What you'll learn

• What Steyvers and Hemmer (2012) investigated about reconstructive memory in everyday settings.
• How schemas, noisy memory traces, and Bayesian reasoning help explain memory errors.
• The study’s key procedures, findings, and conclusions for AO1.
• How to evaluate the study for AO3, including methodology, ethics, and real-world application.

The big idea: memory is reconstructed, not replayed

When you remember something, your mind does not simply “play back” a perfect recording. Instead, it often reconstructs the memory using two sources of information:

• details you actually encoded at the time
• your existing knowledge about what usually happens in that kind of situation

This links strongly to Bartlett’s classic study of reconstructive memory, War of the Ghosts (1932). Bartlett showed that people changed unfamiliar story details so they fitted their cultural expectations. Steyvers and Hemmer (2012) developed this idea using more naturalistic, everyday visual materials.

Why “naturalistic environments” matter

Many memory studies use artificial tasks, such as recalling lists of unrelated words. These are useful because they allow tight control, but they may not reflect everyday memory very well.

Steyvers and Hemmer were interested in memory in naturalistic environments.

This matters because everyday memory usually involves meaningful surroundings. If you try to remember where a mug was in a room, you may use your general knowledge that mugs are often on tables, desks, or kitchen counters.

The study: Steyvers and Hemmer (2012)

Aim

The aim was to investigate how people reconstruct memories from naturalistic scenes, especially how their recall is influenced by:

• the original information they saw
• their prior knowledge about what is typical in that kind of environment
• the shared patterns that appear when many people’s memories are combined

General procedure

The paper reports research using computer-based memory tasks rather than one simple single experiment. The key approach was:

1. Participants viewed meaningful visual material, such as objects in realistic scenes.
2. After a delay or test phase, they reconstructed what they had seen.
3. Their responses were compared with the original scene.
4. The researchers analysed the pattern of errors, including whether errors shifted towards what would usually be expected in that environment.

For example, if an object was shown in an unusual place, participants might remember it as being closer to a more typical position.

What was measured?

The researchers could measure recall accuracy in a more objective way than in Bartlett’s story-recall study. For visual reconstruction tasks, errors can be measured by comparing the remembered response with the original stimulus.

For a location task, this could mean measuring how far the recalled object position was from its true position. For a feature task, it could mean how much the recalled feature differed from the original.

The Bayesian explanation

Steyvers and Hemmer used a Bayesian account of memory.

You do not need to calculate Bayesian models for the Edexcel exam, but you should understand the logic.

Imagine trying to remember where a plant was in a room. Your exact memory trace may be fuzzy. Your schema tells you that plants are often near windows. So your final memory may be pulled slightly towards the window, even if the plant was not originally there.

A simplified version is:

$$estimate=wx+(1-w)\mu$$

where $x$ is the remembered detail from the event, $\mu$ is the typical value suggested by your schema, and $w$ is the weight given to the memory trace.

Key findings

The main findings were that people’s memories were not random. Their errors often showed systematic patterns.

1. Recall was influenced by prior knowledge

Participants tended to reconstruct scenes in ways that fitted their expectations. This supports the idea that schemas guide memory.

For example, if an object belongs in a certain kind of place, people may remember it as being nearer to that typical place than it really was.

2. Memory traces were noisy

A noisy memory trace is an imperfect or uncertain memory of the original event. It does not mean the participant remembers nothing; it means the memory contains some useful information but also error.

When the memory trace is weak or uncertain, people rely more heavily on prior knowledge.

3. Group averages could improve accuracy

Steyvers and Hemmer also explored the idea that combining responses from many people can reduce random error. This is related to the wisdom of crowds effect.

However, this only reduces random individual error. If many people share the same schema-based bias, the group average can still be biased.

Conclusions

Steyvers and Hemmer concluded that memory reconstruction is adaptive. Human memory is not designed to store every detail perfectly. Instead, it uses meaningful knowledge to make likely reconstructions of the past.

This supports reconstructive memory theory and gives it a more modern, quantitative basis than Bartlett’s original work.

AO2: applying the study

In an exam scenario, you might be asked to apply Steyvers and Hemmer to everyday memory.

Suppose someone briefly sees an office and later reports that a stapler was on the desk. If the stapler was actually on a shelf, this could be explained by schema-based reconstruction. Their “office schema” makes a stapler-on-desk memory seem likely.

AO3: evaluating Steyvers and Hemmer

Strength: higher ecological validity than artificial memory tasks

A major strength is that the study used meaningful, naturalistic material. This is more realistic than memorising random syllables or unrelated word lists.

This increases ecological validity, which means the findings may be more relevant to everyday memory.

Strength: objective measurement

The study used measurable reconstruction errors, such as differences between original and recalled visual details. This is more objective than relying only on qualitative interpretation.

This improves reliability because researchers can compare responses in a consistent way.

Strength: real-world applications

The findings are useful for understanding eyewitness memory. Witnesses may confidently reconstruct details based on schemas, especially when the original event was brief, stressful, or unclear.

This does not mean eyewitnesses are “lying”. It means their memory may be a sincere reconstruction.

Weakness: still not fully real life

Although the materials were more naturalistic than word lists, the task was still controlled and computer-based. Real memories often involve emotion, movement, conversation, stress, and personal consequences.

So, the study improves ecological validity, but it does not completely recreate real-world remembering.

Weakness: schemas vary between people and cultures

People do not all have identical schemas. Prior knowledge may depend on culture, age, occupation, and personal experience.

For example, a chef, a child, and an office worker may have different expectations about where objects are usually found. This can limit generalisability.

Weakness: alternative explanations

Memory errors might not only be caused by schema-based reconstruction. They could also be affected by attention, visual perception, guessing, or interference from other stimuli.

A strong evaluation answer should acknowledge that reconstructive memory is a good explanation, but not the only possible one.

Ethics

This type of study is generally low risk, but ethical issues still matter under the British Psychological Society Code of Ethics and Conduct (2009).

Researchers should ensure:

• informed consent
• the right to withdraw
• protection from psychological harm
• confidentiality of participant data
• debriefing after the task

If the full aim was not explained beforehand to reduce demand characteristics, this would involve mild deception. That can be acceptable if participants are fully debriefed and no harm is caused.

Comparison with Bartlett (1932)

Bartlett’s classic study showed reconstructive memory using a story from another culture. Participants changed details to make the story more familiar and logical.

Steyvers and Hemmer supported the same general theory, but with more modern methods:

• more naturalistic visual materials
• quantitative measurement of errors
• computational modelling
• focus on how prior knowledge and memory traces are combined