Loftus & Palmer wanted to study the phenomenon known as reconstructive memory. Bartlett (1932) argued that established knowledge called schemas influence cognitive processing and demonstrated that cultural schemas could distort memory. He argued that humans try to find meaning in what they experience and if something seems unfamiliar they will try to fit the experience into existing schemas.

Loftus & Palmer performed a number of experiments where they demonstrated that people’s memory can be manipulated by post-event information and wording of a question. Loftus has also demonstrated that it is possible to plant a false memory in participants and that they will claim that they have actually experienced the event. Loftus’ research has drawn attention to the problems of eyewitness testimony where people may give false evidence because of leading questions in the courtroom.

The researchers refer to the problems of leading questions in eyewitness testimony in their introduction and argue that some questions are more liable than others to influence estimates than others. Leading questions are defined by Loftus as a question which either by form or content suggests to a witness which answer is desired. Previous research has demonstrated that people’s memory for details after a car accident is inaccurate and that there is a tendency to overestimate the duration of a complex event. It seems that it is particularly difficult to estimate the speed of a moving car (Marshall, 1969). Since previous research had shown that estimation of speed was liable to distortion Loftus and Palmer hypothesized that people’s memory for details of a complex event could be distorted if they were asked to estimate how fast the car was going. Therefore, they set up two experiments where participants were shown videos of traffic accidents and after that, they had to answer questions about the accident. The participants were asked about the speed of the car in different ways. For example, participants were asked:” About how fast were the cars going when they smashed into each other?” or they were asked: “About how fast were the cars going when they hit each other?” This was based on the assumption that ‘hit’ and ‘smashed’ have different connotations and activate different cognitive schemas about the severity of the accident.

The researchers conducted two experiments. Both experiments had student participants.

The aim of the research was to investigate whether the use of leading questions would affect the estimation of speed.

The researchers predicted that using the word ‘smashed’ would result in a higher estimation of speed than using the word ‘hit’. The independent variable was the intensity of the verb used in the critical question and the dependent variable was the estimation of speed. 45 students participated in the experiment. They were divided into five groups of nine students. Seven films of traffic accidents were shown and the length of the films ranged from 5 to 30 seconds. These films were taken from driver’s education films.  The study was an independent samples design; each participant watched all 7 films.

When the participants had watched a film they were asked to give an account of the accident they and seen and then they answered a questionnaire with different questions on the accident with one question being the critical question where they were asked to estimate the speed of the cars involved in the accident. The participants were asked to estimate the speed of the cars. They were asked the same question but the critical question included different words. One group of participants was asked, “About how fast were the cars going when they hit each other?" The critical word "hit’" was replaced by ‘collided’, ‘bumped’ or ‘smashed’ or’ contacted’ in the other conditions.

The mean estimates of speed were highest in the ‘smashed’ condition (40.8 mph) and lowest in the ‘contacted’ group (31.8 mph). The researchers calculated a statistical test and found that their results were significant at p ≤ 0.005.

The results indicate that the critical word in the question consistently affected the participants’ answers to the question. The researchers argued that it may be that the different speed estimates are the result of response-bias, i.e. the participants are uncertain about the exact speed, and therefore a verb like "smashed" biases their response towards a higher estimate. It may also be that the way the question is formed results in a change in the participant’s mental representation of the accident, i.e. the verb "smashed" activates a cognitive schema of a severe accident that may change the participant’s memory of the accident. This distortion of memory is based on reconstruction so that it is not the actual details of the accident that are remembered but rather what is in line with a cognitive schema of a severe accident. This interpretation is in line with Bartlett’s suggestion of reconstructive memory due to schema processing.

In conclusion, it seems that participants’ memory of an accident could be changed by using suggestive questions.

The experiment was conducted in a laboratory and the participants were students. Lab experiments may be problematic in the sense that they do not necessarily reflect how people remember in real life. There may be a problem of low ecological validity. Support for this point could be that the films shown in the experiment were made for teaching purposes and therefore the participants did not experience the same kind of emotion that they would have experienced if it had been a real accident.

However, a strength of the experimental method is that confounding variables can be controlled so that it is really the effect of the independent variable that is measured. This was the case in this experiment and Loftus and Palmer could rightfully claim that they had established a cause-effect relationship between the independent variable (the intensity of the critical word) and the dependent variable (estimation of the speed). The fact that the experiment used students as participants has also been criticized because students are not representative of a general population.  In addition, they were most likely young and inexperienced drivers, so this may have influenced their ability to estimate the speed of the cars.  That being said, most people would have a problem with estimating the speed of a car...

In order to investigate if the differences in speed estimation in the first experiment could be due to the form of the question (schema processing), the researchers performed a second experiment. The aim of the experiment was to investigate if participants who had a high-speed estimate in the first part of an experiment would say that they had seen broken glass in the second part of the experiment. The researchers hypothesized that this would happen.

150 students participated in this experiment. They were randomly allocated to conditions. They were shown a 1-minute film depicting a multiple car accident lasting around 4 seconds. After seeing the film the participants answered a questionnaire. First, they described the accident in their own words, and then they had to answer a number of other questions. Fifty participants were asked:” About how fast were the cars going when they smashed into each other?” Fifty participants were asked:” About how fast were the cars going when they hit each other?” The remaining fifty participants were not asked to estimate speed.

After one week the participants came back to the laboratory to answer some questions about the accident. There was one critical question this time in a list of a total of 10 questions and it was placed randomly in the list in the questionnaire. The critical question was:” Did you see any broken glass?” The participants simply had to answer “yes” or “no”. In fact, there was no broken glass in the accident the participants had seen but the researchers assumed that broken glass was associated with high speed.

The mean estimate of speed by the participants who had the critical question: "About how fast were the cars going then they smashed into each other?” was 10.46 mph. The mean estimate of speed by participants who had the critical question:” About how fast were the cars going when they hit each other?” was 8.00 mph.

In the ‘smashed’ condition 16 participants said yes to having seen broken glass compared to 7 in the ‘hit’ condition. 6 participants in the control condition answered ‘yes’ to the question. 34 participants in the ‘smashed’ condition answered ‘no’ to the question compared to 43 in the ‘hit’ condition. Although most of the participants accurately reported no broken glass, more of the participants in the ‘smashed’ condition said they saw broken glass.

A Chi-square test was calculated because the data was nominal. It was significant at p ≤ 0.025. Therefore, it was concluded that smashed leads to more ‘yes’ responses as well as higher speed estimates.

The researchers argued that the results of the second experiment provided further support for the theory of reconstructive memory and schema processing. The wording of the critical question led to higher speed estimates in the first part of the experiment and this also had consequences for how participants answered in the second part of the experiment. Loftus and Palmer suggest that participants are influenced by the perception of the event but also of the post-event information provided by the critical question. The researchers argue that this information may be integrated in such a way that it is difficult to say where it came from when the participants try to recall the event. The verb used in the critical question provides further information to the participant about the accident. The word ‘smashed’ gives the participant the idea of an accident that is severe and therefore he or she is more likely to think that there was broken glass.

The results of this experiment can be interpreted in terms of Bartlett’s theory of reconstructive memory, i.e. people tend to change details of an event when they try to remember it. This is probably also what happened when the participants in Loftus and Palmer’s study tried to remember the original information when they were given information about the speed of the cars through the use of either ‘hit’ or ‘smashed’. The participants may have used their past knowledge of serious car accidents to make the decision of whether or not they had seen broken glass (schema processing).

This study can also be accused of lacking ecological validity and therefore it may be difficult to generalize the findings to real life. The comments made on the first experiment also apply to the second one.

And finally, a good video clip to help to understand the key concepts of this study.