EXPLORATION 1.3: Do People Use Facial Prototyping?

`Exploration 1.3: Do People Use Facial Prototyping?

Name(s): Dionne Torres and Brianna Oakley

A study in Psychonomic Bulletin and Review (Lea, Thomas, Lamkin, & Bell, 2007) presented evidence
that “people use facial prototypes when they encounter different names.” Participants were given two
faces and asked to identify which one was Tim and which one was Bob. The researchers wrote that their
participants “overwhelmingly agreed” on which face belonged to Tim and which face belonged to Bob,
but did not provide the exact results of their study.

STEP 1. Ask a research question. We have collected data from classes at Hollins this fall to investigate
the research question of whether students have a tendency to associate certain facial features with a
name.

STEP 2. Design a study and collect data. Each student in our sample was shown the same two pictures
of men’s faces used in the research study. You were asked to assign the name Bob to one photo and the
name Tim to the other. Each student then submitted the name that he or she assigned to the picture on
the left.
1. Identify the observational units in this study.
[The students that participated in the study.]

2. Identify the variable. Is the variable categorical or quantitative?

[The variable is the students’ assignments of Tim and Bob to the two pictures of men shown. It is categorical

The parameter of interest here is the long-run probability that a Hollins student would assign the
name Tim to the face on the left.

3. State the null and alternative hypotheses to be tested when the data are collected. Express these both
in words and symbols. (Hint: Think about the parameter and the research question of interest here.)

[The null hypothesis is that the probability of assigning the name Tim to the face on the left is equal to 0.5. H0: =

Consider these two photos:

Melissa A. Lea/Robin D. Thomas/Nathan A. Lamkin/Aaron Bell © Psychonomic Bulletin & Review 2007, 14 (5), 901-907.

4. Do you think the face on the left is Bob or Tim? Tim

pg. 1
 EXPLORATION 1.3: Do People Use Facial Prototyping?

The responses from the STATs classes are posted on the course schedule page. In particular, 30
students said the names were Tim on the left and Bob on the right and 6 students said Bob was on
the left and Tim on the right. Use the collection of responses to answer the next few questions.

STEP 3. Explore the data.

5. How many students put Tim as the name on the left? How many students participated in this study
(sample size)? What proportion put Tim’s name on the left?

Number who chose Tim for the picture on the left: [30]
Sample size: [36]
Proportion who chose Tim for the picture on the left (write as a decimal:) [0.8333]

When we conduct analyses with binary variables, we often call one of the outcomes a “success” and the
other a “failure,” and then focus the analysis on the ‘success’ outcome. It is arbitrary which outcome is
defined to be a success, but you need to make sure you do so consistently throughout the analysis. In this
case we’ll call “Tim on left” a success because that’s what previous studies have found to be a popular
choice.

STEP 4. Draw inferences beyond the data. You will use the One Proportion applet to investigate
how surprising the observed class statistic would be, if students were just randomly selecting which
name to put with which face.

6. Before you use the applet, indicate what you will enter for the following values:
(a) Probability of success: [0.50]
(b) Sample size: [36]
(c) Number of repetitions: [50000]

7. Conduct this simulation analysis. Make sure the Proportion of heads button is selected in the
applet and not Number of heads.
(a) Indicate how to calculate the approximate p-value: i.e., you would count the number of simulated
statistics that equal [0.8333] or more.

pg. 2
 EXPLORATION 1.3: Do People Use Facial Prototyping?

(b) Report the approximate p-value. Insert a screenshot justifying your answer.
Approximate p-value = [0]

Screenshot:

(c) Use the p-value to evaluate the strength of evidence provided by the sample data against the null
hypothesis, in favor of the alternative that students really do tend to assign the name Tim (as the
researchers predicted) to the face on the left.
[The p-value, 0, is less than 0.01. This indicates strong evidence against the null hypothesis.]

The p-value is the most common way to evaluate strength of evidence, but now we will explore a
common alternative way to evaluate strength of evidence. The goal of any measure of strength of
evidence is to use a number to assess whether the observed statistic falls in the tail of the null
distribution (and is therefore surprising if the null hypothesis is true) or among the typical values we see
when the null hypothesis is true.

8. Check the Summary Stats box in the applet.

(a) Report the mean (average) value of the simulated statistics. [0.500]

(b) Explain why it makes sense that this mean is close to 0.5.

[It makes sense that the mean is close to 0.50 because there are only two options — assigning the name Tim

(c) Report the standard deviation (SD) of the simulated statistics. [0.083]

pg. 3
 EXPLORATION 1.3: Do People Use Facial Prototyping?

a. Report (again) the observed class value of the statistic. (What proportion of the students in the
Hollins STAT classes put Tim’s name on the left?) p̂ = [0.8333]

b. Calculate how many standard deviations the observed class value of the statistic is from the
hypothesized mean of the null distribution, 0.50. In other words, subtract the 0.50 from the
observed value, and then divide by the standard deviation. In still other words, calculate:

(0.8333 (p̂ ) – 0.50) = (p̂ – π) ≈ [4.015]

0.083

Your calculation in #8e is called standardizing the statistic. It is telling us how far above the mean the
observed statistic is in terms of the ‘how many standard deviations.’

Definition
The p-value is estimated as the proportion of simulated statistics in the null distribution that
To standardize a statistic, compute the distance of the statistic from the (hypothesized) mean
of the null distribution and divide by the standard deviation of the null distribution.

Once you calculate this value, you interpret it as “how many standard deviations the observed statistic
falls from the hypothesized parameter value.”

The next question is how to evaluate strength of evidence against the null hypothesis based on a
standardized value. Here are some guidelines:

Guidelines for evaluating strength of evidence from standardized values of statistics

Standardizing gives us a quick, informal way to evaluate the strength of evidence against the null
hypothesis. For standardized statistics:
between – 1.5 and 1.5 little or no evidence against the null hypothesis
below – 1.5 or above 1.5 moderate evidence against the null hypothesis
below – 2 or above 2 strong evidence against the null hypothesis
below –3 or above 3 very strong evidence against the null hypothesis

The diagram in Figure 1.13 illustrates the basis for using a standardized statistic to assess strength of
evidence.

Figure 1.3.4 Positions of standardized statistics for a bell-shaped distribution.

pg. 4
 EXPLORATION 1.3: Do People Use Facial Prototyping?

The figure can be summarized by the following key idea.

Key Idea
Observations that fall more than 2 or 3 standard deviations from the mean can be considered
in the tail of the distribution.

STEP 5: Formulate conclusions.

9. Let’s examine the strength of evidence against the null.
(a) Based on the standardized statistic, z, in #8e and these guidelines shown above, how much
evidence do the class data provide against the null hypothesis?
[The class data provides very strong evidence against the null hypothesis.]

(b) How closely does your evaluation of strength of evidence based on the standardized value
compare to the strength of evidence based on the p-value in #7c?

[The evaluation of strength based on the standardized value is the same for the strength of evidence based on

Now, let’s step back a bit further and think about the scope of inference. We have found that in
most classes, the observed data provide strong evidence that students do better than random guessing
which face is Tim’s and which is Bob’s. In that case, do you think that most students at your school
would agree on which face is Tim’s? Do you think this means that most people can agree on which face
belongs to Tim? Furthermore, does this mean that all people do ascribe to the same facial prototyping?

STEP 6: Look back and ahead.

10. Based on the limitations of this study, suggest a new research question that you would investigate
next.

[In my school, I feel that students would have a similar level of agreement on whose face is whose. I feel tha

Extensions
11. In #5 you recorded the proportion of students in your class who put Tim’s name with the photo on
the left. Imagine that the proportion was actually larger than that (e.g., imagine the proportion had
been 34/36 ≈ 0.944.)

(a) How would this have affected the p-value? Smaller

pg. 5
 EXPLORATION 1.3: Do People Use Facial Prototyping?

(Note: No p-value is ever exactly zero. The theoretical p-value for 30 successes in 36 trials is

actually 0.00003480, but the applet can only show a few of those decimal places. 😊)

(b) How would this have affected the absolute value of the standardized statistic? Larger

(c) How would this have affected the strength of evidence against the null hypothesis? Stronger

12. Suppose that only 14 of the 36 students in our class had put Tim’s name on the left. Observe that
then less than half of the students in the class would be putting Tim’s name on the left (in other
words, suppose p̂ = 14/36 ≈ 0.3889 < 0.5).
(a) What can you say about the standardized value of the statistic in this case? Explain. (Hint:
You know how to calculate this value.)
[The standardized value in this case would be smaller. It would be -1.338.]

(b) What can you say about the strength of evidence against the null hypothesis in this case? Will
there be any evidence against the null hypothesis? Explain.

[The value is between -1.5 and 1.5 so it indicates little to no evidence against the null hypothesis. The p-valu

pg. 6