Research

A deep dive into every document behind the charts.

Search the corpus

Find any word or phrase across 30 million words.

Search the full text of every sermon, publication, and book in the corpus. Results show keyword-in-context snippets with source links.

Fair-use snippets only. Our search tools are not designed to reconstruct full documents and are rate-limited to prevent it. Snippets are truncated for copyright; follow the source link for the complete text. Citation links are provided for you to copy, not click — we do not guarantee any external link will remain live. The scholarship is ours; the preservation is not.
Era:
❦

Source coverage

What each source covers, across the record.

A bar per source, drawn between its earliest and latest document in the corpus. Hover a bar to see the exact span and document count. The modern era fans out because more sources are active at once; the early record runs through a narrower pipe.

❦

Where every word comes from

The frozen corpus behind every number on this site.

Every count, rate, and percentage on TrendZion is derived from a frozen corpus snapshot. Counts do not change between releases — when we add material, we cut a new release. The current release is shown below.

Loading the catalog…
❦

The word inspector

Every sermon that used a word.

Pick a word. We’ll list every document in the corpus that contains it, grouped by source, sortable by count or year. Click any row to expand the passages in context — 25 words before, the word itself, 25 words after.

❧

The library

Books by prophets, apostles, and seventies.

Books published by leaders of the Church, 1835 to present, whose full text is loaded into our corpus. Prophet works in green, apostle works in gold, seventy works in indigo.

Year Title Author Calling Status
❧

What’s missing, and why

An honest list of the corpus’s holes.

This is the section that matters most to careful readers. A corpus that claims to be complete is usually lying. These are the things we do not have, that we know we do not have, and that would change the picture if we did.

Time-range gaps

  • 1845–1852, eight years with almost nothing. The Joseph Smith Papers close at his death in June 1844; the Journal of Discourses does not begin until 1853. The exodus, Winter Quarters, and the first years of settlement in the Great Basin are essentially absent from this site.
  • Before 1828. The earliest Joseph Smith Papers document we have is from 1828. His childhood and the events leading to the First Vision in 1820 are not in the corpus because they were not recorded contemporaneously.
  • 1887–1896. Reliable Conference Report scans on the Internet Archive begin in 1897; the Journal of Discourses concluded in 1886. The decade between is thin, with only scattered material we were able to ingest.
  • The current moment. The corpus is refreshed with each General Conference cycle, but whatever was published in the last few weeks may not yet be here.

Scope — we have only what was centrally published

  • No local teaching. Stake conferences, sacrament meetings, devotionals, BYU speeches, firesides, and the week-in-week-out teaching across thousands of wards are not in the corpus. This site reflects the published top of the pyramid only.
  • No lesson manuals or study guides. Come, Follow Me, Teachings of Presidents of the Church, Preach My Gospel, Gospel Principles, institute manuals, seminary material — none of it. These may be the largest body of prophetic language that a typical member actually encounters, and we have none of it.
  • No Ensign, Liahona, or Church News feature articles, only talks proper.
  • No interviews, press statements, or First Presidency letters, unless the specific item happens to appear in one of our four sources.
  • Women’s voices are under-represented by institutional history. Until the women’s session was folded into General Conference in 2014 — and integrated further afterward — women rarely spoke at the events the corpus draws from. The raw counts look like a gender gap; the cause is organizational history, not our filtering.

Quality — data you can see through

  • OCR noise in the 1897–1970 Conference Reports. These come from Internet Archive scans and the quality varies year to year. Some years are clean, others have glued-together words, missing letters, and spurious hyphens from column breaks. We did not correct them by hand.
  • Two versions of the Journal of Discourses exist. The Wikisource transcription (faithful to the 26-volume original) and the Church’s internally curated excerpts. We use the Wikisource original. Passages may read more rough-hewn than in modern Church publications.
  • Non-English content is absent. Since the 1960s, some conference talks have been delivered in Spanish, Portuguese, and other languages and then translated into English. We have the English of record but not the original delivery.
  • Undated items. Four hundred seventy-one documents in the underlying database — mostly early Journal of Discourses sermons whose date was not transcribed — lack a year and therefore do not appear in the year-by-year charts. They are in the word counts behind the scenes but cannot be plotted.

Methodology

  • Matching is by pattern, not by meaning. When we count “keys” we catch every literal instance of the word. We do not distinguish “keys of the priesthood” from “keys of knowledge” without reading the surrounding context — which is exactly what the concordance view above is for.
  • Merged concepts (★) are editorial. Several rows in the top 40 combine related words under a single concept (for example, Christ combines “Christ, Jesus, Savior, Saviour, Redeemer, Jehovah, Messiah”). These merges are hand-curated and visible on hover on the landing page. They are our call, not a property of the data.