The Coffee Shop

network analysis · sociology · political science · genealogy

~39,500 chairs · 5,500 heads of school · 1950 → 2025

A small voice in the dark moves more than it knows. Chirp anyway.

Imagine a coffee shop that opened in 1950 and never closed.

Every chair is a sociologist or political scientist. Some have been there since the doors opened — Coleman in one corner, Parsons at the bar in the early years, Merton in the back booth, Massey by the window watching people walk past. Others arrived later — Mendelberg ordered her first espresso in the mid-90s and has been here ever since. The newest baristas just clocked in last year.

The shop has ~39,500 chairs and they're all full.

Most conversations are citations — someone overhearing what someone else said and writing it down. But some conversations are something rarer: the kind where one scholar trains another, where a sentence in 1953 ends up shaping a sentence in 2018 because the second person learned how to think from the first.

This page is a map of those conversations.

What you're looking at

~39,500 dots. Each dot is an author who published in a flagship sociology or political-science journal between 1950 and 2025, or meets the citation-impact bar in the expansion crawl.
Vertical position = the year of their first publication. 1950 at the top, 2025 at the bottom.
Color = field & tier. Reds for sociology, blues for political science. Darker = top-flight journal.
Lines = mentor → mentee edges.
- Bold purple — Wikipedia-verified advisor/student pairs from top-500 author bios. Gold-standard.
- Thin gray — inferred via Sinatra-style co-authorship asymmetry (≥5yr gap, junior's early career).
Big black-outlined dots are "heads of school" with 25+ descendants. There are 5,517 of them; the largest single lineage (Peter H. Rossi) spans 4,491 descendants.

Click any dot to isolate that scholar's intellectual lineage — mentors in purple flowing up, mentees in green flowing down.

A note on missingness

Anglophone journals only. Bourdieu doesn't have a chair here. Habermas doesn't either.
Co-authorship inference misses solo influence. Coleman, Granovetter, and Tilly trained whole generations through reading, not co-writing. Their footprint here looks smaller than their actual influence.
Recent students are sparse. Famous mentors have 40+ named PhD students on their CVs, but many haven't yet accumulated enough citations to land in the universe. They'll show up in 5–10 years.
Pre-1980 author disambiguation is noisier. OpenAlex's record-linking gets more uncertain for older publications, so some of the largest pre-1980 trees may absorb name-collisions ("John Smith, 1955" vs "John Smith, 1985"). Spot-checking the very largest trees is recommended before drawing strong conclusions about any one descendant chain.

The viz makes this gap visible. Click any famous-but-childless author and you'll see the citation-vs-genealogy callout.

Method, briefly

Authors + citations + journals: OpenAlex Works API; curated 87-journal whitelist, top-cited works 1950–2025.
Inferred mentor edges: OpenAlex co-author pairs in same whitelist; Sinatra-style asymmetry rule (≥5yr first-pub gap, early-career joint paper).
Verified mentor edges: Wikipedia infobox fields (doctoral_students, doctoral_advisor, notable_students) on top-500 authors.
CV-verified edges (rolling): Direct CV scrapes for non-Wikipedia authors; regex on "DISSERTATION ADVISING" / "PhD Students" sections, role-tagged (chair / cochair / committee).
Layout: Width-weighted DFS tree rooted at earliest-mentor per component; Y = first publication year; tilt-fold for over-wide subtrees. Python + NetworkX + Plotly Scattergl.

The whole pipeline is resumable. Every time I sit at the coffee shop I add a few more chairs.

Latest session — 2026-05-10

16 CV-verified mentor → mentee edges, sourced from CVs, lab pages, and dissertation pages of an initial 17-scholar Latino-politics roster plus three calibration anchors (Mendelberg, Meier, Pérez).

Tali Mendelberg → Claire Willeck (chair, Princeton). Diss: Active Civics: How Civic Education Shapes Political Engagement. Won 2024 APSA E. E. Schattschneider Award. Committee: Mendelberg, Markus Prior, Andy Guess, Rocio Titiunik.
Efrén Pérez → 5 current PhDs in REPS Lab at UCLA: Ramona Alhambra, Gustavo A. Mártir Luna, Kasheena G. Rogbeer, Sydney Tran, Tricia Huynh.
Michael Jones-Correa → Angie Ocampo-Roland (chair, Penn 2021) — inferred, pending CV verification.

One correction worth flagging. First-pass surface search guessed Jason Casellas as Sam Chapa's dissertation chair at Houston. Her actual chair is Lydia Tiede. The verification pass caught it from her dissertation page — exactly the failure mode the audit step is designed to catch.

187 dissertation-cited authors extracted from the Worked Cited section of my dissertation. 12 of the 17 Phase-1 roster members appear in it; 25 cited-3×-or-more authors are Tier-2 promotion candidates (Barreto leads at 8, Segura 6, Masuoka and Pérez 5).

Expansion crawl — 2026-05-10

Sat down at the coffee shop and counted everyone in the room. Filter: US-affiliated authors with i10-index ≥ 2 publishing in any of 121 OpenAlex topics under "Political Science and International Relations" + "Public Administration."

Crawl pulled 43,953 unique authors. Three filter tiers:

Broad — 43,423. Max recall; includes cross-discipline scholars.
Subfield-proxy — 20,466. Top topic in PoliSci/IR/Soc+PoliSci/PubAdmin.
Concept-tight — 9,724. Top concept = Politics / Political science / PubAdmin / IR. Top 15 by h-index all real political scientists.

Top of the concept-tight list: Fukuyama, Keohane, Norris, John C. Campbell, Ken Meier, Bozeman, B. Guy Peters, Skocpol, Moynihan, Erikson, Iyengar, John W. Meyer, Tarrow, Apple.

Top institutions: Yale (87), Columbia (81), Wisconsin (76), UNC + UT Austin + American (73), Georgetown (71), NYU (70), Stanford + GW (69), Ohio State (67), Princeton + Cornell (66), Duke + Penn State (65), UCLA + Rutgers (62), Hopkins (61).

What this enabled. Originally ~7,000 dots filtered to a 87-journal flagship whitelist. Expansion added authors who publish outside that whitelist but still meet the impact bar.

Push to 1950 — 2026-05-12

Extended the corpus back another 30 years. A delta pull on the 87-journal flagship whitelist returned 85,472 pre-1980 works and 42,196 unique pre-1980 authors. After merging with the existing 1980+ universe and dropping junk records, the author dataset grew from 15,787 → 40,012.

Re-trimming the top-10,000 master tier brought in 1,022 senior figures who didn't have enough post-1980 citations to qualify before — including Barney Glaser + Anselm Strauss (grounded theory), Harold Garfinkel (ethnomethodology), Peter L. Berger, Howard S. Becker, Alvin W. Gouldner, Thomas Luckmann, Rosabeth Moss Kanter, Otis Dudley Duncan, Elihu Katz, Fritz Heider. A focused coauthor crawl on those 1,022 newcomers added their lineage edges to the graph.

Final shape: ~39,500 dots plotted, 5,517 heads of school, ~54,400 visible mentor → mentee lines. Largest single lineage tree spans 4,491 descendants (Peter H. Rossi, sociology of evaluation research). The 1950s, 1960s, and 1970s decades now anchor the top of the canvas; Coleman, Parsons, Merton, Lipset, Dahl appear where they belong.

The delta approach saved hours. A full re-pull of the 1950–2025 range would have taken ~3–5 hours; pulling only the 1950–1979 slice and merging it into the existing dataset took ~15 minutes. The bobbish principle in action: don't fetch what you already have.

Flagged for soon, not now

Non-US scholars. UK + Canada + Australia (~half the US count), Continental Europe (~similar magnitude), Latin America, East and South Asia, Africa. Going global multiplies the universe by roughly 2.5–3×, with uneven OpenAlex coverage outside the Anglophone core.

Books in the corpus. Currently the pipeline is journal-only. Adding monographs would surface scholars whose work lives mostly in books — Skocpol, Putnam, Theda Bell, Lipset — and would lift book-forward fields like political theory. Plausible 30K–60K additional title-level records to integrate.

Click-into-lineage UI. The current viz isolates one scholar's lineage on click. The next step is a side-panel "neighborhood" view that pulls grad students, coauthors, and book authors for the clicked scholar from a lazy-loaded per-author JSON. Architecture is planned; data scaffolding is in place.

The interactive network visualization has been retired.