In 1974 Loftus and Palmer conducted a study to test their theory; that leading questions can have a significant affect upon later recall especially that of eye-witness testimony. Their study involved participants watching a short video of a car crash. Subsequent questions were then asked, including estimating the speed of the vehicle involved in the accident. The question was "about how fast was the car travelling when it smashed into the other car?" Certain participants received this question, others had the same question but the word 'smashed' was replaced with either contacted, hit, bumped or collided. Results revealed a correlation; the more extreme the verb (smashed), the higher the estimated speed. In fact, the mean estimated speed of the 'contacted' group was 32mph, whereas the 'smashed' condition was estimated at 42mph.
Such results can be interpreted in two ways. A participant may have been undecided between two speeds and the presence of the word 'smashed' may bias their opinion toward the higher speed. Or, and slightly further fetched is the idea that the form of the question could alter a participant's memory representation of the incident.
Our study aimed to reinforce and support Loftus' and Palmer's theory producing a similar trend in results, with the hope of reinforcing further to the criminal justice system to what extent eye-witness testimonies are unreliable. The null hypothesis was that the wording of the question will have no effect on the answer given. The alternative hypothesis is that the more extreme the verb used is, the higher the estimation of speed reported will be.
Method
Participants
30 non-psychology students from Keele University were selected at random from an opportunity sample to take part in this experiment. Within this experiment were 16 females and 14 Males. The mean age of the participants was 21.86 (SD= 15.75)
Materials
During the experiment the materials used consisted of a video that can be obtained from the following website
, this video was shown to the participants and stopped after the first 9 second of showing, and a questionnaire consisting of 7 questions based on the video for participants to answer, a copy of the questionnaire used can be found in the appendix.
Design
Participants were randomly assigned to each of the three conditions upon entering the experimental room. Participants only provided data for one of the three conditions meaning an independent design was employed. Participants were tested separately so as conferring between participants did not occur, they were also requested not to speak of the experiment to any other potential participants upon completion of the experiment.
The independent variable in this study was the 4th question “How fast did you perceive the car to be travelling when it…into the other car?” in the questionnaire and the word used either, “smashed”, “bumped” or “collided”. The DV was the response of the participant to the question. The entire test was conducted under experimental conditions.
Procedure
Participants were separated and one by one entered the experiment room with the experimenter to take part in the study. Participants were then asked to sign a consent form (see appendix) and asked if they had any further questions, all participants were informed of their right to withdraw themselves or their data from the experiment at any time, they were also aware that participation was 100% voluntary. After the consent form was signed the experiment began and all noises and visual distractions were blocked from the experiment room by closing doors, windows and curtains etc.
Once seated approximately 2 metres away from the computer participants were talked through the procedure by the experimenter and asked once again if they had any questions or wished to withdraw from the study. Only the experimenter and the participant were in the room.
The experimenter then started the video clip and stopped it after 9 seconds. The participant was then given a questionnaire. The participant read the sheet and answered the questions by writing on the sheet. The questionnaire was then collected by the experimenter and gave the participant a debriefing sheet explaining the purpose and aim of the experiment; participants were once again given the opportunity to withdraw their data and information.
This experimental procedure was carried out 30 times with 30 different participants. 3 different questionnaires were distributed (10 of each) between the participants at random without them knowing this. All questionnaires were identical apart from one word in one question. The question “How fast did you perceive the car to be travelling when it…into the other car?” varied and in the first condition the questionnaire asked “How fast did you perceive the car to be travelling when it bumped into the other car?” In the second condition the questionnaire asked “How fast did you perceive the car to be travelling when it collided into the other car?” and in the third condition the questionnaire asked “How fast did you perceive the car to be travelling when it smashed into the other car?” These three words are different in severity smashed being the most severe and bumped being the least. It was predicted that the more severe the word used the higher the speed estimated by the participant.
Results
In order to present the results into an easily accessible format, the means and standard deviations need to be calculated. The raw data for each condition were added together then divided by N = 10 (Number of Participants) to give the mean. Then each raw data score was subtracted by the mean then squared, all these scores were then added together then divided by N-1 (9) and finally square rooted to provide the standard deviation. These results are present in Table 1:
Table 1:
Means and Standard Deviation’s of the estimated vehicle speeds under the leading question conditions (mph)
Verb used in leading question
Bumped Collided Smashed
Mean 8.6 11.7 18.5
SD 3.169 2.541 4.836
It is clear to see that in all three conditions the values are within one standard deviation of the mean. As the standard deviations are close to zero, it reflects how all the raw data clusters tightly around the mean, indicating there are few anomalies and shows that the data is reliable and thus can be linked to the hypothesis. Although the collided condition deviates slightly more. The mean scores indicate the average estimated speed of the vehicles in all three conditions. It reflects the original assumptions that the estimated speed will in increase with the severity of the verb used. This is clearly shown here where the average speed for the smashed condition is 8.6 and then 18.5 mph, showing there’s an almost 10 mph difference between the two means.
Table 2
ANOVA summary table for the effect of leading questions on eyewitness testimony
Source Sum of squares df Mean squares F p
Between-groups 512.8 2 256.4 19.284 <0.05
Within-groups 359 27 13.296
There was a significant difference in the estimated speed of the car (mph) across the bumped, collided and smashed conditions, F (df 2, df 27) = 19.286 MS error= 13.296 p. <0.05. Analytical comparisons showed that there was no significant difference between the estimated speeds of the bumped and collided conditions: F (df 2,df 27)= 1.3 p >0.05, and that there was a significant difference between the collided and smashed conditions: F (df 2,df 27)= 13.65 p< 0.05. There was also a significant difference between the final Smashed and Bumped conditions: F (df 2, df 27)= 17.54 p< 0.05.
Discussion
Results were consistent with Loftus and Palmer’s (1974) research; that leading questions can affect a participant’s response. It appears from the results of this study that the findings of Loftus and Palmer, that the wording of a question can alter responses, are still applicable to the modern day.
However, this study is limited in some aspects. For example, for issues of time and practicality, it was impossible to follow up the participants after a considerable amount of time and question them about the car incident clip they watched, as Loftus and Palmer (1974) did in their original study. In addition, Loftus and Palmer (1974) showed seven film clips of car incidents. Again, this was not possible in this small scale replication due to time and practical limitations.
It is also debateable as to whether the choices of verb (bumped, collided and smashed) were appropriate for this study. Loftus and Palmer (1974) used 5 different verbs; smashed, collided, bumped, hit, and contacted. It may have been more effective in this study to have used verbs that vary more in severity. In particular, it can be argued that the verbs ‘bumped’ and ‘collided’ are too similar in severity. It’s also well known that people tend to interpret some words differently and his may too have affected the results of the study. One way of overcoming this could have been to ask the participants to rate the words in order of severity once they have completed the experiment, by doing this we could have gained complete certainty that the words used were effective.
The laboratory setting has also been criticised for creating an artificial environment. This raises issues of ecological validity, as participants are unlikely to act the same in the laboratory as in court. In addition, the emotional states involved when witnessing a real life car incident had been eliminated as its unlikely that by watching a video that the normal levels of stress encountered from viewing a crash would occur. This may also have affected the accounts given by the eye-witnesses.
One possible explanation for the results found and explaining the behaviour discovered in psychological terms could be because firstly the information is achieved primarily from the perception of the event. Then possibly external information after the event has occurred interferes with the original perception of the event. These two factors then integrate with one another to such an extent it is difficult to distinguish what was recalled at the time of the event or afterwards. Simply there is one memory to hold all of this information so it’s likely that inference will occur. In terms of the current study, participants formed a representation of the crash.
The study successfully supports the hypothesis that the more extreme the verb used is, the higher the estimation of speed reported will be. However there are many areas which could have been looked it to further in order to ensure that the results found are even more reliable.
References
Loftus, E. F., & Palmer, J. C. (1974). Reconstruction of Automobile Destruction: An Example of the Interaction Between Language and Memory. Journal of verbal learning and verbal behaviour 13, 585-589. Washington: Academic Press
Hinton, P. Statistics explained (2004). Statistics explained (2nd edition). East sussex: Routledge.
Carlson, N.R. (1994). Physiology of behaviour (5th edition). London: Allyn and Bacon.
Appendix
Calculations for carrying out an unrelated one-way ANOVA:
- Calculate means
Bumped mean (sum of observations/ n) = 8.6
Collided mean (sum of observations/ n) = 11.7
Smashed mean (sum of observations/ n) = 18.5
Grand mean (mean of 3 means) = 12.93
- Between-groups variance
- Between-groups sum of squares
- Between-groups degrees of freedom
Degrees of freedom = a – 1
3 – 1 = 2
- Between-groups variance
Between-groups variance = sum of squares/ df
512.8/ 2 = 256.4
- Within-groups variance
- within-groups sum of squares
Bumped
Collided
Smashed
Within-groups sum of squares = 359
- within-groups degrees of freedom
Degrees of freedom = a (n – 1)
3 (10 – 1) = 27
- within-groups variance
within-groups variance = sum of squares / df
359/ 27 = 13.30
- ANOVA summary table
Car incident Questionnaire
Consent and information form
You will now watch a short video clip, after which you will be asked to fill in a short questionnaire concerning the video clip.
The only personal information you will be required to give is your age and gender. All results will be anonymous.
If you wish to withdraw your data at any point of the experiment or after you will be allowed to do so.
Please ask the experimenter now if you have any questions concerning the study.
If you are happy to continue and participate in the experiment, please sign and print your name below:
NAME (Block capitals)………………………….
Signature:………………………………
Date:
Car incident Questionnaire
- Are you: Male / Female ? (Please circle)
- How old are you? ……………
- What colour were the cars?................................
- How fast did you perceive the car to be travelling when it collided into the other car?........................................ (mph)
- Was there any broken glass on the floor? Yes / No (Please circle)
- Was the incident in a city / built up area? Yes / No (Please circle)
Thank you for taking part in the experiment, please can you now hand your questionnaire to the experimenter.
Debrief
Thank you for taking part in this study.
The purpose of the study was to investigate the effects of leading questions on eye witness testimony.
We hypothesised that the wording of certain questions would affect participant responses.
During the experiment you were asked to answer numerous questions, some of which were simply a distracter as we were only concerned with question four (The speed of the vehicle).
If you would like any information about the results of the study or if you wish to withdraw your responses at any time, please email: