All statistical plots from the listener perception study. 85 listeners, 36 speakers. ← Back to the main project page.
Section 1: Overall Perception Accuracy
1A — Overall accuracy scatter. r = 0.287, R² = 0.083, p = 0.089.
1B — Accuracy broken down by listener gender.
1C — Accuracy broken down by listener sexual orientation.
1D — Accuracy broken down by listener familiarity with gay individuals.
1E — Correlation comparison barplot across demographic groups.
Section 2: Individual Listener Accuracy
2A — Distribution of per-listener accuracy scores.
2B — Listener accuracy vs. self-reported characteristics.
Section 3: Inter-Rater Reliability
3A — Inter-rater correlation matrix across all 85 listeners. ICC(2,k) = 0.964.
3B — Distribution of listener agreement by speaker.
3C — ICC broken down by listener demographics.
Section 4: Consensus vs. Accuracy
4A — Speaker consensus (SD of ratings) vs. absolute prediction error. r = 0.269, p = 0.113.
4B — Mean-based error vs. individual listener MAE by speaker.
Section 5: Speaker Characteristics & Clustering
5A — Speaker readability quadrants: agreement (SD) vs. accuracy (absolute error).
5B — Listener accuracy broken down by speakers' actual self-reported score.
5C1 — Hierarchical clustering dendrogram of speaker rating profiles.
5C2 — PCA variance diagnostics for speaker rating profiles.
5C3 — K-means clustering diagnostics (elbow plot).
5C4 — DBSCAN k-distance plot (2D PCA space).
5C4 — Clustering comparison in 2D PCA space.
5C5 — DBSCAN k-distance plot (50% variance).
5C5 — Clustering in 50% variance PCA space.
5C6 — DBSCAN k-distance plot (70% variance).
5C6 — Clustering in 70% variance PCA space.
5D1 — PC1 score vs. speakers' actual self-reported sexuality.
5D2 — PC1 score vs. mean listener rating per speaker.
5D3 — PC1 extremity vs. prediction error.
5E — Higher principal components vs. mean listener rating.
5F — Rating SD (disagreement) vs. PC2 score.
5G — Rating SD vs. non-consensus distance.
Section 6: Listener Response Bias
6A — Distribution of per-listener mean ratings across all speakers. Mean = 2.73.
6B — Listener religiosity vs. rating bias. r = 0.040, p = 0.721.
6C — Distribution of per-listener rating range (max − min across speakers).
Section 7: Group Interactions
7A — Gender × orientation interaction on perception accuracy.
7B — Match effects: accuracy when listener and speaker share orientation vs. not.
Section 8: Signal Detection Theory
8A — Distributions of d' (sensitivity) and c (criterion) across listeners. Mean d' = 0.11, Mean c = −0.08.
8B — d' (sensitivity) broken down by listener demographic group.
Section 9: Error by Speaker
9A — Mean-based prediction error for each individual speaker.
9B — Individual listener MAE for each speaker.
Speaker Rating Histograms
Distribution of listener ratings (1–5) for each individual speaker.
Speaker JC1
Speaker LQ2
Speaker PP3
Speaker MM4
Speaker HZ5
Speaker GA6
Speaker DM7
Speaker AB8
Speaker AI9
Speaker BO10
Speaker BR11
Speaker CA12
Speaker DH13
Speaker EX14
Speaker GE15
Speaker GO16
Speaker HY17
Speaker JF18
Speaker LA19
Speaker LS20
Speaker NO21
Speaker PE22
Speaker PL23
Speaker PR24
Speaker PS25
Speaker SI26
Speaker TU27
Speaker TY28
Speaker VI29
Speaker WCH30
Speaker WCL31
Speaker WH32
Speaker WT33
Speaker XX34
Speaker RU35
Speaker QI36