Piano Repertoire Explorer

Try this: in the Genres panel above, click the ◎ next to waltz to isolate only waltzes, then press ▶ Play below — watch the most-published waltz composers rise and fall across the century. The same works for polkas, nocturnes, marches, or études.

Arrangement culture was central to domestic music: four-hand reductions of symphonies and operatic fantasias were how most households met the concert repertoire. A high arrangement share in a market is a sign of that market consuming "serious" music at home in domestic dress — not a sign of triviality.

Uses the current market/genre/gender/format filters. ~ of works in view have a publisher city that isn't yet in the normalization table (or none recorded) — grouped as "Other/unknown", not dropped. Publication place ≠ where the music was played.

What this is. The Piano Repertoire Explorer documents the piano music that was actually published and circulated in nine national/linguistic markets between 1820 and 1910 — German trade (Hofmeister XIX / RILM, the largest single source), France (BnF catalogue + Gallica), the UK (Jisc Library Hub), the USA (Library of Congress), Spain (BNE), Norway, Russia (NLR), and scattered others. It is built to counter a concert-hall historiography that remembers a handful of masterpieces and forgets the vast domestic repertoire.

What the numbers are — and are not. A piece is one work (one row of the master table); an evidence record is one documented sighting of it (a catalogue listing, a copyright deposit, a reprint, an advertisement). A piece can carry several records, though in this dataset most carry only one. None of it is a record of sales or of playing — publication ≠ sales ≠ playing. The German trade is far better documented than the others, so absence of evidence in a thin market is not evidence of absence. No "salon vs. serious" judgement is baked into the counts: the 264,096 works are all catalogued piano music in the window (solo, four-hand and two-piano), and how to carve that into salon, teaching, concert and other repertoires is a question the data is meant to help answer, not one it presumes. Out-of-scope records the catalogues mixed in — chamber duos (violin/flute/cello with piano), voice-and-piano songs, and Hofmeister's unresolvable "Idem…" follow-on lines — are excluded from display (~12,400 rows, also counting complete-opera editions and binder\'s "bound volume" shelf records), and Jisc records whose unknown dates parse as "1800" are treated as undated. A work first printed before 1820 stays only if 1820–1910 evidence documents its in-window circulation (dated to its first in-window sighting); pre-window library holdings are excluded.

The canon-formation layer (Trésor des pianistes, Peters Salon-Album, Pauer's Alte Meister, "The World's Best Composers", Bie's history) is kept strictly separate: those 635 rows are selections asserting what ought to be played, and they never enter the repertoire counts (a further ~170 anthology picks were works already in the market, kept on the market side and cross-linked rather than duplicated). Compare the two on the Canon vs Salon tab. Evidence classes: A documented quantities · B systematic publication proxies · C qualitative. Format & derivation classifications and city→country normalization are heuristic and labelled as such.

Notes for scholars.

The window effect. Documentation density grows enormously across the century: only 5,260 works carry evidence in 1829–33 against 22,820 in 1873–77. A "Top 20" placing in the early decades therefore rests on a far smaller documented market — Louise Farrenc's multiple placings around 1831–35, for instance, reflect genuine BnF/Gallica registrations of her early piano works inside a thinly documented window, not a mass market that later vanished. Treat early-decade rankings as rankings of what survives.

Composer-name consolidation. Names are normalized in the database and again at display: abbreviation variants merge only where unambiguous, and a curated table resolves bare initials for high-traffic composers (Schumann "R." → Robert; Cramer "H." → Henri; Krug "D." → Diederich; the full 370-candidate sweep is published as name_initial_sweep_v41.csv). Deliberately not merged: Schubert "F." (Franz vs the violinist François-Louis Schubert), the Strauss brothers under short prefixes, and Fahrbach Sr/Jr (the sources themselves don't say which). Farrenc "L." resolves to Louise on frequency (a four-row "Léonce" exists). Jean-Louis Gobbaerts and his pseudonym "Streabbog" (the surname reversed) are shown as one composer. "Camille Schubert" is a French salon pseudonym, unrelated to Franz.

Title families. The ranking views (Top 20, Pieces, Canon decades) merge cross-language editions of the same piece — Lieder ohne Worte = Songs without Words = Romances sans paroles — because translated titles otherwise split a hit's score against itself, punishing precisely the most-translated pieces. Merged this way, Mendelssohn's Songs without Words is the corpus's most-documented publication family (weighted score 253, four times Chopin's most-cited Nocturne). Browse always lists every edition separately.

Coverage gaps are real and instructive. Streabbog's La Violette op. 99 — one of the century's best-selling teaching pieces — is absent here because it circulated through Schott/Belgian channels our sources don't capture, while his 300+ French dépôt-légal pieces are present. Italy is absent entirely (its union catalogue is technically unreachable). The UK harvest covers roughly 28,000 of a ~100,000-record pool, which cautions against reading the thinness of, e.g., domestic Messiah/Hallelujah piano arrangements (~65 records here) as proof of absence in British parlors.

Against received wisdom. Cross-checked against the scholarship (Loesser; Temperley, ed., The Lost Chord; D. Scott, The Singing Bourgeois), the corpus confirms the canonical salon best-sellers — Songs without Words, the Prière d'une vierge, Aufforderung zum Tanz, Les cloches du monastère, the Home Sweet Home and Last Rose of Summer variation families, and the vast Trovatore-arrangement economy — but corrects the backward projection of today's canon: Für Elise barely registers (18 records; first published 1867), Sinding's Rustle of Spring (1896) and Sullivan's Lost Chord (a song) belong to other repertoires, and Kotzwara's Battle of Prague had spent its vogue before this corpus's window opens.

Ask Claude about this data. The Ask Claude ↗ button (top right) opens a Claude chat pre-loaded with instructions to fetch this site's machine-readable digest (corpus_digest_v41.json: totals, decade/genre/market distributions, top families, method caveats) so you can interrogate the dataset conversationally in your own Claude account. The digest link works once the site is hosted (e.g. GitHub Pages); from a local copy, point Claude at the hosted URL instead.

What this is. The Piano Repertoire Explorer maps the piano music people actually bought and played at home, 1820–1910: 264,096 published works across nine markets (German trade, France, Britain, the USA, Spain, Norway, Russia and others), reconstructed from national-library and music-trade catalogues — set against a separate canon-formation layer of just 635 anthology and revival selections that decided what "deserved" to survive (see Canon vs Salon). Standard histories remember a handful of concert masterpieces; this is everything else. One caveat governs every number here: publication ≠ sales ≠ playing — counts are circulation evidence, not proof of popularity.

How to use it. Start with Top 20 by Year (the opening tab): press ▶ Play on the bottom timeline and watch the most-documented pieces rise and fall across the century — or scrub to any year. The same timeline drives Time-Lapse, Geography and the Dissemination Map. Browse the Data is the full catalogue: search anything, filter by market, gender, format or genre (the checkboxes above work on every tab), and click any row to drill in. The FORMULA panel is the methodology made visible: it shows exactly how evidence types are weighted into a "popularity" score, and you can change the weights — or press ↺ Reset — to test your own assumptions. Composer Careers traces one person's documented output year by year; Women's Share gives floor estimates for women's music; Sources & Methods is the scholars' page, with full source criticism.

A few things the data shows. The most-documented publication family of the century is not a virtuoso showpiece but Mendelssohn's Songs without Words (weighted score 252 across 141 editions in three languages) — a reminder that the salon hits sat beside, not instead of, canonical works. Behind it come Bądarzewska's Prière d'une vierge, Weber's Aufforderung zum Tanz and Lefébure-Wély's Cloches du monastère — exactly the list the social histories describe. Meanwhile some received wisdom fails the evidence test: Für Elise barely registers (18 records; first published only in 1867 — its fame is a twentieth-century construction), and Sinding's Rustle of Spring (1896) had barely begun. At least 3,380 works here are by women — a floor, not a ceiling, since most composers remain gender-unresolved; our companion project Women at the Keys documents 477 women composers of piano music in depth.

Ask Claude about this data. The Ask Claude ↗ button (top right) opens a chat with Claude — in your own account — pre-loaded with instructions to fetch this site's machine-readable digest (totals, decade/genre/market distributions, top pieces, method caveats), so you can interrogate the dataset conversationally: "which genres grew fastest after 1870?", "compare the German and French markets in the 1880s." Claude sees the summary digest, not all 264,000 rows; for row-level questions, use Browse and its CSV export.