1. Bibliographic Information

1.1. Title

Cognitive Conceptions of Learning

1.2. Authors

Thomas J. Shuell, State University of New York at Buffalo

1.3. Journal/Conference

This article was published in the Review of Educational Research, a highly respected academic journal in the field of education research, known for publishing comprehensive reviews of research literature. Its focus on educational research implies that the paper's findings and discussions are particularly relevant to pedagogical practices and the understanding of how learning occurs in educational settings.

1.4. Publication Year

1986

1.5. Abstract

The paper explores the recent shift in psychological and educational thinking towards cognitive psychology and its impact on understanding learning. It notes that while cognitive psychology is now mainstream, a focused concern on learning itself (factors influencing changes in human performance, knowledge structures, and conceptions) has only recently emerged within this framework. The article aims to examine current cognitive thinking on learning and how it can guide future research in both learning and instruction. It discusses similarities and differences between behavioral and cognitive conceptions of learning, addressing key issues such as the active nature of learning, the emphasis on understanding (comprehension), the crucial role of prior knowledge, the cumulative nature of human learning, and the importance of cognitive analyses of performance. Several prominent cognitive theories of learning are presented to exemplify cognitive psychology's influence on learning research.

1.6. Original Source Link

/files/papers/693fd513a078743fa50a04df/paper.pdf (Published)

2. Executive Summary

2.1. Background & Motivation

The core problem the paper addresses is the diminished focus on learning itself within cognitive psychology during the 1960s-1980s, despite cognitive psychology becoming the mainstream in psychological and educational thought. Historically, learning was central to behavioral psychology, but cognitive psychologists shifted their attention to memory systems and information processing rather than the mechanisms of knowledge acquisition.

This problem is important because learning is fundamental to many human endeavors, including teaching, child-rearing, and training. A comprehensive understanding of learning is crucial for developing effective educational and instructional practices. Without a clear cognitive conception of learning, research in these areas lacked direction and theoretical grounding. The paper highlights a gap where cognitive psychology acknowledged learning's importance but failed to develop a robust, data-backed cognitive theory of learning.

The paper's entry point is the observation that a new era of research on learning within cognitive psychology began around 1975, often utilizing information-processing perspectives and computer models. This resurgence provides an opportunity to articulate a cognitive conception of learning that enriches understanding beyond traditional behavioral views.

2.2. Main Contributions / Findings

The paper makes several primary contributions by:

• Defining Cognitive Learning: It articulates a distinct cognitive conception of learning that emphasizes knowledge acquisition and knowledge structures over mere behavioral changes, stressing understanding and an active, constructive role for the learner.

• Contrasting with Behaviorism: It systematically outlines the similarities and, more importantly, the differences between behavioral and cognitive approaches to learning, particularly regarding what is learned (behavior vs. structured knowledge) and the factors influencing it (reinforcement vs. learner strategies and prior knowledge).

• Highlighting Key Cognitive Influences: It identifies five significant ways cognitive psychology has influenced learning research:

  1. Learning as an active, constructive process.
  2. The presence of higher-level processes (e.g., metacognition) in learning.
  3. The cumulative nature of learning and the critical role of prior knowledge.
  4. Concern for knowledge representation and organization in memory.
  5. The use of cognitive process analysis for performance and task understanding.

• Presenting Cognitive Theories of Learning: It describes several influential cognitive theories of learning as examples, including:

  • Bruner's discovery learning.
  • Ausubel's subsumption theory (meaningful verbal learning).
  • Wittrock's generative learning.
  • Bransford and Franks' decontextualization.
  • Rumelhart and Norman's accretion, tuning, and restructuring modes.
  • John Anderson's ACT theory (declarative to procedural knowledge transition, generalization, discrimination, strengthening).

• Outlining Future Research and Educational Implications: It discusses critical areas for future research, such as identifying specific variables affecting cognitive learning, the interplay between knowledge and learning (including domain-specific vs. domain-independent aspects), and phases of learning. It also highlights practical implications for education, advocating for active learning and teachers understanding students' prior knowledge and cognitive processes.

  These findings collectively address the gap by providing a coherent framework for understanding learning from a cognitive perspective, offering a foundation for future research and improved instructional practices.

3. Prerequisite Knowledge & Related Work

3.1. Foundational Concepts

To fully understand the paper, a reader should be familiar with the following foundational concepts:

• Behaviorism (Behavioral Psychology): A school of thought in psychology that emerged in the early 20th century, primarily concerned with observable behavior and how it is learned through stimulus-response associations, reinforcement, and punishment. It largely disregarded internal mental states. Key figures include B.F. Skinner (operant conditioning) and Ivan Pavlov (classical conditioning). In this paper, behaviorism serves as a contrast to cognitive psychology.
• Cognitive Psychology: A branch of psychology that emerged as a reaction to behaviorism, focusing on internal mental processes such such as perception, memory, thinking, problem-solving, and language. It views the mind as an information processor.
• Information Processing: A cognitive paradigm that conceptualizes the human mind as a system that processes information, similar to a computer. It involves encoding, storing, retrieving, and manipulating information.
• Learning: The process by which an individual acquires new knowledge, skills, behaviors, or values, or modifies existing ones. The paper discusses both behavioral and cognitive definitions. The behavioral definition focuses on observable, enduring changes in behavior due to experience. The cognitive definition expands this to include changes in knowledge structures and conceptions.
• Knowledge Structures: Organized representations of information in an individual's memory. These are not just isolated facts but interconnected networks of concepts, principles, and procedures. Examples include schemas and production systems.
• Schema (plural: Schemata): A mental framework or organized pattern of thought or behavior that helps individuals to interpret and process information. Schemas represent generalized knowledge about concepts, events, or situations. For example, a "restaurant schema" might include knowledge about ordering food, eating, and paying the bill.
• Metacognition: "Cognition about cognition," or thinking about one's own thinking. It involves awareness, understanding, and control of one's cognitive processes, such as planning, monitoring, and evaluating one's learning strategies.
• Declarative Knowledge: "Knowing that." Factual knowledge and information that can be stated or declared. For example, knowing that "Paris is the capital of France."
• Procedural Knowledge: "Knowing how." Knowledge of how to perform specific tasks or actions. For example, knowing how to ride a bicycle or solve a math problem. This knowledge is often difficult to articulate verbally.
• Active Learning: A learning approach where learners are directly involved in the learning process rather than passively receiving information. This includes activities like problem-solving, discussions, and self-directed inquiry.
• Constructive Process (in learning): The idea that learners actively construct their own understanding and knowledge of the world through experiencing things and reflecting on those experiences. They build new knowledge upon the foundation of prior learning.
• Generative Learning: A theory that proposes that learners actively generate connections between new information and their existing knowledge to create meaning and enhance comprehension.

3.2. Previous Works

The paper references several key prior studies and theories crucial for understanding the evolution of learning concepts:

• Ebbinghaus (1913, originally 1885) - Research on Memory and Forgetting: Hermann Ebbinghaus conducted pioneering experimental studies on memory using nonsense syllables to study rote learning and the forgetting curve. His work established a tradition of empirical research on learning, primarily within a behavioral framework, focusing on simple forms of memorization.
• Gestalt Psychologists (1910s-1930s): A school of psychology that emphasized that the whole of experience is greater than the sum of its parts. They were more interested in perception and how elements are organized into meaningful wholes. While they discussed learning, they often interpreted it through perceptual principles of organization. They are seen as forerunners of cognitive psychology due to their focus on internal mental structures.
• Bartlett (1932) - "Remembering: A Study in Experimental and Social Psychology": Frederick Bartlett's work highlighted the constructive nature of memory, arguing that people reconstruct memories based on their existing schemas and cultural context, rather than simply retrieving exact copies. This challenged the behavioral view of memory as a passive storage unit.
• Tolman (1932) - Purposive Behavior in Animals and Men: Edward Tolman introduced the concept of latent learning and cognitive maps, suggesting that animals (and humans) learn more than just stimulus-response connections; they develop internal representations of their environment, even without immediate reinforcement. This was a significant step away from strict behaviorism.
• Ausubel (1962, 1963) - Subsumption Theory of Meaningful Verbal Learning: David Ausubel's theory focused on how learners integrate new meaningful verbal information into their existing cognitive structures. He proposed concepts like advance organizers (introductory materials that provide a framework for new information) and subsumption (new information being absorbed under broader, more inclusive concepts).
• Bruner (1957, 1961) - Discovery Learning: Jerome Bruner advocated for discovery learning, where learners actively explore and discover relationships and concepts for themselves, rather than passively receiving information. He emphasized going beyond the information given and the importance of generic coding for transferability of learning.
• Paivio (1969, 1971) - Dual-Coding Theory: Allan Paivio's theory proposed that information is processed and stored in memory in two distinct, interconnected systems: a verbal system (for linguistic information) and an imaginal system (for non-linguistic, visual information). This highlighted the role of imagery in learning and memory.
• Rothkopf (1965, 1970) - Mathemagenic Behaviors: Ernst Z. Rothkopf introduced the concept of mathemagenic behaviors (literally "behaviors that give birth to learning"). These are activities engaged in by learners (e.g., questioning, reviewing, underlining) that facilitate the acquisition of knowledge from instructional materials.
• Wittrock (1974, 1978) - Generative Learning Model: Merlin C. Wittrock's model proposed that learning is a generative process where learners actively construct meaning by relating new information to their existing knowledge and experience, often by generating verbal and imaginal elaborations.
• Miller, Galanter, & Pribram (1960) - "Plans and the Structure of Behavior": This influential book introduced the concept of TOTE (Test-Operate-Test-Exit) units, suggesting that behavior is guided by hierarchical plans and internal feedback loops, moving away from simple stimulus-response chains towards more complex, goal-oriented mental activities.
• Rumelhart and Norman (1978, 1981) - Accretion, Tuning, and Restructuring: Their work proposed three distinct modes of cognitive learning within a schema-based theory of long-term memory. This theory is detailed later in the paper.
• Anderson, J.R. (1982, 1983) - ACT (Adaptive Control of Thought) Theory: John R. Anderson's comprehensive cognitive architecture and learning theory accounts for the acquisition of both declarative and procedural knowledge, and the transition between them, through mechanisms like knowledge compilation, generalization, discrimination, and strengthening. This theory is also detailed later in the paper.
• Bransford & Johnson (1972) and Dooling & Lachman (1971): These studies empirically demonstrated the crucial role of prior knowledge and contextual information in comprehension and retention of prose, showing that meaningfulness is not inherent in the material but constructed by the learner based on what they already know.

3.3. Technological Evolution

The field of learning research has undergone a significant evolution:

1. Early 20th Century - Behaviorism Dominance: Learning research was dominated by behaviorism, focusing on observable stimulus-response connections, largely through animal studies (e.g., Ebbinghaus, Pavlov, Thorndike, Skinner). The goal was to discover universal laws of learning applicable to all organisms, emphasizing reinforcement and contiguity. Learning was primarily seen as the acquisition of behaviors or associations.

2. Mid-20th Century - Transition to Early Cognitive Views: Beginning in the 1930s (Bartlett, Tolman) and accelerating in the 1950s-1960s (Bruner, Ausubel, Miller, Galanter, Pribram), cracks appeared in the behavioral paradigm. Researchers started questioning if simple S-R models could explain complex human learning. There was a growing realization that learners were not passive but actively organized information and that internal mental processes mattered. Verbal learning research, in particular, began incorporating cognitive interpretations.

3. Late 1960s to 1970s - Rise of Cognitive Psychology, Decline in Learning Focus: Cognitive psychology became the mainstream, driven by the information-processing metaphor. However, during this period, the focus shifted largely to understanding memory systems, knowledge representation, and human information processing mechanisms rather than learning as a process of change. The emphasis was on how knowledge is stored rather than how it is acquired. This led to a "demise of interest in learning per se," as noted by the paper.

4. Post-1975 - Resurgence of Cognitive Learning: A renewed interest in learning within cognitive psychology emerged, often incorporating information-processing models and computer simulations (e.g., AI research in learning). This new wave specifically sought to define and investigate cognitive conceptions of learning, emphasizing understanding, knowledge structures, and active construction.

   This paper's work fits into this timeline at the post-1975 resurgence stage. It reviews and synthesizes this new wave of cognitive learning research, articulating its distinct features and implications, setting the stage for future directions.

3.4. Differentiation Analysis

Compared to behavioral conceptions of learning, the cognitive conception presented in this paper introduces several core differences and innovations:

• Nature of the Learner:
  • Behavioral: Learner is passive, responding to environmental stimuli and reinforcement.
  • Cognitive: Learner is active, constructive, and goal-oriented, actively selecting, organizing, and interpreting information.
• What is Learned:
  • Behavioral: Behavior itself, or associations (bonds) between stimuli and responses. Internal mental states are often deemed irrelevant.
  • Cognitive: Knowledge and knowledge structures, with an emphasis on meaning and understanding. Behavior is seen as a result of learning, not what is learned. Knowledge is represented as complex, organized structures (e.g., schemata, production systems), not just simple associations.
• Factors Influencing Learning:
  • Behavioral: Primarily external environmental factors like reinforcement, punishment, contiguity, and practice. Focus is on changing the environment.
  • Cognitive: Internal factors like prior knowledge, metacognitive strategies (planning, monitoring), goal setting, attention, and the way new information is encoded and integrated. Focus is on changing the learner or their internal processes.
• Emphasis:
  • Behavioral: Learning how to perform a task. Rote learning or memorization of simple elements.
  • Cognitive: Understanding (comprehension) of complex relationships. Meaningful learning that builds on existing knowledge.
• Complexity of Learning:
  • Behavioral: Primarily studied simple forms of learning (e.g., conditioning, memorization of nonsense syllables).
  • Cognitive: Focused on complex forms of learning encountered in real-life (e.g., problem-solving, concept acquisition, comprehension of text), where understanding is paramount.
• Role of Prior Knowledge:
  • Behavioral: Limited concern for prior knowledge, typically viewed in terms of transfer or proactive inhibition based on stimulus-response similarity.
  • Cognitive: Prior knowledge is central, serving as the foundation upon which new knowledge is constructed. It influences how new information is interpreted and acquired. The concept of schemas highlights this.
• Analytical Approach:

  • Behavioral: Focus on atheoretical, functional relationships between stimulus and response.

  • Cognitive: Cognitive process analysis seeks to identify the mental activities and knowledge structures that mediate the stimulus-response relationship, explaining how learning occurs internally.

    In essence, the cognitive conception of learning moves beyond observable actions to delve into the mind's internal workings, providing a richer, more nuanced explanation of how humans acquire, organize, and use knowledge.

4. Methodology

The paper is a theoretical and review article, not an empirical study. Therefore, its "methodology" involves a systematic examination and synthesis of existing literature to construct a cognitive conception of learning. The author's approach is to:

1. Define Learning: Establish common ground on the definition of learning while highlighting key differences in emphasis between behavioral and cognitive views.
2. Trace the Transition: Describe the historical shift from behavioral to cognitive orientations and the reasons for the temporary decline and subsequent resurgence of interest in learning within cognitive psychology.
3. Identify Core Influences: Articulate the key ways cognitive psychology has reshaped the understanding of learning.
4. Present Illustrative Theories: Detail several prominent cognitive theories of learning to exemplify the principles discussed.
5. Discuss Implications: Outline directions for future research and implications for educational practice.

4.1. Principles

The core idea behind the cognitive conception of learning is that learning is an active, constructive, cumulative, and goal-oriented mental process driven by the learner's internal activities and prior knowledge, aiming for understanding and knowledge acquisition rather than mere behavioral modification. The theoretical basis is rooted in the information-processing paradigm, viewing the mind as a system that processes, organizes, and stores information.

Key principles include:

• Activity of the Learner: Learners are not passive recipients but actively engage with information. This involves selecting stimuli, organizing material, generating responses, and using learning strategies.
• Understanding/Comprehension: The goal of cognitive learning is to make sense of information, to extract meaning, and to build coherent mental representations, rather than just memorizing facts or performing tasks.
• Role of Prior Knowledge: New learning is heavily dependent on and integrated with what the learner already knows. Prior knowledge acts as a framework (schema) that influences how new information is interpreted and acquired.
• Cumulative Nature: Learning is a progressive process where new knowledge builds upon existing knowledge structures, leading to more complex and organized representations.
• Higher-Level Processes (Metacognition): Learning involves metacognitive or executive processes that regulate and monitor one's own learning, such as planning, goal-setting, and self-assessment.
• Knowledge Representation: Cognitive learning emphasizes how knowledge is organized and stored in memory, typically as complex knowledge structures (e.g., networks of propositions, schemas, production rules).
• Cognitive Process Analysis: Understanding learning involves analyzing the specific mental operations or processes (e.g., encoding, inferring, mapping, applying) that individuals use when performing cognitive tasks.

4.2. Core Methodology In-depth (Layer by Layer)

The "methodology" section in this review paper describes the fundamental tenets and influential theories that constitute the cognitive conception of learning.

4.2.1. Learning as an Active Process

Cognitive approaches fundamentally view learning as an active, constructive, and goal-oriented process. This contrasts sharply with behavioral views, where the learner is often seen as passively responding to environmental cues. The paper highlights several ways learners are active:

• Metacognitive Processes: Learners engage in metacognitive processes like planning, setting goals and subgoals, predicting outcomes, and monitoring their own learning. For example, a student planning how to study for an exam is engaging in metacognition.

• Active Selection of Stimuli: Learners selectively attend to certain aspects of their environment, distinguishing between the functional stimulus (what the learner actually responds to) and the nominal stimulus (what the experimenter intended).

• Organization of Material: Learners actively try to organize new material into meaningful patterns, even when explicit organizational cues are absent. For instance, in a free-recall task, participants often group words by category even if presented randomly.

• Generation of Responses: Learners construct or generate appropriate responses rather than merely emitting pre-learned ones. This includes generating elaborations or connections between new and old information.

• Use of Learning Strategies: Learners employ various conscious and unconscious strategies (e.g., rehearsal, elaboration, mnemonics) to facilitate encoding, storage, and retrieval of information.

  The paper acknowledges a "learning paradox" related to constructivism: how can a learner acquire a new, more complex cognitive structure without already possessing a more advanced structure to build it from? While Bereiter (1985) suggests solutions, the paper notes a lack of empirical support for them.

4.2.2. Higher-Level Processes in Learning

Cognitive conceptions recognize metacognition as central to learning. Metacognition refers to executive processes that regulate learning and cognition. It involves two main types of activities:

• Regulation and Orchestration: This includes planning learning activities, predicting information, guessing, and monitoring the learning process. For example, a student might plan to read a chapter, predict what key ideas will be covered, and then monitor their comprehension as they read.
• Metacognitive Knowledge: This is knowledge about one's own cognitive processes and knowledge. Flavell and Wellman (1977) suggest four classes:

  • Tasks: Knowledge about how task characteristics influence performance (e.g., "this essay requires more detailed planning").

  • Self: Knowledge about one's own strengths and weaknesses as a learner (e.g., "I'm good at remembering faces, but bad at names").

  • Strategies: Knowledge about the effectiveness of different learning strategies (e.g., "re-reading is less effective than active recall").

  • Interactions: Knowledge of how these three types of knowledge interact to influence cognitive performance.

    An example of the hierarchical nature of learning is Sternberg's (1984a, 1984b) componential theory of knowledge acquisition. It proposes metacomponents (executive processes) that regulate performance components (processes for task execution like encoding and comparison) and knowledge-acquisition components. The three knowledge-acquisition components are:

1. Selective encoding: Identifying relevant information and sifting out irrelevant details from the environment.

2. Selective combination: Integrating selected information in a meaningful way.

3. Selective comparison: Relating newly encoded or combined information to old information already stored.

   These components are influenced by moderating variables such as the number of occurrences, variability of contexts, location of cues, importance of information, and density of information.

4.2.3. The Role of Prior Knowledge

Cognitive conceptions emphasize the cumulative nature of learning, where new information gains meaning through its connection to prior knowledge. This is a significant departure from traditional verbal learning research, which focused on simple associations.

• Schema Theory (R.C. Anderson, 1984): This theory highlights that organized, structured, and abstract bodies of information (schemata) that a learner possesses heavily influence how new material is interpreted, understood, and acquired. Bransford and Johnson (1972) and Dooling and Lachman (1971) demonstrated that activating relevant prior knowledge significantly impacts comprehension and retention.
• Transfer: Unlike the behavioral view of transfer based on stimulus-response similarity, cognitive learning views transfer as a more complex process involving the transformation of knowledge to establish "boundary constraints" for identifying "sameness" and "uniqueness" in novel information (Bransford & Franks, 1976). Learning involves changing the form of one's knowledge to enable new discoveries.
• Domain-Specific Knowledge: More recent research emphasizes domain-specific knowledge (e.g., Chi, Glaser, & Rees, 1982), showing that experts and novices solve problems differently due to their specialized knowledge structures. While there's debate on its relative importance compared to domain-independent strategies, both are recognized as crucial. This links directly to the prior knowledge emphasis.

4.2.4. The Question of What is Learned

This is a fundamental distinction between behavioral and cognitive views.

• Behavioral: Learner acquires associations or bonds between stimuli and responses, or the internal mechanisms are irrelevant (Skinner).
• Cognitive: Learner acquires meaning and knowledge. The emphasis is on understanding, not just performance. Behavior is the result of learning, not learning itself. Knowledge is typically represented by complex knowledge structures (e.g., networks of information specifying relationships among facts and actions) rather than simple associations.
  • Other representations: Scandura (1970, 1977) and Siegler (1983) suggest rules as units for knowledge.
  • Multiple knowledge representations: Declarative knowledge ("knowing that") and procedural knowledge ("knowing how") are distinguished, with potentially other forms (Gagné & White, 1978) and multiple memory systems (Tulving, 1985).

4.2.5. Cognitive Process Analysis

Cognitive psychology has introduced the idea of analyzing performance and cognitive abilities by breaking them down into the underlying cognitive processes. This has been applied to various tasks, including:

• Tests of mental ability (intelligence, inductive/deductive reasoning).

• Instructional tasks (learning geometry, physics, reading, arithmetic).

  For example, Sternberg (1977) proposed that analogical reasoning involves six cognitive processes:

1. Encoding: Understanding the terms of the analogy.
2. Inferring: Identifying the relationship between the first two terms.
3. Mapping: Discovering a higher-order rule relating the first and third terms.
4. Applying: Generating a fourth term based on the inferred relationship and mapping.
5. Justification: (Optional) Selecting the best answer if choices are provided.
6. Response: Translating the solution into an overt response. Such analyses help understand both learning processes and effective instructional techniques.

4.2.6. Cognitive Theories of Learning (Examples)

The paper then details several specific cognitive theories of learning:

Early Conceptions

• Bruner's Discovery Learning (1957, 1961): Learning occurs by coding something generically to maximize transferability. Conditions for this include a set to learn, appropriate motivation, prior mastery, and diversity of training.
• Ausubel's Subsumption Theory (1962, 1963): Focuses on meaningful reception learning. New, potentially logical information is subsumed (incorporated) into the learner's existing, hierarchically organized cognitive structure. Key factors: availability of existing cognitive structure, use of advance organizers, and discriminability of new material from existing structure. Retention is influenced by repetition, length of time subsuming concepts have been present, exemplars, and multi-contextual exposure.
• Wittrock's Generative Learning (1974, 1978): Learners construct meaning by generating verbal and imaginal elaborations that relate new information to prior knowledge. Learning involves making inferences, applying them, testing them, and seeking feedback.
• Bransford and Franks' Decontextualization (1976): Understanding (comprehension) involves decontextualization, where knowledge acquired in a specific context becomes more abstract and applicable to various situations. This occurs through encountering relevant examples, which helps clarify concepts.

Rumelhart and Norman (1978) - Three Modes of Learning

This theory accounts for learning within a schema-based theory of long-term memory, proposing that learning is not unitary but occurs in three qualitatively different ways:

1. Accretion: The most common form, involving the encoding of new information into existing schemata. This occurs when new material is consistent with existing schemata and is added without changing the overall organization of knowledge. It's akin to memorization or schema instantiation (Resnick, 1984, similar to Piaget's assimilation). Accretion benefits from study, mnemonic aids, and deep processing. It can be tested by recall and recognition. It tends to have high interference from related topics and low transfer.
2. Restructuring (Schema Creation): The process of creating new schemata or reorganizing existing knowledge. This can occur even without new information, simply by reorganizing what's already known. Similar to Piaget's accommodation (Resnick, 1984). Two ways restructuring occurs:
   • Schema induction: Learning by contiguity, where co-occurrence of schemata leads to new schema formation.
   • Patterned generation: Creating a new schema by copying and modifying an old one (analogical processes). Restructuring is fostered by examples, analogies, metaphors, and Socratic dialogue. It's tested by conceptual tests and problem-solving questions.
3. Tuning (Schema Evolution): The slow, gradual refinement of existing schemata through repeated use in different situations. This process lasts a lifetime. Tuning is best achieved through practice. It's measured by speed and smoothness of performance, especially under stress. Tuning leads to low interference and high transfer for general knowledge, but low transfer for specific (tuned) knowledge.

John Anderson's ACT (Adaptive Control of Thought)

ACT (or $ACT*$ in its current version) is a comprehensive computer program (theory) that models the acquisition of procedural knowledge, such as solving geometry proofs. It posits a single set of learning processes for all skill acquisition.

• Declarative vs. Procedural Knowledge:

  • Declarative knowledge is represented as a network of propositions (statements of relationships among concepts/events).
  • Procedural knowledge is represented as a system of productions (IF-THEN rules specifying conditions for actions and the actions themselves).

• Knowledge Acquisition Flow: Knowledge in a new domain always starts as declarative knowledge. Procedural knowledge is learned by making inferences from this declarative knowledge.

• Three Stages of Learning Procedural Knowledge:

  1. Declarative Stage: New information is encoded probabilistically into a network of existing propositions as declarative knowledge. This knowledge has little direct control over behavior but is interpreted by general problem-solving procedures.
  2. Knowledge Compilation Stage: Declarative knowledge is transformed into higher-order procedures (productions). This increases efficiency. This is where productions are formed.
  3. Procedural Stage (Tuning): The productions are refined through an adaptive production system. This involves three learning mechanisms:

• Learning Mechanisms in ACT:

  1. Generalization: Production rules become broader in their applicability. The ACT system searches for similarities between production rules and creates a new, more general production rule that captures common features. This is an inductive process.

     • Example: P1. IF the goal is to do an addition problem, THEN add the numbers in the rightmost column. P2. IF the goal is to do an addition problem and the rightmost column has already been added, THEN add the numbers in the second column. (Simplified examples from the text, representing declarative knowledge of steps.) These might compile into: P3. IF the goal is to do an addition problem, THEN the subgoal is to iterate through the columns of the problem. P4. IF the goal is to iterate through the columns of an addition problem and the rightmost column has not been processed, THEN the subgoal is to iterate through the rows of the rightmost column and set the running total to zero. (More complex, compiled productions) Later, if the student learns subtraction and acquires a similar production: P5. IF the goal is to do a subtraction problem, THEN the subgoal is to iterate through the columns of the problem. The similarity between P3 and P5 leads to generalization: P6. IF the goal is to do an LV problem, THEN the subgoal is to iterate through the columns of the problem. Here, LV is a "local variable" representing specific instances. The original productions (P3, P5) still apply in specific contexts. Transfer is facilitated if common components are taught across different procedures.

  2. Discrimination: Production rules become narrower in their applicability, preventing incorrect applications. This requires the learner to experience both correct and incorrect applications, similar to needing positive and negative exemplars in concept learning.

     • Example: P7. IF the goal is to iterate through the rows of a column and the top row has not been processed, THEN the subgoal is to add the digit of the top row into the running total. If this is overgeneralized to subtraction problems, discrimination is needed. The learner must learn to distinguish when P7 is appropriate (addition) versus when a subtraction production is appropriate.
     • Two types of discrimination:
       • Action discrimination: Learning a new action based on feedback.
       • Condition discrimination: Restricting the conditions under which an old action is performed.

  3. Strengthening: Successful productions are reinforced, while unsuccessful ones are weakened. This mechanism modifies the probability associated with a production based on positive and negative feedback.

     Generalization and discrimination are seen as the inductive components of ACT.

4.3. Implications for Future Research

The paper outlines several challenges for future research in cognitive learning:

• Variables Affecting Learning: Little is known about the specific variables (e.g., environmental events, instructional designs) that precisely influence cognitive learning. Future research needs to develop operational definitions for these variables (e.g., encoding, combination of information, relating new to old information, evaluation) so they can be systematically investigated. Sternberg (1984a) identified variables like number of occurrences, variability of contexts, location of cues (contiguity), importance of item, and density of items that affect verbal concept acquisition.
• Knowledge and Learning:
  • Investigate how the transition from novice to expert occurs, especially regarding qualitative changes in knowledge structures.
  • Explore the interrelationship between domain-specific and domain-independent learning processes.
  • Clarify the relationship between different types of knowledge (e.g., declarative and procedural). The idea that all knowledge might be procedural ("knowledge how") that can sometimes be interrogated to produce declarative knowledge ("knowledge that") is a key area.
• Phases of Learning: Empirical evidence is scarce for phases in complex, meaningful learning. Research should identify the stages learners go through and how different variables (e.g., mnemonics, organizational strategies) might be more facilitative during specific phases or stages.

5. Experimental Setup

This paper is a review and theoretical exposition of cognitive conceptions of learning, not an empirical research paper. As such, it does not describe a specific experimental setup, novel datasets, or evaluation metrics employed by the author. Instead, it synthesizes findings and theoretical proposals from numerous existing empirical and theoretical works within cognitive psychology and educational research.

The author refers to various types of empirical evidence throughout the paper to support points, such as:

• Studies demonstrating the active nature of learners (e.g., Underwood, 1963 on functional stimulus; Shuell, 1969; Tulving, 1968 on organizational patterns in free recall).

• Studies on the role of prior knowledge (Bransford & Johnson, 1972; Dooling & Lachman, 1971).

• Research on expert/novice differences (Chi, 1978; Chi, Glaser, & Rees, 1982).

• Studies supporting stages of psychomotor and verbal learning (Fleishman & Hempel, 1954, 1955; McGuire, 1961; Underwood, Runquist, & Schulz, 1959).

  However, the paper itself does not report new experimental results. Its "validation" comes from the coherence and explanatory power of the synthesized cognitive framework in accounting for diverse phenomena observed in human learning, and its potential to guide future empirical research.

6. Results & Analysis

As a theoretical review paper, this article does not present novel experimental results in the form of data tables or figures generated by the author. Instead, its "results" are the comprehensive synthesis and articulation of a cognitive conception of learning, derived from an analysis of existing literature. The author's analysis strongly validates the shift towards a cognitive understanding of learning by highlighting its explanatory power for phenomena that behavioral theories struggled with.

6.1. Core Results Analysis

The paper argues that cognitive psychology has fundamentally changed how learning is understood:

• Shift in Focus: Cognitive psychology has moved the focus from observable behavior to internal mental processes and knowledge structures. This allows for a deeper understanding of how changes occur in a learner's knowledge and performance.

• Active Learner: The cognitive conception convincingly demonstrates that learners are not passive, but active and constructive. This is supported by evidence like learners spontaneously organizing material or selecting functional stimuli, which behaviorism struggled to explain.

• Importance of Understanding: The paper successfully establishes understanding (comprehension) as a central goal of learning, contrasting with behavioral emphasis on performance. This is evident in theories like Ausubel's subsumption theory and Bransford and Franks' decontextualization.

• Crucial Role of Prior Knowledge: The analysis underscores prior knowledge as a critical determinant of learning, moving beyond simple transfer models. Schema theory and research on expert-novice differences provide strong evidence for this.

• Complex Knowledge Representation: Cognitive theories offer sophisticated ways to represent what is learned (e.g., propositions, schemas, production rules), which are far more nuanced than stimulus-response bonds. Anderson's ACT theory is a prime example, providing a computational model for knowledge acquisition.

• Hierarchical and Phased Learning: The paper highlights the hierarchical nature of cognitive processes (metacognition) and suggests that learning may proceed through distinct phases (declarative to procedural, concrete to abstract), implying that different instructional strategies might be effective at different stages.

  The advantages of the cognitive conception over behavioral models are its ability to explain complex human learning phenomena, the internal mechanisms of knowledge acquisition, and the crucial role of the learner's internal state. Its primary "disadvantage" at the time of publication was its relative newness and the need for more systematic empirical investigation of its proposed variables and processes, as acknowledged in the "Implications for Future Research" section.

6.2. Data Presentation (Tables)

The paper is a review and theoretical work and does not contain any tables of empirical results or data generated by the author's own experiments.

6.3. Ablation Studies / Parameter Analysis

The paper, being a theoretical review, does not include ablation studies or parameter analysis as these are typical of empirical research validating a specific model or algorithm. Instead, the author presents and analyzes different cognitive theories of learning by detailing their components and mechanisms, effectively performing a conceptual "deconstruction" of each theory to highlight its contributions to the overall cognitive conception of learning. For example, Anderson's ACT theory describes generalization, discrimination, and strengthening as core learning mechanisms, and the author explains how these mechanisms contribute to the acquisition and refinement of procedural knowledge. This is a conceptual analysis of theoretical components rather than an empirical study of their individual contributions.

7. Conclusion & Reflections

7.1. Conclusion Summary

The paper concludes that cognitive psychology has begun to seriously engage with human learning research, moving beyond its earlier focus on memory systems. It synthesizes a new cognitive conception of learning characterized by an active, constructive, cumulative, and goal-oriented learner, with a strong emphasis on understanding, prior knowledge, and complex knowledge structures. While distinct from behavioral approaches, it acknowledges common concerns. The paper presents several cognitive theories (e.g., Ausubel's subsumption, Wittrock's generative, Rumelhart & Norman's modes, Anderson's ACT) as examples of this evolving perspective. Ultimately, the paper highlights that learning is multifaceted, with different types of learning and learning outcomes potentially requiring different theoretical principles. It calls for integrating these multiple aspects to address complex classroom learning and instructional challenges.

7.2. Limitations & Future Work

The author points out several limitations and suggests future research directions:

• Lack of precise operational definitions for variables: Little is known about specific variables (e.g., environmental events, instructional interventions) that influence cognitive learning. Future research needs to define these variables more precisely for systematic investigation.
• Complex learning situations: There's a lack of research on variables affecting core cognitive functions (like encoding, combining disparate information, relating new to old information) within complex learning contexts.
• Relationship between generalization in behavioral and cognitive contexts: While cognitive theories use generalization, its nature might differ when applied to structured relationships (analogies) compared to unidimensional stimuli. This is an empirical question.
• Knowledge and Learning:
  • Need to understand the transition from novice to expert, especially the qualitative changes involved.
  • Further research is needed on how domain-specific and domain-independent learning processes interact.
  • Clarifying the relationship between different knowledge types (e.g., declarative vs. procedural), particularly the idea that procedural knowledge might be fundamental.
• Phases of Learning: There is very little empirical evidence on the phases learners go through when acquiring complex, meaningful material. Future work should identify these phases and determine how different variables or strategies might be effective at different stages.
• Theories of Learning from Instruction: The paper hints at the need for dedicated theories of learning from instruction that account for the unique characteristics of instructional situations, rather than simply applying general learning theories.
• Addressing Misconceptions and Buggy Algorithms: More research is needed to understand student misconceptions and systematic errors ("buggy algorithms") to inform effective instruction.

7.3. Personal Insights & Critique

This paper provides an excellent, foundational overview of the shift from behaviorism to cognitive psychology in the context of learning. Published in 1986, it serves as a crucial historical document, articulating the then-emerging cognitive conception of learning that has largely become the bedrock of modern educational psychology.

Inspirations and Transferability:

• Emphasis on the Active Learner: The paper's strong advocacy for the active, constructive nature of learning is a profound insight that remains central to effective pedagogical practices today. It inspires educators to design learning experiences that engage students deeply, rather than treating them as passive recipients. This principle is highly transferable to any domain requiring deep understanding, from complex scientific fields to practical skill acquisition.

• Role of Prior Knowledge: The consistent highlighting of prior knowledge as the foundation for new learning is critical. It underscores the importance of diagnostic assessment, activating prior knowledge, and addressing misconceptions in instruction across all subjects. This insight is universally applicable in teaching and training.

• Metacognition's Importance: The discussion of metacognition as a higher-level process is prescient. Developing metacognitive skills (planning, monitoring, evaluating one's own learning) is now a widely recognized goal in education, equipping individuals to become lifelong learners.

• Complexity of Learning: The paper successfully argues against a monolithic view of learning, suggesting that different types of learning and outcomes might involve different processes. This encourages a nuanced approach to instructional design, tailoring methods to specific learning goals (e.g., memorizing facts vs. understanding concepts).

  Potential Issues, Unverified Assumptions, or Areas for Improvement:

• The "Learning Paradox": While the author acknowledges Bereiter's (1985) work on the learning paradox, the challenge of how fundamentally new, complex cognitive structures emerge from simpler ones remains a deep philosophical and empirical problem. Cognitive theories often describe how existing structures are modified or combined but less definitively how truly novel structures are generated.

• Operationalizing "Meaning" and "Understanding": While cognitive psychology stresses meaning and understanding, their precise operational definitions and measurement remain challenging. The paper touches on this by saying "an operational definition is not readily available." This difficulty affects the rigorous empirical investigation of cognitive learning compared to the more straightforward measurement of behavioral changes.

• Integration of Domain-Specific and Domain-Independent: The paper correctly identifies the need to understand how domain-specific knowledge interacts with domain-independent strategies. This tension continues in research, and a fully integrated theory is still an active area of development. The paper lays the groundwork but doesn't offer a complete solution.

• Focus on Individual Learning: While the paper mentions tutorial interactions (e.g., Socratic dialogue) as a means of restructuring, the primary focus remains on the individual learner's internal cognitive processes. More recent developments in sociocultural theory and situated learning might critique this as underemphasizing the social and contextual aspects of learning.

• Technological Context: Being from 1986, the computer models (like ACT) discussed are advanced for their time but represent an early stage of AI and computational cognitive modeling. Modern AI and machine learning offer new metaphors and tools for understanding learning, potentially enriching or challenging some of these early cognitive conceptions.

  Overall, Shuell's paper is a seminal work that effectively champions the cognitive perspective on learning, providing a robust framework that continues to influence research and practice in education and psychology. Its identified limitations and future research directions remain relevant and have indeed spurred decades of subsequent inquiry.