Measuring Behavior Change

A Quick Review: Measuring Behavior Change

• Core idea: Learning is reflected by measurable changes in a behavior across multiple dimensions.

• Key task: Identify a learned behavior by observing systematic changes in the behavior’s characteristics.

• Dimensions commonly tracked:

  • Errors: whether the number of errors decreases.

  • Topography: the form/shape of the behavior, which can become more variable or more consistent.

  • Intensity: the force or magnitude of the behavior (e.g., pressure, effort).

  • Speed: how fast the behavior is performed.

  • Latency: the delay between a cue and the start of the behavior.

  • Rate: the frequency or number of occurrences per unit time (e.g., letters per minute).

  • Fluency: smoothness and flow of performance, often related to correctness rate.

• Note on rate vs. errors: If you hold rate constant, changes in fluency or errors may reflect quality; increasing rate with more errors can indicate a trade-off unless rate is controlled.

• Example framing: In a task like maze exploration, higher rate and reduced errors indicate learning, while slower latency and more fluent responses indicate better performance.

• Practical takeaway: Learning is not inherently good or bad; it is a change in behavior relative to prior performance.

Data Metrics and Experimental Signals

• Measured signals include: Letters per Minute, Intensity of Pressures (in grams), Topography, Trial progression (Trial 1 vs. Trial 15), Average Time Scores, Looks similar to frequency but corrected for incorrect responses, Speed, Latency, and Response Delay (in seconds).

• Example scales observed in sessions:

  • Trials scale: 1 through 16 (and beyond in multi-session designs).

  • Intensity scale: e.g., 13, 17, 21, 25, 29, 33, 37, 41, 45, 49, 53, 57.

  • Topography and Latency measures captured across trials.

• Specific task measures include:

  • “Number of re-entry errors” in an 8-arm radial maze (categories observed: 0, 1–3, 4–7, 8–10; training sessions may shift distribution).

  • “Average Time Scores” and speed metrics across trials.

  • Fluency and Letters per Minute as proxies for cognitive and motor fluency.

• Conceptual note: Some graphs may show composite variables like “Looks like Frequency but adjusted for incorrect responses,” highlighting the idea that errors are accounted for in rate-like measures.

• Observational insight: Data often compare early trials (Trial 1) with later trials (Trial 15) to infer learning trajectories.

• A note on experimental design visuals: Phase-like progress (e.g., A/B or Training vs. Control) may be embedded in figures showing changes across trials, days, or conditions.

Identifying a Learned Behavior: Demonstrative Changes

• A learned behavior can be demonstrated by measuring at least one of the following changes:

  • A DECREASE in Errors

  • More VARIABLE Topography

  • DECREASED Intensity

  • SLOWER Speed

  • LONGER Latency

  • INCREASED Rate

  • DECREASED Fluency

• Interpretation notes:

  • Decreased errors strongly signals improved accuracy.

  • Increased variability in topography can indicate exploratory refinement or strategy change.

  • Decreased intensity or slower speed may reflect task-optimization or energy conservation in learning.

  • Longer latency can reflect more deliberation or processing time prior to response.

  • Increased rate typically signals quicker responding or higher throughput when appropriate.

  • Decreased fluency can imply a more conservative or effortful performance if rate remains high.

• Important caveat: When rate is held constant, some changes (e.g., decreased fluency) may resemble increased errors; rate control helps distinguish true learning from mere pacing.

Review of Exercise from Last Class: Concrete Examples

• A DECREASE in Errors:

  • Playing an instrument

  • Parallel parking

• More VARIED Topography:

  • Making your Mom’s ‘recipe’

  • Explaining a complex idea

• DECREASED Intensity:

  • “Work Smarter, Not Harder”

  • Making friends

  • Learned helplessness (as a cautionary note in interpretation)

• SLOWER Speed:

  • Cleaning

  • Finishing the job your boss assigned you

• LONGER Latency:

  • Responding to the question “What’s wrong?”

  • Answering questions on exams with tricky wording

• INCREASED Rate:

  • Studying with flashcards

  • Class attendance

• DECREASED Fluency:

  • If rate is held constant, this can resemble increased errors; but if rate is not held constant, fluency can vary inversely with errors.

• Summary principles:

  • Remember learning is reflected by a change in behavior.

  • Learned behavior is not inherently good or bad; it is a differential change.

The Study of Learning & Behavior: Part 2

• Focus: Sources of the data we study in learning and behavior research.

Sources of Data

• Types covered:

  • Anecdote: First- or second-hand report of personal experience.

  • Case Study: Detailed study and description of a single case (often clinical).

  • Descriptive Study: Descriptive data from a group to describe its members.

  • Experiment: Measures the effects of one or more independent variables on one or more dependent variables.

Anecdotes

• Definition: First- or second-hand reports of personal experience (e.g., “my cats come running when they hear the pop-top on the can”).

• Advantages:

  • Can inspire ideas for case studies, descriptive studies, or experiments.

• Limitations:

  • The plural of anecdote is not data – “The plural of anecdote is not data” (Raymond Wolfinger).

  • Anecdotes are not systematic and are open to many interpretations.

Case Studies

• Definition: Detailed study and description of a single case; common in clinical settings (e.g., self-injurious behavior, music).

• Advantages:

  • More systematic than anecdotes.

• Limitations:

  • Time-intensive.

  • Generalizations about behavior are weak.

  • Cannot establish causation.

  • Much data come from reports rather than direct observations.

Descriptive Studies

• Definition: Descriptive study describes a group by obtaining data from its members (e.g., which group solves a riddle better: 6-year-olds vs. 10-year-olds).

• Advantages:

  • More explanatory power than case studies; larger samples.

• Disadvantages:

  • Can suggest hypotheses but cannot test them.

The “True” Experiment

• Core definition: A research design that measures the effects of one or more Independent Variables (IV) on one or more Dependent Variables (DV).

• Key terms:

  • Independent variable (IV): The variable the researcher controls; expected to affect the DV.

  • Dependent variable (DV): The outcome measured; expected to vary with the IV.

• Formal representations:

  • IV affects DV: DV = f(IV) + \epsilon

Between-Subjects Experiments

• Design concept: The IV varies across two or more groups of subjects (between-subjects or between-treatment/group design).

• Terms:

  • Experimental Group: Exposed to the IV.

  • Control Group: Not exposed to the IV.

  • Random Assignment OR Matched Sampling: Methods to create equivalent groups.

• Example: Teaching 2nd graders outside (Experimental Group) vs. in a classroom (Control Group) on basic academic abilities.

Matched Subjects Design

• Definition: Subjects are matched on variables that might affect the DV, then split into two or more groups.

• Rationale: Ensures pre-existing differences are balanced across groups.

• Illustrative example: Sleep-deprived vs. typical sleep groups; match on average sleep duration and processing speed to control confounds.

Within-Subjects Experiments

• Design concept: The IV varies at different times for the same subject (repeated measures).

• Typical use: Group of subjects; can also be a single-subject design (N = 1).

• Example: The effect of a Learning Unit on students’ knowledge of a subject.

ABA Reversal Design (Single Subject)

• Type: A within-subjects design where behavior is observed before (A) and after (B) an experimental manipulation.

• Reversal: The original condition (A) is restored, sometimes followed again by the manipulation (B).

• Purpose: To determine whether observed changes are due to the manipulation or other factors (medication vs. placebo distinction).

• Notation: A A B B (with reversal implemented to test causality)

Animal Research and Human Learning

• Benefits:

  • Animals allow greater experimental control.

  • Provides essential insights into learning and behavior that can inform human research.

• Criticisms:

  • Ethical concerns: animal rights objections to using animals in research.

  • Some argue simulations or computer models could replace animal work, though this is not always feasible or sufficient.

• Takeaway: A balanced approach uses both animal and human studies to understand learning and behavior.

• Preview topic: Pavlovian (classical) conditioning will be explored in the next session.