Artistic style, marked by unique textures and patterns, has been studied across disciplines. However, previous research neglected quantitative, time-variant analyses of the making process due to technological limitations. While computers can synthesize artistic styles, studies often focus on pre-made products rather than the creation process itself. This paper introduces a computational method to evaluate the dynamics of personal artistic style, focusing on the transitions between motor states during the creation process. By tracking the movement of a sculpting tool, the researchers aim to identify clusters of motor features and analyze relationships between these clusters, providing insights into the 'language of making'. This approach allows for a quantitative observation of the entire making process, independent of the aesthetic qualities of the final product. This is computationally straightforward but nearly impossible manually due to the volume and resolution of data generated during creation. The study aims to provide both a computational method and a proof-of-concept, analyzing the unique and common motor features constructing an individual's artistic process. This enables quantitative observation of the making process as an interplay of various actions, irrespective of the final artifact's aesthetics.

The researchers review existing literature on artistic style, highlighting diverse perspectives. Davis defines style as a polythetic set of similar yet varying attributes in artifacts, suggesting that style explains attribute similarities. However, Davis’s analysis focuses on pre-made artifacts and lacks access to the casual motor features generating these attributes. The authors argue that to study style as a language, information about the temporal relationship between attributes, techniques, and gestures is needed. They discuss the limitations of current digital agents in co-creative processes, often losing crucial aspects of creative processes, and the need for a computational model of an artist's motor style for collaborative technology. The review also touches upon anthropological perspectives on style, noting Boas's focus on consistent aesthetic elements in culture and contrasting it with Malinowski's perspective of style projecting cultural signals onto material artifacts. Schapiro's definition of style as constant elements in the art of individuals or groups, and Wiessner’s connection of style variations to individual behavior and cognition, are also mentioned. The researchers then briefly discuss style's role in art history and the digital world, including digital tracking techniques connecting style to motor skills and the use of digital tools for stylization. Finally, they distinguish between 'dynamic style' as time-dependent motor acts in creation and 'static style' as the visual features of a finished artifact.

The study employed a magnetic motion tracking system (MMTS) to track the 6 degrees of freedom (6DOF) location of a carving knife at 60 Hz during clay-relief creation. Two studies were conducted: the first with five novice participants, each performing three simple geometric tasks; and the second with a single skilled artist creating four complex works using two different models. Data analysis involved continuous wavelet transform (CWT), and unsupervised machine learning (k-Means and Gaussian mixture model, GMM). Feature extraction focused on tool height, Euler angles (representing tool orientation), and high-frequency features indicating local tool behavior. These features were normalized before clustering. Two strategies were used for establishing the features space: GMM-based analysis and k-Means based analysis. The number of clusters for regular features and high frequency features was determined using Elbow and Silhouette criteria. The primary tool states were identified using k-Means, clustering the normalized feature space. Comparison between working sessions utilized the Jensen-Shannon divergence to quantify differences based on the distribution of working states and transitions between them. The analysis of the novice participants' sessions focused on identifying consistent working states among the participants. The analysis of skilled sculptor's sessions explored the differentiation between sets of techniques based on the subject matter, comparing techniques used in tasks involving flower translations and geometric patterns.

The study yielded two key findings. First, the analysis of novice participants' work revealed consistent individual motor skills despite their limited experience. The consistency was observed through the similarity in the morphology of the participants' motor action spans and the distances between these spans. This highlights the potential of the proposed computational method to identify personal style even in individuals with limited artistic training. Second, the study of the skilled sculptor's work demonstrated a correlation between the subject matter (flowers versus geometric patterns) and the motor techniques employed. The analysis showed a clear separation between sessions based on the subject matter, indicating that different models lead to different sets of preferred motor actions. This suggests that the analysis of motor actions during creation can provide insights into the cognitive processes involved in artistic creation. The findings were based on an unbiased analysis of the participants' motor actions, with correlations identified solely based on the sets of actions and their relationships. This indicates that groups of motor actions (techniques) as represented by technical features can be linked to higher-level representations and classifications applied by the artist.

The results support the proposed approach of observing style as transitions between working states, providing a summary of the making process. The findings highlight the consistency and individuality of makers' styles and the influence of different models and sub-techniques on the making process. The computational framework proved robust, showcasing the potential of the proposed approach. The main contribution is the definition of dynamic style as time-dependent motor functions producing static style in a finished work, enabling quantitative observation of the making process, independent of aesthetic qualities. The study opens up avenues for future research on the connection between time and style, style-technique-skill relationships, and the link between maker perception, image, and aesthetics. It also addresses the relationship between cultural and personal aesthetics.

This study successfully demonstrates the feasibility of analyzing artistic style through a computational approach focusing on the dynamics of motor actions during the creative process. The findings suggest that individual stylistic features are present even in novices, and that experienced artists employ distinct motor techniques based on their chosen subject matter. The methodology provides a rigorous, quantitative framework for future investigations of artistic style, potentially bridging the gap between the subjective nature of artistic creation and computational analysis. Future research directions include exploring the connection between time and style, the interaction between skill, technique, and style, and the influence of mental imagery on motor actions. The ethical implications of computationally modeling and imitating individual styles also warrant further discussion.

The study's limitations include the relatively small sample size, particularly in the novice group. The focus on a specific clay-relief technique and sculpting tool may limit the generalizability of the findings to other art forms and tools. The subjective nature of artistic style, influenced by various cognitive, emotional, and contextual factors, is only partially captured by the quantitative analysis of motor actions. The data concerning the artists' mental models and subjective associations were not directly measured but rather inferred from the analysis of the motor activities. Furthermore, the study focused on two distinct categories of subjects (flowers and patterns). A broader range of artistic subjects would be needed to achieve more generalized conclusions. Finally, only one skilled artist was studied in detail.