizontal at the same angle. Since similar triangles have proportional sides, the rise-to-run ratio is the same across them. The studio calls this explanatory process scoring: scratch by scratch, the needle opens the surface until new clay can find purchase.

Out in the world, problems do not always arrive with ready labels of the techniques that solve them. To prepare the student, the studio assigns scrambled practice sets once several schemas within a topic have formed. For example, a student working through factoring may meet four problems in mixed order—3x² − 6x, x² − 9, x² + 6x + 9, and x³ − 8—each requiring a different technique. Without a label announcing the method, the student must inspect each problem before summoning the appropriate schema, whether greatest common factor, difference of squares, perfect square trinomial, or difference of cubes. This is interleaved practice, the mixing of related techniques in a single session. Each new problem, in demanding fresh identification of the relevant schema, helps to build skills of discrimination not required for blocked practice (used when a new schema is first forming); interleaving is also more likely to produce more durable performance than blocked practice alone (Rohrer & Taylor, 2007).

A student who learns to factor quadratics and then returns to it the next day, a week later, and a few more times at widening intervals finds the schema easier to summon each time, until it surfaces without prompting. Whereas massed practice—such as drilling a concept for an hour in a single sitting—typically produces a fleeting fluency, spaced practice produces stronger long-term retention than the same total hours crammed into one sitting (Cepeda, Pashler, Vul, Wixted, & Rohrer, 2006).

Schema formation is typically slow, yet once in a while a single charged moment can do the work of weeks: a striking demonstration, a compelling paradox, or an intense exchange concentrates attention so sharply that the schema sets at once (Cahill & McGaugh, 1998; Greve et al., 2017).

Once a schema is firmly held, the teacher challenges the student to extend it to new cases whose underlying principle is preserved. For example, a student's first encounter with aⁿ is usually of the form n multiplications of a, where n is a counting number: a³ means a · a · a. But the idea of "n multiplications" breaks down when the exponent is zero, a negative integer, or a unit fraction (1 divided by a counting number). The schema for counting-number exponents is extended by naming a property the extension must preserve: for any exponents m and n, aᵐ · aⁿ = aᵐ⁺ⁿ. From this single requirement, with a ≠ 0 and m a counting number, the values of a⁰ and a⁻ᵐ follow as proven consequences once we divide both sides through by aᵐ: a⁰ = 1 because aᵐ · a⁰ = aᵐ⁺⁰ = aᵐ; and a⁻ᵐ = 1/aᵐ because aᵐ · a⁻ᵐ = aᵐ⁻ᵐ = a⁰ = 1. By the same principle, for a > 0 (so the root is real) and n a counting number, raising a to the power 1/n must yield the number whose nth power is a: the nth root of a. Each extension is a small, reasoned step, the schema growing outward by principle (Wu, 2011; Ellis et al., 2022).

While interleaved practice teaches a student to choose the correct schema for a given problem, complexity deepens when a single problem calls for the coordination of several schemas at once. A student fluent in adding numeric fractions and fluent in solving equations may still stall when an equation like 2/(x+1) + 3/x = 5 requires both schemas to act in concert: the fractions must first be combined, or the denominators cleared, before the equation becomes tractable. Neither schema working in isolation can reach the answer. Structured practice that asks several schemas to operate together (Schwartz, Bransford, & Sears, 2005) cultivates what Hatano and Inagaki (1986) called adaptive expertise: the capacity to coordinate multiple schemas at once in novel situations, beyond the routine expertise of executing one schema at a time.

Still more demanding than coordinating schemas within a single problem is integration-as-transfer: the carrying of insight across domains, generally reliant on (even if not assured by) a strong knowledge base in each domain (Haskell, 2001). A student fluent in mathematics, for instance, has not thereby acquired the capacity to read a Shakespearean sonnet with care. Cross-domain integration cannot be forced; it tends to surface when a student has gone deep in two areas and begins, alone or with a teacher's prompt, to see one in the reflected light of the other. The studio aims to build what Haskell calls a culture of transfer, steadily inviting the student to seek connection across domains.

A central practice at the studio is providing immediate, formative feedback—catching a wobble while the clay is still wet and responsive. When a student makes a fresh mistake, the teacher identifies the source of the error and offers a correction. For example, if a student writes (x+2)² = x² + 4, the error reveals an incomplete schema for binomial expansion: the distributive structure of squaring a sum has not quite taken hold. The teacher reviews the error, names any misconceptions that may underlie it, and shows how the correction follows from the structure of the expression. Precise feedback, given in the course of formation and repair, is among the most consequential pedagogical acts (Hattie & Timperley, 2007).