Running head: HOMEWORK 4 1 Individual Predictors in Relation to Ability to Detect Fake Smiles
Julia Buck
University of Central Oklahoma
Author Note
Julia Buck, Department of Psychology, University of Central Oklahoma
The research is supported by a Sensation, Perception, & Action course at the University
of Central Oklahoma
Correspondence concerning this article should be address to Julia Buck, Department of
Psychology, University of Central Oklahoma, 100 North University Drive, Edmond, OK 73034.
Email: [email protected]
Abstract
Signal Detection Theory (SDT) is applicable in a psychological setting whenever there are two
stimuli that can be discriminated against. (Stanis. & Toro., 1999) The present study investigated
the application of SDT and the predictor variables that would influence a decision variable when
the provided sample completed a “Spot The Fake Smile” test. Students from a Midwestern
university completed a yes/no task in which they were to make judgments on the authenticity of
a series of 20 smiles. It was hypothesized that the type of smile, confidence and optimism
ratings, and knowledge of genuine “Duchenne Smiles” would significantly influence a
participant’s ability to detect fake smiles, as measured by d-prime (d’), as well as response bias
quantified by β. A one-way independent samples t-test was performed to determine if smile type
would have a significant effect on the sample’s d’ and β. Results suggest that type of smile has a
HOMEWORK 4 2 significant effect on a individual’s ability to avoid smile deception. Respectively, an independent
samples t-test was performed to test if knowledge of Duchenne Smiles would significantly affect
the sample’s dependent variables. Results showed no significant affect of Duchenne Smile
knowledge on smile type; implicating that knowledge of these smiles may not influence a
person’s ability to determine the authenticity their daily interactions based upon smiling
behavior. Finally, a Pearson’s Product Coefficient was run to determine if confidence and
optimism levels could be seen as accurate predictor variables in one’s ability to detect fake
smiles. Results showed no significant relationship between either of these variables towards
one’s ability to detect fake smiles, suggesting that personality traits may be of insignificant
importance when measuring a person’s decision and judgment abilities within deciphering
personal interactions.
HOMEWORK 4 3 Individual Predictors in Relation to Detection of Fake Smiles
The foundation of Signal Detection Theory (SDT) in psychology lies in the application of
discriminating between two different types of stimuli as stated by Stanis. & Toro. (1999). Their
studies state that SDT may be applied across a variety of curriculums from lie detection to
recognition memory. Conceptually, SDT may be utilized in situations were two possible stimulus
may be discriminated. (1999) According to Stanis. & Toro. (1999), discriminated variables
include signals (stimuli) versus noise (no stimuli). For our research, the area of interest involved
the application of SDT in an individuals ability to discriminate between authentic and false
smiles, and what variables would show a significant effect on an individuals ability to avoid, or
fall victim to, deception.
For our study we utilized a yes/no task form of signal detection. This involved a series of
signal and noise trials pertaining to a series of false and authentic smiles. As explained by Stanis.
& Toro. (1999), individuals base their response on a decision variable that is determined in each
trial. Their studies relate that if a decision variable were significantly high then a subject would
respond with a “yes”, wherein a signal would have been presented. The measured value that
determines the decision variable is called the criterion (Stanis. & Toro., 1999) For our research,
the predictors that would significantly determine the criterion were the focus of the study. To
measure the variables that would affect deception to fake smiles, we determined four predictor
variables: smile type (false/authentic), confidence level, optimism level, and knowledge of
Duchenne Smile’s. These predictors were to be measured by our sample’s mean d-prime (d’), the
measurement used to reflect yes/no task values. Stanis. & Toro. (199) wrote that SDT states that
d’ is unaffected by sensitivity, known as response bias, based upon the constancy of two
assumptions: (1) the signal and noise distributions are equally normal, and (2) the signal and
HOMEWORK 4 4 noise distributions have the same standard deviation. For our study we utilized β, the
measurement often used to quantify response bias to determine if these assumptions were met for
our predictor variables. For β a value of 0 relays that no response bias exists; therefore, when
subjects favor either the yes or no response, β=1 (Stanis. & Toro., 1999).
The Duchenne Smile, according to Gunnery, Hall, & Ruben (2012) can be referred to as a
genuine or, for our research, a authentic smile. As defined by their study, the Duchenne smile is
commonly characterized by activation the cheek raising muscle, the orbicularis oculi, bringing
crow’s feet to outer endpoints of the eye. (Gunnery et al., 2012). Ekman, Friesman & Davidson
(1990) suggest that a Duchenne Smile may be discriminated against other smiles based upon the
time of the smile and it’s correlation with speech and motor behavior, and not strictly from the
foundation of the muscles producing the smile. Furthermore, Gunnery et al. (2012) relate that in
the opposite, non-Duchenne Smile, the eye muscle movement is minimal. Based upon the
content of their study, they explain the distinction between these two types of smiles is important
in nature, because it may help explain the foundations of smiling in everyday situations such as
greeting a stranger, or be used in determining underlying feelings of discomfort in an individual
producing a non-Duchenne Smile. For our research, we were interested if prior knowledge of
these Duchenne Smiles could act as a significant predictor in determining our measured decision
variable; wherein the criterion is determined at the detection of false versus authentic smiles.
Research done by Ratcliff & Starns (2013) involving confidence judgments in memory
recognition suggests that confidence is a key variable is decision making. For our study we were
determined to expand on similar research and determine if confidence can be seen as significant
variable within SDT. Furthermore, we added the added our participant’s optimism level to
determine if these two personality traits can be seen as predictors of an individuals ability to
HOMEWORK 4 5 avoid deception of fake smiles. Research in this area could provide useful in determining the
individual characteristics that may inhibit a person’s ability to accurately perceive the underlying
motivations of their surrounding peers.
To measure our decision variable, participants were prompted to complete a “Spot The
Fake Smile” test provided by BBC online, in which they responded to a series of 20 trials in
which a fake or authentic smile was presented. Participants were to judge whether or not the
smile was genuine in nature. We recorded the number of hit (correct responses), misses
(incorrect responses), correct negatives (responding “genuine” when false), and false alarms
(responding “fake” when authentic) pertaining to our sample’s ability to detect fake smiles.
Using these recordings to calculate our sample’s d’, we hypothesized that the type of smile
(false/authentic) would have a significant effect on our sample’s d’. For individual assessment,
we had each participant rate their confidence and optimism level, as well as assess their
knowledge of Duchene Smiles as predictors in accordance to our criterion d’. It was
hypothesized that confidence would have a significant relationship with the sample’s d’,
optimism rating would show as a significant predictor of d’, and knowledge of Duchene Smiles
would as well show a significant relationship with the d’. To supplement our measurement of d’
second statistical tests were run for each variable wherein we measured the response bias (β) of
each predictor variable. We hypothesized that smile type would not affect our sample’s mean β.
Accordingly, it was hypothesized that optimism, confidence level, and knowledge of Duchenne
Smiles would have no significant affect on our sample’s mean β.
HOMEWORK 4 6 Methods
Participants
Sixty-nine participants completed this study. Participants were all undergraduate students within
the Psychology department of a Midwestern university. Participants were selected based upon
enrollment in a Sensation, Perception, & Action course, and received credit for their
participation. Participants were fully informed as to the nature of the study prior to the
beginning of the experiment.
Materials
To assess participant’s self-rated confidence and optimism level, scales provided by British
Broadcasting Corporation (BBC) online were utilized
(http://www.bbc.co.uk/science/humanbody/mind/surveys/smiles/). Scaling for confidence level
was sectioned into seven levels (1=low, 7=high). Scaling for optimism was also sectioned into
seven levels (1=optimistic, 7=pessimistic). To measure participants’ ability to detect fake smiles,
a computer generated, online “Spot The Fake Smile” test provided by BBC online
(http://www.bbc.co.uk/science/humanbody/mind/surveys/smiles/index_
1.shtml?gender=&age=&occupation=&country=&education=&outlook=1&confidence=5&progr
amme=) was presented on an overhead projector to the students in a classroom setting. This test
presented a series of 20 videos, lasting approximately 5 seconds each. These brief videos
depicted an individual from the shoulders up either smiling authentically or falsely. The provided
answers to these video were either “Genuine” or “Fake”. To measure our dependent variable, this
test automatically output the number of correct and incorrect answers, along with the prospective
correct answer for each smile type. To generate a d’ for each participant, participant’s found their
hit (H) and false alarm rate (F). As explained by Stanis. & Toro. (2012), H was found by diving
HOMEWORK 4 7 the number of hits by the total number of signal trials (n =10), and F was found by dividing the
number of false alarms by the number of noise trials (n = 10). In order to derive each
participant’s d’ based on these values, a Signal Detection application was used
(http://wise.cgu.edu/sdtmod/). The IBM SPSS Statistics analysis program was used to determine
our sample’s mean β and run independent samples t-tests, and Pearson’s Product Coefficient
analysis on the obtain values in order to determine if variables had an affect on our sample’s d’
and the response bias correlated to each predictor variable.
Procedures
An experimental within-subjects design was used to test our hypothesis that smile type
would affect our samples ability to detect fake smiles. In this design, each participant was
exposed to the same set of varying smile samples. For our second hypothesis regarding predictor
variables and their influence on detection of fake smiles, an experimental between-subjects
designed was used. In this instance, students were assigned to conditions based upon their own
self-assessed confidence and optimism ratings; therefore, no confederates were necessary.
Testing took place in the participants’ natural Sensation, Perception, & Action classroom
setting. Participants were tested as a group in accordance with this setting, and testing took
approximately 10 minutes before students resumed their normal class time lecture. Participants
were instructed by their professor to rate their confidence and optimism levels according to the
scales provided by BBC online. Upon completing their self-assessed rating scores, participants
observed and recorded a response in their notebooks in accordance to the 20 sample smiles
provided by the BBC online “Spot The Fake Smile Test”. Students were instructed to respond
with either “genuine” or “fake” after watching each sample smile.
HOMEWORK 4 8 Following the completion of the “Spot The Fake Smile Test”, participants were provided
with the correct responses to each sample at their professor’s discretion. Being provided the
correct responses, students were to next record their number of hits, misses, false alarms, and
correct negatives in accordance with each sample smile. Participant’s then calculated their H and
F rates. Participants were to utilize these rates to determine their d’ and β at their own discretion
outside of the classroom setting, and return with these scores during the next class session two
days later. These values were then recorded by the professor’s teaching assistant to be used as
data for analysis.
Since participants were aware as to the nature of the experiment regarding detection of
smile type, no debriefing was necessary in correspondence to this variable. After completing the
“Spot The Fake Smile” task, participants were informed as to why they were prompted to
provide their confidence and optimism ratings.
Results
A one-way independent t-test was performed to determine the significance of smile type
in relation to ability to detect fake smiles. Results suggested that there is a significant effect of
smile type on the samples d’, t(68) = 12.879, p < .00001. In this case, we reject our null
hypothesis and state that smile type does have a significant effect on ability to detect fake smiles.
An independent samples t-test was run to show if knowledge Duchenne Smiles could
significantly predict d’, t(67) = -.726, p = .470, and did not show a significant relationship
between the predictor variable and d’. A Pearson Product Coefficient was run to predict the
relationship between our participant’s self-assessed confidence rating and d’. Results predicted
no significant relationship between confidence and d’, r = .109, p > .05. A second Pearson
Product Coefficient was run to predict the relationship between our participant’s self-assessed
HOMEWORK 4 9 optimism level and d’. Results showed no significant relationship between optimism and d’, r =
.023, p > .05.
A one-sample t-test was performed to measure smile type’s (false/authentic) effect on our
sample’s mean β, t(68) = 17.40, p < .00001, and showed a significant affect on smile type to
detect deception. Our sample mean was significantly different from β = 1. In this case we reject
our null hypothesis and state that smile type does have a significant effect on mean β.
To test the predictor variables (confidence level, optimism, knowledge of Duchenne
Smile), an independent samples t-test was run, t(66) = -.69, p = .67, and showed no significant
affect of predictor variables on ability to detect deception. Therefore we retain our null
hypothesis that optimism, confidence level, and knowledge of Duchenne Smiles effect mean β. A
Pearson Product Coefficient was run to show any relation between predictors and our sample’s
mean β. Optimism, confidence level, and knowledge of Duchenne Smile showed no significant
relationships to mean β; although, there was a significant relationship between knowledge of
Duchenne Smiles and confidence level, β ≠ 1. All other means were significant different from β
= 1.
Discussion
The focus of this study was to determine the variables that could be used to significantly
predict the detection of fake smiles. Conceptually, it was hypothesized that the type of smile
(false/authentic) would have a significant affect on our sample’s calculated mean d’.
Furthermore individual variables across participants such as confidence rating, optimism level,
and knowledge of Duchene Smile were considered in relationship to detection of fake smiles.
Firstly, based upon these variables, it was hypothesized that confidence rating would have a
significant effect on d’. Second, optimism level was hypothesized to have a significant effect on
HOMEWORK 4 10 d’. Then at last, that knowledge of Duchenne Smiles would be a significant predictor variable in
relation to our sample’s d’. In relation to response bias, it was hypothesized that smile type
would have a significant effect on our sample’s mean β; furthermore, it was hypothesized that
our predictor variables would have a significant effect on our sample’s response bias.
There was significant output from the independent t-test to support our initial hypothesis
that smile type has a significant effect on ability to detect fake smiles. Although, the output from
the independent samples t-test and Pearson’s Product Coefficient did not show a significant
effect on our participants individual conditions of confidence, optimism, and knowledge of
Duchenne smiles. Implications of these results suggest that the ability to detect fake smiles is
influenced strictly by the stimulus, and personal characteristics and variables have no influence.
Therefore, an individual may not be able to rely on their own personal characteristics in relation
to detecting deception in their everyday interactions, and may need to focus primarily on the
signals provided from the stimulus provider. For future research to support these implications,
studies expanding on inter/intra-personal characteristics and deception would need to be
researched.
Limitations to this study were seen as participants were active Psychology students;
possibly having preconceived notions of deception and knowledge of Duchenne Smiles.
Furthermore, response bias may be seen in that those with prior knowledge of Duchenne Smiles
may naturally choose a higher confidence rating, therefore affecting the correlation between
these variables.
HOMEWORK 4 11 References
Ekman, P., Davidson, R., & Friesen, W. (1990). The duchenne smile:
emotional expression and brain physiology II. Journal of Personality
and Social Psychology, 58(2), 342-353.
Gunnery, S., Hall, J., & Ruben, M. (2012). The deliberate duchenne smile: individual
differences in expressive control. Journal of Nonverbal Behavior, 29-41. doi:
10.1007/s10919-012-0139-4
Ratcliff, R., & Starns, J. (2013). Modeling confidence judgments, response times, and
multiple choices in decision making: Recognition memory and motion
discrimination. Psychological Review, 120(3), 697-719. doi: 10.1037/a0033152
Stanislaw, H., & Todorov, N. (1999). Calculation of signal detection theory measures.
Behavior Research Methods, Instruments, & Computers, 31(1), 137-149.