BiologyNeuroscience

How Reliable is fMRI as a Neuroimaging Tool to Study Visual Experiences in Dreaming?

Abstract

The nature of dream is private. It is impossible to render the dreaming experience public until scientists discovered the association between dreaming and activation of the brain, followed by the invention of neuroimaging tools, such as functional magnetic resonance imaging (fMRI) and electroencephalogram (EEG), to study brain activities during dreaming. Hence, the construction of knowledge about dreaming heavily relies on objective data produced by neuroimaging tools.

However, these neuroimaging tools may not be as reliable as scientists thought, and objective data produced by tools may not be able to represent the subjective and private experience. Therefore, the basis on which the whole kingdom of knowledge of dreaming built is shattered.

This paper discusses the epistemic problems arise from using fMRI as neuroimaging tool to study of visual elements of dreams. It may raise awareness of the scientists about the limitation of objective tools in studying subjective experience, thus encouraging scientists to take a more sceptical view towards knowledge of dreaming.

1 Introduction

Dreaming, a subjective experience during sleep, is so prevalent yet unfathomable. Since Ancient Greece and Ancient Rome, humans have been intrigued by the mystery of dream. Scientific studies have shown that everyone dream, even those claim they do not (Herlin et al., 2015). However, objective analysis of dream is hindered by its private nature. The academic knowledge of dreaming amounted to a public accumulation of private experiences (Kroker, 2007).

The initial study of dream concentrated on the interpretation of dreams, and later, the focus shifted to the reasons behind dreaming largely due to the influence of Sigmund Freud’s theories. In modern days, scientists attempt to utilise objective tools such as electroencephalography (EEG), electrooculogram (EOG), electromyogram (EMG) and functional magnetic resonance imaging (fMRI), facilitated by machine-learning-based analysis, to decode visual content of dream, and hence break the private nature of dream.

However, this neural decoding process is highly complicated, and epistemic problems could arise at various stages of knowledge construction. Therefore, the reliability of the neuroimaging tools in studying dreams is called into question, and the basis upon which the kingdom of knowledge of dreaming is built would be shattered. My essay arguing against the epistemic reliability of fMRI, a typical neuroimaging tool, in the study of visual elements of dreams would commence with an introduction to the process of neural decoding by using fMRI. Then, I would address the reliability and limitations of fMRI itself before moving on to explain why such an objective method could not reflect subjective experience comprehensively and accurately. Furthermore, I would explore the epistemic challenges pertaining to data verification in the abovementioned process. To conclude my essay, I would use a diagram to summarise the central problems associated with three major stages of knowledge construction in the study of visual elements of dreams, namely-collection, interpretation and verification of data.

2 Neural decoding of visual imagery during sleep

Before proceeding to cast doubt on the reliability of fMRI, I shall first detail the process of neural decoding of visual imagery during dreaming. Firstly, scientists employed fMRI to track brain activity of participants during sleep stage 1 or 2. Then, participants were awakened on an average of every 342.0 seconds and were asked to give a verbal report of visual experience before awakening. Following that, scientists analysed verbal reports using a lexical database to create systematic labels for visual contents. The hypothesis behind this experiment is that contents of visual imagery during sleep are represented at least partly by visual cortical activity patterns shared by stimulus representation, thus the fMRI data obtained before each awakening were labelled with a visual content vector and then sent for decoder training (Horikawa et al.,2013). After collecting huge amount of fMRI data and verbal reports, a database would be established in the decoder, which correlates fMRI patterns to elements of visual imagery experience by dreamers during dreaming. Subsequently, by just obtaining the fMRI images of a dreamer during sleeping, scientists would know the visual contents of his or her dream. In such way, breaking the private nature of dreams is rendered possible because the subjective visual experience during dreaming is represented by the objective fMRI pattern.

3 Reliability of fMRI

3.1 Reliability of statistical methods of fMRI

Since its invention more than 20 years ago, functional magnetic resonance imaging (fMRI) has become a popular tool for understanding the human brain, with some 40,000 published papers according to PubMed. Despite the popularity of fMRI as a tool for studying brain function, the statistical methods it uses have rarely been validated using real data (Eklund et al., 2016). The validation of work has been done by a group of scientists led by Anders Eklund from Linköping University in Sweden. They reported that the three most popular fMRI software packages for fMRI analysis – SPM, FSL, and AFNI- suffer from spatial autocorrelation, a basic flaw, that the fMRI signal tends to be similar (correlated) across nearby regions, which leads to elevated false-positive rates, from expected 5 percent up to 70 percent, when it comes to finding activations associated with tasks or stimuli i.e. finding which brain area ‘lights up’ during particular tasks (Neuroskeptic,2016). The study has subversive effect on the confidence that scientists have in fMRI. Given such a high false positive rate, previous studies on dreams based on fMRI analysis worth rigorous local scepticism.

3.2 Limitation of fMRI in studying brain activity

fMRI reveals neural activity indirectly by measuring the BOLD (Blood Oxygenation Level Dependent) signal. This method is based on the fact that haemoglobin carrying an oxyhaemoglobin in the bloodstream emits a different MR signal than deoxyhaemoglobin.  When parts of the brain become active, for example, when a person is carrying out a cognitive task, the active parts of the brain use up more oxygen than relatively inactive parts (Nguyen, 2009). Moreover, the fMRI signal cannot easily differentiate bottom-up signals from top-down signals, and it may potentially confuse excitation and inhibition. The magnitude of the fMRI signal cannot be quantified to reflect accurately differences between brain regions, or between tasks within the same region (Logothetis, 2008). Some scientists may argue that fMRI is currently the best tool to study brain activity during dreaming, for it is non-invasive, relatively inexpensive, and when complemented with EEG, its accuracy can be further improved. Still, the standard required for establishing direct link between neural activity and pattern produced by neuroimaging tools should not be compromised due to current technological constraints. Therefore, fMRI, one of the most commonly used neuroimaging tools, may not be as reliable in tracking brain activities as scientists assume.

4 Reliability of using an objective tool to study subjective experience

Dream, by definition, is a hallucinatory experience during sleep. All forms of experience exhibit subjective character because each organism has a unique point of view from which no other organism can gather experience (Nagel, 1974). So as a type of experience, dream entails subjective character with no exception. Despite various attempts by scientists to reduce subjective experience during dreaming to a set of signals or neuroimages, some philosophers of science are sceptical to the reliability of measuring subjective experience by employing objective tools. In my opinion, while real and often describable, subjective experience cannot be objectively measured.

First of all, what subjective experience entails is more than the object perceived by senses. While the perceived object, be it tangible or abstract, could be measured by objective scientific instrument, the emotions that it provokes and its meaning to the person still remain undetectable. For example, when someone listens to a piece of music, the music’s note, pitch, speed, volume and the listener’s ear vibration and heartbeat can be measured by scientific instruments, but the listener’s aesthetic experience cannot (Cycleback, 2014). Similarly, when a dreamer dreams of a river, the object, the river, may be generated as the result of neural decoding, but the emotional aspects of the dreamer when he or she saw the river during dreaming is not measurable. Perhaps the river evoked nostalgia, or reminded the dreamer of his or her hometown where he or she spent the carefree childhood. The above examples illustrate that there is always an important fraction of subjective experience that preserve its private and unfathomable nature. Thus, objective tools can never examine subjective experience thoroughly.

In addition, it is naïve to just assume that the same results obtained from neural decoding equates to identical visual experience during dreaming. Given two identical fMRI images from 2 different participants (which is highly impossible in reality but let us just assume), it is probable that the participants saw different objects in dreams. This is because scientists have only discovered that certain fMRI pattern could correlate to a specific object, but the correlation may not be unidirectional. Let the fMRI element be “p” and the visual content experienced by dreamers be “v”, we have:

pcorrelates to v1;

pcorrelates to v2, etc.

After several rounds of decoder training machine learning, when the input is p1, the software will generate an output v1. However, it is still possible that an unknown element “p3” is also correlated to v1; and p2 correlates to not only v2 but also “v4”. The correlation between fMRI results and visual imagery during dreaming is neither linear nor unidirectional; from one fMRI image, more than one results may be possible, and some elements are not even included in the database. Hence, fMRI as an objective tool cannot consistently produce accurate results.

5 Epistemic problems associated with data verification

5.1 Dreamers as the final authority

The only possible way to evaluate the accuracy and trustworthiness of the decoding results from fMRI pattern is to verify them with the dreamers. However, it is impossible to further assess the reliability of the verbal reports given by the dreamers. Dreamer is the owner and final authority on his or her own dreams (Ullman and Zimmerman, 1979). Theoretically, the dreamer could lie to the researchers and misguide them to false correlations between fMRI pattern and visual content of dreams. It could be argued, however, that even if few participants intentionally lie about the actual visual content they experienced during dreaming, the effect would be masked by enormous amount of data gathered from people who are honest about their dreams. Nonetheless, the verbal reports may still be inaccurate due to the problem of dream recall, and this problem is applicable to every participant.

5.2 The problem of dream recall

The problem of dream recall is mainly associated with memory. At least 95% of the dream are not remembered because certain brain chemicals necessary for converting short-term memories into long-term ones are suppressed during REM sleep (Allan and Robert, 1997). Hence, even the participants were awakened every 342.0s on average, much dream content had been lost during this short period. Moreover, the memory of the dreamers is always prone to confabulation, a memory disturbance. As the result, the dreamers may describe visual content that they never “saw” in the verbal report, albeit without conscious intention to deceive. Therefore, the dreamers, the final authority, cannot provide rigorous enough verification to the decoding results; this give rise to an epistemic problem because scientists have no other way to substantiate the accuracy of the decoding results but to fully rely on dreamers’ inaccurate verbal report.

Conclusion

In conclusion, fMRI is not a reliable tool to study visual elements experienced by dreamers during dreaming. At least, its reliability is more fragile as scientists usually assume. First and foremost, just as the recently published impactful research led by Anders Eklund suggests, the reliability of statistical methods of fMRI is questionable. Moreover, whether it is possible to utilise objective tools to study subjective experience is still debatable. Within my scope of argument, it can be seen that some components of subjective experience cannot be measured quantitatively. Besides, scientists do not have a watertight way to verify the results obtained from fMRI data input. The one and only authority, the dreamers, may lose part of memory about the dream content and even they could remember a small fraction of their dreams, the memory is easily distorted or even faked. As the diagram illustrates, the three epistemic problems undermine three stages of knowledge construction regarding the use of fMRI respectively, and hence challenged the value and the usefulness of knowledge about neural decoding of visual elements during dreaming.

6 Bibliography

  1. Herlin, Bastien, Smaranda Leu‐Semenescu, Charlotte Chaumereuil, and Isabelle Arnulf. “Evidence that non‐dreamers do dream: a REM sleep behaviour disorder model.” Journal of sleep research 24, no. 6 (2015): 602-609.
  2. Kroker, Kenton. The sleep of others and the transformations of sleep research. University of Toronto Press, 2007.
  3. Horikawa, Tomoyasu, Masako Tamaki, Yoichi Miyawaki, and Yukiyasu Kamitani. “Neural decoding of visual imagery during sleep.” Science 340, no. 6132 (2013): 639-642.
  4. Eklund, Anders, Thomas E. Nichols, and Hans Knutsson. “Cluster failure: why fMRI inferences for spatial extent have inflated false-positive rates.” Proceedings of the National Academy of Sciences (2016): 201602413.
  5. Neuroskeptic. “False-Positive fMRI Hits The Mainstream.” Neuroskeptic. July 15, 2016. Accessed January 11, 2017. http://blogs.discovermagazine.com/neuroskeptic/2016/07/07/false-positive-fmri-mainstream/.
  6. Nguyen, J. “Neuroimaging.” HOPES. April 09, 2012. http://web.stanford.edu/group/hopes/cgi-bin/hopes_test/neuroimaging/.
  7. Logothetis, Nikos K. “What we can do and what we cannot do with fMRI.” Nature 453, no. 7197 (2008): 869-878.
  8. Nagel, Thomas. “What is it like to be a bat?.” The philosophical review 83, no. 4 (1974): 435-450.
  9. Cycleback, David Rudd. “Subjectivity and Your Unique Experience.” Subjectivity and Your Unique Experience. Accessed January 11, 2017. http://www.cycleback.com/subjectivityofsubjectivity.html.
  10. Ullman, Montague. Working with dreams. Dell Pub. Co., 1980.
  11. Allan, Hobson J., and McCarley Robert. “The brain as a dream state generator: an activation-synthesis hypothesis of the dream process.” Am J Psychiatr 134 (1997): 1335-1348.

About the Author

Yutong Cai, Singapore, 19

Yutong spent her first 15 years in Chengdu, a city known as the “Land of Abundance” in China. At the age of 15, she was rewarded with a government scholarship to study abroad in Singapore.
She often represents her school in various local and international science and math competitions.
Beside a passion for STEM subjects, Yutong loves classical Chinese dance and philosophy.

Leave a Reply

Your email address will not be published. Required fields are marked *