Social Network Analysis
The network outputs in the shared folder include yearly and decade-based temporal summaries, cross-layer overlap measures, node-level centrality metrics, tie-strength summaries, and a set of fourteen static visualisations. Several broad patterns are especially clear.
First, the three layers expand at different historical moments. The coauthorship layer grows rapidly in the 1950s, especially in 1955 and 1957, reflecting Yang’s most active period of formal collaboration. The correspondence layer expands most sharply in the 1960s, with the largest single-year intake of new correspondents in 1965. The photo layer has its most dramatic expansion much later, especially in 1992, 1999, and 2002, capturing a more public and institutionally visible phase of Yang’s later life.

- Figure 2. Coauthorship grows earliest, correspondence accelerates in the 1960s, and the photo layer expands strongly in the 1990s and early 2000s.
Second, the overlap between layers is extremely low. The shared-node overlap measured by Jaccard similarity is only 0.035 between coauthorship and correspondence, 0.016 between coauthorship and photo, and 0.019 between correspondence and photo. In other words, each layer captures a largely distinct social world. The participation summary in the network outputs shows 392 relationships that appear in only one layer, 16 that appear in two layers, and only 1 that appears across all three.

- Figure 3. The low overlap confirms that publication data alone would miss most of the social and archival relationships visible elsewhere in the collection.
Third, the strongest recurring ties differ by layer. In the composite tie-strength table, the most prominent relationships are with T.D. Lee, T.T. Chou, Tai Tsun Wu, C.P. Yang, and Du Zhili. T.D. Lee and T.T. Chou dominate in coauthorship, Tai Tsun Wu is particularly prominent in correspondence, and Du Zhili is the strongest figure in the photo layer. T.T. Chou is especially important because this is the one tie that appears across all three layers in the processed network.

- Figure 4. The strongest ties do not come from one homogeneous network; they come from different layers with different social meanings.
These patterns support the central claim of the project. Formal collaboration, direct communication, and public or archival co-presence are related but not interchangeable. A single-channel model based only on publication records would flatten the complexity of Yang’s transpacific network history.