chap 6 | EXP

Three Building Blocks of Within-Subjects Design

01 Randomization, 02 Within-Subjects Fundamentals, 03 Carryover & Counterbalancing :

Getting participants into conditions without letting bias creep in.

Randomization

The process of assigning participants to conditions purely by chance — it is the foundation every other technique in this lecture builds on.

Randomization

Subjectivity Rule of Randomization

The assignment method itself must remove any human judgment from deciding who receives which condition — otherwise it is not truly random

The point in time at which randomization occurs is important — it should happen at a well-defined step in the procedure, consistently applied to every participant.

Timing Rule of Randomization

rules of randomization

it cannot be appllied subjectively , timing matters

Assignment relies entirely on an unbiased chance procedure — no structure is imposed on how conditions fall out.

Simple Randomization

Four Common Methods of Simple Randomization

Flip a coin, Use a random-number generator, Draw from a table of random numbers, Partition a numeric range into the number of conditions

Problem with Simple Randomization

With pure chance, group sizes can end up unequal — sometimes badly so, especially with smaller samples

Solves the imbalance problem - it guarantees that any imbalance between conditions never exceeds a specified amount.

Blocked Randomization

How Blocked Randomization Works

Write out all permutations of N conditions taken B at a time — this fixed-size group is the "block." At the start of each block, one ordering is selected at random. Block sizes should be greater than 2

Example Permutations for Block Size of 4

AABB ABAB BAAB BABA BBAA ABBA

Risk of No Blocking

Group sizes and composition can drift apart by chance.

Guarantee of Blocking

Within each block, conditions stay balanced.

First stratify participants by a measured factor; then randomize independently within each stratum.

Stratified Randomization

Stratify First Step

Participants are split into groups ("strata") based on a factor measured in advance — for example, baseline attention span.

Randomize Within Stratum Step

Randomization is applied separately inside each stratum — it can itself be either simple or blocked. '

Pure chance — coin flips, random-number generators. Easiest to run, but group sizes can end up unequal.

Simple Randomization

Chance within fixed-size blocks. Guarantees balance never drifts by more than the block size allows.

Blocked Randomization

Group first by a known factor, then randomize (simply or blocked) within each group.

Stratified Randomization

Which design- Different groups of participants receive different levels of the IV. Each participant serves in only one condition of the IV. Independent samples or matched samples are used in each condition.

Between-Subjects Design

Which Design-One group of participants receives every level of the IV. Each participant serves in all conditions of the IV. The same sample is used in each condition.

Within-Subjects Design

What design is this - Peppermint and digit span

Between-Subjects (Independent Samples Example)

What design is this - Ball toss, matched on height

Between-Subjects (Matched Samples Example)

What design is this - Chocolate and letter-cancellation (sustained attention) task; levels of processing

Within-Subjects (Same Sample Example)

Participants Serve as Their Own Controls

Comparing the treatment effect means examining how each participant's performance changes across treatments — not comparing different people to each other.

Compare Within a Person, Across Conditions

Each participant's response at one level of the IV is compared to their own response at a different level of the IV.

Repeated-Measures Design Name Origin

The name reflects the method: each participant's behavior is measured repeatedly, once per condition.

Advantages of Within-Subjects Design

Controls Participant Variables, Reduces Error Variance, More Powerful Design, Requires Fewer Participants

Controls Participant Variables (Advantage)

Removes many participant variables that might otherwise systematically vary with the DV — each person is compared to themselves.

Reduces Error Variance (Advantage)

Because individual differences no longer contaminate the comparison, variability in the data goes down.

More Powerful Design (Advantage)

Lower error variance makes real treatment effects easier to detect statistically.

Requires Fewer Participants (Advantage)

Since every participant contributes data to every condition, recruitment needs shrink.

Disadvantages of Within-Subjects Design

More Demanding on Participants , Attrition / Mortality, Carryover (Order) Effects

More Demanding on Participants (Disadvantage)

Sitting through every condition takes more time and effort than experiencing just one.

Attrition / Mortality (Disadvantage)

Participants may drop out partway through, resulting in a loss of data for the conditions they never reached.

Carryover (Order) Effects (Disadvantage)

Experiencing one condition can change how a participant responds to the next.

The cost of repeated testing — and the ordering techniques built to keep that cost from wrecking your results.

(Carryover Effects & Counterbalancing)

occur when a previous treatment changes the behavior observed in a subsequent treatment

Carryover effects

Scores in later conditions of within subjects design can?

come out higher or lower than they otherwise would, simply because the participant was already exposed to an earlier condition.

Confound Threat of Carryover Effects

If carryover effects are not controlled, they become a confounding variable — you can no longer tell whether the IV or the order caused the result

Sources of Carryover

Learning , Fatigue , Habituation , Sensitization , Contrast , Adaptation

what source of Source of Carryover is Learning a task in the first treatment may affect performance in the second.

Learning (Source of Carryover)

what source of Source of Carryover is Fatigue from earlier treatments may affect performance in later treatments.

Fatigue (Source of Carryover)

what source of Source of Carryover is Repeated exposure to a stimulus may lead to unresponsiveness to that stimulus.

Habituation (Source of Carryover)

what source of Source of Carryover is Exposure to one stimulus may make a subject respond more strongly to another.

Sensitization (Source of Carryover)

what source of Source of Carryover is Subjects may compare treatments against each other, which may affect their behavior.

Contrast (Source of Carryover)

what source of Source of Carryover is If a subject undergoes adaptation, earlier results may differ systematically from later ones.

Adaptation (Source of Carryover)

Exposing different participants to different orders of conditions.

Counterbalancing

Result Without Counterbalancing

Same order every time — order effects stack up unevenly.

Result With Counterbalancing

Different orders across participants — order effects cancel out.

Different treatment sequences are presented to the same participant — everyone eventually experiences every order.

Within-Subject Counterbalancing

Different treatment sequences are presented to different participants — the group as a whole covers every order.

Within-Group Counterbalancing

Requirement of Within-Subject Counterbalancing

Each participant experiences each condition of the experiment at least twice, using a different order each time.

Balance order effects by presenting one sequence, then its exact opposite.

The ABBA Method

Why the ABBA Method Works

A straight A→B run lets fatigue build only on B. Mirroring the order cancels that drift out. Fatigue rises left to right, but each condition gets one early slot and one late slot — so it nets out.

Three Requirements of Within-Group Counterbalancing

Each treatment must be presented to a participant an equal number of times; 2. Each treatment must occur an equal number of times at each position; 3. Each treatment must precede and follow each of the other treatments an equal number of times '

Factorial Growth of Possible Orders

The number of possible orders is a factorial of the number of conditions — it explodes fast (e.g., 2=2, 3=6, 4=24, 5=120, 6=720).

Why Incomplete Counterbalancing Exists

Complete counterbalancing quickly becomes impossible to run in full — which is exactly why incomplete counterbalancing exists.

All possible treatment sequences are presented — at least one participant receives each order of the conditions.

Complete Counterbalancing

Advantage of Complete Counterbalancing

Offers the best control for order effects of any method covered here.

Disadvantage of Complete Counterbalancing

Requires a large number of participants once there are more than 4 conditions — usually reserved for 4 or fewer.

Only a portion of all possible treatment sequences is presented — a practical fallback once complete counterbalancing is out of reach.

Incomplete Counterbalancing

Random Selection Method of Incomplete Counterbalancing

Sequences are drawn at random from the full set of possible orders.

Rotation Selection Method of Incomplete Counterbalancing

A starting order is chosen at random, then systematically rotated to generate the remaining sequences used.

Scores in one condition are affected to the same extent, regardless of which condition preceded it. Use a within-groups design with counterbalancing.

Equal Carryover Effects

Scores in one condition depend on which condition preceded it. Counterbalancing will NOT eliminate this — use a between-subjects design instead (independent or matched samples).

Differential Carryover Effects

Design Path for No Carryover Effects

use Simple Within-Subjects Design. No special ordering technique is required — proceed with repeated measures as designed.

Design Path for Equal Carryover Effects

use Within-Groups Design + Counterbalancing. Counterbalancing is effective here — use complete counterbalancing if 4 or fewer conditions, incomplete if more.

Design Path for Differential Carryover Effects

use Between-Subjects Design. Counterbalancing cannot fix this — switch to independent or matched samples instead.

According to the diagram, when should you use complete counterbalancing?

If 4 or fewer conditions (under Equal carryover / Within-Groups Design)

According to the diagram, when should you use incomplete counterbalancing?

If more than 4 conditions (under Equal carryover / Within-Groups Design)

What are the two design options to switch to when counterbalancing cannot fix Differential carryover?

1. Independent samples, 2. Matched samples