Steyvers & Hemmer (2012) Reconstructive Memory (Edexcel A-Level Psychology): Revision Notes

Steyvers & Hemmer (2012) Reconstructive Memory

Background

Previous research into reconstructive memory used experimental designs that deliberately created conditions likely to produce memory errors. For example, participants might be asked to recall objects from a photograph of an office that contained no books or computers. Researchers argued that removing highly probable items from such scenes would induce false memories, suggesting memory was unreliable and prone to error.

However, Steyvers and Hemmer challenged this interpretation. They proposed that these errors occurred primarily because of experimental manipulation of the environment, such as withholding expected objects from images. This approach did not reflect how memory functions in everyday, naturalistic contexts. They argued that memory should be studied in more realistic settings where environments are not artificially altered to elicit specific responses. This would provide better insight into how episodic memory (memory for specific events) and semantic memory (general knowledge about the world) interact during recall.

Participants

The investigation involved three separate groups:

• 22 participants completed the expectation test to assess prior knowledge
• 25 participants completed the perception test to identify visible objects
• 49 participants took part in the main memory experiment

All participants were recruited from the experimental participant pool at the University of California, Irvine.

Aim

The research aimed to investigate the interaction between episodic and semantic memory in naturalistic environments. Specifically, the researchers wanted to examine how prior knowledge (stored in semantic memory) influences the reconstruction of memory when recalling photographs of normal everyday settings such as hotels, kitchens and offices.

Procedure

Initial testing

The first phase assessed participants' prior knowledge and expectations about naturalistic scenes. This was essential to understand what objects people typically expected to find in different environments.

One group of 22 participants completed an expectation test. They were asked to list objects they would expect to find in five naturalistic scenes: office, kitchen, hotel, urban and dining. Participants entered their responses on a computer, spending at least one minute per scene. The frequency with which objects were named provided a measure of prior expectations for each scene type.

A separate group of 25 participants viewed 25 images of the five scenes (five images of each scene type). They completed a perception test in which they named all objects they could see in each image. This served as a control to ensure that objects not identified later in the memory test had actually been visible in the images, rather than simply overlooked.

The researchers analysed the frequency with which objects were named in both tests. The top ten most frequently recalled objects were identified, along with low-frequency recall objects. Results showed that high-frequency objects tended to be iconic items for each scene.

The experimental memory condition

Using the same participant pool, 49 participants who had not taken part in either previous test were randomly selected. Ten stimulus images from the prior tests were chosen for the experiment – two from each scene type, specifically those that had elicited the most objects named in the perception test. Two sets of five images were created, with one image from each scene type per set. Participants viewed only one set of five images to avoid carry-over effects from seeing multiple images of the same scene type.

The key manipulation involved exposure duration. Participants viewed the five images for either 2 or 10 seconds to control for how long they had to encode the information. Four possible trial time orderings were created, and participants were randomly allocated to one of these sequences.

For example, one participant might see:

• Kitchen scene: 10 seconds
• Hotel scene: 10 seconds
• Urban scene: 10 seconds
• Dining scene: 2 seconds
• Office scene: 2 seconds

The exposure duration manipulation aimed to alter the extent to which participants relied on prior knowledge during episodic memory retrieval. The researchers hypothesised that recall from short exposure of 2 seconds would rely more heavily on prior knowledge, as the event would have had limited time to be encoded as an episodic memory.

The trials were randomised and participants completed free recall – recalling objects they remembered from each scene in their own time and order. Free recall refers to recall of stimulus material in any order and without memory cues.

The researchers recorded all objects recalled by participants and the order in which they were remembered. Responses were normalised to remove plurals (treating 'chairs' as 'chair') and additional descriptive content (treating 'silver car' as 'car').

Findings

Analysis of errors

Recall of objects that had also been listed in the perception test was recorded as accurate.

The mean number of objects recalled during free recall was:

• 2-second exposure duration: 7.75 objects
• 10-second exposure duration: 10.05 objects

Steyvers and Hemmer analysed error rates based on whether objects had high or low probability of recall according to the expectation test. They found that incorrect recall of highly probable objects was only 9%, whilst incorrect recall of low probability objects was 18%.

This indicates that when participants view scenes representative of naturalistic contexts that are not experimentally manipulated, memory of such scenes is quite accurate. Where scenes did not accurately represent real-life contexts, such as a dining scene without a tablecloth, the error rate increased to 19%.

The effect of prior knowledge

The effect of prior knowledge (semantic memory) was assessed by comparing the correct number of objects guessed in the expectation test to those objects actually recalled in the two experimental conditions (2-second and 10-second duration exposure). The cumulative accuracy of object guesses based on the expectation test was over 55% compared to semantic memory under initial testing, and the actual recall in both conditions was much higher, exceeding 80%. This suggests that episodic memory played a substantial role in recall.

Interestingly, longer duration improved recall overall. With short exposure to the picture, seven objects were correctly recalled on average, compared with nine objects recalled with longer exposure to the picture.

Conclusions

The research demonstrates that in recall of naturalistic scenes, prior knowledge drawn from semantic memory can contribute to accurate recall in episodic memory tasks, when such scenes are unmanipulated. People draw on general knowledge as effective guesses of what is expected to be seen in particular contexts. Prior knowledge contributes substantially to recall of naturalistic environments, but this does not occur at the expense of accuracy. In fact, people are more likely to notice novel items more readily than previous research might suggest.

When using unaltered naturalistic contexts, guesses can be effective because of the high probability of the objects being present. This guessing frees up cognitive resources to be better spent focusing on novel and unexpected objects in a scene. In this sense, both recall of inconsistent and consistent objects benefits from using a more ecologically valid approach.

Evaluation

Strengths

Ecological validity: Steyvers and Hemmer are strong advocates for increased ecological validity in memory research to enable generalisation of findings to everyday memory use. Their use of naturalistic scenes (photographs of hotels, kitchens, offices, urban and dining settings) represents a more realistic approach than many laboratory studies. This allows for better understanding of how memory functions in real-world contexts.

Important controls: Several controls were implemented to strengthen the research design. Participants only viewed one image from each of the five scenes rather than multiple images of each scene type. This prevented interference from a previously viewed scene of a similar nature affecting subsequent recall. Time orderings were manipulated using a Latin square design, and participants were randomly allocated to one of the time ordering sequences to control for order effects.

Practical applications: The research has important implications for eyewitness testimony in the justice system. It suggests that, contrary to previous research, prior knowledge from semantic memory can enhance recall of episodic events and may even allow greater cognitive effort to be spent on recognising unexpected features of a context. This implies that eyewitnesses are effective when recalling from familiar contexts and effective at encoding novel features.

Weaknesses

Limited naturalism: Whilst Steyvers and Hemmer acknowledge their research is not as naturalistic as it could have been, there are clear limitations. The study used photographs rather than exposing participants to real environments, and employed laboratory conditions rather than real settings. As such, the research attempts to establish greater ecological validity in memory research but does so without fully compromising generalisability and operationalisation of concepts. It represents a step towards greater realism rather than complete naturalism.

Participant sample: All participants were recruited from the University of California, Irvine experimental participant pool, suggesting they were likely university students. This raises questions about the generalisability of findings to other age groups and populations. University students may have different prior knowledge structures or memory capabilities compared to other demographic groups.

Artificial recall conditions: Although the stimuli were naturalistic, the recall conditions remained artificial. Participants knew they would be tested on their memory, which may have affected their encoding strategies. In real-world situations, people do not typically know in advance which events they will need to remember, which could affect how memory operates naturally.