Brazilian Portuguese Nonce Word Acceptability Study

An R-based analysis investigating whether word structure (syllable length and initial segment type) influences acceptability judgments among native Brazilian Portuguese speakers, using a nonce-word forced-choice task.

Table of Contents

1. Project Overview
2. Hypotheses
3. Data
4. Analysis & Outputs
5. Prerequisites
6. Installation
7. Usage
8. Script Breakdown
9. Interpretation of Results
10. Extending & Customizing
11. Data Source
12. License

---

Project Overview

We analyze acceptability judgments for nonce words varying by:

• Length: number of syllables (e.g., monosyllabic vs. disyllabic)
• Initial segment: type of onset (e.g., stop vs. fricative)

Responses are binary (accept vs. reject). We visualize the contingency with mosaic plots and test for associations using chi-square tests.

---

Hypotheses

• Null hypothesis (H₀): Word structure (length & initial segment) does not influence acceptability judgments.
• Alternative hypothesis (H₁): Word structure does influence acceptability judgments.

---

Data

• File: bp-nonce.csv
• Columns:
  • response: "accept" or "reject"
  • length: e.g. "1" (monosyllabic) or "2" (disyllabic)
  • initial: initial segment category, e.g. "stop", "fricative", etc.

---

Analysis & Outputs

1. Contingency table of response × length × initial.
2. Mosaic plot faceted by length and initial.
3. Chi-square test of response vs. initial.
   • p ≈ 0.59 → fail to reject H₀ at α = 0.05.
4. Chi-square test of length vs. initial.
   • p = 1.00 → no association (as expected).

---

Prerequisites

• R ≥ 4.0
• RStudio (optional)
• Internet access (to install any missing packages)

R Packages

• ggplot2
• ggmosaic

The script will install missing packages automatically.

---

Installation

1. Clone the repository:

   git clone https://github.com/yourusername/bp-nonce-acceptability.git
   cd bp-nonce-acceptability

2. Place bp-nonce.csv in the project root.

---

Usage

Run the analysis script:

# In R or RStudio, set working directory and source:
setwd("path/to/bp-nonce-acceptability")
source("bp_nonce_analysis.R")

This will:

• Print the contingency tables
• Display the mosaic plot
• Print chi-square test results

---

Script Breakdown

• Load libraries (ggplot2, ggmosaic)

• Read data from bp-nonce.csv

• Summarize with xtabs(~ response + length + initial)

• Mosaic plot:

  ggplot(bp) +
    geom_mosaic(aes(weight=1, x=product(response), fill=response)) +
    facet_grid(length ~ initial) + …

• Chi-square tests:

  • chisq.test(xtabs(~ response + initial, data=bp))
  • chisq.test(xtabs(~ length + initial, data=bp))

---

Interpretation of Results

• Response vs. Initial: p = 0.59 → no significant association → fail to reject H₀.
• Length vs. Initial: p = 1.00 → variables are independent (by design).

Conclusion: No evidence that initial segment type influences acceptability. Word structure does not appear to affect judgments in this dataset.

---

Extending & Customizing

• Test response vs. length directly.
• Add other predictors: stress pattern, coda type, etc.
• Use logistic regression (glm(response ~ length + initial, family="binomial")).
• Refine visualization colors, labels, or facet layouts.

---

Data Source

• Brazilian Portuguese nonce-word experiment: collected via [describe methodology or link here].
• Dataset: bp-nonce.csv (private dataset, see project repository).

---

License

This project is licensed under the MIT License. See LICENSE for details.