The Verbal Overshadowing Effect in Face Recognition: A Critical Examination

Bhavneet Kaur Ahuja

Department of Psychology, Trent University

PSYC 2019H: Basic Research Methods & Stats

Tuesday 12:00

Dr. Kevin Peters

March 21, 2022

In the first order, The robber was correctly identified in the photo line-up by 36.11% of

participants in the Experimental Group, who completed the task of face verbalization. In

contrast, 56.72% of participants in the Control Group, correctly identified the robber, who

completed an unrelated writing task. Thus, accuracy was 20.61% worse in the Experimental

Group, which suggests that the accuracy of being able to recall is worse when face

verbalization is involved.

This difference was statistically significant, X2(1) = 5.93, p = 0.015. Accuracy was 20.61%,

95%CI [4.36%, 36.85%] worse in the Experimental Group, which suggests that the accuracy

of being able to recall is worse when face verbalization is involved.

In the second order, The robber was correctly identified in the photo line-up by 39.19% of

participants in the Experimental Group, who completed the task of face verbalization but

were also made to perform an extra visual task and the task was delayed by 10 mins. In

contrast, 53.95% of participants in the Control Group, correctly identified the robber, who

completed an unrelated writing task. Thus, accuracy was 14.76% worse in the Experimental

Group, which suggests that the accuracy of being able to recall is worse when face

verbalization is involved.

This difference was not statistically significant, X2(1) = 3.28, p = 0.701. Thus, accuracy was

14.76%, 95%CI [-1.03%, 30.54%] worse in the Experimental Group, which suggests that the

accuracy of being able to recall is not really affected when face verbalization is involved.

Calculating the Cramer’s V measure of Effect Size for Order 1 and Order 2.

For Order 1, Cramer’s V= 0.21, therefore, it is a small to medium effect size (Cohen,1992).

For Order 2, Cramer’s V= 0.15, therefore, it is a small effect size (Cohen,1992).

25
20
15
10
5
0
Experimental Control
Conditions
Order 1 Order 2
Fig 1: Order 1 and Order 2 accuracy rates of recognising the robber
following visual recollection and visual recollection accompanied
with face verbalization.

Discussion

In all studies, the influence of face verbalization on visual recall was investigated. The

experimental group participants were assigned to write a description of the robber for 5

minutes after showing them a 44-sec video of a bank robbery and then they were made to

perform the face recognition task in which they had to pick the robber out of a photo line-up.

According to the study's findings, individuals in the experimental group did not recognise the

robber as precisely as those in the control group, demonstrating that verbalising the robber

did not improve their visual memory/recall ability.

The results of my experiment did not support my prediction for Order 1. I predicted that the

results of the study will differ from the results of the original study by Engstler-Schooler

(1990) when the participants belong to younger age group that is between the age of 15-20

years since, the younger participants might have less accurate visual memory than older

adults and that may produce less exact results. But on the contrary the results were similar to

that of the original study by Engstler-Schooler (1990), Their study showed that when items

that need to be recalled are faces, vocal practise worsens rather than aids memory
performance. Similarly, the results of my study showed that the ability of being able to recall

is worse when face verbalization is involved. The chi-square test for this study revealed that

the result is statistically significant, having a statistical value of 5.93 (Schooler & Engstler-

Schooler, 1990) which is exactly the same as my study.

Dodson, Johnson & Schooler replicated the experiment by Schooler & Engstler-Schooler in

1997, with the goal of learning more about the verbal overshadowing effect and how it affects

visual memory and facial recognition. In their experiment, they changed the order of the

filler/distracter task and the writing task to see if removing the delay in describing the robber

had any effect on the expected results of the study. The results of this study were also

statistically significant, and the change made to the order of the experiment did not have any

effect on the expected results (Dodson.et al, 1997).

The results of my experiment did not support my prediction for Order 2. I predicted that the

results of my study will differ from the results of the original study by Dodson, Johnson &

Schooler, 1997 when we change the intervals at which the filler tasks are performed to

observe whether delaying or doing a task early has an impact on the findings. But on the

contrary the results were similar to that of the original study by Dodson, Johnson & Schooler,

(1997), Their study showed that the change made to the order of the experiment did not have

any effect on the expected results (Dodson.et al, 1997). Similarly, our results showed that

change made to the interval at which filler tasks are performed did not have any effect on the

expected results that is the accuracy of being able to recall is not really affected when face

verbalization is involved. The results of my study were not statistically significant, X2(1) =

3.28, p = 0.701. Thus, Cramer’s V= 0.15, therefore, it is a small effect size (Cohen,1992).
Previous research shown that when identifying objects visually and vocally, visual memory is

more reliable than verbal recall. Even though both visual and verbal recall are independent

processes that occur in the brain, combining the two does not substantially aid in visual

recollection. Memory is more accessible when it is visual rather than verbal (Hall et al.,

2018). These findings are consistent with the findings of the Order 1 and Order 2

experiments.

The replication of this study was crucial in confirming the findings of the first experiment, as

well as identifying and correcting previous errors, and opening up new opportunities for

further research into the subject of verbal overshadowing. According to our study’s result we

can say that the way police officers interview witnesses of crimes, the witnesses may miss

key information since, verbal description can hinder recognition accuracy. So, instead of an

on-the-spot verbal interview the police officers could rather take a interview through a purely

visual task (drawing) which would be even more effective in order to collected as much

information as possible. As per the findings of our experiment, Verbalization could affect

accuracy of recall of information.

Further, we could extend research in this area by taking into account the mistakes made by

the participants while performing the Face recognition task that is while picking the robber

out of a photo line-up. The conditions were not ideal for this in our study thus we do not

know if this could have had a potential effect on the studies result.
 References

Dodson, C. S., Johnson, M. K., & Schooler, J. W. (1997). The verbal overshadowing effect:
Why descriptions impair face recognition. Memory & Cognition, 25(2), 129-139.
doi:http://dx.doi.org/10.3758/BF03201107

Hall, E., Bainbridge, W., & Baker, C. (2018). Comparing memory based on visual recall,
visual recognition, and verbal recall. Journal of Vision, 18(10), 825.
https://doi.org/10.1167/18.10.825

Schooler, J. W., & Engstler-Schooler, T. (1990). Verbal overshadowing of visual memories:

Order 1 R Syntax and Output

In R type: > prop.test(x = c(38,26), n = c(67,72), correct = FALSE)

Output (To know if the difference is statistically significant or not)

2-sample test for equality of proportions without continuity correction

data: c(38, 26) out of c(67, 72)

X-squared = 5.9312, df = 1, p-value = 0.01488

alternative hypothesis: two.sided

95 percent confidence interval:

0.04362043 0.36848571

sample estimates:

prop 1 prop 2

0.5671642 0.3611111

What percentage of participants in each group correctly identified the robber?

In R type: 26/72*100

= 36.11% in the Experimental Group

In R type: 38/67*100

= 56.72% in the Control Group

What is the difference between these two percentages?

In R type: 36.11-56.72

= - 20.61%.

Order 2 R Syntax and Output

In R type: > prop.test(x = c(41,29), n = c(76, 74), correct = FALSE)

Output

2-sample test for equality of proportions without continuity correction

X-squared = 3.2811, df = 1, p-value = 0.07008

alternative hypothesis: two.sided

95 percent confidence interval:

-0.01030681 0.30547039

sample estimates:

prop 1 prop 2

0.5394737 0.3918919

What percentage of participants in each group correctly identified the robber?

In R type: 29/74*100

= 39.19% in the Experimental Group

In R type: 41/76*100

= 53.95% in the Control Group

What is the difference between these two percentages?

In R type: 39.19 - 53.95

= - 14.76%