Predicting the recall of TV-commercials using inter-subject correlation calculated from EEG data

Predicting the recall of TV-commercials using

inter-subject correlation calculated from EEG data

Leidelmeijer, M.

, Diepen, van, R.

Department of Research, Alpha.One, Rotterdam, the Netherlands

2

Faculty of Natural sciences, Mathematics and Informatics, University of Amsterdam, Amsterdam, the Netherlands

E-mail: marijnleidelmeijer@hotmail.com Received 27/03/2020

Accepted for publication xxxxxx Published xxxxxx

Abstract

TV-commercials are one of the most important sources for companies to create brand awareness and could induce enormous financial benefits. Making an effective TV-commercial however, is difficult and therefore increasingly more brain research is carried out in this discipline. Current research is executed to test if the remembering of TV-commercials is correlated with brain activity measured while watching these commercials and if EEG data is predictive of the remembering of them. This is tested by showing 12 commercial TV-advertisements to 33 participants while measuring their brain activity through Electroencephalography (EEG). The EEG signals of all participants are then combined and inter-subject correlation (ISC) was calculated. This is an approach to analyze EEG data of naturalistic stimuli such as videos and calculates the correlation coefficients between the EEG time series of the participants in the corresponding brain locations. Subsequently the same commercials were shown to 25 other subjects and to them a survey was given to find out about the remembering of the commercials one week later. The percentage of free recall and recognition of each commercial was then calculated and the obtained EEG data were processed to find consistency in brain activity between subjects. A significant correlation was found between ISC and the remembering of commercials, in the free recall condition, after a correction was performed for the length of the commercial. These findings suggest that EEG data could predict the free recall of TV-commercials.

Keywords: Advertising, Effectiveness, Electroencephalography (EEG), Inter-subject correlation (ISC), Memory, Neuromarketing, Prediction, Recall, Recognition, Remembering, TV-commercials

1. Introduction

Advertisements have been used for marketing communication since the 16th century and with the arrival of television it didn’t took long before TV advertisements were used for the first time: in 1941 (O’Barr, 2010). Since then, commercial television broadcasting has grown enormously and now, most privately owned television networks are financed by means of advertising revenue. On the other hand, companies create TV-commercials to inform people about their product and to create brand awareness and the financial benefits for brands and organizations due to this form of advertising are immense (Khakhubia, 2018). Conventional marketing methods have been constantly used in the past decades to determine consumer behavior. Making an effective TV-commercial however, is difficult and therefore

neuromarketing made an entrance. In neuromarketing, neuropsychology is applied to marketing research, using brain data to understand how consumers make purchasing decisions (Kumar, 2015). Brain data is believed to be more reliable compared to conventional methods such as surveys because in this case there is no conflict between the subconscious and conscious parts of the brain: people lack to fully describe their own preferences and with traditional methods the true preferences will therefore remain hidden (Nisbett and Wilson, 1977). In addition the neuroscience methods are faster and thus also cheaper, since predictions can be made with smaller samples (Boksem, 2015). This means that for companies it could be very valuable to use brain activity measurements. Moreover, the performance effect of the advertisements could already be predicted before airing them, instead of evaluating them afterwards.

2

Electroencephalography (EEG) is one of the techniques that is used to monitor activity in the brain. This technique uses electrodes that are placed on the scalp with the help of an EEG cap, to record human brain activity. It records voltage differences between two or more points with the use two different electrodes. The difference in electrical potential between one active exploring site and one reference is measured and generates observable EEG waveforms (Britton et al. 2016). Besides functional Magnetic Resonance Imaging (fMRI) and eye tracking, EEG is one of the most important measurement methods in neuromarketing research.

In neuromarketing research it is important to not only look at the neural activity during the showing of commercials, but also to link these measurements to behavior and then translate the individual measurements of the subjects to population level. The latter is tested before by Berns and Moore (2011) with the use of an fMRI study. They used 15 seconds clips of songs of unknown artists and obtained fMRI data of participants about their response to music. This response was compared to the amount of sales of the songs (which showed the popularity at population level) and they found that activity in reward-related regions such as the Orbitofrontal cortex and Ventral Striatum (Nucleus Accumbens) was predictive of future purchasing. Non-hits were associated with activity in both systems and hits were associated with Nucleus Accumbens activity only. To investigate if the same predictions could be made for naturalistic stimuli such as movies and television, Dmochowski et al. (2013) showed popular television content to a group of individuals and recorded their neural activity and found that this could predict the behavior of larger groups. They found that naturalistic stimuli evoke reliable activity in the brain with both EEG and fMRI. They recorded the neural activity of a group of individuals and found that it could predict the behavior of larger groups. They compared scene-by-scene neural activity of 16 participants to Twitter tweet frequency that related to the same scenes of the television content. To examine the neural activity, Dmochowski used a measuring method named inter-subject correlation (ISC). This is an approach to analyze EEG data of naturalistic stimuli and calculates the correlation coefficients between the EEG time series of the participants in the corresponding brain locations. Briefly, ISC calculates the overall level of synchrony of stimulus-driven responses between subjects at all brain regions (Hasson et al., 2004; Madsen et al. 2018) This method has been linked to attentional modulation and engagement (Dmochowski et al. 2012; Poulsen et al. 2016). If participants are engaged with the given content, the neural responses are reliable and correlated between participants, however if not, no reliable neural response arises (Kelly et al. 2016; Ki et al. 2016). ISC increases during movie parts containing high arousal, but it is strongest during naturalistic and familiar situations. Therefore, this method is especially effective to analyze context-dependent, naturalistic stimulus conditions, such as during watching movies and commercials (Kauppi et al., 2010). Cohen et al. (2018) have done more research into this

phenomenon and found that ISC of EEG data that was evoked by educational videos could explain neural engagement of students. They showed the videos to two groups of students and found that more information was retained if students had higher neural engagement and that EEG could therefore be predictive of academic performance in relation to the videos. This information is useful because attention and engagement play a key role in formation of both unconscious and conscious memories (Chun et al. 2007) and is necessary for encoding and retrieval of information (Muzzio et al. 2009).

In this research the EEG technique is used to investigate TV-commercials of varied brands, since TV-TV-commercials are a different kind of stimuli compared to music, movie trailers and educational videos. The research involves investigating whether the EEG measurements could help to predict the remembering of these TV-commercials. This has never been tested before and could eventually help companies to convey the message from their commercials as effectively as possible. Additionally, neuroscience could help to discover why people behave in a certain way since self-report is less reliable and thus less significant (Boksem, 2015). Instead, EEG measures the brain activation instantly and unconscious rather than afterwards and conscious. Remembering can be separated into two different forms of memory retrieval. The first form is recognition and memories retrieved that way are based on associations. The more external cues are given, the easier the brain creates a relationship between this familiar stimulus and the original stimulus that is encrypted in the brain. The second form is recall and memories retrieved this way are obtained without any external cues. The latter form of memory retrieval is more difficult since it involves fewer cues (Hall et al. 1976; Hollingworth 1913). Moreover, for brands and organizations it is important that their TV-commercial will be both well remembered and engaging, since this is related to the chance that people will want to purchase their product (Boksem, 2015). For this reason both above stated forms of memory retrieval are included in this research as well as the exposure time. Goldstein and McAfee (2011) found that the exposure time of display advertisements as well has an effect on the remembering of them. They found that there is a strong increase in recognition and recall between 0-60 seconds exposure time, followed by a continuous, but weaker increase after 60 seconds. These findings suggest the same result for the duration of TV-advertisements in combination with the remembering of them and the data should therefore be corrected for the length of the commercials if a significant relation will be found between these two parameters. The findings about memory performance and engagement while watching TV-commercials, in combination with the earlier mentioned findings that suggest that brain activity could predict population-wide success of naturalistic stimuli, brings us to the hypothesis of current research. It is hypothesized that there will be a positive relation between the length of the commercials and remembering of them as well as a significant relation between the ISC and the remembering. It

is also suspected that the level of inter-subject correlation during the watching of commercials correlates with the level of remembering (both after recall and after recognition), which can be explained due to attentional engagement. Altogether, brings us to the following research question: Is it possible to predict the remembering of TV-commercials with data of neural inter-subject correlation using EEG? The goal of current research is to sort TV-commercials based on how well they are remembered after one week and to link that data to inter-subject correlation obtained with EEG. Therefore, participants will be confronted with 12 TV-commercials from varied companies that will be shown consecutively. EEG activity will be recorded continuously and inter-subject correlation will be calculated on the EEG signal of all participants using the approach developed by Dmochowski et al. (2012). After this, the same commercials will be shown to a different group of people (n = 25) to examine how well the commercials are remembered, followed by a survey about the remembering one week after. The remembering of the commercials is tested one week after viewing, due to a transition period of one week from Short Term Memory in the hippocampus to Long Term Memory in the frontal cortex (Frankland and Bontempi, 2005). In this way, behavior measures can be linked to the EEG activity of all participants to assess the effectiveness of the TV-commercials.

It is expected that the results of the survey are expected to show that there is a difference in the degree of remembering for each TV-commercial. It is also expected that commercials that will be remembered best one week after showing them, will also evoke most ISC.

Materials and Method

Participants

Thirty-three native Dutch speaking participants (15 male, 18 female, age 19-63; M = 38 years, Sd = 10.2) were recruited from the Alpha.One register or asked via social media such as LinkedIn and WhatsApp. All participants were paid €40 to participate in this research and provided written informed consent before the study. For the memory test, Twenty-five participants (14 male, 11 female, age 20-57; M = 24 years, Sd = 18.9) were recruited via WhatsApp and mouth-to-mouth contact. One participant was excluded due to inattention during the watching of the commercials. None of the participants had a history of neurological illnesses or any form of brain damage and were using medication that was psychiatric related. All of them had normal or corrected-to-normal vision.

Stimulus and timing parameters

The TV-commercials were presented using NeuroBS Presentation (Neurobehavorial Systems, Inc., San Francisco Bay Area). All commercials were Dutch and obtained from partners of Alpha.One and www.reclameregister.nl and had

the same size (1920x1080) and resolution (96 dpi). No subtitles were added to the commercials and all stimuli were presented directly after each other. The commercials lasted between 27 and 105 seconds each.

EEG measure procedure

After entering the Alpha.One office, participants were led to the EEG lab. The EEG lab is a 1.20*1.20 meter acoustically shielded box/room with a 19-inch PC monitor in it. Here they were given a verbal instruction about experiment and the tasks they were going to perform. At the same moment an EEG cap was applied to the scalp of the participants and the electrodes were tested separately before the beginning of the experiment. Subsequently the electro-oculogram (EOG) was adjusted to record the eye-tracking during the viewing of all commercials. The participants were asked to sit still and watch all TV-commercials that were shown on the monitor. Each trial started with a calibration of the eye-tracking that was fixed around the eyes. Eye movements were monitored, but not analyzed in this manuscript. The trial was divided into three runs of 11 minutes of the same 12 TV-commercials. Each run had another random order. Between the three runs and afterwards the participants were asked to write down all commercials they liked the most. This was only requested to ensure that participants continued to pay attention.

Memory test measure procedure

The same commercial reel as described above was sent to the participants via E-mail using a link in Google drive. In order to control for primacy and recency effects the commercials were randomized for every participant. The participants were asked to open the video on a computer only and watch the commercials without pause and with a normal sound volume. The link was deleted directly after the viewing to prevent viewing it for a second time. One week after the viewing, the participants were sent an online survey (Limesurvey; www.limesurvey.com, Hamburg, Germany) via E-mail to probe which commercials were remembered best. A total of 39 questions were composed to test memory for all 12 commercials. 3 of them were open recall questions about all commercials and 36 of them (the same 3 for all 12 commercials) were questions that focused on the main occurrence during each commercial (see Appendix - Survey). The latter 36 questions were therefore given accompanied by a screenshot from the commercials that could not be associated with the brand at once. Subjects were not informed about the occurrence of the memory test one week later and were not told to actively remember the commercials.

Recording characteristics and instruments

To record EEG activity from all scalp sites (10-20 International placement), 32 electrodes were placed in a stretch cap (Electro-Cap International, Eaton, OH) and Parker Signa Gel (Parker Laboratories Inc., Fairfield, USA) was applied for conductance between scalp and electrodes.

4

Both right and left bone behind the ears were used as reference point for the electrodes. To record the ocular movements, two Electro-oculogram (EOG) channels were applied: above and below the left eye electrodes for the vertical eye movements and on the left and right outer canthi’s for the horizontal movements. The EEG data was collected by BioSemi Active Two amplifier hardware and ActiView707-Lores software (BioSemi, the Netherlands).

Data reduction and preprocessing

The EEG data were processed using Brain Vision Analyzer (BVA) software (Brain Products, Gilching, Germany). During the preprocessing of the data, a down sample to 256 Hz was adjusted and a high-pass filter cut-off of 1 Hz was used and the EEG data were filtered, this time with a 50 Hz low-pass filter. Each trial was cut into 1-second non-overlapping epochs and for artifact removal, each segment that contained amplitude differences outside the range of 0.5μV/200ms and 150μV/200ms were rejected as well as segment jumps larger than 30μV/ms. Further, the data were corrected for eye movements using the EOG channels and independent component analysis, which is implemented in the BVA software. Channels that contained only noise were set to NaN.

EEG preprocessing

To determine the neural similarity (ISC) between participants while watching the commercials, the results of the EEG data were transferred to Matlab R2019b (Mathworks, Massachusetts, USA). The analysis of the ISC was computed according to the method of Dmochowski et al. (2012 & 2014) (see https://www.parralab.org/isc/). This included the capturing of correlated components across the data set of all subjects with a component analysis. In the end it came down to 12 averaged inter-subject correlations (for every TV-commercial one ISC) in total.

Behavioral data

For the statistical analysis, the memory performance of each subject was examined and the commercials were labeled as forgotten or remembered. For the free recall rate, one point was given for each brand name that was remembered correctly by a participant. The total number of times a commercial was recalled was divided by the number of subjects who participated in this part of the study. For the recognition rate, the participants had to rate how well they remembered the advertisement on a scale from 1 to 5, as well as to write down a description about the commercial. The total number of times a commercial was recognized was divided by the number of subjects who participated in this part of the study times five (since this was the maximum score). After this, the results of the memory performance were combined between subjects and both the remembering rate after recall and recognition rate were calculated for further analysis using Microsoft Excel (2010). To get a better visualization of the results on the behavioral data, the duration of all 12 commercials was plotted in the same chart and commercials were sorted on ISC from low to high.

Statistical analysis

Hereafter, the preprocessed data obtained from the EEG experiment was analyzed using R studio 1.0.136. Three multiple regression analyses (one for the free recall rate (1), one for the recognition rate (2) and one for the brand recall after recognition rate (3)) were calculated within this study. This was done to find out how much of the remembering of TV-commercials was significantly linearly related to the ISC. Duration was also included in the regression as an independent variable since previous research suggests a high correlation with memory. To test for Multicollinearity a Pearson correlation was calculated between the ISC and duration of the commercials (4). Since both variables were not correlated, a multiple regression was calculated subsequently to measure the relation between the free recall rate corrected for duration as dependent variable and both independent variables ISC and duration (5). The correction was done by dividing the recall rate per commercial by the length of the corresponding commercial, leading to a remembering rate per second for each commercial. A second way to analyze the data was by conducting a Pearson correlation analysis, to measure the linear correlation between the ISC and the survey results of the free recall rate. Thereafter, the ISC-recall correlation was calculated on the data that were corrected for the duration of the commercials (6).

Results

EEG and memory results

A clear difference in ISC per commercial was found between all 12 commercials (mean = 0.036, Sd = 0.008) and the average of all 3 trials in all 33 participants is shown in figure 1 from smallest to largest. Performance on the memory test showed that the percentages of remembering of the average memory retrieval after recognition and after free recall were respectively 59% and 28%. The results are plotted in a graph, as well as the duration per commercial (figure 2).

5

Figure 3: Pearson correlation between ISC and duration of all 12 commercials. On the X-axis the duration (s) is

shown, on the Y-axis the ISC value. All 12 commercials are 0 20 40 60 80 100 120 0 50 100 150 200 250 C ar ne xt 1 Ee n te ge n… O hr a zo m er 3 D isn e yl an d 4 O nt de k z o rg 5 Vo lk sb an k 6 St aa ts lo te ri j 7 A lbe rt H ei jn 8 Jum bo 9 N ik e 10 Ene co 11 P lus 12 Du ra tio n (s ) Re m em b ere d (% )

Graph with duration of commercials and

remembering rate corrected for duration

Duration (s)

Recall (%)

Recognition (%)

Statistical analysis

The results of the multiple regression analysis showed that the duration is significantly related to the free recall rate per commercial (p = 0.007). Within the same analysis, there was found a p-value of 0.051 for the relation between ISC and the free recall rate. The multiple regression concerning the recognition rate however, showed no significant relationship with ISC (p = 0.219), but there was with duration (p = 0.046). The same result was found in the relationship between the brand recall after recognition and both ISC (p = 0.320) and duration (p = 0.005). The correlation between the independent variables (ISC and duration) was close to zero (r = 0.078, p = 0.810, t = 0.246, df = 10) (figure 3).

The multiple regression analysis that was calculated with free recall rate corrected for duration as dependent variable showed a significant p-value in relationship with ISC (0.012), however no significant relationship with the duration of the TV-commercials (0.815).

One more Pearson correlation was calculated within present study, to verify above stated results. The calculated correlation between the between ISC and the free recall rate corrected for the duration of the commercials (figure 4) appeared to be highly significant (r = 0.724, p = 0.008, t = 3.317, df = 10).

Conclusion

In summary, this study was executed to test if calculating the inter-subject correlation is a robust method to predict the remembering of TV-commercials. More specifically, it was tried to find a link between inter-subject correlation and memory retrieval of 12 Dutch TV-commercials. There seems to be a difference between the memory retrieval after free recall and after recognition: the memory retrieval rate after recognition was twice as much as after free recall. Further, the relationship between the two independent variables (ISC and duration) showed no significant effect, meaning that there is no potential multicollinearity risk by calculating the multiple regressions. The multiple regressions showed that all three types of memory retrieval had a highly significant relationship with duration (p = 0.007, p = 0.005 and p = 0.046). These findings led to the conclusion that duration of TV-commercials is useful in the prediction of the recall rate, recognition rate and recall after recognition rate.

The most important finding of the multiple regression analysis was that only the free recall rate showed a strong trend towards significance in relationship with ISC (p = 0.051). This meant that ISC could potentially be useful as a predictor for the free recall rate of TV-commercials. Therefore, this was investigated again using a different

Commercial ISC

Carnext 0.0246 Een tegen eenz. 0.0266 Ohra zomer 0.0299 Disneyland 0.0313 Ontdek zorg 0.0325 Volksbank 0.0337 Staatsloterij 0.0344 Albert Heijn 0.0378 Jumbo 0.0383 Nike 0.0425 Eneco 0.0462 Plus 0.0528

Figure 4: Pearson correlation between ISC and corrected recall for all 12 commercials. On the X-axis the

recall (x100%) is shown, on the Y-axis the duration (s). All 12 commercials are plotted with white dots, correlation is shown with a red trend line.

Figure 2: Graph with the remembering and duration per commercial.

The percentage recall (light blue line) and recognition (dark blue line) per commercial, combined with the duration (s) (bar plot). The commercials are ranged from lowest ISC (1) to highest ISC (12).

r = 0.078 p = 0.810

r = 0.724 p = 0.008

Figure 1: Table with ISC per commercial. Left column all 12

commercials, right column the calculated average inter-subject correlation.

6

method in which the free racall rate was corrected for the duration of commercials. Now, a strong relationship was found between ISC and the free recall rate (p = 0.012) within the multiple regression analysis. On the other hand, the significant relationship between free recall rate and duration had completely disappeared (p = 0.815). These findings were verified by the Pearson correlation between ISC and free recall rate corrected for duration (p= 0.724). Hence, it can be concluded that there is a strong association between these two variables (ISC and free recall) and that the correlation is significantly different from zero. On the other hand, no significant effect was found between the ISC and the recognition rate of commercials after correcting the data for duration. Altogether, these findings suggest that inter-subject correlation calculated from EEG data is a steady predictor for the remembering of TV-commercials in the form of free recall, however lacks to predict the recognition of TV-commercials.

Discussion

In present study, it is investigated whether the remembering of TV-commercials could be linked to ISC calculated from EEG data. The results show that it is possible to predict memorability of TV-commercials using brain data that is obtained with EEG. The results show that the most important predictor of commercial recall is its length. This was found in all three survey conditions, concerning free recall, recognition and brand recall after recognition. Besides the length of a commercial, inter-subject correlation is a predictor of commercial memorability as well, however only in the form of free recall. For this reason, this research provides more evidence that ISC is a strong method for doing marketing research and to assess TV-commercials using EEG measurements. On the other hand, this study shows that ISC is no steady predictor for the remembering after recognition or brand recall after recognition.

As mentioned before, the length of commercials differed and the duration of them turned out to be significantly related to the memorability. This suggested that commercials that last longer will be remembered better, which was in line with the earlier findings of Goldstein and McAfee (2011). This however, does not mean that longer commercials are more effective than shorter commercials. As mentioned before, brain data is believed to be more reliable compared to conventional methods such as surveys because in this case there is no conflict between the subconscious and conscious parts of the brain (Nisbett and Wilson, 1977). The relationship between duration and memorability is therefore not as important as the realtionship between ISC and memorability. Moreover, evidence has been provided that ISC is a steady measure method to predict the remembering of TV commercials. This was in line with the earlier findings of Chun et al. (2007) and Muzzio et al. (2009), explaining that memory performance is highly related to attentional engagement, which in turn is a good indicator of the effectiveness of a commercial (Dmochowski et al. 2012;

Poulsen et al. 2016; Cohen et al. 2018). It is important noting that the inter-subject correlations that were calculated in this study are averaged from a range of an entire commercial and that they are therefore different for each commercial. For example, the ISC of Eneco is calculated from a range of 35 seconds, while the ISC of Staatsloterij is calculated from a range of 105 seconds. It could be that ISC decreases as time goes on, due to lower engagement over time, which could cause aberrant inter-subject correlations for each commercial. On the other hand, as earlier research showed, the memorability increases over time due to longer exposure to the stimuli. These two factors may be in conflict and therefore it was a logic step to correct the memorability of the TV-commerials for their length in this study. Although a significant effect was found that was in line with the expectations, this phenomenon is unknown and should therefore be investigated more extensively in future research. Besides a significant effect between ISC and free recall, no significant effect was found between ISC and recognition. A possible explanation for this is that the survey about the remembering of the commercials contained a low amount of cues. Only one screenshot per commercial was shown and it could be that participants lacked to identify that specific part of the commercial. An example is Jumbo: 67% of the participants were able to remember the commercial after free recall, however 23% of them lacked to recognize the screenshot later on during the survey. This indicates that the screenshot was insufficient and future research should therefore focus on a larger survey containing significantly more commercial-specific questions. The survey questions about the remembering of commercials were separated into free recall and recognition and as expected, there was a difference in results between these two variables. The highest survey result for remembering was after recognition (mean = 59% for recognition and mean = 28% for recall). As stated in the earlier research, it was already known that memory retrieval through recall is more difficult than through recognition. However, after calculating the multiple regressions, the recognition scored unexpectedly lower than recall in terms of relationship with duration (respectively p = 0.056 versus p = 0.007). This could be since the exposure to commercials during the memory task of present study was only one time, without the question to actively remember them. This could have caused lower scores on the memory task, since repetition of learning tasks would increase the score on such tasks (Cermak, 1969; Campbell et al. 2003). Further, it is worth noting that the content of some commercials was completely remembered after viewing the screenshot (score 5/5 with a clear description of the content), but that participants attributed the commercial to the wrong brand or even to their main competitor. An example is the OHRA commercial, which was attributed to ANWB by 13 of 24 participants. Only one participant was able to recall the name of OHRA after recognition of the screenshot. An explanation for this could be that this commercial (see appendix screenshot OHRA) contained cues that are often associated to ANWB. This company is very well-known in

the Netherlands for offering highway information, roadside assistance and general assistance during car travelling. OHRA on the other hand, is an insurance company that offers various types of insurance and is therefore not immediately associated with car breakdown. This is in line with the earlier findings from Hall et al. (1976) explaining that the more external cues are given, the easier the brain creates a relationship between the presented stimulus and a brand name that is encrypted in the brain combined with these cues.

Lastly, in this research it is assumed that the results of the calculated ISC of an EEG study are applicable population-wide. This assumption is made due to previous research by Berns and More (2011) and Dmochowski et al. (2013) that showed that there is a relationship between ISC and the population-wide success of naturalistic stimuli. Nevertheless this is not necessarily the case for TV-commercials and should be investigated more extensively. A possibility to link ISC to the population-wide success of TV-commercials could be by selecting commercials that have a clear nationwide behavioral measure. An example would be the calculated ISC of commercials about donor registration correlated to the amount of registrations in the period after airing the campaign.

The results of current research are promising, nevertheless more research needs to be done to verify the outcome of it. Future research should therefore focus on the same research question, however with more data points to make the results of it more reliable. If the results of larger researches are in line with current research, this could mean that smaller sample sizes in this research area are adequate, which would be time-effective as well as cost-effective.

In summary, a significant correlation between neural inter-subject correlation obtained with EEG and the remembering in terms of free recall of TV-commercials was found in this research, however not for the remembering in terms of recognition. Current study shows that EEG measures capture unique information regarding attentional engagement of individuals when watching TV-commercials and therefore this method can be used as a neural marker for assessing the effectiveness of TV-commercials. As such, this study provides the first evidence that EEG activations in response to marketing stimuli are related to the memorability rate of commercials and that the remembering of TV-commercials can be predicted by calculating the neural ISC. This can be cost effecitve for brands and other companies, since the more accurate the succesfulness of a TV-commercial can be estimated, the greater the revenue of the product a company offers. Hence, this is an important step in neuromarketing research that could lead to a better understanding of the reaction of the brain on naturalistic stimuli and about how to make TV-commercials as effective as possible.

Acknowledgements

I would like to thank all colleagues at Alpha.One for the fun and educational time during my internship. This research is performed at the Alpha.One office. Alpha.One is a neuromarketing company situated in Rotterdam. They are specialized in analysis of brain processes using eye tracking, fMRI and EEG. The company is an exclusive research partner of Rotterdam School of Management (RSM, Erasmus University) in the field of consumer neuroscience, and together they offer neuromarketing research services to create and improve advertisements, designs and packaging.

References

- Berns, G.S., Moore, S.E., A neural predictor of cultural popularity (2011), Journal of Consumer Psychology, doi:10.1016/ j.jcps.2011.05.001

- Britton, J.W., Frey, L.C., Hopp, J.L., Korb, P., Koubeissi, M.Z., Lievens, W.E., Electroencephalography (EEG): An Introductory Text and Atlas of Normal and Abnormal Findings in Adults, Children, and Infants (2016),

Chicago: American Epilepsy Society, ISBN-13:

978-0-9979756-0-4

- Boksem, M.A.S., Smidts, A., Brain responses to movie trailers predict individual preferences for movies and their population-wide commercial success (2015), Journal of

Marketing Research Vol. LII (August 2015), 482–492

- Cermak, L., (1969), Repetition and encoding in short-term memory, Journal of Experimental Psychology 82, p. 321-326, doi:10.1037/h0028146.

- Cohen, S.S., Madsen, J., Touchan, G., Robles, D., Lima. S.F.A., Henin, S., Parra, L.C., (2018), Neural engagement with online educational videos predict learning performance for individual students, Neurobiology of

Learning and Memory, Vol. 155 (November 2018), 60-64

- Chun, M.M., Turk-Browne, N.B., Interactions between attention and memory (2007), Current opinion in

Neurobiology 2007, 17: 177-184, DOI

10.1016/j.conb.2007.03.005

- Dmochowski, J. P., Sajda, P., Dias, J., & Parra, L. C. (2012). Correlated components of ongoing EEG point to emotionally laden attention - a possible marker of engagement?. Frontiers in human neuroscience, 6, 112. doi:10.3389/fnhum.2012.00112

- Dmochowski, J. P., Bezdek, M. A., Abelson, B. P., Johnson, J. S., Schumacher, E. H., & Parra, L. C. (2014). Audience preferences are predicted by temporal reliability of neural processing. Nature communications, 5, 4567. doi:10.1038/ncomms5567

- Frankland, P. W.; Bontempi, B. (2005). "The organization of recent and remote memories". Nature Reviews

Neuroscience. 6 (2): 119–130. doi:10.1038/nrn1607

- Fries, Pascal (2001), “Modulation of Oscillatory Neuronal Synchronization by Selective Visual Attention,” Science, 291 (5508), 1560–63

- Goldstein, D.G., McAfee, R.P., Suri, S. (2001), The Effects of Exposure Time on Memory of Display Advertisements, EC '11: Proceedings of the 12th ACM

conference on Electronic commerce 2011, p. 49–58,

8 - Hasson U, Nir Y, Levy I, Fuhrmann G, Malach R (2004),

Intersubject synchronization of cortical activity during natural vision. Science 303:1634-1640

- Hall, J.W., Grossman, L.R., Elwood, K.D. (1976), Difference in encoding for free recall vs. recognition,

Memory & Cognition 1976, 4 (5), 507-513

- Hollingworth, H. L. (1913), Characteristic differences between recall and recognition, The American Journal of

Psychology, 24(4), 532–544.

https://doi.org/10.2307/1413450

- Kauppi, J., Jääskeläinen, I.P., Tohka, J., Inter-Subject Correlation of Brain Hemodynamic Responses During Watching a Movie: Localization in Space and Frequency (2010), Frontiers in Neuroinformatics 2010 4: 5, doi: 10.3389/fninf.2010.00005

- Kawasaki, Masahiro and Yoko Yamaguchi (2013), “Frontal Theta and Beta Synchronizations for Monetary Reward Increase Visual Working Memory Capacity,” Social Cognitive and Affective Neuroscience, 8 (5), 523– 30

- Khakhubia, N., TV Advertising as the Destination Marketing Tool (2018), Doi: 10.19044/esj.2018.c7p1 - Ki, J. J., Kelly, S. P., & Parra, L. C., Attention Strongly

Modulates Reliability of Neural Responses to Naturalistic Narrative Stimuli (2016). The Journal of neuroscience :

the official journal of the Society for

Neuroscience, 36(10), 3092–3101.

doi:10.1523/JNEUROSCI.2942-15.2016

- Kumar, S., Neuromarketing: The New Science of Advertising, Universal Journal of Management 3(12): 524-531, 2015, DOI: 10.13189/ujm.2015.031208 - Madsen, J., Margulis, E.H., Simchy-Gross, R., Parra,

L.C., Music synchronizes brainwaves across listeners with strong effecs of repetition, familiarity and training,

Nature Scientific Reports 2019 9:3576,

https://doi.org/10.1038/s41598-019-40254-w

- MATLAB. (2018). 9.7.0.1190202 (R2019b). Natick, Massachusetts: The MathWorks Inc.

- Muzzio, I.A., Kentros, C., Kandel, E., What is remembered? Role of attention on the encoding and retrieval of hippocampal representations (2009), The

Journal of Physiology 2009 Vol 587, Issue 12,

https://doi.org/10.1113/jphysiol.2009.172445

- Nisbett, R.E., T, DeCamp Wilson (1977), Telling More Than We Can Know: Verbal Reports on Mental Processes, Psychological Review, 84 (3), 231–59

- O’Barr, W.M., The Rise and Fall of the TV Commercial (2010), Advertising & Society Review, vol. 11, issue 2 - Poulsen, A.T., Kamronn, S., Dmochowski, J., Parra, L.C.,

Hansen, L.K., EEG in the classroom: Synchronized neural recordings during video presentation (2016), Nature

Scientific Reports 7:43916, DOI: 10.1038/srep43916

- Unknown 1, 2013, https://inside-the- brain.com/2013/03/07/what-is-attention-and-where-is-it-in-the-brain/

- Wang, L., Test and Evaluation of Advertising Effect Based on EEG and Eye Tracker (2019), Translational Neuroscience, vol. 10, issue 1, DOI: https://doi.org/10.1515/tnsci-2019-0003

Appendix

Survey

Open recall:

- Q1: Schrijf zo veel mogelijk merknamen/afzenders op die je je nog herinnert van de reclamereeks - Q2: Beschrijf heel beknopt de storyline van elke

reclame die je je nu nog kan herinneren

- Q3: Schrijf bij elk van de antwoorden die je hebt gegeven bij vraag 2 op hoe bekend je bent met de bedrijven/afzenders van de reclames die je bij vorige vraag hebt opgeschreven.

(kies uit: niet bekend – redelijk bekend - heel bekend)

Main occurrence per commercial (screenshots, see below): - Q4: Hoe goed herinner jij deze reclame op een

schaal van 1 tot 5?

- Q5: Beschrijf de storyline van de reclame zo beknopt mogelijk

- Q6: Noem de naam van het merk/de afzender van deze reclame

Screenshots

Eneco

Albert Heijn Ohra Carnext Disneyland Jumbo

Eén tegen eenzaamheid

Plus Staatsloterij Ontdek zorg Nike