Contribution
Rules out two simple explanations of the location updating effect within the original paradigm. Doorway trials did not take longer, so retention interval is not the cause. A 3-second delay before the probe did not shrink the effect, so it is not a transient cost of updating, the explanation Radvansky & Copeland (LIT-1392) had given. The remaining account is the event horizon model’s: retrieval interference.
Key insight
The effect sits on the object being carried, and that is what makes it an interference effect rather than a decay or deactivation effect. “The idea that information from a prior event model is less available is consistent with the decline in availability of information in the dissociated condition as that object is not carried to the new event model. However, the result of central interest here is the effect of a shift in the associated condition, which lends itself to a competition-based account” (p. 2126). “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” (p. 2125).
Assumptions
- A 3-second pause is enough for updating to finish. Justified by reading-time effects at narrative boundaries being under a second (p. 2123). If updating took longer, the delay would not test the disruption account.
- Interference and deactivation are the two live accounts. Other explanations (context change, task demands of navigating doorways) are not tested here.
- The event horizon model as stated in LIT-1384.
Key results
- Experiment 1 (N = 48; 46-inch touchscreen, joystick, headphones with footsteps, lights off; 48 probes, with 12 “yes” probes after a shift and 24 without, plus fillers). Errors on associated probes .113 after a shift against .046 without, F(1,47) = 12.81, p = .001; dissociated .149 against .188, p = .20. Shift × association interaction on errors p = .02 (the paper reports F = 9.35 with p = .02); on response time F(1,47) = 14.34, p = .001, with associated 1,543 against 1,275 ms. Travel time: no-shift 15.6 s, shift 14.4 s, p = .088, so “If anything, in this study, the retention interval was longer in the no shift condition” (p. 2122).
- Experiment 2 (N = 57; the same with a 3-second dimmed pause before each probe). Associated errors .114 against .072, F(1,56) = 7.42, p = .009; dissociated .197 against .217, F < 1. Interaction on errors p = .04; response times associated 1,649 against 1,354 ms, p < .001. “Again, this shows that walking through doorways causes forgetting, and that this is largely confined to the associated probes” (p. 2123).
- Between experiments (N = 105). No interaction of experiment with shift or association on errors or response times (Fs < 1). The delay changed nothing.
- Working memory. The load is two objects, yet “it is typically the case that memory error rates double or triple in this research paradigm when a person has moved from one event to another” (p. 2127). What that says about working memory is left open: “the nature of such processes is unclear at this time” (p. 2127).
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | The doorway effect on the carried object replicates with joystick, touchscreen and headphones | moderate (same lab, same paradigm) | Experiments 1 and 2 |
| C2 | It is not a retention-interval effect | moderate | travel times recorded, if anything longer without a shift; a room-size manipulation cited from the authors’ 2015 work |
| C3 | It is not a transient disruption from updating | moderate, given the 3-s assumption | Experiment 2 and the between-experiment analysis |
| C4 | The effect on the carried object is retrieval interference between two event models | interpretive | by elimination of C2 and C3, and from the event horizon model |
| C5 | The effect does not reach the object just set down | weak (nulls on few trials) | both experiments, p ≥ .20 |
Concepts
- Location updating effect, associated / dissociated, event horizon model: as in LIT-1392 and LIT-1384.
- Retention-interval account: more forgetting because more time passes on doorway trials.
- Updating-disruption account: a temporary drain on resources while the event model is rebuilt at the boundary, after reading-time costs at narrative shifts (Zwaan et al. 1995).
Connections
It follows LIT-1392 and LIT-1384, and cites companion papers from the same lab not held here (Radvansky, Tamplin & Krawietz 2010; Radvansky, Pettijohn & Kim 2015; Pettijohn & Radvansky 2015, on room size and single objects; Lawrence & Peterson, imagined travel). It restates LIT-1384’s return result accurately: “there was no difference in the pattern of data compared to the shift condition” (p. 2126). It does not cite event segmentation theory.
Bearing on the record
- CASE-136. The case’s bullet is accurate: similar travel times, a 3-second delay with no change, and interference among event models that share elements. It should add that the effect was on the carried object only, in both experiments.
- CLAIM-421 (activation interruption). This is the paper that sets the two mechanisms against each other, and it argues against the one CLAIM-421 uses for the doorway. Deactivation of the prior event model is granted for the object set down, which showed no reliable doorway effect here. The doorway effect proper, on the carried object, is assigned to “a competition-based account”. If the record wants the doorway effect as an activation interruption, it is reading against the authors. The case’s own quotation from A37 (“lost accessibility because the current situation is being organized through a different event model, one in which the intention’s earlier contextual supports are no longer active”) is closer to the deactivation account the paper calls secondary.
- Proposed THEORY-888 takes this paper as a primary source.
- No instruction for machine-learning practice.
Limitations
- Same lab, same paradigm, undergraduates; the shift condition has 12 “yes” probes per person, so about 6 per cell for associated probes.
- The 3-second delay tests only a disruption that ends within 3 seconds.
- Interference is reached by elimination of two alternatives; it is not tested against deactivation by a condition where they predict opposite results.
- Travel-time figures differ between the text (15.6 and 14.4 s) and Table 1 (cell means of 13.47 to 17.18 s, one printed as “17,178”), and the Experiment 1 interaction is printed with a p-value that does not match its F. Small, but the reporting is loose.
Open questions
- What would interference predict that deactivation does not? The paper names none; a condition where the carried object is put down and picked up again in the new room, or a probe for an object that was in both rooms without being carried, might separate them.
- Do the effects survive balanced trial counts and an independent lab? (LIT-1387 is the record’s one attempt, with different conditions.)