Table 1.
Participant characteristics (N = 20).
Fig 1.
Example stimulus from the self-paced reading task.
Note. This example is taken from the experimental stimulus set (Stimulus ID: mzt.ekn.II). Numbers below each chord indicate the harmonic information content (IC) values calculated by the Information Dynamics of Music (IDyOM). Higher IC values indicate lower conditional probability given the preceding context—that is, events that are less expected based on the statistical regularities learned by the model. This example illustrates a pattern commonly observed in our stimulus set: harmonic IC values tend to decrease toward the end of progressions. This pattern likely reflects the conventional cadential structure of our stimuli, in which accumulating context combined with stylistically predictable harmonic patterns may constrain model predictions for later events, making them more predictable (lower IC values). We note that this is not a universal property of the model; unexpected events can receive high IC values regardless of position.
Fig 2.
Distribution of harmonic information content (IC) across all stimuli in the self-paced reading task.
Note. N = 368 chords (46 progressions × 8 chords/progression). The prominent spike (0.39) represents all first chords, which received identical harmonic IC values from the Information Dynamics of Music (IDyOM) as they lack preceding context. These first-chord data were excluded from subsequent analyses to accommodate temporal dependencies in the statistical modeling. The remaining distribution shows variability in harmonic predictability across the stimuli.
Fig 3.
Schematic illustration of the self-paced reading task.
Note. Participants viewed chord progressions one chord at a time using a moving-window presentation method. At trial onset, only the key signature was displayed. Each spacebar press revealed the next chord at a new position to the right while the previous chord disappeared. The key signature remained visible throughout. Reading time (RT) for each chord was measured as the duration between successive spacebar presses. After viewing the complete progression, participants judged whether a probe chord had appeared in the sequence.
Table 2.
Descriptive statistics and normality tests for sight-reading performance metrics.
Fig 4.
Comparison of sight-reading performance across tonal and atonal conditions.
Note. Box plots show the distribution of (A) pitch error rate, (B) rhythm precision (beat-normalized mean absolute deviation; bn-MAD), and (C) rhythm consistency (coefficient of variation; CV) scores. Individual data points are overlaid, with gray lines connecting each participant’s performance across conditions to visualize within-subject changes.
Fig 5.
Individual differences in reading time (RT) distributions for the self-paced reading task.
Note. Box plots show the distribution of (A) raw RTs in seconds and (B) log-transformed RTs for each participant.
Fig 6.
Individual participant slopes for the effect of harmonic information content (IC) on log-transformed reading time.
Note. Each panel shows data from one participant with individual observations (gray dots), linear regression fit (blue line), and 95% confidence interval (shaded area). All participants demonstrated positive slopes, indicating consistent processing costs for harmonically less predictable (high harmonic IC) chords.
Table 3.
Linear Mixed model results predicting log-transformed reading times: Harmonic IC × sight-reading performance metrics.
Fig 7.
Model-predicted main effects of sight-reading metrics on log-transformed reading time from the self-paced reading task.
Note. (A) Pitch error rate effects, (B) beat-normalized mean absolute deviation (bn-MAD) effects, and (C) coefficient of variation (CV) effects. Lines represent predictions from linear mixed models with 95% confidence intervals (shaded areas).
Table 4.
Three-way interaction results: Harmonic IC × sight-reading performance metrics × NESI score.