Subliminal messaging – we’ve all heard about it. But it doesn’t really work, right?
New research from Valentin Dragoi’s lab at the University of Texas at Houston suggests that subliminal images can change our brain activity and behavior.
What does subliminal messaging entail? Subliminal messages are words or images presented below our conscious awareness. Usually we think of short frames cut into a video feed, where the subliminal message appears so quickly (usually less than one tenth of a second!) that our minds do not register their appearance. On the other hand, supraliminal messages are presented for longer periods of time, such that we can consciously see them.
One of the most famous examples of subliminal messaging was conducted in the 1950’s. To test if subliminal messages could sway behavior, short messages stating, “Drink Coca-Cola” and “Hungry? Eat Popcorn” were played in a movie theater during a film. Although the results were later proved fraudulent, the researcher, James Vicary, claimed that presenting these suggestive messages increased concession sales.
So, does subliminal messaging actually affect us?
Importantly, and thankfully, subliminal messaging is not capable of brainwashing. However, there is evidence dating all the way back all the way to the 1960’s, which suggests that showing subliminal images improves behavioral performance.
Though the behavioral change has been established through multiple publications, we still don’t know how the brain processes subliminal images.
Sorin Pojoga, a member of Valentin Dragoi’s lab, set out to determine the neural underpinnings of subliminal images and how this brain activity changes our behavior. In a recent study, Pojoga and colleagues designed a task which would subliminally expose rhesus macaques to a set of natural images while recording from neurons in the primary visual cortex.
The natural images (such as a photograph of an animal) were imbedded in an oriented grating for either two or five consecutive frames, 33.3 and 83.3 ms respectively. The authors found that subjects could easily identify the image when it was imbedded for five frames. However, subjects only guessed the correct image at chance level when the images were presented for two frames, meaning the images were below the subjects’ threshold for detection, making them subliminal.
The first clue
The authors had found that they could present images below the level of conscious detection. But the question remains if and how the brain registers these images. The authors addressed this in a follow-up experiment. Subjects performed tasks to test whether brain activity or behavioral performance changed due to subliminal image presentation.
Again, animals fixated on a central point on a computer screen while an oriented grating appeared. At a random point in each trial, a novel natural image was presented for two consecutive frames instead of the oriented grating. The inserted natural image was presented in its original form or rotated 5-20 degrees. Researchers could then align the trials at the time the subliminal natural image was present and analyze the data for any changes in neural activity.
Pojoga used a linear discriminant analysis (LDA) to decode neural responses to the natural images. LDA is a supervised learning model that attempts to classify data into different categories. In this case, researchers tested whether the LDA could separate trials based on which orientation the natural image was presented at. For 90% of the trials, they input firing rates from when the natural image was displayed and “trained” the model by pairing firing rates with the correct image orientation. Once the model was trained, researchers could then input the firing rates from the remaining 10% of trials and test the model, by letting the model categorize each trial by image orientation.
They found that LDA performance was significantly higher than chance level, meaning the neurons encoded the orientation of the natural image. Researchers also computed the LDA for a portion of the trial when the subliminal image was not being presented and found that LDA performance was not different from chance levels. This analysis further supports the conclusion that neurons were indeed processing the natural images.
The underlying circuitry
With evidence that neurons are in some way processing the embedded natural images, the group reasoned that processing subliminal images should serve a purpose. They performed a follow-up experiment to determine if natural images, which were previously presented subliminally, would later improve neural processing during supraliminal presentation.
Subjects performed an orientation discrimination task. While animals fixated on a central point of a computer screen, a natural image was shown in the periphery. After a brief delay, the image was shown again. Subjects used a response lever had to signal whether the two images were shown at the same orientation, or if the second was shown at an angle (when the images were different, animals were trained to release the lever, but when the images were the same, subjects were expected to continue holding the lever). Importantly, 50% of the natural images used in this task were novel, but the other half were shown subliminally as natural images in the previous fixation task (described above). The natural images which were previously shown subliminally and now presented supraliminally in the discrimination task are referred to as “exposed” stimuli.
The authors found that neurons extract more information about exposed stimuli compared to unexposed stimuli, measured by a mutual information analysis. Additionally, they used d’ analysis and found that neurons are more sensitive to exposed stimuli.
Similar to the LDA performed during the two-frame presentation of the natural image, researchers found that LDA performance during the orientation discrimination task was significantly above chance level. Consistent with their hypothesis, the group found that LDA performance was significantly higher for exposed natural images compared to unexposed stimuli.
Together, these results suggest that subliminal priming (previously showing images subliminally) allows for improved image processing when those stimuli are later made supraliminal.
It takes two to tango (or converse)
Through multiple analyses, it’s clear that single neurons show improved image processing for exposed images compared to unexposed images. But how does that happen if we’re not consciously aware of seeing the images in the first place? The authors argue that repeated viewing of subliminal stimuli activates groups of cells at the same time. As the old adage says, cells that fire together, wire together. So by activating groups of neurons simultaneously, subliminally stimulation could increase communication between neurons.
Pojoga and colleagues measured connectivity between cells by computing cross correlograms. This analysis measures the timing of spikes between two cells. The higher the number of coincident spikes, the stronger the coupling between those two neurons. The group hypothesized that higher coincident activity would occur for exposed images, signifying an increase in signaling between cells.
The group found that while both exposed and unexposed natural images produce strong cross correlograms, cross correlograms were significantly higher for exposed images compared to unexposed images. This suggests that repeated subliminal stimulation of natural images improves communication between groups of cells, making them better able to signal the image if it is shown supraliminally.
With evidence for single neuron changes and support for a network-level mechanism, does any of this actually affect behavior?
Researchers used the same discrimination task described above, and simply computed the percentage of correct responses across all orientations for exposed compared to unexposed images. They found that animals performed significantly better on trials with exposed images. Additionally, this improvement for exposed stimuli tracked across orientations as well. For low-orientation-change trials, which are most difficult for the animal to get correct, performance was better for exposed compared to unexposed images. This suggests that single neuron and network-level changes in stimulus encoding are strong enough to alter perception.
The authors found other evidence supporting behavioral changes as well. Animals took a shorter amount of time to make a response for exposed stimuli, suggesting a higher confidence in their decisions. Additionally, they found that behavioral performance was highly correlated with increased coincident spikes (measured by cross correlograms), suggesting that when neurons communicate better among themselves, improved behavior follows.
Am I being brainwashed?
The short answer here – no!
As shown in this paper, subliminal imagery can significantly alter neural activity and behavior. But that doesn’t mean that if a car company flashes a subliminal message saying, “Buy this car,” that you’re instantly going to drive to a dealership and make an extravagant purchase.
Some experts suggest that subliminal messaging must be “goal-relevant” to a person. Showing a subliminal message saying “Drink Coca-Cola” won’t make you thirsty. But if you’re already thirsty and you see the same subliminal message, and you’re already thirsty, you’re more likely to buy the suggested brand. This influence is probably why subliminal messaging in advertising is banned in many countries.
Subliminal messages live in our everyday lives, where we may never notice them. While behavioral studies over the years have suggested subliminal messaging can modulate our choices, this new study shows how these behavioral changes occur on a single neuron and network level.
Now you know – subliminal messaging is not a myth! Whilst it’s not going to brainwash you, maybe the next time you pick up your favorite drink, take a moment to ask yourself, “Why is this my favorite brand?”