Discovery Lab Answers

The decline effect: what the term means and where it does — and does not — appear

The “decline effect” in parapsychology refers to a pattern in which a strong early result weakens as more trials are run, more studies accumulate, or replication is attempted. It is one of the oldest observations in the field, and it operates at two scales that are worth keeping distinct:

• Within-study (position) declines — scoring falls across the runs, columns, or turns of a single experiment.
• Chronological (long-term) declines — the effect size of a whole research area appears to shrink across decades.

A note on scope before the substance: the sources retrieved for this question are a small, uneven slice — two mid-1940s Journal of Parapsychology issues, one 2023 critical essay, and one 2025 micro-PK pilot. The share of the published decline-effect literature that ESP-Nexus holds on this question has not been measured, so read what follows as what these particular sources report, not as a settled reading of the field.

Within-study declines are documented, and old

The 1940s Journal of Parapsychology material shows within-study position effects were treated as an established, expected feature of the data. The 1943 issue reports significant “diagonal” and “horizontal” declines in pooled PK dice work, and even notes a “double decline” pattern characteristic of certain conditions ^[2]. The 1946 issue similarly reports declines across turns and top-bottom declines within record columns, some significant, some not ^[4]. These are internal-to-the-experiment patterns — declines within the structure of a single dataset, not across a literature over time.

Long-term chronological declines are more contested in these sources

Here the picture is genuinely divided across the retrieved sources.

Storm (2023) engages the long-term/chronological decline critically, treating it — alongside long-term experimenter psi — as a problem that can be “reified,” i.e. talked into existence ^[1]. Two specific points he raises:

• He notes that a foundational ganzfeld database, when checked, shows only a weak, non-significant negative relationship between effect size and study year (he reports r(26) = -0.20, p = .155, one-tailed, for a 28-study set), and reports similarly weak, non-significant correlations for the MDL study sets ^[1].
• He states that declines over time have not been reported in more recent RNG meta-analyses ^[1].

His argument is essentially that parapsychologists have “made too much of these long-term effects,” appealing to them too often — a caution against treating a chronological decline as an established phenomenon ^[1].

Dechamps, Iovine and Maier (2025) approach the decline from the opposite framing — as something real enough to be worth engineering against. Their micro-PK pilot tests whether aligning outcomes with participants’ implicit 50% expectations can prevent a decline; they describe systematic decline as “commonly observed in parapsychology,” while flagging that their own results are preliminary and require replication ^[3].

So within the retrieved sources you have one paper treating chronological decline as possibly a reification artifact ^[1] and another treating it as a real effect to be counteracted ^[3] — a direct difference in how two contemporary research groups frame the same phenomenon.

Proposed causes

Across the sources and the site’s synthesis, the candidate explanations for declines include fatigue, loss of motivation, ordinary statistical regression to the mean, flaws in study conduct, participants’ implicit probability expectations ^[3], experimenter-psi accounts, and — as a live possibility raised rather than settled — something about psi itself. ESP-Nexus’s Decline Effect page lays out this multi-factorial menu.

A relevant counter-case

Not every psi finding shows a decline. ESP-Nexus’s Sheep–Goat Effect page notes that a long-span assessment found that effect statistically robust and stable, with neither study quality nor effect size declining over 46 years — explicitly running counter to decline claims. And its Meta-Analysis in Parapsychology synthesis reports that pooled ganzfeld summaries continued to report effects modestly above chance with no sign of decline over roughly four decades. Whether a decline appears therefore depends heavily on which effect and which database you examine.

Bottom line

In the retrieved sources, within-study position declines are old and documented ^[2][4]; chronological across-the-literature declines are treated inconsistently — as a possibly over-claimed or reified pattern by Storm (2023) ^[1], as a real effect worth counteracting by Dechamps et al. (2025) ^[3], and as absent in specific ganzfeld and RNG meta-analytic databases ^[1]. There is no single settled answer here, and the sources themselves say so.

For the fuller treatment, see the Decline Effect methodology page.