PSY 350 CH. 6 Methodological Control in Experimental Research

🧠 Overview

This chapter explains how researchers maintain control in experimental psychology, focusing on two major types of experimental designs:

• Between-Subjects Designs

• Within-Subjects Designs

It also covers:

• How to create equivalent groups.

• How to handle order effects.

• How to prevent experimenter and participant bias.

• Special developmental research designs (cross-sectional, longitudinal, and sequential).

⚖ Between-Subjects Designs

Definition

Participants are divided into different groups, with each group experiencing only one level of the independent variable (IV).

When It’s Used

• When the IV is a subject variable (e.g., gender, introversion vs. extroversion).

• When deception or learning effects make repeated exposure impossible.

Example

In the Sigall and Ostrove (1975) study on attractiveness and sentencing:

• Groups saw “Barbara Helm” as attractive, unattractive, or no photo.

• Crime type varied (burglary vs. swindle).

• Attractiveness helped in burglary but hurt in swindle.

• Because participants couldn’t “unsee” one version, the design had to be between-subjects.

Advantages

• Participants are naïve to other conditions.

• No carryover or learning effects.

Disadvantages

• Requires many participants.

• Risk of nonequivalent groups (differences due to individual traits, not IV).

⚙ Creating Equivalent Groups

Purpose

Ensure differences between groups are due to the independent variable, not personal characteristics.

1. Random Assignment

Every participant has an equal chance to be in any condition.

Difference from Random Selection

• Random Selection = choosing participants from a population.

• Random Assignment = placing participants into conditions after selection.

Goal

Distribute personal differences (e.g., intelligence, anxiety) evenly across groups.

Blocked Random Assignment

Ensures equal numbers in each condition (e.g., using computer randomizer).

Example

Memory study with 2-second vs. 4-second presentation rates:

• Equal anxious participants per group → equivalent results.

• Unequal distribution → nonequivalent groups and possible Type II error (missed real effect).

2. Matching

Used when small samples make random assignment risky.

Steps

1. Measure participants on a matching variable (e.g., anxiety).

2. Pair or cluster participants with similar scores.

3. Randomly assign one of each pair to each group.

Conditions for Matching

• The variable must correlate with the dependent variable.

• The variable must be measurable and practical to collect.

Examples

• Fung & Leung (2014): Matched autistic children on IQ and verbal fluency to test therapy dogs.

• Goosens et al. (2014): Matched 9-year-olds by vocabulary to test “retrieval practice.”

Challenges

• Hard to decide which variables to match.

• Multiple matching variables complicate logistics.

🔁 Within-Subjects Designs

Definition

Each participant experiences all levels of the independent variable. Also called a repeated-measures design.

Advantages

• Requires fewer participants.

• Eliminates individual differences between conditions.

• Increases statistical sensitivity (smaller variance).

Example

Golf study:

• Same golfers hit both brands of golf balls.

• Removes skill differences between groups.

Main Problem: Order Effects

Performance changes due to the order in which conditions are experienced.

Types of Order Effects

• Progressive effects: steady changes (e.g., practice or fatigue).

• Carryover effects: earlier conditions influence later ones (e.g., noise studies).

Example

• Predictable vs. unpredictable noise → performance differs depending on order (carryover).

🔄 Controlling Order Effects: Counterbalancing

Definition

Present conditions in different orders to different participants.

1. Testing Once per Condition

Complete Counterbalancing

• All possible sequences used (number of orders = n!).

• Example: 3 conditions → 3! = 6 orders.

Partial Counterbalancing

• Use a subset of all possible orders.

• Methods:

  • Random selection of orders.

  • Random order per subject.

  • Balanced Latin Square – ensures each condition appears in each position and follows every other condition equally.

Example

Reynolds (1992) – Chess expertise:

• 6 games, randomized order per participant.

2. Testing More Than Once per Condition

Used in perception or attention research.

Reverse Counterbalancing

• Conditions in one order, then reversed (e.g., A–B–C–D → D–C–B–A).

• Used by J. Ridley Stroop (1930s).

Block Randomization

• Each block contains all conditions in randomized order before repeating.

• Prevents predictability.

Example

Hagemann et al. (2008): Tae kwon do referees judged fights twice (red vs. blue gear).
Red uniforms biased scoring higher. Used block randomization with counterbalancing.

🧩 Developmental Research Designs

1. Cross-Sectional Design

• Between-subjects approach.

• Compares different age groups at one time.

Problem: Cohort Effects

• Differences may reflect historical environment, not age.

• Example: Comparing 45-, 65-, and 85-year-olds (school, culture, technology differences).

2. Longitudinal Design

• Within-subjects approach.

• Studies the same group over time.

Problems

• Attrition (dropouts).

• Ethical concerns – ongoing consent as people age.

3. Cohort Sequential Design

• Combines both methods.

• Tests multiple cohorts over time.

Example

Schaie’s Seattle Longitudinal Study (1956–2005):

• New cohorts added every 7 years.

• Found that mental ability decline starts after age 60 and depends on health and education.

Classic Case: Terman’s Longitudinal Study (1921–2010)

• 1,470 gifted children (“Termites”).

• Tracked over decades.

• Results: Gifted individuals were well-adjusted, successful, and loyal to the project.

• High participant retention (92–98%).

• Showed importance of long-term measurement and participant engagement.

👩‍🔬 Controlling for Bias

1. Experimenter Bias

When the researcher’s expectations influence results.

Classic Example

Rosenthal & Fode (1963):

• Experimenters told to expect positive or negative ratings.

• Subtle cues (tone, expression) affected participant responses.

• Even rats ran faster when experimenters thought they were “maze-bright.”

Other Findings

• Experimenter’s race, gender, or demeanor can affect participant behavior.

Controls

• Mechanize procedures (automated tests, computers).

• Use research protocols (standardized scripts).

• Double-Blind Procedure:

  • Neither participants nor experimenters know who’s in which condition.

  • Example: Williams & Bargh (2008) – warm vs. cold coffee influencing personality perception.

2. Participant Bias

Definition

When participants alter behavior because they know they’re being studied.

Types

• Trying to guess the hypothesis.

• Acting how they think they “should.”

• Hawthorne Effect – increased productivity simply because people know they’re observed.

Origins

Western Electric Plant, Hawthorne (1924–1933):

• Worker productivity increased not from lighting or breaks but from being observed.

• Later historians question the interpretation—may have been politically distorted.

✅ Key Takeaways