Dynamics and the Dead

The Grateful Dead were an American rock band formed in 1965 in the San Francisco Bay Area. The band was known for its unique and eclectic style, which fused elements of rock, folk, bluegrass, blues, reggae, country, improvisational jazz, psychedelica, and space rock—and for live performances of long musical improvisation. "Their music," writes Lenny Kaye, "touches on ground that most other groups don't even know exists." These various influences were distilled into a diverse and psychedelic whole that made the Grateful Dead "the pioneering Godfathers of the jam band world." [wikipedia]

Introduction: This project examines dynamics in the music of the Grateful Dead, specifically looking at variations in tempo and loudness, and how these variations contributed to the musical experience. Inspired by the band's progressive taping policy which encouraged audience recordings of the 2000+ live concerts, the data set was taken from the Grateful Dead Archive. The two songs chosen for analysis were Morning Dew and Scarlet Begonias -> Fire on the Mountain (in Grateful Dead nomenclature, a setlist arrow represents a seamless transition jam). The versions of the songs were picked from among my personal favorites spanning many years of the band's career, ones that I (very subjectively) feel best encapsulate different eras of the band.

Materials and Methods: The audio waveforms were imported to Sonic Visualiser and vamp plugins were used to extract tempo and loudness data. Specifically, the VAMP SDK Amplitude Follower was set to monitor the waveform voltage, taking readings every quarter of a second. The Queen Mary University Tempo and Beat Tracker was set to estimate the beats-per-minute over 5-second intervals, based on a hybrid of two algorithms developed by Davies and Ellis. The extracted tempo and loudness data was normalized for a comparable mean and standard deviation, and graphically represented as the Scape Plot method developed by Sapp.

These plots provide a convenient and readable means of data visualization on both small and large scales, illustrated in Figure 1 to the right. The process begins with dividing a performance into a series of beats, in this case with the tempo tracker. Each beat in the series is represented by a “brick” in the horizontal “foundation layer” of the plot. Sequential groupings of beats are then stacked on top of that layer in a pyramid shape. The second layer of bricks contains groups of two beats each, the third layer contains groups of three beats, and so on. As the number of beats covered in each brick increases, fewer bricks are needed in that layer to represent the entire sequence, until one final brick at the top is covering the whole performance.

Plots of the loudness are called dynascapes, and tempo plots are timescapes. By placing them together, it is possible to compare both properties simultaneously. The color scheme of the plots is as follows: green is the average value, red is higher values, and blue is lower values, with the intermediate spectrum of colors between them.

Conclusion: A connection was found between tempo and loudness, with the properties frequently changing together, though not always in the same direction. Note how the colored regions line up vertically on the tempo and loudness plots. Future work would include taking full use of the scape plots' capabilities to examine the correlation between multiple performances.

The University of Southern California does not screen or control the content on this website and thus does not guarantee the accuracy, integrity, or quality of such content. All content on this website is provided by and is the sole responsibility of the person from which such content originated, and such content does not necessarily reflect the opinions of the University administration or the Board of Trustees