There are numerous ways of non-invasively creating images of neuronal activity in humans which can be split into two categories; metabolic-based and electrophysiological-based (Burle et al., 2015). The two methods that will be addressed in this comparative analysis can be categorized into each one of these classifications; ERP taking a more electrophysiological approach and fMRI from the metabolic based approach. The purpose of extrapolating neuroimaging data is to not only identify the function of a concentrated group of neurons but also to support and be used as evidence for cognitive theories. Therefore, this comparative analysis will also investigate the application of both theories to the topic of recognition memory.
Event-Related Potential (ERP) technique is increasing in popularity in both the clinical and non-clinical fields as it can measure the electrophysiological functioning across the networks of neurons of the brain, which are time locked in response to sensory, cognitive and/or motor stimuli using electroencephalography (EEG) which involves the placing of electrodes on the scalp (Kropotov, 2016; Luck, 2014). Typically, there is an average of 10 peaks per second (10 Hz), however, instabilities caused by the uniformity of groups of cortical pyramidal cells produce an observable charge separation in post-synaptic potential dipoles that can be detected at the scalp. The lining up of these cells are perpendicularly oriented in the direction of the cortical surface, allowing for the examination of the components of the ERP brain wave. Consequently, leading to the identification of the structural components of the brain involved as well as the timings of the response to be deciphered (Tatum, 2014). The waves that are initially produced are “noisy” thus there is a need for filtering and editing to eliminate unrelated variables and to isolate fluctuations in electrical activity caused by the stimulation; this produces an ERP wave (Widmann & Schröger, 2012).
Functional Magnetic Resonance Imaging (fMRI) is an effective non-invasive method of exploring regions of high metabolic activity in the brain creating a blood-oxygen-level-dependent (BOLD) contrast through the use of a strong magnetic field produced by an electromagnet which typically has a field strength of 1.5-3 teslas (T) in the research and medical fields but can reach strengths of 7T (Huettel, Song & McCarthy, 2014). The electro-magnet examines the localized increase and/or decrease of oxygenated blood flow in regions of high neural activity by identifying the charge difference of oxygenated and deoxygenated blood due to the hydrogen protons found in hemoglobin (Buxton, 2013). Prior to stimulation the magnetic nuclei are randomly oriented, however once exposed to the electromagnet, alignment is positioned in the direction of the magnetic field (Bergger, 2002). The high strength of the magnet is necessary as the higher the degree of magnetism the more affiliated the nuclei become, which produces a signal that can be detected and differentiated by the fMRI so that the structural component(s) causing the activity can be distinguished. It, however, is important to note that this alignment is not the intentional measure of the fMRI but is a consequential response. The hemodynamic response causing the high concentration of the nuclei in the activated region is due to the increased stimulation of populations of neurons causing them to necessitate oxygenated blood. Blood flow is therefore used as a proxy for neural activity as it infers that the differential activation is in response to some presented stimulus or due to a task the individual is asked to partake in whilst in the fMRI scanner.
One of the main advantages of using fMRI as a diagnostic tool is the exceptional spatial resolution (approx. 1mm3) (Chen & Ugurbil, 1999) that it offers despite the occasional compromise of a poorer temporal resolution in favor of higher accuracy in the form of smaller voxel sizes (Crosson et al., 2010). The exceptional spatial resolution, relative to ERP, is highly advantageous as it allows the location of the collection of neurons that are responding to a stimulus to be identified. However, there are numerous disadvantages to fMRI. The main one being that the smaller the voxel size the poorer the temporal resolution. This is a disadvantage as, despite the ability to locate where in the brain simulation occurs, the relatively inferior temporal resolution cannot illustrate an accurate time frame whereby the response occurred (Glover, 2011). Causation can therefore not be precisely inferred as the activation of regions detected by the fMRI may not necessarily be in response to the presented stimulus as it cannot be time-locked to the event.
On the other hand, the comparatively superior temporal resolution provided by the ERP method is the main advantage of using the technique as brain activity can be tracked to the degree of milliseconds (Howe, Knott & Conway, 2017). The ability to track brain activity to this measure allows the researcher to see the behavior of electrochemical activity as it occurs in real time allowing an unequivocal assessment of brain response and function. The relatively poorer spatial resolution of the EEG is a disadvantage as it only allows for the general region of activity to be identified whereas fMRI allows for a more localized detection of brain activity.
It may be possible for one to argue that the limitation to blood vessels of fMRI decreases the advantageous nature of its spatial resolution. Nevertheless, blood vessels are spread out over the majority of the cortex, therefore, most researchers are of the opinion that this argument is redundant. Another argument that may be relative to the validity of the fMRI technique is the use of blood flow as a proxy for interpreting neuronal activation as they are linked (Anderson, Mizgalewicz & Illes, 2012). This, therefore, assumes that neural activity is linked to the metabolic response and thus associated with blood flow (Cardoso, Sirotin, Lima, Glushenkova & Das, 2012). Supposing this key assumption is correct, the limitation to blood vessels is not a disadvantage as all brain activity requires a hemodynamic response of some fashion as without it an individual is declarable “brain dead” (Mintun et al., 2001) making blood flow a valid proxy for brain activity.
The conduction of numerous trials of EEG data is needed in order to isolate the neurocognition ERP wave by creating an average result and discarding unrelated data (Coles & Rugg, 1996). This is a disadvantage as it means that it is a very time-consuming process unlike the singular process required in a fMRI trial. Furthermore, the cost efficiency of ERP relative to fMRI is an advantage of the technique as generally the establishment of an EEG laboratory approximately has an expenditure of £70-100K which is comparatively a more reasonable disbursement than the £3-8 million that a fMRI lab requires to construct. This makes an EEG machine more readily accessible than a fMRI machine.
Both techniques can be applied to the investigation of recognition memory. Recognition memory is a subcategory of Explicit memory that can be defined as “the conscious, intentional recollection of factual information, previous experiences, and concepts” (Ullman,2004). There are two subdivisions to recognition memory; the ‘Recollection’ process and the ‘Familiarity’ process (Medina, 2008). The recollection process entails the meticulous, intentional means of evoking the memory of a previously experienced event (Mandler, 1980), whereas the Familiarity process is an automatic response that lacks the recall element yet one is subjected to a feeling of recognition at the instance of the event (Jacoby, 1991).
There are two theoretical models on how these two processes interact. The dual process model suggests that there may be two unconnected and independent memory systems for each of the subsets of Recognition Memory. On the other hand, the single process model supposes that both subsets are operated by one memory system but that they are on a continuum and that the ‘Recollection’ subclass accesses a higher and stronger memory on the continuum than the ‘Familiarity’ subclass; Hence the more vivid detail in the ‘Recollection’ subgroup (Earhard & Landry,1976). The main methods of measuring Recognition Memory is the ‘Old-New Recognition Task’ and the ‘Forced Choice Recognition Task’. Both tasks involve identifying the correct target word/object which is an item that the participant has previously been exposed to. The only difference between the two tasks is that the ‘Old-New Task’ involves a ‘yes/no’ response; ‘yes’ being for an old word and ‘no’ being for a new item, whereas in the ‘Forced Choice’ task they need to identify it in a distraction group of up to four items (Rabinowitz & Graesser, 1976).
Using the ERP technique to investigate Recognition Memory is set up in the same manner as outlined at the start of this essay. EEG is an effective method of studying Recognition Memory as ERP studies of the topic have found that during the familiarity subpart there is an electrical activation that peaks at 300–500 ms in the central frontal lobe whereas in the recognition process the peaks are at 400-600ms in the left parietal lobe (Ratcliff, Sederberg, Smith & Childers, 2016; Rugg, 1995; Rugg & Curran, 2007). ERP evidence corroborates the dual process explanation model of recognition memory as the increase in oscillations (4-8 Hz) for both processes occurred in different locations of the cortex and do not have a clear regressional relationship linking them indicating that two separate systems are responsible for the occurrence of ‘Familiarity’ and ‘Recollection’ (Jacobs, Hwang, Curran & Kahana, 2006). This is further support for Curran’s (2000) observation of a difference in the FN400 component of familiarity and recollection indicating that there is a plausible reason to believe that the dual process theory is correct.
The ERP technique has many positive contributions as it has enabled the expansion of research into recognition memory through the growth of insight into the effect of emotions on recall ability. ERP has identified that there is more activation in the posterior parietal lobe during recognition of negative stimuli than positive ones; This is a finding consistently found through developmental studies showing that it is the same from older childhood to adulthood (Leventon, Stevens & Bauer, 2014; Carver, Meltzoff & Dawson, 2006). However, a limitation to using the ERP technique to study recognition memory is that due to the time locked nature required of understanding ERP waveforms, it may be difficult to analyze data in relation to recognition memory as there is no obvious measurable event that is the cause of recall and an EEG has a poorer spatial ability to localize ERP effects relative to other methods (Tulving & Craik, 2005). Thus, the ERP technique may not be the most beneficial method in the application of recognition memory.
The fMRI process of investigating recognition memory is more complex than ERP as it is a two-part process whereby the BOLD response is measured for the ‘recollection’ condition and the ‘familiarity’ stimuli separately. fMRI studies can supply more accurate location data than an EEG analysis could not derive. A larger hemodynamic response in five regions of the cortex were identified by the fMRI to be in relation to recognition; ‘Left and right lateral inferior parietal cortex, medial parietal cortex, left dorsal middle frontal gyrus, and left anterior prefrontal cortex’ (Donaldson, Petersen & Buckner, 2001). On the other hand, evidence supporting a separate memory process for the ‘Familiarity’ aspect presents itself in the form of ‘encoding activity in the anterior parahippocampal gyrus, primarily the peripheral cortex’ in familiarity-based recognition judgments (Ranganath et al., 2004). This supports the dual approach theory as there were two separate regions of activation that did not traverse into the other subset of recognition memory.
The application of other theorems; e.g. the novelty theory, is also a positive of fMRI as it allows for the expansion of knowledge into cognitive neuroscience fields, such as recognition memory. fMRI studies have furthered research by identifying higher levels of activity in the aforementioned regions in response to novelty effects; i.e. the presence of novelty effects increased both recall and familiarity (de Chastelaine, Mattson, Wang, Donley & Rugg, 2017; Kishiyama & Yonelinas, 2003). The poorer temporal reliability of fMRI studies invalidates single-subject studies without a test-retest element as recognition memory responses are not time-locked it is problematic to ascertain that the haemodynamic change is due to the presented stimulus; therefore, corroboration from either another task or a different neuronal imaging technique may be required (Brandt et al., 2013).
Research may, therefore, benefit from a dual approach to research and diagnoses as it is apparent that both techniques complement each other. The integration of both techniques may, therefore, have many benefits in cognitive neuroscience research as it would allow for the strengths of both techniques to unify and collectively generate, in theory, more precise conclusions. In terms of recognition memory, it is evident that the dual process approach is the more widely accepted model for both methods as the ERP technique demonstrates the differences in timings and the fMRI technique illustrates the difference in location of activation for each process. In conclusion, both methods have their own individual strengths and limitations and the choice of which technique to use lies entirely on the foci of one’s future research; whether the interest lies more in when or where activation occurs. Both methods individually provide sufficient insight, but a multi-modal approach would further substantiate and validate any further research and would be a route to further research that would be highly recommended.