Neuromarketing research in the last five years: a bibliometric analysis

Abstract Neuromarketing (NM) is an application of neuroimaging and physiological tools to record the neural correlates of consumers’ behaviour (e.g., decision-making, emotion, attention, and memory) toward marketing stimuli such as brands and advertisements. This study aims to present the current tools employed in the empirical research in the last five years. In this article, we have followed the Preferred Reporting Items for Systematic Review and Meta-analyses (PRISMA) framework and a bibliometric analysis to select empirical and review papers that used NM tools in the last five years. We have extracted and analysed twenty-four documents from the Scopus database to answer our study questions. We found that electroencephalography (EEG) is the most popular neuroimaging tool in neuromarketing research, wherein has been used almost thirteen times. Followed by eye-tracking (ET) and galvanic skin response (GSR) as the most physiological tools, wherein have been applied almost four times for each tool. We hope that this study provides valuable insights into the common NM tools used in marketing research.


ABOUT THE AUTHORS
Ahmed H. Alsharif is a PhD student of Azman Hashim International Business School (AHIBS) at Universiti Teknologi Malaysia, Malaysia. His main research interests are neuromarketing, consumer behaviour, advertising, emotional and cognitive processes. He has published several research papers in the field of neuromarketing and consumer behaviour.
Nor Zafir Md Salleh is a Senior Lecturer of Marketing at AHIBS, Universiti Teknologi Malaysia. Her current research interests include culture and heritage, service innovation, Islamic tourism, marketing.
Rohaizat Baharun is a professor of marketing at AHIBS, Universiti Teknologi Malaysia. He became a visiting professor for local and international universities. His current research interests include educational marketing, retailing management, and customer behaviour.
Alharthi Rami Hashem E is a PhD student at AHIBS, Universiti Teknologi Malaysia, Malaysia. His research interests are product placement, brand placement and integration, brand attitudes, implicit association test (IAT), persuasion knowledge.

PUBLIC INTEREST STATEMENT
The rising interest in exploring the neural and physiological reactions of consumers toward marketing stimuli such as advertisements and brands paid scholars and practitioners to employ physiological (e.g., ET, EMG, GSR, ECG) and neurophysiological (e.g., fMRI, EEG, ERP) tools in marketing research to explore, understand, analyze, explain, and predict the consumer's behaviour such as decision-making, the emotional (e.g., emotion) and cognitive (e.g., attention and memory) processes toward marketing stimuli. This study provides the global research trend in the neuromarketing field according to the outputs of publications such as the most productive countries, academic institutions, journals, authors, number of publications and citations.

Introduction
Techniques and methodologies are highly important in the hyper-competitive environment, this has prompted companies to seek new effective methods to better understanding and predicting consumer behaviour. Thus, scholars have investigated how marketing research can benefit from these techniques to develop marketing practices and advertising research. Wherein, the fMRI investigation showed that most of the consumers' behaviour and decision-making are made beyond awareness that highly contributed to the purchase decisions (Agarwal & Dutta, 2015;Alsharif, Salleh, Baharun, & Effandi, 2021;Brierley, 2017). That study was a threshold for a new approach of studying consumer behaviour by using neuroscience technology in business research, which has called "neuromarketing". NM is located on the borderline between marketing, neuroscience, and psychology (Alsharif et al., 2021a;Alvino et al., 2020).
Traditional research methods were largely used self-report (e.g., surveys, focus groups) to study the responses of consumers (e.g., decision-making) toward marketing stimuli such as advertisements and brands (J. Harris et al., 2018). Self-report relies on the consciousness of the consumer behaviour and overlooks the unconscious; thereby, mismatching between what consumer says and does. Therefore, traditional research methods provide inaccurate and unreliable information about consumer's behaviour (Alsharif et al., 2021b;Alsharif, Salleh, Baharun, & Effandi, 2021), which has led to infer that most products and ads have failed in the first year (Jordao et al., 2017;Vecchiato et al., 2015). Accordingly, the majority of consumer's behaviour (e.g., decision-making, perception) occurs unconsciously, which is not possible for predicting by traditional research methods (Alsharif, Salleh, Baharun, Hashem, et al., 2021;Alsharif, Salleh, Baharun, & Effandi, 2021).
The main goal of NM is a better understanding of the neural correlates of emotion, attention, memory, and decision-making in advertising campaigns (Alsharif, Salleh, Baharun, Hashem E et al., 2021;Alvino, 2019).
Therefore, NM has enormously received attention from an academic and industrial environment to employ these tools in consumer research to study concealed consumer reactions such as emotional and cognitive responses toward marketing stimuli (Hamelin et al., 2017). From an academic perspective, the publications number of NM has been remarkably increased since the last decade . From an industrial perspective, the number of NM companies (e.g., Millward, Emsense) has been increased (Rawnaque et al., 2020). Consequently, NM research has attracted companies and scholars to use NM tools in their research to overcome the limitations/restrictions of the orthodox methods and solving the marketing issues (Alsharif, Salleh, Baharun, & Safaei, 2020;Morin, 2011;Sebastian, 2014). Consequently, the contribution of neuroscientific methods became significant to enrich research about consumer behaviour in a new millennium of marketing that led several authors to speak about the benefits of NM (Lee et al., 2017;Ramsoy, 2015;Songsamoe et al., 2019).
Several studies provide an overview of the most popular NM techniques which can be used in marketing research, for example, fMRI. However, there is a lack of investigation in NM studies that were employed physiological and neuroimaging techniques in the last five years to guide scholars and practitioners. Hence, the current study presents the current trend of NM research and the most NM tools (e.g., physiological and neurophysiological) used in marketing research in the last five years. We reviewed the literature to address the following research questions: Answering these questions will give us a comprehensive knowledge of the current tools and studies employed in empirical NM research. To this end, this study seeks to combine as many directions as possible, besides, common research subjects are investigated deeply based on subdomains to accomplish a precise, concrete, and concise conclusion. The key contributions of the current study are summarized as follow.
• The first study has reviewed, analysed, and discussed the latest studies that used neuromarketing tools in empirical research between 2015 and 2019.
• The first bibliometric analysis that shed light on the most trend global contributions in the neuromarketing field from 2004 to 2019.
In this vein, this study paper provides the methodology and data collection techniques in section 2. Section 3, presents the analysis of bibliometric, the studies used physiological and neurophysiological tools used in marketing research. Sections 4 and 5 present the discussion and conclusion of the study.

Methods
To answer the RQ1, RQ2, bibliometric analysis has been used in the current study to understand the global trends in NM, a bibliometric analysis provides the global trends such as the most productive journals, authors, countries and academics/institutions, and the most-cited articles in relevant topics based on the outputs of the publication from 2004 to 2019. To answer the RQ3, we have followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework recommended by Moher, et al. (Moher et al., 2015) to select empirical articles that used neurophysiological and physiological tools in marketing research in the last five years. The current study is characterized by extracting articles from the Scopus database relevant to our study. However, we also followed instructions recommended by Block and Fisch (Block & Fisch, 2020) to present an impactful bibliometric analysis and assess the structure of a particular research area. This process will help us to better understand the advancement of the NM field. We first have used the following query "neuromarketing" to select relevant articles from the Scopus database. Figure  1 shows the accumulative and annual publications in the NM domain from 2004 to 2019. We found 479 publications based on abstract, title, and keywords.
As depicted in Figure 2, we followed the PRISMA framework, which includes four stages: (i) identification as recording identified through database searching, (ii) screening the record documents, (iii) eligibility documents, and (iv) selecting studies, as follows: • Empirical articles published between 2015 and 2019 were included.
• We excluded any other documents (e.g., conference papers, reviews) were excluded.
• We excluded any publications in the non-English language.

Figure 2. PRISMA Flow chart.
A total of thirty-four documents were extracted and after reviewed documents one by one, ten documents were excluded; therefore, a total of twenty-four documents were selected for this study based on the article questions. According to literature, the first article in NM was carried out in 2003and published in 2004by McClure, et al. (McClure et al., 2004. In addition, we have classified the selected documents based on the tool that used in the study, which includes three categories as tabulated in Table 1.  fMRI. Functional magnetic resonance imaging; EEG. Electroencephalography; ERP. Event-related potential; Hr. Heart rate; GSR. Galvanic skin response; ET. Eye tracking, IAT. Implicit Association Test; EMG. Electromyography. Hence, this review includes empirical articles that used physiological and neurophysiological tools to investigate the consumers' reactions toward marketing stimuli (e.g., advertising, branding, product) within a marketing context. We will read, analyse, and summarized these articles in Table  8 to accomplish the purpose of this review paper.

Leading countries and academic institutions
We have reviewed and analyzed the twenty-four documents from the Scopus database, as tabulated in Table 2. Table 2 demonstrates that almost 50% of the total papers were published by the USA, Spain, Italy, UK, Germany, which lead to infer that these countries are key players in the improvement of NM research. As we can observe from Table 1, the USA is the top producing country in the NM field with almost seventy-seven papers, four of these papers published by Emory University. Spain is considered the second most productive country with forty-eight papers. Followed by Italy, UK, Germany with thirty-eight, thirty-three, thirty papers, respectively. China and Japan were placed in the 6th, 7th, and 8th position in the list of the ten most productive countries with twenty-six, twentyone, and twenty-one papers, consecutively. Finally, countries that published less than twenty papers such as Malaysia and Turkey with nineteen and seventeen papers.
Although Italy is in the third position among the most ten productive countries around the world, its Universita Degli Studi di Roma La Sapienza is considered as the most productive academic institution among other countries institutions with twenty-four papers. Followed by Bucharest University of Economic Studies with twelve papers, then Zhejiang University in China with nine papers. Seven and six papers were published by Universidade da Coruña and the University of Oxford. Emory University, Zeppelin University, Toyama Prefectural University, University Sains Malaysia were contributed with four articles for each one. Finally, Galatasaray Üniversitesi has published three papers.

Leading authors
We have found that the ten most productive authors in the NM topic were affiliated with three countries (e.g., Italy, UK, and China) as shown in Table 3. The total number of publications by these authors was 107 documents. In addition, the four most productive authors in the list are Babiloni, F., Vecchiato, G., Cherubino, P., and Astolfi, L. from Italy with a total of 22, 17, 10, and 9 documents, 508, 383, 147, and 297 citations, respectively. Lee, N. from the UK, is the most fifth productive author with a total of 9 documents since 2007, 473 citations, and seven h-index. Interestingly, Toppi, J. is the least productive author within the top ten list, with 226 citations and eight h-index.

Journal outcomes
The findings indicate that seven publisher agencies owned the top ten most productive journals, as tabulated in Table 4. Springer Nature has published three documents of them. Followed by Frontiers Media S.A which published two documents. The rest were published by APA, Cogent OA, Elsevier, Emerald, and the Advertising Research Foundation. Moreover, the most productive journal was Advances In Intelligent System and Computing with 9 documents. Frontier in Neuroscience has published eight documents with 128 citations. Journal Of Economic Psychology has published the most cited document among the top ten list with 250 citations, while Communications In Computer And Information Science journal was published the least cited document with 6 citations. The answer to the RQ1, the most-cited journal is the Journal Of

Keywords analysis
The keywords analysis is an insightful quantitative device to investigate specific subjects in marketing according to the presumption that keywords provide a coherent explanation to the content of the documents which has been used a lot recently (Alsharif, Salleh, Baharun, Hashem, et al., 2021;Wang & Chai, 2018). Bibliometrics is following four analysis methods such as co-word association, co-word cluster, co-word frequency, and burst word monitoring. Importantly, co-word analysis by examining the co-occurrence keywords. Wherein the connection between two keywords is presented by a numerical value which shows the relationship between both of them, and the higher this value, the link strength (Goyal & Kumar, 2021;Ravikumar et al., 2015;Saha et al., 2020). The link strength between two keywords represents the number of times where these keywords occurred manifested in the same article. The total number of these links refers to the total number of these two keywords that occur together. In VOSviewer, five were set as the minimum occurrences of a keyword to be presented, which means that keywords will appear on the bibliometric map one-two keywords occur together in a document more than five times.
The author keywords co-occurrence analysis carried out in this study involved 1073 keywords from 479 articles in 311 journals. However, a large number of these keywords have been used once. About 1073 keywords were used once, 220 keywords were used twice, and 101 keywords were used triple. The number of total keywords decreased to 1000 by re-labelling them. The keywords were inserted in VOSviewer to map the literature with a minimum of five occurrences, and only 43 keywords met the thesaurus, as depicted in Figure 3. Scholars usually employ cooccurrence analysis as it is an effective method to address the research trends on a particular topic by exploring existing documents Khudzari et al., 2018), including the marketing field (Wang & Chai, 2018). Table 5 illustrated the summary of the most occurrences from the selected keyword search. in the Scopus database and the total link strength of the linked keywords to the keyword search. In addition, it was expected to have a strong association with devices aspect such as "EEG" which is second most topics examined concerning NM (46 occurrences, 97 total link strength). Further, neuroscience and consumer neuroscience were observed to be highly associated with NM with 39 occurrences, 97 total link strength, and 31 occurrences, 86 total link strength, respectively. The fMRI device is also had a strong association with NM (25 occurrences, 60 total link strength).

Leading citations
We used citation analysis to identify the most popular articles within the NM environment. It has been suggested that citation analysis counts the number of times other articles cite a particular article to identify the popularity and impact of the article in the scientific field (Baker et al., 2020). Based on the "total times cited count" provided by the Scopus database, we have analyzed the citation of 479 documents. The findings indicate that the documents that have more than 200 citations were published by three different publishers such as Nature Review Neuroscience, Elsevier, and Wiley in five different journals, as tabulated in Table 6. The most cited document was published by Ariely, Dan in 2010, with 388 citations. While the least cited document in the list was published in 2012 by Falk, with 151 citations.

Co-citation network analysis and data clustering
To address the intellectual development of the scope, it has adopted the co-citation network analysis. Consequently, several clusters were identified to conduct the content analysis. To study the structure of the subject "neuromarketing", we use the VOSviewer program to analyze the co-citation network. The initial findings result is 19 documents with at least 80 citations. To visualize the map of the co-citation, VOSviewer formed a random cluster map that was too complex to understand. Therefore, we identify the ten top documents based on recommended methodology by Kumar and Ranjani (Kumar & Ranjani, 2018), who identified the leading 10 papers from each cluster. In addition, we used the weighted citation count provided by VOSviewer to ensure a high-quality analysis of documents. As shown in Figure 4, the analysis results in four clusters with a high correlation between them. Among the four clusters, the red group is the largest that is led by Morin, C., (2011). Followed by the green cluster with Ariely, D. and Berns, G.S., (2010) as the most dominant study and the most cited author in NM literature. The 3rd blue cluster with Reimann et al., (2010). Finally, the yellow cluster is dominated by Nick, lee et al. (2017). It should be noted that despite that the fact that these clusters address different aspects of NM, they are highly interrelated and complement each other.

Discussion
This review provides an overview of the studies that employed NM tools in empirical research between 2015 and 2019 to answer research questions (RQ.1, RQ.2, and RQ.3). In addition, it has been used bibliometric analysis to answer RQ.1 and RQ.2 and know the global trends in NM topics based on the outputs of the Scopus database publications from 2004 to 2019. The study revealed that there is a clear interest in NM tools by both academia and industrial environments, which included four main categories: (i) neuroimaging tools, (ii) physiological tools, (iii) self-report tools, and (iv) behavioural measurements tools. According to the literature, there are concerns about the operational restrictions and the implementation cost of neuroscience tools (fMRI), besides, ethical issues such as invasive of consumer's privacy and confidentiality, and autonomy to evaluate consumer's perceptions, preferences and decision-making and the difference between initial estimates and future estimates for neuroscience tools.
We have found that the most generally co-occurring keywords were EEG, fMRI, ERP, and ET, besides, were the closest words from NM as shown in Figure 3. We have reviewed and analyzed the selected articles and we found that the EEG is the most neurophysiological tool that used in articles (approximately thirteen times), followed by the ERP used almost seven times, then the fMRI applied nearly five times. The GSR and ET are the most physiological tools that were used in NM research (almost 4 times for each tool).
The highest number of publications was seven articles in 2018, the total number of citations was 51 times. Additionally, most of the productive countries based on selected articles were China, Malaysia, South Korea, Lithuania, Germany, Brazil, UK, Ukraine, Japan, and Denmark. Based on the authors, among the top 15 authors, South Korea and Malaysia with 3 authors for each country. Lithuania, China, Ukraine with 2 authors per country, and Japan, Germany, and Italy with one author for each country. Chew, et al. (Chew et al., 2016) had the most cited article, with 26 citations. This research has provided an overview of the empirical articles that used NM tools in the last five years.

Selected article analysis
After identifying and evaluating the papers published on the Scopus database. We found twenty eligible articles for this study. Figure 4 shows the network of collaboration between the twelve productive countries in the NM field. Table 7 shows that the cluster of countries which were produced empirical articles in the last five years. We can observe that Malaysia and Germany countries have the largest total link strength, followed by Brazil with 302 total link strength. While Ukraine has the lowest total link strength with only two links. For avg. norm. citations, Italy has the highest avg. norma. citations with 3.43, followed by USA and Denmark with 2.33 avg. norm. citations for each country. At lowest avg. norm. Citations were given to Brazil. Table 8 shows the analysis and summary of the selected articles for this review study.

Implication of the research findings for theory and practice
Theoretically, the current findings can be divided into three folds, as follow: Firstly, physiological tools such as ET, GSR, EMG, and ECG can provide beneficial insights about the physiological correlate of consumer behaviour, while neuroimaging tools such as fMRI and EEG enables to capture the neural correlates of emotion processes(e.g., pleasure, motivations, and arousal), cognitive processes (e.g., attention, recall, recognition, and memory) toward marketing stimuli such as advertisements and brands. Secondly, marketers and advertisers will be able to identify the attractiveness and aversion aspects in advertising campaigns before using them in the real world; therefore, recognize the aspects that cause aversion attitude toward advertising and fix them, additionally, reinforce the strengths aspects that drive the consumer to pleasure and making decisions. Third, the majority of the studies focused on detecting the impact of advertising campaigns on consumers' behaviour; therefore, the ability to predict consumer behaviour. In addition, there are some studies focused on spokesperson's gender, ads appeal, warm colour temperature and cool colour temperature, celebrity face, social initiatives (i.e., anti-smoking), and public health in advertising campaigns. Therefore, these three folds together can explain the neural and physiological correlates of motivation, reward, and perception processes of interest for advertising. An application of this research might offer measurable explanations of how advertising works in consumers' minds which is reflected in the bodily responses of consumers, thereby, creating attractive advertising in several domains such as political, social and business.

Conclusion and future work
NM plays a vital role whether in social advertisements or commercial advertisements. This study was conducted to concentrate on studies that employed the NM tools in practical research from 2015 to 2019, wherein authors found that the EEG is more common in experiments in the NM field because it has excellent temporal accuracy and less cost. It also used a bibliometric analysis to understand the global research trends in the NM field based on the outputs of academic publications such as the most-cited article and journal. Along with a literature review to evaluate and analyze the previous studies that employed NM tools in empirical research such as neuroimaging, physiological, and behavioural tools. It was analyzed 24 empirical articles from the Scopus database and it was noticed that NM research has increased globally since 2004. The authors identified, reviewed, and analyzed the relevant articles to know how these studies used the most common words. Then these studies were summarized and their major contribution and findings were detailed in Table 8.  Andrii, et al. (Andrii et al., 2019) Attractiveness modeling of retail on emotional fatigue of consumers Trying to study emotional fatigue and its impact on choosing the retailer and the purchase decision process.

GSR
The atmosphere in the store and the consumer's state impact the emotional fatigue differently; thereby, the footfall. 15 different retailers shops Goto, et al. (Goto et al., 2019) Can brain waves really tell if a product will be purchased? Inferring consumer preferences from single-item brain potentials Examining if single-item ERPs could predict the consumer's preference for a specific product accurately and reliability.

EEG/ERP
The accuracy of prediction toward consumer's preference to specific products is almost 70.8% by using SI-ERPs, and also the accuracy vary based on ERPs type. LPP and PSW were better than N200 in predicting consumer's preferences. In addition, group-related ERPs could differentiate between high preferred and less preferred significantly.

Alcoholic beverages
Ma, et al. (Ma et al., 2019) The influence of the consumer ethnocentrism and cultural familiarity on brand preference: Evidence of event-related potential Exploring the influence of ethnic affiliation (e.g., Black African people and Chinese people) toward brand preference.

EEG/ERP
The logo of the brand has highly influenced the Chinese participants' preference toward brands.
Brand logo preference  The effects of money on fake rating behavior in e-commerce: Electrophysiological time course evidence from consumers Exploring whether RN or RI strategy is more strongly impact the online fake rating behavior of the consumers.
EEG/ERP RI strategy has more impact on the online fake comments behavior of the consumer.
Online fake comments behavior Wei, et al. (Wei, Wu, Wang, Supratak, Wang, Guo et al., 2018) Using support vector machine on EEG for advertisement impact assessment evaluating advertisement influence and the potential of consumers buying the advertised product.

EEG and Questionnaire
The percentage of purchasing power was relatively high after watching the advertising.
Purchasing intention (Continued) Alsharif et al., Cogent Business & Management (2021) Emotional fatigue Ramsoy, et al. (Ramsoy et al., 2018) Frontal brain asymmetry and willingness to pay Exploring and examining the potential role of the prefrontal asymmetry in calculating willingness to pay.

EEG
Illustrating that asymmetry in the prefrontal cortex in the gamma and beta band was highly related to subsequent WTP responses, and the gamma band was most strongly related to WTP decision.

WTP
Jung, et al. (Jung et al., 2018) The neural correlates of celebrity power on product favorableness: An fMRI study Investigating the neural correlates of celebrity power in influencing the car favorableness in advertising.   Pilelienė and Grigaliūnaitė (Pilelienė & Grigaliūnaitė, 2017) Relationship between spokesperson's gender and advertising color temperature in a framework of advertising effectiveness  Determining the impact of advertising appeal on print/ outdoor advertising effectiveness; thereby, developing the model of the influence of advertising appeal on advertising effectiveness.

Eye-tracking (Tobii Eye
Tracking Glasses), IAT, and Questionnaire Firstly, in the cognitive stage to consumer's response toward print/ outdoor advertising can achieve effectiveness by applying either the emotional or rational appeal. Secondly, in the affective stage, the application of emotional advertising appeal is more effective. Lastly, in the conative stage, the application of rational advertising appeal is more effective.

Ads appeal
Lewinski (Lewinski, 2015) Don't look blank, happy,  fMRI, EEG, ET, IAT, Selfreport, Biometrics (heart rate, breathing, SCR/EDR) The fMRI measures allows to explain the variance in TV advertising much better than traditional measures.

TV Advertising
Cascio, et al. (Cascio et al., 2015) Neural correlates of susceptibility to group Combining between machine learning tools with fMRI data to describe the relationship between brand personality and brain activity.
fMRI It is possible to predict the brand that person is thinking of it based on the correlation between brand personality and brain activity. Nevertheless, this study can be useful for future research trends as it gives previous studies of the research subject/problem. The authors would recommend that future research on NM has to use the fMRI tool to evaluate and predict the purchase processes and decision-making by determining the activity regions in the consumer's brain; thereby, predicting the success or failure of the product or advertisements before put it in the real environment due to that it has an excellent spatial resolution, besides, can record the distal regions in the brain accurately unlike EEG. In addition, we suggest that scholars should focus on the contributions of NM research in other domains such as social sciences, politics.

Limitations of the study
We tried to minimize the shortcoming in methodology, this study comes with a limitation that offers opportunities for future research. We focus on extracting the publications from the Scopus database that were published in the last five years. In addition, we extracted empirical papers that were published in the English language and employed NM tools and overlooked the books, chapter books, proceeding books, notes, reviews, conferences and the non-English language publications, thereby the current study is not fully bias-free.