Chapter 17 Exercises: Fractals, Self-Similarity & Musical Patterns

Part A: Fractal Identification and Basic Concepts

A1. The Koch snowflake is constructed by starting with an equilateral triangle and repeatedly adding smaller equilateral triangles to the middle third of each edge. After four iterations, describe the following properties: - What happens to the perimeter of the figure at each step? Does it converge or diverge? - What happens to the area at each step? Does it converge or diverge? - Explain why the Koch snowflake has a well-defined area but an infinite perimeter, and what this tells us about the relationship between scale and complexity in fractal objects.

A2. The Sierpinski triangle is constructed by repeatedly removing the central quarter-triangle from each remaining triangular piece. After iteration n, what fraction of the original triangle remains? Write a formula for this fraction. As n approaches infinity, what fraction remains? What does this mean about the relationship between the Sierpinski triangle's "size" (in the two-dimensional sense of area) and its "complexity" (in the sense of structural detail)?

A3. Explain in your own words the difference between exact self-similarity (as in the Mandelbrot set, where zooming in reveals exact copies of the whole) and statistical self-similarity (as in a coastline, where zooming in reveals similar but not identical structure). Which type of self-similarity is more common in music? Give a specific musical example of each type if possible.

A4. The fractal dimension D of the Cantor set is log(2)/log(3) ≈ 0.631. Explain intuitively what it means for a set to have dimension 0.631 — more than 0 (like a finite collection of points) but less than 1 (like a line segment). How does the construction of the Cantor set (repeatedly removing middle thirds) produce a dimension between 0 and 1?

A5. In the chapter, the fractal dimension of melodies is discussed as a measure of melodic complexity. The dimension ranges from approximately 1.0 (smooth, stepwise melody) to 2.0 (random, maximally complex). Give an example of a real melody you know and estimate its fractal dimension qualitatively (low, medium, or high) based on its characteristics. Justify your estimate by describing the melody's features (range of leaps, chromatic content, overall smoothness).

Part B: Pink Noise Analysis and Power Spectra

B1. White noise has a flat power spectrum (equal power at all frequencies). Brown noise has a power spectrum that falls as 1/f². Pink noise (1/f noise) falls as 1/f. For each of the following signals, determine which type of noise its power spectrum most resembles, and explain your reasoning: - A melody that takes a random walk (each note is the previous note plus a random step of ±1 semitone) - A melody where each note is chosen completely independently at random from the chromatic scale - A melody where each note is the same as the previous note 80% of the time and changes randomly 20% of the time - A typical Bach chorale melody

B2. Voss and Clarke (1975) found that the pitch sequences of different types of music all fall in the 1/f range. However, they also found systematic differences: some genres have steeper spectra (closer to 1/f²) and some have flatter spectra (closer to 1/f°, white noise). Based on your musical knowledge and the concepts discussed in the chapter, predict which of the following genres would have the steepest 1/f spectrum (most correlated, most predictable) and which would have the flattest (least correlated, most random), and explain your reasoning: Gregorian chant, bebop jazz improvisation, Beethoven symphonies, West African drum ensemble music.

B3. The "Hurst exponent" H is a measure of the degree of long-range correlation in a time series. H = 0.5 corresponds to white noise (no correlation); H = 1.0 corresponds to brown noise (maximum correlation); H in between (0.5 < H < 1.0) corresponds to pink noise (1/f-like correlation). For a musical melody, propose an intuitive interpretation of the Hurst exponent in musical terms. What does a high Hurst exponent (close to 1.0) mean about a melody's structure? What does a low Hurst exponent (close to 0.5) mean?

B4. The following pitch sequence is given (using scale degrees in C major, 1 = C, 2 = D, etc.): 1, 2, 3, 2, 3, 4, 3, 4, 5, 4, 5, 6, 5, 6, 7, 6. This is a simple ascending sequence with local oscillations. Describe the local correlation structure (are adjacent notes correlated?) and the long-range correlation structure (is the overall trend predictable?). Would you classify this as closer to white noise, pink noise, or brown noise? Explain.

B5. Researchers have found that the heartbeats of healthy individuals have approximately 1/f variability in their inter-beat intervals, while individuals with certain cardiac conditions show either more random (white noise) or more regular (brown noise or periodic) variability. Given what you know about the 1/f property in music, propose a hypothesis about why 1/f variability might be a signature of health in biological systems. What does the 1/f property provide that pure regularity or pure randomness does not?

Part C: Algorithmic Pattern Generation

C1. Apply the following L-system rule four times and describe the resulting string. Seed: "F". Rules: F → F+F−F−F+F (where + means "turn right 90°" and − means "turn left 90°" in a turtle-graphics interpretation). What geometric figure does this L-system generate? If you mapped F to "play note C4" and + to "go up a half step" and − to "go down a half step," what would the resulting "melody" sound like?

C2. The Fibonacci sequence (1, 1, 2, 3, 5, 8, 13, 21, ...) appears in the L-system described in section 17.8. The Fibonacci sequence is also connected to the golden ratio φ ≈ 1.618. The time signature 5/4 (used in Holst's "Mars" from The Planets, and Dave Brubeck's "Take Five") divides 5 beats as 3+2 or 2+3 — Fibonacci numbers. Is this connection between Fibonacci numbers and musical time signatures more than coincidental? Research the use of Fibonacci proportions in Bartók's music (particularly his string quartets and the Music for Strings, Percussion and Celesta) and discuss whether the connection is compositional intention, post-hoc analysis, or mathematical coincidence.

C3. Design a simple musical L-system. Choose: - A seed (starting symbol, representing a musical idea — e.g., "A" = play a C4 quarter note) - At least two symbols with musical meanings (e.g., A = C4, B = E4, both quarter notes) - At least two production rules (e.g., A → AB, B → A) - A mapping to actual musical notes

Apply your rules four times and write out the resulting musical sequence. Does it have self-similar structure? Does it sound musical? What would you need to add (harmonic context? Dynamic variation?) to make it more musically compelling?

C4. Conlon Nancarrow (discussed in the case study) created player piano works with tempo ratios of extreme precision — canons where one voice plays at √2 times the tempo of another, or 12/15 of another. These ratios produce canons that converge (reach the same point simultaneously) after a very long time or never. Explain how such tempo ratios create a kind of rhythmic self-similarity over very long time scales. If voice A is at tempo T and voice B is at tempo φT (where φ is the golden ratio ≈ 1.618), when (if ever) do the voices return to their original phase relationship?

C5. The "dragon curve" is an L-system fractal generated by repeatedly folding a strip of paper in half in the same direction. Its L-system representation uses rules that create a complex, self-avoiding path. Describe in musical terms what a "self-avoiding" melody would mean (a melody that never returns to the same pitch-time location). Is this a musical concept that has compositional value? Compare to twelve-tone music, where the composer avoids repeating a pitch class until all twelve have been heard.

Part D: Fractal Dimension and Musical Complexity

D1. The box-counting dimension is one way to measure the fractal dimension of a melody treated as a curve in pitch-time space. To compute it: draw a grid of boxes of size ε × ε over the pitch-time plane; count N(ε) = the number of boxes containing part of the melody; the dimension D is estimated as D = log(N(ε)) / log(1/ε) for small ε. Sketch two melodies — one smooth and stepwise, one jagged with large leaps — and qualitatively compare their box-counting dimensions for a fixed grid size. Which has a higher N(ε)? Why does this mean a higher fractal dimension?

D2. The following melodies are described in words. Rank them from lowest to highest estimated fractal dimension, and justify your ranking: - (a) A C major scale ascending and descending smoothly - (b) The "Ode to Joy" melody from Beethoven's Ninth Symphony - (c) A chromatic scale with random octave jumps inserted at random points - (d) The opening of Chopin's Fantaisie Impromptu (rapid arpeggios with frequent direction changes) - (e) A random walk with steps chosen from {-12, -7, -5, -2, -1, +1, +2, +5, +7, +12} semitones with equal probability

D3. Research has found that different historical periods of Western music have characteristic fractal dimensions in their pitch sequences. Based on your knowledge of music history, predict how fractal dimension might change from (a) Medieval plainchant to (b) Baroque polyphony to (c) Classical sonata to (d) Romantic symphony to (e) twentieth-century serialism. Explain the musical reasons for each change. Does the historical trend support or complicate the idea that music has been moving toward higher fractal complexity over time?

D4. The chapter mentions that heart disease and aging tend to make heartbeat variability less fractal (more regular or more random, but not intermediate). A healthy heart has intermediate fractal complexity. Propose an analogy in music: what is "musically healthy" fractal complexity? Can a piece of music be "too fractal" (too complex, too random) or "not fractal enough" (too repetitive, too predictable)? Give one example of each extreme from actual musical works.

D5. Some music theorists have proposed that musical "genius" correlates with a specific fractal dimension range — that the compositions of Mozart, Bach, and Beethoven cluster around a particular dimension while "lesser" composers deviate from this range. Discuss the methodological problems with this hypothesis. Even if the correlation were found to be statistically significant, would it prove that fractal dimension is the cause of musical greatness? What alternative explanations would need to be ruled out?

Part E: Synthesis and Research

E1. Steve Reich's composition technique of "phasing" (used in works like Piano Phase, 1967) involves two identical repetitive patterns played simultaneously but gradually drifting out of phase with each other. As the patterns move apart, they create a complex texture; as they move back together, they create near-unison. Analyze this technique in terms of self-similarity: what self-similar structures are created when two identical patterns are offset by different fractions of their cycle length? Is the resulting composite pattern more or less self-similar than the original pattern?

E2. The "coastline paradox" says that the measured length of a coastline depends on the scale of measurement: measure with a kilometer ruler and you get one length; measure with a meter ruler and you get a longer length; measure with a centimeter ruler and you get even longer. This happens because finer measurements capture finer features of the coastline's fractal structure. Propose a musical analog of the coastline paradox: a property of a musical work that, when measured at finer and finer scales, gives increasingly large values rather than converging to a fixed number. What does this suggest about the nature of musical complexity as a quantity?

E3. The chapter discusses "sonification" — converting natural fractal data (mountain profiles, coastlines) into musical sequences. Design a complete sonification procedure for the day-by-day temperature record of a city over one year. Specifically: - How would you map temperature values to pitches? (Linear? Logarithmic? Other?) - How would you map the time axis to musical time? (One note per day? Per week?) - How would you handle tempo and rhythm? - What musical "shape" would you expect the result to have, given that weather is approximately 1/f in its long-term statistics?

E4. The concept of multifractal analysis extends fractal dimension from a single number to a spectrum of dimensions, capturing the different scaling behaviors of different parts of a complex system. A multifractal system has regions of high "singularity" (rapid change, high dimension) alternating with regions of low singularity (slow change, low dimension). Propose how multifractal analysis might apply to music: what are the "high singularity" and "low singularity" regions in a musical work? Does Beethoven's music have a more complex multifractal structure than simple folk music? What would this mean perceptually?

E5. Essay: Write a 400–500 word essay addressing the following question: If the 1/f statistical property of music is universal across cultures, does this mean that musical structure is determined by physics and biology rather than by culture and history? What is left for culture to determine, once the 1/f constraint is imposed? Use evidence from the chapter and your own musical knowledge to develop an argument. Your essay should acknowledge the strongest objection to your position and address it.