ACCORDING to a popular view in jazz research, improvisations are composed of non-overlapping patterns that are retrieved from memory during the improvisation process (Pressing, 1988, 1998). In contrast to emphasizing the importance of pre-learned patterns, Johnson-Laird (2002) claimed that jazz improvisations are primarily based on rules or constraints. In an earlier article, Johnson-Laird (1988) also argued that no one (not even beginners) use pre-learned patterns all the time and claimed that “it is easier […] to make up new melodies than to remember a vast array of motifs and to modify them to fit the chord sequence” (p. 211).
Although musicians may rely more on pre-learned patterns when they are playing in a familiar context and rely more on their harmonic understanding when playing in a less familiar context (Goldman, 2013), patterns and other pre-existing materials have several potential benefits in jazz improvisation. For example, pattern use may allow to allocate attention to other musicians during the improvisation process (Norgaard et al., 2016). An extensive and highly organized storage of pre-learned patterns may also lead to better and faster solutions whereas other pre-existing materials (e.g., using a shared chord progression) may decrease cognitive resources required to select and generate materials for variation (Pressing, 1998). According to classic motor chunking theory, the number of transitions between subsequent actions decreases with the acquisition of motor chunks and leads to faster execution of movement sequences and freed cognitive resources (Schmidt & Wrisberg, 2004; Thompson et al., 2019). Assuming that patterns are learned and retrieved from memory as chunks, these benefits also apply to pattern use.
Several studies have given empirical support to the notion that pre-learned patterns play an important role in jazz improvisation. Arguing that patterns are the building blocks of Charlie Parker’s improvisations, Owens (1974) found 97 recurring melodic patterns that formed 64 pattern categories in his extensive study of about 250 solos by saxophonist Charlie Parker. The proportion of notes that started a recurring pattern at any metrical location or the proportion of notes contained in recurring melodic patterns were not reported. More recently, Weisberg et al. (2004) found 3,395 patterns that contained three or more intervals and occurred at least twice in six solos by Charlie Parker, 797 patterns that contained three or more intervals and occurred at least twice in four solos by saxophonist Lester Young, and 308 patterns that contained three or more intervals and occurred at least twice in only one solo by electric bassist Jaco Pastorius. When considering only 4-interval patterns, the authors found 458 recurring interval patterns in Charlie Parker’s solos, 217 recurring interval patterns in Lester Young’s solos, and 60 recurring interval patterns in Jaco Pastorius’s solo (considering only 3-interval patterns, the number of recurring interval patterns was 385 in Parker’s solos, 255 in Young’s solos, and 84 in Pastorius’s solo). The authors also reported that the average proportion of notes contained in recurring 4-interval patterns was 90% in Charlie Parker’s six solos, 74% in Lester Young’s four solos, and 51% in Jaco Pastorius’s solo (Weisberg et al., 2004). Note that Weisberg and colleagues found a much larger number of recurring melodic patterns compared to Owens (1974) even though they used a much smaller sample of solos.
Most recent studies have supported the view that recurring melodic patterns are extensively used in jazz improvisation. According to Norgaard (2014), 82.6% of notes started a recurring 4-interval pattern at any metrical location and 99.3% of notes were contained in interval patterns that included at least three intervals and occurred at least twice in a sample of forty-eight solos by Charlie Parker. In another study, Stehr (2016) investigated pattern use in six solos by Charlie Parker and found 44 pitch patterns that occurred at least once in all solos. In comparison, Stehr found nine pitch patterns that were played in all four solos by saxophonist Lee Konitz and only one pitch pattern that was played in all three solos by saxophonist Warne Marsh. The number of pitch patterns that occurred at least once in two or more solos was 2,497 in Charlie Parker’s solos, 643 in Lee Konitz’s solos, and 48 in Warne Marsh’s solos. In this study, Stehr searched for pitch patterns with at least five notes.
Frieler (2018) investigated a sample of 56 solos by Charlie Parker and found 5,216 recurring interval patterns and 3,864 recurring pitch patterns. Even when recurring patterns were required to occur in at least six solos, the number of recurring interval patterns was 2,123 and the number of recurring pitch patterns was 2,030. Recurring patterns that contained at least three intervals and occurred in at least two solos covered almost all data (a precise proportion not given), which suggests that Parker’s solos were basically composed of recurring patterns only. Norgaard and Römer (2022) investigated pattern use in the Weimar Jazz Database and found that the proportion of notes that started a recurring 4-interval pattern at any metrical location ranged from 41.6% to 63.4% in saxophone solos by Michael Brecker, Steve Coleman, John Coltrane, David Liebman, Charlie Parker, Sonny Rollins, and Wayne Shorter, as well as trumpet solos by Miles Davis (the minimum number of solos per musician was ten). Recently, Norgaard et al. (2023) investigated pattern use in piano improvisations by Kevin Bales and other jazz pianists. According to this study, the proportion of notes that started a recurring 4-interval pattern at any metrical location was 71.8% in Kevin Bales’s improvisations and 80.5% in the improvisations of the other jazz pianists.
Nurmi (2023) investigated pattern use in Paul Chambers’s and Ron Carter’s bass lines when rhythm and subdivision-level differences in melody were removed and found that 16.8% of recurring 4-note chordal pitch class patterns occurred at least twice in two or more bass lines (which means that 83.2% of recurring 4-note chordal pitch class patterns occurred at least twice in only one bass line) and covered 41.2% of data in Paul Chambers’s bass lines. In Ron Carter’s bass lines, 11.2% of recurring 4-note chordal pitch class patterns occurred at least twice in two or more bass lines and covered 13.4% of data. A chordal pitch class pattern refers to a sequence of notes with octave equivalence coded in reference to the current chord (e.g., C, E and G are coded as the root, a major third, and a perfect fifth when the current chord is a C major). Subdivision-level differences in melody refer to “any difference between melodic patterns that still shared the same underlying interval structure at the beat level” (Nurmi, 2023, p. 117). Since removing these differences may sometimes lead to excluding a large proportion of notes in bass lines at slow tempos, I preferred to use the term ‘subdivision-level differences in melody’ instead of the term ‘small differences in melody’ (used in Nurmi, 2023) in this article.
Although the results from Nurmi (2023) indicate that pre-learned patterns were used to a lesser extent in Paul Chambers’s and Ron Carter’s bass lines compared to results from earlier studies (all of which have focused on solos), several factors preclude the direct comparison between these results and results from earlier studies. These include instrument (e.g., double bass vs. saxophone), differences in the length of analyzed bass lines or solos, and not using the same method of measurement to investigate pattern use in bass lines and solos. The proportion of notes that started a recurring pattern at any metrical location (which is often used to measure pattern use in jazz solos) was reported in Nurmi (2023), but overlapping patterns were not effectively removed in this study except when calculating the average length of recurring melodic patterns. It is also important to note that the number of times a specific melodic pattern occurred in a particular bass line was calculated using the following procedure: (1) the number of occurrences of each melodic pattern was calculated separately in each harmonic rhythm category (one chord per bar harmonic rhythm, one chord per two bars harmonic rhythm, and two chords per bar harmonic rhythm) and then (2) the number of times each melodic pattern occurred in each harmonic rhythm category was summed to find out how many times each melodic pattern occurred in the same bass line. A limitation of this calculation method is that it may underestimate the proportion of recurring melodic patterns.
In Nurmi (2023), the main aim of the study was to investigate pattern use and the relationship between time constraints (operationalized as tempo and harmonic rhythm) and the repetition and entropy of melodic patterns. Harmonic rhythm was not taken into account when investigating pattern use to avoid substantially increasing the number of analyses. Although harmonic rhythm influences both bass playing and soloing, there are no studies that have investigated whether harmonic rhythm influences the extent of pattern use. As a difference between analysis of bass lines and solos, bass lines allow to use a mechanical approach to group successive notes into patterns where the first note of each bar is always the first note of a pattern (see Nurmi, 2023, pp. 110-113), which is not an appropriate grouping method with solos (Nurmi, 2023). Instead, the proportion of notes that started a recurring interval pattern at any metrical location (Norgaard, 2014; Norgaard & Römer, 2022; Norgaard et al., 2023) and the proportion of elements contained in recurring patterns at any metrical location (Frieler, 2018; Norgaard, 2014; Weisberg et al., 2004) can be used to analyze pattern use in both solos and bass lines. In contrast to chordal pitch class patterns (where each note is coded in reference to the current chord), the current chord is not taken into account with interval patterns (interval patterns only contain information about the number of semitones between each pair of notes).
The aim of the current study was to investigate the extent of pattern use in Paul Chambers’s and Ron Carter’s bass lines from the 1950s and 1960s and to provide an analysis of methods used in earlier studies. This study extends on my previous work (Nurmi, 2023) by examining basically the same data (one bar from the original data was removed) using a revised method and additional measures. The research questions were: (1) What is the total number of recurring melodic patterns that occurred at least twice in either one or more bass lines or in two or more bass lines by the same musician? (2) What is the proportion of recurring melodic patterns that occurred at least twice in two or more bass lines by the same musician? (3) What is the proportion of data covered by melodic patterns that occurred at least twice in two or more bass lines by the same musician? (4) To what extent do the results differ depending on whether overlapping patterns were removed or not, whether pattern coverage (i.e., the proportion of elements contained in recurring patterns at any metrical location) was used as a measure of pattern use or not, whether the research material was either combined into a single data set or analyzed separately, and whether the metrical location of notes was disregarded or not?
Except for Nurmi (2023), previous studies on pattern use in jazz improvisation have considered solos only. The current results contribute to research on pattern use in jazz improvisation by analyzing bass lines (the role of which is usually to keep time and add low notes to harmony) instead of solos and by providing an analysis of methods used in previous studies. The results are relevant for understanding the relative importance between pre-learned patterns (Pressing, 1988, 1998) and constraints (Johnson-Laird, 1988, 2002) in the creativity of eminent jazz musicians. The results also add to existing research on similarities and differences between language and music (for a review, see Temperley, 2022).
The current research was conducted by analyzing bass line transcriptions from two eminent jazz bassists. The bassists were selected based on their participation in critically acclaimed and highly well-known jazz albums and their influence on the history of jazz. Using the same criteria, a number of other eminent bassists could have been selected as well (e.g., Ray Brown, Oscar Pettiford, and Sam Jones, to name a few). The research material consisted of thirty bass lines by Paul Chambers and twelve bass lines by Ron Carter. Paul Chambers (1935-1969) recorded more than three hundred albums during his unfortunately short career, including many highly well-known and critically acclaimed classic albums such as Miles Davis’s Kind of Blue (1959) and John Coltrane’s Giant Steps (1959) (Palmer, 2012). Ron Carter (b. 1937), most famous for his recordings with the Miles Davis Quintet between 1963 and 1968, has received numerous awards and has played in more than 2,000 recordings including a number of jazz classics (Ouellette, 2013).
Paul Chambers’s bass lines were recorded between 1956 and 1964 and included 6,745 bars of transcriptions (27,750 notes before the reduction process and 26,980 notes after the reduction process). Ron Carter’s bass lines were recorded between 1961 and 1968 and included 2,589 bars of transcriptions (11,502 notes before the reduction process and 10,356 notes after the reduction process). Except for one excluded bar from Paul Chambers’s bass line for So What, the research material was the same as in Nurmi (2023). The average duration of analyzed bass lines was 4 minutes and 39 seconds (median: 3 minutes and 58 seconds) in Paul Chambers’s bass lines and 4 minutes and 42 seconds (median: 4 minutes and 16 seconds) in Ron Carter’s bass lines. In comparison, the median duration of transcribed solos in the Weimar Jazz Database is 1 minute and 27 seconds (Pfleiderer, 2017), which highlights one of the benefits of analyzing bass lines instead of solos: bass lines offer much more data from each recording. Selected bass lines varied in their rhythmic and harmonic complexity, from straight-ahead walking bass lines (usually based on quarter notes with little rhythmic variation and little use of harmonic freedom) to rhythmically adventurous broken time bass lines (which emphasize rhythmically free playing and a high level of harmonic freedom). For a list of transcribed bass lines used in this study, see Nurmi (2023, pp. 98-99).
Seventeen out of the forty-two bass lines were transcribed by the author. The other transcriptions were collected from various sources (Benning, 2018-2020; Bierma, 2019; Fink, 2009-2012; Gregor, 2018; Herridge, 2020; Jazz Bass Transcriptions, 2021; Kolarczyk, 2013; Nabuurs, 2017; Skinner, 2016). The following sheet music was used for chord progressions in bass lines transcribed by the author: Hal Leonard (2004), Sher Music (1988, 1991, 1995), Waters (2011) and Wetzel (1996). To ensure reliability of the research material, all bass lines were checked for accuracy at least twice. Bass solos, bars including inaudible notes, pre-composed material or parts of bass lines based on repeated riffs (e.g., Paul Chambers’s bass line for the head section in So What), and parts of bass lines that primarily contained half notes (e.g., Ron Carter’s bass line for the head section in E.S.P.) were excluded using the procedure proposed by Nurmi (2023, p. 100). The length of analyzed bass lines by Paul Chambers ranged from 95 to 465 bars (M = 224.83, SD = 96.88) and the length of analyzed bass lines by Ron Carter ranged from 104 to 310 bars (M = 215.75, SD = 69.13).
To compensate for the small number of analyzed bass lines by the same musician, rhythm and subdivision-level differences in melody were removed with the reduction procedure used in Nurmi (2023, p. 114) with one modification: with double-stops including the root (as the lower voice) and the perfect fifth of the chord (as the upper voice), the upper voice was disregarded instead of the lower voice. After the reduction process (which converts all note durations to quarter notes based on the metrical location of notes) (see Nurmi, 2023, p. 114), all bass lines were coded as strings of chordal pitch classes (where melodic direction, interval size, and harmonic context are considered) and strings of intervals (where only melodic direction and interval size are considered).
It is important to note that patterns can be modified during the improvisation process (Frieler et al., 2018; Norgaard et al., 2023; Owens, 1974). If variations of pre-learned ideas could not be created during the improvisation process, jazz musicians would be required to learn a huge storage of patterns to avoid excessive repetition of patterns (see Christensen et al., 2019). As a result, the number of patterns required to improvise fluently would be much greater if pre-learned patterns could not be modified during the improvisation process. Although most studies on pattern use in jazz improvisation are based on analyzing interval patterns (which can be transposed to a different pitch level without changing the interval structure and therefore do not require exact replication in terms of pitch), reductions allow more room for variation by allowing patterns with different notes at the subdivision-level to be categorized as the same pattern as long as they share the same notes at the beat-level.
The aim of the current study was to investigate (1) to what extent Paul Chambers and Ron Carter repeated the same chordal pitch class patterns and interval patterns in their bass lines, and (2) to what extent the results differ depending on whether overlapping patterns were removed or not, whether pattern coverage (the proportion of elements contained in recurring patterns at any metrical location) was used as a measure of pattern use or not, whether all bass lines were combined into a single data set or analyzed separately, and whether the metrical location of notes was disregarded or not.
To achieve the first aim of the study, the following questions were answered: (1) What is the total number of recurring melodic patterns that occurred at least twice in either one or more bass lines or two or more bass lines by the same musician? (2) What is the proportion of recurring melodic patterns that occurred at least twice in two or more bass lines by the same musician? (3) What is the proportion of data covered by melodic patterns that occurred at least twice in two or more bass lines by the same musician? In addition, the number of different chordal pitch classes and intervals was reported to find out how many of all possible chordal pitch classes or intervals were actually used in analyzed bass lines (which was considered to provide useful basic statistics of the data). In each of these analyses, all patterns were required to start at the first beat of a bar. The analysis was restricted to chordal pitch class patterns with two to four notes and interval patterns with one to three intervals. The identification of chordal pitch class patterns was performed manually. MeloSpyGUI (Abeßer et al., 2018) was used to identify interval patterns from MIDI files. Rhythm and subdivision-level differences in melody were removed using the reduction process outlined earlier. For the rest of this article, the term ‘bass line reduction’ is used to refer to bass lines where rhythm and subdivision-level differences in melody were removed.
The total number of recurring melodic patterns that occurred at least twice in either one or more bass lines or two or more bass lines by the same musician was calculated using a pattern calculator for Python programmed by the author (also used when calculating the proportion of recurring melodic patterns and the proportion of data covered by recurring melodic patterns).
The proportion of recurring melodic patterns that occurred at least twice in two or more bass line reductions was calculated according to the formula below, where ‘a’ is the total number of different melodic patterns that occurred at least twice in two or more bass line reductions by the same musician and ‘b’ is the total number of different melodic patterns that occurred at least twice in one or more bass line reductions by the same musician. Note that ‘a’ is a subset of ‘b’.
$$f_{a} = \frac{a}{b} \times 100$$
The proportion of data covered by melodic patterns that occurred at least twice in two or more bass line reductions was calculated according to the following formula, where ‘c’ is the total number of instances of melodic patterns that occurred at least twice in two or more bass line reductions by the same musician and ‘d’ is the total number of instances of melodic patterns in all bass line reductions by the same musician. Note that ‘c’ is a subset of ‘d’.
$$f_{b} = \frac{c}{d} \times 100$$
To calculate the probability that equal or greater results could have occurred by chance, a randomization test was conducted using Python code written by the author. All melodic patterns that occurred in any bass line reduction by the same musician were included in a list of possible values. New bass lines were then created by choosing random patterns from this list (each pattern was allowed to occur multiple times). The length of each randomly generated bass line was the same as that of the original bass line reduction. For example, if a bass line reduction contained one hundred 4-note melodic patterns, a randomly generated bass line also contained one hundred 4-note melodic patterns. When all randomly generated bass lines were created, the program calculated the proportion of recurring melodic patterns that occurred at least twice in two or more bass lines by the same musician and the proportion of data covered by melodic patterns that occurred at least twice in two or more bass lines by the same musician. This test was repeated 5,000 times. After all tests were performed, the program calculated (1) the proportion of replicates that gave an equal or larger proportion of recurring melodic patterns that occurred at least twice in two or more randomly generated bass lines compared to the observed results and (2) the proportion of replicates that gave an equal or larger proportion of data covered by melodic patterns that occurred at least twice in two or more randomly generated bass lines compared to the observed results.
The second aim of the study was to investigate how the choice of methods may have influenced the results of previous studies. To achieve this aim, I examined to what extent the results differed depending on whether overlapping patterns were removed or not, whether pattern coverage was used as a measure of pattern use or not, whether the research material was combined into a single data set or analyzed separately, and whether the metrical location of notes was disregarded or not. In each analysis, rhythm and subdivision-level differences in melody were removed using the reduction process outlined earlier. The analysis was restricted to interval patterns only. Most previous studies have used either the proportion of notes that started a recurring interval pattern at any metrical location (Norgaard, 2014; Norgaard & Römer, 2022; Norgaard et al., 2023; Nurmi, 2023) or the proportion of elements contained in recurring patterns at any metrical location (Frieler, 2018; Norgaard, 2014; Weisberg et al., 2004) to investigate pattern use in jazz improvisation. Despite these similarities, there are various factors (e.g., whether the metrical location of notes was disregarded or not) which may influence the results.
The consequences of overlapping patterns were examined by comparing the average proportion of notes that started an interval pattern that occurred at least twice at any metrical location in the same bass line reduction (1) using the database mode in MeloSpyGUI (in which case overlapping patterns were not removed) and (2) the partition mode in MeloSpyGUI (in which case overlapping patterns were removed) using the following settings: the minimum length of patterns was three intervals and the minimum number of occurrences was two. All bass line reductions were analyzed separately, and the results were then averaged (i.e., the number of sources was one). All duplicates from the metrical position list were removed, which means that each note was allowed to start only a single pattern. If these duplicates are not removed, the proportion of notes that started a recurring pattern at any metrical location could exceed 100%. After removing duplicates from the metrical position list, all patterns contained three intervals.
The consequences of using pattern coverage (the proportion of elements contained in recurring patterns at any metrical location) as a measure of pattern use were examined by comparing pattern coverage (where each note was allowed to start one or more patterns and the metrical location of notes was disregarded) with the proportion of data covered by patterns that occurred at least twice in one or more bass line reductions or at least once in two more bass line reductions (where each note was only allowed to start a single pattern and all patterns were required to start at the first beat of a bar). Overlapping patterns were removed in both analyses and had no effect on the results. Pattern coverage was calculated using the partition mode in MeloSpyGUI (see Frieler, 2017, p. 83). When the number of sources was one, pattern coverage was calculated separately for each bass line reduction. The proportion of data covered by recurring patterns was calculated using Python code programmed by the author.
The consequences of combining all bass line reductions into a single data set were examined by comparing (1) the proportion of notes that started a recurring 3-interval pattern at any metrical location when all bass line reductions by the same musician were combined into a single data set and overlapping patterns were not removed (calculated using the database mode in MeloSpyGUI), (2) the proportion of notes that started a recurring pattern at any metrical location and contained three or more intervals when all bass line reductions by the same musician were combined into a single data set and overlapping patterns were removed (using the partition mode in MeloSpyGUI), (3) the average proportion of notes that started a recurring 3-interval pattern at any metrical location (where all bass line reductions by the same musician were analyzed separately and overlapping patterns were not removed; using the database mode in MeloSpyGUI), and (4) the average proportion of notes that started a recurring pattern at any metrical location and contained three or more intervals (where all bass line reductions by the same musician were analyzed separately and overlapping patterns were removed; using the partition mode in MeloSpyGUI). In each case, each note was only allowed to start a single pattern (all duplicates from the metrical position list were removed).
The consequences of disregarding the metrical location of notes were examined by calculating the proportion of notes that started one or more patterns that occurred at least once in two or more bass line reductions at any metrical location (1) when the metrical location of notes was disregarded (which means that instances of the same pattern were allowed to start from different metrical locations) and each note was allowed to start one or more patterns and (2) when patterns were considered to be similar only if they shared the same sequence of intervals and started from the same metrical location. These analyses were performed using the partition mode in MeloSpyGUI and Python code written by the author (used to calculate the proportion of notes that started one or more patterns at any metrical location when all instances of the same pattern were required to start from the same metrical location). In each analysis, overlapping patterns were removed, the minimum length of patterns was three intervals, the minimum number of occurrences was two, and the minimum number of sources was two. When all instances of the same pattern were required to start from the same metrical location, all values in the metrical position and value columns (in the results file produced with MeloSpyGUI) were combined into a single column, which allows to take into account both variables at once. It is important to note that the partition mode only requires that recurring patterns occur at least once in two or more musical works when the minimum number of sources is two. Similarly, when the minimum number of occurrences is three, the partition mode gives a list of patterns that occurred at least three times in all musical works together (which means that a pattern does not have to occur at least three times in any single musical work).
I also compared the total number of different interval patterns that occurred at least once in two or more bass line reductions (1) when overlapping patterns were removed and all instances of the same pattern were required to start from the same metrical location and (2) when overlapping patterns were removed but all instances of the same pattern were not required to start from the same metrical location with (3) the total number of different interval patterns that occurred at least twice in two or more bass line reductions when there were no overlapping patterns and all patterns were required to start from the first beat of a bar. In each analysis, the minimum number of occurrences was either one or two, the minimum number of sources was two, and pattern length was either three intervals or three or more intervals. The total number of different recurring interval patterns was calculated using the partition mode in MeloSpyGUI (in which case, all duplicates from the values column were removed) and Python code programmed by the author (used when all patterns were required to start from the first beat of a bar and when all instances of the same pattern were required to start from the same metrical location).
With chordal pitch class patterns (where each note was coded in reference to the current chord), five possible relationships between any pair of notes and their melodic direction were considered: repetition of the same note (twelve possible relationships between any pair of notes in chromatic scale), ascending or descending melodic direction where the interval between two subsequent notes was less than one octave (twelve possible relationships with ascending melodic direction and twelve possible relationships with descending melodic direction), and ascending or descending melodic direction where the interval between two subsequent notes was more than one octave (twelve possible relationships with ascending melodic direction and twelve possible relationships with descending melodic direction). With chordal pitch class patterns, intervals larger than two octaves were not distinguished from intervals larger than one octave. As a result, the number of possible chordal pitch classes is 60 (which could be used to generate 604 = 12,960,000 different 4-note chordal pitch class patterns). The number of different chordal pitch classes that were actually used in Paul Chambers’s bass line reductions was 50 (ascending and descending diminished twelfth, ascending and descending minor thirteenth, ascending major thirteenth, ascending major fourteenth, descending minor ninth, descending minor tenth, descending perfect eleventh, and descending minor fourteenth were not used in Paul Chambers’s bass line reductions) and 59 in Ron Carter’s bass line reductions (the only chordal pitch class not used in any bass line reduction by Ron Carter was descending minor fourteenth).
With interval patterns (which present the number of semitones between each pair of notes), all possible relationships between any pair of notes and their melodic direction were considered: repetition of the same note, ascending or descending melodic direction where the interval between two subsequent notes was less than one octave, ascending or descending melodic direction where the interval between two subsequent notes was more than one octave but less than two octaves, and so on. When considering intervals with 36 half steps at maximum, the number of possible intervals is 73 (36 possible ascending intervals, 36 possible descending intervals, and the repetition of the same note), which could be used to generate 733 = 389,017 different 3-interval patterns. The number of different intervals actually used was 43 in Paul Chambers’s bass line reductions (the size of the largest ascending or descending interval was 27 half steps) and 51 in Ron Carter’s bass line reductions (the size of the largest ascending or descending interval was 31 half steps).
In Paul Chambers’s bass line reductions, 29.42% of recurring 4-note chordal pitch class patterns occurred at least twice in two or more bass line reductions and covered 46.69% of data. In Ron Carter’s bass line reductions, 17.89% of recurring 4-note chordal pitch class patterns occurred at least twice in two or more bass line reductions and covered 18.54% of data. Both the proportion of recurring melodic patterns that occurred at least twice in two or more bass line reductions and the proportion of data covered by melodic patterns that occurred at least twice in two or more bass line reductions were greater when analyzing interval patterns compared to when analyzing chordal pitch class patterns. In Paul Chambers’s bass line reductions, 43.69% of recurring 3-interval patterns occurred at least twice in two or more bass lines and covered 62.98% of data. In Ron Carter’s bass line reductions, 30.04% of recurring 3-interval patterns occurred at least twice in two or more bass lines and covered 31.05% of data. All results from analyses of 4-note chordal pitch class patterns and 3-interval patterns were equal or greater than expected by chance (p < .001). See Tables 1 and 2 for full results.
These results differ from Nurmi (2023), although both studies used basically the same research material (the only difference was that one bar from the original research material was excluded in the current study). The difference between the results was caused by the fact that Nurmi (2023) used a calculation method in which the number of times a specific melodic pattern occurred in a bass line reduction was first calculated separately in each harmonic rhythm category, and then the number of times each melodic pattern occurred in each harmonic rhythm category was summed to determine how many times each melodic pattern occurred in the same bass line reduction. The problem with this calculation method is that it can underestimate the proportion of recurring melodic patterns.
Both the proportion of recurring melodic patterns that occurred at least twice in two or more bass line reductions and the proportion of data covered by melodic patterns that occurred at least twice in two or more bass line reductions increased as pattern length decreased (see Figures 1 and 2). The proportion of recurring 2-note chordal pitch class patterns and the proportion of recurring 1-interval patterns ranged from 51.39% to 82.76%, indicating that a large proportion of patterns started with a limited number of note choices. Also note that 56.31% to 70.58% of recurring patterns with four notes or three intervals occurred at least twice in only one bass line reduction by Paul Chambers, and 69.96% to 82.11% of patterns with four notes or three intervals occurred at least twice in only one bass line reduction by Ron Carter, depending on the type of melodic pattern. This shows that most recurring melodic patterns occurred in only one bass line reduction rather than across different bass line reductions and supports Frieler’s (2014) finding that recurring melodic patterns usually occurred in the same solo rather than across several solos.
Table 1. Total number and proportion of recurring chordal pitch class patterns.
| 4 notes a | 3 notes a | 2 notes a | 4 notes b | 3 notes b | 2 notes b | |
|---|---|---|---|---|---|---|
| patterns 1 | 183 | 167 | 98 | 44 | 71 | 74 |
| patterns 2 | 622 | 372 | 130 | 246 | 229 | 144 |
| instances 1 | 3149 | 4726 | 6082 | 480 | 1088 | 1972 |
| instances 2 | 6745 | 6745 | 6745 | 2589 | 2589 | 2589 |
| % patterns | 29.42% (70.58%) p < .001 |
44.89% (55.11%) p = .013 |
75.38% (24.62%) p > .99 |
17.89% (82.11%) p < .001 |
31.00% (69.00%) p < .001 |
51.39% (48.61%) p > .99 |
| % data |
46.69% p < .001 |
70.07% p < .001 |
90.17% p < .001 |
18.54% p < .001 |
42.02% p < .001 |
76.17% p < .001 |
Note. patterns 1 = total number of different chordal pitch class patterns that occurred at least twice in two or more bass line reductions by the same musician; patterns 2 = total number of different chordal pitch class patterns that occurred at least twice in one or more bass line reduction by the same musician; instances 1 = total number of instances of chordal pitch class patterns that occurred at least twice in two or more bass line reductions by the same musician; instances 2 = total number of instances of chordal pitch class patterns in all bass line reductions by the same musician; % patterns = proportion of recurring chordal pitch class patterns that occurred at least twice in two or more bass line reductions by the same musician (proportion of recurring chordal pitch class patterns that occurred at least twice in only one bass line reduction by the same musician is shown in brackets); % data = proportion of data covered by chordal pitch class patterns that occurred at least twice in two or more bass line reductions by the same musician. a Paul Chambers’s bass line reductions. b Ron Carter’s bass line reductions.
Table 2. Total number and proportion of recurring interval patterns.
| 3 intervals a | 2 intervals a | 1 interval a | 3 intervals b | 2 intervals b | 1 interval b | |
|---|---|---|---|---|---|---|
| patterns 1 | 187 | 110 | 24 | 73 | 78 | 22 |
| patterns 2 | 428 | 174 | 29 | 243 | 162 | 36 |
| instances 1 | 4248 | 5724 | 6644 | 804 | 1723 | 2502 |
| instances 2 | 6745 | 6745 | 6745 | 2589 | 2589 | 2589 |
% patterns |
43.69% (56.31%) p < .001 |
63.22% (36.78%) p > .99 |
82.76% (17.24%) p > .99 |
30.04% (69.96%) p < .001 |
48.15% (51.85%) p = .92 |
61.11% (38.89%) p > .99 |
| % data | 62.98% p < .001 |
84.86% p < .001 |
98.50% p > .99 |
31.05% p < .001 |
66.55% p < .001 |
96.64% p > .99 |
Note. patterns 1 = total number of different interval patterns that occurred at least twice in two or more bass line reductions by the same musician; patterns 2 = total number of different interval patterns that occurred at least twice in one or more bass line reduction by the same musician; instances 1 = total number of instances of interval patterns that occurred at least twice in two or more bass line reductions by the same musician; instances 2 = total number of instances of interval patterns in all bass line reductions by the same musician; % patterns = proportion of recurring interval patterns that occurred at least twice in two or more bass line reductions by the same musician (proportion of recurring interval patterns that occurred at least twice in only one bass line reduction by the same musician is shown in brackets); % data = proportion of data covered by interval patterns that occurred at least twice in two or more bass line reductions by the same musician. a Paul Chambers’s bass line reductions. b Ron Carter’s bass line reductions.
Fig. 1. Effects of pattern length on proportion of recurring patterns in Paul Chambers’s bass line reductions: (a) chordal pitch class patterns and (b) interval patterns. % patterns = proportion of recurring melodic patterns that occurred at least twice in two or more bass line reductions by the same musician; % data = proportion of data covered by melodic patterns that occurred at least twice in two or more bass line reductions by the same musician.
Fig. 2. Effects of pattern length on proportion of recurring patterns in Ron Carter’s bass line reductions: (a) chordal pitch class patterns and (b) interval patterns. % patterns = proportion of recurring melodic patterns that occurred at least twice in two or more bass line reductions by the same musician; % data = proportion of data covered by melodic patterns that occurred at least twice in two or more bass line reductions by the same musician.
The average proportion of notes that started a 3-interval pattern occurring at least twice at any metrical location in the same bass line reduction was about 17 to 36 percentage points greater when overlapping patterns were not removed compared to when they were removed (see Table 3).
Table 3. Consequences of not removing overlapping patterns.
| % notes a | range + SD a | % notes b | range + SD b | |
|---|---|---|---|---|
| overlapping patterns not removed | 76.07% | min.: 58.28% max.: 90.54% SD = 8.79 |
54.53% | min.: 29.09% max.: 72.64% SD = 12.42 |
| overlapping patterns removed | 39.93% | min.: 26.39% max.: 49.85% SD = 5.82 |
37.58% | min.: 22.36% max.: 49.83% SD = 7.39 |
Note. % notes = average proportion of notes that started a 3-interval pattern that occurred at least twice at any metrical location in the same bass line reduction. a = Paul Chambers’s bass line reductions. b = Ron Carter’s bass line reductions.
Pattern coverage analysis indicated a much greater extent of pattern use compared with results based on the average proportion of data covered by recurring patterns that started on the first beat of a bar. The average pattern coverage (when the number of sources was one) gave similar results to the proportion of data covered by recurring interval patterns that started on the first beat of a bar (see Table 4).
Table 4. Consequences of using pattern coverage as a measure of pattern use.
| Paul Chambers’s bass line reductions | Ron Carter’s bass line reductions | |
|---|---|---|
average pattern coverage (number of sources: 1) |
90.54% (79.50% to 97.50%) SD = 4.61 |
76.70% (54.00% to 90.00%) SD = 9.77 |
average pattern coverage (minimum number of sources: 1) |
99.21% (97.60% to 100%) SD = 0.67 |
95.17% (89.20% to 98.40%) SD = 3.00 |
average pattern coverage (minimum number of sources: 2) |
98.68% (95.70% to 99.80%) SD = 1.05 |
92.18% (84.80% to 97.70%) SD = 4.43 |
average proportion of data (number of sources: 1) |
71.72% (48.42% to 90.69%) SD = 10.87 |
48.99% (22.12% to 64.02%) SD = 12.69 |
average proportion of data (number of sources: 2) |
61.76% (42.86% to 82.15%) SD = 10.06 |
29.73% (10.58% to 47.47%) SD = 10.27 |
| proportion of data (min. sources: 1) | 92.10% | 71.42% |
Note. Average pattern coverage = average proportion of elements contained in patterns that occurred at least twice at any metrical location and contained three or more intervals; average proportion of data = average proportion of data covered by 3-interval patterns that started at the first beat of a bar and occurred at least twice in one or more bass line reductions; proportion of data = proportion of data covered by 3-interval patterns that started at the first beat of a bar and occurred at least twice in the same bass line reduction or at least once in two or more bass line reductions. Range is shown in brackets.
A potential explanation for the difference between these results is that pattern coverage does not distinguish between patterns that share the same intervals but start from a different metrical location, and it also allows each note to start any number of patterns. Overlapping patterns were removed in all analyses and had no effect on the results. However, these results should be considered with caution because the methods of measurement did not use the same pattern length (3-interval patterns were compared with patterns that contained three or more intervals) or, in some cases, the same number of sources (e.g., data where the minimum number of sources was two was compared with data where the number of sources was two). Another issue is that patterns allowed to start at any metrical location were compared with patterns that were required to start on the first beat of a bar.
When overlapping patterns were not removed, the proportion of notes that started an interval pattern that occurred at least twice at any metrical location was about 19 to 26 percentage points greater when all bass line reductions were combined into a single data set compared to when each bass line reduction was first analyzed separately and the results were then averaged. When overlapping patterns were removed, the proportion of notes that started an interval pattern that occurred at least twice at any metrical location was about 30 percentage points greater when all bass line reductions were combined into a single data set compared to when each bass line reduction was first analyzed separately and the results were then averaged. See Table 5 for full results.
These results indicate that combining all musical works into a single data set may overestimate the proportion of recurring patterns simply because recurring patterns are only required to occur either at least once in two or more musical works or at least twice in the same musical work (instead of at least twice in the same musical work or at least twice in two or more musical works) when all musical works are combined into a single data set. Note also that the plausibility that a pattern was pre-learned and retrieved from memory during the improvisation process increases with the number of times a pattern occurs in the same musical work, the number of times a pattern occurs across different musical works, and the length of the recurring pattern (Nurmi, 2023, p. 113; see also Frieler, 2020, p. 139). If a short pattern occurs only once in two or more musical works, there is little evidence to support the claim that this pattern was pre-learned and retrieved from memory during the improvisation process.
Table 5. Consequences of combining all bass line reductions by the same musician into a single data set.
| % notes (three intervals) | % notes (three or more intervals) | |||
|---|---|---|---|---|
| combined into a single data set (overlapping patterns not removed) | bass line reductions analyzed separately (overlapping patterns not removed) | combined into a single data set (overlapping patterns removed) | bass line reductions analyzed separately (overlapping patterns removed) | |
| Paul Chambers | 95.14% | 76.07% (58.28% to 90.54%) SD = 8.79 |
n/a * | 39.93% (26.39% to 49.85%) SD = 5.82 |
| Ron Carter | 80.59% | 54.53% (29.09% to 72.64%) SD = 12.42 |
67.75% | 37.58% (22.36% to 49.83%) SD = 7.39 |
Note. % notes (three intervals) = proportion of notes that started a 3-interval pattern that occurred at least once in two or more bass line reductions or at least twice in the same bass line reduction at any metrical location; % notes (three or more intervals) = proportion of notes that started a pattern that occurred at least once in two or more bass line reductions or at least twice in the same bass line reduction at any metrical location and contained three or more intervals; * = due to high computational demands of this analysis, Paul Chambers’s bass line reductions were not analyzed. Range is shown in brackets.
When overlapping patterns were removed and the metrical location of notes was disregarded, the proportion of notes that started one or more patterns that occurred at least once in two or more bass line reductions at any metrical location was about 11 to 18 percentage points greater compared to when overlapping patterns were removed and all instances of the same pattern were required to start from the same metrical location. Disregarding the metrical location of notes also increased the number of different interval patterns that occurred at least once in two or more bass line reductions. Overlapping patterns were removed in all of these analyses and had no effect on the results. See Table 6 for full results.
Table 6. Consequences of disregarding the metrical location of notes.
| Paul Chambers’s bass line reductions | Ron Carter’s bass line reductions | |
|---|---|---|
| % notes (metrical location of notes disregarded) ** | 55.15% | 60.81% |
| % notes (all instances of the same pattern were required to start from the same metrical location) ** | 44.44% | 42.83% |
| pattern count (metrical location of notes disregarded) ** | 4,965 different patterns that occurred at least once in two or more bass line reductions and contained at least three intervals | 2,038 different patterns that occurred at least once in two or more bass line reductions and contained at least three intervals |
| pattern count (all instances of the same pattern were required to start from the same metrical location) ** | 3,634 different patterns that occurred at least once in two or more bass line reductions and contained at least three intervals | 1,412 different patterns that occurred at least once in two or more bass line reductions and contained at least three intervals |
| pattern count (metrical location of notes disregarded) ** | 885 different 3-interval patterns that occurred at least once in two or more bass line reductions | 884 different 3-interval patterns that occurred at least once in two or more bass line reductions |
| pattern count (all instances of the same pattern were required to start from the same metrical location) ** | 604 different 3-interval patterns that occurred at least once in two or more bass line reductions | 599 different 3-interval patterns that occurred at least once in two or more bass line reductions |
| pattern count (all patterns were required to start from the first beat of a bar) * | 428 different 3-interval patterns that occurred at least twice in one or more bass line reductions | 243 different 3-interval patterns that occurred at least twice in one or more bass line reductions |
| pattern count (all patterns were required to start from the first beat of a bar) ** | 187 different 3-interval patterns that occurred at least twice in two or more bass line reductions | 73 different 3-interval patterns that occurred at least twice in two or more bass line reductions |
Note. % notes = proportion of notes that started one or more patterns that occurred at least once in two or more bass line reductions at any metrical location and contained three or more intervals; pattern count = number of different patterns; * = minimum number of sources: one; ** = minimum number of sources: two.
Although most earlier studies have removed overlapping patterns, the metrical location of notes has been disregarded in nearly all studies (i.e., these studies have not distinguished between patterns that share the same notes or intervals but start from a different metrical location). Changing the metrical placement of notes can affect the perception of melodies to the extent that patterns which share the same notes and rhythm but differ in the metrical placement of these notes and rhythm values are not perceived as the same pattern (Acevedo et al., 2014; Povel & Essens, 1985; Sloboda, 1983). As a result, if two note sequences share the same sequence of intervals or chordal pitch classes but they start from a different metrical location, they should not be considered as the same pattern.
The results obtained when using different methods are also presented visually in Figures 3 to 7.
Fig. 3. Consequences of not removing overlapping patterns. % notes = average proportion of notes that started a 3-interval pattern that occurred at least twice at any metrical location in the same bass line reduction.
Fig. 4. Consequences of using pattern coverage as a measure of pattern use. Average pattern coverage = average proportion of elements contained in patterns that occurred at least twice at any metrical location and contained three or more intervals; average proportion of data = average proportion of data covered by 3-interval patterns that started at the first beat of a bar and occurred at least twice in the same bass line reduction; proportion of data = proportion of data covered by 3-interval patterns that started at the first beat of a bar and occurred at least twice in the same bass line reduction or at least once in two or more bass line reductions. Overlapping patterns were removed in all analyses.
Fig. 5. Consequences of combining the research material into a single data set. % notes = proportion of notes that started a pattern that occurred at least once in two or more bass line reductions or at least twice in the same bass line reduction at any metrical location (all bass line reductions by the same musician were combined into a single data set); aver. proportion = average proportion of notes that started an interval pattern that occurred at least twice in two or more bass line reductions at any metrical location (all bass line reductions by the same musician were analyzed separately); length = pattern length.
Fig. 6. Consequences of disregarding the metrical location of notes. % notes = proportion of notes that started one or more patterns that occurred at least once in two or more bass line reductions at any metrical location and contained three or more intervals; metrical location of notes taken into account = all instances of the same pattern were required to start from the same metrical location. Overlapping patterns were removed in both analyses.
Fig. 7. Consequences of disregarding the metrical location of notes when counting the number of recurring patterns. Start from the same metrical location = all instances of the same pattern were required to start from the same metrical location; * = number of different patterns that occurred at least once in two or more bass line reductions and contained three or more intervals; ** = number of different 3-interval patterns that occurred at least once in two or more bass line reductions; *** = number of different 3-interval patterns that occurred at least twice in one or more bass line reductions when all patterns were required to start from the first beat of a bar; **** = number of different 3-interval patterns that occurred at least twice in two or more bass line reductions when all patterns were required to start from the first beat of a bar. The minimum number of sources is two except when stated otherwise. Overlapping patterns were removed in all analyses.
The current results support the view that learning a large vocabulary of patterns is not a necessary requirement for expert level skills in jazz improvisation (Nurmi, 2023; Stehr, 2016) and suggest that recurring melodic patterns were used in Paul Chambers’s and Ron Carter’s bass lines from the 1950s and 1960s to a smaller extent compared to results from earlier studies, most of which investigated solos by saxophone players (Frieler, 2018; Norgaard, 2014; Norgaard et al., 2023; Norgaard & Römer, 2022; Weisberg et al., 2004). However, differences in pattern length, instrument, the use of different methods of measurement to analyze bass lines and solos, and the length of analyzed bass lines and solos may all influence the results. These differences preclude the direct comparison of results between studies using a different pattern length or different method of measurement, between studies analyzing double bassists and musicians playing other instruments, and between studies analyzing jazz bass lines and jazz solos. Whereas most previous studies have analyzed 4-interval patterns or patterns with three or more intervals, the current study focused on patterns with two to four notes or one to three intervals. This decision makes it difficult to compare the current results with those from previous studies because the proportion of recurring patterns decreases as the pattern length increases. Another factor is that the number of possible note choices may differ between musicians playing a different instrument, at least to some extent, which influences the probability of repeating the same sequences of notes or intervals. Comparing the extent of pattern use also requires using the same methods of measurement to analyze both bass lines and solos. Finally, the extent of pattern use tends to increase with sample size (the number of analyzed musical works) and the length of analyzed musical works (Nurmi, 2023; see also Norgaard & Römer, 2022). As a solution to this problem, Norgaard and Römer (2022) suggested comparing only results based on the same total number of notes. Note that splitting individual musical works into shorter segments with the same number of notes may lead to problems when used for this purpose because segment length also influences the results and because a small segment length “may lead to smaller variability of results compared to when using non-splitted data” (Nurmi, 2025, p. 11).
The current results also suggest that disregarding overlapping patterns and the metrical location of notes, allowing each note to start any number of patterns, combining all musical works by the same musician into a single data set, and requiring recurring patterns to occur only at least once in two or more musical works can overestimate the proportion and the number of recurring patterns. These findings indicate that earlier studies may have overestimated the extent of pattern use in jazz improvisation, which also calls into question whether solos of Charlie Parker or any other world-renowned jazz musician were built exclusively or almost exclusively from pre-learned patterns (cf. Frieler, 2018; Norgaard, 2014; Weisberg et al., 2004). Although most previous studies have removed overlapping patterns, almost all of them have disregarded the metrical location of notes and many have also combined all musical works into the same data set (instead of analyzing each musical work separately) or have required recurring patterns to occur either at least once in two or more musical works or at least twice in one or more musical works. It is also often unclear whether each note is allowed to start only a single pattern or not.
Unfortunately, the current study did not use the same pattern length and, in some cases, the same number of sources (the number of musical works where a pattern is required to occur) when comparing results acquired by using pattern coverage and the proportion of data covered by recurring patterns. Therefore, the results acquired by comparing these methods of measurement should be considered with caution. However, it is noteworthy that results from pattern coverage analysis indicated a much greater extent of pattern use compared to results acquired by using almost all other methods of measurement. Also note that Ron Carter’s bass lines from 1963 to 1967 were often very complex and the results from pattern coverage analysis indicated much more repetition of melodic patterns than expected by playing these bass lines in the absence of other instrumental parts (for transcriptions and analysis of Ron Carter’s musical style, see Nurmi, 2018). Taking these things into account, the proportion of notes that start a recurring pattern at any metrical location may produce more credible results but it is important that all instances of the same pattern are required to start from the same metrical location, all recurring patterns are required to occur at least twice in two or more musical works, and pattern use is measured in each musical work separately instead of combining all musical works by the same musician into a single data set. Further analysis is needed to understand why the results from pattern coverage analysis differed so much from most other results reported in the current study.
According to Nurmi (2023), Paul Chambers and Ron Carter often relied on a limited set of target notes, and they used pre-learned approach-note patterns extensively. Based on the current results, both bassists also often relied on a limited number of note choices at the first two beats of a bar. These findings indicate that eminent jazz bassists may be able to avoid excessive repetition of melodic patterns even if their note choices are constrained at the level of target notes, the first two beats of a bar, and approach-note patterns. In addition, these findings suggest that the use of a limited set of target notes, short patterns at the beginning of a bar, and approach-note patterns may facilitate the generation of non-repetitive bass lines and support the view that rules or constraints play an important role in jazz improvisation (Johnson-Laird, 1988, 2002). On the other hand, Paul Chambers and Ron Carter also used longer recurring patterns, which indicates that constraints and patterns both played an important role in Paul Chambers’s and Ron Carter’s bass lines. This finding is in line with Norgaard (2011) who found that experienced jazz musicians described using both pre-learned patterns and internalized rules or constraints.
Finally, the current results suggest that expert-level jazz improvisation may differ from linguistic skills in the size of vocabulary required to achieve fluency. Brysbaert et al. (2016) estimated that native speakers of US English know about 42,000 uninflected words and about 11,000 word families on average at the age of twenty and about 48,200 uninflected words and about 13,400 word families by the age of sixty. Although productive knowledge (the number of words actually used) is less than half of the number of known words (Brysbaert et al., 2016), the current study found only 183 four-note chordal pitch class patterns that occurred at least twice in two or more bass line reductions and 622 four-note chordal pitch class patterns that occurred at least twice in one or more bass line reductions by Paul Chambers. In Ron Carter’s bass line reductions, the total number of 4-note chordal pitch class patterns that occurred at least twice in two or more bass line reductions was 44 and the total number of 4-note chordal pitch class patterns that occurred at least twice in one or more bass line reductions was 246. The total number of interval patterns that occurred at least twice in two or more (or one or more) bass line reductions was similar to the number of recurring chordal pitch class patterns. These findings suggest that vocabulary may play a more important role in language compared to jazz improvisation.
According to Temperley (2022), “it is […] hard to know whether these patterns are genuine schemata in the minds of composers and listeners or whether they simple arise by chance as clusters of highly probable events” (p. 156). Referring to this problem, the current article relied on the assumption that short melodic patterns that occur only once in two or more musical works provide little evidence to support the claim that such patterns were actually pre-learned and retrieved from memory during the improvisation process. However, it is important to note that the total number of recurring chordal pitch class patterns and interval patterns in Paul Chambers’s and Ron Carter’s bass lines was probably underestimated because of small sample sizes (even though the small number of analyzed bass lines was compensated to some extent by allowing differences in rhythm and subdivision-level changes in melody and by requiring recurring melodic patterns to occur not more than at least twice in two or more bass lines).
This article is an extension of the author’s dissertation, “Temporal constraints and creativity in bass lines of eminent jazz musicians,” completed at the University of Jyväskylä in 2023, and funded by the Eino Jutikkala Fund of the Finnish Academy of Science and Letters, Oskar Öflunds Stiftelse sr, Jenny and Antti Wihuri Foundation, and Alfred Kordelin Foundation.
[1] Correspondence can be addressed to: Mikko Nurmi, Department of Music, Art and Culture Studies, University of Jyväskylä, P.O. Box 35, FI-40014 Jyväskylä, Finland. Email: mikko.m.nurmi@gmail.com.
Supplementary materials to this article are available at https://osf.io/h4y8x.
Abeßer, J., Frieler, K., Pfleiderer, M., Zaddach, W.-G., & Höger, F. (2018). MeloSpyGUI (Version 1.4.1) [Computer software]. https://jazzomat.hfm-weimar.de/index.html
Acevedo, S., Temperley, D., & Pfordresher, P. Q. (2014). Effects of metrical encoding on melody recognition. Music Perception, 31(4), 372–386. https://doi.org/10.1525/mp.2014.31.4.372
Benning, M. (2018-2020). Paul Chambers bass line transcriptions on Bass on Top, It’s a Blue World, and Jazzmen Detroit. Retrieved October 6, 2021, from http://www.mikebenning.com
Bierma, P. (2019). Paul Chambers bass line transcriptions on A Foggy Day, All of You, Blue Train, C-Jam Blues, and Tenor Madness.
Brysbaert, M., Stevens, M., Mandera, P., & Keuleers, E. (2016). How many words do we know? Practical estimates of vocabulary size dependent on word definition, the degree of language input and the participant’s age. Frontiers in Psychology, 7, Article 1116. https://doi.org/10.3389/fpsyg.2016.01116
Christensen, W., Sutton, J., & Bicknell, K. (2019). Memory systems and the control of skilled action. Philosophical Psychology, 32(5), 693–719. https://doi.org/10.1080/09515089.2019.1607279
Fink, D. (2009-2012). Paul Chambers bass line transcriptions on I Can't Give You Anything but Love, I’ll Remember April (Excerpt), and Israel. Retrieved April 6, 2021, from http://www.jazzcapacitor.com
Frieler, K. (2014, September 26–27). Pattern usage in monophonic jazz solos [Paper presentation]. International Jazzomat Research Workshop, Weimar, Germany.
Frieler, K. (2017). Computational melody analysis. In M. Pfleiderer, K. Frieler, J. Abeßer, W.-G. Zaddach, & B. Burkhart (Eds.), Inside the Jazzomat: New perspectives for jazz research (pp. 41–84). Mainz, Germany: Schott. https://schott-campus.com/jazzomat/
Frieler, K. (2018, May 4). Patternology. Melodic pattern usage in Charlie Parker's solos [Video]. Ircam, Paris, France. https://medias.ircam.fr/x54b408
Frieler, K. (2020). Miles vs. Trane: Computational and statistical comparison of the improvisatory styles of Miles Davis and John Coltrane. Jazz Perspectives, 12(1), 123–145. https://doi.org/10.1080/17494060.2020.1734053
Frieler, K., Höger, F., Pfleiderer, M., & Dixon, S. (2018). Two web applications for exploring melodic patterns in jazz solos [Paper presentation]. 19th International Society for Music Information Retrieval Conference, Paris, France.
Goldman, A. (2013). Towards a cognitive-scientific research program for improvisation: Theory and an experiment. Psychomusicology: Music, Mind, and Brain, 23(4), 210–221. https://doi.org/10.1037/pmu0000020
Gregor, E. (2018, September 2). Paul Chambers bass line transcription on Mr. P.C. [Video]. YouTube. https://www.youtube.com/watch?v=7XW-MUbf6OQ
Hal Leonard (2004). The Real Book Volume 1. Milwaukee, WI: Hal Leonard.
Herridge, M. (2020, November 21). Paul Chambers bass transcription on Cool Struttin’ [Video]. YouTube. https://www.youtube.com/watch?v=8dGhgLWC_M8
Jazz Bass Transcriptions (2021). Paul Chambers full bass score on Syeeda’s Song Flute and Woody’n You. Retrieved October 19, 2021, from https://www.jazzbasstranscriptions.com
Johnson-Laird, P. N. (1988). Freedom and constraint in creativity. In R. J. Sternberg (Ed.), The nature of creativity: Contemporary psychological perspectives (pp. 202–219). New York, NY: Cambridge University Press.
Johnson-Laird, P. N. (2002). How jazz musicians improvise. Music Perception, 19(3), 415–442. https://doi.org/10.1525/mp.2002.19.3.415
Kolarczyk, T. (2013, January 31). Ron Carter bass line transcription on Autumn Leaves [Video]. YouTube. https://www.youtube.com/watch?v=qC9jgyhG9b0
Nabuurs, P. (2017, October 2). Paul Chambers bass line transcription on Moment’s Notice [Video]. YouTube. https://www.youtube.com/watch?v=nj4XB-cwsOM
Norgaard, M. (2011). Descriptions of improvisational thinking by artist-level jazz musicians. Journal of Research in Music Education, 59(2), 109–127. https://doi.org/10.1177/0022429411405669
Norgaard, M. (2014). How jazz musicians improvise: The central role of auditory and motor patterns. Music Perception, 31(3), 271–287. https://doi.org/10.1525/mp.2014.31.3.271
Norgaard, M., Bales, K., & Hansen, N. C. (2023). Linked auditory and motor patterns in the improvisation vocabulary of an artist-level jazz pianist. Cognition, 230, Article 105308. https://doi.org/10.1016/j.cognition.2022.105308
Norgaard, M., Emerson, S. N., & Dawn, K. (2016). Creating under pressure: Effects of divided attention on the improvised output of skilled jazz pianists. Music Perception, 33(5), 561–570. https://doi.org/10.1525/mp.2016.33.5.561
Norgaard, M., & Römer, U. (2022). Patterns in music: How linguistic corpus analysis tools can be used to illuminate central aspects of jazz improvisation. Jazz Education in Research and Practice, 3(1), 3–26.
Nurmi, M. (2018). The art of Ron Carter 1963-1967: Transcription and analysis of selected bass lines. New York, NY: Ron Carter Books.
Nurmi, M. (2023). Temporal constraints and creativity in bass lines of eminent jazz musicians [Doctoral dissertation, University of Jyväskylä]. JYX Digital Repository. http://urn.fi/URN:ISBN:978-951-39-9695-6
Nurmi, M. (2025). The relationship between time constraints and creativity in bass lines of eminent jazz musicians: An exploratory study [Supplemental material]. Psychology of Aesthetics, Creativity, and the Arts. Advance online publication. https://doi.org/10.1037/aca0000788.supp
Ouellette, D. (2013). Ron Carter: Finding the right notes. New York, NY: Retrac Productions.
Owens, T. (1974). Charlie Parker: Techniques of improvisation [Doctoral dissertation, University of California].
Palmer, R. (2012). Mr P.C.: The life and music of Paul Chambers. Sheffield, UK: Equinox Publishing.
Pfleiderer, M. (2017). The Weimar Jazz Database. In M. Pfleiderer, K. Frieler, J. Abeßer, W.-G. Zaddach, & B. Burkhart (Eds.), Inside the Jazzomat: New perspectives for jazz research (pp. 19–40). Mainz, Germany: Schott. https://schott-campus.com/jazzomat/
Povel, D.-J., & Essens, P. (1985). Perception of temporal patterns. Music Perception, 2(4), 411–440. https://doi.org/10.2307/40285311
Pressing, J. (1988). Improvisation: Methods and models. In J. A. Sloboda (Ed.), Generative processes in music: The psychology of performance, improvisation, and composition (pp. 129–178). New York, NY: Oxford University Press. https://doi.org/10.1093/acprof:oso/9780198508465.001.0001
Pressing, J. (1998). Psychological constraints on improvisational expertise and communication. In B. Nettl (Ed.), In the course of performance: Studies in the world of musical improvisation (pp. 47–68). Chicago, IL: University of Chicago Press.
Schmidt, R. A., & Wrisberg, C. A. (2004). Motor learning and performance: A problem-based learning approach (3rd ed.). Champaign, IL: Human Kinetics.
Sher Music (1988). The New Real Book. Petaluma, CA: Sher Music.
Sher Music (1991). The New Real Book Volume 2. Petaluma, CA: Sher Music.
Sher Music (1995). The New Real Book Volume 3. Petaluma, CA: Sher Music.
Skinner, J. W. (2016). A comparative analysis of three jazz bassists’ walking bass lines [Doctoral dissertation, University of Northern Colorado]. https://digscholarship.unco.edu/dissertations/329
Sloboda, J. A. (1983). The communication of musical metre in piano performance. The Quarterly Journal of Experimental Psychology, 35(2), 377–396. https://doi.org/10.1080/14640748308402140
Stehr, M. W. (2016). Bird's words and Lennie's lessons: Using or avoiding patterns in bebop (Publication No. 10247656) [Doctoral dissertation, University of Nebraska]. ProQuest Dissertations.
Temperley, D. (2022). Music and language. Annual Review of Linguistics, 8, 153–170. https://doi.org/10.1146/annurev-linguistics-031220-121126
Thompson, J. J., McColeman, C. M., Blair, M. R., & Henrey, A. J. (2019). Classic motor chunking theory fails to account for behavioural diversity and speed in a complex naturalistic task. PLoS ONE, 14(6), Article e0218251. https://doi.org/10.1371/journal.pone.0218251
Waters, K. (2011). The studio recordings of the Miles Davis Quintet, 1965–68. New York, NY: Oxford University Press. https://doi.org/10.1093/acprof:osobl/9780195393835.001.0001
Weisberg, R. W., Brinkman, A. R., Folio, C. J., Dick, A., Fleck, J. I., Niederberg, B., & Barrett, F. (2004, April 15–18). Toward a cognitive analysis of creativity: Improvisation in jazz [Paper presentation]. Conference on Interdisciplinary Musicology, Graz, Austria.
Wetzel, P. (1996). Bill Evans Fake Book. New York, NY: Ludlow Music.