The “transfer problem” identified by David Perkins and Gavriel Salomon represents one of the most significant challenges in education. Their research highlighted how knowledge and skills acquired in classroom settings often fail to transfer effectively to real-world contexts where they should be applied.

Key Findings from Perkins & Salomon’s Research

Types of Transfer

Near Transfer: Application of learning to contexts very similar to the original learning situation
- Generally occurs more readily
- Example: Learning algebra in class and then solving similar algebra problems on homework
Far Transfer: Application of learning to contexts substantially different from the original learning situation
- Much more difficult to achieve
- Example: Using statistical concepts learned in math class to evaluate real-world health claims

Mechanisms of Transfer

Perkins & Salomon identified two primary mechanisms for transfer:

Low Road Transfer: Automatic transfer based on practiced routines that become triggered in similar situations
- Based on extensive, varied practice to near-automaticity
- Depends on similarity of surface features between contexts
High Road Transfer: Mindful, deliberate abstraction of principles that can be applied in new contexts
- Requires explicit reflection, abstraction, and connection-making
- Depends on recognizing deeper structural similarities despite surface differences

Why Transfer Often Fails

Their studies identified several reasons why classroom learning frequently doesn’t transfer:

Inert Knowledge Problem: Knowledge remains “inert” because it’s learned in isolation from application contexts
Compartmentalization: Students mentally separate school learning from real-world problem-solving
Lack of Sufficient Practice: Skills aren’t practiced to the level needed for automatic application
Missing Bridging Contexts: Insufficient intermediary experiences between classroom and real-world
Overly Abstract Learning: Concepts taught without concrete examples and applications
Lack of Metacognitive Awareness: Students not taught to recognize when knowledge is applicable

Empirical Evidence

Several classic studies demonstrate the transfer problem:

Gick & Holyoak’s Analogical Reasoning Studies: Only 10-30% of subjects could apply a solution principle from one problem to an analogically similar problem without hints
Detterman’s Critical Thinking Research: Students failed to apply critical thinking skills across domains despite explicit instruction
Bransford & Schwartz’s “Preparation for Future Learning”: Traditional measures underestimate transfer effects; better transfer occurs when measured as enhanced ability to learn in new domains

Implications for Education

Perkins & Salomon suggested several approaches to improve transfer:

Hugging: Making learning situations more similar to application contexts (supporting near transfer)
- Using authentic problems, simulations, case studies
- Practicing in varied contexts that mirror real-world complexity
Bridging: Explicitly helping learners make connections between learning and application (supporting far transfer)
- Teaching general principles alongside specific examples
- Encouraging metacognitive reflection about how and when to apply knowledge
- Making abstractions explicit through compare/contrast activities

Their work suggests that transfer doesn’t happen automatically but must be deliberately designed into educational experiences. The persistence of the transfer problem helps explain why classroom education often fails to translate directly into career performance, despite students having technically “learned” the material.

Key Studies on the Transfer Problem

Gick & Holyoak’s Analogical Reasoning Studies

Mary Gick and Keith Holyoak conducted a series of influential studies in the early 1980s examining people’s ability to use analogical reasoning to solve problems.

The Classic Tumor Problem Experiment:

Initial Story: Participants read about a general who needed to capture a fortress. Multiple roads led to the fortress, but each road had mines that would detonate if a large force traveled on it. The solution was to divide the army into small groups that converged on the fortress simultaneously from different roads.
Target Problem: Participants then encountered a medical problem about how to use radiation to destroy a tumor without damaging surrounding tissue.
Key Finding: Only 10-30% of participants spontaneously recognized they could apply the “convergence solution” (using multiple weak radiation beams from different angles) from the military story.
With Hints: When explicitly told to consider the previous story as potentially helpful, 80-90% successfully applied the solution.

This demonstrated that people often fail to access relevant knowledge they possess when solving new problems, unless explicitly prompted to make the connection.

Detterman’s Critical Thinking Research

Douglas Detterman conducted studies in the 1990s challenging the notion that critical thinking is a general skill easily transferred across domains.

Methodology and Findings:

Students received explicit instruction in critical thinking skills, including evaluating evidence, detecting bias, and logical reasoning.
When tested on critical thinking tasks in the same domain as their instruction, students performed well.
However, when presented with structurally similar problems in unfamiliar contexts or domains, performance dropped dramatically.
Even when told the problems were analogous, many students struggled to apply the same thinking principles.

Detterman concluded that critical thinking skills are highly domain-specific and do not readily transfer without extensive bridging instruction and practice across multiple contexts.

Bransford & Schwartz’s “Preparation for Future Learning”

John Bransford and Daniel Schwartz proposed a reconceptualization of transfer in their influential 1999 paper.

Key Contributions:

They argued traditional measures of transfer were too narrow, focusing only on direct application of prior knowledge to solve new problems (“sequestered problem solving”).
Instead, they suggested measuring transfer as “preparation for future learning” (PFL)—how prior experiences enhance the ability to learn in new situations.

Their Alternative Approach:

Dynamic Transfer Assessment: Rather than testing if students could immediately solve new problems, they tested how quickly students could learn new material with minimal instruction.
Findings: While direct transfer was often limited (supporting Gick & Holyoak’s results), they found significant differences in how efficiently students could learn new but related concepts.
Example: Students who had previously studied biological systems didn’t automatically transfer specific solutions to engineering problems, but they learned new engineering concepts more quickly than peers without the biology background.

This perspective suggests that education might better prepare students not by teaching them solutions to specific problems they’ll encounter, but by developing their ability to learn efficiently when faced with new challenges.

These three research programs collectively highlight the complexity of knowledge transfer while suggesting that education could be redesigned to better facilitate the application of learning across contexts—an insight particularly relevant to the gap between classroom education and workplace performance.

Data and Key Findings on the Transfer Problem

Quantitative Evidence of the Transfer Problem

Laboratory Studies

Barnett & Ceci (2002) Meta-analysis: Aggregated data from 126 transfer studies showed an average effect size of only d = 0.28 for far transfer, compared to d = 0.63 for near transfer
Gick & Holyoak’s Convergence Studies: Across multiple experiments, unprompted transfer rates consistently ranged from 8% to 30% in analogical reasoning tasks
Australian Mathematics Study (Haskell, 2001): Only 18% of students could transfer mathematical procedures to slightly modified problems outside their original context

Classroom-to-Workplace Transfer

Levy & Murnane (2004): Analysis of labor statistics showed that while college graduates possessed theoretical knowledge, 68% reported requiring substantial on-the-job training to apply concepts to workplace tasks
Business School Research by Pfeffer & Fong (2002): Found near-zero correlation between MBA grades and subsequent career success metrics, suggesting limited transfer of classroom learning
Healthcare Education (Norman, 2009): Medical students who scored in the top quartile on exams only performed 5-10% better on standardized patient assessments than middle quartile peers

Domain-Specific Studies

Physics Education Research: Harvard FCI (Force Concept Inventory) studies by Mazur showed students who earned As in physics courses could often solve formula problems but failed to apply principles to real-world scenarios 50-65% of the time
Writing Instruction (Beaufort, 2007): Longitudinal studies tracking writing development found less than 30% of writing strategies transferred from composition courses to discipline-specific writing tasks
Programming Education (Robins et al., 2003): Only 22% of novice programmers could successfully apply learned algorithms to novel problems without substantial guidance

Factors Affecting Transfer Rates

Meta-analyses of Educational Interventions

Billing’s Review (2007): Explicit instruction in transfer strategies improved transfer rates by 30-45% compared to control groups
Hattie’s Synthesis (2009): Interventions focusing on metacognition showed effect sizes of d = 0.69 for improving transfer, while content-only approaches averaged d = 0.31

Temporal Patterns

Bahrick & Hall (1991): Longitudinal studies found only 15% retention of college mathematics after 5 years for students who didn’t continue using the material
Schmidt & Bjork (1992): Spaced practice with varied examples improved long-term transfer by 35-60% compared to massed practice of similar examples

Individual Differences

Bransford’s Expertise Studies: Novices transferred surface features (20-30% success rates), while experts transferred underlying principles (60-80% success rates)
Working Memory Correlations: Research by Engle shows working memory capacity correlates with transfer success (r = 0.42), suggesting cognitive load factors affect transfer ability

These data points collectively confirm the robustness of the transfer problem while highlighting conditions under which transfer is more likely to occur. The evidence consistently shows that transfer is not an automatic outcome of learning but requires specific instructional design and learner approaches to bridge the gap between knowledge acquisition and application.