The Development and Validation of the Ontological Addiction Scale

The scale was developed by the present authors, three of whom are experts in Buddhist philosophy and practice. The development of the scale utilised the components model of addiction (Griffiths, 2005), which is applicable to all forms of addiction. The components model of addiction asserts that an addiction must meet six core criteria: salience, mood modification, tolerance, withdrawal, conflict and relapse. However, for the purposes of the present study, the mood modification and conflict components were substituted with euphoria and dysphoria, as these better captured their intended roles as polar opposites relating to affective valence, and appeared to fit better with the notion of ego-addiction. For the content of these six addiction component categories, a component breakdown of the Buddhist eight mundane concerns was used (Nagarjuna, 2005). The eight mundane concerns are as follows:

These eight concerns embody the positive and negative aspects of four underlying components (Table 1); material wealth or possessions; sensations; reputation amongst kin and social circles; and wider reputation. Each component has two polarities: gain and loss. It should be stressed here that both the Griffiths’ components model of addiction and the eight mundane concerns of Buddhism were employed purely to ensure a suitable range of items covering the various facets of ontological addiction. Consequently, they were not intended as distinct, empirical components or dimensions of responses on the scale.

By creating four corresponding items for each of the six addiction components, a prototype scale was formulated using the resulting 24 question-categories as a guide. In order to enable the OAS to be refined according to initial findings, an additional seven candidate items were included as alternatives for some of these items or as additional items deemed worthy of examining. This was to enable weaker items to be identified and rejected according to the results. The prototype scale items are shown in Table 2.

The present investigation was a cross-sectional cohort study employing a correlational design.

Sample size calculations were performed using G*Power3 (Faul et al., 2007). A figure of 200 participants was determined based on a power of 0.95, a significance threshold of 0.05 and a small to medium effect size (r = 0.25) for a correlational design employing two-tailed tests. Sample size requirements for an exploratory factor analysis (EFA) are notoriously difficult to assess because they depend not only the number of extracted factors, but on the anticipated factor loadings. However, guidance suggests that for an anticipated single-factor solution, this is also an acceptable figure for an EFA of the variables under examination (Goldberg & Velicer, 2006). A total of 210 participants (125 males and 85 females) were recruited via the Prolific online recruitment system and therefore comprised a self-selected convenience sample. Each participant received a payment of £2.50 following participation in the survey (mean duration 20 min 25 s; SD = 10.8 min). The mean age for males was 25.1 years (SD = 7.33; range = 18–48 years) and the mean age for females was 29.9 years (SD = 10.80; range = 18–65 years). Of these participants, 187 reported their ethnicity as “White” (89.1%), two as “Black” (0.95%), five as “Asian” (2.38%), three as “mixed” (1.43%) and ten as “other ethnic group” (4.76%). Three participants responded “not stated” (1.43%). Of the 188 participants for whom location data was available, 45 were based in Poland (23%), 39 in Portugal (20.7%), 23 in the UK (12.2%), 11 in Italy (5.9%), nine in the USA (4.8%), eight in Greece (4.3%), six in Canada (3.2%), six in Hungary (3.2%) and six in Mexico (3.2%). The remaining 35 participants (18.6%) were located in France, Ireland, Spain, Latvia, Germany, South Africa, Belgium, Chile, Germany, Israel, Netherlands, Czech Republic, Denmark, Estonia and Sweden. Inclusion criteria for the survey were that participants should be English-speakers aged over 18 years, who did not have psychotic symptoms, neurological conditions, alcohol use disorders and/or drug use disorders. These conditions were excluded in line with standard practice for scale validation studies using general population samples (Boateng et al, 2018).

In terms of assessing convergent and divergent validity for OA, ego-centeredness, non-attachment and non-attachment to self were deemed to be the most applicable constructs. Therefore, the following criterion measures were selected.

For the primary criterion measure of ego-centeredness, an adapted version of the 60-item short-form of the Five-Factor Narcissism Inventory (FFNI-60; Krizan & Herlache, 2017; Sherman et al., 2015) was used. The FFNI-60 comprises items relating to vulnerable and grandiose narcissism. It contains items such as “I deserve special treatment” and “It really makes me angry when I don’t get what I deserve”. Respondents rate their agreement on a five-point Likert scale (“disagree strongly”, “disagree a little”, “neither agree nor disagree”, “agree a little”, “agree strongly”) and scores range from 60 to 300, with higher scores indicating higher levels of narcissistic traits. For the purposes of the present study, items from the Acclaim-Seeking, Arrogance, Entitlement, Exhibitionism, Grandiose Fantasies, Need for Admiration, Shame and Reactive Anger subscales were selected to reflect dysfunctional ego-centeredness. The resulting shorter form (FFNI-32), with scores in the range of 32 to 160, was used as a measure of ego-centeredness that is believed to reflect processes central to OA. The Cronbach’s alpha in the present study was 0.88.

To assess non-attachment, the 30-item Non-Attachment Scale (NAS) was used. This assesses “release from mental fixations” (Sahdra et al., 2010) and employs a six-point Likert scale from 1 (“strongly disagree”) to 6 (“strongly agree”). For the present study, the eight-item, short form of the NAS (NAS-SF) was used (Chio et al., 2018). The scale contains items such as “I find I can be calm and/or happy even if things are not going my way” and “I can accept the flow of events in my life without hanging onto them or pushing them away”. The NAS has total scores ranging from 8 to 48, with higher scores indicating higher levels of non-attachment. The Cronbach’s alpha in the present study was 0.80.

Non-attachment to self was assessed using the Non-Attachment to Self Scale (NTS; Whitehead et al., 2018). The NTS uses a seven-point Likert scale with questions such as “I can let go of unhelpful thoughts about myself”. The total score ranges from 7 to 49, with higher scores indicating higher non-attachment to self. The Cronbach’s alpha in the present study was 0.77.

To examine the relationship between OAS scores and mental health, three well-validated measures of depression, anxiety and self-esteem were used. The nine-item Patient Health Questionnaire (PHQ-9; Kroenke et al., 2001) provides a brief assessment of depressive symptoms. Respondents are asked if they have experienced symptoms such as “little interest or pleasure in doing things” and “feeling down, depressed or hopeless”. Responses are rated on a four-point Likert scale (not at all, several days, more than half the days, nearly every day). The scale has a Cronbach’s alpha of 0.81 and total scores are in the range of 0 to 27, with higher scores indicating higher levels of depressive symptomatology. The Cronbach’s alpha in the present study was 0.86.

The seven-item Generalised Anxiety Disorder Scale (GAD-7; Spitzer et al., 2006) provides a brief assessment of general anxiety. It has seven items including “feeling nervous, anxious or on edge” and “trouble relaxing”, which are scored on the same, 4-point Likert scale. The scale has a Cronbach’s alpha of 0.83 with total scores in the range of 0 to 21, with higher scores representing greater general anxiety. The Cronbach’s alpha in the present study was 0.87.

The Rosenberg Self-Esteem scale (RSES; Rosenberg, 1965) is a 10-item scale that assesses self-esteem (e.g., “On the whole, I am satisfied with myself” and “At times, I think I am no good at all”). Participants rate items on a four-point Likert scale (strongly agree, agree, disagree and strongly disagree). The scale has a Cronbach’s alpha of 0.77 and scores range from 0 to 30, with higher scores indicating higher levels of self-esteem. The Cronbach’s alpha in the present study was 0.91.

The OAS-31 prototype scale is detailed in Table 2. It comprises 31 items across six domains—salience, euphoria, tolerance, withdrawal, dysphoria and relapse. Items are rated on a five-point Likert scale (never, rarely, sometimes, often and always), with scores in the range of 0 to 124. The Cronbach’s alpha in the present study was 0.902 for run 1 and 0.920 for run 2 (see the “Procedure” section).

Data for the test and retest conditions of the OAS item responses and totals, as well as the RSES, FFNI-32, NAS-SF, NTS, PHQ-9 and GAD-7, were analysed using SPSS Version 25 (SPSS Inc., Chicago, IL, USA). Scale scores for the FFNI-32, NAS-SF and NTS were normalised such that the lowest score was always zero, and thereby consistent with negative endpoint descriptors such as “not at all”, “nothing” or “never” in the scales employed. Relationships between these variables were examined using a correlation matrix. If a variable showed no significant skew or kurtosis, and no significant departure from normal distribution, then Pearson tests were used; otherwise, Spearman tests were employed. The data for the OAS items were also subjected to an exploratory factor analysis to examine the scale’s factor structure.

Descriptive statistics for all measures taken are shown in Table 3. Age, PHQ-9 scores and GAD-7 scores showed significant skewness, and Kolmogorov–Smirnov and Shapiro-Wilkes tests showed significant departure from normal distribution. No other variables showed significant skewness or kurtosis, and Kolmogorov–Smirnov and Shapiro-Wilkes tests showed no significant departure from normal distribution for any of the other measures. Correlations of PHQ-9 and GAD-7 scores therefore employed Spearman’s tests, while other scale measures used Pearson tests.

Participants completed the OAS-31 prototype once at the beginning of the test battery, and once at the end. Therefore, in order to examine the factor structure, an exploratory factor analysis was performed on each of these datasets. The factor structure and variance is detailed in Table 4. In order to establish the most appropriate number of factors to retain, a scree plot of factor loadings is usually examined, using the “Kaiser criterion” of retaining factors with Eigenvalues of greater than 1. However, Velicer et al. (2000) argue that this criterion is not appropriate and is prone to result in over-extraction of factors. Horn’s method of parallel analysis (Horn, 1965) offers a more appropriate method by including a baseline from random data generated based on the relevant number of variables and observations. An online parallel analysis engine (Patil et al., 2007) was therefore used to compute optimal eigenvalue cut-offs for each factor, and these were included in the scree plots for this analysis.

Examination of these plots (Figs. 1 and 2) reveals a three-factor solution based on the point at which the EFA scree plot lines meet those of the baseline. Loadings of these factors are presented in Table 5. Item-total correlations were also computed and Cronbach’s alpha was calculated for each run.

Through examining the factor loadings, it appeared clear that a single interpretable factor predominated, with generally high, positive loadings for both runs. The mean factor loadings for this were 0.49 for run 1 and 0.53 for run 2, and suggests that the sample size was adequate for this analysis (Goldberg & Velicer, 2006). This factor, accounting for 26.3% and 30.3% of the variance respectively, was deemed to represent the central construct of ontological addiction. Factor 2 appeared for the most part to reflect the affective valence of the items. For example, in run 1, high positive loadings were evidenced on E1 (“Felt uplifted when you were praised?”) and E5 (“Felt elated when things were going well?”), while high negative loadings appeared to be associated with strong negative affect, particularly items on the relapse subgroup, such as R3 (“Felt regret after having given a gift?”) and R6 (“Felt regret about doing something good for somebody?”). Notably, they did not appear to load very heavily on the dysphoria subgroup. A similar pattern was evidenced for run 2, only in this instance the signs for factor 2 were reversed, with positive loadings indicating strong negative affect, and vice versa. Euphoria subgroup items tended to have higher loadings, as did items from the relapse subgroup, though, again, the dysphoria subgroup items did not feature strongly. This factor accounted for 9.7% and 11.1% of the variance for the respective runs.

The Cronbach’s alpha figures for run 1 (α = 0.902) and run 2 (α = 0.920) of the OAS prototype showed excellent internal consistency and reflected the predominance of a single factor. Therefore, the use of item-total correlations was deemed acceptable to refine the scale by identifying and removing weaker items.

For each of the OAS-31 runs, a reliability analysis was performed in which Cronbach’s alpha figures were calculated for the scale with each item omitted alongside item-total correlations. Figures for Cronbach’s alpha become higher when weaker items are omitted, and lower when stronger items are left out. The aim of this part of the analysis was to arrive at a version of the OAS which contained four items per subgroup, retaining stronger items while dispensing poorer performing items. The dysphoria and withdrawal OAS subgroups did not contain alternative items, and so item-total correlations and adjusted Cronbach’s alpha figures were examined to ensure that they were satisfactory (see Table 6).

For the item-total correlations, all items showed satisfactory figures overall. Items D3 (“Felt low when you encountered financial or material loss?”) and D4 (“Felt low when you encountered difficult circumstances?”) showed somewhat weaker correlations in run 2 than in run 1. Items W3 (“Found it hard to give something away?”) and W4 (“Found it hard to live more simply?”) had poor correlations for run 1, but notably improved in run 2.

Figures for all subgroups are shown in Table 6. The salience, euphoria, tolerance and relapse subgroups included alternative items that were also evaluated in order to allow the omission of weaker items. Although the four-factor structure described in Table 1 was used as a guide to the content of items within each subgroup, not all questions applied to a specific factor. Some items, such as S2 (“Thought about how others see you”), spanned two components, while others, such as E4 (“felt good when you experienced fewer challenges”), were general enough to apply to all four components.

Finally, for the regret subgroup, item R6 (“Felt regret about doing something good for somebody”) performed most poorly in both runs and so this was flagged for removal. Items R3 (“Felt regret about giving something away”) and R4 (“Stopped helping others because it was causing discomfort or inconvenience”) also performed poorly, but since R3 was an alternate formulation of the better-performing R5 (“Felt regret after having given a gift”), R3 was eliminated. A similar process was then used to select the strongest two items per subgroup. These items were then marked for inclusion in an OAS-12 Short Form version of the scale (Table 6).

Correlations between measures are shown in Table 7. All correlations were significant to the p < 0.001 level.

The OAS is grounded in the Buddhism-derived system of psychopathology, which is based on the view that what we think of as our individual self—and for that matter everything else—is psychologically constructed and has no inherent existence (Ducasse et al., 2019). Selfhood is seen as more akin to a mentally constructed narrative or story in which individuals are placed as agents surviving and managing life in a potentially dangerous and hostile environment. Due to its importance to physical and species survival, it is only natural that selfhood appears intensely real and that such basic concerns of survival and well-being take priority. However, the reification of self can lead to concerns and preoccupations becoming increasingly self-centred (Van Gordon et al., 2018a). It is this attachment to self that the OAS seeks to gauge, on the basis that it reflects the root of psychological suffering. This paradigm underlying the development of the OAS—as well as the corresponding findings from this validation study—suggests key mechanisms through which mental illness arises and reinforces itself. However, the underlying OAT model also offers remedies through the use of Buddhist-based meditation practises designed to undermine the illusion of self (Goldberg et al., 2018).

OAT elucidates a three-phased approach to the treatment of OA based on first examining, and then transforming this imputed (i.e. “made up”) self: “(i) becoming aware of the imputed self, (ii) deconstructing the imputed self and (iii) reconstructing a dynamic and dual approach” (Shonin et al., 2016, p. 665). The first stage involves simply becoming aware that there are no rational grounds for believing that the self inherently exists in the way that most individuals think it does. This may involve exercises in which the psychology of selfhood is logically examined and explored, using scientific argument to demonstrate principles of non-self or emptiness, but its main purpose is to cultivate proficiency in meditative awareness, particularly concentration-based meditation (Pāli: samatha). This enhances awareness, calmness and clarity, and allows the individual to experience the contents of their mind in a more equanimous and dispassionate way, consequently bringing about the conditions for confronting maladaptive, ego-centred beliefs.

In the second phase, the individual is taught to engage in a range of spiritual practises that run directly counter to self-interest, such as compassion, loving-kindness, patience, generosity and death awareness. Importantly, they then shift their focus to vipassanā (Pāli) or “insight meditation”. Here, the conditions created by the concentration techniques employed in the first stage are now turned toward a deeper metaphysical examination of the causes, intrinsic properties and existence of phenomena. The goal of this is to cultivate the realisation that neither an individual nor anything else has an intrinsic self (Van Gordon et al., 2016), and that both form and emptiness are ultimately one.

In the third phase, having undermined the core beliefs underpinning ontological addiction, the individual may undergo a transformation of perspective from one which is ego-based, to one grounded in “non-self”. Here, subject-object duality is transcended and a new, fluid and dynamic “true self” can arise which encompasses both the individual and the whole. This newly centred awareness still contains an emulated self in order to function in the world, but there is now recognition of its inherent nonexistence, and the inseparability and interdependence of all form, human and otherwise.

It is in this third phase that the Buddhist concept of emptiness (Sanskrit: śūnyatā) is of particular importance because it articulates an essential truth about the nature of the world. For practical purposes, emptiness can be considered equivalent to non-self and refers to the boundless, undifferentiated “ground” from which all distinctions, duality and manifest forms arise. It is only recently through developments such as OAT that the relevance of these insights—particularly emptiness theory (Van Gordon et al., 2019, 2017a, b)—to understanding and treating mental illness is becoming understood in Western society (Shonin et al., 2014, 2016; Van Gordon, et al., 2018a, b).

In particular, second-generation mindfulness interventions, such as mindfulness awareness training (MAT), have played an important role in recognising the importance of Buddhist principles such as emptiness, non-self and non-attachment, by facilitating deeper levels of metaphysical enquiry concerning selfhood through vipassanā, śūnyatā and other forms of meditation prescribed in the second and third phases of OAT-based treatment (e.g. compassion, loving-kindness, patience, generosity and death awareness) (Shonin et al., 2016; Van Gordon et al., 2015). Śūnyatā meditation, in particular, has invited growing interest because it involves cultivating a direct experience of conscious reality at its deepest level, where self and other dualities can be profoundly transcended. Indeed, evidence from studies of advanced meditators appears to confirm that śūnyatā meditation—in which a state of emptiness-of-self is sought—is more effective than mindfulness for cultivating compassion and non-attachment to self, as well as deeper and more profound mystical or revelatory spiritual experiences that can further undermine self-attachment (Van Gordon et al., 2019, 2017a, b, 2018a, b). This aside, second-generation interventions generally are already proving effective not only for the treatment of mental illness (Ducasse et al., 2019) and for managing physical pain (Van Gordon et al., 2016, 2017a, b), but for more conventional addictions such as problem gambling (Shonin et al., 2013, 2016).

Further research is clearly needed to establish the clinical utility of OAT and the various Buddhist contemplative practices constructs to which it pertains. However, as interventions which seek to undermine the belief in an inherently existing self continue to gain traction in Western medicine, there is a growing need for tools which specifically assess OA. The OAS-24 and OAS-12 Short Form should therefore be a helpful addition to other Buddhism-derived non-attachment scales currently in use (Sahdra et al., 2010; Whitehead et al., 2018).

A limitation of the present study is that the psychometric properties of OAS content were not always similar between test and retest runs, with some items showing marked differences from one run to the next, although variations due purely to chance are to be expected with such a large number of questions. However, this underlines the fact that many different possible selections of these questions would perform equally well and—were the study to be repeated on another independent sample—the findings might suggest a somewhat different selection of items for these scales. Nonetheless, the methodology employed herein served the crucial purpose of identifying weaker and stronger items that could guide the selection of items for the two shorter forms of the OAS.

The test–retest interval in this study was also somewhat short at around 15 min on average; a longer interval would have been desirable to give a better idea of how stable responses to the OAS are over time. However, the retest regimen did provide some crucial evidence of convergent validity of the scales’ psychometric properties. Finally, it must be remembered that though this sample was suitable for the purposes of this study, this was a relatively small-scale study and the results should be treated with caution. Ideally the OAS should be independently tested in a clinical population with a larger retest interval.