T-Pattern Detection and Analysis (TPA) With THEMETM: A Mixed Methods Approach

Abstract

This work, which was started in the early 1970s, was inspired by social interaction analysis based on direct observation and careful coding of behaviors according to a list of behavioral (mostly ethological) categories, especially the ethological work of N. Tinbergen, K. Lorenz, and K. von Frisch, for which they shared a Nobel Prize in 1973 in Medicine or Physiology but also H. Montagner’s ethological analyses of interactions in social insects and children. S. Duncan’s psychological and linguistic research on turn-taking in human interactions provided great inspiration, and so did Chomsky’s work on syntactic structure and Skinner’s probabilistic real-time functional analysis and their consequent debate. A hypothesis concerning numerous kinds of temporal and spatial natural and especially biological structures, the T-pattern is a hierarchical self-similar fractal-like structure that recurs with significant translational symmetry on a single discrete dimension, initially real time. It also points to profound self-similarity across many levels of biological spatio-temporal organization, as it seems characteristic of molecular structures such as genes and a multitude of recurrent motives on DNA and its 3D generalization corresponding to (3D) folded proteins. Developed initially to facilitate empirical analysis, the T-pattern and its detection algorithms were first presented in AI (Magnusson, 1981) and Applied Statistics (Magnusson, 1983) through THEME (3 k Fortran IV) software using an evolution algorithm. It is now over 300 k lines of code, runs under Windows, and, more recently, uses parallel processing for increased speed. This has allowed abundant detection of hidden structure in numerous kinds of biological phenomena at highly varied scales, from human behavior at timescales of days (Hirschenhauser et al., 2002; Hirschenhauser and Frigerio, 2005) to interactions of many individual neurons simultaneously registered at a temporal resolution of 10–6 s in neuronal networks in rat brains to ongoing work on T-patterns in DNA molecules at a spatial nano-scale. T-pattern detection and analysis (TPA) thus mix qualitative and quantitative analyses, as T-patterns themselves are artificial categories composed of recurring coding categories with special real-scale statistical relations between their instances. After their detection, T-patterns are thus analyzed much as are other behavioral categories.