Network analysis reveals insights about the interconnections of Judaism and Christianity in the first centuries CE

The study investigates how Jews and Christians in late antiquity were interconnected through shared literary traditions, challenging the older "parting of the ways" model that assumed clean separation after an early split. The authors propose that textual parallels and literary interactions can serve as proxies for historical connections between religious communities. They introduce network analysis—a method widely used to model complex systems—as a tool to visualize and analyze inter-religious literary connections across centuries and geographies. Focusing on the Babylonian Talmud (3rd–7th centuries CE) and select Palestinian rabbinic sources alongside Christian texts from the 1st–6th centuries CE, the paper aims to assess Jewish knowledge of Christian traditions, characterize the nature (polemical vs non-polemical) of engagements, and identify influential texts (nodes) within this inter-religious literary network. By mapping temporal and spatial relationships, the authors seek to reveal patterns and generate new directions for research on Jewish-Christian interactions.

The article surveys two strands of prior scholarship: (1) Earlier approaches to Jewish-Christian relations often ignored Christian parallels in rabbinic literature or focused primarily on Palestinian rabbinic works, leading to an underappreciation of broad, multi-layered interactions. Recent studies have demonstrated varied literary interactions, including polemics, shared concepts, appropriations, and parodies (e.g., Boyarin, Kalmin, Schäfer, Zellentin, Bar-Asher Siegal). (2) The use of networks in humanities and religious studies has grown, including social and semantic networks (e.g., Clark; Schor; Elwert; GEHIR project). Within rabbinic studies, scholars have modeled social circles, created visualizations of geography and economy, and built citation networks (Hezser; Lapin; Satlow & Sperling). However, prior network studies typically focused on persons or places, not on inter-religious literary interactions based on textual parallels. The authors identify a gap: no previous study has used networks to represent and analyze literary interactions between Jewish and Christian traditions by modeling shared motifs and references across corpora.

Design: The authors created small-scale, exemplary inter-religious networks in which nodes represent literary traditions (motifs reflecting theological, practical, or hagiographical aspects) identified in rabbinic or Christian sources, and directed edges represent demonstrated familiarity by one corpus with traditions of the other. Edges are annotated by type (polemical vs non-polemical), directionality (source of familiarity), and weight (level of certainty). Nodes are annotated by religious origin, geography, and time. Visualization strategies include polar/time-circle and spatial-temporal layouts with color-coding and edge thickness for certainty. PageRank (Google’s algorithm, implemented in Matlab) was used to rank node influence in one network. Data sources and selection: The primary Jewish corpus is the Babylonian Talmud (BT). Two main Talmudic cases were chosen: (1) BT Hullin 87a (from the heretics narrative corpus engaging biblical exegesis), and (2) rabbinic traditions engaging contemporaneous Christian monastic traditions (as identified in Bar-Asher Siegal, 2013). Additional rabbinic parallels include the Palestinian Talmud (PT) and Avot deRabbi Natan (ARN), selected based on analogous literary topoi. Christian sources span the New Testament and later authors (2nd–6th centuries), including Eastern and Western traditions (e.g., Cyril, Ambrose, Basil; Apophthegmata Patrum, Jerome, Paphnutius). Edge weighting and certainty: Three levels of certainty (visualized as solid, dashed, finely dashed lines) reflect the strength of evidence for shared traditions, derived from philological analysis and prior publications (Bar-Asher Siegal 2013, 2019). Examples: a strong shared tradition ties BT Hullin 87a with Christian debates about the Holy Spirit connected to Amos, involving Cyril, Ambrose, Basil; weaker connections include a 3-day fast motif paralleling pre-Easter fasts; very low certainty connects BT Hullin and BT Shabbat via the term euangelion but with different meanings. In the monastic network, all edges were assigned equal weights due to higher and more uniform certainty (later dating, distinct monastic literary features). Directionality generally points from Christian to Jewish sources in this study due to its focus on identifying Christian traditions within rabbinic texts. Annotations and visualization: Nodes color-coded by religious origin and geography; edges color-coded by polemical content; nodes positioned along temporal and spatial axes to reveal dynamics; edge arrows encode direction and weight. For the monastic network, PageRank quantified the influence of nodes as hubs of shared tradition. Scope and availability: The networks are intentionally small (proof-of-concept). All data to reconstruct the networks are contained within the visualizations and supplementary text (Supplementary Text 1) that lists the connections and justifications.

- Hullin network (BT Hullin 87a): Visualizing connections reveals BT Hullin 87a as a hub aggregating multiple Christian traditions. The temporal positioning—particularly references to 4th–5th century debates about the Holy Spirit (Cyril, Ambrose, Basil)—demonstrates that BT Hullin 87a post-dates the emergence of those debates (cannot be earlier than 4th century CE). Color-coding by interaction type shows a pattern: engagements with earlier Christian sources (e.g., New Testament) are predominantly polemical, while later Christian traditions elicit non-polemical references. A mixed edge (BT Shabbat 115a on euangelion) straddles both polemical and non-polemical meanings, aligning temporally with the hypothesized transition from polemical to non-polemical engagement. Geographical coding indicates earlier Christian traditions referenced are primarily Eastern; later traditions come from both East and West. - Monastic network: Apophthegmata Patrum serves as a major hub, with multiple rabbinic passages showing familiarity with monastic traditions. PageRank results (node ranks) emphasize Apophthegmata Patrum as most influential (0.3683), followed by Paphnutius/Jerome (0.1737), while several Talmudic and rabbinic nodes share equal lower ranks (0.0763: BT Avodah Zarah, Avot deRabbi Natan, BT Berachot 18b, BT Yoma 87a, BT Shabbat 33b, PT Nedarim 11). Both Palestinian and Babylonian rabbinic sources exhibit awareness of Eastern and Western Christian monastic traditions, and shared familiarity suggests possible intra-rabbinic connections. - Weight distributions: The Hullin network shows a wider edge-weight distribution than the Monastic network, reflecting more variable certainty in the former versus uniformly higher certainty in the latter. - Overall patterns: Network analysis reveals a temporal shift from polemical to non-polemical attitudes, a temporal-spatial familiarity correlation (early Eastern; later East and West), and identifies influential Christian texts as key hubs for rabbinic awareness.

The findings address the central question of Jewish knowledge of Christian traditions by providing a visual, quantitative-qualitative framework that captures the complexity of inter-religious literary interactions. Networks enable detection of influential texts, types of connections, and temporal-spatial patterns at a glance—capabilities difficult to achieve through traditional narrative analysis alone. The Hullin network supports dating inferences (terminus post quem based on 4th–5th century debates), and reveals a shift in rabbinic attitudes: early Christian (e.g., New Testament) engagements are largely polemical, while later engagements are often non-polemical. Geographic patterns indicate that references in rabbinic texts initially align with Eastern traditions and later encompass West as well. The Monastic network highlights the Apophthegmata Patrum as a key hub and demonstrates cross-geographical awareness by both Babylonian and Palestinian rabbinic sources. Together, these networks illustrate how inter-religious texts can serve as anchors for understanding intra-rabbinic connections and editorial relationships. The discussion emphasizes that, although small and exploratory, this approach can uncover novel patterns, propose new research questions (e.g., centrality, clustering, shortest paths), and serve as a generative framework to predict or infer unobserved links in larger future datasets. It also reflects on the methodological necessity of combining philological expertise with computational tools, especially given issues of certainty, dating, and sampling biases.

The study introduces and demonstrates a novel methodology—network analysis of inter-religious literary interactions—to map and analyze the complex, multi-layered relationships between Jews and Christians in late antiquity. Using two proof-of-concept networks (Hullin and Monastic), it identifies key hubs (e.g., Apophthegmata Patrum), reveals temporal and geographic patterns (shift from polemical to non-polemical engagement; early Eastern familiarity followed by broader East/West), and supports dating of rabbinic texts based on references to Christian debates. The approach offers a panoramic view that complements and extends traditional philological methods, generating new questions and directions for research. Future work should scale up data collection (potentially via computational text similarity tools), implement bi-directional networks, and apply advanced network analyses (centrality, clustering, path analyses) and generative modeling to simulate knowledge transfer and predict yet-unidentified connections between sources.

- Small, proof-of-concept networks may not represent the broader Jewish-Christian literary landscape; insights could be skewed by limited sampling. - Edge certainty relies on human philological judgment; assigning weights is debatable and not quantitatively calibrated. - Dating uncertainties: while some references permit lower bounds (e.g., 4th–5th century debates), the timing and routes of transmission into rabbinic texts often remain unclear. - Sampling bias: selection of parallels is guided by existing scholarship and identifiable topoi; historical focus on polemical passages may bias findings, though the study attempts to mitigate this. - Directionality is mostly one-way (Christian to Jewish) due to study design; bi-directional interactions remain to be mapped. - Automation limits: fully automated text-mining currently cannot replace nuanced philological evaluation of parallels; language diversity and subtle literary features complicate detection. - Geographical attribution reflects composition sites of Christian sources, not necessarily the locations from which rabbinic authors learned them.