The ten tests are for revision, not composition. They rest on one procedure: read the join cold, covering the preceding unit and reading the transition and the next two hundred words as though the book began there, since a writer reading in sequence arrives already oriented and will never find the fault. Run tests 1 to 4 (Anchor, Dimension count, Launch and Technique fit) first, because there is no point tuning an echo at a join whose anchor never arrives. Record a sentence, not a verdict: “the anchor is a place-noun at word thirty-four” is useful, and “fails test 1” is not. On a full manuscript, run the transition inventory (in the checklist) first, then the tests on the dozen joins it flags. Each test is failed on purpose by a book worth reading.
Test 1 — Anchor
Read cold: at what word can you say when, where and whose vantage? Is any delay doing something?
Failure. An orphaned cut, unless the delay meets the three conditions of drift.
Exception. Benjy’s section of The Sound and the Fury.
Test 2 — Dimension count
How many of these change: time, space, protagonist, causality and intentionality?
Failure. An expensive gap treated as a cheap one, or the reverse.
Exception. A disorientation the book means.
Test 3 — Launch
What does the last sentence before the gap do for the far side?
Failure. The departure simply stops, so the arrival carries the whole bill.
Exception. A cut meant to surprise.
Test 4 — Technique fit
Which technique is this, and was it chosen or defaulted to?
Failure. A habit: usually the numerical time-stamp.
Exception. A book whose single technique is its form (Dracula).
Test 5 — Momentum
At what word does the new unit’s own business begin?
Failure. Throat-clearing: words telling the reader what she would have assumed.
Exception. A narrator whose company is the point (Tom Jones).
Test 6 — Gap content
Write in one sentence what happens inside the gap. Will the reader need it?
Failure. A false ellipsis: a consequence will later arrive without its cause.
Exception. An omission the book later declares (“Time Passes”).
Test 7 — Redundancy
Is the reader oriented more than once?
Failure. A double anchor or a recap.
Exception. A reader returning after a long absence (serial publication).
Test 8 — Simultaneity
Does the reader know whether this unit comes after the last one or alongside it?
Failure. Unmarked simultaneity: she will assume “after”.
Exception. A chronology meant to be unrecoverable (Pedro Páramo).
Test 9 — Echo
Could the join do a second job, and if it does one, will a first-time reader notice?
Failure. Inert connective tissue, or an echo too faint to land.
Exception. A book written for rereading (Ulysses).
Test 10 — Consistency
Across the manuscript, does the technique vary with the gap?
Failure. One technique in more than two joins in three.
Exception. The same as test 4 (Technique fit).
Four refinements sharpen the tests. In test 1 (Anchor), give the join to a reader who has not read the previous scene and ask her to mark where she knew where she was; her mark will usually fall later than yours. Test 2 (Dimension count) catches expensive joins that feel cheap because the prose is fluent, and its count should govern everything else: none or one dimension needs only a resumed referent, two a clear anchor in the first clause, three an explicit anchor settling at least two of them at once, and four or five a paratextual or compound anchor, or a rethink of whether the gap needs to be so expensive. Test 3 (Launch) most often solves a join that has resisted rewriting at the arrival. In test 8 (Simultaneity), write “after” or “during” in the margin at every handoff, then check whether the text says which; where it does not, one phrase (that same afternoon) usually suffices.
Reading the results together
Failures cluster, and the cluster points to the first move. Test 1 (Anchor) alone means an under-anchored arrival: put a stronger anchor type in the first clause. Tests 1 (Anchor) and 2 (Dimension count) together mean an expensive gap treated as a cheap one: add a paratextual or compound anchor, or move the gap. Tests 1 (Anchor) and 3 (Launch) together put the problem at the departure: rewrite the launching line to name the destination. Tests 4 (Technique fit) and 10 (Consistency) mean defaulting: re-cut six important joins with three different techniques. Test 5 (Momentum) alone means throat-clearing: delete downward from the gap until deletion causes loss. Test 7 (Redundancy) alone means a double anchor or recap: delete the second orientation and trust frame recall. Tests 2 (Dimension count), 5 (Momentum) and 10 (Consistency) together are not a transition problem but a structural one, and the gaps are in the wrong places, which is The Chapter Break’s business (Chapter 6).
Test 9 (Echo) is the one most likely to make a manuscript worse when over-applied, because it rewards the writer’s ingenuity, not the reader’s experience. Run tests 1 to 8 (Anchor, Dimension count, Launch, Technique fit, Momentum, Gap content, Redundancy and Simultaneity) on every flagged join, and test 9 (Echo) on the ten joins that matter most and on no others.
See also: the Reference protocol, with each test’s procedure and a table of combined failures; the Reference failure catalogue, where each pattern named in the table is set out with its exception.