Contribution
The first demonstration of what later papers call the location updating effect. It carries the associated/dissociated effect of narrative comprehension (Glenberg et al. 1987: an object a protagonist carries stays more available than one put down) into a task people perform themselves, in a desktop virtual environment, and adds a spatial-shift manipulation: moving through a doorway into a new room, against walking the same distance across one large room. The shift harmed memory for the carried object, not for the one set down.
Key insight
A room change is treated as a change of situation. “Entering a new room is a major change in the current situation, and this requires a person to update their situation model of the environment” (p. 1155). What surprised the authors is where the cost fell: on the information that should have been most available. “Thus, making a spatial shift disrupted processing of currently relevant information” (p. 1154).
Assumptions
- Virtual space stands in for real space. Cited work is taken to show that virtual and real spaces are represented “essentially identical” or with “only small deviations” (p. 1151). Tested only later (LIT-1384, Experiment 2).
- The probe task measures availability in a situation model. The task is a two-item short-term memory task with verbal probes (colour plus shape names). The authors grant that people “need only rehearse the names of the two objects” (p. 1155) and argue the effect matters all the more for that.
- The shift and no-shift conditions differ only in the doorway. In Experiment 2 the no-shift trials used large rooms made by “tearing down the walls” (p. 1153), with half the room darkened and an invisible barrier until the object was exchanged. Each room had “a different pattern on the walls, to emphasize changes in location” (p. 1152), so a shift also changed the visual context.
Key results
- Experiment 1 (N = 41 after 10 dropped; 51 probes, all after a room change; no no-shift condition). Associated probes were answered faster and more accurately than dissociated ones: 1,440 against 1,814 ms, error .20 against .33 (Table 1). The size of the effect correlated with dissociated response times (r = .74), not associated ones (r = .09): “a difficulty in accessing information about a dissociated object rather than in maintaining information about the associated object” (p. 1152).
- Experiment 2 (N = 54 after 11 dropped; 15 no-shift trials, 5 per
probe type, and 36 shift trials, 12 per type; Table 2).
- Associated probes: error .05 without a shift, .14 with one, F(1,53) = 20.99, p < .001; response time 1,433 against 1,588 ms, F(1,53) = 6.01, p = .02.
- Dissociated probes: error .19 against .22, F = 1.07; response time 1,736 against 1,682 ms, F < 1. No effect.
- Shift × probe interaction on response time, F(1,53) = 4.33, p = .04; on errors, marginal, p = .09. No main effect of shift on response time.
- The shift effect and the associated/dissociated effect did not correlate across people (r = .01), read as “qualitatively different processes” (p. 1154).
- The authors’ summary. “There was fairly clear evidence that moving through a doorway from one region to the next made information that would otherwise be highly available less available. Information at a relatively low level of availability was less affected” (p. 1154).
- Alternatives argued against (p. 1155): encoding specificity (rarely found with recognition; the carried object was in the new context too; it should have hit both objects); verbal rehearsal alone; a mismatch between visual experience and verbal probes; recency (the dissociated object was the last one seen, yet it was the associated one that suffered).
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | In a desktop virtual task, a carried object is remembered better than one just set down (the associated/dissociated effect) | moderate | Experiments 1 and 2, both significant |
| C2 | Walking through a doorway, against walking as far within a room, lowers accuracy and slows recognition for the carried object | weak to moderate | Experiment 2 only: 5 no-shift trials per condition; interaction p = .04 on RT, p = .09 on errors |
| C3 | The doorway does not affect memory for the object just set down | weak (a null on few trials) | Experiment 2, F ≈ 1 |
| C4 | The effect reflects situation-model updating at a spatial shift | interpretive | General Discussion; no test separates updating from other accounts |
| C5 | Encoding specificity and recency cannot explain the pattern | moderate as argument | the pattern falls on the associated object, which was in the new context and was not the last seen |
Concepts
- Associated / dissociated object: the object a person is carrying, against the one just set down; after Glenberg et al. (1987).
- Spatial shift: moving through a doorway into a new room; the no-shift control is crossing an equal distance within a large room.
- Situation model: “mental representations that act as mental simulations that capture the functional relations among the entities in an event” (p. 1150).
- Foregrounding: elements of a situation model more available than others, here by “functional interaction between the person and the objects” (p. 1154).
Connections
It comes out of narrative-comprehension work on spatial shifts and situation models (Glenberg, Meyer & Lindem 1987; Zwaan & Radvansky 1998; Rinck & Bower 1995) and the authors’ earlier video-game studies. The interference account the same lab later gives (LIT-1384, LIT-1385) is not in this paper; here the explanation is the effort of updating: “Situation model updating requires a relatively large amount of cognitive effort and coordination, which can slow processing and allow for more errors to occur” (p. 1155). That transient-cost account is the one LIT-1385 later tests and rejects. It does not cite event segmentation theory (LIT-1382).
Bearing on the record
- CASE-136. The case, and the summary of LIT-1392, say memory “for objects just picked up or set down” was worse after a doorway. The paper says otherwise: the shift harmed the carried object and left the one set down unchanged (C2, C3). The finding is narrower and stranger than the case states. It is the most available item that suffers.
- CLAIM-421 (activation interruption). The case is offered as an intention that is held but not retrievable. Nothing here concerns intentions or goals. The probe is a colour–shape name, the memory load is two objects, and errors on the carried object rose from 5% to 14%. The popular gloss of forgetting why one entered a room is not tested. The one point that fits CLAIM-421’s distinction is that the object is held, since most probes are still answered correctly, but slower and less reliably after a shift. That is a loss of availability without loss of the item, which is TERM-055’s distinction in miniature. It is read off the data, not argued by the paper.
- Proposed THEORY-888 (doorway effect as retrieval interference) takes this paper as its first source, for the pattern, not for the explanation.
- THEORY-688 (doorway options). No relation beyond the word. Here the doorway is an event boundary in a person’s memory task, not a subgoal in learning.
- No instruction for machine-learning practice.
Limitations
- The decisive experiment has 5 no-shift trials per probe type per person. A 9-point change in error rate is less than one trial per cell. McFadyen et al. (LIT-1387) make this objection to the paradigm.
- Shift and no-shift trials differed in room size, in the darkened half and invisible barrier of the no-shift rooms, and in wall pattern, which changed at every doorway. The doorway is not isolated from a change of visual context.
- The test is short-term recognition of verbal labels, in a desktop environment, by undergraduates. 21 of 116 participants recruited across the two experiments were dropped or lost, 13 for not following instructions.
- The situation-model interpretation is not tested against an alternative that predicts the same pattern; the alternatives argued against are memory accounts that ignore context.
Open questions
- Does the effect hold when the rooms on either side of the doorway look the same, and when shift and no-shift trials are equally many and unpredictable? (LIT-1387 did this, passively, and found no effect on hits.)
- Why should the carried object, part of the new situation too, be the one that suffers? The later interference account (LIT-1384) is one answer.
- Does anything like it happen to an intention or goal, rather than to an object label?