Replication of Study: What Makes Words Special? Words as Unmotivated Cues (Edmiston & Lupyan, 2015, Cognition)

Replication Author: Jenna Brooks (j8brooks@ucsd.edu), Noah Khaloo, Sihan Yang, Reeka Estacio
Date: November 24, 2024

---

Overview

This repository contains materials, code, and documentation for a replication of Edmiston & Lupyan (2015). The original study examined whether verbal labels (e.g., "dog") activate broader conceptual knowledge compared to environmental sounds (e.g., a dog bark) in an image recognition task. This replication was conducted online using jsPsych and hosted on Prolific.

---

Resources

• Pre-Registration
• Experiment Paradigm

---

Introduction

This study investigates the "label advantage": whether verbal labels lead to faster, more abstract conceptual activation than environmental sounds. Participants were presented with auditory cues (labels or sounds) followed by an image, and asked to judge if they matched the same basic category. Reaction time was the primary measure.

---

Methods

Power Analysis

Sample size (n = 50) was selected to achieve 80% power based on an a priori analysis using the simr package.

Participants

• Recruited via Prolific
• English speakers
• No hearing impairments

Materials

• Cues: Spoken words and environmental sounds for 6 categories (e.g., bird, guitar)
• All audio normalized to 600ms
• Stimuli from: http://sapir.psych.wisc.edu/stimuli2/

Procedure

• jsPsych implementation of Experiment 1A
• Participants respond using keyboard
• Audio check included before trials
• 384 test trials per participant
• 50% of trials were matches

---

Design Overview

• Design: 2x2 within-subjects
  • Cue Type: Label vs. Sound
  • Match: Yes vs. No
• DV: Reaction Time (RT)

---

Analysis Plan

Pre-Registered Plan

• Exclude RTs < 250ms or > 1500ms
• Use Linear Mixed-Effects Regression (with lme4)
• Use Chi-square tests for model comparisons

Hypotheses

• Labels will elicit fastest RTs
• Congruent sounds faster than incongruent
• Incongruent sounds elicit slowest RTs

---

Deviations from Original Study

• Conducted online, not in lab
• Keyboard used instead of controller
• Audio environment controlled through instructions only
• Slight variations in instructions due to lack of access to originals

---

Results Overview

Sample

• 50 participants collected
• 42 passed QA (accuracy > 90%)

Data Cleaning Steps

• Removed trials with RT < 250ms or > 1500ms
• Excluded incorrect or unmatched responses
• Created new congruency column
• Final dataset filtered and structured for modeling

Confirmatory Analysis

• Modeled RTs for "Yes" responses on matching trials
• Condition (label, congruent sound, incongruent sound) was main predictor
• Mixed-effects model included random effects for participants and items
• Pairwise contrasts evaluated condition differences

---

📂 Folder Structure

├── data/
│   └── complete/           # Raw CSV data from participants
├── prereg/
│   └── preregistration.md  # Pre-registered plan
├── scripts/
│   └── writeup_jenna_brooks.qmd          # R analysis and write up
├── README.md              
└── index.html               # Paradigm Design