Source
Radvansky & Copeland (2006), LIT-1392; Radvansky, Krawietz & Tamplin (2011), LIT-1384; Pettijohn & Radvansky (2016), LIT-1385; each read in full (NOTE-1083, NOTE-1082, NOTE-1081). McFadyen et al. (2021), LIT-1387, read in full in NOTE-1079, for the limits. Page numbers are the journals’.
What was actually shown
- The pattern. In desktop virtual rooms, a probe for the carried object drew more errors and slower answers after a doorway than after crossing a large room (2006 Experiment 2: errors .05 to .14); the object just set down was not affected. “Thus, making a spatial shift disrupted processing of currently relevant information” (LIT-1392, p. 1154). The same pattern held in both 2016 experiments, “largely confined to the associated probes” (LIT-1385, p. 2123).
- Alternatives that could have won and did not.
- Retention interval: doorway trials were not slower to travel; if anything the no-shift trials took longer (LIT-1385, Experiment 1).
- A passing cost of updating: a 3-second pause before the probe left the effect unchanged (LIT-1385, Experiment 2), against the 2006 paper’s own effort-of-updating explanation.
- Loss of the encoding context: returning to the room where the object was picked up did not help (A′ .90 against .92 for a plain shift, p = .20), and two new rooms were worse than going and returning (LIT-1384, Experiment 3). Memory tracked “the number of new areas entered” (LIT-1384, p. 1640).
- The account. On the event horizon model the carried object is in two event models, the room where it was picked up and the room it is carried into, and a recognition probe must choose between them: “The effect is a result of the competition between these two event models during retrieval, even though they are both pointing to the same response” (LIT-1385, p. 2125).
- The limits. In head-mounted virtual reality with passive movement and identical rooms, doorways did not lower hits (Bayes factors 3–6 for the null); under a counting load they raised false alarms to recombined probes, which the authors read as interference (LIT-1387, Experiments 1 and 2). Watching or walking a curtained hallway did not affect memory for a 16-item spatial arrangement (Experiments 3 and 4).
What this does not say
- That people forget their intentions when they enter a room. No study here tests a goal or intention. The task is recognizing colour–shape labels for two objects, and the effect is a few percentage points of errors.
- That the effect is robust. It replicates within one lab and its paradigm, with few trials per cell (as few as 5 or 6), and with room changes that also changed wall pattern and room size. The one independent multi-experiment test in the record changed several things at once and found no loss of hits. Independent positive studies exist (Lawrence & Peterson 2016; Seel et al. 2019, cited in LIT-1387) and are not held here.
- That interference has been shown. It is reached by eliminating two alternatives. No condition yet separates it from deactivation of the prior event model by an opposite prediction.
- That it is event segmentation theory’s mechanism. EST (THEORY-886) predicts that what precedes a boundary becomes harder to retrieve because the model is reset. The cost here falls on what is carried across the boundary.
- That location is irrelevant. The return result says the encoding room does not reinstate access. On the authors’ account each room entered is still a new event model, so the unit is room-bound. Whether a change of task without a doorway does the same (A37’s proposal at CASE-136) has not been tested in these papers.
Connections
- CASE-136 describes these studies; this states the account they support and its limits.
- CLAIM-421 uses the doorway effect to illustrate activation interruption. The authors’ mechanism is interference, not deactivation; the claim’s defeater already lets the illustration fall without harm.
- TERM-055: the effect is a loss of availability with the item still held, which is that term’s distinction, at a small scale.
- THEORY-886: the segmentation account it is often read through, with a different memory mechanism.