From movies to real life, polygraph machines, otherwise known as lie detectors, have been used as a means to seek the truth since 1921. However, the legitimacy of these machines has been long questioned in the scientific community. The purpose of this experiment is to determine the accuracy of polygraph machines by building a model machine to monitor physiological responses to lying. Specifically, the machine monitors skin conductivity using a breadboard, a few wires, three LEDs, and an Arduino nano chip connecting the breadboard to a computer. The Arduino software enabled a computer to graph the levels of the test subject’ sweat on a serial monitor. Heart rate was also measured using a monitor that clips onto the test subjects’ middle fingers. The test subjects were then asked a series of questions that would force them to lie. Their skin conductivity and heart rate were recorded and analyzed to determine when the test subjects lied. Five different age groups were tested ranging from children ages 8 and under to seniors over 55 years old; two males and two females were present in each age group. Out of the twenty tests that were conducted, the interrogator was able to correctly guess when the subjects lied in 6 tests, giving the polygraph machine an accuracy rate of 30%, which was much lower than predicted. Additionally, the polygraph machine had 0% accuracy when measuring participants from the 13 to 17-year-old age bracket.
The name “lie detector” can be very deceptive, as this can lead people to believe that the machine has the ability to peer into people’s thoughts and determine precisely when they lie. What the polygraph machine is actually capable of, however, is monitoring three measures of physiological response: heart rate, blood pressure, and galvanic skin conductivity. Galvanic skin conductivity is the change in the electrical resistance of the skin caused by emotional stress. Our anxiety is directly proportional to the activity of the eccrine sweat glands, so the increased nervousness felt by a person will result in them sweating more, making their skin momentarily more conductive to electricity. This increase in conductivity is measured by attaching electrodes to a person’s fingertips. The fluctuations in the amount of electricity that is being conducted through the fingers will be graphed on a polygraph. The activity in the eccrine sweat glands are not under conscious control, and so are a direct reflection of one’s emotional state. Insight into how the subjects are feeling can therefore be obtained by measuring human reaction corresponding to psychological change, such as the galvanic skin response.
The most common method of interrogation under the polygraph machine would be the CQT, or the Control Questions Test. Under this interrogation method, the interrogator will ask the subject a series of questions that would and wouldn’t be related to the reason of questioning. For example, if a person is being accused murder, the CQT may involve questions such as “Do you have any siblings?” or, “Have you ever been in a relationship?” These are control questions which serve as a baseline reference for the questions that actually have relevance, such as “Were you at home during the night of the murder?” Theoretically, with these questions, the guilty subject should show an increased amount of stimulation shown through skin conductivity, heart rate, or blood pressure when compared to the irrelevant questions. The interrogator will be looking for sudden surges in the graphs plotted by the machine to determine if the subject is lying.
There are other confounding factors that affect polygraph machines aside from the aforementioned physiological variables, including the neurological characteristics of the individual test subjects. Some people lie out of habit and do it without experiencing any discomfort at all; these people are called pathological liars. Trained liars are similar, although their ease of dishonesty comes from consistent training and habit forming. According to a study done, prefrontal cortices in the brains of pathological liars are composed of up to 26% more white matter than the average person, making them, “more likely to make connections between different memories and ideas”, which can enable them to lie more consistently with ease. The exact opposite is true for the average person, who might respond more dramatically to psychological stimuli under the pressure of interrogation. As such, this person might become a false positive due to their anxiety, regardless of whether they were telling the truth, whereas the more consistent liars pass without breaking a sweat.
The accuracy of polygraph machines have been a controversial question due to the complicated science behind it. However, an even bigger problem is how to test the accuracy of lie detectors; it is near impossible to simulate conditions for testing guilty liars in real-life situations. Differences in the environment for the test subjects will affect the results of the experiment, and controlling every aspect of the test environment is the ultimate flaw in every study trying to discern the accuracy of lie detectors. The current consensus for polygraph accuracy is that it is correct in around 65% of cases. Thus, the null hypothesis of this experiment is that the hand-built polygraph machine would also enable the interrogator to accurately determine the lies in about 65% of the tests conducted.
Materials and Procedure
Using the diagram of the board circuitry (Figure 1), the 220 Ohm resistor was soldered to ground (labeled on GND on the Arduino Nano microcontroller). Including a resistor is important, as their purpose is to regulate the amount of current that flows through the circuit. This will keep the LEDs from being destroyed and limiting the amount of current flowing through the test subjects fingers.
Figure 1: Board Circuitry of the Polygraph Machine
The green, yellow, and red LEDs were placed on the board and connected together by soldering another wire to ground (GND) and the board. A cable wire was soldered to analog (labeled A0 on microcontroller) and was extended so that the resistor was connected to it. Another cable wire was soldered to the 5V pin. All the LED ground pins were connected. A cable wire was soldered onto D2 on the board and connected to the positive leg of the green LED. A second
cable wire was soldered to D3 on the board and connected to the positive leg of the yellow LED. Lastly, a cable wire was soldered to D4 and connected to the positive leg of the red LED. The pins on the circuit that needed wires soldered to them were highlighted in yellow for guidance (figure 2).
Figure 2: Highlighted Pins of where the wires need to be soldered
One end of the Arduino USB cable was inserted into the microcontroller and the other end was inserted into the USB port of a laptop. Code for the serial monitor output was written using the Arduino 1.8.9 program, which allowed users to visualise the polygraph. (see figures 3 and 4). The polygraph was plotted in real time, so it showed the exact time at which the test subjects’ sweat levels were rising or falling.
Figure 3: Arduino Code
Here, the code identifies the three different LEDs and turns all three of them on and off at the start of the program. ‘HIGH’ in the code means on for the LEDs, and ‘LOW’ means off.
Figure 4: Arduino Code Cont.
The code enabled different LEDs to turn on depending on the input value; how much the test subject was sweating. If the input value was greater than 10, then the green LED connected to pin 2 would turn on. If the input value was greater than 35, then the yellow LED connected to pin 3 would turn on. If the input value was greater than 55, then the red LED connected to pin 4 would turn on.
The experimental procedure was as follows: the serial monitor on the Arduino was opened. Five age groups were determined, with the first group consisting of children under 8, the second group 9 to 12 year olds, the third group 13 to 17 years olds, the fourth group 18 to 54 year olds, and the fifth consisting of seniors aged 55 and above. Each age group contained two males and two females. The test subjects were interrogated individually, with each test being recorded by a screen monitor camera. A heart rate monitor was clipped on to the middle finger of the test subject, while the pointer and thumb fingers of each hand pinched the extended wires from the board, as seen in figure 5. This pinching is necessary to ensure a complete circuit and to allow sweat levels to be measured.
Figure 5: Diagram of polygraph machine set up
The test subject was asked to pick a number between 1 and 8 and refrain from telling the interrogator the number they had chosen. The subject was then asked a series of yes or no questions relating to their number: “Was 1 the number chosen? Was 2 the number chosen? Was 3 the number chosen?” and so on. These questions were asked for each number from 1 to 8, and the test subject was asked to answer with “no” for each question, forcing them to lie one out of eight times. After the interrogation ended, the graphs plotted on the serial monitor on the computer were analyzed, and the heart rate of the test subject was displayed on the monitor. Sudden elevations and shifts in the graph were specifically looked for, to determine when the participants lied and which number was chosen. The differences between baseline plots and shifts when the subject was definitively lying are shown in figures 6 and 7 respectively.
Figure 6: Normal Shifts on Serial Monitor
Figure 7: Abnormal Shifts that Indicate Lying
The interrogator would then guess as to which number the subject chose, depending on the point at which they lied. The subject would inform the interrogator if they were correct or not, and this would be recorded on a chart. This process was repeated for all 20 test subjects.
Results and Discussion
In order to measure the accuracy of the polygraph machine, several dependent variables were measured: the participants’ heart rate, galvanic response, and blood pressure. To ensure that no other confounding variables could affect the data, all test subjects were asked the same question. The independent variable would be the number of times the interrogator guessed the instance of lying correctly from the polygraph and heart rate data. The magnitude of skin conductivity of the test subjects was recorded on polygraphs as the y-axis data, and the x-axis is time.
The guessed number was recorded in a chart along with the test subject’s chosen number. The times in which the interrogator was correct is highlighted in green in the table below. Out of the 20 trials that were conducted, the polygraph machine was accurate in 6. The calculated percentage of accuracy was therefore 30%, 35% lower than the expected percentage of accuracy.
Figure 8: Table showing number chosen by the subject and number guessed by the interrogator
The polygraphs plotted by the machine while interrogating children aged 8 and under had more distinct elevations in skin conductivity and heart rate than that of other age groups (see figure 8). Determining when and at which number they might have lied was therefore much easier. This is probably due to the psychology of young children, as their brains lack the cognitive ability to lie easily without showing symptoms of guilt: anxiety, sweating, and increased heart rate. These reactions are easily picked up by the polygraph machine, leading to the correct number being guessed.
Figure 8: Polygraph results of a test subject from the 8 and under age category
It became harder to determine when the test subject was lying in the three older groups. The polygraph was equally accurate with both the 9 – 12 and over 55 age groups. Both test groups had polygraphs that would spike in random spots that were not when they lied. These led the interrogator to incorrectly guess when they lied and what their number was. The main difference between the graphs of these two groups is that the serial input was much lower for the seniors, meaning that they sweat less than the 9 – 12 age group (figure 9).
Figure 9: Test Subject 1 Age 55 and Over Polygraph
The age group of 13 – 17 year olds were the hardest to determine when they were lying. The polygraphs from this group had slopes that remained relatively constant throughout the interrogation, making it difficult to pinpoint the time when they lied (figure 10). This could be because the brains of teenagers are more conditioned to lie without feeling guilt.
Figure 10: Test Subject 4 Age 13 – 17 Polygraph on Computer
For all test subjects, their heart rate steadily increased as the interrogation continued, so it was not as helpful in telling when the test subject lied. The reason that sudden elevations could not be observed could be due to the lack of sensitivity with the heart rate monitor, as it was bought at a local pharmacy instead of those used in hospitals with higher precision and accuracy. The data from the experiment was displayed on a bar graph showing % of accuracy for each age group.
Figure 11: Accuracy of Polygraph Machine
These results are evidence supporting the inaccuracy of polygraph machines. The usage of these machines should be discontinued, since the ability for the machine to correctly predict lies depends solely on how much a person responds to guilt and anxiety, which varies from person to person. Despite this, the polygraph machine is widely employed by the police, organizations such as the FBI and CIA, and employers who are looking to weed out unsuitable employees. It is administered 2.5 million times each year in the US, and if the results of this experiment were to be applied to this statistic, then the machine would be incorrect in 1.75 million of those cases. The consequences of this inaccuracy are too great to allow the polygraph test to remain as one of the hallmarks of global security.
The results from this experiment may not be concrete as there was a small sample size and the blood pressure of the test subjects was not monitored, unlike with a real polygraph machine, because the equipment needed for this was not accessible. Like with any other study testing the accuracy of these machines, the correct environment needed to make the data reproducible could not be perfectly recreated. To improve on this experiment, a larger sample size of test subjects would be used to make results more reliable, avoid possible anomalies, and allow for greater analysis. Other research has recently been conducted by the University of Pennsylvania, which compares the ability of detecting lies using polygraph machines and brain scans. A future experiment could explore this in more depth, to see if brain scans are more reliable than the machines. The experiment could also be developed by using iMotions, an emotion recognition software which combines data like heart rate, blood pressure, skin conductivity, and eye movement into one, neat report. This would be very helpful in determining how the test subject is feeling and when they might have lied by analyzing their behavior in greater detail.
The polygraph machine was made in the hopes that it could detect lies, but is potentially nothing more than a data plotter which serves as an index of the subject’s emotional state. In simple terms, it measures the magnitude of anxiety felt by the test subject under interrogation conditions, but provides no insight into true internal thought. After making a basic version of a polygraph machine, it was proven that the machine had an accuracy rate of 30%, far too low for the machine to be considered legitimate. People of all age groups responded differently to being interrogated with different reactions in galvanic skin conductivity or heart rate, causing the accuracy of the machine would fluctuate. Despite our current technological development, feat of knowing when someone is lying remains nearly impossible to determine in a systematic way.
The experimenter would like to thank the 20 people who gave their time to be participants in this experiment.
1. Volyk, A. (2019). History of the Polygraph: Lie Detector, Polygraph Examiner, Polygraph Testing, Lie Detection. [online] Argo-a.com.ua. Available at: http://www.argo-a.com.ua/eng/history.html [Accessed 17 Jun. 2019].
2. The Truth About Lie Detectors (aka Polygraph Tests). American Psychological Association. August 5, 2004. Accessed June 23, 2019. https://www.apa.org/research/action/polygraph.
3. Farnsworth, Bryn, Ph.D. What Is GSR (galvanic Skin Response) and How Does It Work? Imotions. July 17, 2018. Accessed June 23, 2019. https://imotions.com/blog/gsr.
4. What is the skin conductance response ? News & Updates – MIT Media Lab. 1993. Accessed June 23, 2019. https://www.media.mit.edu/galvactivator/faq.html.
5. Legg, T, Ph.D. Pathological Liar: How to Cope with Someone’s Compulsive Lies. Healthline. August 27, 2018. Accessed June 23, 2019. https://www.healthline.com/health/pathological-liar.
6. Knapp, Alex. Do Liars Really Have Different Brains? Forbes. June 12, 2012. Accessed June 23, 2019. https://www.forbes.com/sites/alexknapp/2012/06/11/do-liars-have-really-different-brains/#7599113a632c.
7. Barber, Nigel, Ph.D. Do Lie Detectors Work? Psychology Today. March 7, 2013. Accessed June 23, 2019. https://www.psychologytoday.com/us/blog/the-human-beast/201303/do-lie-detectors-work.
8. Vitelli, Romeo, Ph.D. When Does Lying Begin? Psychology Today. November 11, 2013. Accessed June 23, 2019. https://www.psychologytoday.com/us/blog/media-spotlight/201311/when-does-lying-begin.
9. Harris, Mark. The Lie Generator: Inside the Black Mirror World of Polygraph Job Screenings. Wired. October 30, 2018. Accessed June 23, 2019. https://www.wired.com/story/inside-polygraph-job-screening-black-mirror/.
10. Davis, Heather A. Brain Scans Detect Lies More Accurately than the Polygraph. Penn Today. December 22, 2016. Accessed June 23, 2019. https://penntoday.upenn.edu/research/brain-scans-detect-lies-more-accurately-than-the-polygraph.
11. Image of Board Circuitry: Electronics for Everyone, “ Arduino Lie Detector”, 2016. https://www.instructables.com/id/Arduino-Lie-Detector/
All images by author, unless otherwise specified.
ABOUT THE AUTHOR
Gopiga Dass is a rising sophomore at Lower Moreland High School in Huntingdon Valley, Pennsylvania. She has a passion for math and biology, and enjoys taking part in science competitions, including the Governor’s STEM competition and the Pennsylvania Junior Academy of Science. Her hobbies include playing the flute for her school’s wind ensemble and riding her bike with her sister.