A composer generates a melody by selecting each note independently and uniformly at random from all 12 pitch classes — maximizing Shannon entropy. What does information theory predict about listener engagement with this melody?
AMaximum engagement, because each note carries maximum information and surprises the listener
BModerate engagement, because listeners can form partial expectations from the equal distribution
CLow engagement, because without patterns, listeners cannot form expectations to be fulfilled or violated
DHigh engagement initially, declining only after the listener memorizes the pattern
Maximum entropy means no predictable structure — every note is equally surprising. Without structure, listeners cannot form expectations, and the anticipation-and-resolution cycle that drives musical engagement cannot occur. Information theory predicts peak engagement in an intermediate entropy zone where expectations form and are sometimes confirmed, sometimes violated. Option A confuses 'high information content per note' with 'high engagement' — these come apart when there is no pattern for the brain to model.
Question 2 Multiple Choice
In tonal music, the leading tone resolving to the tonic has very high probability of occurring. In information-theoretic terms, this resolution has:
AHigh information content, because it is a significant musical event
BLow information content, because it is highly predictable
CZero entropy, because the entire passage is deterministic once the leading tone sounds
DHigh entropy, because resolution can occur at many different moments
Information content = −log₂(p). A high-probability event has low information content: if p is close to 1, −log₂(p) is close to 0. The leading tone resolution is expected — it carries little 'news.' This is why its eventual arrival feels satisfying rather than surprising: the resolution confirms the expectation rather than violating it. A sudden chromatic pitch, by contrast, has high information content because it is rare (low p) and therefore genuinely surprising.
Question 3 True / False
A serialist composition organizes all pitches according to a deterministic tone row, so the composer's entropy is zero. Yet listeners unfamiliar with the row will experience high entropy in the piece.
TTrue
FFalse
Answer: True
This is the key distinction between structural entropy (from the composer's perspective) and perceptual entropy (from the listener's perspective). The row imposes complete determinism on pitch selection, so the composer's model has zero entropy. But a listener who cannot perceive the row — because serialism does not match auditory pattern recognition — has no usable statistical model and therefore experiences the sequence as nearly random: high perceptual entropy. Information theory thus separates the encoder's structure from the decoder's experience.
Question 4 True / False
Information content and entropy are the same thing — a melody where each note has high information content necessarily has high entropy.
TTrue
FFalse
Answer: False
Information content is a property of a specific event: −log₂(p(xᵢ)) for one occurrence. Entropy H(X) = −Σ p(xᵢ) log₂ p(xᵢ) is the *expected* information content — an average over the entire probability distribution. A single surprising event (high information content) embedded in a mostly predictable piece does not make the whole piece high-entropy. Conversely, a piece can have high entropy without any single event being particularly surprising — they are all roughly equally probable.
Question 5 Short Answer
Explain the 'optimal entropy zone' concept: why do both very low-entropy (highly predictable) and very high-entropy (highly random) music tend to disengage listeners, while intermediate entropy engages them most?
Think about your answer, then reveal below.
Model answer: Listener engagement depends on the formation and resolution of expectations. Very low entropy means the listener's predictive model succeeds nearly perfectly every time — no surprises, nothing for the prediction engine to do, and the music becomes boring. Very high entropy means no patterns can be detected, so no expectations form — the listener cannot engage with anticipation or surprise because there is nothing to anticipate. At intermediate entropy, patterns are detectable enough for expectations to form, but violations and confirmations occur in a ratio that sustains interest. The brain is actively predicting and sometimes rewarded, sometimes surprised.
This model has empirical support in music cognition research: it predicts why music with some predictability (tonal structure, repeating rhythms) holds attention, and why both rigid minimalism and dense atonality can exhaust listeners in different ways. It also explains why familiarity breeds enjoyment up to a point before satiation sets in — repeated exposure lowers entropy from the listener's perspective as the piece becomes more predictable.