After the Doorway: Distinguishing Forgotten Objects from False Recognition

Imagine seeing a yellow sphere and a purple prism. Later, someone asks whether you saw a yellow prism. Both the color and the shape are familiar, but their combination is not. In this illustrative example, answering yes would not mean that nothing remained in memory. It could reflect difficulty keeping familiar features attached to the correct objects. A memory experiment therefore needs to ask not only which genuine matches stop being recognized, but also which nonexistent combinations begin to seem right.

Jessica McFadyen and colleagues brought that distinction to an investigation of the doorway effect in 2021. Across four experiments, they used immersive virtual reality, corridor videos and actual walking. They did not find a uniform pattern in which crossing a boundary removed recently encountered information. In the virtual experiment with an additional working-memory task, the important change was an increase in accepting mismatched color-shape combinations. The study makes the meaning of a memory error central to interpreting what a doorway has changed. [1]

What the virtual rooms were designed to control

The first experiment conceptually replicated earlier doorway research under modified conditions. Participants viewed a building through a virtual-reality headset and used controllers to handle objects. They remained seated while the system moved them through the virtual space. At a table, an object collected earlier was put down and a new object was picked up. Participants needed to remember both the object just deposited and the one they were carrying toward the next table.

Rooms were visually similar. That reduced the possibility that a difference attributed to a doorway actually reflected encountering a substantially different visual environment, such as walls changing color. The route was also arranged so that participants could not reliably anticipate the doorway condition while initially encoding the object. Their knowledge of the coming transition was therefore less able to change how they studied it.

These controls sharpened a particular comparison, but did not make the experiment identical to every everyday room transition. Increasing experimental control can also remove features that contribute to an effect outside that arrangement. The question was whether crossing a boundary would impair performance under these specified conditions, not whether all possible changes in location have the same consequence. [1]

Why a false alarm is not just a missing hit

Partway through the journey, a text probe named a color and a shape. Participants answered yes when the description matched either the deposited object or the carried object, and no otherwise. Crucially, negative probes recombined features from those two objects. The individual features were not completely novel. Rejecting the probe required recognizing that the particular color and shape had not belonged together.

A yes response to a genuine match is a hit. A yes response to a mismatched combination is a false alarm. Both matter for evaluating recognition, but they describe different behaviors. If only an overall accuracy score is considered, their contributions can become difficult to distinguish. The researchers used signal-detection measures to examine sensitivity and response criterion, the tendency to favor yes or no answers.

Sensitivity can decline because genuine objects receive fewer hits, because misleading probes receive more false alarms, or through a combination of changes. Describing every decline as the same kind of forgetting would lose that distinction. The recombined probes were useful precisely because they tested whether familiar information could be accepted in an incorrect arrangement. [1]

Adding a second task changed the error pattern

Forty people began the first experiment. After participants who stopped because of motion sickness and those excluded for poor performance were removed, twenty-nine remained. Recognition performance was high. The analysis did not demonstrate a significant effect of crossing a doorway on sensitivity or response criterion. Bayesian comparisons also supported an absence of an effect under the particular statistical assumptions used, although very high accuracy raised a possible ceiling concern.

The second experiment added spoken backward counting between handling the object and encountering the memory probe. Participants usually subtracted six or seven, with adjustments for some individuals' arithmetic performance. They were consequently maintaining the objects while performing another demanding activity. The methods report a final sample of forty-five. Because different people participated in the two experiments, the addition of load should not be described as a within-person manipulation across the complete series. [1]

With this additional task, crossing a doorway was associated with reduced sensitivity and a greater tendency to say yes. Examining the responses separately showed that the change was driven particularly by false alarms to mismatched probes. Hits to the genuine deposited and carried objects did not show the same significant doorway pattern. That finding fits an interference account better than a simple story in which earlier contents have been wiped away.

It nevertheless remains a behavioral interpretation. Preserved hit rates do not directly demonstrate that every memory representation remained unchanged. The useful conclusion concerns the observed error: participants more often accepted an incorrect conjunction of familiar features. The experiment did not measure a literal clearing operation in the brain. [1]

Returning through a corridor to rebuild a layout

The third and fourth experiments changed the memory task. Separate samples of twenty-six participants studied sixteen butterfly photographs in a four-by-four grid for thirty seconds. In one experiment, they then watched a first-person video of a journey down a corridor and back. In the other, participants made the walk themselves. Curtains divided the corridor on boundary trials; comparison trials lacked those dividing curtains. Backward counting accompanied both versions.

Afterward, participants received the shuffled photographs and had forty-five seconds to reconstruct their original positions. This was no longer a yes-or-no judgment about a color-shape description. The test concerned spatial relationships between the items. It would therefore be misleading to call these four experiments identical repetitions that differed only in the medium used to display a doorway.

Neither corridor experiment demonstrated a significant overall difference between the boundary and no-boundary conditions. In the actual-walking experiment, participants tended to perform better in whichever condition came first. Fatigue and interference from earlier layouts were possible explanations, but were not separately established as causes. Counterbalancing which condition began the task helped prevent that order pattern from being mistaken for a general curtain effect. [1]

What a conceptual replication can establish

The findings do not erase every earlier observation linking spatial boundaries with memory. A conceptual replication asks a related question with altered procedures. Those alterations can expose limits to generalization as well as remove potential confounds. Here, passive virtual movement, visually similar rooms, an additional counting task and the distinction between recognition and reconstruction all matter for interpretation.

The experiments did not form a complete design in which every possible moderator was independently varied and isolated. For example, the contrast between the virtual and corridor tasks involved more than simply replacing a simulated doorway with a physical curtain. The memory test changed too. Suggestions about the importance of active navigation or a boundary's visual salience are therefore research possibilities, not a license to assign each outcome to one proven cause.

The evidence supports a more specific lesson than either doors erase memory or doors never matter. Under one loaded virtual condition, incorrect combinations became harder to reject without a corresponding significant reduction in hits to the real objects. Under other tested conditions, an overall boundary effect was not demonstrated. Memory performance must be interpreted through the answers the task requires and the errors it permits.

Returning to the opening example, a yellow prism can seem familiar for a reason even when that exact object was never seen. Failing to preserve the conjunction is not identical to failing to recognize every component. Understanding memory around boundaries therefore requires more than counting doorways. It requires explaining what was learned, what the test asked, and which kind of answer changed. [1]

Sources

[1] McFadyen, J., Nolan, C., Pinocy, E., Buteri, D., and Baumann, O. (2021). Doorways do not always cause forgetting: a multimodal investigation. BMC Psychology, 9, 41. https://link.springer.com/article/10.1186/s40359-021-00536-3

The source study is licensed under Creative Commons Attribution 4.0: https://creativecommons.org/licenses/by/4.0/ This is an original explanatory account and assessment, not an official translation. The opening objects are a teaching example rather than participant data.