Time perspective, mindfulness, and psychological symptoms: psychometric properties of the Turkish short Zimbardo time perspective inventory

Abstract

Background

Our perspective on time constitutes a fundamental aspect of psychological functioning, making the assessment of time perspective critically important. This study aimed to examine the validity and reliability of the Turkish version of the 18-item Zimbardo Time Perspective Inventory – Short Form (ZTPI-Short), which includes the “Future Negative” subscale, and to explore the relationship between mindfulness and time perspective.

Methods

A descriptive, cross-sectional online survey was conducted, and participants were reached through a convenience sampling procedure. A total of 367 participants were included in the analysis. The factor structure of the inventory was evaluated through Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA) and exploratory structural equation model (ESEM).

Results

The mean age of participants was 25.81 (SD = 0.741), and 61.9% were female. Among the participants, 13.9% reported having a physical illness, and 14.7% reported having a psychiatric disorder. EFA supported a six-factor solution explaining 59% of variance. CFA showed limited fit improved by theory-justified residual covariance; ESEM yielded substantially better fit (CFI = 0.960, TLI = 0.896, RMSEA = 0.069, SRMR = 0.026). Under MLR, information criteria favored six over five factors. Ordinal α exceeded 0.70 for most subscales; ω ≥ 0.70 for five subscales, with Present Hedonistic marginal (ω = 0.678) yet supported by mean loading (0.690) and discriminant evidence. Fornell–Larcker and HTMT (< 0.85; highest 0.673 for Past Negative–Future Negative) indicated discriminant validity. Furthermore, mediation analyses demonstrated that mindfulness significantly mediated the relationships between past negative time perspective and depression, and between future negative time perspective and anxiety.

Conclusion

The study provides evidence for the validity and reliability of the Turkish version of the ZTPI-Short. These results underscore the importance of integrating both temporal orientation and mindfulness into psychological assessment and intervention strategies aimed at reducing depression and anxiety symptoms.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40359-025-03699-5.

Introduction

The experience of time constitutes one of the most intricate and fundamental aspects of human psychology. Unlike the classical senses—such as hearing, sight, touch, and taste—which rely on specialized receptors, there is no discrete sensory organ for perceiving time. Nevertheless, the brain operates as a kind of “time machine” that integrates temporal information across experiences [1]. Time is pervasive in human consciousness, providing a sense of continuity, coherence, and meaning. As Loewald aptly noted, all psychic functions are inherently temporal in nature [2]. Our lived experience—and the psychic processes that shape it—emerge within a dynamic interplay of past, present, and future.

Humans typically experience time as a continuous flow, often living through a succession of “presents” that are qualitatively distinct from both the past and the future [3]. Time perspective refers to the largely unconscious cognitive process by which experiences are organized into temporal categories—past, present, or future—thus providing coherence and structure to one’s subjective narrative [4]. In other words, from the time perspective, subjective experiences have a coherent narrative around the axis of past, present, and future, and the construction of a continuous self becomes possible.

The clinical relevance of how individuals experience time—and the emotional valence they assign to different temporal dimensions—has long been a central concern in philosophy and has now been substantiated by empirical research. Numerous studies have demonstrated significant associations between time perspective and a range of psychological symptoms, including depression, anxiety, and substance use disorders [5–8]. With this increasing awareness, measurement methods such as the Adolescent and Adult Time Inventory-Time Attitudes Scale and the Temporal Focus Scale have been developed to assess time perspective [9, 10]. However, due to the narrow-focused nature of these measurements, Zimbardo and Boyd introduced the Zimbardo Time Perspective Inventory (ZTPI) to the literature, which consists of 56 questions and five dimensions, addressing time perspective in broader terms, such as affect, cognition, and behavior [4]. ZTPI conceptualizes five distinct temporal orientations: Past-Positive, Past-Negative, Present-Hedonistic, Present-Fatalistic, and Future. The Past-Positive dimension reflects a nostalgic and affirmative view of the past, whereas the Past-Negative dimension entails a pessimistic and aversive perception of past experiences. Present-Hedonistic orientation emphasizes pleasure-seeking, while Present-Fatalistic reflects a belief that life is controlled by fate and external circumstances. Lastly, the Future dimension captures a goal-directed mindset characterized by planning, self-discipline, and delayed gratification.

Despite wide use, the ZTPI has been criticized for its length and variable factor structure across languages and samples. In response, shorter forms, including 36 items, 25 items, 20 items, 17 items, and 15 items, by removing items with the lowest factor loadings, were proposed [11–15].

While the original model included a singular Future Time Perspective (FTP), later research emphasized the need to differentiate between its positive and negative aspects on this rationale the Swedish ZTPI (S-ZTPI) therefore added Future-Positive and Future-Negative, yielding six subscales [16, 17]. On the other hand, Koštál et al. introduced a six-factor version, with 18 items which incorporates a 3-item Future-Negative dimension to preserve theoretical breadth while improving psychometrics [18]. However, this version showed high correlation between the Past-Negative and Future-Negative subscales highlighting the conceptual overlap between these temporal domains and suggesting the need for further empirical investigation [18].

Based on these considerations, the primary aim of the present study was to examine the validity and reliability of the 18-item, six-factor ZTPI, adapted by Kostal et al., within a Turkish sample, with a particular focus on evaluating the discriminant validity of the Future-Negative subscale.

Additionally, our perception of temporality is closely linked to mindfulness, as the past is perceived as either positive or negative, and the future is perceived as either positive or negative, in the present moment [19]. Within the present context, we may feel nostalgia for the past, or the future may be experienced as a moment to be welcomed. Mindfulness means being fully present in the present moment and observing thoughts and feelings without judgment [20]. Studies have found that mindfulness interventions can be effective in reducing symptoms of anxiety and depression [21]. The second aim of this study was to investigate the effects of time perspectives and mindfulness on anxiety and depression. By demonstrating the relationship between the ZTPI subscales and mindfulness, depression, and anxiety, we aimed to provide additional evidence for the scale’s convergent and criterion validity.

Methods

Participants and procedure

A descriptive, cross-sectional online survey was conducted, and participants were reached through a convenience sampling procedure. Upon receiving approval from the hospital’s ethics committee, 408 medical students enrolled at Gazi University Faculty of Medicine between February 2025 and August 2025 were invited to participate in a web-based, cross-sectional study conducted via SurveyMonkey. Participants were contacted via e-mail, and those who accessed the provided link were presented with an information sheet outlining the study’s purpose and objectives, followed by a consent form that confirmed their voluntary participation. Only students who provided informed consent were included in the study. Electronic consent was obtained from all participants, with assurances that responses would remain confidential and would not be disclosed to third parties. The survey was administered anonymously.

Measures

ZTPI - Short TR: The Zimbardo Time Perspective Inventory (ZTPI) is a 56-item questionnaire developed by Zimbardo and Boyd to assess fundamental time-related dimensions of human behavior [4]. The present study employed the short version tested by Kostal et al. In addition to the original five-subscale model, Kostal and colleagues evaluated an alternative six-factor model that incorporated a three-item “Future Negative” subscale, as proposed by Carelli et al. [17] Although the five-factor model demonstrated slightly better fit indices, the authors acknowledged the high correlation between the Past Negative and Future Negative dimensions, which raised concerns regarding the dimensionality. Nevertheless, they emphasized that the six-factor model remained theoretically and empirically justifiable [18]. Therefore, the current study utilized six subscales: Past Negative, Past Positive, Present Hedonistic, Present Fatalistic, Future Positive, and Future Negative. There are three items for each time perspective, and there are no reverse-coded items. Scale scores were calculated separately.

Depression, Anxiety, Stress Scale 21 (DASS-21): The DASS-21 was developed by Lovibond and Lovibond by selecting items from the DASS-42 to shorten the time [22]. DASS21 contains seven items for each scale. The scale has been shown to have acceptable validity and reliability in Turkish society [23]. In the current study, Cronbach’s alpha values for depression, anxiety, and stress were found to be 0.879, 0.848, and 0.895, respectively.

Mindful Attention Awareness Scale (MAAS): The 15-item Mindful Attention Awareness Scale (MAAS), developed by Brown, was used to assess dispositional awareness and attentiveness to present experiences [24]. The Turkish validity and reliability study of the scale was conducted by Çatak et al. [25]. In the current study, Cronbach’s alpha values for MAAS were found to be 0.879.

Translation Procedure:

Following the receipt of formal permission from the original author, the translation and cultural adaptation process was initiated. The Turkish translation adhered to the methodological guidelines proposed by Wild et al., employing the back-translation technique to ensure linguistic and conceptual equivalence [26]. The Turkish adaptation of the ZTPI-Short was initiated after obtaining formal permission from the original author. The process aimed to achieve semantic, idiomatic, and conceptual equivalence, following the core principles of cross-cultural adaptation. The original English scale was translated into Turkish by a single bilingual translator who is a native Turkish speaker with full professional proficiency in English. To verify the accuracy and conceptual fidelity of this initial translation, a critical quality control step was employed: back-translation. The Turkish version was independently translated back into English by a second bilingual translator, a native English speaker with full professional proficiency in Turkish. This translator was blinded to the original English ZTPI-Short. The research team then meticulously compared the back-translated version with the original scale. This comparison revealed only minor and non-substantive discrepancies, which primarily involved phrasing rather than core meaning. These minor points were reviewed and resolved by the research team to reach a consensus on the final Turkish wording. The expert judgment of the research team, which included specialists in the field, was central to this harmonization process, ensuring the final version was conceptually sound and linguistically appropriate.

Sample Size

Following the recommendation of a sample-to-item ratio between 1.2 and 10, as demonstrated in the validity and reliability tests by Anthoine et al., a minimum of 180 participants were determined for inclusion in the study [27].

Data Analysis

All analyses were conducted in R (version 4.4.3) using the lavaan, psych, semTools, and mirt packages. Data suitability was evaluated with the Kaiser–Meyer–Olkin (KMO) test, Bartlett’s test of sphericity, and item-level sampling adequacy indices. The total sample (N = 367) was randomly split for exploratory (EFA; n = 184) and confirmatory factor analysis (CFA; n = 183).

Analyses consisted of (i) exploratory factor analysis, (ii) confirmatory factor analysis in independent subsamples, (iii) convergent/criterion validity tests, (iv),,reliability and (v) item-level diagnostic statistics.

EFA was performed using polychoric correlations and the Weighted Least Squares Mean and Variance adjusted (WLSMV) extraction method, oblimin rotation, with factor retention guided by parallel analysis and the scree test. CFA was estimated with the robust weighted least squares method (WLSMV) based on polychoric correlations, and competing models were compared using χ² difference testing and information criteria (AIC, BIC). Because the ZTPI items are 5-point Likert-type, WLSMV was preferred in CFA. Since information criteria are not provided in WLSMV, AIC/BIC were calculated under MLR for sensitivity purposes only, and the main model evaluation was performed using WLSMV fit indices. Model fit was assessed using CFI, TLI, RMSEA (90% CI), and SRMR. Modification indices were considered only when theoretically justified. Exploratory structural equation modeling (ESEM) with geomin rotation was conducted as a sensitivity analysis.

Reliability was examined using Cronbach’s α, ordinal α, and McDonald’s ω. Convergent and discriminant validity were evaluated via AVE, the Fornell–Larcker criterion, and HTMT ratios. Item quality was assessed with corrected item–total correlations and item response theory analyses (graded response model). Finally, mediation analyses were performed with Hayes’ PROCESS macro (Model 4). A p-value of < 0.05 was considered statistically significant.

Results

Sample characteristics

408 participants were contacted, 400 of them gave their consent to participate in the study, and 33 of them were not included in the analysis because they filled out the survey data incompletely, 367 were included in the final analysis (response rate: 89.9%). The mean age of participants was 25.81 (SD = 0.741), and 61.9% were female. Among the participants, 13.9% reported having a physical illness, and 14.7% reported having a psychiatric disorder. In terms of living arrangements, 24.5% reported living alone. Additionally, 59.5% of the participants described their income level as below average.

Explanatory factor analysis

An exploratory factor analysis (EFA) was conducted using polychoric correlations and the Weighted Least Squares Mean and Variance adjusted (WLSMV) extraction method with oblimin rotation, appropriate for the ordinal nature of the data and the expectation of correlated latent factors. The suitability of the data for factor analysis was evaluated using the Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy and Bartlett’s test of sphericity. The Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy was 0.67, and Bartlett’s test of sphericity was statistically significant (X² (153) = 1550.6, p < 0.001), indicating that the data were suitable for factor analysis. The examination of the anti-image correlation matrix indicated diagonal values between 0.48 and 0.80.

The number of factors and items to retain was determined based on the following criteria: the eigenvalue cutoff rule, Cattell’s scree test, parallel analysis, item loadings of at least 0.40 on a single factor, and theoretical interpretability of the extracted factors [28–30].

Based on eigenvalues greater than one and Cattell’s scree test, a five- to six-factor solution was considered appropriate. To confirm this, a parallel analysis was conducted using 100 randomly generated datasets. The results revealed that the first six eigenvalues from the actual data exceeded those from the simulated data, supporting the retention of a six-factor structure. Accordingly, an exploratory factor analysis was conducted using polychoric correlations and WLSMW extraction method with oblimin rotation to allow for correlated factors. The analysis yielded six factors accounting for approximately 59% of the total variance. Standardized factor loading was generally strong, with all items loading ≥ 0.50 on their intended factor (Table 1).

Table 1.

Results of explanatory factor analysis (N = 184)

	Factor 1	Factor 2	Factor 3	Factor 4	Factor 5	Factor 6
Item1	0.775
Item 2	0.970
Item 3	0.663
Item 13		0.633
Item 14		0.756
Item 15		0.841
Item 16			0.533
Item 17			0.877
Item 18			0.778
Item 10				0.831
Item 11				0.806
Item 12				0.530
Item 7					0.713
Item 8					0.874
Item 9					0.631
Item4						0.744
Item5						0.599
Item6						0.809

The values ​​presented in the table represent standardized factor loadings. All loadings are statistically significant (p < 0.001). Factor 1: Past Negative, Factor 2: Future Negative, Factor 3: Future Positive, Factor 4: Present Hedonistic, Factor 5: Present Fatalistic, Factor 6: Past Positive

Confirmatory factor analysis

Confirmatory factor analysis (CFA) was conducted on the holdout subsample (n = 183) using the robust weighted least squares estimator (WLSMV) with polychoric correlations, which is recommended for ordered categorical indicators. The five-factor CFA under WLSMV estimation showed poor fit, X²(125) = 639.14, p < 0.001, CFI = 0.839, TLI = 0.803, RMSEA = 0.150 (90% CI [0.139, 0.162]), SRMR = 0.123.The initial six-factor model demonstrated inadequate fit, X² (120) = 317.36, p < 0.001, RMSEA = 0.101 (90% CI [0.087, 0.114]), SRMR = 0.083, CFI = 0.826, TLI = 0.728. Modification indices indicated a large residual covariance between item7 and item8, two items belonging to the same subscale and sharing highly similar content (MI = 81.12). Allowing this covariance to be freely estimated resulted in a notable improvement in model fit, X² (119) = 217.87, p < 0.001, RMSEA = 0.086 (90% CI [0.072, 0.101]), SRMR = 0.073, CFI = 0.873, TLI = 0.837, although the fit indices remained marginal.

However, negative error variances continued to be observed in some indicators, and item9 was also flagged by modification indices, suggesting local misfit and associations with multiple factors. To account for these small but systematic cross-loadings, an exploratory structural equation model (ESEM) with geomin rotation was estimated as a sensitivity analysis. The ESEM solution demonstrated substantially better fit, X² (59) = 85.17, p = 0.015, CFI = 0.960, TLI = 0.896, RMSEA = 0.069 (90% CI [0.043, 0.093]), SRMR = 0.026, while preserving the expected six-factor structure. Cross-loadings were generally small, and the factor configuration was consistent with the theoretical model.

As anticipated, fit indices obtained with WLSMV differed from those based on robust maximum likelihood (MLR), which treats items as continuous. To enable comparison of non-nested alternatives, the models were also estimated with MLR. Information criteria confirmed the superiority of the six-factor solution over the five-factor alternative (AIC = 8749.6 vs. 8788.7; BIC = 8916.7 vs. 8939.8). Model fit indices for CFA and ESEM solutions under WLSMV and MLR estimation are shown in Table 2.

Table 2.

Model fit indices for CFA and ESEM solutions under WLSMV and MLR Estimation

Model	Estimator	X2(df)	CFI	TLI	RMSA	SMRM	AIC	BIC
5factor CFA	MLR	X2(124) = 226.67 P < 0.001	0.886	0.860	0.067 [0.053–0.081]	0.070	8788.7	8939.8
6factor CFA	MLR	X2(119) = 177.54 P < 0.001	0.935	0.917	0.052 [0.035–0.067]	0.063	8749.6	8916.7
5factor CFA	WLSMV	X2(125) = 639.14 P < 0.001	0.839	0.803	0.150 [0.139–0.162]	0.123	-	-
6factor CFA	WLSMV	X2(119) = 217.87 P < 0.001	0.873	0.837	0.086 [0.072–0.101]	0.073	-	-
6factor ESEM	WLSMV	X2(59) = 85.17 p = 0.015	0.960	0.896	0.069 [0.043–0.093]	0.026	-	-

A post-hoc power analysis using RMSEA indicated that the ESEM model (χ²(59) = 85.17, RMSEA = 0.069 [0.043–0.093], N = 184) had a power of 0.447 to reject the close-fit hypothesis (H₀: RMSEA = 0.05), but a power of 0.861 to reject the poor-fit hypothesis (H₀: RMSEA = 0.10; left-tailed). These results suggest that while the sample had limited sensitivity to detect very close fit, it provided strong evidence that the model was not poorly fitting.

Convergent and discriminant validity

Convergent Validity: Convergent validity was evaluated according to the Fornell and Larcker criterion, and the average variance explained (AVE) values ​​are expected to be above 0.50. While AVE values ​​were found above 0.50 for five factors (0.519–0.618), the AVE value for the F4 factor was calculated as 0.495. However, since the McDonald’s Omega value for this factor was 0.678, the average factor loading was 0.690, and the minimum loading was 0.550, convergent validity was at an acceptable level. The MC Donald omega, AVE and average loading values ​​for each factor are shown in Table 3.

Table 3.

Convergent validity

Factor	AVE	Mc Donald’s ω	Mean loading	Minimum loading
F1	0.519	0.736	0.703	0.527
F2	0.586	0.782	0.759	0.646
F3	0.618	0.796	0.784	0.721
F4	0.495	0.678	0.690	0.550
F5	0.514	0.761	0.709	0.555
F6	0.550	0.722	0.735	0.605

Factor 1: Past Negative, Factor 2: Future Negative, Factor 3: Future Positive, Factor 4: Present Hedonistic, Factor 5: Present Fatalistic, Factor 6: Past Positive

AVE average variance explained

Discriminant Validity: For discriminant validity, the Fornell-Larcker criterion and HTMT (Heterotrait-Monotrait) ratios were examined. According to the Fornell-Larcker criterion, the square root of the AVE of a factor is expected to be greater than its correlations with other factors. The analysis results are presented in Table 5, and this criterion was met for all factors. Fornell-Larcker analysis demonstrates that the scale’s six-factor structure possesses discriminant validity. The √AVE values ​​for all factors are greater than the correlations of the relevant factors with other factors. The highest correlation is 0.632 between F1 and F2, which is theoretically significant and reflects the conceptual closeness of the factors. All other correlations are reasonable, supporting the distinction between the factors.

Table 5.

Heterotrait-Monotrait (HTMT) ratio of correlations for discriminant validity assessment

	F1	F2	F3	F4	F5	F6
F1	NA	0.673	0.257	0.116	0.308	0.293
F2	NA	NA	0.475	0.156	0.252	0.457
F3	NA	NA	NA	0.226	0.125	0.161
F4	NA	NA	NA	NA	0.204	0.304
F5	NA	NA	NA	NA	NA	0.167
F6	NA	NA	NA	NA	NA	NA

HTMT (Heterotrait-Monotrait ratio) values are shown above the diagonal. Factor 1: Past Negative, Factor 2: Future Negative, Factor 3: Future Positive, Factor 4: Present Hedonistic, Factor 5: Present Fatalistic, Factor 6: Past Positive

HTMT (Heterotrait-Monotrait) analysis results show that the six-factor structure of the scale has discriminant validity. HTMT values ​​were found below the 0.85 limit value for all factor pairs. The highest HTMT value was calculated as 0.673 between F1 and F2, reflecting a theoretically expected relationship. HTMT values ​​were below 0.50 for all other factor pairs, indicating excellent discrimination. When these findings were evaluated together with the Fornell-Larcker criterion, it was concluded that the discriminant validity of the scale was supported by strong evidence (Tables 4 and 5).

Table 4.

Fornell-Larcker criterion assessment: discriminant validity with square root of AVE (Diagonal) and factor correlations (Off-Diagonal)

	F1	F2	F3	F4	F5	F6
F1	0.720	0.632	−0.279	0.088	0.254	−0.272
F2	0.632	0.765	−0.466	−0.104	0.119	−0.447
F3	−0.279	−0.466	0.786	0.199	−0.112	0.154
F4	0.088	−0.104	0.199	0.704	0.063	0.253
F5	0.254	0.119	−0.112	0.063	0.717	0.120
F6	−0.272	−0.447	0.154	0.253	0.120	0.742

Diagonal elements (in bold) represent the square root of average variance extracted (√AVE). Off-diagonal elements represent factor correlations. Factor 1: Past Negative, Factor 2: Future Negative, Factor 3: Future Positive, Factor 4: Present Hedonistic, Factor 5: Present Fatalistic, Factor 6: Past Positive

Reliability

When ordinal alpha coefficients, which are more appropriate for Likert-type ordinal data, were examined for reliability analysis, it was observed that the F1, F2, F3, F4, and F6 subscales all showed values ​​above the 0.70 threshold (αlpha ordinary = 0.717–0.814). This finding indicates that internal consistency is adequate given the categorical structure of the scale. Second, McDonald’s omega coefficients based on structural equation modeling showed values ​​above the 0.70 threshold for the F1, F2, F3, F5, and F6 subscales (ω = 0.722–0.785). Given that omega coefficients provide more reliable reliability estimates than alpha, these findings support an overall acceptable level of internal consistency for the scale. The F4 factor, with an omega value of 0.678, falls within the acceptable range (0.65–0.70) recommended by Dunn et al. [31] Furthermore, the average factor loading of 0.690 and robust evidence of discriminant validity support the psychometric adequacy of this factor [32].

Item- level analysis

Corrected item–total correlations (CITC) were computed separately within each subscale to assess the consistency of individual items with their respective constructs. CITC values ranged from 0.46 to 0.75, comfortably above the recommended threshold of 0.30. This indicates that all items contributed meaningfully to their intended subscale, with no items showing evidence of redundancy or weak alignment.

To complement these classical indices, item response theory (IRT) analyses were conducted using graded response models (GRM) for each of the six subscales. The GRM provides two key sources of evidence: (a) item discrimination parameters (a), which reflect how well an item differentiates between individuals at different levels of the latent trait, and (b) category threshold parameters (b1–b4), which reflect the points on the latent continuum at which respondents are equally likely to endorse adjacent response categories.

Across subscales, discrimination parameters ranged from a = 1.16 (moderate) to a = 5.37 (very high), showing that most items were highly sensitive in distinguishing respondents with differing levels of the underlying time perspective dimensions. Threshold parameters (b1–b4) were ordered for all items, supporting the proper functioning of the Likert response format and suggesting that response categories were used as intended by participants.

Finally, item-fit statistics (S–X²) indicated that the GRM models fit the data adequately. Most items yielded nonsignificant S–X² values, and where significant values occurred, the effect sizes (RMSEA of S–X²) suggested acceptable local fit. Together, the CITC and GRM results provide converging evidence that the items are psychometrically sound, discriminative, and appropriately scaled, supporting the overall adequacy of the Turkish ZTPI-Short at the item level. The item level analysis results of the items are shown in Table 6.

Table 6.

Item level analysis

	CITC	a	b1	b2	b3	b4	S–X²|	df| |	RMSEA	p-value
Item1	0.691	2.590	−2.134	−0.636	0.017	1.195	4.618	5	0.000	0.464
Item2	0.750	5.368	−1.531	−0.447	−0.112	0.880	6.705	5	0.043	0.243
Item3	0.604	1.852	−1.930	−0.718	−0.263	1.563	8.476	7	0.034	0.292
Item4	0.495	1.576	−3.616	−2.198	−1.099	0.913	4.495	5	0.000	0.481
Item5	0.549	1.859	−2.322	−1.044	0.276	1.891	5.440	6	0.000	0.489
Item6	0.587	2.902	−2.273	−1.194	−0.659	1.176	6.922	5	0.046	0.227
Item7	0.567	2.169	−1.043	0.414	1.375	2.291	7.798	5	0.055	0.168
Item8	0.591	2.501	−0.705	0.656	1.279	2.489	8.866	5	0.065	0.115
Item9	0.467	1.324	−1.079	0.407	1.738	3.036	0.907	7	0.000	0.996
Item10	0.554	2.266	−2.803	−1.843	−0.852	1.246	7.193	5	0.049	0.207
Item11	0.601	3.046	−1.603	−0.489	0.496	2.075	3.409	4	0.000	0.492
Item12	0.477	1.310	−2.689	−1.076	0.203	2.612	6.553	7	0.000	0.477
Item13	0.490	1.440	−1.616	0.332	1.063	2.801	14.695	8	0.068	0.065
Item14	0.626	2.634	−1.404	−0.196	0.477	1.562	3.286	5	0.000	0.656
Item15	0.667	3.080	−1.383	−0.289	0.246	1.471	10.168	6	0.062	0.118
Item16	0.465	1.158	−3.621	−2.496	−1.256	2.058	4.049	6	0.000	0.670
Item17	0.648	3.317	−1.712	−0.895	−0.132	1.595	6.076	4	0.053	0.194
Item18	0.655	2.603	−1.937	−0.936	−0.220	1.710	7.492	4	0.069	0.112

Criterion‑Related validity

Following the establishment of the factor structure, correlation analyses were conducted to examine the associations between the ZTPI-Short TR subscales and psychological constructs, including depression, anxiety, stress, and mindfulness. A strong and statistically significant correlation was observed between the Past Negative subscale and depression, as well as between the Future Negative subscale and anxiety. Furthermore, both the Past Negative and Future Negative subscales were positively and significantly correlated with stress. A low but statistically significant positive correlation was also found between mindfulness and Future Positive. Also significant negative correlation was found mindfulness between past negative. These findings provide preliminary evidence for the construct validity of the ZTPI-Short TR. The detailed correlation coefficients are presented in Table 7.

Table 7.

Correlation table

	Anxiety	Depression	Stress	Mindfulness
F1	0.492**	0.560**	0.546**	−0.351**
F2	0.503**	0.592**	0.540**	−0.425
F3	−0.166**	−0.299**	−0.211**	0.208**
F4	0.015	−0.030	0.009	−0.017
F5	0.028	0.117*	0.006	−0.048
F6	−0.213**	−0.273**	−0.293**	0.135*

Factor 1: Past Negative, Factor 2: Future Negative, Factor 3: Future Positive, Factor 4: Present Hedonistic, Factor 5: Present Fatalistic, Factor 6: Past Positive, *<0.05, **<0.001

Mediation analysis

Mediation analyses were conducted to examine whether mindfulness mediated the relationship between time perspective dimensions and psychological outcomes (depression, anxiety, and stress). In Model 1, where depression was the dependent variable, Past Negative time perspective significantly predicted depression directly (β = 0.64, SE = 0.07, p < 0.001). A significant indirect effect was also observed through mindfulness (β = 0.20, Bootstrap SE = 0.04, 95%CI [0.13, 0.28]). In Model 2, with anxiety as the dependent variable, Future Negative time perspective was a significant direct predictor of anxiety (β = 0.53, SE = 0.07, p < 0.001). Additionally, the indirect effect through mindfulness was significant (β = 0.14, Bootstrap SE = 0.03, 95%CI [0.09, 0.20]). In Model 3, where stress was the outcome, both Past Negative (β = 0.69, SE = 0.07, p < 0.001; indirect effect: β = 0.18, Boot SE = 0.04, 95%CI [0.11, 0.27]) and Future Negative (β = 0.66, SE = 0.08, p < 0.001; indirect effect: β = 0.21, Boot SE = 0.05, 95% CI [0.13, 0.32]) dimensions demonstrated significant direct and indirect effects via mindfulness. Overall, these findings indicate that mindfulness statistically accounts for part of the associations between negative time perspectives and psychological distress, without establishing causal direction. The detailed results are presented in Table 8.

Table 8.

Mediation analysis results

Dependent Variable	Direct Effect			Indirect Effect
	Beta	SE	p	Beta	Boot SE	Boot CI
Model 1. Depression is a dependent variable
Past Negative	0.64	0.07	< 0.001	0.20	0.04	0.13–0.28
Model 2. Anxiety is a dependent variable
Future Negative	0.53	0.07	< 0.001	0.14	0.03	0.9 − 0.20
Model 3. Stress is a dependent variable
Past Negative	0.69	0.07	< 0.001	0.18	0.04	0.11–0.27
Future Negative	0.66	0.08	< 0.001	0.21	0.05	0.13–0.32

Discussion

The primary aim of this study was to validate the Zimbardo Time Perspective Inventory–Short Form (ZTPI-Short), which provides a concise assessment of multidimensional temporal perspectives, within a Turkish cultural context. The Turkish version of the ZTPI-Short comprises 18 items and includes six subscales: Past Negative, Past Positive, Present Hedonistic, Present Fatalistic, Future Positive, and Future Negative. Exploratory factor analysis (EFA) supported the six-factor structure, which explained 53.0% of the total variance, with item loadings ranging from 0.519 to 0.815. CFA findings provided partial support for the scale’s hypothesized six-factor structure but indicated limited model fit. The five-factor solution exhibited poor fit, while the six-factor model was found to be better but still inadequate. Modification indices indicated local dependencies, particularly among items with content similarity (e.g., Items 7 and 8), and adjusting the error covariance improved the fit. However, persistent negative error variances and cross-loadings demonstrated the limitations of the strict CFA approach. ESEM, implemented as a sensitivity analysis, yielded significantly better fit values, supporting the fit of the factor structure to the theoretical model. Furthermore, information measures obtained under MLR confirmed the superiority of the six-factor solution over the five-factor alternative. These results suggest that the six-factor structure is generally valid, but there are content overlaps and local misfits among some items, and that the ESEM approach provides a more appropriate framework. Furthermore, internal consistency reliability was within acceptable limits across all subscales, indicating that the ZTPI-Short Turkish version is a psychometrically sound instrument for assessing time perspective in Turkish populations.

The Heideggerian philosophical perspective, which asserts that human consciousness is fundamentally oriented toward the future, has significantly influenced the increased scholarly interest in the future time perspective (FTP) within the broader time perspective literature [33]. Conceptualizations of FTP have often emphasized future-oriented cognition as a hallmark of human evolution—particularly the capacity for planning and goal-directed behavior [34, 35]. However, it has been highlighted in recent years that considering thinking about the future as a capacity to plan and shape, while ignoring the negative effects of thinking about future events, may be a reductionist perspective on future time perspective [36]. The finding that fear of future threats is a strong predictor of PTSD and depressive symptomatology has raised concerns that the future time perspective may be ambivalent, with positive/negative consequences [37].

Based on these concerns, Carelli et al. expanded the original ZTPI by introducing an 8-item subscale assessing Future Negative, resulting in the development of the Swedish Zimbardo Time Perspective Inventory (S-ZTPI) [17]. When both five-factor and six-factor models (including the Future Negative subscale) were tested using the standard-length ZTPI, model fit indices were found to be comparable. Kostal et al. later developed a short version of the ZTPI by selecting the three items with the highest loadings from the S-ZTPI’s Future Negative subscale, in addition to the original five dimensions. Their analyses indicated that although the five-factor solution yielded better fit indices, the six-factor model still demonstrated acceptable fit [18]. Similarly, Molinari et al. confirmed the six-factor structure in a sample of Italian adolescents and young adults (X²(369) = 574.75, CFI = 0.90, RMSEA = 0.04, SRMR = 0.07), supporting the validity of the Future Negative construct in this population [38]. Lee et al. aimed to develop a shortened version of the S-ZTPI in a South Korean sample but found that a four-factor structure was more appropriate than a six-factor structure [39]. In our study, we observed that the six-factor structure had better fit values. Findings in this area suggest that the factor structure of the ZTPI-Short is sensitive to both cultural and sample characteristics. For example, a six-factor model was validated in a sample of adolescents and young adults in Italy [38], whereas a four-factor solution provided the best fit in a South Korean sample with a broader age distribution [39]. Consistent with the Italian study, our university student sample also supported the six-factor structure. Several factors may account for these discrepancies. First, as Nurmi emphasizes, negative themes about the future may vary by life stage; adolescents and young adults may perceive the future as a dual domain encompassing both hope and anxiety, whereas this distinction may become less pronounced in more heterogeneous age groups [16]. Second, according to Hofstede’s cultural dimensions theory, Turkey and Italy share relatively similar scores on uncertainty avoidance and long-term orientation, both of which may reinforce the coexistence of future-oriented planning and anxiety. At the same time, South Korea differs considerably on these dimensions [40]. Finally, nuances in translation and linguistic adaptation could attenuate the negative affective connotations of the “Future Negative” items in certain languages. Taken together, the validation of the six-factor structure in the Turkish sample aligns with the Italian findings in terms of both age-related characteristics and cultural–linguistic proximity, further underscoring the ZTPI-Short as a culturally sensitive measurement instrument.

Reliability analyses based on coefficients appropriate for ordinal data provided support for the internal consistency of the scale. Ordinal alpha coefficients exceeded the recommended 0.70 threshold across most subscales, indicating satisfactory internal consistency given the categorical response format. Similarly, McDonald’s omega coefficients derived from the structural model were above 0.70 for the majority of subscales, further confirming the robustness of the measurement. Although, the McDonald’s omega coefficient for Factor 4/Present Hedonistic was slightly below the conventional cut-off (ω = 0.678). Although this value suggests somewhat weaker reliability, it still falls within the “marginal but acceptable” range proposed by Dunn et al. [31]. This pattern may reflect the heterogeneous content of these subscales and the limited number of items, both of which are known to attenuate internal consistency [14, 41].

Evidence for the validity of the six-factor structure was further reinforced through the evaluation of convergent and discriminant validity. The Fornell–Larcker criterion provided the first indication of adequacy. For most factors (F2, F3, F5, F6), the square root of AVE exceeded the recommended threshold, indicating that items effectively converged to represent their underlying latent constructs. However, for F4 (Present Hedonistic), the √AVE fell slightly below the conventional 0.70 criterion, suggesting that the items within this factor may require refinement in future studies to more accurately capture the intended construct.

As expected, depression was positively correlated with Past Negative and negatively correlated with Past Positive, supporting prior research findings [42]. Similarly, anxiety was positively associated with Future Negative and negatively associated with Future Positive. Despite a strong correlation between Past Negative and Future Negative, which may raise concerns about discriminant validity, further psychometric analysis supported their distinctiveness [17, 18]. The highest correlation emerged between Past Negative (F1) and Future Negative (F2) (r = 0.632), raising potential concerns about their empirical distinctiveness. However, HTMT ratios provided more stringent evidence, with all values well below the conservative 0.85 benchmark. The highest HTMT value (0.673) was again observed between F1 and F2, consistent with their conceptual relatedness. This pattern indicates that, although theoretically linked, the factors maintain sufficient empirical distinction. Importantly, the moderate correlation between F1 and F2 is not merely a statistical artifact but reflects a meaningful theoretical overlap, whereby individuals who perceive their past negatively are also likely to hold a pessimistic view of the future. The ESEM framework is particularly suitable in this context, as it allows factors to remain statistically distinct while accommodating theoretically meaningful inter-factor correlations. Furthermore, in mediation analyses that included the Future Negative variable as a covariate in the Past Negative → Depression pathway, the effect of Past Negative remained substantial and consistent. This suggests that, at least within the Turkish sample, Past Negative and Future Negative involve distinct phenomenological temporal concepts.

Finally, mediation analyses suggested that mindfulness statistically accounted for part of the association between depression and past negative time perspective, as well as between anxiety and future negative time perspective.

Rumination and worry, two thought patterns with distinct temporal processes, are known to play a significant role in the development of depression and anxiety [43, 44]. Mindfulness exercises are known to be associated with decreased rumination outcomes [45]. Rumination and worry may result from a lack of attentional control [46]. Mindfulness, the antithesis of rumination and worry, may prevent past or future thoughts from becoming automatic by keeping the focus of attention on the present [47]. On the other hand, individuals with high mindfulness are likely to experience less rumination and worry because they consciously observe their experiences rather than getting lost in thoughts [48]. The effects of mindfulness on rumination and worry may illuminate the mediating role of time perspectives between depression and anxiety. The findings of Rönnlud et al. that a high level of mindfulness supports a more balanced time perspective by focusing less on negative aspects of the past and negative expectations about the future may provide insight into the results of our study [49].

This study has several limitations that must be acknowledged. Initial limitations stem from the self-report nature, cross-sectional design, and recruitment method, which preclude causal inferences and the generalizability of our findings to the broader Turkish-speaking population. Future validation of the ZTPI-short TR in clinical populations is still necessary. Another limitation is that we did not assess other relevant psychometric properties of the scale, such as test-retest reliability and predictive validity. Additional studies should consider addressing this limitation. Although the translation procedure followed a forward–backward workflow by bilingual experts, no formal cognitive debriefing or pilot testing was conducted with target participants. This represents a limitation, as item comprehension and cultural equivalence were not systematically verified. Future validation studies should incorporate cognitive interviewing and pretesting steps to strengthen the evidence for full cultural validity.

Conclusion

In conclusion, the present study provides initial evidence for the validity and reliability of the Turkish version of the ZTPI-Short. While our findings support a six-factor solution and highlight the distinctiveness of the Future Negative subscale in this cultural context, caution is warranted given the mixed evidence across international studies. Mediation analyses suggested that mindfulness statistically accounted for part of the associations between negative time perspectives and psychological distress, although causal interpretations cannot be made due to the cross-sectional design. Practically, the six subscales can be interpreted independently in Turkish settings, but the use of composite indices or cut-off scores requires further validation against external behavioral and clinical outcomes.

Authors’ contributions

MCK collected the data. NNT designed and conducted the research strategy, analyzed the data, and wrote the manuscript; ASB and HG consulted in the concept analyses during designing the research strategy and writing the manuscript; İE reviewed and edited the manuscript.

Funding

The authors have not declared a specific grant for this research from any funding agency in the public, commercial, or not-for-profit sectors.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Ethics committee approval for the study was obtained from Gazi University Ethics Committee in accordance with the Declaration of Helsinki (date: 14.01.2025, no: 2025 − 101/01). The participants were informed about the purpose of the study, provided written consent, and were assured that they could withdraw from the study at any time without coercion.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.