Within-Subjects Experimental Designs and Counterbalancing Techniques

Overview of Within-Subjects Designs
Definition and Core Concepts
  • Repeated measures design where each participant completes every condition of the independent variable (IVIV).

  • Each participant acts as their own control baseline.

Key Advantages
  • Reduced Error Variance: Controls for individual differences and noise between subjects.

  • Economy: Requires fewer overall participants than between-subjects designs.

  • Extraneous Control: Offers greater control over participant-related extraneous variables.

Confounds and Threats to Validity
  • Order Effects: When the presentation sequence confounds experimental outcomes.

    • Practice Effects: Performance improvements from repeated exposure or familiarity.

    • Fatigue Effects: Performance decline due to tiredness, boredom, or cognitive load.

    • Carry-Over Effects: Lingering impacts of one condition spilling into subsequent conditions.

Complete Counterbalancing
General Principles
  • Balances practice and order effects by testing conditions across all possible sequence orders.

  • Ensures every potential order permutation is represented.

Specific Techniques
  • ABBA Design:

    • Conditions are presented in a forward order, followed by the reverse order.

    • Example: English then Spanish, followed by Spanish then English.

  • Block Randomization:

    • Every condition is placed into a block with randomized internal ordering.

    • Ideal when participants complete each condition multiple times.

    • Evenly distributes fatigue and prevents participants from anticipating upcoming conditions.

  • All-Orders Block Assignment:

    • Generates every unique permutation (n!n!) as a distinct sequence.

    • Example: 33 conditions produce 3!=63! = 6 distinct blocks.

    • Each participant receives one specific order, with all orders balanced across the total sample.

Incomplete (Partial) Counterbalancing
General Principles
  • Selects a strategic subset of possible sequences when complete counterbalancing is impractical.

Random Order with Rotation
  • Begins with a randomly selected sequence and systematically rotates condition positions for successive participant groups.

  • Example with 44 conditions (2424 total possible permutations):

    • Group 1: BCAD

    • Group 2: CADB

    • Group 3: ADBC

    • Group 4: DBCA

Latin Square Design
  • A structured matrix where every condition:

    • Appears at each ordinal position exactly once.

    • Precedes and follows every other condition an equal number of times.

  • Sequence Construction Rules:

    • Initial Sequence Formula: Generated using the sequence 1,2,n,3,(n−1),4,(n−2),5,…1, 2, n, 3, (n - 1), 4, (n - 2), 5, \dots

    • Step Progression: Derive each following sequence by adding 11 to each value of the previous row.

    • Even Conditions (nn): Requires nn sequences (e.g., 44 conditions need 44 sequences: ABDC, BCAD, CDBA, DACB out of 2424 possible).

    • Odd Conditions (nn): Requires 2n2n sequences (e.g., 55 conditions require generating 55 base sequences plus their 55 exact reversals, totaling 1010 sequences).