Concept atlas
Seth's dictation spans twenty-two years, held in the corpus as 19 book divisions. Which concepts run from beginning to end, which belong to a single division, which year an idea first appears — these two charts lay that out.
All figures are computed from this site's concept annotation table (AI labels each block of original text with the concepts it discusses) — they are neither Seth's words nor literal word frequencies. Definitions and known biases are set out at the foot of the page underMethod and limits.
Concepts × books
The top 120 concepts by associated blocks × 19 book divisions. Colour is not occurrence count butdensity lift: how many times denser this concept is in this book than across the whole corpus. Read a row for one concept's distribution; read a column for what a book is about.
Timeline axis
The same top 40 concepts × 22 years (1963–1984). Each session's year is inferred by the five-level priority set out under "Method and limits" below. 1983 — 1 year in all — is a blank column because the corpus holds no dictated book from that year, not because nothing was said then.
Three rankings
These three rankings are computed overall 383 concepts, not only the 120 plotted above.
Universal concepts
The 15 most evenly spread across the 19 books (highest normalised entropy h; h=1 means perfectly even). These are the bedrock terms Seth uses from beginning to end.
| Concept | h | Blocks |
|---|---|---|
| spontaneity | 0.956 | 531 |
| simultaneous time | 0.945 | 735 |
| dream reality | 0.941 | 1273 |
| creativity | 0.940 | 1136 |
| intuition | 0.937 | 120 |
| death | 0.923 | 724 |
| psychic gestalt | 0.923 | 410 |
| inner self | 0.922 | 1221 |
| probabilities | 0.918 | 982 |
| inner reality | 0.916 | 178 |
| consciousness | 0.915 | 504 |
| value fulfillment | 0.908 | 494 |
| spacious present | 0.902 | 192 |
| illness and healing | 0.896 | 1448 |
| seth material | 0.896 | 68 |
Exclusive concepts
The 15 most concentrated (lowest h, with at least 30 associated blocks). Each belongs almost entirely to one book — its home ground.
Strongest pairings
The 20 concept × book pairs with the highest density lift. A lift of 69 means the concept is 69 times denser in that book than across the whole corpus.
Method and limits
Where the data comes from
The figures are computed from this site's concept annotation table chunk_concepts: each block of original text is labelled with the concepts it discusses (from two sources — extraction out of the block, and retrieval back by concept). So "associated blocks" here meansthe number of text blocks an AI judged to discuss this concept, not keyword occurrences: paraphrases count in, and so do the AI's mistakes. Only blocks with speaker='seth' are counted (9,948 in all); narration by Jane and Rob is excluded. The English original and the Chinese translation are merged into one book division, and blocks in both languages count, so a bilingual book holds roughly twice the blocks of a single-language one.
How the two numbers are computed
- · density lift =
(hits in this book ÷ blocks in this book) ÷ (total hits for this concept ÷ blocks in the corpus). 1 = as dense as the corpus average; above 1 = denser in this book. It cancels out book length, so a slim book such as The Magical Approach (88 blocks) can rank fairly. - · distribution entropy h = the concept's distribution entropy across the 19 books ÷ ln 19. 1 = perfectly even across all 19; 0 = entirely inside one.
- · Lift on the timeline axis works the same way with "year" in place of "book"; a year with no dictated-book material is treated as 0 (drawn blank) rather than dividing by zero.
How each session's year was decided
All 1,296 sessions, tried in the following order of priority, with the count reached at each level reported as it is:
- · Session date present in the corpus: 487 sessions
- · Parsed from the date heading on the first line (both English and Chinese headings recognised): 620 sessions
- · Borrowed from the same session's other-language edition: 84 sessions
- · Linearly interpolated by real session number between sessions of known date, then clamped into the book's period: 84 sessions
- · None of the above available, falling back to the book's opening year: 21 sessions(all of them chapters of the English Seth Speaks, which carries chapter numbers only and so offers no session number to interpolate on)
Known bad data and limits
- · Two session dates in The Nature of Personal Reality disagree between the Chinese and English editions (which one is right has not been checked, and no judgement is made here); one date heading in The Early Sessions Book 8 looks like a typesetting slip. None of the three has been corrected by hand, so the corresponding years are off by one cell.
- · 1983 — 1 year in all — is a blank column on the timeline because the corpus holds no dictated Seth book from that year (not because Seth was silent, and not because data went missing).
- · The heatmaps plot only the top 120 concepts by associated blocks; the three rankings above cover all 383.
- · The concept labels are AI-generated and will both miss and mislabel. Use these charts for trends and comparisons, not as exact counts; to get down to a specific sentence, open the concept page and read the excerpts.