You are on page 1of 22

Quantitative Case-Studies in Musical Composition: the Development of

Creativity in Popular-Songwriting Teams


Quantitative Case-Studies in Musical Composition:

The Development of Creativity in Popular-Songwriting Teams Richard W. Hass, Robert W. Weisberg,

and Jimmy Choi

Temple University
Abstract

In the present study, we examined the development of expertise in the careers of three

collaborative songwriting teams, Rodgers and Hart, Rodgers and Hammerstein, and George and

Ira Gershwin. The primary goal of the study was to extend to collaborative songwriting teams the

results of previous research (Hass & Weisberg, 2009), which examined the career development of

two seminal composers of American popular music, Irving Berlin and Cole Porter, who wrote both

music and lyrics. A quantitative case-study methodology was employed here, using the proportion

of hit songs per year within collaborators’ active careers as a measure of aesthetic significance (hit

ratio). Regression analysis revealed significant positive hit-ratio trends over the careers of the

Rodgers-Hart team, and the Gershwin team. In contrast, the Rodgers-Hammerstein team began at

a high level of success and exhibited no significant hit-ratio trends. That team began their

collaboration when both members were mature professionals, with years of experience writing for

Broadway. In addition, both members of the team had collaborated with other individuals earlier in

their careers, and both exhibited the expected significant increases in hit ratio with other

collaborators prior to their teaming up. Such results are evidence that a process of learning to write

hits occurs early in the careers of collaborative songwriting teams, as it does in the careers of

composers who both write music and lyrics. Implications of the results and future directions for

study are discussed.


4

Quantitative Case-Studies in Musical Composition:

The Development of Creativity in Popular-Songwriting Teams Many recent studies of

creativity have focused on the development of creative expertise in various domains (e.g. Feist,

1993; Hass & Weisberg, 2009; Kozbelt, 2004; 2005; Simonton, 2003a, Weisberg, 2006). As in

many other developmental discussions, the issue has arisen as to whether high-level creativity is

positively influenced by learning and practice (c.f. Hass & Weisberg, 2009). Ericsson (1999), for
example, proposed that many of the most eminent innovators of the modem era (i.e. Picasso;

Einstein; Watson and Crick) can be characterized as expert-level performers in domains that

require creative thinking. This link between expert performance and creative productivity is the

foundation of the expertise-based perspective on creativity (cwhere creativity is viewed as the

result of normal cognitive processes of individuals with a high level of expertise within a

particular domain (Hass & Weisberg, 2009; Weisberg, 2006). Thus, one goal of research on

creativity from the expertise-based perspective is to examine the influence of learning and

expertise on the development of high-level creativity.

In order to examine the creative process in eminent individuals, research from the expertise-

based perspective has focused primarily on creative products (c.f. Martindale, 1990; Simonton,

2003b; 2007; Weisberg, 1995; 2004; 2006), which allows for the collection of objective

quantitative data from both living and deceased creators. For example, Weisberg (2004) examined

the 45 preliminary sketches leading up to the creation of Picasso’s 1937 painting Guernica.

Weisberg found that sketches were clustered temporally according to content, with composition

sketches typically occurring first, followed by groups of sketches of the different characters that

appeared in the final painting. This methodology, which can be called a quantitative case-study

approach, has been successfully utilized to examine cases of in a variety

Song writing Teams - 5

of domains). In the study of musical composition, this approach has been employed in the

examination of lifetime creativity of one or a few eminent individuals (e.). This variant of the

quantitative case study is most pertinent to understanding the role played by learning in the

development of high-level creativity.

Recently, Kozbelt (2004) conducted a quantitative case-study of the career development of 18

classical-music composers. The study was designed to replicate results of an early study of 10
classical-music composers (Simonton, 1977). Kozbelt examined the same 10 composers from

Simonton’s study and added an additional eight composers to increase the power of the analysis.

Kozbelt found that none of the composers in the sample had completed a masterwork (defined

according to a scoring system which utilized the number of recordings of each work and its

inclusion in a guide for listening to classical music; Halsey, 1976) earlier than 10 years into their

careers. That result was consistent with previous research by Hayes (1989) and Weisberg (1999),

and more broadly with what has been called the 10-year rule of expert-performance (c.f. Chase &

Simon, 1973; Ericsson, 1996; Ericsson & Chamess, 1994; Howe, Davidson & Sloboda, 1998).

High-level performance in any domain is made possible by deliberate and intensive practice of

highly complicated, domain-specific skill sets over an extended period of time (Ericsson, 1996).

The discovery of the 10-year rule in the careers of classical music composers lent support to the

expertise-based perspective on creativity.

Kozbelt (2004) also found that when the ratio of major-works to minor-works was calculated

for successive 5-year blocks in each composer’s career (the “hit ratio”; c£, Simonton, 1977), the

early blocks were marked by significantly lower hit ratios than later blocks. That

Songwriting Teams - 6

result clearly demonstrated that the 10-year rule of expert-performance may also be applicable

to the development of high-level creativity (see also, Kozbelt, 2005; Weisberg, 1999).

However, challenges to the expertise-based view have been raised by Simonton (e.g.

1977, 1999, 2003a). Simonton’s original analysis suggested that the hit ratios of the classical-

music composers he examined remained constant across their careers. In addition, Simonton found

a significant correlation between the number of major and minor-works in each composer’s career.

Those findings have formed one foundation of his chance-configuration theory of creative

production (c.f. Simonton, 2003a), which states that a creator can increases the chances of
producing a successful product only by increasing the quantity of his or her output. In a more

recent examination of a larger sample of composers, Kozbelt (2008) found that chance factors may

influence the creativity of “one-hit-wonder” composers who, unlike, for example, Mozart, did not

produce a consistent pattern of high-quality work.

According to chance-configuration theory, the creative process is analogous to the process of

genetic evolution, a two-stage process involving the random variation of creative ideas, and the

selective retention of ideas that prove successful. Due to the random nature of idea generation and

combination, a creator cannot learn to produce quality output by gaining expertise; he or she can

increase the chance of success only by producing a large quantity of products.

As noted, Kozbelt’s (2004) results were opposite to those of Simonton (1977) and therefore

contradicted the Darwinian view. Hass and Weisberg (2009) outlined several potential reasons for

the inconsistency between Simonton’s (1977) and Kozbelt’s (2004) results. First, both Simonton

and Kozbelt used samples that included composers from various periods all of whom composed

pieces in various genres. It is not clear how to combine pieces from very different genres when

trying to assess changes in a composer’s compositional skills over an

Songwriting Teams - 7

entire career. In addition, each author used different sources to classify compositions as major

or minor works. Any of these issues might have contributed to the opposite conclusions drawn by

Simonton and Kozbelt.

Hass and Weisberg (2009) carried out a study designed to eliminate the potential problems

brought about by genre on measurement of aesthetic success of musical compositions. Rather than

examine classical composers, Hass and Weisberg studied two eminent composers of 20th-century

American popular music, Cole Porter and Irving Berlin. Popular songs are relatively standard in

length and form (c.f. Middlton, 1990; Wilder, 1972), which eliminated the problems of genre that
arise in the study of classical composers.

One additional incentive for focusing on American popular music was that an objective and

face-valid index of aesthetic success (the number of recordings per song) was readily available for

all of the songs. The composers in question were writing songs for the purpose of drawing people

into theatres, having singers record the songs, and selling sheet music and recordings to the public.

Hence, the number of recordings of a given song is a face-valid measure of a composer’s success.

Hass and Weisberg (2009) collected information on the number of recordings for each song

from the All Music Guide’s (AMG) online data base (www.allmusic.com). Songs were labeled as

hits based on recording-based criteria. Each composer’s career was marked by an initially low hit

ratio, followed by a substantial increase in hit ratio 10 to 20 years into the career. These results

provided support for Hayes’s (1989), Kozbelt’s (2004), and Weisberg’s (1999) results and, more

broadly, supported the expertise-based view on creativity.

Many of the songs from what has been called the Golden Age of American popular music

(1900-1950; cf., Wilder, 1972) were written by collaborative songwriting teams. In an effort to

further characterize the possible roles of practice and learning in the creation of popular songs,

Songwriting Teams - 8

the current case study is focused on three collaborative teams: George and Ira Gershwin,

Richard Rodgers and Lorenz Hart, and Richard Rodgers and Oscar Hammerstein. Previous

research by Weisberg (e.g. 1999, 2006) suggested that collaborative songwriting teams develop in

a fashion similar to individual composers. Weisberg investigated the career of the Beatles, from

1957, when John Lennon and Paul McCartney met and the latter joined the former’s musical

group, to 1966, when, due to the frenzy brought on by Beatlemania, the Beatles stopped

performing in public. Those years could be looked upon as times of extensive practice. By the end

of 1961 (i.e., approximately half-way into their performance careers), the Beatles had performed
nearly 600 times, and the Lennon-McCartney team had written 20 songs, many of which were

never recorded by the Beatles. In addition, during that time their shows featured a majority of

cover songs, mainly American rhythm and blues songs (Price, 1997). From 1962 to 1966, the

Beatles performed at least an additional 600 times, and mostly performed their own material.

Thus, it can be said that the Beatles spent a significant amount of time practicing their craft.

Furthermore, in 1962-1966 (i.e., the second half of their performing careers), Lennon and

McCartney wrote 37 songs together, almost all of which were recorded, indicating a significant

increase in the quality of their work over the years 1957-1966. Weisberg (2006) concluded that the

10-year rule was evident in the collaborative partnership of Lennon and McCartney, since the duo

started writing songs in 1957 and began their most creative period around 1967.

The goal of the current study was to examine whether collaborative teams working in

American popular music showed the same developmental trajectory found in studies of individual

composers (i.e. Hass & Weisberg, 2009; Kozbelt, 2004) and of the Beatles (Weisberg, 1999).

Method
Data Collection

Exhaustive lists of songs by each collaborative team were compiled using various sources

cited below. Efforts were made to include only songs for which lyrics and music both survive and

which were published. Generally, songs were composed and published within the same year, and

so the date of publication was used to date each song. The number of times a song was recorded

was tabulated using the All Music Guide database (www.allmusic.com), which represents a

cumulative frequency of recordings. The number of different artists who recorded each song was

also tabulated. However, that measure was found to be highly correlated with the recording-count

measure; only the recording-count measure was used as the basis for analysis.

Recording counts were then corrected for the presence of full-cast recordings that might have
biased the data. A song might have an inflated recording count because every song in a show

appears on multiple cast albums. Table 1 lists the average number of cast albums per songwriting

team, as well as the show with the maximum amount of recordings. The number of cast-album

recordings of the source show for each song was subtracted from the song’s total. Songs were then

classified as hits if the adjusted recording count was ten or more.

The active years in the careers of each collaborative team were then numbered. Years in which

the team wrote fewer than 4 songs were considered inactive years, and were combined with the

surrounding years in the following fashion: the first song (alphabetically) was assigned to the

previous year, the second to the following year, and so forth. The rationale for that dating system

was to remove years in which productivity was near zero, which could result in highly variable

hit-ratios.

For each team in each active year, the following variables were tabulated: (1) the number of

songs {output), (2) the number of hits (adjusted recording count more than 10), (3) the number of

nonhits, (4) the ratio of hits to total output {hit ratio).


Rodgers, Hart, and Hammerstein
10

Rodgers began his composing career when he began collaboration with Hart in 1919. Neither

Rodgers nor Hart worked with anyone else between 1919 and 1942 (Suskin, 1990). The

development of Rodgers-Hart team is particularly “clean” and, therefore, greatly informative as a

test of our perspective, since both members of the team began the collaboration as relative novices

and worked exclusively with each other. Hart died in 1943, succumbing to the effects of alcohol

abuse. Examination of the creative development of the Rodgers-Hammerstein team is more

complicated, because both Rodgers and Hammerstein were experienced when they teamed up.

Hammerstein had worked with several different composers over 22 years before teaming with

Rodgers. Thus, in order to examine the development of Rodgers and Hammerstein, it is necessary
to carry out two analyses: one examining Rodgers’s whole career (i.e., his work with Hart plus that

with Hammerstein) and the second examining Hammerstein’s whole career (i.e., his pre-Rodgers

work plus that with Rodgers).

Before discussing the analyses and results, some theoretical issues should be addressed.

Assume that the pattern found in previous studies of individual composers (e.g. Hass & Weisberg,

2009; Kozbelt, 2004)—an increasing hit ratio—is, at least in general form, representative of the

typical development of a collaborative team. That assumption leads to interesting possibilities

concerning the pattern of career development over two long-term collaborations, as shown in

Figure 1. On the one hand, it might be that each collaboration represented a unique learning

experience, with no transfer from one to another. If so, the second collaboration should, like the

first, begin with very low hit ratios and only over time develop into a success (see Figure la). On

the other hand, the expertise developed in the first collaboration might transfer to the second, in

which case high levels of hit ratios would be seen from the beginning of the second collaboration

(see Figure lb). Rodgers’s career, encompassing two longterm collaborations, allows us to examine

those possibilities.

Songwriting Teams -11

After Hammerstein died, Rodgers went on to work on several other musicals: No Strings

(1962; lyrics by Rodgers); Do I Hear a Waltz? (1965; lyrics by Stephen Sondheim); Two by Two

(1970; lyrics by Martin Chamin); Rex (1976; lyrics by Sheldon Hamick); I Remember Mama

(1979; lyrics by Martin Chamin and Raymond Jessel). We decided for the sake of completeness to

include those collaborations. One interesting question was whether Rodgers’s status at that point in

his career—the “dean” of Broadway composers—might have resulted in anything he worked on

resulting in hits.

Songs
A complete list of Rodgers’s songs (695 in total), compiled by Suskin (1990), was used as the

database for analysis.

Rodgers and Hart. A separate song list was compiled for the Rodgers-Hart collaborative team,

which included only works that were written by both members of the team during the years of

their partnership (other collaborations - less than 4% of the total - were ignored). Works for which

neither music nor lyrics survived (less than 1% of the total) were excluded. The total number of

Rodgers and Hart songs was 320.

Rodgers and Hammerstein. Suskin’s (1990) book was also used to compile the list of songs

written by Rodgers and Hammerstein. During their collaboration, neither member of the team

worked with any other collaborators. Music and lyrics for all songs written during that period

survive. Instrumental works (less than 1% of the total), written as parts of musicals, were excluded

from the analysis. Together, the Rodgers-Hammerstein team wrote 171 songs.

Data Analysis

Rodgers and Hart wrote 60 hits and 260 non-hits and were active for a total of 19 years.

Between 1943 and 1959 Rodgers and Hammerstein were active for 10 years, and wrote 58 hits
and 113 non-hits.
12

Figure 2 depicts the hit ratios per active year for Rodgers and Hart, and Rodgers and

Hammerstein. Figure 2 resembles Figure lb more than that in Figure la: it appears that there was

transfer of creative expertise between Rodgers’s collaboration with Hart and his collaboration with

Hammerstein. That transfer becomes clearer when one examines the hit ratios in more detail.

Hit-Ratios. The hit ratios per active year for the Rodgers-Hart team increased throughout the

first three quarters of their career and then declined. A regression model was created with year and

year quadratic as predictors and hit ratio as the dependent. Year quadratic was created by

squaring the z-scores of the year variable. The regression was significant, F(2, 16) = 6.18,/? = .01,
adjusted R = 0.37. Year (3 = 0.01 and year quadratic P = -0.10, significantly predicted hit ration’s

lees than 0.05. The presence of a positive linear and negative quadratic trend suggests that

Rodgers and Hart’s career is best described by a backward, inverted J-function, as can be seen on

the left side of Figure 2. No significant correlation between hits and non-hits was found, r(l 8) = - .

16,/? = .95.

The hit ratios for Rodgers and Hammerstein show no such early increase. A regression model

was created with year quadratic as the predictor variable and hit-ratio as the dependent variable.

The regression was not significant, F(l, 8) = 1.11,/? = .18, adjusted R2 = 0.11. There was a

significant negative correlation between hits and non-hits in the career of Rodgers and

Hammerstein, r(9) = -.69, p = .03. Those results, from the expertise perspective, indicate transfer

from the early collaboration to the later (i.e. Figure lb).

Hammerstein: Many + Rodgers

Hammerstein worked with many composers prior to teaming up with Rodgers in 1943

(Hammerstein, 2008). A complete list of Hammerstein’s pre-1943 work was also compiled using
13

Suskin’s (1990) list. Between the years of 1920 and 1942, Hammerstein wrote lyrics for 340

songs with composers other than Rodgers.

Data Analysis

Between the years of 1920 and 1942, Hammerstein wrote 30 hits and 310 non-hits with other

composers before teaming up with Rodgers, and was active for a total of 19 years.

Including his work with Rodgers, Hammerstein was active for a total of 29 years. Throughout

his entire career, Hammerstein wrote lyrics for 88 hit songs and 423 non-hits.

Figure 3 depicts a possible quadratic trend in the hit-ratios of the active years in

Hammerstein’s early (pre-Rodgers) career. Year quadratic served as the predictor in a regression

model with hit-ratio as the dependent. The regression was marginally significant, F(l, 6) = 3.23, p =
.09, adjusted R2 = 0.11. Year quadratic (3 = -0.004, predicted hit-ratio, p = 0.09. A marginally

significant correlation between hits and non-hits was found in Hammerstein’s early career, r(18) =

0.41,/? = .08.

Figure 3 depicts hit-ratios for the active years in Hammerstein’s entire career, including work

with Rodgers. A regression model was created using year as the predictor and hit-ratio as the

dependent. The regression was significant, F(l, 27) = 11.02,/? lees than0.01, adjusted R2 = 0.27.

Year P = 0.001, significantly predicted hit-ratio, p lees than0.01. There was also no correlation

between the number of hits and non-hits throughout active years in Hammerstein’s entire career,

r(28) = - 0.20,/? = 0.31.

In conclusion, the careers of both Rodgers and Hammerstein as discussed so far show similar

patterns: an increase in hit-ratio as a result of collaboration, resembling the idealized trend

depicted in Figure lb.

Rodgers and Hammerstein Pre-Collaboration Years

14 Rodgers pre-Hammerstein worked with a single collaborator (Hart), while Hammerstein

worked with many collaborators (c.f., Hammerstein, 2008). Thus, we can compare the success

levels for Rodgers and Hart with those for Hammerstein’s pre-Rodgers career to gain some insight

as to the role of experience in song-writing, per se, versus experience in collaborative song

writing.

We can examine the hit ratios collapsed over the last 10 years of Rodgers pre- Hammerstein

(i.e., Rodgers and Hart; 152 songs, 42 hits, hit ratio = .32) and Hammerstein pre- Rodgers (144

songs, 12 hits, hit ratio = .09). Those overall hit ratios are different (X (df= 1, N = 296) = 18.46, p

lees than .001), which indicates that it is not simply years of collaboration that are important,

although they do play some role, since we have seen in the regression analysis of Hammerstein’s

pre-Rodgers years that the hit ratios did show an increase over those years. Thus, the members of a
long-term collaborative team are acquiring some specific sort of experience working with each

other, above what each member of the team is learning about writing lyrics or melodies,

respectively.

Rodgers after Hammerstein

As noted earlier, after Hammerstein’s death Rodgers worked on several additional projects: No

Strings (1962; lyrics by Rodgers); Do I Hear a Waltz? (1965; lyrics by Stephen Sondheim); Two

by Two (1970; lyrics by Martin Chamin); Rex (1976; lyrics by Sheldon Hamick); I Remember

Mama (1979; lyrics by Martin Chamin & Raymond Jessel). As shown in Figure 2 (far right), those

projects were generally not successful. Figure 4 presents the hit ratios per active year in Rodgers’s

entire career. A regression model was created with year quadratic as predictor and hit ratio as the

dependent. The regression was significant, F{ 1, 32) = 8.58, p lees than

0.01, adjusted R2 = 0.19. Year quadratic p = -0.11, significantly predicted hit-ratio, p lees than

0.01.
Song writing Teams - 15

There was no correlation between hits and non-hits during the active years of Rodgers’ entire

career, r(33) = -0.28, p = 0.11.

Discussion

The results of the analysis of the careers of Rodgers, Hart, and Hammerstein revealed that all

three individuals were affected by their partnerships. Rodgers and Hart can be said to have

developed their creative expertise together, throughout the early years of their collaboration. While

Rodgers and Hammerstein began their collaboration with immediate success, both members of the

team exhibited significant development in earlier collaborations with others. By the time of

Rodgers and Hammerstein’s collaboration in 1943, each member of the team was at the peak of

his creative careers.

Hammerstein’s career did not fit a curvilinear trend. There are several possible reasons for that
result. The two most direct causes may be his lack of a stable collaboration early in his career, and

the fact that, like Hart, Hammerstein’s career was interrupted by his death. As Simonton (1977)

pointed out, physical illness can affect the quantity and quality of creative output. Hart, suffering

from alcoholism, presumably felt the effects of illness over a longer period of time than did

Hammerstein, who died of cancer in 1960 at age 65 after a short-lived battle with the disease.

In summary, the increases in hit ratios throughout the formative years in the careers of

Rodgers, Hart, and Hammerstein suggest that all three men experienced creative development in

the years leading up to their creative peaks. That result is consistent with the 10-year rule of the

development of expertise and more generally supports the expertise-based perspective on

creativity.
George and Ira Gershwin
16

The Gershwins collaborated over a 20-year period (1918-1937). George began writing popular

songs in addition to his more elaborate instrumental works in 1913. Between 1913 and 1925

George collaborated with several lyricists. However, most of George’s best known popular songs

featured lyrics written by Ira.

Ira began his career as a lyricist writing songs with George in 1918, however until 1924, their

collaborations were sporadic. Between 1925 and George’s death in 1937, and apart from Ira’s

work on the ZeigfeldFollies of1936 (music by Vernon Duke), the pair published popular songs

exclusively with each other. After George’s death, Ira wrote lyrics for Broadway and Hollywood

until 1954.

When George and Ira Gershwin began to work together, it was Ira’s initial professional

experience, which is comparable to the situation for Rodgers and Hart. However, Ira’s older

brother George had already worked for approximately 5 years with several other lyricists. Thus,

we begin our analysis of the Gershwins with Ira, since all of his early experience came as part of
that team. We then turn to George’s career, in which he wrote songs, concurrently, with several

collaborators (see Carnovale, 2000). Based on the analysis of Hammerstein’s pre-Rodgers career,

which also involved multiple collaborations, we might expect to find an increase in hit ratios for

George, although it might be slower than that found for Ira, who was working within a consistent

collaboration. Second, in the early years when George and Ira began to work together consistently,

George was still working with other lyricists. It will be interesting to compare George’s hit ratios

with Ira versus those with other composers for those years, to see if there was anything about the

Gershwins’ developing consistent collaboration that resulted in more successful results.


Songs
17

A list of songs written by Ira Gershwin without George was compiled using Suskin’s (1990)

book and the ASCAP database (www.ascap.com/ace/). Again, songs for which music and lyrics

did not survive (less than 5% of the total) were excluded. Also excluded from analysis were 12

songs that Ira wrote lyrics for after George’s death, as they could not be accurately dated. A

complete list of George Gershwin’s popular songs compiled by Camovale (2000) was used as the

database for analysis of his career. George Gershwin’s instrumental works (i.e. “Rhapsody in

Blue”) were also excluded.

Data Analysis

Ira and George. Between 1918 and 1937 the Gershwin brothers composed a total of 163

songs (68 hits). During that time, they were active for 15 years, however, the years 1918 through

1920 yielded only one song each, so they were collapsed into one composite year. Figure 5

presents the hit ratios per active year in the Gershwins’ collaboration plus Ira’s collaborations

post-George. Examining the Gershwins’ collaboration first, presented in the left-hand curve in

Figure 5, we see that their development parallels that for Rodgers and Hart: an initial lull in hit

rate, followed by increasing success. A regression model was used to fit the hit ratios to a
* 2
logarithmic function. The regression was significant, F(l, 8) = 6.70, p = .03, adjusted R = 0.39.

Year P = 0.04, significantly predicted hit-ratio, p = 0.03. There was also a marginally significant

correlation between hits and non hits, r(19) = 0.51 ,p lees than 0.09.

Ira after George. As noted, Ira continued to write lyrics for various composers after George’s

death in 1937. For the purposes of continuity, the current analysis examined Ira’s activity after

George’s death, between 1938 and 1954. During that time Ira was active for 8 years, with a variety

of partners, and wrote lyrics for 74 songs (8 hits).

The right-hand curve in Figure 5 presents the hit ratios for Ira’s career after George (1938-

1954). A regression model was created with year and year quadratic as predictors and hit
18

ratio as the dependent variable. The regression was not significant, F(3, 17) = 0.51, p = 0.63,

adjusted R2 = -.17. There was no correlation between the number of hits and non-hits in Ira’s

career after George, r(7) = -0.17,/? = 0.69.

The overall pattern in Ira Gershwin’s career, in contrast to that of Rodgers, looks more like that

in Figure 1(a). There was little transfer on Ira’s part from the Gershwin team to the subsequent

collaborations, as well as no development in the later collaborations. Evidence for the lack of

transfer is seen if we compare the hit ratios for the eight active years that Ira worked post- George

(a total of 74 songs written, with 8 hits; hit ratio of .11) with those for the last eight years of the

Gershwins’ collaboration (131 songs written, 58 hits; hit ratio of .44), those two overall hit ratios

are different (X2(df= 1, N = 205) = 24.26,/? lees than .0001).

George without Ira. Figure 6 presents the hit ratios for George’s career, with Ira and with other

lyricists. As mentioned earlier, George had started working with other lyricists in 1913 and began

to collaborate with Ira in 1918. From 1918 - 1925, George worked with Ira and with other

lyricists. From 1925 until his death, George worked only with Ira. Between 1913 and 1925,

George Gershwin wrote the music for 202 songs (10 hits) with lyricists other than Ira. The years
between 1913 and 1916 yielded only 9 songs, so the songs were broken into 2 composite years:

1913-1914 (5 songs) and 1915-1916 (4 songs). None of those songs had recording counts greater

than zero and grouping them together caused no variation in the model. Together with the two

composite years just described, George was active for a total of 10 years, with other collaborators.

The left-hand curve in Figure 6 depicts hit ratios for George’s work without Ira between 1913 and

1925. A simple regression model was constructed with year as the predictor and hit ratio as the

dependent. The regression was significant, F(l, 8) = 9.72,/? = 0.01, adjusted R2 = 0.49. Year (3 =

0.001, significantly predicted hit-ratio,/? = 0.01. A marginally significant correlation was also

found between the number of hits and non-hits in George’s career, r(9) =
19

0.62, p = 0.06. However, it should be noted that the increase in hit ratios was small in absolute

terms, with the maximum hit ratio for George in the pre-Ira years equal to . 12.

As discussed above, during the early years of the Gershwins’ collaboration, from 1918- 1925,

George also worked independently of Ira. The George-without-Ira collaborations produced 10 hits

out of 193 total songs, a hit ratio of .05. In those same years, the Gershwin team produced 10 hits

out of a total of 32 songs, a hit ratio of .31. Those hit ratios were significantly different, AY#1),N =

225) = 23.13, plees than .001. As can be seen from Figure 6, however, the Gershwins still showed

clear signs of development. So it seems that George’s teaming up with Ira resulted in “new

learning,” which quickly outstripped the experience he had acquired over the early years of hit-or-

miss partnerships.

Discussion

As in the careers of Rodgers, Hart, and Hammerstein, the Gershwins showed evidence of

development throughout the formative years of their collaboration. Despite some variation, the

early collaborative career of the Gershwin team was marked by initial lulls in hit ratio, followed

by substantial increases. That result is consistent with results reported by Hass and Weisberg
(2009), and more generally supports the 10-year rule of creative development. Comparison of the

two curves in Figure 6, and the analysis comparing the years in which George worked both with

Ira and other lyricists, illustrate that George’s work before teaming up with Ira lacked the

consistent success rate of their collaborative efforts, and is further evidence of a long period of

creative development for Gershwin as a composer of popular music.

General Discussion and Conclusions

Two of the three teams examined (the Gershwins and Rodgers and Hart) exhibited increasing

hit-ratios throughout the formative years of their careers. The Rodgers-Hammerstein team

exhibited no such trend. However, when the careers of Rodgers and Hammerstein were
20

analyzed separately, increasing trends were found for each member, and it was concluded that

Rodgers and Hammerstein had developed separately. Thus, the data suggest that when the

members of collaborative songwriting teams begin their careers together, the team develops

creative expertise as a unit. However, if collaboration begins subsequent to individual

development, further development may not be observed.

There are similarities among the careers of the Gershwins, Rodgers and Hart, and Porter, and

Berlin that are worth noting. Taken together with the examinations of Rodgers and Hammerstein

separately, there is mounting evidence that the development of high-level creativity in writing

popular songs is similar to the development of high-level performance in various other domains.

Several authors have made the claim that creative individuals should be viewed as high-level

performers in a domain that requires successive creative acts (e.g. Ericsson, 1999; Hayes, 1989;

Weisberg, 2006). The positively increasing trends found across the early careers of the Gershwins,

Rodgers, Hart, and Hammerstein bear striking resemblance to the development of musical

performers (c.f. Howe, Davidson, & Sloboda, 1998), chess players (c.f. Ericsson & Chamess,

1994), and more generally support the ideas presented by the expertise- based view of creativity.
Collaboration as Something More than Conjoint Expertise: Collaboration as a Social

Relationship

Our analyses of long-term collaborative teams have uncovered at least two different types of

collaborators: (1) collaborators with no experience prior to the collaboration (Rodgers and Hart;

Ira Gershwin); (2) collaborators with experience in multiple collaborations prior to a longterm

collaboration (Hammerstein pre-Rodgers; George Gershwin pre-Ira). For the former, there was an

initial period with no hits, followed by a relatively quick increase in hit ratios (see Figure 2

(Rodgers and Hart) and Figure 5 (the Gershwins)). For those collaborators who participated in
21

many collaborations, there was an increase in success over time, but that increase was much

smaller than that attained by comparable long-term collaborations. If we assume that those results

are not due to the specific individuals involved, they lead to the question of what might be

happening in long-term collaborations that sets those interactions apart from multiple short-term

collaborations.

One possibility is that the social relationship between the two team members plays a role in the

success of a collaboration, and here long-term collaborations might have an advantage. For

example, long-term team members might learn ways of constructively criticizing each other’s

contribution, which might result in an increase in quality of the output of the team. Perhaps when

composers and lyricists change partners frequently, there is a “distance” between the members of

each pair, which might preclude an open exchange of ideas. A long-term collaboration is also a

social interaction, and aspects of that interaction might be important. Ira Gershwin’s relative lack

of success after George’s death also indicates that long-term collaboration may not result in skills

that are universally transferable.

The Question of the Relation Between Quantity and Quality

One analytic issue deserves brief mention here. Simonton has presented positive correlations
between quality and quantity of work over a creator’s career as support for the chance-

configuration view (c.f. Simonton, 1977; 1999). In the analyses carried out here, we have found

inconsistent relationships between quality and quantity of output (see also Hass & Weisberg, 2009;

Kozbelt, 2004; 2008). That lack of consistency leads one to question the usefulness of the

correlation between quantity and quality in discriminating between the chance- configuration and

expertise views.
Future Directions
22

Assume for the sake of discussion that the present results and those of other studies provide

support for the hypothesis that composers and lyricists need significant amounts of time to learn to

write hits. We have collected no evidence concerning what those composers and lyricists might be

learning in those formative years of their careers. Simonton (e.g. 1980; 1994) has attempted to

account for the aesthetic success of over 15,000 classical music themes by examining the

originality of each theme, based on first six notes of each theme. However, Simonton’s analysis

did not examine any music-theoretical features of the themes that have contributed to their

success. It might be informative to examine the use of various melodic patterns or formulas (c.f.

Weisberg, et al, 2005) as well as harmonic structures (i.e. chord progressions) throughout the

careers of composers. One might also examine changes in melodic formulas over the course of a

composer’s career. Also, research on the perception and cognition of music has revealed that many

parameters of musical stimuli are salient to music listeners. For example, melodies that conform to

listener’s expectations can be more easily perceived and conceptualized than melodies that violate

expectations (c.f. Dowling, 1991; Huron, 2006; Krumhansl, 1990; Narmour, 1990; Temperley,

2007). Hit songs may feature more memorable melodies while non-hits may be more mundane

and, thus, less memorable. In addition, music- theoretical information can provide psychologists

with coding systems that would allow for the analysis of musical works. A similar set of analyses
could be carried out on the lyrics of songs (e.g., Simonton, 1989; Martindale, 1990), to determine

factors that change over a lyricist’s career and that might contribute to the production of hits.

Despite the limitations just discussed, the current study has provided more substantial evidence

that musical creativity develops in a fashion similar to domain-specific expertise. The view that

creativity is made possible by the normal cognitive processes of individuals with a high level of

expertise in a particular domain has the potential to enhance educational practices in

23 addition to providing valuable information to professional and amateur songwriters. That is,

if a creative individual understands the normal thought processes involved in high-level creativity,

he or she might apply such knowledge to individual creative endeavors.

You might also like