![Sample our Education journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5932/TOC-Subject-Sample-2021-Education.jpg"})
Abstract
The purpose of this research was to study the creative process in real-life settings and in a multicomponential perspective of creativity. Relations between the creative process and other important variables (mood, personality, and the creative product) were investigated. The data analyzed were collected in four applied art schools (n = 41) in Switzerland, during mandatory workshops. The creative process was operationalized through 2 subprocesses: generation (idea production) and selection (idea evaluation), both repeatedly measured across the workshops. The trajectories of theses subprocesses were modeled with a commonly used statistical model for longitudinal data (Latent Growth Models). Results showed that generation had an overall decreasing pattern through time, whereas selection had an inverted U-shaped pattern. Important individual differences in both subprocesses and related variables were also found, many of which had strong predictive validity. Indeed, process and personality variables explained 70% of the variance of the evaluation of the final product.
Notes
1To avoid confusion between these two related notions, the term subprocess will be exclusively used to mention such specific cognitive functions.
2Second-order factors are factors that load on other, lower order factors.
3Generation and exploration has been extensively studied by Finke and colleagues (e.g., Finke et al. Citation1992), but in experimental settings. And some longitudinal study of creativity and affect does exist (e.g., Amabile et al., Citation2005), but with a focus on idea production and no measure of any variable similar to selection subprocesses. Hence, to our knowledge no study combines the generation and selection framework as synthesized by Bink and Marsh (Citation2000) with a longitudinal, ecological approach.
Note. n = sample size; t = number of measurement waves; SD = standard deviation.
4The construction rational and empirical testing of these scales are described in detail in Fürst (Citation2012); only a brief overview is presented here. The Big 5 scales were a French adaptation of the best marker in English (e.g., John & Srivastava, Citation1999; Saucier, Citation1994). Tested in a first sample of undergraduates (n = 111), the reliability of these scales was satisfactory (mean Cronbach alpha's = .72) and the factor correlation pattern between factors was similar to that of the classical Big 5 instruments. Recent retesting of these scales in another undergraduate sample (n = 254) showed good convergent validity with the NEO-FFI (Costa & McCrae, Citation1989): the correlations between the analogous factors of the two scales were about .90 (except for extraversion, which correlated at .75, probably because our scale focuses more on sociability and less on energy than that of the NEO-FFI's).
5A more natural scaling of the time process was also tested. It consisted of respecting the unequal intervals of time between repeated measures. In the photography workshops, this means that rather than using a fixed interval of 8/10 = .8, intervals were used that respected the actual timing of data collection. The two measures a day were collected approximately at 12 noon and 3 pm, which correspond to the 12th and 15th hour of the day. This second scaling hence attributed the values 12/(24*5) and 17/(24*5) to the data of the first day, (12 + 24)/(24*5) and (17 + 24)/(24*5) to those of the 2nd day, etc. This second type of scaling was statistically inferior to the first (i.e., model fits were worst), hence it was not used.
Note. parm. = parameter; est. = estimation; s.e. = standard error of estimation; p-val. = p-value. @0 = parameter fixed at zero (and not estimated).
Note. s.e. = standard error; iiv = intraindividual variability. The group of reference for this analysis is the designers' (i.e., the effect of “decorators” and “illustrators” predictors represent the differences between these groups and the designers).
6The quadratic slope of generation was not included in this model because of its extreme collinearity with the three other process variables (about 99% of shared variance).