Design foundations: Towards a model of style grammar in creative drawing
A style grammar is a principled rule set that governs the organization of very complex ideas. It allows for the examination of underlying structures which are often times obscured. Style grammars have been developed for many fields such as writing, fashion and architecture but to date there is no style grammar for creative drawing. The research identifies the necessary visual features and core traits associated with each feature towards developing such a model for creative drawings. Then operational measures are defined using the computer to extract and measure the core traits of those features towards developing a model of style grammar in drawing. These visual features include line, tone, and depth. Core traits include line length, line width, line expressiveness, local tone, global tone, texture, pattern, outline, shape, and position. A multidimensional scaling (MDS) using input from 27 subjects, 10 art experts and 17 novices, supported the overall list of visual features and added the dimension of smudge to the list. A second MDS sort discusses issues with images and large art categorical sorts from the standpoint of both human perception and machine measures that were obtained using feature extraction. It was concluded from the results of the second MDS that large art categories were too broad to be useful in evaluating measures to develop the model. Further analysis was run using only drawings from three artists, two impressionists to compare similarity and one expressionist for dissimilarity to determine if the machine measures of the core traits of the visual features were able to differentiate smaller groupings of consistent drawing styles. Using the computer allowed for systematic and objective procedures to be used to obtain measures. The multinomial logistic regression showed high significance for all the traits except marginal significance for line length and no significance for depth. Binomial logistic regressions run on each pair of artists showed high significance for all the traits except depth. The combined positive results of the first MDS card sort and the binomial and multinomial regression analysis provide proof of concept and offer strong support towards the development of a model of style grammar for creative drawings. Implications for teaching drawing using the identified visual features and core traits are offered. The outcomes and analysis provided in this research currently support a general practice rule in design reuse and intelligent borrowing that suggests first smudge, then depth, then tone, and then line quality are the most significant elements to use for style comparison. Discussions for future research including improved measures and other types of perception testing are provided towards further development of the model.