Electroencephalography in Consumer Research A cognitive psychological and neurological review of consumer research using EEG

(1)

Electroencephalography in Consumer Research

A cognitive psychological and neurological review of consumer research using EEG

August 2016

Yuhee Kim (10859969)

Supervisor: Dr. M.A.S. (Maarten) Boksem

Co-assessor: Dhr. dr. I.G. (Ilja) Sligte

Literature Thesis

MSc in Brain and Cognitive Science, Cognitive Neuroscience Track

(2)

This paper examines research in three main areas: electroencephalography (EEG), neuroscience, and consumer behavior. Its main purpose is to provide a basic knowledge of research on consumer neuroscience for marketers who find neuroscience methodology to be difficult and inaccessible. Chapter 1 describes EEG as a neuroimaging tool. In this chapter, I provide a general overview of EEG, including its historical background, strengths and weaknesses, prominent features, components, and equipment. Chapter 2 describes cognitive and neuropsychological approaches to the fundamental cognitive functions related to consumer behavior in which most marketers are interested. In this chapter, I discuss the main cognitive functions underlying information processing and decision-making as outlined in neuroscience research using EEG. Chapter 3 describes a practical approach to consumer neuroscience. In this chapter, I elaborate on how the results of studies on cognitive neuroscience can be utilized for marketing purposes and to conduct consumer neuroscience research using EEG. In the conclusion, I summarize the previous chapters and analyze the advantages and limitations of neuromarketing using EEG for future research, evaluating the value of EEG for use in consumer neurosciences research for business purposes.

As neuroscience techniques have rapidly advanced, researchers in the psychological and physiological sciences have adopted neuroimaging methods to investigate brain and cognition in considerable depth (N Lee, Broderick, & Chamberlain, 2007). Then, neuroscience tools have been utilized in various consumer studies to improve our understanding of the market. It is known that 95 percent of thinking is unconscious, but conventional market research has focused on the conscious mind only, as no methods had been developed to capture unconscious thought. However, in recent years, neuroimaging has been developed, which allows researchers to observe brain activity, including the unconscious thoughts and feelings of consumers (Zaltman, 2003). Thus, the subdiscipline of consumer neuroscience known as neuromarketing was created.

The unanswered questions of all marketers are, “How do consumers process information?” and, “What drives them to make certain choices?”. In response to these questions, the Neuromarketing Science and Business Association (NMSBA) was established to promote neuromarketing in the real business world. The official objectives of the NMSBA are to provide science-based resources and build an international network of neuromarketing researchers.

(4)

Currently, this organization consists of members in 90 countries who attend an annual neuromarketing world forum. In recent years, major brands have partnered with members of the NMSBA to initiate practical attempts to converge marketing with neuroscience. For example, Porsche (Bell, 2015) and Coca Cola (Dooley, 2013) implemented neuroscience techniques to develop their products. Hyundai, Frito-Lay, Yahoo, and eBay briefly disclosed their applications of neuroscience as part of their marketing strategy and services in a Forbes interview (Burkitt, 2009). As more and more businesses began to explore the possibilities of neuromarketing, Nielsen, a global market research company, enthusiastically acquired neuromarketing startups like Neurofocus and Innerscope and organized an internal neuroscience research division (Dooley, 2015). Now, Nielsen provides an exclusive service evaluating TV advertisements with the combined measures of neurometrics, biometrics, facial coding, eye tracking, and self-report. According to Nielsen, their new service helps clients see a fuller picture of the consumer’s mind.

In addition, papers and articles with a focus on consumer neuroscience have been published to compensate for the weaknesses of conventional marketing research methods. Traditionally, survey or focus group interview methods have been used to probe consumers’ minds, but these methods have the limitation that their results are easily distorted by external factors like culture, social norms, or even interviewers’ behaviors. Despite the time and effort involved, traditional research methods are inadequate to measure consumer thinking (Zaltman, 2003). Thus, researchers have recently been motivated to employ neuroscience methodology in consumer research.

However, there are ethical concerns and skepticism about neuromarketing in general. This discipline is becoming more familiar as more studies are openly conducted according to the regulations of the NMSBA, but only a few companies have made their results public, and most are reluctant to reveal their neuromarketing applications. General criticism of neuromarketing may be found in the description of it as “finding a buying button in the brain”(Nick Lee et al., 2007), or the perception that it involves searching for ways to manipulate consumers subconsciously. Moreover, even marketing academics consider neuroscience and cognitive psychology as intimidating subjects due to their lack of experience with related techniques (Nick Lee et al., 2007). Better understanding of neuromarketing and its foundation is required to eliminate prejudice and alleviate growing concern.

This paper particularly focuses on consumer neuroscience research in which EEG is employed. EEG is a technique that has been used since the initial stage of neuromarketing research and is still widely adopted by many researchers. In spite of its great accessibility and usability, its value has been underestimated because of its weakness in terms of spatial resolution and its inability to detect subcortical activations in the brain. However, its strengths in terms of

(5)

temporal resolution and minimal cost make it appropriate for business applications, and its weaknesses can be complemented by high-density EEG systems. In this paper, an overview is provided of consumer neuroscience research using EEG that encourages marketers to use it. After reading this paper, readers will be more familiar with consumer neuroscience and EEG and the possibilities and advantages of combining these things in future research.

(6)

Chapter 1

An Overview of Electroencephalography (EEG)

The aim of this chapter is to provide a brief overview of EEG, including its historical background, advantages and disadvantages of using it in consumer research, classification of EEG and ERP, and how the EEG device works. This basic knowledge of EEG will help readers understand the concepts presented in the next chapter.

Historical Background of EEG

Communication between brain cells produces tiny amounts of electrical activity. EEG measures and records these potential changes in the brain. EEG originated from physiological experiments measuring electricity in the body as a whole, which was achieved by Luigi Galvani (1737 – 1798), who attempted to record electrical charges and muscle activity in living frog tissue. Like Galvani, the earliest scientists were initially interested in physiological potentials. Only later did scientists interest themselves in electrical activity in the brain. Richard Caton (1842 – 1926) measured electrical potentials in the brains of rabbits and monkeys, recording a frequency response from the surface of the gray matter. Then, in 1924, Hans Berger (1873 – 1941), the first electroencephalographer of the human brain, observed electrical charges in the cortex of a 17-year-old boy who was undergoing brain tumor surgery. After many tests, he invented scalp electrodes made of foil and tested them on his son Klaus and a 37-year-old volunteer. EEG records of decent quality were obtained, especially in the bald area of the volunteer’s head. Berger published the first report on human brain EEG in 1929 and continued to develop EEG recording devices and techniques. His contributions to experimental and clinical EEG are significant; subsequent scientists benefited from his works. Since that time, EEG devices have been modernized and computerized for use in modern neuroscience labs. In addition, automatic recording, mapping, and analysis of EEG data has become easier for both academic and commercial purposes (Collura, 1993).

Advantages and Disadvantages of EEG

The most common way to detect an EEG signal in human subjects is through use of scalp electrodes, which can record the electrical fields produced by postsynaptic potentials in the grey matter of the brain. A modern EEG device has two great advantages as a brain imaging tool. Firstly, it is a noninvasive method with great time resolution. Scalp electrodes can catch the moment at which electrical events take place at a resolution of a single millisecond without

(7)

inserting electrodes into the brain, unlike other electrical recording devices. It is a participant-friendly tool that is painless, comfortable, safe, and easy to apply on healthy subjects. Secondly, the modern EEG device is relatively inexpensive and easy to operate compared to other devices. Functional magnetic resolution imaging (fMRI), for example, is an expensive measuring device; it costs a lot to maintain the equipment, whereas EEG does not require any extra maintenance cost despite its high usability.

The main disadvantage of EEG recordings is their poor spatial resolution. Since the scalp electrodes are situated a few centimeters away from the neurons where real neuronal communication take place, they do not capture electrical activity of individual neurons, but electrical fields produced by large populations of neurons from various sources. Since electrical fields spread in all directions, strong electrical activity can be detected by several close electrodes, but EEG alone cannot identify the precise source of the activity. However, high-density EEG systems combined with advanced spatial analysis methods have been developed to improve spatial resolution. In addition, simultaneous EEG and fMRI application has also emerged as an alternative.

EEG Classification

The most prominent feature of EEG is its ability to detect oscillations at multiple frequencies. Most EEG studies rely on finding a distinct frequency band depending on the purpose of the experiment. The brain wave spectrum consists of five bands: delta, theta, alpha, beta and gamma waves, each of which is associated with a unique state ranging from alert wakefulness to deep sleep. Delta waves are oscillations between 0.5 – 4 Hz (the slowest brain waves) that are present in the unconscious human mind. They are dominant in infants and children and appear during the deep, dreamless sleep stage in adults. Delta waves and muscle artifacts are similar in appearance; thus, careful observation is required. Theta rhythms include oscillations between 4 and 7 Hz, which appear during deep meditation and light sleep such as the drowsy or REM states. Theta waves are also evident during times of emotional stress such as when a subject is feeling frustration or disappointment. Alpha waves are found in the frequency range of 8 to 13 Hz; these were the first EEG oscillations described by Berger (1929). Alpha waves lie at the base of the conscious mind and indicate relaxed awareness. They are closely related to concentration, learning, and memory. “Alpha blocking” occurs during times of cognitive or emotional awareness. Beta waves occur within the range of 13 – 30 Hz and appear throughout the day if the subject is awake. These waves are involved in critical and logical thought and increase during focusing on the outside world or when solving concrete problems. Lastly, gamma wave

(8)

oscillations indicate brain waves between 30 and 100 Hz; these are the fastest waves. They are associated with high-level cognitive functioning such as information processing.

ERP Classification

ERP is a time-locked EEG activity indicating a response to specific stimuli (Sur & Sinha, 2009). It is a high-resolution measure of neuronal activity changes to sensory, motor, or cognitive events within a specific time frame. ERP components are named according to their latency times and amplitudes. Most ERP studies investigate the roles and functions of these components.

“P” indicates a positive peak. The numbers P50, P100, P200, P300, and P600 stand for positive peak deflects occurring 50, 200, 300, and 600 milliseconds after the stimulus is presented. P50 is identified as “sensory gating” and P100 (or P1) is observed when visual stimuli with attention are processed. P200 (or P2) is associated with sensation-seeking behavior and P300 (or P3), the ERP component that has been researched most, appears when stimuli are distinct and easily distinguishable from another. The amplitude of P300 increases when the stimulus is processed with great attention. Lastly, P600 appears in language processing.

“N” reflects a negative peak. N100 (or N1) appears when an unexpected stimulus is presented and N200 (or N2) is shown when stimuli are changed. Lastly, N400 is known to be associated with semantic incongruity.

ERP research has been conducted to study cognitive functions of healthy subjects as well as to diagnose psychiatric and neurotic disorders. Currently, ERP is being used to help researchers understand consumer behaviors in depth.

EEG Device

An EEG recording system consists of electrodes with conductive media, amplifiers with filters, A/D converters, and a recording device (Teplan, 2002). Electrodes can be applied on the surface of the head directly using adhesive tape, but they may also be fixed in the holes of a customized EEG cap (Figure 1). In this case, a conductive gel bridges the gap between the electrodes and the scalp to reduce resistance and ensure a clean signal. However, because of the gel application, conventional EEG recording procedures require substantial amounts of time to prepare and also cause inconvenience to participants. As an alternative, a recording system with dry electrodes was invented (Figure 2). This new system is appropriate for long-term recording with short preparation time. It is more participant-friendly because it uses dry sensors which do not require gel application. Most importantly, it yields almost identical results compared to the

(9)

conventional EEG system. Therefore, it is expected to be employed more often in future research (Gargiulo et al., 2010).

(10)

Chapter 2

Cognitive Psychological and Neuroscientific Approaches to Fundamental Cognitive Functions Related to Consumer Behavior

The aim of this chapter is to explore the relation between cognition and consumer behavior. Cognitive processes related to consumer behavior such as information processing and decision-making are related to three fundamental cognitive components: attention, emotion, and memory. In addition, neuroscience research related to these three components is described in preparation for the next chapter.

Part 1. Cognitive Psychological Approach to Cognitive Functions and Consumer Behavior

Bartels and Johnson (2015) outline four reasons why cognition research and consumer research need to be combined. First, consumer choices, no matter how trivial they are, influence consumers’ lives more than they imagine. Therefore, understanding the fundamentals of consumer choice is essential for market researchers. Second, collaboration between cognitive science and consumer studies can improve people’s lives. For example, knowing the key concepts of cognition may provide better insights to designers about user experience (UX), resulting in better products and services more closely related to consumer needs. Third, consumer research can be based on large amounts of data compared to the amount of data available in a laboratory setting. Lastly, the consumer setting is a natural domain in which to research cognitive processes. According to Sloman (2015), the Editor in Chief of Cognition, cognition researchers have recently shifted their focus from cognitive functions to the roles of those functions in other fields like consumer decision-making. In future research, studying cognition will be a necessary step to understanding consumers and their choice behavior.

Two cognitive processes, information processing and decision-making, are main focuses in both cognitive science and consumer research (Bartels & Johnson, 2015). In the next section, I examine them and their underlying cognitive components.

Information Processing

Information processing theory is based on how individuals can gain, store, change, and use information and obtain knowledge and skills. Generally, attention and memory are considered as two main cognitive components of information processing theory, according to Donald A. Norman (1976).

(11)

Attention is an initial stage of information processing, the definition of which has changed over time. An early psychologist, Edward Bradford Titchener (1867 – 1927), emphasized the importance of consciousness in understanding attention and asserted that attention determines conscious status. Later scientists focused more on awareness than consciousness and defined attention as a cognitive function that enhances some information and inhibits other information selectively for future processing (Smith & Kosslyn, 2013).

Attention is divided into four types depending on its characteristics. First is selective attention, which means focusing on one aspect of information and ignoring the rest. Second is divided attention, which indicates focusing on more than one activity or aspect of experience at the same time. Third is sustained attention, which involves staying focused on a specific activity for an extended period of time. Last is executive attention, which implies focusing on the information immediately at hand. “Executive attention” includes planning, focusing, detecting, revising, and handling difficulties.

After the attention step, information needs to be encoded into the memory for processing. Eliasmith (2001) defines memory as the “general ability, or faculty, that enables us to interpret the perceptual world to help organize responses to changes that take place in the world”. Smith and Kosslyn (2013) describe memory as “the ability to remember the people, places, and things encountered in the course of daily life… a fundamental form of cognition that guides behavior”.

In cognition research, there are three types of memory. First is short-term memory, which is a memory system that stores a limited amount of information for a limited amount of time, usually a maximum of thirty seconds (Dempster, 1981). This type of memory disappears quickly unless the individual makes an extra effort to keep it. Second is working memory, which is important not only for decision-making and problem-solving, but also for writing and speaking language (Baddeley, 2006, 2007). Last is long-term memory, a permanent information storing system that can hold enormous amounts of information. To convert short-term or working memory to long-term memory, individuals must rehearse or repeat the information and make a robust mental connection for later use.

Decision-making

Decision-making has been an important topic in various disciplines such as mathematics, sociology, psychology, economics, political science, and so on (Buchanan & O’Connell, 2006) because it happens in different contexts dozens of times a day. Many researchers have demonstrated the power of emotion in the decision-making process (Han & Lerner, 2009). Thus, emotion must be considered as a valid cognitive component in decision-making.

(12)

Emotion interacts with decisions as both an input and an output. In other words, decisions and their consequences create certain emotions such as joy or regret; an emotion or the expectation of an emotion may affect the next decision (Wilson & Gilbert, 2005). There are two categories of emotions related to decision-making: expected and immediate emotions (Loewenstein & Lerner 2003, Han & Lerner, 2009).

Expected emotions are predictions of the emotions that will be experienced as a consequence of decisions. They cannot be experienced when a decision is made, but will be experienced when outcomes become actualized in the future. Expected emotions definitely play a critical role in decision-making. Many decision-making theories in economics consider expected emotions only in their models. However, this practice has some drawbacks which could impair decision-making theories. First, individuals are not good at anticipating future emotions that may result from present decisions (Gilbert & Wilson, 2000). Second, expected emotions may distract a decision-maker from critical factors that must be considered at the time of decision-making. Nevertheless, expected emotions are considered as important in decision-making theory.

Immediate emotions are real emotions that are experienced at the time when a decision is made. There are two types of affective influences underlying immediate emotions. First are anticipatory influences, also called integral influences, which arise from thinking about possible outcomes. This initially seems similar to expected emotions, but it differs fundamentally in that the factors affecting immediate emotions (modulated by integral influences) have no impact on expected emotions. For example, expected emotions are sensitive to probability, whereas emotions subject to integral influences are sensitive to time and vividness of outcome. Second is incidental influence, which stems from factors unrelated to the decision such as the weather or previous mood. Though incidental influences may seem negligible, several studies demonstrate the power of incidental influences on decision-making.

Making decisions based on emotion may not sound appropriate and reasonable. However, a neuroscience study proved that emotions actually help us to make “rational” choices (Damasio, 1994). This study involving patients with damage in the ventromedial prefrontal cortex found strong impairments not only in terms of emotional reaction, but also in rational thinking and performance of intelligent tasks. This supports the assertion that emotion is an essential mental component that guides our decision-making substantially (Brosch, Scherer, Grandjean, & Sander, 2013).

Attention and memory also play important roles in decision-making. People with cognitive deficits perform less well than healthy subjects in decision-making tasks (Suleman & Kim, 2015). Emotions are related to information processing as well. Finally, unconscious processing of stimuli still creates positive emotions (Herr & Page, 2004).

(13)

In conclusion, attention, emotion, and memory are interrelated fundamental cognitive components for information processing and decision-making.

Part 2. Neuroscientific Approach to Cognitive Functions and Consumer Behavior

Researchers in cognitive psychology have identified the cognitive processes and functions that induce certain behaviors. Similarly, researchers in cognitive neuroscience investigate cognitive processes by studying the neural mechanisms that generate them (Frank & Badre, 2015). Moreover, neural mechanism models provide new insights and perspectives into and refine previously developed cognitive models (Ratcliff & Frank, 2012).

Many laboratory studies show that different features of brain oscillations indicate response to diverse cognitive processes (Bas & Schurmann, 2001). Therefore, the following section provides a review of neuroscience experiments involving attention, emotion, and memory and elaborates on their functionally relevant EEG/ERP signals in detail.

Attention

Selective attention

Attentional modulation of gamma band responses has been demonstrated in studies involving auditory or visual–spatial stimuli. One study of the human auditory system confirmed that a 40-Hz response is elicited by steady state or transient stimulation, and that this number increases when subjects pay attention to select stimuli (Tiitinen et al., 1993). Another study using rotating stimuli identified higher gamma power between 35 – 51 Hz with a moving visual– spatial stimulus (Gruber, Müller, Keil, & Elbert, 1999). Lastly, another study involving visual stimuli with a working memory template demonstrated that early gamma activity is involved in allocation of attention to a selected object (Herrmann & Mecklinger, 2001).

ERP studies associated distinct ERP components with selective spatial attention (Hillyard & Anllo-Vento, 1998) or contrasted selective attention with consciousness (Koivisto & Revonsuo, 2007). Both studies confirmed enhanced P1 and N1 amplitudes with selective attention. Thus, P1 and N1 components can be used as indicators of selective sensory attention. In addition, patterns of ERP components are changeable depending on individual characteristics and age (Mueller, Brehmer, von Oertzen, Li, & Lindenberger, 2008). Careful experimental design is required when subjects are diverse in age.

(14)

One divided attention study examined the reactions in the brain when mental resources are distributed during performance of a highly demanding cognitive task. The results showed a decrease in alpha rhythm when attention is divided. The opposite results were found during rest periods (Molteni, Contini, Re, Caffini, Bianchi, Spinelli, & Torricelli, 2010). It was also found that EEG classification accuracy under divided attention conditions is superior to eye tracking (Rodrigue, Son, Giesbrecht, Turk, & Höllerer, 2015).

Another early ERP study on divided attention suggested that the demands on the brain lead to increases in latency of P300 peaks because of the increased cognitive load (Joseph, 1984). A later study in a dual task setting confirmed that the peak amplitude of P300 decreased progressively as the working memory load increased, reflecting reallocation of attention (Watter, Geffen, & Geffen, 2001). Therefore, when the latency and amplitude of P300 change, it may be taken as evidence of divided attention.

Sustained attention

One study testing the EEG index for sustained attention was based on the ratio, which was utilized to detect task engagement. The authors assumed that sustained attention can increase beta power and decrease alpha rhythm. The results confirmed that the relation between alpha and beta waves changes with incremental changes in task engagement, especially in the frontal cortex (Coelli et al., 2015).

Another study of ERP components and sustained attention found an influence of aging on sustained attention ability. Older adults maintained sustained attention better than younger participants; in addition, larger amplitudes for P2 and P3, components related to resource allocation, were observed (Staub, Doignon-Camus, Bacon, & Bonnefond, 2014). Another sustained attention study tested if caffeine improves sustained attention. The results revealed that larger frontal P2 and parietal P3 components are associated with caffeine intake; however, no correlation was found with the behavioral results. The results of this study still have value for those exploring the potential of ERP in drug research (Ruijter, Ruiter, & Snel, 2000).

Executive attention

One study involving ERP tested if executive attention can be trained by practice or whether it matures with age. Adults exhibited N2 component activity in the frontoparietal areas for incongruent trials, unlike 4-year-olds, who showed no frontal negativity. After training the executive attention function, brain activity in areas where initial negative components were observed had moved from the prefrontal to the dorsal frontal areas. This suggests that executive attention modulates different brain areas with maturation and/or training (Rueda, Rothbart,

(15)

McCandliss, Saccomanno, & Posner, 2005). Similarly, a study of patients with brain injury compared to healthy subjects found negativity components in the left ventrolateral prefrontal cortex as well as the right dorsal anterior cingulate cortex in both groups (Rogers, Donnelly, & Wilson, 2015).

Emotion

Studies on emotion and its connection with EEG mostly focus on measuring the different valence and arousal levels embedded in various emotions. The international affective picture system (IAPS) that provides standarized emotional images at different valence and arousal levels is often used to provide stimuli.

Many studies have focused on finding a major frequency band. In one study with five different basic emotions, intention, anxiety, aggression, sadness, and joy, alpha EEG coherence elicited by each emotion was measured and the results showed an association of aggression and joy with increased alpha coherence, whereas anxiety and sadness decreased alpha coherence (Hinrichs & Machleidt, 1992). Another study revealed that emotional valence is better represented in beta rather than alpha activity, especially in the right temporal area of the brain. More beta activation was detected with positive stimuli than with negative stimuli, which suggests that beta waves may modulate the valence of emotion (Ray & Cole, 1985). More recently, the role of gamma waves has been studied. The results revealed that gamma activity is correlated with induced worry in posterior areas, which is associated with negative emotions (Oathes et al., 2008). Morover, a significant increase in prefrontal theta oscillations with emotional regulation has being recently identified (Ertl, Hildebrandt, Ourina, Leicht, & Mulert, 2013).

Several frequency bands were measured together in another study that demonstrated an association between an elevated theta power band and positive emotional film clips in the bilateral frontal cortical regions. By contrast, the alpha power band decreased over the left frontal areas by comparison to the right areas. Frontocortical alpha asymmetries indicate that the left and right frontal regions are engaged differently in terms of the valence of emotion (Aftanas, Lotova, Koshkarov, Makhnev, Mordvintsev, & Popov, 1998). This has been an important topic of study, serving as “(1) an individual difference variable related to emotional responding and emotional disorders, and (2) a state-dependent concomitant of emotional responding”, according to Coan and Allen (2004). They reviewed a large amount of literature to find evidence of frontal EEG asymmetry as a moderator and mediator of emotions. Their results supported the critical role of this factor in emotional regulation. However, careful consideration is necessary when studying frontal EEG asymmetry because many studies have examined the

(16)

correlates of variation in asymmetry by means of self-report responses, which is technically regarded as manipulable information (Davidson, 2004).

Results of one meta-analysis showed that affective valence and arousal can modulate ERP components. In particular, the effects of arousal on ERP components are consistent across studies, whereas those for valence are inconsistent and reported at several different latency ranges (Olofsson, Nordin, Sequeira, & Polich, 2008). Early posterior negativity (EPN) is the cortical ERP component sensitive to processing of emotional stimuli. It develops around 150 ms and maximizes around 260~280 ms after onset of stimulus (Markus et al., 2003). In the IAPS, an increased EPN is identified with pictures of pleasant and unpleasant scenes compared to neutral images (Schupp, Flaisch, Stockburger, & Junghöfer, 2006).

Memory

Short- and long-term memory

Tracing alpha synchronization and desynchronization is essential to measuring memory using EEG (Klimesch, 1996). Synchronous alpha oscillations from large cell assemblies reflect a resting state or state of funtional inhibition. Therefore, alpha waves are suppressed during memory processing, and different neural networks may start to oscillate.

In memory research, alpha bands are divided into two subcategories, lower and upper bands, which are independent from each other and respond differently in memory processing. One study found an association between long-term memory tasks and task-specific desynchronization in the upper alpha band in the posterior thalamic system and a positive correlation between alpha desynchronization and long-term memory performance. The results confirmed that the upper alpha band plays a role in encoding information into long-term memory and reflects a significant correlation between memory performance and actual memory processing (W Klimesch, Doppelmayr, Pachinger, & Ripper, 1997). Interstingly, there was a 1-Hz alpha frequency difference between good and bad performers during the recall task in a study in which higher alpha frequencies were observed in good performers. This may suggest the invariable nature of alpha frequency (Klimesch, 1997).

The role of theta activation has also been investigated. The results showed that theta synchronization is positively correlated with encoding of new information (Klimesch, 1999). In addition, engaging in a short-term memory task leads to theta band synchronization in the anterior limbic system (Klimesch, 1996).

In other studies, alpha–theta phase coupling, in which increased power is measured at several different frequency bands simultaneously, has been examined. The results reveal that theta and upper alpha acitivities are negatively correlated in the EEG phenomena of good memory

(17)

performers. As memory load increases, theta power in the frontal midline is activated, whereas alpha power in the parietal regions is deactivated. Moreover, beta coherence activation in frontotemporal regions has recently been identified with increasing memory load (Stokić, Milovanović, Ljubisavljević, Nenadović, & Čukić, 2015). Furthermore, a strong relation between theta activity at Fz and gamma activity at F3 and Fp1 during short-term memory processing has also been identified. This may indicate that gamma amplitude could be modulated by frontal theta oscillations (Schack, Vath, Petsche, Geissler, & Moller, 2002).

Another study focused on the direction of EEG coherence during memory processing. The results suggested that performance of short-term memory tasks increases bidirectional frontoparietal EEG coherence in the gamma band, unlike performance of a control task, in which only parietal to frontal flow prevailed. Thus, there may be evidence of increased power flow from the frontal to parietal lobes during memory processing (Babiloni et al., 2004).

In sum, the results of previous research indicate that, during performance of memory tasks, theta power in the frontal area is increased (Stokić et al., 2015), and that this power could be accompanied by increased gamma power (Schack et al., 2002). This results in an increased flow of gamma power from the frontal lobe to the parietal lobe and more even communication from both directions between frontal and parietal areas (Babiloni et al., 2004).

Working memory

Working memory is composed of two parts: central executive function and short-term storage. Functional image studies revealed that executive functions are related to frontal areas, whereas storage processes are associated with parietal areas (Baddeley, 2000). The hypotheses of many working memory studies are related to findings about upper alpha–theta phase synchronization in many previously conducted short- and long-term memory studies, which assume that increasing the working memory load will cause an increase in phase coupling between theta activation in frontal areas (Onton, Delorme, & Makeig, 2005) as well as a decrease in upper alpha waves in right posterior areas. The results confirmed the existence of upper alpha–theta synchronization during both retention and retrieval working memory processes. In addition, upper alpha waves were found to be more sensitive in the retention stage, whereas upper alpha–theta synchronization activity was more evident in the retrieval stage. This hints that theta activity is associated with central executive function (Sauseng, Klimesch, Schabus, & Doppelmayr, 2005) whereas upper alpha is related to storage processes. The presence of alpha–theta synchronization in working memory processes suggested that long- and short-term memory share some properties in terms of processing (Schack, Klimesch, & Sauseng, 2005).

(18)

Conclusion

In this chapter, by outlining the cognitive psychological and neurological background of the discovery of cognitive processes and functions related to consumer behaviors, I presented several insights that researchers in consumer neuroscience and neuromarketing can apply in their research on real consumer behaviors.

In response to academic and practical demands, workers in consumer research and cognition have come together to improve our understanding of consumer behaviors in more depth. Information processing and decision-making, which are important topics in both consumer research and cognition studies, have been investigated in terms of three main cognitive components: attention, emotion, and memory, all of which play critical roles in cognitive processes. Neuronal characteristics of those functions and their interactions have been elaborated in many studies, which may inspire consumer neuroscientists to utilize cognition research in their consumer studies.

(19)

Chapter 3

A Review of Marketing Research Using EEG Based on Marketing Mix Models

The aim of this chapter is to review studies in consumer neuroscience in which EEG is employed in the product and promotion research areas. Attempts have been made to match cortical activations to marketing elements. This hidden information that consumers themselves are not even aware of or cannot explicitly articulate has great power over consumer behavior. Basic information about products and promotions, two important elements in the 4P marketing mix model, are described in this chapter, and related research is analyzed accordingly.

Part 1. A Scope of Consumer Neuroscience Research

Marketing research is “the systematic and objective search for, and analysis of, information relevant to the identification and solution of any problem in the field of marketing" (Green & Tull, 1970). Marketing research concerns every activity in the market as a whole and all elements involved in market operation. Its scope is wide, comprehensive, and diverse. In this paper, I categorize research into that related to products and that related to promotion, based on the four fundamental elements (Smidts et al., 2014) in the marketing mix model created by E. Jerome McCarthy (Anderson & Taylor, 1995).

First, a product is defined as an item produced for the satisfaction of target consumers that involves goods and services. Product research often includes information about product quality and performance, physical and psychological characteristics, competitiveness compared to rival products, and chances of success or failure in the market.

Second, promotion is any activity that enhances brand awareness and sales. The most common and influential method of promotion is advertising. Advertising is a powerful way of communicating using different media such as television, radio, print, the Internet, and so on. A substantial portion of most promotion budgets is spent on advertising; thus, it is also important for marketers. Advertising research includes studies about the objectivity of advertising, media and media selection, and the effectiveness of advertising messages.

Part 2. Product-related Research

Consumer neuroscience research related to brands, products, and services are described in this section.

(20)

Good brand names should be memorable, create positive images, and enhance product preferences (Aaker, 2012). Consumers build their very first impressions of products from brand names. Thus, selecting a good brand name is an important part of forming a brand identity (Kohli & LaBahn, 1997). In recent years, many brands have emerged in global markets. In this international context, deciding on a brand name became even more important; however, it is also difficult because of the unfamiliarity of most original brand names in local markets. For example, in China, to attract local customers, international companies use both original English names and translated Chinese names, but it is a challenge to translate correctly and naturally due to differences in the language systems of English and Chinese. Therefore, marketers have investigated if the translated Chinese name and original English name of a product represent brand identity equally to consumers and are processed similarly despite language differences. A brand name is usually a proper noun that reflects the characteristics of the brand and is considered as a verbal stimulus embedded with meaning. It can also be associated with more complex linguistic processes such as semantic processing. One study confirmed that EEG coherence in the theta, alpha, and beta frequency bands plays a critical role in language processing. In addition, increased intra-hemispheric beta coherence has been associated with high-imagery verbs (Weiss & Mueller, 2003).

To understand how a brand name in different languages is processed in the brain, some researchers investigated neuronal processing of the same brand name in English and Chinese using EEG (Cheung, Chan, & Sze, 2010). EEG coherence of the inter- and intra-hemispheres in the theta and beta bands was mainly considered, as these areas are known to be related to language processing (Weiss & Mueller, 2003). The results demonstrated that reading brand names increased intra-hemispheric beta coherence in both hemispheres in both languages. Given that high-frequency coherence tends to reflect complex linguistic subprocesses such as semantic processes, increased intra-hemispheric beta coherence from reading brand names may reflect the fact that brand names are processed as semantic information regardless of the language. This finding can be interpreted as follows: brand names are high-imagery words associated with brand identity whose attributes and features may elicit meaning for consumers beyond just a name. Interestingly, Chinese brand names showed higher theta coherence than English brand names in the frontal and temporal cortical regions. This finding may show either that the main effect of language differences can be modulated by theta coherence, or that Chinese brand names are quickly encoded into memory by theta activity modulation, as discussed earlier (chapter 2, part 2, Memory).

The results of previous studies suggested that beta activation can distinguish if brand names in different languages are processed similarly on the semantic level. Moreover, distinct theta

(21)

activation from processing of Chinese brand names may suggest that Chinese brand names are easier to memorize than English brand names.

Brand Extension

Brand extension is defined as applying existing and well-known brand attributes to new product categories (Keller & Aaker, 1992). Consumers may transfer their established attitudes from earlier experiences with a brand to a new product if the new product is cognitively categorized as a member of the existing brand (Cohen, 1982). Some physiological methods, such as tracking eye movements or reaction time, have been employed to measure consumer responses to brand extension, and a neurophysiological study using an ERP component was performed (Ma, Wang, Shu, & Dai, 2008). An earlier study confirmed the occurrence of prominent P300 activation when pictures in the probe set and their order of presentation were consistent with those in the memory set (Zhang, Wang, Li, & Wang, 2003). In another study, a neurophysiological response similar to P300 showed larger activation when the non-target stimuli shared some perceptual similarity with the target stimuli (Azizian, Freitas, Watson, & Squires, 2006). Therefore, Ma et al (2008) speculated that P300 can be an indicator of categorization processing if consumers extend previously established attributes to a new product category. Fifteen beverage brands were chosen as prime stimuli and twenty product names (including five beverage products) from a familiar product category were selected as target stimuli. Pairs of brand and product names, for example, Coke and juice or Pepsi and telephone, were presented during continuous EEG recording, and P300 was detected in both the beverage and non-beverage conditions. P300 was more positive in the beverage condition than in the non-beverage condition, especially at the right parietal and occipital sites. As in early studies, this study also detected large amounts of P300 in the congruence condition, in which a pair was made of a beverage brand name and a beverage product. In addition, N400, a negative component of ERP, which is sensitive to different semantic categories (Heinze, Muente, & Kutas, 1998), was observed after P300 with non-beverage pairs. Therefore, N400 could be useful as a physiological index to evaluate the fitness of a product for brand extension.

In recent research, therefore, both P300 and N400 have been used as neuronal indexes to measure consumers’ perceptions of the similarity of attributes between existing brands and new product categories.

Product Preference

One study attempted to find selective EEG frequencies and channels that can distinguish product preference (Yilmaz, Korkmaz, Arslan, Güngör, & Asyali, 2014). In this study, 10 females

(22)

and 5 males viewed 16 images of women’s shoes of different styles and colors. All participants were asked to press a button for the particular shoe they liked. This data was later divided into “like” and “dislike” categories. The study concluded that 4 channels (F7, T6, CZ, O1) and 2 frequencies in the theta band (4 and 5 Hz) possessed the discriminative power for determining preference. This could be in line with the results of the aforementioned study (in chapter 2, part 2, Emotion; (Ertl et al., 2013), which confirmed a significant increase in prefrontal theta oscillations when negative emotions decreased. However, a better conclusion may have been drawn if the study had been designed properly. Although the behavioral results for the stimuli, women’s shoes, were similar in both genders, the gender-specific stimuli may have had a cognitive influence; instead, neutral stimuli should have been chosen and the same number of subjects should have been employed in each experiment.

A study using gender-neutral stimuli was conducted in order to identify cortical activity associated with like/dislike conditions using the Emotive EPOC, a high-resolution dry headset mentioned earlier (in chapter 1, EEG device; Khushaba et al., 2013). In this study, participants were required to choose one cracker out of three options that varied in shape (circle, triangle, and square), topping (salt, poppy seed, and plain), and flavor (wheat, dark rye, and plain). During EEG recording, participants indicated their most- and least-liked crackers by clicking a mouse to provide a preference ranking. EEG segments corresponding to the moment of clicking were analyzed to measure changes in spectral power in all frequency bands. Also, the phase locking value (PLV), which identifies in what brain regions activity is taking place and in which EEG bands phase synchronization is most evident, was calculated. The PLV results reflect previous knowledge that active cognitive processes take place with activation of theta, alpha, and beta frequencies in the frontal and occipital regions. As previously mentioned, this has relevance for attention and emotion (see chapter 2). Interestingly, PLV values for the occipital and parietal channels were highest, which implies an interaction between those regions in preference decision-making. Lastly, mutual information about the EEG features and preference was analyzed to identify features of the stimulus that may have magnified or attenuated the EEG power in association with preferences. Similar to the results of the previous study using women’s shoes (Yilmaz et al., 2014), theta band power exhibited the highest level of mutual information over the left occipital region, which is known to be related to visual preferences (Kawasaki & Yamaguchi, 2012). In addition, delta oscillation, which has been identified as a sign of stimulus-elicited activity on the brain’s rewards circuit (Knyazev, 2007), was apparent on the left frontal and right temporal channels. A similar pattern was also detected in the alpha, beta, and gamma power bands. These results indicated a substantial advantageous impact of flavor and topping over shape.

(23)

In conclusion, product preference studies imply the importance of activation of the theta band, which is also a critical EEG band in emotion and memory studies.

Service Recovery Evaluation

In service markets, how consumers respond to service failure and the influence of service recovery on their judgment and level of satisfaction are important. Recent studies in this field have investigated the role of emotions in service failure and service recovery. However, efforts to measure emotional responses in these negative situations by means of self-report surveys or post-interview methods are not satisfying. Therefore, a study using neurophysiological approaches to measure consumers’ emotional responses during a service recovery situation was performed (Boshoff, 2012). This study focused on the influence of the physical appearance of the service provider on customers’ evaluations of service. Traditionally, studies based on social identity theory and similarity–attraction theory insisted that similarities between the physical features of service providers and consumers would influence the latter’s evaluation of service positively due to the importance of personalized social categories (Strauss, Barrick, & Connerley, 2001). In other words, hiring workers similar in physical appearance to customers in terms of characteristics such as ethnic origin and gender would encourage consumers to categorize service providers as in-group members and strengthen the relationship between service provider and consumer. In some studies, researchers measured consumers’ neurophysiological responses while interacting with a service provider of different or the same gender or race by means of EEG, galvanic skin response, and facial muscle movement. Participants watched four scenarios about service failure and recovery situations that were identical aside from changes in the gender and race of the service provider. The results showed no significant changes in EEG signal under the gender condition nor in the condition in which ethnic origin was varied; however, distinct frontal asymmetry was detected when participants were exposed to a scene in which a travel agent was listening to a customer’s complaints. Frontal asymmetry activation during active listening reflects emotional engagement in the situation, in line with the previous discussion (in chapter 2, part 2, Emotion; Aftanas et al., 1998), but it is not clear if the differences in hemispheric activation were caused by the service provider’s active listening or the customer’s complaints. It would be interesting to determine which hemisphere is more dominant during such an experiment. Unfortunately, no cortical changes were definitely associated with different gender or race conditions in this service recovery situation. However, there are several important insights in this study. Its results provide evidence that people prefer not to report about their negative emotions during evaluation of service satisfaction. Lastly, this study demonstrates that neurophysiological

(24)

methods can detect and measure what traditional methods cannot, especially when distinctions are too subtle or issues too sensitive.

Real Cases

In a 2009 Forbes interview, Dean Macko, a manager of brand strategy at Hyundai Motors America, said that the company determined consumer preferences regarding car exteriors prior to manufacturing using EEG. In addition, PepsiCo’s Frito-Lay switched its potato chip packaging to a shiny plastic bag when it was discovered that this kind of packaging resulted in more activity in the area of the brain associated with emotions.

Part 3. Promotion-related Research

Despite the rapid expansion of new media and channels that may replace television, TV still plays a major role in increasing brand awareness for consumers (Rubinson, 2009). Therefore, delivering marketing messages on TV is still important to marketers. Within a limited timeframe, TV commercials must grab viewers’ attention, create a positive impression, and encourage viewers to become loyal customers. In most cases, marketers conduct advertising pretests to determine the effectiveness of TV commercials before their release due to the correlation between the success of the product and the success of the commercial.

Memorability of TV Commercials

Early studies regarding brain activity and TV commercials were initiated to investigate the relationship between EEG data and behavioral memory data. Appel et al. (1979) tried to match brain activity with different types of commercials and found that commercials producing greater brain activation in any hemisphere are more easily recalled. An initial hypothesis assumed that frequently recalled commercials would induce lateralized hemispheric activation; however, the study failed to detect any such activation. Weinstein et al. (1980) tried to compare EEG activation caused by different advertising media such as magazines and television. Increased beta activation in the left hemisphere caused by magazine advertising was observed, but no significant results were found for beta activity and memory. Later, Rust et al. (1985) also analyzed EEG data of four frequency bands – delta, theta, alpha, and beta – to compare different neuronal responses elicited by advertisements in different communication channels including newspapers, magazines, radio, and television. The results revealed that advertisements in newspapers and magazines resulted in greater theta activation, while radio and TV advertisements resulted in higher alpha activation. However, the relation between brain

(25)

activation and recall was inconsistent. Early studies at this point reported inconsistent findings and only shared partial results.

Rothschild et al. (1986) conducted research that extended the previous findings of Appel et al. (1979) confirming variations in the alpha frequency band during processing of TV commercials. As mentioned previously (in chapter 2, part 2, Attention and Memory), alpha is a critical frequency band that modulates attention and affects memory. Thus, Rothschild et al. (1986) focused on detecting “alpha blocking”, in which an association was observed between alpha activity reduction and attention arousal at the onset of a stimulus, and “alpha attenuation”, which means slow recovery of alpha activity until a new stimulus appears. The results showed an association between highly recalled commercials and low alpha activation overall. However, the opposite occurred with forgotten advertisements. This negative correlation between alpha band activation and remembered commercials reflects the fact that memorability has a relationship with attention that can be traced by alpha activation. In a later study using short-period EEG data for analysis, remembered commercials elicited alpha blocking with left hemisphere dominance, whereas forgotten commercials were associated with an attenuated alpha power with right hemisphere dominance (Rothschild & Hyun, 1990). Therefore, early commercial studies focused on identifying the role of alpha activation in the memory process.

Modern EEG studies using the high-resolution technique and sophisticated analysis have extended the scope of previous studies. Astolfi et al. (2008) hypothesized that the differences between remembered TV commercials and forgotten TV commercials may be evident not only in the alpha band, but also in bands of other frequencies (Astolfi et al., 2008). The results indicated theta and alpha activations in the prefrontal right area, the anterior cingulate area, and the right and left parietal areas and gamma and beta activations in the frontal and parietal areas in a comparison between the remembered (RMB) and forgotten (FRG) datasets. These results are in line with the finding of bidirectional frontoparietal EEG coherence in the memory study cited in chapter 2, part 2, Memory. No specific frequency band that was sensitive to memorability was detected in this study, but it confirmed statistical differences between the RMB and FRG conditions and higher spectral amplitude in the RMB condition than in the FRG condition. Moreover, unlike early studies that only considered occipital areas, the study of Astolfi et al. (2008) detected activations in various cortices, especially the parietal cortex, which is known to process complex visual stimuli (Rothschild, 1988) and play an important role in information storing (Ioannides et al., 2000). The results also support the notion that prefrontal and parietal areas interact during high-level cognitive processing and suggest that these regions may be related to transfer of short-term memory to long-term memory storage. Lastly, this

(26)

study shed light on the role of the right cortices, which had been underestimated in memory processing (Knutson, Rick, Wimmer, Prelec, & Loewenstein, 2007).

In addition to localization of cortical activations during successful memory encoding, recent studies applied a functional connectivity approach to improve our understanding of the organized behaviors of different cerebral regions (Horwitz, 2003) in memory. Mathematical tools used in graph theory have been implemented to investigate brain networking and extraction of global and local network structures. In graph theory, the efficiency of the functional network decreases if the organization of the network relates to higher cognitive functions. The opposite would be true for low cognitive function. A study was conducted evaluating different cortical networks estimated according to differences in the global and local properties between remembered (RMB) and forgotten (FRG) commercials (De Vico Fallani et al., 2008). The results showed that in the RMB condition, which may require higher cognitive processing, decreased global and local efficiency was evident in the estimated cortical networks. Specifically, major differences appeared in the beta and gamma bands for the global efficiency index and the alpha band for the local efficiency indexes. In contrast, forgotten commercials were associated with increased global and local efficiency, which indicates a low level of neural engagement. As we saw in the study of Astolfi et al. (2008), the relation between the alpha, beta, and gamma bands in memory formation already found but this study provided further information about the connectivity between the estimated cortical areas.

Emotion in Commercials

Preceding studies suggested that the prefrontal and frontal cortex (PFC and FC) are involved in generating emotions, and that the left and right hemispheres are differently lateralized for approach to and withdrawal from emotions (Davidson, 2004). Emotions that viewers experience during the observation of a commercial could therefore influence their attention and memory, which may in turn influence the effectiveness of the commercial.

An experiment to measure the likability of commercials revealed significant increases in activity in the theta and gamma bands for “like” commercials, whereas higher beta band activation was detected with “dislike” commercials (Vecchiato et al., 2010). This study emphasized the role of theta activity over the left frontal areas in TV commercial testing. A year later, Vecchiato et al. (2011) investigated the modulation of the power spectral density (PSD) of EEG activations in the FC and PFC during the observation of emotion-inducing TV advertisements (Vecchiato et al., 2011). They assumed that EEG frontal activity could differ in reaction to pleasant or unpleasant advertisements because of lateralized differences between hemispheres for different emotional experiences. The results indicated a greater association of

(27)

the left frontal hemisphere with pleasant emotions, whereas right frontal hemisphere is related to unpleasant emotion. This finding was supported by scores for frontal asymmetry that indicated differences between the like and dislike conditions for both the theta and alpha bands. Particularly noteworthy is the negative correlation between an imbalance in theta activity between hemispheres and the pleasantness score, whereas the alpha index was positively correlated with this score. This finding revealed that the spectral index is able to gauge the degree of pleasantness induced by TV commercials. The result is in line with those of frontal asymmetry studies previously mentioned in chapter 2, part 2, Emotion.

Likability studies match EEG data to likability scores obtained from viewers during interviews. The design of such studies violates the basic assumption that conscious responses are not valid or representative of real responses. However, the studies described herein were able to replicate the results of emotion studies and even provide further information about the correlations between the theta and alpha bands.

Real Cases

In a 2009 Forbes interview, an advertisement for Cheetos, one of the famous snacks of Frito-Lay, included a prank that was evaluated negatively by female participants in focus group interviews. However, EEG testing revealed that they actually liked it, but did not answer honestly. Therefore, the snack food marketers at Frito-Lay decided to air the commercials regardless of the results of the focus group interviews. Yahoo also pretested a commercial using a neuroscience technique before launching and decided to air a commercial that activates brain areas related exclusively to memory. eBay wanted to persuade e-shoppers to use its online payment service more often. Finally, brain research convinced marketers at eBay that the word “speed” turns users on more than the words “safety” and “security”; therefore, “speed” became the main theme in eBay ad campaigns.

Conclusion

This text has explored consumer neuroscience research involving EEG from cognitive psychological and neuroscientific perspectives. In chapter 1, basic features of EEG were summarized. EEG is a noninvasive, economical method with good time resolution which is more suitable for business purposes than any other brain-imaging tools. EEG studies often focus on finding a certain frequency band or identifying ERP components depending on the purpose of

Electroencephalography in Consumer Research A cognitive psychological and neurological review of consumer research using EEG