Home

Methods

Studies

Themes


Background

How accurately do we remember the details of a complex event?  More specifically, how accurate can we be when asked how long an event took, how fast were the cars going or how much time elapse between the horn sounding and the moment of collision?

It is well documented that most people are very inaccurate in reporting details such as time, speed and distance (Bird 1927; Whipple 1909).  Most people have the tendency to overestimate the length of time a complex event took (Block1974; Marshall 1969; Ornstein 1969).  The judgement of speed is particulalry difficult, and practically every automobile accident results in huge variations from one witness to another as to how fast a vehicle was actually travelling (Gardner 1933).  In one test given to U.S. Air Force personell, who knew in advance that they would be watching a moving vehicle, estimates ranged from 10 to 50 mph.  The car they were watching was travelling at 12 mph (Marshall, 1969, p 23).

Given the inaccuracies in estimates of speed it seems there are variables that are potentially powerful in terms of influencing these estimates.  The present study was conducted to investigate one such variable, namely, the phrasing of the question used to elicit the estimate of the speed of a moving vehicle.  Some questions are clearly more suggestive than others.  This fact of life has resulted in the concept of a leading question and in legal rules indicating when leading questions are allowed (Supreme Court Reporter, 1973).  A leading question is one that either by its form or content, suggests to the witness what answer is desired or leads him to the desired answer.

Using the words "smashed" and "hit" when asking participants about the speed of the moving vehicle, it has been suggested by Fillmore (1971) and Bransford and McCarrell (in press) that smashed and hit may suggest differential rates of movement.  Furthermore, the two verbs may may also indicate a difference in the likley consequences of the event to which they are referring.  The impact of the accident is apparently gentler for hit than for smashed.

Authors

Elizabeth F. Loftus and John C. Palmer

Method

Both parts of this study were experiments, each of the had an Independent Variable that changed.

Samples

Experiment 1

Forty five students participated in groups of various sizes.

Experiement 2

One hundred and fifty students participated in groups of various sizes.

Procedure

Experiment 1

The participants watched seven films, taken from a longer film for Driver Education, each of the segments lasted between 5 and 30 seconds.  Following each film, the students were given a questionnaire.  This asked them to give an account of what they had seen, in their own words, and then there was a series of more specific questions.  The critcal question was, nine participants were asked "About how fast were the cars going when they hit each other?"  Equal numbers of the remaining participants were asked the same question with the verb changed to "smashed", "collided", "bumped", and "contacted".  The whole process lasted for about one hour and a half.  Films were shown in a different order to each of the groups of participants.

Experiment 2

A film depicting a multiple car accident was shown.  Following the film the students were given a questionnaire that first asked them to describe, in thier own words, what they had seen, and then to be asked a series of more specific questions.  The critical question was the question that asked about the speed of the cars.  Fifty students were asked  "About what speed were the cars going when they smashed into each other?"  Gifty students were asked "About how fast were the cars going when the hit each other?"  Fifty student were not asked about the speed of the car at all.

One week later, the students returned and, without seeing the film again, they answered a series of questions about the accident.  The critical question was "Did you see any broken glass?"  Obviously, the answer would be either "Yes" or "No".  This question was contained in a list of 10 questions and its position was moved around at random.  Just to be clear, there was no broken glass visible in the film of the accident.

Results

Experiment 1

Speed estimates for the verbs used in Experiment 1

Verb

Mean Speed Estimate

Smashed

40.5

Collided

39.3

Bumped

38.1

Hit

34.0

Contacted

31.8


An analysis of variance was carried out with verbs as the fixed effect and the participants and films as random effects, yielding a signigficant quasi F ratio, F'(5,55) = 4.65 p<0.005.

Four of the seven films were staged films and they were created to educate drivers, and the speeds of these cars was known.  One car was travelling at 20 mph, one at 30 mph and two at 40 mph.  The mean estimate of speed of these four film were 37.7mph, 36.2 mph, and 36.1 mph respectively.  In agreement with previous work, people are not very good at judging how fast a vehicle was actually travelling.

Experiment 2

The mean estimate of speed for participants who were asked the crical question with the verb smashed was 10.46 mph and with the verb hit was 8.00 mph.  These means are statistically significantly different, t(98) = 2.00, p< 0.05

Distribution of "Yes amd "No" responses to the question "Did you see Broken glass?"

Verb Condition

Response

Smashed

Hit

Control

Yes

16

7

6

No

34

43

44

An idependent chi-square test on these responses was significant beyond the 0.25 level.

The important thing about the table above, is that if you double the numbers in the table it will give you the probability of these answers being given.  So, the probability of  saying "yes" P(Y) to the question about broken glass is 32% or 0.32, when the verb is smashed is used, and P(Y) for hit  is 14% or 0.14.  Therfore, smashed leads both to more "Yes" responses and to higher speed estimates.  It appears to be the case that the effect of the verb is mediated, at least in part, by the speed estimate.  The question now arises: Is smashed doing anything else besides increasing the estimate of the speed?  To answer this, the function relating to P(Y) to speed estimate was calculated separately for smashed and hit.  If the speed estimate is the only way in which the verb is mediated, then for a given speed estimate P(Y) should be independent of the verb.

P(Y) to "Did you see broken glass, separated out on speed estimates

Speed estimate mph

Verb condition

1-5

6-10

11-15

16-20

Smashed

0.09

0.27

0,41

0.62

Hit

0.06

0.09

0.25

0.50

The values in the table above for the P(Y) at different speeds for the different verbs does not seem to produce similar values,  Some values are closer together, than others, but it is entirely possible that smashed has other effects besides increasing the estimate of speed.  This will be discussed in the next section.

(Please Note: I carried out a t-test on the data above (Unrelated and one-tailed) and found that there was a highly significant difference between the values for "smashed" and the values for "hit".  You can check my maths here.)

Discussion

We have demonstrated something that had been known for a long time, namely, that the way a questionnaire is asked can enourmously influence the answer that is given.  In this instance, the question, "About how fast were the cars going when they smashed into each other?" led to higher estimates of speed than thesame question asked with the verb smashed replaced by hit.  Furthermore, this seemingly small change had consequences for how questions were answered a week after the original event occurred.

As a framework for discussung these results we would like to propose that two kinds of information go into one's memory for some complex occurrence.  The first is the information that is from the original event; the second is external information supplied after the fact.  Over time, information from these two sources may be integrated in such a way that we cannot tell from which the specific detail is recalled.  All we have is one memeory.

Discussing the present experiments in these terms, we propose that the participant first forms a memory of the accident he has just seen on film.  The experimenter then, while asking "About how fast were the cars going when they smashed into each other?" supplies a piece of extra information, namely, that the cars indeed did smash into each other.  When these two pieces of information are integrated, the participant has a memory of an accident that was more severe than in fact it was.  Since broken glass is commensurate with a severe accident, the subject is more likley to think that broken glass was present.

There is some connection between the present work and earlier work on the influence  of verbal labels on memory for visually presented form stimuli.  A classic study in psychology showed that when participants are  asked to reproduce a visually presented form, their drawings tend to err in the direction of a more familiar object suggested by a verbal label initially associated with the to-be-remembered form (Carmichael, Hogan & walter, 1932).  More recently, Daniel (1972) showed that recognition memory as well as reproductive memory, was similarly affected by verbal labels, and he concluded that the verbal label causes a shift in the memory strength of forms which are better representatives of the label.

When the experimenter asks the participant the question "Aboult how fast were the cars going when they smashed into each other?", he is effectively labeling the accident as a smash.  Extrapolating the conclusions of Daniel to this situation, it is natural to conclude that the label, smash, causes a shift in the memory representation of the accident in the direction of being more similar to a representation suggested by the verbal label.



  Designed by Tony, Powered by IONOS