Objective Measurement of Experiences in Tourism and Hospitality: A Systematic Review of Methodological Approaches and Best Practices

Dissecting Experiences

Experiences reflect the multifaceted nature of human perception, cognition, and emotions. Experience is an umbrella concept which encompasses the richness and diversity of sensations, thoughts, feelings, and interactions that individuals encounter. Table 1 summarizes the selected constructs perspectives related to experience theory originating from emotion psychology, cognitive psychology, and neuroscience, which provides a foundation for the discussion of the measurements. This theoretical perspective focuses on the individual as the unit of analysis and assumes that psychological processes have both subjective and biologically mediated components. In the following theoretical review, we first distinguish between remembered and lived experiences. We then argue that emotions make a key link between lived and remembered experiences, and break emotions down into their components. In a seminal position statement based on the work of behavioral economist Daniel Kahneman, Zajchowski et al. (2017) stated that leisure, tourism, and hospitality experiences should be understood as two distinct but related phenomena: lived experience and remembered experience. While lived experience reflects the sense that we are experiencing something—perceiving, judging, and reacting to a continuous stream of sensory information—remembered experience is a temporally distinct episode reconstructed from such information which has been stored in memory. An important question in psychology over the past decades has been how lived experiences lead to remembered experiences, and how they differ.

OPEN IN VIEWER

Table 1.

Selected Constructs Under the Umbrella of Experience.

Construct	Sometimes Also Called*	Measurement Approaches	Definition	Seminal Reference(s)
Remembered Experience	Recalled experience, reflection, reconstructed experience	Self-report	A temporally distinct episode reconstructed from information obtained over the course of lived experience which has been stored in memory	Zajchowski et al. (2017)
Lived Experience	Subjective immediate experience, immediate conscious experience, on-line experience (in a non-technology context), stream of consciousness	Self-report (indirectly); Objective methods (directly)	A sense of experiencing something—perceiving, judging, and reacting to a continuous stream of sensory information	Zajchowski et al. (2017)
Affect	Mood, feelings, emotionality, positivity/negativity (in a between-persons, trait-level context), emotional characteristics of people	Self-report (indirectly); Objective methods (directly)	“A broad rubric that refers to all things emotional” (p. 247); Stable, trait-level degree to which certain emotions or emotion categories are more or less easily triggered, and therefore more or less frequent	Rosenberg (1998)
Emotion	Emotional reactions	Self-report (indirectly); Objective methods (directly)	“Acute, intense, and typically brief psychophysiological changes that result from a response to a meaningful situation in one’s environment” (Rosenberg, 1998, p. 250); necessary for memory formation and behavior processes to be triggered	Barrett (2017a); Mauss and Robinson (2009); Rosenberg (1998)
Core Affect	Immediate feelings, immediate emotional reaction	Self-report (indirectly); Objective methods (directly)	Pre-conscious attractive or aversive response to meaningful stimuli; most often understood to comprise dimensions of arousal, reflecting intensity; and valence, reflecting attractive-/aversiveness	Barrett (2017a)
Arousal	Intensity, activation, strength, emotional energy, engagement	Self-report (indirectly); Objective methods (directly)	The relative experienced intensity of a core affect or emotion response	Barrett (2017b)
Valence	Pleasure, positivity/negativity, desirability	Self-report (indirectly); Objective methods (directly)	The extent to which a core affect or emotion response is experienced as relatively pleasant or desirable in the lived experience	De Wijk et al. (2022)
Happiness	Subjective happiness, subjective wellbeing, wellbeing, hedonic and/or eudaimonic and/or evaluative well-being, subjective quality of life	Self-report	The cognitive and affective judgments an individual makes about how well their life is going	Diener et al. (2002)
Stress	Pressure, anxiety, apprehension	Self-report (indirectly); Objective methods (directly)	The physiological and psychological reaction that occurs when an individual perceives a situation as potentially overwhelming, threatening, or beyond their ability to cope effectively	Petrick et al. (2021)
Attention	Focus, concentration	Self-report (indirectly); Objective methods (directly)	The mental process of directing one’s awareness and cognitive resources toward a specific task, object, or information	M. Li et al. (2020)
Visitor Movement	Visitor location patterns, visitor spatial behavior, (individual) visitor flows	Self-report (indirectly); Objective methods (directly)	Change over time in physical location of visitors to a specific bounded geographic area	East et al. (2017)

Note. *These constructs have unfortunately been referred to by multiple, occasionally overlapping or vague terms. Our first column should be seen as the most common and/or most distinctive term for the given construct, whereas the second column is a non-exhaustive list of terms (probably) used to refer to that construct in allied psychology or tourism, leisure, and hospitality literatures.

Bastiaansen et al. (2019) summarized the state of the art of this research as it applies to leisure, tourism, and hospitality. To make sense of their conclusions, it is important to assert that both lived and recalled experiences contain multiple constructs reflecting the psychology of consciousness; these include but are not limited to thoughts, feelings, and motivations. Within constructs related to feelings, a distinction must be made between affect, “a broad rubric that refers to all things emotional” (Rosenberg, 1998, p. 247); emotions, which are specific, brief, conscious reactions to personally relevant stimuli; and “core affect,” a term proposed by Barrett (2017a) to describe the initial, pre-conscious reaction of valence and arousal toward a stimulus. Subsequently, the mind and body construct a full-fledged experience of emotion.

Bastiaansen et al. (2019) proposed a process model of experience, which argues that sensory input from outside and within the body, as well as awareness of one’s own cognition, triggers mental models and core affect reactions to create a continuous sense of experiencing (i.e., lived or immediate experience). To manage this information stream, mental models for periods of time (e.g., breakfast, commute) are used to segment consciousness into temporally distinct experiential episodes. Within each, core affect ebbs and flows. When core affect reaches a certain threshold level of arousal, yet another set of mental models is triggered—namely, emotion (Barrett, 2017a). Mental models for emotion create the experience of being emotional in an individual and lead to that experiential episode being encoded in memory.

It is well understood that the experience which is remembered is different from the lived experience. There are systematic biases that occur in the encoding of a lived experience to memory. Gilbert (2006) and Wirtz et al. (2003) found that people expect and recall more intense emotions than they report experiencing near the moment in question. Also, people tend to attend to and recall negative experiences more than positive experiences (the “negativity bias”) despite also remembering positive experiences as more positive than they were (the “positivity bias” or “rosy view”). These and other known episodic or autobiographic memory biases (for a review see Conway & Pleydell-Pearce, 2000) hinder the straightforward use of self-reports in experience research and call for careful consideration concerning the use of subjective and objective ways of measuring experiences.

Measurement Approaches to Experiences: The Necessity for Objective Measurement

Experiences, as understood within the realms of tourism and hospitality research, are intricate constructs, deeply rooted in human cognition, perception, and emotional responses. During the early stages of experience research in tourism and hospitality, researchers often relied on subjective measurements through self-report methods. Participants would rate their experiences using open-ended questions or affective items on a Likert scale (Hosany & Prayag, 2013; Kozak & Rimmington, 2000). Over time, researchers have explored alternative subjective methods to measure experiences, aiming to address some limitations of the traditional approaches (e.g., recall bias and lack of ecological validity; S. Li et al., 2018a). One such alternative is the diary methods, which have been employed to understand tourists’ emotions and experiences longitudinally, as demonstrated by Nawijn et al. (2013) and Lin et al. (2014), who asked their participants to record their emotions daily throughout their holidays. Another approach is the Experience Sampling Method (ESM), which requires participants to provide real-time reports of their emotions and experiences at random intervals throughout the day (Larson & Csikszentmihalyi, 2014). Kahneman et al. (2004) built upon the strength of ESM and introduced the Day Reconstruction Method (DRM), which involves asking participants to recall their experiences by breaking down their day into episodes and evaluating the emotions they felt during each episode.

Despite the widespread use and development of subjective measurement methods, major drawbacks persist. A notable limitation of subjective measurements is the influence of cognitive biases on the data collected (Cattell et al., 1949; S. Li et al., 2015). Hazlett and Hazlett (1999) argued that the cognitive effort needed to describe emotions can impact the accuracy of self-report measures, which is especially relevant when recalling emotional experiences, as memory distortions can lead to discrepancies between the lived experience and the remembered experience. Given these cognitive biases, there is a pressing need for more objective methods. Technologies such as electroencephalography (EEG), which focuses on the electrical activity of the brain, offer insights into cognitive states and emotional processes, providing a more direct and continuous measure of experiences (Luck & Kappenman, 2017). This real-time data can bridge the gap between perceived and actual experiences.

The second major drawback of subjective measurement of experience is social desirability bias, which refers to a systematic error in self-report measures resulting from the desire of respondents to avoid embarrassment and project a favorable image to others (Fisher, 1993). Steenkamp et al. (2010) examined the issue of socially desirable response tendencies in survey research, particularly in the context of marketing, and argued that this bias introduced extraneous variation in scale scores and, as a result, compromised the validity of survey data, which is also evident in self-report of experiences in tourism and hospitality research (S. Li et al., 2018a). This further underscores the importance of objective measurements. Eye-tracking, as an example, emerges as a powerful tool in this context. By objectively measuring visual behaviors, eye-tracking bypasses the pitfalls of social desirability biases by offering a genuine record of where individuals focus their attention, making it especially valuable in assessing the effectiveness of visual stimuli in tourism and hospitality settings (Wang & Sparks, 2016).

Lastly, self-report methods could not capture continuous emotional experiences or account for the dynamic and comprehensive nature of experiences (S. Li et al., 2022, Micu & Plummer, 2010). Traditional self-report methods only capture a snapshot of participants’ experience at a given moment in time, failing to provide a comprehensive understanding of participants’ entire experience throughout an event. Limited sensitivity to subtle emotions can also hinder the ability of subjective measurement methods to fully capture the complexity of participants’ experiences (Barrett, 2006). These limitations highlight the imperative for objective measurements. Location-tracking methods, such as GPS, offer a dynamic record of participants’ movements, behaviors, and interactions within an environment. By capturing real-world data continuously, GPS provides a holistic, moment-to-moment view of experiences, allowing researchers to understand the intricacies of participants’ journeys in tourism and hospitality contexts (Kumar et al., 2013). Such granularity in data is often missing in traditional subjective measures.

Recognizing these gaps and the potential of technology-driven solutions, researchers have increasingly turned to objective measurements to gain a more comprehensive understanding of experiences. When it comes to the objective measurement of experiences, a distinction can be made between physiological and behavioral measures (Bastiaansen et al., 2019). Physiological measures capture bodily properties of the processes that give rise to experience (Jacobs, 2015), and can therefore be treated as an indirect measure of experience.

Physiological measures can be distinguished into two different types: measures that capture processes related to the Central Nervous System (CNS) and measures that reflect activity in the Peripheral Nervous System (PNS). Judging from the overview presented in Appendix A (see the online supplemental material), Electro-Encephalography (EEG; Berger, 1929) has been by far the most used CNS method in tourism, while functional Near-InfraRed Spectroscopy (fNIRS; e.g., Slevitch et al., 2022) and functional Magnetic Resonance Imaging (fMRI; e.g., Munoz-Leiva & Gomez-Carmona, 2019) have been used more sporadically. The limited use of fMRI in tourism and hospitality studies might be attributed to its stringent requirements. Participants must remain still inside the scanner’s bore hole, and face significant restrictions on head movement, and researchers have limited options for presenting stimuli. Therefore, the application of fMRI may be limited to specific phenomena and contexts within tourism and hospitality experience research, holding potential usefulness particularly when combined with other methods. In sum, given the large predominance of EEG as a neuroscience research tool in tourism, most of the discussion in the present work focuses on EEG.

EEG is the most widely used non-invasive technique for recording activity from the human brain in action. It focuses on the electrical activity of the brain. More specifically, EEG is sensitive to the excitatory and inhibitory inputs to the large pyramidal cells in the cortex (the outer layer of the brain, which supports most of the conscious processing; Luck & Kappenman, 2017). The EEG signal is made up of a superposition of many neural processes, some of which subserve basic bodily functioning (e.g., maintaining heart rate, blood pressure, or body posture), and some of which subserve cognitive and affective processes that overlap both in time and in the brain areas in which these processes take place.

Generally, there are two different ways to isolate components of the EEG signal that relate to specific cognitive or affective processes. One is to have study participants engage in brief tasks (i.e., trials) that typically take several seconds (e.g., watching pictures of a destination) and to average the time segments of the EEG signal that correspond to those trials. The resulting signals, the so-called Event-related Potentials (ERPs), reflect specific cognitive and/or affective processes and have been thoroughly described in the literature. Differences in ERPs between two experimental conditions (e.g., an attractive vs. unattractive destination) can then be used to infer differences in cognitive or affective processes between the experimental conditions.

A different way of analyzing EEG signals is to look at the different frequency bands that the signal contains (e.g., by applying specific band-pass filters to the signal). Under certain conditions, changes in power (e.g., the average strength of the signal) in a specific frequency band can be related to specific cognitive or affective processes (Bastiaansen et al., 2012; Herrmann et al., 2016). The idea behind this type of analysis is that brain cells inherently tend to fire rhythmically and that the brain rhythms that result from that property are reflected in EEG recordings by sine-wave-like oscillations in different frequency bands. The advantage of this approach compared to the ERP approach is that it does not require the repetition of many trials; however, the drawback is that the coupling of frequency-specific EEG changes to cognitive and affective phenomena is much less clearly defined.

With regards to capturing processes from the PNS, most measures that have been used in tourism and hospitality are related to electrodermal and cardiovascular activity. For electrodermal activity, skin conductance (SC) is the most prevalent measure that has been used both in fundamental psychology (Kreibig, 2010) and in tourism and hospitality (S. Li et al., 2022). SC is related to the activity of the sweat glands, which has been empirically coupled to emotional arousal (Boucsein, 2012). To measure SC, electrodes are generally placed on the phalanges of participants’ non-dominant hand to detect fluctuations in SC, although other electrode positions have been used as well (e.g., wrist and foot). Particularly, the amplitude of subtle, yet characteristic, bursts of SC (having a steep increase and slow decrease) has been used as a proxy for the level of arousal of an emotional response to certain stimuli and events. To validly detect such SC responses, it is necessary to deconvolute the recorded SC signal into a tonic, long-term driver of the signal, and a phasic, short-term driver of the signal. The latter of these two drivers is a more accurate reflection of sweat gland activity than the aggregate SC signal and is, thus, a more appropriate proxy of emotional arousal (Benedek & Kaernbach, 2010). Importantly, SC does not distinguish between positive and negative emotions (Boucsein, 2012).

For cardiovascular activity, two measures that have been used in the field of tourism and leisure consist of heart rate (HR) and heart rate variability (HRV). HR is the frequency of the heartbeat across a given time interval; HRV is a measure describing changes in HR over time. HR(V) can be measured using electrocardiography (ECG), where electrodes are placed on the skin to measure electrical changes resulting from the activity of the cardiac muscle, or using photoplethysmography (PPG), where infrared light is emitted into the skin to detect variations in blood volume, which can be related to HR in turn. While fluctuations in HR have mostly been associated with negative emotions (e.g., Brosschot & Thayer, 2003), several studies have related fluctuations in HR with positive emotions as well (see Kreibig, 2010 for review). Fluctuations in HRV have mostly been linked to emotion regulation resulting from the interplay between the sympathetic (activating) and parasympathetic (inhibiting) nervous systems (Appelhans & Luecken, 2006). In sum, while HR and HRV have all been coupled to emotional saliency or arousal, it is not evident whether a clear distinction can be made between positive and negative emotions on the basis of these measures alone. Therefore, in working with peripheral physiological measures to capture the experience emotion, it has been suggested to supplement these measures with others (e.g., the Experience Reconstruction Method, see Strijbosch et al., 2021a) that allow for relating physiological proxies of emotional arousal to either pleasant or unpleasant experiences.

Another popular experience-capturing process from the PNS route is facial Electromyography (fEMG), which measures subtle facial muscle contractions and provides insights into people’s experiences (Bolls et al., 2001). By monitoring the electrical signals produced when specific facial muscles are activated, fEMG can reveal the lived experiences behind these contractions. For example, the zygomatic major muscle is associated with positive experiences, while the corrugator supercilii muscle indicates negative experiences (Lang et al., 1993). In tourism and hospitality research, fEMG has proven to be a valuable tool for understanding consumers’ emotional responses to marketing materials and sensory stimuli (Jelinčić et al., 2022; S. Li, 2019). Alternatively, instead of putting sensors on participants’ faces, some researchers have adopted a less intrusive approach, namely using software (e.g., FaceReader) that leverages neural network algorithms to detect emotions based on facial muscle movements (Terzis et al., 2013). By evaluating video footage of participants’ facial movements, the software can detect discrete emotions, allowing researchers to examine both emotional arousal and valence (Zaman & Shrimpton-Smith, 2006).

Regarding the behavioral measurement of people’s experiences, two methods stood out: eye-tracking and location-tracking. As the tourism experience is intangible, visualization of experience components is important to depict its images in delivering and communicating with customers. Scott et al. (2019) emphasized that eye tracking could enhance experience design by testing consumer focuses and reactions to hotel interiors, layouts, or other visual elements that imply hotels’ messages and values. Eye-tracking technology has been treated as an objective and accurate method for measuring attention (Berto et al., 2008). When people look at an element of information, they take time to read and process it. Therefore, a positive relationship exists between fixation time (duration) and people’s attention and cognitive efforts (Hutton & Nolte, 2011). Fixation counts, fixation duration, point order, and patterns of saccades are commonly measured in eye-tracking studies to examine what elements are more eye-catching, how long people spend time viewing them, what information attracts attention first, and how much space they view (Smit et al., 2015; Wang & Sparks, 2016).

There are a few benefits of using eye-tracking technology: First, an eye-tracker can objectively measure people’s relative attention and eye movements. It records participants’ actual visual behaviors instead of their memory of the process (Rayner et al., 2001). It is especially useful for assessing the effectiveness of multiple stimuli, such as visitor guides, menus, and flyers (Wang & Sparks, 2016). A large amount of information decreases the reliability of people’s memory of what they read, so using an eye-tracker can adequately compensate for the drawbacks of self-reported data from a methodological perspective (Shen et al., 2020). Furthermore, it is difficult for participants to manipulate their visual behaviors and eye movement data, and thus these data are more valid and reliable than self-reports (Smit et al., 2015).

As for location-tracking, two primary methods exist: Global Positioning System (GPS) and Bluetooth. The most common method is GPS, which derives location using signals from multiple satellites (Kumar et al., 2013). Many smartwatches and nearly all smartphones have GPS receivers which can combine the GPS signal with triangulation based on distance from multiple cell towers to improve location accuracy where satellite signal is limited. GPS-equipped devices can be lent out to research participants to track their location (e.g., Mitas et al., 2022). Participants can also use their own devices to record their own data, and then share this data with researchers (e.g., Mitas et al., 2023).

GPS signals are usually unable to penetrate buildings and generally unsuitable for tracking location in indoor spaces. Instead, some indoor experiences allow location tracking using Bluetooth transmitters, also known as “beacons” (e.g., Mitas et al., 2023). Beacons transmit a signal containing a specific identification number at a (usually) fixed frequency. These signals are then usually received by an application on a smartphone or other Bluetooth-receiving device worn by the individual whose location the researcher wishes to track. This methodology allows two different approaches to location tracking. A simpler method is based on the nearest beacon received. A more complex method involves triangulating between beacons. The “nearest beacon” method divides the space being recorded into polygons based on beacon location. These polygons are mutually exclusive in terms of which beacon is considered nearest, but not necessarily exhaustive in terms of covering the entire space. When a participant’s beacon reception indicates a certain beacon a is their nearest beacon (strongest signal), and the signal of a meets a minimum strength threshold, the location for that data point is assigned to the polygon representing an area near or around a. Thus, each data point is then mapped to whichever was its nearest beacon or deemed to not be near any beacon within a certain threshold.

Triangulating between beacons follows the same logic as GPS, wherein the signal strength from three different beacons—the ones with the greatest strength at each moment are selected—is used to infer the presumed location of the receiver-holding participant. This allows a much more accurate mapping of actual participant locations in the space being mapped, as it infers actual location in two dimensions rather than univariate proximity to a transmitter location.

Table 2 provides a summary of the six objective measurement methods reviewed. The comprehensive breakdown encompasses measurements, intrusiveness, captured time windows, instruments/devices, costliness, strengths, and weaknesses of various tools from general literature. More detailed information on each of these measurement methods, their applications in the experience industry, and best practices, will be addressed in the section Key Methodological Issues and Best Practices of Employing Objective Measurements.

OPEN IN VIEWER

Table 2.

Objective Measurement Characteristics.

Measurements	Intrusiveness	Captured Time Window	Instrument/Device	Costliness	Strengths	Weaknesses	Contexts and Sample References
Brain Activity (BA)	Medium to high	Seconds to minutes	Brain imaging (e.g., fMRI, EEG) or brain activity monitor	High ($60,000 and up)	High temporal precision, sensitivity, directly measures substrate of psychological experience	While technically feasible outside of a lab setting, the isolation of psychological phenomena is only possible in-lab	Emotional engagement in hotel experiences (Fronda et al., 2021); affective destination image in tourism marketing (Bastiaansen, Straatman, et al., 2022); emotional contagion of virtual tourism experiences (Zhang et al., 2023); evaluation of design of restaurant dishes (Munoz-Leiva & Gomez-Carmona, 2019)
Electrodermal Activity (EDA)	Low	Seconds to hours	Wearable, needs to be with electrodes strapped or stuck to undersides of fingers, palm, or sole of foot for valid measurement	Medium (devices plus software start around $500-1000 per participant at a time)	High temporal precision, sensitivity	Purely a measurement of arousal, in-lab understanding of skin conductance responses has not yet been fully integrated with analyses of continuous mobile measurements, possibility of motion artifacts	Loss aversion in hotel choice (Chark & King, 2022); emotional responses to marketing stimuli (Di-Clemente et al., 2022); hotel environment exploration (Fronda et al., 2021); emotional responses during visits (Kou et al., 2023), in gastronomic experiences (Mengual-Recuerda et al., 2020), and at an event (Strijbosch et al., 2021a)
Cardiovascular Activity (CA)	Low to high	Minutes to hours	Wearable, such as wristband or ring; needs to use chest electrodes for greater precision, heart rate variability, or high-movement applications	Low (devices plus software start around $100 per participant at a time)	Temporal precision, ease of measurement, breadth of platforms available, some prospect of measuring	High chance of motion artifacts, severely confounded by physical activity	Stress and mood of virtual tourism (Baldwin et al., 2021) and onsite experiences (Petrick et al., 2021); emotional engagement in hotel experiences (Fronda et al., 2021); emotional responses to marketing stimuli (Marchiori et al., 2018); heart rate and responses in sports (Nikonovs et al., 2015)
Facial Expression (FE)	Low to medium	Seconds to hours	Facial electromyography (fEMG) or facial expression analysis software	Low for facial expression analysis software and medium for fEMG (devices plus software start around $500-1000 per participant at a time)	Allows measurement of valence, arousal, and some specific emotions	Context must be held constant in research designs	Reactions to marketing stimuli (S. Li et al., 2018a, 2018b); sensory features of dining experiences (Gluchowski et al., 2021); emotional responses during onsite visits (Jelinčić et al., 2022) and in virtual tourism (Zhang & Xiao, 2023); emotional contagion and perceptions between tourists and hosts (Zhang et al., 2023)
Eye-tracking (ET)	Low	1–2 hr without a power bank	Mobile eye tracking, head-stabilized eye tracking, remote eye tracking, and integrated or embedded systems	Low to medium; varies based on different devices and brands ($1,000–$18,000)	High precision in capturing gaze direction, which can help understand attention and interest	Might be influenced by external distractions (e.g., sunlight, overheating) and it requires proper calibration	Visual attention to marketing stimuli (Scott et al., 2020) and in experiences (Hong et al., 2022; S. Li et al., 2022); Attention to forest landscape (Wu et al., 2022) and service recovery (Boyol Ngan & Yu, 2019); visual comfort in hotel lobby (Geng et al., 2023)
Location-tracking (LT)	Very low	Seconds to weeks	Passive data logger or tracker (e.g., a GPS tracker, a Bluetooth beacon)	Low (devices plus software start around $100 per participant at a time)	Long-term passive data collection with minimal participant burden	Typically requires post-processing, and might lack fine-grained temporal precision or specificity	Tourist movement at destination (Huang et al., 2020; Galí & Aulet, 2019); cruise tourist movement (Navarro-Ruiz et al., 2020); experience of cycling and motorcycling journeys (Scuttari, 2021)

Literature Search Strategy and Study Selection

We conducted a comprehensive search of electronic databases, focusing on Web of Science and EBSCO (including databases such as Academic Search Complete, Business Source Complete, and Hospitality & Tourism Complete) in April of 2023, to identify relevant articles (Liberati et al., 2009). Our search strategy aimed to capture articles published within the last decade that employed objective measures in the context of tourism and hospitality experiences. We used a combination of keywords related to objective measurement, the tourism and hospitality industry, and experiences or emotions. The search query used was as follows:

(skin conductance OR skin sensor* OR electrodermal OR HRV OR heart-rate OR eye-track* OR EEG OR Electroencephalo* OR EMG OR Electromyogra* OR Facial OR GPS) AND (touris* OR vacation OR travel* OR hospitality OR hotel OR event-planning OR restaurant OR airline) AND (experien* OR emotion*)

We applied this search strategy to the titles, abstracts, and keywords of the articles in the selected databases. After removing duplicates, the initial search resulted in 1,563 articles. Two reviewers independently screened the titles and abstracts of these articles to assess their eligibility for inclusion, following the PRISMA-ScR guidelines (Tricco et al., 2018). Any disagreements between the reviewers were resolved through discussion or consultation with a third reviewer. To be included in the review, each study underwent evaluation based on four requirements: incorporation of at least one objective measurement tool; reporting of original research; publication within peer-reviewed journals in the last 10 years; as well as being written in the English language. As a result of reviewing the titles and abstracts, the researchers identified a total of 125 articles that met the specific inclusion criteria. The inter-rater agreement was calculated through agreement percentage, yielding a value of 96.80%, which indicates a good agreement level (Landis & Koch, 1977).

Data Extraction and Quality Assessment

The full text of each article was downloaded and essential information such as the author(s), title, journal, publication year, research context, research design, sample size, and the measurements used, were retrieved from the full text. One reviewer took on the responsibility of extracting this data, while a second reviewer checked the information for accuracy and completeness. This approach follows the best practices suggested by Levac et al. (2010). Based on the information from the full texts, 23 studies were excluded from the review due to the study not conducting original research (n = 4), the focus not being on tourism or hospitality (n = 11), or the main text was written in a language other than English (n = 8), which brought the total number of articles down to 102.

We also assessed the methodological quality of the studies included. Both reviewers independently evaluated each included study to ensure the reliability of the quality assessment. The quality assessment process enabled us to identify potential methodological issues and biases in the studies included, which informed our synthesis of the findings and our recommendations for best practices (Hong et al., 2018). Three articles were determined to not meet the required quality standards and were subsequently removed from the review, which brought the final number of articles included in the review to 99 (see Appendix A in the online supplemental material).

Number of Publications

The results of the systematic review comprising 99 articles on objective measurement of experiences in tourism and hospitality show a significant increase in the number of publications over the last decade. Most of these articles were published recently, with 21 articles (21.2%) in 2022, 19 articles (19.2%) in 2021, and 16 articles (16.2%) in 2020, which highlights a rising interest in the application of objective measurements in tourism and hospitality settings (Figure 1). Furthermore, nine articles (9.1%) were published in the first 4 months of 2023, suggesting that the upward trend in research on this topic is likely to continue.

OPEN IN VIEWER

Figure 1.

Number of Publications by Year.

Research Context

Destination marketing and urban tourism are the two most extensively researched themes, with 20 and 10 publications respectively (Table 3). Lodging and food and beverage research also demonstrate a significant presence, with hotel online platforms and restaurants being the most studied contexts, accounting for six and five publications, respectively. Recreation, while not as prominent, also contributes to the overall picture of the research landscape with 12 publications. Theme parks, for instance, are a popular context within the recreation category, featuring in five publications.

OPEN IN VIEWER

Table 3.

Research Contexts of the Studies Included in the Systematic Review.

Theme	Context	No. of Publications
Travel	Urban tourism	10
	Heritage tourism	6
	Virtual tourism	6
	Nature-based tourism	5
	Vacationing	2
	Mountain tourism	2
	Ski tourism	1
	Artificial intelligence	1
	International travel	1
	Risk management	1
	Online travel agency	1
Destination	Destination marketing	20
	Local community	2
Lodging	Hotel online platforms	6
	Hotel environments	4
	Hotel selection	2
	Human-robot interaction	1
	Accommodation sharing	1
Recreation	Theme parks	5
	Museums	4
	Zoos	1
	Art activities	1
	Theaters	1
Food and beverage	Restaurants	5
	Food product	3
	Wine tourism	1
	Gastronomy	1
Transportation	Cruise	3
	Airline	2
Events	Cultural events	2
	Festivals	2

Note. The total count exceeds 100 as some articles address multiple contexts.

Research Design

Based on the location of the study and if the study adopted an experimental approach, articles in this review can be categorized into four categories: lab experiments, lab non-experiments, field experiments, and field non-experiments. In total, there are 55 lab-based research publications compared to 44 field-based research publications. Furthermore, there are 60 experimental studies as opposed to 39 non-experimental studies.

Specifically, lab experiments are the most common research design, with 50 publications employing this approach. Lab non-experiments represent the least common design, indicating that experimental design is preferred for studying people’s experiences in a controlled environment. Field non-experiments and field experiments account for 34 and 10 publications, respectively. Field non-experiments being the more frequently used approach indicates the difficulties of employing experiments in a less controlled environment (Falk & Heckman, 2009).

Measurement Type

Per the discussion in the Theoretical Background section, six primary objective measurement types were examined, which include Brain Activity (BA), Electrodermal Activity (EDA), Cardiovascular Activity (CA), Facial Expression (FE), Eye-tracking (ET), and Location-tracking (LT).

Eye-tracking (ET), with 27 publications, was the most popular measurement, as ET is suitable for studying participants’ visual attention in various contexts, offering valuable insights into cognitive processes and decision-making (Bojko, 2013). In addition, all but one ET study (96.3%) were conducted in the lab, partially because most ET equipment used was not suited for field data collection. Location-tracking (LT), with 25 publications, was the most common measurement in field research, accounting for approximately 46.3% of the total field research publications. LT is often used to study tourists’ movement patterns in various tourism and hospitality contexts, and its popularity can be attributed to the ease of implementation and the ability to provide rich, real-time data in naturalistic settings (Shoval & Isaacson, 2007).

In contrast, physiological measurements like Brain Activity (BA) were more commonly used in lab experiments, with 11 out of 15 BA publications being lab experiments. BA data collection typically requires controlled environments and specialized equipment. The average sample size for BA studies was 40.67, the smallest among all measurement types. EDA and CA, on the other hand, can be used either in the lab or in the field, with a slightly higher average sample size of 57.83 and 50.94, respectively. Furthermore, the high standard deviations of the sample size across all six measurement types indicated the heterogeneous nature of the studies in the tourism and hospitality industry and that a good portion of the publications was still in the exploration phase of applying objective measurements.

Table 4 also illustrates that various objective measurements were frequently used in combination with each other and with subjective data collection methods. For instance, Slevitch et al. (2022) integrated BA, EDA, CA, ET, and survey data in their research to examine people’s experiences when viewing virtual reality hotel images. Their findings highlighted the benefits of incorporating multiple objective measurements to supplement subjective measurements in consumer behavior and experience research. In total, out of the 100 publications, 23 (23.0%) employed at least two types of objective measurements and 96 (96.0%) included some form of subjective measurement, indicating that researchers acknowledge the importance of employing a multi-method approach to gain a comprehensive understanding of people’s experiences (S. Li et al., 2018a).

OPEN IN VIEWER

Table 4.

Descriptive Statistics by Measurement Type.

	BA	EDA	CA	FE	ET	LT
Publication Number	15	23	17	21	27	25
Sample Size
Range	12-130	2-160	7-170	7-230	29-133	8-1369
Average	40.67	57.83	50.94	67.05	61.30	268.75
SD	30.28	38.73	51.26	68.05	31.62	325.02
Research Design
Lab expt.	11	10	8	13	25	0
Lab non-expt.	2	0	0	3	1	0
Field expt.	2	4	2	1	1	4
Field non-expt.	0	9	7	3	0	21
Type of Experience
Physical	3	13	8	5	4	25
Virtual	12	11	8	15	24	1
Stage of Experience
Pre	9	12	16	9	13	14
During	14	23	17	21	27	25
Post	13	23	17	17	22	20
Used With
BA	/
EDA	4	/
CA	2	4	/
FE	0	5	3	/
ET	5	2	1	2	/
LT	0	3	2	1	0	/
Survey	12	18	14	17	25	23
Interview	2	7	1	6	2	3

Note. The total count exceeds 100 as some articles address multiple contexts. BA = Brain Activity; EDA = Electrodermal Activity; CA = Cardiovascular Activity; FE = Facial Expression; ET = Eye-tracking; LT = Location-tracking; expt. = experiment.

Regarding the “Type of Experience” section from Table 4, it is apparent that researchers predominantly studied virtual experiences when it came to eye-tracking, with 24 out of 27 studies centered around virtual scenarios. In stark contrast, location-tracking predominantly focused on physical experiences. This dichotomy aligns well with the nature of each measurement method: ET often investigates digital interfaces and screen-based stimuli, making virtual experiences its primary context, while LT’s practical applications in tracking movements thrive in real-world, physical environments. As for the “Stage of Experience” section, most studies spanned all stages—“Pre,” “During,” and “Post”—and the experience in focus. For instance, in ET research, while there were 13 studies exploring the Pre stage, the During and Post stages were evaluated in 27 and 22 studies, respectively. Similarly, in LT-focused research, all three stages were nearly equally represented with 14 studies in the Pre stage, 25 in the During stage, and 20 in the Post stage. This spread indicates that researchers recognize the importance of understanding the entire journey of an individual’s experience, from anticipation to reflection.

The Central Nervous System Route

Most of the methodological issues, as identified for measures pertaining to the nervous system route, fall under Step 2 of the process chart: selecting the most appropriate measurement equipment and experiment design. One general methodological consideration is whether to conduct EEG studies in the EEG laboratory, or in the field using mobile EEG equipment. Lab studies have two main advantages. One is that they allow for the use of high-quality equipment in a carefully designed set-up, yielding high-quality measurements of EEG. The other advantage is that it allows for careful control over the sensory stimuli and mental processes that the study participant is exposed to. Therefore, the complex changes that can be extracted from the EEG signal if properly analyzed, can be clearly and unequivocally related to the cognitive and affective processes that are of interest to the researcher. However, the major drawback of lab studies is that they lack ecological validity. This is to say that people do not normally have their tourism experiences in a dimly lit, and sound-attenuating EEG chamber, devoid of other external stimuli. Ecological validity would be perfectly intact when conducting studies in the field (i.e., at destinations and sites where people are having their tourism experiences). However, for EEG studies this poses two problems. One is the issue of data quality. EEG involves measuring very tiny electrical signals from the human scalp, in the order of several tens of microVolts. These signals are easily disturbed by external electrical sources (e.g., electrical apparatuses that are nearby, high-voltage power cables) but mostly by electrical sources that stem from the participant itself. Electrical activity originating from eye movements, or from the face, head, and neck muscles are typically a Factor 10, or even 100 larger than EEG signals proper, and removing those so-called measurement artifacts from the EEG signal proper is a challenging enterprise that requires both high-quality recordings (not always guaranteed with mobile, wearable recording equipment) and extensive data analysis skills (cf. Klug & Gramann, 2021; Klug et al., 2022). However, by far the largest problem with EEG measurements in the field using mobile equipment is that it is very difficult to meaningfully relate the complex changes in the EEG signal to the highly multidimensional, complex, and not-well-controlled sensory, motor and mental processes that simultaneously take place in individuals that are freely behaving in natural environments. Therefore, only under very exceptional circumstances should mobile EEG be considered as a tool for the objective measurement of tourism experiences. It should therefore come as no surprise that of the 15 studies that used brain imaging techniques in this review, only two involved recording data outside of the lab (i.e., Mengual-Recuerda et al., 2020; Wu et al., 2022). At the same time, attempts at overcoming the different challenges involved in mobile EEG are increasing (Stangl et al., 2023). In sum, with the current state of technological developments, it is considered best practice to design EEG studies for a lab context. Concerns regarding ecological validity can be mitigated, for example, by using immersive media technologies such as Virtual, Augmented or Mixed Reality (VR, AR, / XR), that transpose participants into relevant tourism or hospitality contexts.

Limiting the discussion to lab-based EEG studies for the remainder of this section, the analysis of ERPs imposes clear constraints on experimental design and data analysis. ERP research is widely used in fundamental and applied psychological research, as there is a very clear and well-documented relationship between the ERP component changes on the one hand, and the cognitive and affective processes taking place in the study participants’ brains on the other hand (cf. Luck & Kappenman, 2011). Despite this, only a handful of studies have applied ERP analysis in a tourism or hospitality research context (e.g., Bastiaansen, Straatman, et al., 2022; Bastiaansen, Oosterholt, et al., 2022). Clearly, the constraints on study design are severe, and it may be challenging to design theory-relevant tourism studies that satisfy those constraints.

Beyond ERPs, there is a (much smaller, and less well-documented) literature that has identified cognitive and affective correlates of changes in EEG oscillations (see below for a brief review). By its nature, analyzing EEG oscillations allows for much more flexible study designs, as power changes can be quantified over relatively long time stretches (up to several minutes, possibly even longer), and there is less need for task repetition compared to with ERP studies. As a result, this type of EEG analysis has been more widely applied in tourism research.

In general, oscillations in the theta frequency band (between 4 and 7 Hz) have been related to memory, and working memory processes (e.g., Klimesch, 1999; Scheeringa et al., 2009). Changes in theta oscillations have been reported in various tourism studies (e.g., Refs. No. 23, 37, & 71 in Appendix A). Oscillations in the alpha frequency band have most consistently been related to attention, with less alpha corresponding to more attention, and more alpha corresponding to relaxation or mental idling (van Gerven & Jensen, 2009). In addition, the extent to which alpha oscillations over the frontal parts of the brain are asymmetric (terms Frontal Alpha Asymmetry, or FAA) has been taken as an index of emotional valence or of approach-avoidance motivation (Briesemeister et al., 2013), although the validity of this relationship has been questioned (van der Vinne et al., 2017). Especially questionable is whether short-lasting changes in FAA can be reliably induced by stimuli that are associated with approach or withdrawal, although this may be due to inconsistent use of analysis methods (Smith et al., 2017). Despite the not-so-solid empirical basis for the relationship between FAA and emotional valence motivation, the measure is relatively widely used in EEG research, probably because of its potential usefulness, and because some commercial software packages automatically generate alpha power indices. Altogether, a relatively large proportion of the existing EEG studies have focused on alpha oscillations, either as an index of attention, or as an index of emotional valence/approach-avoidance motivation (Refs. No. 44, 58, 59, 63, 71, 74, & 85).

Finally, EEG oscillations in higher frequency bands, such as beta (13–30 Hz) and gamma oscillations (40 Hz and higher) seem to be more loosely associated with specific cognitive phenomena and have been related to a wide variety of phenomena, such as ADHD (Saad et al., 2018), syntactic analysis of language (Bastiaansen et al., 2010), and movement planning (Zaepffel et al., 2013). Therefore, beta and gamma oscillations are much less readily interpretable measures for tourism research, pending further developments in fundamental cognitive neuroscience research. Only a few papers (e.g., Refs. No. 23 & 37) have reported changes in beta oscillations and, to our knowledge, none have reported changes in gamma band oscillations.

The Peripheral Nervous System Route

Whereas methodological issues pertaining to central nervous system measures mostly fall under Step 2 of the process chart in Figure 2 (selecting the most appropriate measurement equipment and experiment design), methodological issues pertaining to peripheral nervous system measures are more widespread throughout the process chart. Nevertheless, peripheral nervous system route measures also come with methodological issues under Step 2 of the process chart requiring careful consideration. These issues are particularly relevant for facial electromyography (fEMG). One major limitation of studies using fEMG is its sensitivity to motion, particularly when participants move their faces or bodies during the experiment. Consequently, five out of the six articles that employed fEMG were conducted in labs. A similar issue applies to facial recognition software studies, as it is more challenging to record people’s facial expressions when they are on the move. As a result, only four out of 14 facial recognition software studies were conducted in the field. A related limitation for both fEMG and facial recognition software studies is the use of artificial or staged environments (e.g., Refs. No. 3, 20, 26, 30, & 69), which may not accurately represent real-world contexts and could potentially affect participants’ emotional responses. Utilizing staged environments might lead to social desirability bias, where participants modify their responses to meet perceived expectations or adhere to social norms (Paulhus, 2002). Moreover, the controlled settings may not capture the dynamic nature of real-life situations, which are characterized by unpredictable factors and various stimuli that can influence experiences. To overcome these issues, future research should aim to incorporate more ecologically valid methodologies, such as naturalistic observations or in-situ data collection techniques, while ensuring proper electrode placement and providing clear instructions to participants to minimize movement during data collection. Additionally, future studies should compare the use of both fEMG and face-reading software, along with subjective measurement tools, to further examine the validity and reliability of these methods. It is also worth noting that technological advancements such as 5G have significant implications for data collection methods, particularly in terms of cost and effectiveness. The high-speed, low-latency characteristics of 5G technology have indeed enhanced the performance of these measures, enabling real-time, accurate data collection and analysis. This has not only improved the effectiveness of these measures but also reduced the costs associated with data processing and storage and offered the potential for remote data collection (Boccardi et al., 2014).

A methodological issue pertaining to Step 4 of the process chart in Figure 2 (report or control for potential covariate parameters) is that the currently reviewed articles hardly ever report on or control for extraneous variables that are known to affect the psychophysiological signals at hand, such as ambient temperature and humidity in the case of EDA data (S. Li et al., 2022), and respiratory activity during, and caffeine or nicotine intake before, measuring HR-related data, among others (Quintana & Heathers, 2014). While there is enthusiasm for psychophysiological measures as proxies for emotional engagement and experience, this enthusiasm should be tempered by the notion that psychophysiological signals are the end-state of multiple interlocking systems other than emotions alone (Quintana & Heathers, 2014). Controlling extraneous variables by experiment design or statistical analyses would form a solution and is recommended for future work.

A methodological issue that pertains to Step 6 of the process chart (preprocessing the recorded data) mostly comes to the fore in studies using electrodermal activity (EDA) measures. In these studies, we identified a lack of signal deconvolution approaches in the preprocessing steps of the respective EDA data. Based on the current review, EDA signal deconvolution into a tonic, long-term driver, and a phasic, short-term driver does not seem to be a standard procedure in the tourism and hospitality literature as this procedure is only reported in a handful of the reviewed EDA papers (Bastiaansen, Oosterholt, et al., 2022; Mitas et al., 2022; Strijbosch et al., 2021a, 2021b). However, deconvolution of the EDA signal is necessary to get a more accurate representation of sweat gland activity (Benedek & Kaernbach, 2010). EDA recordings yield a cumulative signal of skin conductance responses (SCRs): SCRs that closely follow up on each other superimpose on top of the EDA signal from earlier SCRs which, due to its very slow recovery phase, might not have fully returned to baseline yet. This leads to overestimating SCR amplitudes, which leads to overestimating the level of sweat gland activity and emotional arousal that SCRs are taken as a proxy for in turn. Not deconvoluting the EDA signal to account for these issues thus comes with concerns of internal validity. It is unclear to what extent findings in articles that do not employ signal deconvolution might suffer from overestimated amplitudes because of superimposing SCRs. For future work, approaches to disentangle the phasic from the tonic components as described by Benedek & Kaernbach (2010) or Kyriakou et al. (2019) are therefore highly recommended.

The most important methodological issue that was identified for peripheral nervous system measures in particular falls under Step 8 of the process chart: reporting practices, specifically for studies using electrodermal and cardiovascular measures. With some exceptions, almost all articles that have been reviewed for this review paper come with serious shortcomings when it comes to reporting methodological details. First, while the majority of all papers report on the sites of electrode placement and the sampling frequency, several papers do fail to report on this crucial information (on electrode placement sites: S. Li, 2019; S. Li et al., 2018a; Mengual-Recuerda et al., 2020; Scuttari, 2021; on sampling frequency: Baldwin et al., 2021; Di-Clemente et al., 2022; Fronda et al., 2021; Hoare, 2020; Hruska et al., 2019; Jelinčić et al., 2022; Kou et al., 2023; S. Li, 2019, Li, 2021; S. Li et al., 2018a; Liu & Huang, 2023; Marchiori et al., 2018; Mengual-Recuerda et al., 2020; Nikonovs et al., 2015; Oliveira-Silva et al., 2016; Peng-Li et al., 2022; Scuttari, 2021). Sometimes, this information can be inferred from the devices that have been used for recording the physiological signals at hand, but some accounts fail to report on the used equipment as well (S. Li et al., 2018a; Petrick et al., 2021). Second, most of the papers fail to report on the preprocessing procedure of their physiological signals, such as data screening procedures, artifact detection and correction procedures, signal standardization procedures, and the extraction procedures of measures that are later used in statistical analyses. The current findings are in line with S. Li and colleagues’ (2022) recent review of EDA-related papers (of which 15 papers are shared with the current 23 EDA-related papers). Non-transparency about these procedures not only hinders peer reviewers and readers alike from properly critiquing a study but also hampers future replication studies and meta-analyses. Tourism and hospitality scholars involving themselves with psychophysiological research should therefore better inform themselves about standard reporting procedures (see e.g., Quintana et al., 2016). In turn, peer reviewers should more critically and rigorously assess submissions on these publication guidelines.

The Behavioral Measurement Route

Eye-tracking technology has been applied to measure people’s attention to a variety of stimuli, such as visual comfort in a hotel lobby (Geng et al., 2023), exoticism in destinations (Hong et al., 2022), perceived beauty (Scott et al., 2020), and destination preferences (Ramsøy et al., 2019). We identified some limitations associated with these studies that are related to the process chart from Figure 2 as well. However, before discussing those issues in more depth, the behavioral measurement route comes with a more fundamental methodological issue as well: sampling bias. First, most studies had a sample size of 35 to 60, and very few had about 100 participants (e.g., Ref. No. 8). Student samples and convenience sampling are commonly used in eye-tracking research because of the difficulty of recruiting participants (e.g., Refs. No. 25, 42, & 78). When researchers select participants for a study based solely on their availability or willingness to participate, this approach may not yield a comprehensive and accurate representation of eye movements across various real-life situations and contexts. Convenience sampling and small sample sizes are especially problematic because they may introduce bias into the participant pool and bring challenges for generalizing findings to larger populations. Additionally, different situations and tasks (e.g., reading a visitor guide while planning a 3-day trip, or reading a visitor guide without a task) can lead to variations in behavior. Therefore, a participant’s eye movements in one specific study may not necessarily reflect how they would behave in other everyday situations. To mitigate this limitation and enhance the validity and generalizability of findings, researchers should consider diversifying participant selection criteria, employing more representative samples, or conducting studies across a range of contexts to capture a broader understanding of how eye movements function in various situations. Researchers are also encouraged to recruit visitors on-site as participants and use stratified random sampling to overcome this limitation (Shi et al., 2022).

Beyond this more fundamental issue, there are several issues that can be related to the process chart presented in Figure 2 more clearly. First, one methodological issue associated with the behavioral measurement route pertains to Step 2: selecting the appropriate measurement tool. The issue here, however, is not so much about the eye-tracking equipment, but about supplementary measurement instruments to better contextualize the eye-tracking data. All eye-tracking studies included in this review used surveys instead of interviews. Surveys may not provide sufficient insight into the reasons behind individuals’ attention patterns. Mayr et al. (2009, p. 198) argued that “interpretations of eye-tracking data . . . are often based on assumptions and heuristics about underlying cognitive processes.” Researchers can incorporate interviews as a complementary method to allow participants to express their opinions and provide a more comprehensive understanding of the phenomenon being studied. Another potential limitation is the design of the research. Most eye-tracking studies testing marketing effectiveness analyzed linear stimuli where all participants see the same information (e.g., a photo and a page of an advertisement) within the same amount of time (e.g., Refs. No. 24 & 89). This type of research design may not reveal participants’ natural behavior. Therefore, inviting participants to view stimuli without a specific time limitation and sequence could better reflect their visual processing behavior.

As we transition from the realm of visual stimuli to spatial understanding, location data offer a set of unique challenges due to their structure, as well as some challenges similar to those of the other data types already discussed. The experience of space significantly influences participant behavior, and the technical challenges inherent in location tracking can directly impact the accuracy and reliability of spatial data. These factors, among others, underscore the importance of delving deep into the nuances of location data in experience research. The challenges unique to spatial data may be categorized under data collection, processing, and analysis, thus pertaining to Steps 5, 6, and 7 in the process chart in Figure 2.

Challenges to data collection mostly stem from the fact that location is determined by the sending of remote signals through the air to devices that participants wear. The transmitters of these signals, whether satellites in space or beacons in the same room, depend on their own technical soundness as well as a clear path for the fidelity of the signal sent. It is the clear path that turns out to be a serious challenge for research in our field. GPS signals are blocked by walls, awnings, roofs, overhangs, and even dense tree canopies. In these situations, most smartphones and other powerful GPS trackers still receive a signal, but its noise increases, and accuracy degrades. The problem of finding a clear path for the signal is much more serious for Bluetooth. Not only are Bluetooth signals completely blocked by glass, steel, or concrete, but they also reflect off these surfaces, sometimes causing inaccurate reception of signals. Furthermore, even a human body partly blocks the signal, so, when standing near a beacon but wearing a phone on the opposite side of their body, a “nearest beacon” may instead register as a beacon further away whose reflected signal hit the phone better.

The second set of challenges arises in processing data. Related to the problem of transmitter signal path, both GPS and Bluetooth signals are inherently noisy when received by a smartphone, and again this issue is worse with Bluetooth. GPS tracking applications in general smooth the inputs sufficiently to visualize data with reasonable reliability (e.g., a walk in a straight line actually looks straight on a map), but two receiving devices of the same type may record the data with slightly different spatial offsets (e.g., two people walking the same straight line may produce data that maps as two straight lines a few meters apart). The resulting noise sometimes leads to data loss (e.g., Ref. No. 17) or maps that separate participants who were in fact near each other (e.g., Ref. No. 60). With Bluetooth, smoothing as part of any triangulation software solution is crucial, and often needs to be fairly heavy-handed to produce realistic data.

The structure of spatial data is best understood as a series of (at least) two-dimensional layers (e.g., Ref. No. 11). Each layer has a North-South (latitude) and East-West (longitude) dimension, values of a variable of interest with an associated location, and sometimes also elevation. When a GPS or Bluetooth signal is used to derive location, the variable of interest in the generated layer is time. In simple terms, there is a series of data points generated by a specific participant that can be mapped to a particular location at each (usually) second. To create meaningful visualizations for experience research, additional variables and geospatial data layers are needed. For example, location data can be integrated with additional variables relevant to individuals’ experiences, such as skin conductance (Ref. No. 60), mode of transport (Ref. No. 17), or self-reported affect (Ref. No. 16). These may be displayed using point symbology or aggregation into clusters, or they may be transformed into continuous field representations using weighted kernel densities (Ref. No. 11). Baseline Geographic Information System (GIS) layers which contain information about the location, such as a street map, land cover, or aerial photos, are useful to put data into context (e.g., Refs No. 12). Bluetooth location tracking is relatively simple in this stage of data processing, as location algorithms and layer showing—for example, a map of a room—are both conceptualized as flat, two-dimensional planes in a local coordinate system referenced to the room or building (Ref. No. 99).

When working with geospatial data, such as locations recorded by GPS or geographical data including roads, building footprints, administrative boundaries, or land cover, it is important to realize that these data are often recorded in different coordinate systems. When combining them for analyses or mapping, they need to be re-projected to common coordinate systems (e.g., Ref. No. 97). The GPS data are recorded in ellipsoidal coordinates as latitude and longitude in degrees, while many local and state government data are projected into flat two-dimensional planes with features recorded as North-South and East-West coordinates in meters. If the collected data include embedded information about the coordinate system, many GIS software automatically re-project the data to a common coordinate system. Care should be taken that this is done correctly and that there are no shifts between features from different data sets. For example, paths that participants have cycled should fit onto traces of cycle paths on map layers.

Finally, analyzing spatial data presents the same challenges of data density, volume, and nestedness as the other methods discussed above. These can be circumvented by using simple between-person summaries of location or movement data (e.g., Ref. No. 10), but often it is more complex, within-person phenomena that are most valuable in using location to measure experience. Spatial data features another source of spurious associations in the data, however, namely spatial autocorrelation. This refers to the phenomenon that data points closer to one another are more likely to be related to an outcome variable of interest (e.g., skin conductance). For example, imagine three skin conductance response data points (A, B, and C) from a roller coaster ride. Data Points A and B are 1 m apart on the roller coaster track, while Point C is located another 10 m further along the track. Data Points A and B are likely to be more similar, because they are likely to be closer to one another in terms of the speed, acceleration, and scenery of the rollercoaster than Points A and C. This could lead to spurious associations if one spatially autocorrelated variable is used to predict another. Note that such associations would occur, albeit at different magnitudes, whether or not the data were nested within participants.

There are three ways to deal with spatial autocorrelation. The most relevant is to simply design studies in such a way that predictors are not spatially autocorrelated. For example, a common and effective design is to predict an outcome—such as skin conductance or heart rate variability—based on location within certain polygons in the study area (e.g., Refs. No. 11 &16). A second approach is to break the study area down into very small polygons for mapping, summary, and network statistics (e.g., Ref. No. 17). In the resulting polygons, the spatial autocorrelation itself becomes a variable of interest and thus visualizes the localization of spatial association. This procedure is known as Local Indicators of Spatial Autocorrelation (LISA; Anselin, 2010). A final possible approach is to control spatial autocorrelation within the model. For a simple bivariate regression model, geographically weighted regression is an option (Kim & Nicholls, 2016). More complex multilevel models require the addition of spatial autocorrelation matrices, which can be extremely computationally intensive (Bjørnstad, 2018).