Network Analysis Finds Shifts in the History of Modern Architecture

1. Abstract

Architectural historical texts implicitly create networks of relations between the objects, people, and texts they examine. Histories of European modern architecture examine a consistent core of these entities. From each of four English-language histories of European modern architecture (by Le Corbusier, Reyner Banham, Anthony Vidler, and Jean-Louis Cohen), we construct a graph of the objects, people, and texts mentioned in the same context. Different modeling resolutions can be used; we add undirected, unweighted edges between all nodes that occur on the same printed pages. We construct our graphs at the page-level because the data comes from book indices. In actor-network theory, which is popular in many fields including architectural history, as in our work, both humans and cultural artifacts can be network constituents. Networks have been used in other digital art history, for example in social analysis and mapping projects. To our knowledge, ours is the first work on architectural history that explicitly constructs networks from texts and quantifies aspects of them.

These four graphs provide overlapping but distinct views of modern architecture spaced throughout the last century. Network analysis, in the form of eigenvector centrality and triangle counting, allows us to make observations about both the form and content of these histories. This process allows comparisons between how the different histories relate groups of objects, and we find that some of the texts reexamine objects in multiple contexts much more frequently. Our results compare and contrast how the works construct historical importance, which is determined through close readings of the histories in question and supplemented by other histories of modern architecture. We quantify the importance of different people, objects, and texts in each of the networks, providing an empirical comparison between stated notions of historical importance and importance derived from the structures of the texts. Network analysis provides additional evidence to augment rather than supplant traditional ideas of historical importance.

We find that certain architects that the histories portray as important, like Le Corbusier and Walter Gropius, are in fact the most central in the four graphs. In the later histories, we find that modern architects are more often mentioned with the work they influenced rather than with the works to which they understood themselves to be responding. For example, Le Corbusier wrote about his response to Greek architecture, while the later histories focus on his relation to buildings other architects designed during his early career. Methodologies for comparing networks derived from multiple texts are used to investigate architectural history in this case study, but they are also useful for any situation with multiple views of relationships between similar entities. We expect the methodology to afford larger historiographical studies in other areas of architecture and history.

The data is publicly available: https://github.com/gyauney/modern-architectural-history-indices

Gregory Yauney (gjy24@cornell.edu), Cornell University and David Mimno , Cornell University

Theme: Lux by Bootswatch.