Wanting and Liking: Testing the Factor Structure of The Temporal Experience of Pleasure Scale in Major Depression and Community Samples

Abstract

The Temporal Experience of Pleasure Scale (TEPS) is a multidimensional self-report measure that has been used to improve understanding of anticipation (“wanting”) and consummation (“liking”) of reward. The TEPS has been used to assess anhedonia in clinical depression, but its factor structure has not yet been confirmed in this population. This seems important given mixed findings on the model fit and factor structure of the TEPS in other clinical and community samples. To remedy this, the current study used confirmatory factor analysis to test models of the TEPS items across three studies: (a) in adults with major depression (n = 334), (b) in youth with major depression (n = 305), and (c) in a community sample (n = 320). In summary, the model fit of the two-factor TEPS scales was adequate in depressed and community Australian samples. Nevertheless, some items may require removal or revision based on cultural preferences for pleasurable experiences.

Keywords

depression anticipatory pleasure consummatory pleasure Temporal Experience of Pleasure Scale wanting liking confirmatory factor analysis

Anhedonia refers to a lack of positive emotion in response to experiences that otherwise elicit a subjectively rewarding, pleasurable response. Anhedonia is one of the core criteria of major depression, and a feature of depressive disorders in general (American Psychiatric Association, 2013; World Health Organization, 1993). A number of self-report measures have been developed over time to specifically assess anhedonia and the capacity for reward responses (Leventhal et al., 2006; Leventhal et al., 2015; Rizvi et al., 2016). More recently, measures have been informed by research showing distinctions in reward response based on anticipating (wanting) and consuming (liking) reward (Berridge & Robinson, 2003). One of these measures, the Temporal Experience of Pleasure Scale (TEPS; Gard et al., 2006) has become widely used in the study of anticipatory and consummatory pleasure. The initial validation of the TEPS demonstrated its construct validity in assessment of these two aspects of reward response in samples of undergraduate students from the United States.

Since its development the TEPS has been used to study reward response in people with depressive disorders, and has been shown to be sensitive to detecting differences between those with and without depression (Hallford & Sharma, 2019). Despite this wide use, no study has assessed the psychometric properties of the TEPS items nor its factor structure in people with clinical depression. Such an investigation is important, given there have been mixed findings regarding model fit in samples with other psychiatric disorders and community samples. If the model fit in samples of people with depression is poor, then the TEPS may be misspecifying how wanting and liking operates in those with depression. For example, underestimating differences with healthy controls or failing to capture distinct associations between the subscales with related variables, such as motivation, decision making, or behavior.

With regard to previous studies in clinical samples, Favrod et al. (2009) indicated that a two-factor model of the TEPS was an excellent fit to responses from a sample of individuals with schizophrenia using confirmatory factor analysis (CFA). Simon et al. (2018) reported an acceptable, but somewhat poorer, fit of a two-factor model in a joint clinical (schizophrenia/schizoaffective disorder) and healthy control sample. In contrast, Garfield et al. (2016) reported poor fit for a two-factor model of a state version of the TEPS in a sample of individuals with opiate dependence. This fit was improved by removing two items and correlating some residuals.

With regard to community samples, studies have reported mixed findings on the factor structure. In contrast to the initial validation study in English-speaking college students from the United States (Gard et al., 2006), Ho et al. (2015) reported inadequate fit of a two-factor model in samples of adults from the United Kingdom and Australia, as well as an inadequate fit for a four-factor model. They suggested that this was due to a lack of distinction in the latent variables, and that items may have loaded onto multiple factors. Notably, their test of a one-factor model incorporating all items was a poorer fit to the data than multifactorial models. Stratta et al. (2011) found that a three-factor model parsing anticipatory pleasure into abstract and contextual subscales was the best solution for the TEPS items in exploratory factor analysis in Italian teenagers. However, this model was not confirmed in CFA. Chan et al. (2012) found an inadequate fit for the two-factor model in a sample of Chinese college students, but reported an excellent fit using a four-factor model that further parsed the anticipatory and consummatory subscales into abstract (e.g., When something is coming up in my life, I really look forward to it; abstract anticipatory) and contextual item subscales (e.g., The sound of crackling wood in the fire is very relaxing; contextual consummatory). They did note, however, that the abstract and contextual consummatory reward subscales could not be clearly distinguished conceptually. Despite this, subsequent studies by Z. Li et al. (2018) and Zhou et al. (2019) replicated the finding of a superior fit for a four-factor model in Chinese adults. Notably, Z. Li et al.’s (2018) study also used an adult sample from the United States, however, in this sample the four-factor model was not superior to the two-factor model. Furthermore, in Zhou et al.’s study, error terms were correlated in order to achieve good fit in the four-factor model.

In summary, no studies have used CFA to test the factor structure of the TEPS in a clinically depressed sample, although a reasonable to good model fit has been found for a two-factor model in other clinical samples. In contrast, studies in community samples have shown mixed findings on model fit for a two-factor structure. Evidence indicating the superiority of a four-factor model seems limited to Chinese samples and has not been replicated in English-speaking samples.

The aim of this study was to examine the model fit of the TEPS in clinically depressed and community samples. To do this, responses to TEPS items from two clinical samples with a current major depressive episode (MDE) were examined: one of adults aged 18 years and older, and one of youth aged between 15 and 25 years. In addition, responses from one community sample of adults aged 18 years and older were used. Given the findings of previous research, three models were tested and compared in each sample: the original two-factor model, a four-factor model, and a one factor model. In addition, differences between the clinical and community samples on anticipatory and consummatory pleasure were planned as being tested, pending model invariance between the groups. Based on previous findings in clinical samples, it was hypothesized that a two-factor model (anticipatory; consummatory) would be the best fit to the data, and that both subscales on the TEPS would be significantly lower in those with a current MDE relative to the community sample.

Study 1: CFA in Adults With Major Depression

Method

Participants and Procedure

The sample comprised participants who completed the measures for this study as part of a broader baseline questionnaire in a separate, repeated measures study (Hallford et al., 2019; Hallford, Austin, et al., 2020). The TEPS was added to this baseline questionnaire after that study had already commenced. A total of 334 participants completed the TEPS along with other measures. This provided a good case to item ratio (>18:1; Kline, 1998) and adequate sample size with which to accurately estimate parameters (Gagne & Hancock, 2006). The inclusion criteria for that study were ≥18 years of age, residing in Australia, currently experiencing a MDE as determined by the electronic–psychological assessment system (e-PASS; Nguyen et al., 2015) and a score of ≥10 on the Patient Health Questionnaire–9 (PHQ-9; Kroenke et al., 2010) indicating at least moderate depression. The mean score in the current sample was 17.4 (SD = 3.3), indicating moderately severe depressive symptoms.

The sample was on average middle-aged, with a mean age of 46.9 (SD = 12.7; range 18-82). The sample comprised 84.7% women and 14.1% men, with 1.2% reporting their gender identity as “other.” The vast majority identified as Caucasian/White European (94.3%), while 1.2% identified as Asian, and the remaining 4.5% was distributed among other ethnicities. Regarding highest education level attained, 0.3% of participants endorsed primary school, 21.3% high school, 25.1% diplomas/certificates, 39.2% bachelor degrees, and 14.1% postgraduate degrees. Over half of the participants were in paid work (60.1%). Most of the sample reported being on antidepressant medication (86.8%).

Ethics approval was obtained from the University Human Research Ethics Committee prior to data collection commencing. The participants were recruited through advertising on social media platforms (e.g., Facebook, Instagram), and online groups and forums based in Australia. Participants responded to advertisements by following a link to an online survey which contained a plain language statement. Informed consent was provided by participants indicating their agreement through clicking on the continue button after the plain language statement. They then answered questions regarding eligibility and proceeded to complete the larger questionnaire of which the TEPS was part. No incentive for participation was offered.

Materials

Temporal Experience of Pleasure Scale (Gard et al., 2006)

The TEPS consists of 18 self-report questions that assess for trait anticipatory pleasure (10 items) and consummatory pleasure (8 items). Participants are provided with the prompt: “Below you will find a list of statements that may or may not be true for you. Please read each statement carefully and decide how true that statement is for you in general. Please respond to all items.” Participants give their responses on a 6-point scale (1 = very false for me to 6 = very true for me). The item responses are averaged in each subscale so that higher scores indicate a stronger tendency to anticipate or experience pleasure. Table 1 shows the content of all TEPS items.

Table 1.

Descriptive Statistics for the Two-Factor Temporal Experience of Pleasure Scale Across the Three Studies.

TEPS items and which latent factor they predict in the models (two-factor; four-factor)	Sample 1: Adults with MDD (n = 334)				Sample 2: Youth with MDD (n = 305)				Sample 3: Community sample (n = 320)
	M (SD)	Item-total r within subscale	Skewness (SE = 0.13)	Kurtosis (SE = 0.26)	M (SD)	Item-total r within subscale	Skewness (SE = 0.14)	Kurtosis (SE = 0.27)	M (SD)	Item-total r within subscale	Skewness (SE = 0.13)	Kurtosis (SE = 0.27)
1. When something exciting is coming up in my life, I really look forward to it (A; AA)	3.5 (1.4)	.62	0.2	−0.8	3.5 (1.3)	.57	−0.05	−0.9	5.0 (1.1)	.57	−1.3	1.8
2. The sound of crackling wood in the fireplace is very relaxing (C; CC)	4.3 (1.5)	.60	−0.6	−0.7	4.4. (1.5)	.64	−0.7	−0.5	4.8 (1.2)	.63	−1.0	0.4
3. When I think about eating my favorite food, I can almost taste how good it is (A; CC)	3.6 (1.6)	.62	.02	−1.2	3.6 (1.7)	.74	−0.03	−1.3	4.8 (1.1)	.62	−0.9	0.4
4. I love the sound of rain on the windows when I’m lying on my warm bed (C; CA)	4.8 (1.4)	.53	−1.1	0.1	4.9 (1.4)	.60	−1.2	0.5	5.1 (1.1)	.62	−1.4	1.7
5. The smell of freshly cut grass is enjoyable to me (C; CA)	3.7 (1.6)	.67	−.01	−1.1	3.2 (1.6)	.54	0.2	−1.1	4.2 (1.5)	.63	−0.5	−0.7
6. I enjoy taking a deep breath of fresh air when I walk outside (C; CA)	4.0 (1.5)	.65	−0.3	−0.9	3.9 (1.5)	.65	−0.2	−1.0	4.9 (1.2)	.65	−1.0	0.5
7. I don’t look forward to things like eating out at restaurants (R; A)	3.3 (1.6)	.47	0.05	−1.1	3.5 (1.7)	.49	−.09	−1.2	3.9 (1.8)	.38	−0.4	−1.3
8. A hot cup of coffee or tea on a cold morning is very satisfying to me (C; CA)	4.3 (1.6)	.57	−0.6	−.07	4.0 (1.6)	.53	−0.4	−0.9	4.4 (1.6)	.56	−0.9	−0.2
9. I love it when people play with my hair (C; AC)	3.3 (1.9)	.48	0.1	−1.5	4.2 (1.7)	.43	−0.6	−1.2	3.9 (1.8)	.55	−0.3	−1.2
10. I really enjoy the feeling of a good yawn (C; CC)	3.0 (1.5)	.55	0.3	−0.8	3.1 (1.8)	.62	−0.3	−0.9	4.1 (1.4)	.64	−0.4	−0.9
11. When I’m on the way to an amusement park, I can hardly wait to ride the roller coaster (A; AC)	1.8 (1.3)	.31	1.7	2.2	3.1 (1.8)	.46	0.3	−1.2	3.5 (1.8)	.53	−0.08	−1.3
12. I get so excited the night before a major holiday I can hardly sleep (A; AC)	2.8 (1.6)	.58	0.5	−0.9	3.1 (1.7)	.66	0.2	−1.2	4.1 (1.7)	.67	−0.5	−0.9
13. I appreciate the beauty of a fresh snowfall (C; CA)	3.5 (1.8)	.59	−.07	−1.3	3.7 (1.8)	.58	−0.2	−1.3	4.6 (1.4)	.64	−0.9	0.00
14. When I think of something tasty, like a chocolate chip cookie, I have to have one (A; AC)	3.4 (1.6)	.50	0.1	−1.1	3.4 (1.6)	.54	0.07	−1.2	4.0 (1.6)	.59	−0.4	−0.9
15. Looking forward to a pleasurable experience is in itself pleasurable (A; AA)	3.2 (1.4)	.68	0.2	−0.8	3.3 (1.5)	.66	0.1	−0.8	4.6 (1.2)	.66	−0.8	0.08
16. I look forward to a lot of things in my life (A; AA)	2.3 (1.1)	.58	0.9	1.0	2.6 (1.4)	.68	0.7	−0.1	4.6 (1.2)	.63	−0.7	−0.1
17. When ordering something off the menu, I imagine how good it will taste (A; CC)	3.1 (1.4)	.68	0.1	−1.0	3.1 (1.6)	.72	0.2	−1.1	4.4 (1.3)	.67	−0.7	−0.3
18. When I hear about a new movie starring my favorite actor, I can’t wait to see it (A; AC)	2.7 (1.4)	.56	0.6	−0.4	3.1 (1.6)	.56	0.3	−1.0	4.0 (1.5)	.58	−0.4	−0.7
TEPS A Total	3.0 (0.8)	—	0.1	−0.4	3.3 (0.9)	—	0.05	−0.4	4.3 (0.8)	—	−0.5	−0.05
TEPS C Total	3.9 (0.9)	—	0.1	−0.4	4.0 (0.9)	—	−0.3	−0.2	4.5 (0.8)	—	−0.4	−0.1

Note. All statistics correspond to the original two-factor model. Labels for items on the four-factor model are provided purely to indicate which subscale they corresponded to in analyses. TEPS = Temporal Experience of Pleasure Scale; MDD = major depression; SE = standard error; A = anticipatory subscale; C = consummatory subscale; AA = anticipatory abstract; AC = anticipatory contextual; CA = consummatory abstract; CC = consummatory contextual; R = reverse-scored item. All item-total correlations are significant at the p < .001 level.

Electronic–psychological assessment system (Nguyen et al., 2015)

The e-PASS is self-report clinical assessment system designed for online completion. It has 11 items corresponding to a diagnosis of a MDE from the Diagnostic and Statistical Manual of Mental Health Disorder–Fifth edition (DSM-5; American Psychiatric Association, 2013). The questions asked about the frequency of symptoms over the last 2 weeks using a scale from 0 (not at all) to 4 (every day), and one item relating to severity using a 0 (no interference/distress) to 8 (extremely significant interference/distress). An algorithm with branching logic is used to score the e-PASS in concordance with DSM criteria. If participants endorse at least five symptoms as occurring either more days than not or every day, at least one of which is disturbance in mood or anhedonia, and score three or more on the interference/distress question, they are categorized as experiencing an MDE. Diagnosis on the e-PASS for MDE has been shown to correspond with diagnoses made using the structured, face-to-face Mini-International Neuropsychiatric Interview (MINI; κ = .57; Nguyen et al., 2015; Sheehan et al., 1998). To further assess the criterion validity of the e-PASS, a sample of 50 individuals (aged 18 to 65 years) completed the e-PASS online and the MINI over the telephone as part of a study not otherwise related to this series of studies. The concordance between diagnosis on the e-PASS and the MINI was 94% (n = 47/50), indicating the e-PASS had high agreement with a structured interview confirming diagnosis of major depressive disorder.

Patient Health Questionnaire–9 (Kroenke et al., 2010)

The PHQ-9 comprises nine self-report items referring to DSM criteria for a MDE (American Psychiatric Association, 2013). Items refer to the frequency of the symptoms over the last 2 weeks on a scale of 0 (not at all) to 3 (nearly every day), and answers are summed to provide a severity score. The PHQ-9 has shown good psychometric properties for identifying probable depression and assessing the severity of depressive symptoms (Kroenke et al., 2010). The computerized version retains these excellent psychometric properties (Erbe et al., 2016). The internal reliability was acceptable in the current study (Cronbach’s α= .82).

Data Analytic Approach

The CFAs were conducted using the Lavaan package (Rosseel, 2012) in R version 4.0.2 (RStudioTeam, 2020) using a diagonal weighted least square with variance adjusted robust estimator with Pearson correlations to estimate the model. This approach provides less biased estimations of factor loadings in ordinal data compared with maximum likelihood estimation and is robust to nonnormality in the distribution of item responses with samples of this size (C. H. Li, 2016; Moshagen & Musch, 2014). To assess model fit we used the following indices: The chi-square value (CMIN) and corresponding p value (set at p < .05), the relative chi-square statistic (CMIN/degrees of freedom [df]; using a conservative guide of <3.0 for acceptable model fit suggested by Kline (1998), the root mean square error of approximation (RMSEA), the standardized root mean square residual (SRMR), and the comparative fit index (CFI). The latter three fit indices were assessed relative to guidelines provided by Hu and Bentler (1999) of scores that, if approximated, indicate good model fit: RMSEA ≤ .06, SRMR ≤ .09, and CFI ≥ .95. Marsh et al. (2004) contends that such cutoffs should not be used in a hypothesis-testing way, and therefore interpretation of model fit was also guided by criteria that an RMSEA between .06 and .10 and a CFR between .90 and .95 were acceptable, but scores outside of this criteria indicated models were likely to be in need of substantial improvement. McDonald’s ω_t was used to assess the internal reliability of scale items. The TEPS subscales were correlated using Pearson correlation coefficients. Other preprocessing analyses were conducted in SPSS 26.0.

The two-factor model loaded items onto the anticipatory and consummatory subscales consistent with those prescribed by Gard et al. (2006). Consistent with Chan et al. (2012), the four-factor model loaded items onto anticipatory abstract, anticipatory contextual, consummatory abstract, and consummatory contextual subscales, and excluded the single reverse-scored item on the TEPS. Table 1 shows the content of each TEPS item and corresponding subscale.

Results

The results from the CFA (see Table 2, for all fit indices across studies) indicated that the two-factor model was an acceptable fit to the data. The four-factor model had a similar fit to the two-factor model, with indices within acceptable ranges. Notably, the consummatory abstract and consummatory contextual latent variables, and anticipatory context and consummatory context variables had very high standardized covariances (.93, p < .001 and .89, p < .001, respectively). This indicated that most of the variance in responses was shared, and they could not be clearly discriminated. The one-factor model was then tested and was an inadequate fit to the data on several indices. Given this model was nested within the two-factor model, a chi-square test could be conducted to statistically compare their relative fit. Unsurprisingly, this indicated the one-factor model was a significantly poorer fit than the two-factor model, Δχ²(1) = 111.2, p < .001 (ΔRMSEA = .015, ΔSRMR = .009, ΔCFI = .044).

Table 2.

Fit Indices for Confirmatory Factor Analyses.

Model	CMIN	df	CMIN/df	p	RMSEA	RMSEA 90% [CI]	SRMR	CFI
Study 1
Two-factor model	374.1	134	2.79	<.001	.073	[.065, .082]	.069	.903
Four-factor model	338.1	113	2.99	<.001	.077	[.068, .087]	.068	.906
One-factor model	485.3	135	3.59	<.001	.088	[.080, .097]	.078	.859
Study 2
Two-factor model	284.5	134	2.12	<.001	.061	[.051, .071]	.061	.934
Four-factor model	337.9	113	2.99	<.001	.081	[.071, .091]	.072	.895
One-factor model	455.9	135	3.37	<.001	.088	[.080, .097]	.081	.860
Study 3
Two-factor model	362.9	134	2.70	<.001	.073	[.064, .082]	.069	.927
Four-factor model	341.7	113	3.02	<.001	.080	[.070, .089]	.066	.925
One-factor model	557.5	135	4.12	<.001	.099	[.091, .108]	.086	.865

Note. CMIN = chi-square value; df = degrees of freedom; RMSEA = root mean square error of approximation; CI = confidence interval; SRMR = standardized root mean square residual; CFA = confirmatory factor analysis.

In summary, the two-factor model was the best fit to the data statistically and its latent variables were more clearly distinguished than the four-factor model, which showed evidence for collinearity. Descriptive statistics and item-total correlations for the two-factor model are shown in Table 1. The correlation between the two subscales was significant, r(334) = .54, p < .001. The internal reliability was acceptable for the anticipatory (ω = .81) and consummatory subscales (ω = .80). Figure 1 indicates that all items loaded higher than .40 onto their latent variables with the exception of Items 7, 9, and 11 which had loadings <.40. Item 7 is a reverse-coded item, which typically show weaker correlations with other scale items and poorer loadings with latent variables (Swain et al., 2008; Van Sonderen et al., 2013; Woods, 2006). The reasons for poor loadings on Items 9 and 11 are less clear, although Items 11 references anticipation of riding rollercoasters at amusement parks, which is likely to be an infrequent event for respondents and could possibly be confounded with feelings of anxiety about this experience. Notably, while these items did not load adequately on their subscales, removing them made the model fit worse on some indices, CMIN = 292.9 (df = 89, p < .001), CMIN/df = 3.29, RMSEA = .083, 90% confidence interval [CI; .073, .094], SRMR = .069, CFI = .907.

Figure 1.

Two-factor model of the Temporal Experience of Pleasures Scale in adults with major depression showing standardized factor loadngs and error terms.

Study 2: CFA in Youth With Major Depression

Participants and Procedure

The sample of youth were recruited as part of a repeated measures study, separate from that the sample used in Study 1 (Hallford, Carmichael, et al., 2020). Recruitment for this study was still underway at the time of writing. Based on the same sample size estimates from Study 1, we aimed to include at least 300 people that completed the TEPS items. At the time the data was extracted, there were over 300 participants available with this data, therefore proving a final sample of n = 305. The recruitment method and inclusion criteria was the same as Study 1, except that participants were required to be 15 to 25 years old (assessed through self-report).

The sample had a mean age of 19.3 (SD = 3.0; range 15-25). The sample comprised 76.1% women and 15.1% men, 0.7% intersex, and 8.2% reported their gender identity as “other.” The vast majority identified as Caucasian/White European (83.5%), while 6.3% identified as Asian, and the remaining 10.2% was distributed among other ethnicities. Regarding highest education level attained, 4.9% of participants endorsed primary school, 55.7% high school, 17% diplomas/certificates, 20.3% bachelor degrees, and 2% postgraduate degrees. Just under half of the participants were in paid work (46.9%). Most of the sample reported having antidepressant treatment sometime in their past (60.7%).

Materials

The TEPS and e-PASS were used again in this study, but the PHQ-9 was not.

Results

The data analytic approach for CFA was repeated from Study 1. The results indicated that the two-factor model was an acceptable fit to the data. Again, the four-factor model did not fit the data as well as the two-factor model, and the covariances between abstract and contextual latent variables within anticipatory and consummatory pleasure indicated collinearity (both .85, p < .001). The one-factor model was an inadequate fit to the data on several indices. A chi-square test indicated this was a significantly poorer fit than the two-factor model, Δχ²(1) = 171.4, p < .001, and outcomes on the fit indices were notably worse (ΔRMSEA = .028, ΔSRMR = .020, ΔCFI = .074).

In summary, the two-factor model was again the best fit to the data. Descriptive statistics and item-total correlations for the two-factor model are shown in Table 1. The correlation between the two subscales was again significant, r(305) = .48, p < .001. Most items in the two-factor model loaded higher than .40 onto their latent variables (see Figure 2). The exceptions to this were again Items 9 and 11, and removing these items resulted in marginally poorer fit on some indices, CMIN = 238.0 (df = 103, p < .001), CMIN/df = 2.31, RMSEA = .066, 90% CI [.037, .061], SRMR = .061, CFI = .934. The internal reliability was acceptable for the anticipatory pleasure (ω = .86) and consummatory subscales (ω = .79).

Figure 2.

Two-factor model of the Temporal Experience of Pleasures Scale in youth with major depression showing standardized factor loadngs and error terms.

Study 3: CFA in a Community Sample of Adults and Group Comparisons

Participants and Procedure

The participants in this study (n = 320) were recruited through advertising on social media platforms (e.g., Facebook, Instagram) and completed the TEPS along with other measures that were part of a broader cross-sectional study. The sample size estimates from Studies 1 and 2 were used again. The inclusion criteria were ≥18 years of age and residing in Australia. The procedure for online recruitment was otherwise consistent with Studies 1 and 2. The mean age of the sample was 32.4 (SD = 11.6; range 18-84). The sample comprised 62.5% women and 37.2% men, with 0.3% reporting their gender identity as “other.” The vast majority identified as Caucasian/White European (66.2%), while 20.8% identified as Asian, and the remaining 12.9% was distributed among other ethnicities. Regarding highest education level attained, 14.5% endorsed high school, 19.5% diplomas/certificates, 44% bachelor degrees, and 22.1% postgraduate degrees. Most were in paid work (85.3%).

Materials

The TEPS was used again in this study, but not the e-PASS nor the PHQ-9.

Data Analytic Strategy

The analytic strategy for CFA was replicated from Study 1 and 2. Prior to assessing differences in TEPS scores between the clinical and community samples, multigroup analysis was conducted, also using Lavaan package (Rosseel, 2012) in R version 4.0.2 (RStudioTeam, 2020).

Results

The results indicated that the two-factor model was an acceptable fit to the data. Once again, the four-factor model did not fit the data as well as the two-factor model on some indices and collinearity was indicated between several latent variables, including consummatory abstract and consummatory contextual (.89, p < .001). The one-factor model was again an inadequate fit to the data. A chi-square test indicated this was a significantly poorer fit than the two-factor model, Δχ²(1) = 194.6, p < .001, and outcomes on the fit indices were notably worse (ΔRMSEA = .026, ΔSRMR = .017, ΔCFI = .062).

Descriptive statistics and item-total correlations for the two-factor model are shown in Table 1. The correlation between the two subscales was significant, r(320) = .54, p < .001. As indicated in Figure 3, again most items loaded higher than .40 onto their latent variables, with the exception of Items 7 and 11. and removing these items resulted in approximately the same fit with marginally higher or lower scores on some indices, CMIN = 305.8 (df = 103, p < .001), CMIN/df = 2.96, RMSEA = .079, 90% CI [.068, .089], SRMR = .068, CFI = .932. The internal reliability was acceptable for the anticipatory pleasure (ω = .84) and consummatory subscales (ω = .83).

Figure 3.

Two-factor model of the Temporal Experience of Pleasures Scale in a Community Sample showing standardized factor loadngs and error terms.

Comparison Between Clinical and Community Samples

The test for configural invariance of the two-factor model across the three groups indicated that the multiple-group model was an adequate fit to the data, CMIN = 1023.8 (df = 402, p < .001), CMIN/df = 2.54, RMSEA = .070, 90% CI [.064, .075], SRMR = .066, CFI = .921. When testing for metric invariance (i.e., constraining the factor loadings across the groups), the model fit was adequate, CMIN = 970.5 (df = 434, p < .001), CMIN/df = 2.23, RMSEA = .062, 90% CI [.057, .068], SRMR = .072, CFI = .932, but a robust chi-square difference test showed that it was a significantly worse fit than the configural model, Δχ²(32) = 46.6, p = .046. These results indicated that the factor structure was invariant across the groups, but the factor loadings differed across the groups. Given that items contributed differently to the latent variables based on group, the mean scores between the groups could not be validly compared.

Discussion

This study provided evidence of adequate fit of the original two-factor model of the TEPS within clinical samples with major depression and a community sample. The two-factor model appeared adequate for adults with major depression (Study 1) and for youth with major depression (Study 2). The one-factor model was not an adequate fit, and comparatively poorer than the two-factor model. This indicates that the TEPS items can be clearly statistically distinguished into a better fitting model when anticipatory and consummatory items are loaded onto separate latent variables. The four-factor model was either not superior or inferior to the two-factor model across the three studies. The only studies to show a superior fit for a four-factor model in community samples have been in Chinese populations (Chan et al., 2012; Z. Li et al., 2018; Zhou et al., 2019), and therefore this might be related to cultural differences or an artefact of translation of the items. In the current study, a four-factor model did not effectively distinguish between the subscales either, with high covariance between the latent variables in all three studies. Furthermore, the mixture of anticipatory and consummatory pleasure items between the subscales obscures their conceptual distinction. As Chan et al. (2012) suggests, it appears necessary to define what is being experienced for consummatory pleasure and it may not be definable in an abstract sense using the TEPS items.

Although the two-factor model was an adequate fit, across the three studies three items were found to load poorly onto their latent variables. Removing these items had varying effects on the models, but in general lead to marginally worse model fit. As a first step, the reverse-scored Item 7 could be reworded positively (i.e., “I look forward to things like eating out in restaurants”) to potentially improve its loading onto the anticipatory subscale. Future studies might test replacements for the other problematic items which could be psychometrically evaluated alongside the original items. Researchers could approach test redevelopment with the aim of making it culturally specific, or cross-culturally invariant dependent on their aims. For example, in the case of Item 9, also found to be problematic in a previous community sample (Ho et al., 2015), a culturally invariant item might refer to liking the feeling of warm water against one’s skin during bathing. In the case of Item 11, riding rollercoasters may be replaced with a less culturally dependent item, such as excitement about seeing one’s favorite music played live (e.g., classical, jazz, rock, or doom metal). Of course, this culturally invariant approach would require testing across culturally diverse samples.

Notably, no item error terms were correlated in any of the models in this study, although modification indices did indicate that model fit would be improved by doing so. These indices were not interpreted as suggestive of measurement problems per se, but that the items and latent variables shared common aspects of positive emotion or behavior that would give rise to shared variance. Given we had no a priori notions about which items should be correlated or whether cross-loadings should be modelled, we sought to avoid doing this just to improve model fit and the items were instead left as independent predictors which still resulted in an adequate fit.

Given that the groups were noninvariant in terms of their factor loadings, the scores on the TEPS subscales could not be accurately compared in this study. Although previous studies have shown that people with depression score lower on both TEPS subscales than healthy controls (Y. Li et al., 2015; Wu et al., 2017; Yang et al., 2017), the current findings suggest it is possible that factor loading differences may have confounded those results somewhat. It seems important therefore to test that measures of anticipatory and consummatory pleasure have measurement invariance in future studies when group characteristics differ on important dimensions. This invariance would then allow for more accurate comparisons of group differences, as well as reasons for any observed impairments, such ability to imagine detail and use of mental imagery for positive future events in depression (Hallford et al., 2018; Hallford, Barry, et al., 2020).

Strengths of this study include being the first to assess this measure of anhedonia in major depression and the use of two large samples incorporating adults and youth. Limitations included having primarily well-educated Caucasian/White Europeans in the clinical samples, and therefore there is some uncertainty about how the findings will generalize to people that identify otherwise. Furthermore, the community sample was not assessed for major depression, and therefore this group was likely to include some participants with clinically significant depressive symptoms. Other factors known to be associated with anhedonia, but not assessed in this study, included substance use (e.g., Garfield et al., 2016) and schizophrenia (American Psychiatric Association, 2013). Therefore, the influence of such factors or comorbidities remains unknown.

In conclusion, these findings support the TEPS two-factor model as adequate for use in those with major depression and community samples. Several of the items did not operate adequately though and were poor predictors of the latent constructs. Future studies may develop and test more culturally invariant or culturally appropriate items dependent on context.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

David John Hallford

References

American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.). Author.

Berridge

K. C.

Robinson

T. E.

(2003). Parsing reward. Trends in Neuroscience, 26(9), 507-513. https://doi.org/10.1016/S0166-2236(03)00233-9

Chan

R. C.

Shi

Y. F.

Lai

M. K.

Wang

Y. N.

Wang

Kring

A. M.

(2012). The Temporal Experience of Pleasure Scale (TEPS): Exploration and confirmation of factor structure in a healthy Chinese sample. PLOS ONE, 7(4), Article e35352. https://doi.org/10.1371/journal.pone.0035352

Erbe

Eichert

H. C.

Rietz

Ebert

(2016). Interformat reliability of the Patient Health Questionnaire: Validation of the computerized version of the PHQ-9. Internet Interventions, 5(September), 1-4. https://doi.org/10.1016/j.invent.2016.06.006

Favrod

Ernst

Giuliani

Bonsack

(2009). Validation of the Temporal Experience of Pleasure Scale (TEPS) in a French-speaking environment. Encephale, 35(3), 241-248. https://doi.org/10.1016/j.encep.2008.02.013

Gagne

Hancock

G. R.

(2006). Measurement model quality, sample size, and solution propriety in confirmatory factor models. Multivariate Behavioral Research, 41(1), 65-83. https://doi.org/https://doi.org/10.1207/s15327906mbr4101_5

Gard

D. E.

Gard

M. G.

Kring

A. M.

John

O. P.

(2006). Anticipatory and consummatory components of the experience of pleasure: A scale development study. Journal of Research in Personality, 40(6), 1086-1102. https://doi.org/10.1016/j.jrp.2005.11.001

Garfield

J. B.

Cotton

S. M.

Lubman

D. I.

(2016). Psychometric properties, validity, and reliability of the Temporal Experience of Pleasure Scale state version in an opioid-dependent sample. Drug and Alcohol Dependence, 161(April), 238-246. https://doi.org/10.1016/j.drugalcdep.2016.02.011

Hallford

D. J.

Austin

D. W.

Raes

Takano

(2019). Computerised memory specificity training (c-MeST) for the treatment of major depression: A study protocol for a randomised controlled trial. BMJ Open, 9(2), Article e024508. https://doi.org/10.1136/bmjopen-2018-024508

10.

Hallford

D. J.

Austin

D. W.

Takano

Fuller-Tyszkiewicz

Raes

(2020, February 27). Computerised Memory Specificity Training (c-MeST) for Major Depression: A randomised controlled trial. Behaviour Research and Therapy, 136(January), Article 103783. https://doi.org/10.1016/j.brat.2020.103783

11.

Hallford

D. J.

Austin

D. W.

Takano

Raes

(2018). Psychopathology and episodic future thinking: A systematic review and meta-analysis of specificity and episodic detail. Behaviour Research and Therapy, 102(March), 42-51. https://doi.org/10.1016/j.brat.2018.01.003

12.

Hallford

D. J.

Barry

T. J.

Austin

D. W.

Raes

Takano

Klein

(2020). Impairments in episodic future thinking for positive events and anticipatory pleasure in major depression. Journal of Affective Disorders, 260(January), 536-543. https://doi.org/10.1016/j.jad.2019.09.039

13.

Hallford

D. J.

Carmichael

A. M.

Austin

D. W.

Takano

Raes

Fuller-Tyszkiewicz

(2020). A study protocol for a randomised trial of adjunct computerised memory specificity training (c-MeST) for major depression in youth: Targeting cognitive mechanisms to enhance usual care outcomes in mental health settings. Trials, 21(1), Article 85. https://doi.org/10.1186/s13063-019-4036-6

14.

Hallford

D. J.

Sharma

M. K.

(2019). Anticipatory pleasure for future experiences in schizophrenia spectrum disorders and major depression: A systematic review and meta-analysis. British Journal of Clinical Psychology, 58(4), 357-383. https://doi.org/10.1111/bjc.12218

15.

P. M.

Cooper

A. J.

Hall

P. J.

Smillie

L. D.

(2015). Factor structure and construct validity of the temporal experience of pleasure scales. Journal of Personality Assessment, 97(2), 200-208. https://doi.org/10.1080/00223891.2014.940625

16.

L. T.

Bentler

P. M.

(1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1-55. https://doi.org/https://doi.org/10.1080/10705519909540118

17.

Kline

R. B.

(1998). Principles and practice of structural equation modeling. Guilford Press.

18.

Kroenke

Spitzer

R. L.

Williams

J. B.

Lowe

(2010). The Patient Health Questionnaire Somatic, Anxiety, and Depressive Symptom Scales: A systematic review. General Hospital Psychiatry, 32(4), 345-359. https://doi.org/10.1016/j.genhosppsych.2010.03.006

19.

Leventhal

A. M.

Chasson

G. S.

Tapia

Miller

E. K.

Pettit

J. W.

(2006). Measuring hedonic capacity in depression: A psychometric analysis of three anhedonia scales. Journal of Clinical Psychology, 62(12), 1545-1558. https://doi.org/10.1002/jclp.20327

20.

Leventhal

A. M.

Unger

J. B.

Audrain-McGovern

Sussman

Volk

H. E.

Strong

D. R.

(2015). Measuring anhedonia in adolescents: A psychometric analysis. Journal of Personality Assessment, 97(5), 506-514. https://doi.org/10.1080/00223891.2015.1029072

21.

C. H.

(2016). Confirmatory factor analysis with ordinal data: Comparing robust maximum likelihood and diagonally weighted least squares. Behavior Research Methods, 48(3), 936-949. https://doi.org/10.3758/s13428-015-0619-7

22.

Mou

Jiang

Yang

Shen

Jin

Dai

Liu

Mao

Zhang

Yuan

(2015). A comparative study of anhedonia components between major depression and schizophrenia in Chinese populations. Annals of General Psychiatry, 14(1), Article 24. https://doi.org/10.1186/s12991-015-0061-3

23.

Shi

H. S.

Elis

Yang

Z. Y.

Wang

Lui

S. S. Y.

Cheung

E. F. C.

Kring

A. M.

Chan

R. C. K.

(2018). The structural invariance of the Temporal Experience of Pleasure Scale across time and culture. Psych Journal, 7(2), 59-67. https://doi.org/10.1002/pchj.207

24.

Marsh

H. W.

Hau

K. T.

Wen

(2004). In search of golden rules: Comment on hypothesis-testing approaches to setting cutoff values for fit indices and dangers in overgeneralizing Hu and Bentler’s (1999) findings. Structural Equation Modeling, 11(3), 320-341. https://doi.org/10.1207/s15328007sem1103_2

25.

Moshagen

Musch

(2014). Sample size requirements of the robust weighted least squares estimator. Methodology: European Journal of Research Methods for the Behavioral and Social Sciences, 10(2), 60-70. https://doi.org/10.1027/1614-2241/a000068

26.

Nguyen

D. P.

Klein

Meyer

Austin

D. W.

Abbott

J. A.

(2015). The diagnostic validity and reliability of an internet-based clinical assessment program for mental disorders. Journal of Medical Internet Research, 17(9), Article e218. https://doi.org/10.2196/jmir.4195

27.

Rizvi

S. J.

Pizzagalli

D. A.

Sproule

B. A.

Kennedy

S. H.

(2016). Assessing anhedonia in depression: Potentials and pitfalls. Neuroscience & Biobehavioural Reviews, 65(June), 21-35. https://doi.org/10.1016/j.neubiorev.2016.03.004

28.

Rosseel

(2012). Lavaan: An R package for structural equation modeling and more (Version 0.5–12 [BETA]). Ghent University.

29.

RStudioTeam. (2020). RStudio: Integrated development for R. http://www.rstudio.com/

30.

Sheehan

D. V.

Lecrubier

Sheehan

K. H.

Amorim

Janavs

Weiller

Hergueta

Baker

Dunbar

G. C.

(1998). The Mini-International Neuropsychiatric Interview (M.I.N.I.): The development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. Journal of Clinical Psychiatry, 59(Suppl. 20), 22-33.

31.

Simon

J. J.

Zimmermann

Cordeiro

S. A.

Maree

Gard

D. E.

Friederich

H. C.

Weisbrod

Kaiser

(2018). Psychometric evaluation of the Temporal Experience of Pleasure Scale (TEPS) in a German sample. Psychiatry Research, 260(February), 138-143. https://doi.org/10.1016/j.psychres.2017.11.060

32.

Stratta

Pacifico

Riccardi

Daneluzzo

Rossi

(2011). Anticipatory and consummatory pleasure: Validation study of the italian version of the Temporal Experience of Pleasure Scale. Journal of Psychopathology, 17, 322-327.

33.

Swain

S. D.

Weathers

Niedrich

R. W.

(2008). Assessing three sources of misresponse to reversed Likert items. Journal of Marketing Research, 45(1), 116-131. https://doi.org/10.1509/jmkr.45.1.116

34.

Van Sonderen

Sanderman

Coyne

J. C

. (2013). Ineffectiveness of reverse wording of questionnaire items: Let’s learn from cows in the rain. PLOS ONE, 8(7), Article e68967. https://doi.org/10.1371/journal.pone.0068967

35.

Woods

C. M.

(2006). Careless responding to reverse-worded items: Implications for confirmatory factor analysis. Journal of Psychopathology and Behavioral Assessment, 28(3), Article 186. https://doi.org/10.1007/s10862-005-9004-7

36.

World Health Organization. (1993). The ICD-10 classification of mental and behavioural disorders. Author.

37.

Mata

Furman

D. J.

Whitmer

A. J.

Gotlib

I. H.

Thompson

R. J.

(2017). Anticipatory and consummatory pleasure and displeasure in major depressive disorder: An experience sampling study. Journal of Abnormal Psychology, 126(2), 149-159. https://doi.org/10.1037/abn0000244

38.

Yang

X. H.

Tian

Wang

D. F.

Wang

Cheung

E. F. C.

Xie

G. R.

Chan

R. C. K.

(2017). Anhedonia correlates with abnormal functional connectivity of the superior temporal gyrus and the caudate nucleus in patients with first-episode drug-naive major depressive disorder. Journal of Affective Disorders, 218(August), 284-290. https://doi.org/10.1016/j.jad.2017.04.053

39.

Zhou

Liu

Fan

Xia

Zhu

(2019). The Temporal Experience of Pleasure Scale (TEPS): Measurement invariance across gender in Chinese University Students. Frontiers in Psychology, 10, Article 2130. https://doi.org/10.3389/fpsyg.2019.02130