What does Sound, Scent and Proustian Memory conclude, in numbers, about “From "Instant Anchors" to Learning Curves: the Ledger of Evaluative Conditioning”?

Answer: NLP "anchoring" lacks empirical support: of the 33 relevant studies counted by Witkowski, the supportive share is only 0.273. The testable alternative is evaluative conditioning, with a meta-analytic mean effect of d = 0.52 that is larger when participants are aware of the pairing. The Rescorla-Wagner rule turns "a few pairings" into a learning curve: at a learning rate of 0.10 it takes 7 pairings to pass one half, and unpleasant pairings pull the strength back by the same rule. Spacing has an effective region between too dense and too sparse, whose location must be filled in from data. Three red lines hold throughout: nothing covert, nothing aimed at minors, and no describing liking as a state.

Scope of the numbers: Parameters are illustrative and all calculations come from the book's programs. This is not medical, psychological or regulatory advice and claims no effect on body or mind.

中文版

Data table

The numbers below are printed verbatim by the chapter's Python program and can be reproduced by running it. standard-library Python onlyprinted values that match the answer: 0.273, 0.52, 33

A. NLP studies (n33)
supportive9 0.273 [0.151, 0.442]
non-supportive18 0.545 [0.380, 0.702]
uncertain6 0.182 [0.086, 0.344]
B. evaluative conditioning, people per group
meta mean d0.52 -> 60 per group, 120 in all
CI lower d0.466 -> 74 per group, 148 in all
field guess d0.30 -> 176 per group, 352 in all
weak field d0.20 -> 394 per group, 788 in all
Show the full program
# Chapter 6. p63: what the evidence counts say, with intervals.
# A. Witkowski (2010) classified 33 relevant empirical NLP studies:
#    9 supportive, 18 non-supportive, 6 uncertain (body counts, p. 60).
#    Wilson 95 % intervals for each share.
# B. Hofmann et al. (2010): evaluative conditioning, mean d = 0.52,
#    95 % CI 0.466 to 0.582 over 214 studies. People per group for a
#    two-group test (alpha 0.05 two-sided, power 0.80, normal approx.)
#    at d = 0.52 and at smaller field-like values.
import math
from statistics import NormalDist

Z = NormalDist().inv_cdf(0.975)
ZB = NormalDist().inv_cdf(0.80)


def wilson(k, n):
    p = k / n
    den = 1.0 + Z * Z / n
    mid = (p + Z * Z / (2 * n)) / den
    half = Z * math.sqrt(p * (1 - p) / n + Z * Z / (4 * n * n)) / den
    return p, mid - half, mid + half


def per_group(d):
    return math.ceil(2.0 * ((Z + ZB) / d) ** 2 + 1)


if __name__ == "__main__":
    print("A. NLP studies (n = 33)")
    for lab, k in (("supportive", 9), ("non-supportive", 18),
                   ("uncertain", 6)):
        p, lo, hi = wilson(k, 33)
        print("   {:<15} {:>2}  {:.3f}  [{:.3f}, {:.3f}]".format(
            lab, k, p, lo, hi))
    print("B. evaluative conditioning, people per group")
    for lab, ds in (("meta mean", "0.52"), ("CI lower", "0.466"),
                    ("field guess", "0.30"), ("weak field", "0.20")):
        n = per_group(float(ds))
        print("   {:<12} d = {:<5} -> {:>4} per group, {:>4} in all".format(
            lab, ds, n, 2 * n))

Single source and citation

Fan, Y. (2026). Cross-Modal Coupling of Acoustic Rhythms and Olfactory Cues, Proustian Memory Retrieval and Rapid State-Conditioning Anchors (Version 1.0.0), Chapter 6: From "Instant Anchors" to Learning Curves: the Ledger of Evaluative Conditioning. 5D Library. DOI pending

DOI pending · monograph page goes live after publication · 5DL-B-392 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026392,
  author    = {Fan, Yafu},
  title     = {Cross-Modal Coupling of Acoustic Rhythms and Olfactory Cues, Proustian Memory Retrieval and Rapid State-Conditioning Anchors},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

Other chapter answers in this book