Context-dependent memory is better recall when the external environment at retrieval matches the environment at learning. State-dependent memory is the same effect for internal states, such as mood or intoxication.
Both follow from Tulving’s (1974) principle that people encode the surrounding context and their internal state alongside new information, so reinstating either one can aid recall.
The Encoding-Specificity Principle
Both context- and state-dependent memory rest on one underlying idea: Tulving’s encoding-specificity principle. It holds that people store more than the target item when they learn something. They also store a sample of whatever else was present at the time. This includes the setting and their internal state.
A retrieval cue works only to the extent that it overlaps with the information encoded alongside the target. High overlap makes remembering easy. Low overlap can leave the same stored memory temporarily unreachable.
This is the radical part of the idea: a cue’s power comes from the encoding episode, not general association. Tulving and Thomson (1973) tested this directly.
Tulving and Thomson (1973): Recognition Failure of Recallable Words
- Aim: To test whether a retrieval cue’s effectiveness depends on being encoded alongside the target, rather than on its usual strength of association with that target.
- Method: Participants studied target words paired with weak cue words, such as the target BLACK with the weak cue train. At test, they saw strong, familiar associates (white for BLACK) and tried to recognise the targets, then received the original weak cues for cued recall.
- Results: Participants often failed to recognise target words when prompted with the strong, familiar associate, yet recalled the very same words moments later when given the original weak cue from encoding.
- Conclusion: A cue works because of what was encoded with the target, not its normal strength of association. Retrieval is specific to the conditions of encoding, the foundation for both context- and state-dependent memory.
This “recognition failure of recallable words” overturned the standard assumption that recognition always beats recall. It is the theoretical foundation beneath both effects below, and it reframes much everyday forgetting as a retrieval-cue problem: the information is still there, just temporarily out of reach.
Context-Dependent Memory
Context-dependent memory refers to improved recall of specific episodes or information when contextual cues relating to the environment are the same during encoding and retrieval.
In other words, recall is easier when you are in the same environment or situation where you first learned the information.
For example, studying for a test in the room where the exam will be held can boost recall on the day, compared with studying somewhere completely different.
This idea is well established in cognitive psychology and has been demonstrated across numerous experiments.
It’s one of the reasons why techniques like “place-based mnemonics” (also known as the “method of loci”) can be so effective.
Godden and Baddeley (1975)
An interesting experiment conducted by Godden and Baddeley (1975) indicates the importance of setting for retrieval. Baddeley asked 18 deep-sea divers to memorize a list of 36 unrelated words of two or three syllables.
One group did this on the beach and the other group underwater. When they were asked to remember the words half of the beach learners remained on the beach, the rest had to recall underwater.
Half of the underwater group remained there and the others had to recall on the beach.
The results showed that participants recalled far more words when the learning and recall environments matched. Recall was around 50% higher in the matching conditions than in the mismatched ones.
Critical Evaluation
This study has limited ecological validity. The environment was familiar to the divers. But the task was artificial: people rarely learn lists of meaningless words in everyday life.
A further weakness is disruption. Groups who changed environment between learning and recall were disrupted by that change. Groups who stayed in the same environment were not, and this could have affected their recall.
However, it was a controlled experiment so it can be replicated so reliability can be tested.
Godden and Baddeley (1980) then tested whether the same land/underwater manipulation would affect recognition memory rather than recall.
Using a similar diving procedure, they found no context effect at all. Matching or mismatching the environment made no difference to how many words participants correctly recognised.
This recognition-memory null is as important as the original result. It shows that context-dependent memory is chiefly a recall phenomenon, one that mostly disappears once the test itself supplies a strong retrieval cue.
There is further support for the influence of contextual cues. Abernethy (1940) found that students performed better on tests held in the same room where they had learned the material. Scores were also better when the same instructor gave both the teaching and the test.
These studies also do not account for the meaning of the material or the learner’s motivation.
This theory has a real-world application: the cognitive interview used in police interviews. Witnesses are guided to mentally recreate the physical scene, and their emotional and physical state, from the time of the incident.
Reinstating those cues can bring back extra accurate detail. This technique is also central to eyewitness testimony research.
This theory is also hard to disprove: if recall fails, is that because the information was never stored, or because the right cue was missing? (circular argument)
State-Dependent Memory
State-dependent memory refers to improved recall of specific episodes or information when cues relating to emotional and physical state are the same during encoding and retrieval.
State retrieval cues are based on the person’s physical or psychological state when information is encoded and retrieved: alert, tired, happy, sad, drunk, or sober, for example.
People are more likely to retrieve information when they are in the same state as when they learned it.
Goodwin et al. (1969) tested this with an alcohol experiment. Forty-eight male medical students learned a list of words while either drunk or sober.
Recall was tested 24 hours later. Some participants switched state for the retest, with sober learners getting drunk and drunk learners going sober, while others stayed in the same state throughout.
They were randomly assigned to four groups:
- Group 1: (SS) was sober on both days.
- Group 2: (AA) was intoxicated both days.
- Group 3: (AS) was intoxicated on day 1 and sober on day 2.
- Group 4: (SA) was sober on day 1 and intoxicated on day 2.
The intoxicated groups had 111 mg/100 ml alcohol in their blood, and they all showed signs of intoxication.
The participants completed four tests: an avoidance task, a verbal rote-learning task, a word-association test, and a picture-recognition task.
Information learnt while drunk was more available when tested in the same state later. Matching states won.
More errors occurred on day 2 in the AS and SA conditions than in the AA or SS conditions. This pattern did not hold for picture recognition, and the SS participants performed best overall on every task.
This supports the state-dependent memory theory as the performance was best in the participants who were sober or intoxicated on both days.
Critical Evaluation
This study has limited ecological validity. The tasks were artificial. Performance might not reflect how people act in everyday life.
Participants also knew they were taking part in a study, so they might have changed their behaviour to fit the study’s aims (demand characteristics).
As with the divers study, its controlled design means it can be replicated and its reliability checked.
There is further support for the influence of state-dependent cues. Overton (1964) tested two groups of rats: one received a mild barbiturate, the other did not. Both groups then learned to escape an electric shock in a maze.
Rats trained on the drug forgot how to escape when tested drug-free. Recall returned as soon as the drug was reinstated.
Humans differ from rats, so these findings cannot simply be extrapolated. Even so, a real strength of the study is that animals cannot show demand characteristics.
As with the divers study, these tasks do not account for the meaning of the material or the participant’s motivation. The theory also faces the same circularity problem: is poor recall really due to a missing cue, or was the information never stored?
Real-life applications: this idea is used to improve recall in eyewitness memory. Witnesses are asked to describe their mood and emotional state when the incident took place, as part of the cognitive interview technique described above.
Critical Evaluation
Context- and state-dependent memory are genuine effects. But both are modest and tightly bounded. Three questions matter most: how large the effect is, whether it replicates, and whether the underlying logic can be tested cleanly.
How Strong Is the Effect?
Smith and Vela (2001) reviewed the environmental context literature. Their meta-analysis put the average effect at only d ≈ 0.28, a small effect by convention.
Effects shrink further once a stronger cue appears.
This is why free recall shows the effect but recognition often does not. A recognition test hands participants the studied item itself as a cue. That strong cue outshines the weaker, diffuse context cue.
Godden and Baddeley found exactly this pattern. They retested their own divers procedure using recognition rather than recall, and the context effect disappeared.
Two further limits matter. Most demonstrations use artificial word lists, which says little about meaningful, real-world learning.
Tulving’s “informational overlap” is rarely measured independently of the memory performance it explains. This risks circular reasoning. Poor recall gets blamed on low overlap, but low overlap is inferred from the poor recall itself.
Contemporary Research
Recent work asks a sharper question. Not just whether the effect exists, but how large it is, when it appears, and whether it survives being tested again.
Does the Classic Divers Effect Replicate?
- Aim: Murre (2021) tested the robustness of Godden and Baddeley’s original land/underwater effect by replicating their 1975 divers experiment as closely as possible.
- Method: Sixteen divers studied word lists on land or underwater in a heated pool. Each diver later recalled the words on land or underwater, completing all four learning-recall combinations.
- Results: Recall in the same context where the words were learned was not reliably better than recall in the mismatched context. Where a difference appeared, it favoured recall on land after underwater learning.
- Conclusion: The classic divers effect did not replicate under these conditions. The effect looks less robust than its textbook status suggests, and sensitive to details such as pool setting and retention interval.
This null result carries real weight. It is a direct replication of the field’s most famous study.
It lines up with the small pooled effect size that Smith and Vela’s meta-analysis reported.
Virtual Reality Studies of Context
Immersive virtual reality lets researchers manipulate context far more powerfully than a simple room change. Two studies show why that power cuts both ways.
Shin, Masís-Obando, Keshavarzian, Dáve, and Norman (2020) had participants learn material in two distinct virtual worlds, underwater and Mars. Recall was better when the study and test worlds matched.
The benefit was largest for items woven into that world’s storyline.
A later study moved learners between a virtual indoor and outdoor scene using a head-mounted display (Chocholáčková et al., 2023). The scene change alone did nothing. Recall showed no significant effect of context.
So a simple change of scenery is not enough. The context has to be distinctive. It must also be meaningfully bound to the material being learned.
Context Reinstatement in Eyewitness Interviews
Faber, Nielsen, and Berntsen (2023) tested what mental context reinstatement does for eyewitnesses. It helped in one way but not another.
Participants who mentally recreated the original scene reported more details. They also performed better on cued recall.
Their free recall, however, was no more accurate, and the extra detail included information that could not be verified.
Reinstatement changes what people remember. It does not, on its own, make free recall more reliable.
Key Takeaways
- Context vs. State: Context-dependent memory involves the external environment; state-dependent memory involves the learner’s internal condition, such as mood or intoxication.
- Encoding Specificity: Both effects follow Tulving’s principle that a retrieval cue only works if it was encoded alongside the target information.
- Recall, Not Recognition: The effect is strong for free recall but disappears for recognition memory, because a strong, item-specific cue outshines the diffuse context cue.
- Modest Effect Size: A meta-analysis puts the average context effect at only d ≈ 0.28, far smaller than the dramatic swing in the classic divers study.
- Replication Concerns: A 2021 direct replication of the divers study failed to reproduce the original effect, showing it is more fragile than its textbook reputation suggests.
- Cognitive Interview: Police use mental reinstatement of context and state to help witnesses recall more detail.
References
Abernethy, E. M. (1940). The effect of changed environmental conditions upon the results of college examinations. The Journal of Psychology, 10(2), 293-301. https://doi.org/10.1080/00223980.1940.9917005
Godden, D. R., & Baddeley, A. D. (1975). Context-dependent memory in two natural environments: On land and underwater. British Journal of Psychology, 66(3), 325-331. https://doi.org/10.1111/j.2044-8295.1975.tb01468.x
Godden, D. R., & Baddeley, A. D. (1980). When does context influence recognition memory? British Journal of Psychology, 71(1), 99-104. https://doi.org/10.1111/j.2044-8295.1980.tb02735.x
Goodwin, D. W., Powell, B., Bremer, D., Hoine, H., & Stern, J. (1969). Alcohol and recall: State-dependent effects in man. Science, 163(3873), 1358-1360. https://doi.org/10.1126/science.163.3873.1358
Murre, J. M. J. (2021). The Godden and Baddeley (1975) experiment on context-dependent memory on land and underwater: A replication. Royal Society Open Science, 8(11), 200724. https://doi.org/10.1098/rsos.200724
Overton, D. A. (1964). State-dependent or “dissociated” learning produced with pentobarbital. Journal of Comparative and Physiological Psychology, 57(1), 3-12. https://doi.org/10.1037/h0048023
Smith, S. M., & Vela, E. (2001). Environmental context-dependent memory: A review and meta-analysis. Psychonomic Bulletin & Review, 8(2), 203-220. https://doi.org/10.3758/BF03196157
Tulving, E. (1974). Cue-dependent forgetting. American Scientist, 62(1), 74-82.
Tulving, E., & Thomson, D. M. (1973). Encoding specificity and retrieval processes in episodic memory. Psychological Review, 80(5), 352-373. https://doi.org/10.1037/h0020071
