Anatomy of a Lick: Structure & Variants, History & Transmission

paper, specified "long paper"
  1. 1. Klaus Frieler

    Hochschule für Musik "Franz Liszt" Weimar (University of Music “Franz Liszt”)

  2. 2. Höger Frank

    Hochschule für Musik "Franz Liszt" Weimar (University of Music “Franz Liszt”)

  3. 3. Pfleiderer Martin

    Hochschule für Musik "Franz Liszt" Weimar (University of Music “Franz Liszt”)

Work text
This plain text was ingested for the purpose of full-text search, not to preserve original formatting or readability. For the most complete copy, refer to the original conference program.

Oral transmission of musical material plays an important role in many African-American music cultures, such as in blues and in jazz (Berliner, 1994). This does not only pertain to entire songs but also to smaller musical units which are often called licks, formulas, or patterns. Due to the importance of improvisation in jazz, there is a certain need to command a personal vocabulary of patterns, which are musical snippets of a few tones. This greatly facilitates the improvisation process, particularly in fast tempos, by reducing the cognitive load. An overall heightened level of virtuosity became common in jazz with bebop in the 1940s (DeVeaux, 1997; Frieler, 2018). In this context, characteristic “bebop lines” were invented by players like Charlie Parker and Dizzy Gillespie, amongst others. Since the musical features those long lines are rarely found in other Western musical styles, they became a token of jazz. Certain patterns and licks are building blocks of those “bebop lines” and, hence, have become important components of jazz improvisation. Therefore, they deserve closer scrutiny, lending itself to the use of computers as it is not easy to discern patterns by listening or analyzing transcriptions manually.

Pattern mining and search
The international Dig That Lick project (DTL) is dedicated to investigating the usage of patterns and licks in monophonic jazz solos using search algorithms on a large database of jazz solo transcriptions. These transcriptions are created automatically using state-of-the-art melody extraction algorithms based on neural networks and advanced signal processing techniques.(Basaran et al., 2018) The transcriptions are equipped with extensive metadata based on a specifically designed semantic model. N-grams, i.e., melodic sub-sequences, are extracted from the transcriptions using pitch and interval representations and stored in a database. Similarity algorithms, which are grounded in music psychological research, are used to retrieve patterns instances for a given query and similarity threshold. Additionally, exact patterns can be extracted using regular expressions. This system allows tracing patterns and its variants across the whole database while combining it with the available metadata to make further inferences.

A case study
To demonstrate the viability of this approach, we present in the following a small case study exploring a typical bebop pattern. The pattern was found with the help of the Pattern History Explorer (Frieler et al., 2018), which contains over 600 interval patterns in over 11,000 instances pre-mined from the Weimar Jazz Database

Publicly available from
(WJD; Pfleiderer, 2017) for exploration.

The chosen interval pattern [‑1,‑2,‑1,3,3,3,‑1,‑2] (measured in semitones, cf. Frieler, 2017) can be considered a typical bebop pattern with a distinctive recognizable structure. The pattern can be found as patterns M20 and M40 in Owens work on Charlie Parker (1974). The pointwise self-information (logarithm of expected to observed frequency) of this pattern is about 11 bits, which means it occurs about 3000 times more often than it could be expected based on a 0th order Markov model, which shows its significance. In order to find variants of the pattern, it was submitted as a query to the DTL similarity search system currently working with the Weimar Jazz Database which contains 456 solo transcriptions by 78 soloists (Pfleiderer, 2017). A similarity threshold of .7 and a maximum length difference of 2 was used. This resulted in a set of 768 similar (including 12 identical) pattern instances. Next, consecutive stretches of instance locations were filtered by using maximum similarity first, length matching and left-most precedence. This filtering left 184 patterns in total. Aural control of sample instances showed, however, that patterns without the ascending seventh chord (or an inversion of it) in the center are usually not perceived to be similar to the query. After filtering these out, a final set of 100 instances was left. The pattern nuclei were classified by the seventh chord they represent, and prefixes and suffixes of the nuclei were frequency ranked. This allowed to construct unique tags of the form
nn-X-mm, where
nn (
mm) is the frequency rank of the prefix (suffix), and X is the nucleus code: D for a diminished chord [3,3,3], D’ for its first inversion [-9,3,3], H for half diminished chord [3, 3, 4,], 7 for a dominant seventh chord [4,3,3] and 7’ for its first inversion [-8,3,3], and m7 for a minor seventh chord [3,4,3].

Pattern structure and variants
Out of 4
3 = 64 possible ascending seventh chords with inversions (i.e., combinations of ascending minor/major thirds and descending minor/major sixths), only six occurred in our result set as nuclei, with the original [3,3,3] being the most common with 70 instances, followed by its first inversion [-9,3,3] with 14 instances. No sixth was found on any other than the first position. Together, this is a first indication for the stability and specificity of the pattern. For the prefixes, 21 different versions could be found with the original [‑1,‑2,‑1] the far most common with 63 instances. The suffixes are more varied with 26 forms and the original [-1,‑2] the most common with 35 instances. A pattern network using Edit Distance-based similarity of all patterns can be found in Fig. 1. Here, a similarity cut-off of .8 was used and node size is chosen proportional to Freeman centrality. The original pattern (01-D-01) is in the center, as expected.

Figure 1. Similarity network the patterns with similarity cut-off of 0.8 for adjacency.

For further structural analysis, we extracted chord contexts, metrical positions, absolute pitch values and chordal diatonic pitch class information (Frieler, 2017) for the first tones of the nuclei. This showed a remarkable consistency. 55% of all nuclei start on a beat, most frequently on the third and the first beat of a 4/4 bar. The most common chordal diatonic pitch class is the third of the chord, whereas the most common chord is a C
7 chord, followed by G
7, D
7 and F
7. Generally, the dominant seventh chord was the most common chord type with 65% of all instances. From these most common traits, a prototypical version of the pattern can be constructed, which, however, cannot be found as such in our results. The closest to such a prototype is an instance of [‑1,‑2,‑1,‑9,3,3,‑1,‑2] depicted in Fig. 1. The only difference to the virtual prototype is the nucleus D’ instead of D.

Figure 2: Nearly prototypical pattern instance [-1,-2,-1,-9,3,3,-1,-2] (01-D’01), found in m. 6 of Charlie Parker's solo on “My Little Suede Shoes” (1951).

Oral transmission
In Fig. 3 a timeline plot of all instances of the pattern variants found in the WJD is depicted. Striking is the number of instances by Charlie Parker, nearly exclusively with a D or D’ nucleus (cf. Fig. 1). Dizzy Gillespie is another heavy and early user, with four instances along in one solo (on “Be-Bop”, 1945). Sonny Rollins, Dexter Gordon, Sonny Stitt and Phil Woods are also fond of this pattern. Interestingly, more recent post-bop players such as Michael Brecker, Chris Potter and Wynton Marsalis have this pattern in their repertoire. However, the pattern variants are not equally popular across the main jazz styles as annotated in the WJD (χ
2(7) = 91.8
p < 0.001), as it is much more likely to be found in bebop and hard-bop solos than in any other styles. This justifies post-hoc the denomination of the pattern as a “bebop lick”. However, the earliest instance can be found with swing tenor sax player Chu Berry (in his solo on “Body Soul”, 1938).

Figure 3: Timeline plot of all instances of all pattern variants according to the recording year of the containing solo, sorted by performer on the y-axis. Labels and colours correspond to nucleus type, point size represents number of instances.

In this case study, we found several interesting results. Firstly, Charlie Parker and Dizzy Gillespie seem to have been, not unexpectedly, the main users and popularisers of this pattern and its variants, even though they themselves might have taken inspiration for it from the earlier swing players. Secondly, many other players from the bebop era, known to be influenced by Parker and Gillespie (Berliner, 1994; DeVeaux, 1991), also used the pattern quite frequently, indicating a direct transmission. Thirdly, modern post-bop players used it also quite often, which is indicative of their mastery of the bebop tradition, though it might also be a direct and deliberate reference to their bebop forebears (e.g., Michael Brecker using it over “Confirmation”, a well-known composition by Charlie Parker). Fourthly, the pattern variants nearly always appear in a specific metrical configuration and certain harmonic contexts, which indicates that metrical and harmonic aspects might be stored along with the pattern in a player’s memory. However, it can also be adapted to different harmonic context without losing its musical shape which opens further questions about pattern construction and memorisation.

This case study demonstrates that computer-based methods are useful to address research questions at the interface of historical, cultural and psychological aspects, leading to new results which could not have been be gained without the help of digital tools. While the case study builds on the rather small sample of 456 solos contained in the WJD the much larger database of jazz improvisations which is currently under development by the DTL project will very likely corroborate the results and provide further insights.


Basaran, D., Essid, S. and Peeters, G. (2018). Main Melody Estimation with Source-Filter NMF and CRNN.
Proceedings of the 19th International Society for Music Information Retrieval Conference (ISMIR), Paris 2018. Paris.

Berliner, P. F. (1994).
Thinking in Jazz. The Infinite Art of Improvisation. Chicago: University of Chicago Press.

DeVeaux, S. (1991). Constructing the Jazz Tradition: Jazz Historiography.
Black American Literature Forum,
25: 525–60.

DeVeaux, S. (1997).
The Birth of Bebop. A Social and Musical History. Berkeley: University of California Press.

Frieler, K. (2017). Computational melody analysis. In Pfleiderer, M., Frieler, K., Abeßer, J., Zaddach, W.-G. and Burkhart, B. (eds),
Inside the Jazzomat. New Perspectives for Jazz Research. Mainz: Schott-Campus, pp. 41–84

Frieler, K. (2018). A Feature History of Jazz Solo Improvisation. In Knauer, W. (ed),
Jazz @ 100. An Alternative to a Story of Heroes, vol. 15. (Darmstadt Studies in Jazz Research /). Hofheim am Taunus: Wolke Verlag.

Frieler, K., Höger, F., Pfleiderer, M. and Dixon, S. (2018). Two web applications for exploring melodic patterns in jazz solos.
Proceedings of the 19th International Society for Music Information Retrieval Conference (ISMIR), Paris 2018. Paris

Owens, T. (1974). Charlie Parker. Techniques of Improvisation Los Angeles: University of California.

Pfleiderer, M. (2017). The Weimar Jazz Database. In Pfleiderer, M., Frieler, K., Abeßer, J., Zaddach, W.-G. and Burkhart, B. (eds),
Inside the Jazzomat. New Perspectives for Jazz Research. Mainz: Schott-Campus, pp. 41–84

If this content appears in violation of your intellectual property rights, or you see errors or omissions, please reach out to Scott B. Weingart to discuss removing or amending the materials.

Conference Info

In review

ADHO - 2019

Hosted at Utrecht University

Utrecht, Netherlands

July 9, 2019 - July 12, 2019

436 works by 1162 authors indexed

Series: ADHO (14)

Organizers: ADHO