In to the understanding of cognition as of

In
the past three decades, the use of functional magnetic resonance imaging (fMRI)
has exploded within the literature, offering an alternative biological
perspective in the field of psychology. Much of the growth within human
neuroscience research has been made possible by the development of fMRI
(Poldrack & Farah, 2015). As a tool for understanding cognition, it allows
a coarse measure of brain activity while completing a cognitive task (Henson,
2006) and has become the most advanced form of investigating brain activity in vivo
(David et al., 2013). Resultantly, cognitive psychologists and neuroscientists have
embraced this resource as an insight into the biological basis of cognition (Yarkoni
et al., 2011). It facilitates the isolation and localisation of processes in
the brain in the attempt to link specific brain activity with a particular
function (Poldrack, 2006). It is also often used for purposes beyond functional
mapping such as contributing to cognitive theoretical frameworks and hopeful
researchers anticipate that fMRI will have an integral role in behaviour
prediction. However, due to the overwhelming engagement and interest
surrounding fMRI research and its quite recent development, it remains subject
to much criticism. Conducting fMRI research is accompanied by many limitations
that are often overlooked in the pursuit of compelling results. The literature
has repeatedly outlined such limitations and the dangers of not addressing them
while also highlighting the progress that may be possible if such limitations
are acknowledged and amended appropriately. This discussion intends to
critically evaluate what fMRI has contributed to the understanding of cognition
as of yet and what it may contribute to future research if developed properly.

fMRI
remains the least invasive tool compared to other functional imaging techniques
such as positron-emission tomography (PET). Unlike other forms of neuroimaging
and psychophysiology such as electroencephalography (EEG) and magnetoencephalography
(MEG), fMRI boasts unparalleled spatial resolution and can reliably map brain
areas during task engagement (Logothetis,
2008). However, fMRI registers a proxy for neuronal activity and is resultantly
met with criticism. fMRI identifies changes of blood oxygenation brought about
by neuronal activity as measured by the blood
oxygenation level dependent (BOLD) signal. While valid in its assumption to be
linked to neuronal activity, interpretations of this signal can be misguided,
and findings may not be fully representative of true brain activity. For
example, the role of various glial and neuronal cell types in the association
between the BOLD effect and neuronal activity remains ambiguous (Poldrack &
Farah, 2015). Additionally, the BOLD signal reflects general activation of a region
and does not distinguish between excitation or inhibition of local neurons (Lee
et al., 2010, Logothetis et al., 2001). Likewise, the anticipatory nature of
the BOLD signal suggests it is likely to only reflect input activity to an area
rather than illustrating output (Logothetis, 2008). The equating of blood flow
to neural activity has been met with caution and critique yet many within the
field maintain that fMRI is one of the most valuable tools within cognitive
neuroscience to date and worth developing (Farah, 2014).

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!


order now

Nevertheless,
the haemodynamic response limits temporal precision of fMRI due to its sluggish
nature. The BOLD signals temporal resolution cannot match the superiority of
other techniques such as EEG and MEG and therefore cannot accurately register the
neural basis of many fast cognitive processes (Glover, 2011). For example, in
the study of error-processing, brain mapping has localised the anterior
cingulate cortex (ACC) as having an integral role in action monitoring and
responding to erroneous behaviour (Kerns et al., 2004, Carter et al., 1998).
That said, fMRI does not offer sufficient insight into the neurobiological
features of error-processing when compared with EEG (Hajcak et al., 2005,
Gehring et al., 1993), that has temporal resolution capable of identifying event
related potentials (ERPs) by the millisecond. This allows the detection of definitive
neural markers of error-processing such as error related negativity and error
positivity that a functional map of the ACC BOLD signal would fail to register.
Continuing this example, fMRI and ERPs have been paired as complimentary
methods that offer a far superior spatio-temporal mapping of error-monitoring than
either could achieve alone (van Veen & Carter, 2002).

However,
if EEG proves more useful at capturing and understanding cognition than fMRI,
its relevance within cognitive neuroscience is questionable. Are fMRI studies
simply eye-catching illustrations or can they also tell the story of cognition?
Critics have doubted the usefulness of brain mapping within cognitive
neuroscience and argued that fMRI is simply the new phrenology (Uttal, 2001).
Some question the significance of finding correlational relationships between
brain activity localisation and cognition and doubt its ability to enhance the
understanding of how the brain performs cognitive function (Coltheart, 2013). The
correlational analytics in fMRI research are in many cases, grossly inflated
and damage the credibility of fMRI (Vul et al., 2009). Stringent alpha levels
and small sample sizes all pose as threats to correlations in fMRI results and
jeopardise the possibility of true discovery (Yarkoni, 2009). In a review of
234 fMRI studies, Carp (2012) notes that the majority use single group designs
with minimal number of subjects, rendering them at risk to false positive results
if effect sizes are low. While in theory only five percent of studies should
display Type I errors, many software packages designed for fMRI analysis
produce false positives at a rate of up to 70 percent, with one software package
producing consistent errors for 15 years (Eklund et al., 2016). Bennett et al. (2009) report that 20 to 30 percent of
fMRI papers in prominent journals did not use corrected statistical thresholds
and consequently, readers are not informed on the true risk of Type I errors. As
an alarming example, a dead Atlantic salmon showed active voxel clusters when
using uncorrected statistical thresholds (Lieberman
& Cunningham, 2009). Multiple studies focusing on the same issue are
required to overcome the underpowered and misleading nature of many fMRI
studies (Yarkoni et al., 2011). Although uncommon within practice (Gilmore et
al., 2017), cognitive neuroscientists are urged to promote open sharing of data
and homogeneity of methodology and ontology if these issues are to be overcome
and cognition understood (Poldrack et al., 2013). Efforts are currently
underway with open sharing groups such as 1000 Functional Connectomes Project (Yan et al., 2013), OpenfMRI (Poldrack et al., 2013) and the fMRI Data Centre (Gilmore et al., 2017).  Yarkoni’s creation and maintenance of the Neurosynth platform synthesises many
fMRI results to find reoccurring and reliable trends in brain mapping. Such
large-scale initiatives facilitate fMRIs valuable contribution to the
understanding of cognition as both errors and meaningful results are made clear
(Yarkoni et al., 2011). Addressing these issues is a promising solution
to the challenges in neuroimaging and may help delineate cognitive functions
through activation mapping (Poldrack & Yarkoni, 2016).

fMRI
paradigms need to be created and manipulated with extreme care if brain mapping
is to develop and enhance the understanding of cognition. Adaption paradigms
have allowed for brain mapping to explain more than the location of brain
activity by imaging changes in activity or activity patterns while engaged in a
repeated perceptual or cognitive task (Farah, 2014). Additionally, Farah writes
that providing a full map of brain functioning over a period of time is unique
to fMRI and can not only describe location but indicate connectivity between
areas and specific brain systems, a task far beyond the scope of single cell
recordings. That said, conclusions on cognitive function must be made with
caution using a fMRI study alone. Logothetis (2008) warns that studies cannot
assume that cognitive function is exclusively modular and even if that were the
case, it would be naïve to assume cognitive functions can be identified without
correct recursive experimentation. For example, in the analysis by Carp (2012),
223 of 241 fMRI studies used a unique analysis resultantly suggesting many were
unlikely to obtain reliable and replicable results. In fact, Bennett and Miller
(2010) report that when evaluating reliability in fMRI research, no agreement
on adequate reliability or how to measure was found causing wide spread
misinterpretation. If software can become standardised as is the objective of
projects such as the functional Biomedical Informatics Research Network,
reliability within fMRI will ensure sound contributions to the understanding of
cognition. Repeated forward inferencing using standardised and reliable
experimental design will lead to bidirectional associations becoming “de facto”
as has been done with episodic memory (Gilmore et al., 2017). In the field of
clinical cognitive neuroscience, distinct and replicable activation patterns
are being identified and utilised in treatment development. fMRI mapping has
identified associations and dissociations between psychiatric traits which has
steered research away from conventional categorical DSM based research to a transdiagnostic
approach to mental health (Gillan et al., 2017, Mather et al., 2013). Using
fMRI to delineate the cognitive basis of disorders enriches the field of
psychiatry that has conventionally relied on observation and subsequently been
limited in knowledge (Sanislow et al., 2010). According to the Research Domain
Criteria, the integration of functional neuroimaging in psychiatry will
contribute to the understanding of basic cognitive functions ranging from
normal to abnormal (Insel et al., 2010).

While
localising and mapping has purpose in cognitive neuroscience, research has been
fiercely concerned with whether fMRI contributes to cognitive theories. Sceptics
of the technique claim that fMRI has little use in the study of cognition
because fMRI itself can only be understood after a theory of cognition has been
established, rendering it redundant (Harley, 2004). Coltheart (2006), an ardent
critic of fMRI claims it has not yet offered any insight into cognitive theory
just has viewing computer hardware does not offer insight into the workings of
the computer software. Nonetheless, this opinion is not consistent, and many
researchers engage at length to prove that fMRI has and is currently developing
cognitive theories. Advancements in fMRI analysis such as representational
similarity analysis has facilitated the formation and clarification of
cognitive theories, for example in episodic memory (Xue et al., 2010). Using
the example of dual and single process models of recognition memory, Henson
(2006) argues that fMRI can distinguish between competing cognitive theories
when issues cannot be solved by behavioural experiments. Mather et al. (2013)
propose the selectivity in activation to specific tasks can dissociate between
cognitive function and be the basis of distinguishing between competing
theories. The dissociation technique and use of forward inference within fMRI
can be used to pinpoint the type of activity triggered by a specific task and
if a task engages distinct or various neural networks. Such methods are
extremely valuable to the formation and falsification of cognitive theories. That
said, in the effort to find meaning within results, researchers may fall victim
to a “consistency fallacy”, where fMRI results consistent with theory are
assumed as proof of the theory (Coltheart, 2013). Likewise, the improper use of
reverse inference has damaged fMRIs ability to contribute to cognitive theory. Reverse
inference is the method by which the engagement of a particular cognitive
function is inferred based on observed activation in a particular area
(Poldrack, 2006). Poldrack (2006) gives the example of such logical fallacy by
the deductively invalid conclusion that suckling was more rewarding than
cocaine administration in rat pups due to larger BOLD activation of the reward
system during suckling than when ingesting cocaine (cf. Ferris et al., 2005).
However, while imperfect, reverse inference in fMRI data has potential to
contribute to cognition. The union of behavioural and fMRI data can help
justify reverse inferences and areas of high selectivity pose the most valid
areas for reverse inference (Poldrack, 2006). The resources discussed, such as
homogenised neural ontologies and large-scale meta-analyses in Neurosynth, can
promote the establishment of de facto relationships which offer a promising
starting point for reverse inference (Yarkoni et al., 2011). Establishing
reliable and selective associations can allow fMRI reverse inferencing to investigate
new novel hypotheses and enhance fMRIs contribution to cognitive theories.

Additionally,
fMRI research in cognitive theory has advanced from descriptive to predictive
in recent years. Research in predicting decision making has challenged the long-standing
theory of cognitive dissonance (Moran & Zaki, 2013). That is, neuronal
activity has proven capable of predicting a post choice shift in preference in
individuals prior to their response, undermining the notion of post choice
justification as explained by behavioural data according to cognitive
dissonance theory (Moran & Zaki, 2013, cf. Sharot, De Martino, & Dolan,
2009). fMRI research continues to develop in a hope that brain activity mapping
will be used to predict future behaviour as has been done in decision making
research. Another example of such research involves the activation mapping of
the reward centre where enhanced activation of the system during the listening
of a never before heard song was a stronger predictor of sales of the song over
a three-year period than likeability self-reports (Berns
& Moore, 2012). Lie detection and accountability of a crime are also
possible avenues for the future of fMRI if limitations are addressed (Poldrack
& Farah, 2013). Individual differences within moral agency have been
identified within the dorsal frontal and ventral medial prefrontal cortex which
may guide the decisions within the legal system by aiding the prediction of future
criminal behaviour (Greene & Haidt, 2002). Neural markers found within
cognitive fMRI studies have provided better prediction of future behaviour
alone or in combination with behavioural measures than behavioural measures
alone (Gabrieli et al., 2015). This suggests that fMRI studies may make
valuable contributions to the current understanding of cognition and be applied
to policy and practice.

The
present understanding of cognition is a developing science that relies on the
practice of sound experimental design and analysis. fMRI can contribute greatly
to this understanding if limitations are appropriately identified and overcome
within research.  fMRI has and can
continue to facilitate great progress in the understanding of cognition but
must be critically evaluated if this is to come about. While its temporal and
haemodynamic limitations are inevitable, many of the issues can be overcome as
has been proposed throughout the course of this discussion. If these
improvements are made and put into practice, functional brain mapping can
contribute greatly to the understanding of cognitive processes. Furthermore,
cognitive theories can be enriched and distinguished by this technique. While
only emerging, there is much promise of fMRI not only describing brain activity
and inferring cognitive processes, but also predicting future behaviours based
on cognitive fMRI research.