A quantitative analysis of LDS General Conference discourse, 1942 to 2025. Seven metrics across 5,838 talks and 11.1 million words show language drifting toward mainstream Protestant patterns. Nelson is the largest single-era inflection point in 84 years of data, and he named the moment himself.
Each of the seven sub-metrics is rescaled 0–100 across all eras, then averaged. The composite ranges from 0 (peak LDS distinctiveness) to 100 (peak Protestant convergence). Read the methodology section below for the exact formula.
Same metric, zoomed to the Nelson presidency so the 15 numbered events have room to spread out. Each gold dot is hoverable for full details.
Hover any gold dot on either chart for full details. The Mar 18, 2026 opening of ward Sunday School president to women is post-Nelson and not shown on the chart.
Per-talk mention rate of the sitting president (matched by full name and "President <Surname>") versus mentions of "Joseph Smith" or "the Prophet Joseph." Quote-attribution rate counts mentions within 80 characters of a quoting verb (said, taught, wrote, declared, testified, counseled, etc.).
Distinctively LDS terminology (priesthood keys, plan of salvation, dispensation, sealing ordinance, etc.) versus generically Protestant phrasing (born again, personal relationship with Christ, by grace alone, etc.). Y-axis is the ratio of LDS-distinctive mentions per Protestant mention.
Same data, plotted as per-talk frequency rather than ratio. Both Mormon-distinctive and Protestant-style vocabulary increase under Nelson; the ratio narrows because Protestant grows faster, not because LDS shrinks.
Grace share of grace+works language. Grace counted as the words grace, gracious, graciously. Works counted via salvation-context phrases only (works of righteousness, good works, obedience, keep the commandments, endure to the end, etc.) to filter generic uses of "work."
Share of "godhood" language that uses hard-theosis or exaltation framing (becoming gods, celestial kingdom, eternal increase, joint-heirs with Christ) rather than soft-theosis ("Christlike," "divine nature") or Protestant-replacement framing ("personal relationship with Christ").
References to traditionally Catholic/Protestant liturgical observances (Lent, Advent, Holy Week, Good Friday, Palm Sunday, Ash Wednesday) in conference. Context-aware regex filters out the verb "lent," "advent of Christ" (the LDS sense), and "second advent."
Hard-materialist Godhead claims (separate beings, bodies of flesh and bone, three distinct personages, First Vision physicality) versus generic Christian-piety phrasing (personal relationship with Christ, come unto Jesus, trust in the Lord, my Savior).
For each apostle who has at least 5 talks both before and during the Nelson era, we compare their average grace-share, generic-piety-per-talk, and liturgical-references-per-talk between the two windows. This tests whether the cohort-level shift was driven by individuals changing their language, or by a different cohort speaking under Nelson.
Each bar is one apostle's change in average grace-share between their pre-Nelson talks and Nelson-era talks (positive = more grace language under Nelson). Counts of talks shown in tooltip. Speakers with fewer than 5 talks in either window are excluded. Listed in order of magnitude.
Every number on this dashboard is computed in the open. The corpus is the BYU Scripture Citation Index (8,834 talk IDs probed; 5,838 successfully parsed; 1942–2025; 11.1M words). Each talk is attributed to the LDS Church President alive on its conference date, not to its calendar year. Conference dates are the first weekend of April and October.
Concern: the composite Reformation Index could be sensitive to methodology choices. We tested 8 alternative weighting and inclusion schemes. Nelson ranks #1 across all of them, with composite scores ranging from 92 to 99.
Schemes 2-4 multiply individual metrics by 6 (vs default 1) so that single metric carries 50% of the composite. Schemes 5-7 exclude specific eras or metrics. Scheme 8 takes the median rank across single-metric rankings instead of using the composite.
The composite uses min-max scaling. The baseline is the range of values observed across the 11 presidencies Grant through Nelson. The Oaks era (34 talks in October 2025, post-Nelson) is computed and shown but is not part of the scaling baseline because it is a successor cell, not a peer presidency. If Oaks were added to the scaling baseline, Nelson's score would compress from 97 to 89, but the rank ordering does not change.
Regex (short for "regular expression") is a precise pattern-matching language used in software to find text. It is how this dashboard counts how many times a phrase like "President Nelson" or "personal relationship with Christ" appears in millions of words of conference transcripts. A regex like \b(grace|gracious|graciously)\b means "match the word grace, or gracious, or graciously, but only as standalone words" (the \b tokens are word boundaries that prevent false matches inside longer words like "disgrace"). Every metric on this dashboard uses regex to count specific patterns, with documented patterns listed in each method block below so anyone can audit or reproduce the counts.
Each of the seven sub-metrics is min–max scaled across the 11 prophet eras (Grant…Nelson). Sub-metrics with negative polarity (where higher value = more LDS) are inverted to 100−x. The seven scaled values are averaged with equal weight.
No hand-tuning, no era-specific weights, no hardcoded values. Formula is identical to the per-year Reformation Index shown on the chart.
Living-prophet regex: e.g., \b(?:President\s+)?Russell\s+M\.?\s+Nelson|\bPresident\s+Nelson\b. Joseph Smith regex: \bJoseph\s+Smith(?:\s+(?:Jr\.?|Junior))?|\bthe\s+Prophet\s+Joseph\b (filters out unrelated mentions of "Joseph"). Quote attribution = mention within 80 characters of a quoting verb (said, taught, wrote, declared, testified, counseled, etc.).
Denominator: per talk.
Mormon distinctives: dispensation, priesthood keys, plan of salvation, sealing ordinance, melchizedek priesthood, aaronic priesthood, first vision, urim and thummim, Book of Mormon, three degrees of glory, premortal, Council in Heaven, Word of Wisdom, etc. Protestant generics: born again, personal relationship with Christ, sola gratia / fide / scriptura, give your heart to Christ, accept Jesus, walk with the Lord, my Savior, etc.
Denominator: ratio of Mormon to Protestant counts per era.
Grace: \b(grace|gracious|graciously)\b. Works: salvation-context phrases only: works of righteousness, good works, by our works, keep the commandments, obedience, endure to the end, work out your salvation. Generic uses of "work" are excluded.
Reported as grace / (grace + works) percentage and as works:grace ratio.
Hard theosis: become gods, gods in embryo, the Gods, exalted being, King Follett, Snow Couplet. Exaltation: celestial kingdom, eternal increase, joint-heirs with Christ. Soft theosis: Christlike, divine nature, become like Christ. Protestant replacement: personal relationship with Christ, come unto Jesus, my Savior.
Reported as (hard + exaltation) share of all four godhood categories.
Lent: capitalized only, with verb-form filter (rejects "lent his hand," "had lent") and religious-context boost. Advent: capitalized only, rejects "advent of Christ" (LDS sense) and "second advent." Holy Week, Good Friday, Palm Sunday, Ash Wednesday: literal phrase match.
Total references and per-talk rate.
Materialism: body of flesh and bones, three distinct beings/personages, First Vision, two personages, embodied God. Piety: personal relationship with Christ/Jesus/God, come unto Christ, trust in the Lord, my Savior/Redeemer.
Per-talk rates, both lines on chart.
Each talk is attributed to whichever LDS Church President was alive on the conference date. Notable corrections vs the v1 dashboard:
1957 October conference is missing entirely from the BYU corpus. Some 1980s April conferences are short by 5–10 talks. BYU appears to exclude sustaining-of-officers, statistical reports, opening/closing prayers, and choir interludes. This is appropriate for word-frequency analysis. The v1 dashboard's count of 6,251 talks likely included non-doctrinal entries; the audited count is 5,838.
Every talk in the corpus is listed with its BYU URL so you can read the source.