A systematic search was conducted in PubMed, Web of Science (WoS) and Scopus databases. The search was limited to scientific articles published in peer-reviewed journals, written in English and Spanish. To identify all relevant articles published between 2013 and 2023 the search was conducted using the term related to DG (using the operator “or”) and the addition to the term “anxiety” with the operator “and,” in title, abstract or key words. The variety of terminology employed throughout the literature was chosen with the objective of ensuring inclusion of all potentially pertinent studies in the review. The terms referring DG were (‘discounting’, ‘reward delay’, ‘delay of gratification’, ‘delayed reinforcement’, ‘intertemporal choice’ and ‘intertemporal decision’).

The following criteria was set: Only empirical papers published in peer-reviewed journals and focused on human samples were included. They must show a clear description of definitions and assessments of the constructs of interest, hitherto explained, anxiety and delay of gratification or any of their names, provided a proper definition.

As inclusion criteria, Delay of Gratification was defined as any test in which subjects were presented with the option of receiving a smaller, immediate reward or a larger but delayed reward instead. Additionally, Anxiety measurement was included only if a standardised psychometric test was applied, or a direct expression of the subject was recorded.

Apart from not meeting the inclusion criteria, exclusion criteria included non-pair-review-type papers, defining concepts in terms not related to behavioural research, not including clear definition of constructs of interest, or not explicitly addressing the relation between Gratification Delay and Anxiety.

The selection of documents for review was conducted by a single reviewer basing on search strategy (Figure 1). The overall review process was supervised by two independent reviewers.

The following terms were considered: ‘delay discounting’ and ‘temporal discounting’, which refer to the devaluation of reinforcement as it is delayed (Amlung et al., 2019); ‘intertemporal choice/decision’ (Kalenscher and Pennartz, 2008; Marchetti et al., 2014) which, in a more general sense, examines the discrepancy in the perceived value of reinforcement contingent on its temporal positioning; and ‘Delay of Gratification’ or ‘Reward Delay’, which underscores an individual’s capacity to defer gratification (Fernández García et al., 2021).

The final search was conducted on 14 January 2024. A total of 205 results were retrieved from the WoS database, 163 from Scopus, and 52 from PubMed. A total of 157 papers were identified as duplicates and subsequently excluded. The remaining 263 abstracts and methods were then reviewed to ascertain whether they met the inclusion criteria. A total of 177 papers were excluded from further analysis, as they did not meet the pre-established criteria. Among excluded reports, 88 were deemed irrelevant to the topic, primarily due to the non-behavioural application of the term ‘discounting’, 46 studies were found to have employed an invalid measure for anxiety, as defined by established criteria, 23 studies did not studied human subjects, 14 were excluded as they differed in design from those for which this review was conducted, four were not scientific articles, and two studies did not assess delay of gratification. The remaining 86 papers were then subjected to a further assessment to determine their eligibility. A further examination of the reports led to the exclusion of 54 papers on the grounds that they did not explicitly assess the relationship between delayed gratification and anxiety, and a further four on the basis that they addressed concepts that were incompatible with the objective of this review. Three articles were excluded from this review due to unavailability of the necessary resources. The remaining 25 articles were subjected to a process of data extraction, the results of which are presented in Table 1.

The extracted data included the following characteristics: sample size, age and standard deviation, nationality, sex ratio, and diagnosis; study type; measure of anxiety; measure of delay discounting; and main outcomes of interest.

The present review identified 18 studies that employed an observational methodology, six with experimental designs, and a last one of a mixed methods approach.

Of the 25 studies included, only one employed a longitudinal methodology, while the remainder employed a cross-sectional approach. Anxiety was assessed using a variety of instruments, with the most frequently employed being the State–Trait Anxiety Inventory (7; Spielberger et al., 1983) the Depression Anxiety and Stress Scale (4; Lovibond and Lovibond, 1995), the Generalised Anxiety Disorder Scale (4; Johnson et al., 2019), and the Patient Health Questionnaire (4; Arroll et al., 2010). With regard to the assessment of DG, an Ad Hoc design was utilised in the majority of studies (10), followed by the Monetary Choice Questionnaire (5; Kirby and Maraković, 1996) and the Adjusting Delay Discounting Task (4; Koffarnus and Bickel, 2014).

The total number of subjects included in the 25 studies reviewed was 12,728, with an age range of 13 to 69 years and a sex ratio that varied from 0 to 100% female. The samples encompass a diverse range of nationalities, with the most prevalent being the United States and China (Campbell and Egede, 2023; Haines et al., 2020; Jiao et al., 2022; Li et al., 2023; Minhas et al., 2020; Steinglass et al., 2017; Worthy et al., 2014; Xia et al., 2017, 2023; Zampella and Benau, 2022; Zhang et al., 2020) followed by Italy, the United Kingdom, and Vietnam (Ho et al., 2023; Nigro et al., 2017; Robertson et al., 2023).

The ethnical consideration of subjects varied widely among studies. Three of them (Gui et al., 2023; Hurlemann et al., 2019; Levitt et al., 2023) only specified the geographical origin of the sample, rather than nationality or ethnicity, while others chose to specify the ethnic background of the sample- Within that latter type, the majority of subjects were Caucasian, although there were also smaller percentages of Hispanic, Asian, and African American populations (DeRosa et al., 2023; Gerhart et al., 2016; Jenks and Lawyer, 2015; Levin et al., 2018; Tanovic et al., 2018; Zhao et al., 2015). Finally, two studies did not report this information (Morasco et al., 2019; Patt et al., 2021).

Lastly, regarding the clinical aspect of samples, a total of 15 samples were drawn from the general population (Gui et al., 2023; Ho et al., 2023, p. 23; Jiao et al., 2022; Levin et al., 2018; Levitt et al., 2023; Li et al., 2023, p. 23; Patt et al., 2021; Robertson et al., 2023; Tanovic et al., 2018; Worthy et al., 2014; Xia et al., 2017, 2023; Zampella and Benau, 2022; Zhang et al., 2020; Zhao et al., 2015), and only nine reports were included focusing on clinical or subclinical samples with various pathologies (Campbell and Egede, 2023; DeRosa et al., 2023; Gerhart et al., 2016; Haines et al., 2020; Hurlemann et al., 2019; Jenks and Lawyer, 2015; Minhas et al., 2020; Morasco et al., 2019; Nigro et al., 2017; Steinglass et al., 2017).

Of the studies reviewed, 16 provided evidence in favour of the existence of a positive association between anxiety and DD. This evidence suggests that higher levels of anxiety are related to higher discount rate, which means a lower ability of subjects to delay gratification. Significant associations supporting this hypothesis were identified in 12 of the papers (Campbell and Egede, 2023; Gerhart et al., 2016; Ho et al., 2023; Jiao et al., 2022; Levin et al., 2018; Levitt et al., 2023; Morasco et al., 2019; Nigro et al., 2017; Xia et al., 2017, 2023, p. 23; Zampella and Benau, 2022; Zhang et al., 2020), Two studies did not provide the corresponding statistical analysis (Hurlemann et al., 2019; Minhas et al., 2020) while two others reported non-significant associations (Tanovic et al., 2018; Worthy et al., 2014).

Two of the studies reviewed yielded evidence that contradicts the hypothesis proposed in this study. In contrast with the hypothesis proposed in this study, Li et al. (2023) discovered negative, albeit non-significant, correlations between Immediate Choice Ratios (ICR) calculated from the MCQ test and anxiety sensitivity obtained from the Childhood Anxiety Sensitivity Index instrument. This suggests that as anxiety sensitivity increases, the percentage of times the immediate option is selected decreases. Conversely, Jenks and Lawyer (2015) observed no significant increase in the discount rate in either the control or experimental group when anxiety was induced through public speaking in subjects with social anxiety disorder.

With regard to the remaining articles under review, the results are inconclusive (DeRosa et al., 2023). Furthermore, the interactions between factors are complex and require further elucidation. Haines et al. (2020) observed that high state anxiety scores were a reliable predictor of a devaluation of reinforcement in participants. Moreover, anxiety was demonstrated to act as a robust mediator between impulsivity and discounting rates, rather than exerting a direct influence on discounting rates. A statistical analysis of all participants in the Steinglass et al. (2017) study revealed a negative correlation between discounting rates and anxiety. However, the group with a diagnosis of Social Anxiety Disorder exhibited higher delay discounting rates than healthy controls. In contrast, the experimental design proposed by Patt et al. (2021) to investigate intertemporal choices across an experiential and a hypothetical task revealed disparate associations between state anxiety and trait anxiety with these tasks. Specifically, state anxiety was found to be positively associated with the discount rate in the experiential task, while no correlation was observed between the discount rates of the experiential and hypothetical tasks.

The study by Robertson et al. (2023) revealed no correlation between anxiety and discount rate. However, the researchers did identify a relationship between anxiety and temporal representation through language. In contrast, the study by Gui et al. (2023) employed the Drift Diffusion Model (DDM) to investigate delay discounting. Their findings indicated a significant correlation between state anxiety and the weight assigned to monetary differences in decision-making. This suggests that individuals with high state anxiety may devalue monetary differences and exhibit increased impulsivity.

In the study by Zhao et al. (2015) a significant interaction between state and trait anxiety was observed. Individuals with high state and trait anxiety exhibited a tendency to choose delayed rewards, whereas those with high trait anxiety but low state anxiety demonstrated a preference for more immediate rewards.

The demographic characteristics of some studies compromise the external validity of the results. Of the 25 studies reviewed, 6 did not reach a sample size of 100 (Hurlemann et al., 2019; Morasco et al., 2019; Patt et al., 2021; Tanovic et al., 2018; Worthy et al., 2014; Xia et al., 2017) and 4 only slightly exceeded it (Jenks and Lawyer, 2015; Steinglass et al., 2017; Zhang et al., 2020; Zhao et al., 2015) which could jeopardize their representativeness.

A mere six of the papers subjected to review opted for an experimental design (Hurlemann et al., 2019; Jenks and Lawyer, 2015; Tanovic et al., 2018; Worthy et al., 2014; Xia et al., 2017; Zhao et al., 2015). The interpretation of results in observational studies is susceptible to bias due to the lack of control over variables. Additionally, the six papers included in this review have relatively small sample sizes, which could introduce further bias into the overall results.

One of the characteristics extracted from the studies presented in Table 1 is the diagnosis of the sample. It should be noted, however, that two of the nine studies with clinical samples did not carry out an actual diagnostic process of the pathology (Jenks and Lawyer, 2015; Nigro et al., 2017). Rather, the sample was selected or assigned to the control or experimental group based on the scores obtained in validated questionnaires. The remaining studies reviewed did, however, carry out this process (Campbell and Egede, 2023; DeRosa et al., 2023; Gerhart et al., 2016; Haines et al., 2020; Hurlemann et al., 2019; Morasco et al., 2019; Steinglass et al., 2017). The aforementioned variability in sample selection may result in the drawing of conclusions regarding the population in question that are biased in one way or another.

The incorporation of a control group is a crucial element in research, as it enhances the reliability and validity of the findings. A review of the 24 studies revealed that 14 of them lacked a control group, thereby preventing a comparison of the results of their observations (DeRosa et al., 2023; Gerhart et al., 2016; Ho et al., 2023; Levin et al., 2018; Levitt et al., 2023; Li et al., 2023; Nigro et al., 2017; Patt et al., 2021; Tanovic et al., 2018; Worthy et al., 2014; Xia et al., 2017, 2023; Zampella and Benau, 2022; Zhang et al., 2020). Although the majority of the studies were conducted in the general population, the inclusion of a comparative group outside the study population could have been beneficial.

A significant source of bias is the considerable heterogeneity in the assessment of both DD and anxiety across the reviewed studies. On the one hand, the majority of studies employ an ad hoc design for the assessment of DD. For the remaining studies, only two tests are utilised by more than one study. While the fundamental paradigm of the tests is consistent, the specific characteristics vary, which may influence the results, as illustrated in the study by Patt et al. (2021). A similar phenomenon is observed in the anxiety measures. Despite the use of standardised measures in all but one study (Tanovic et al., 2018) the variability in measurement methods considerably reduces the reliability of the results.

A lower DD rate among low-anxious versus high-anxious subjects has been found in 10 out of 15 general population studies (Ho et al., 2023, p. 23; Jiao et al., 2022; Levin et al., 2018; Levitt et al., 2023; Tanovic et al., 2018; Worthy et al., 2014; Xia et al., 2017, 2023; Zampella and Benau, 2022; Zhang et al., 2020). These studies provide evidence in favour of the existence of a positive relationship between anxiety and DD. Furthermore, additional noteworthy findings pertaining to this relationship have been identified in four of the remaining five articles (Gui et al., 2023; Patt et al., 2021; Robertson et al., 2023; Zhao et al., 2015). While these findings do not directly indicate a direct correlation between DD and anxiety, they do contribute valuable insights that enhance the comprehension of this relationship. In relation to the general population, the evidence presented in this review is sufficient to affirm that there is a positive association between DD and anxiety in people with gambling disorder. In other words, elevated anxiety levels are associated with reduced capacity to delay gratification in individuals, a phenomenon that appears to be generalisable across different ethnicities and nationalities, as well as to both sexes. This effect is observed consistently across studies, despite the wide range of anxiety and DD measures employed. These findings provide preliminary evidence for this relationship in the general population, although further research is needed to substantiate these observations.

The three studies on medical illnesses included in this review indicate that anxiety is positive related to DD among these subjects. This finding is analogous to that observed in the general population sample. These results align with those reported in previous research (Craft et al., 2020; Epstein et al., 2021). They suggest that this construct is a fundamental aspect to be considered in the development of more effective intervention protocols that take into account aspects of patients’ psychology.

The close relationship between delay discounting and patients with various addictions is well established in the literature (MacKillop et al., 2011; Weinsztok et al., 2021). Nevertheless, further research is required to substantiate the existence of an altered pattern of delay discounting that is a common feature of all addictive disorders. Furthermore, it would be beneficial to incorporate a range of terminologies beyond DD to ensure the inclusion of diverse approaches to the study of this paradigm. The DD paradigm has facilitated a more nuanced understanding of addictive mechanisms in both substance and non-substance addictions. The direction of this relationship is not yet entirely clear, although this is beyond the scope of this review. However, the results of the three studies included in this review related to addictions are in line with studies specific to this subject (Amlung et al., 2017; Konecky and Lawyer, 2015).

The study of the role of DD among people with social anxiety disorder has yielded conflicting evidence (Armstrong and Hoge, 2024; Boyd et al., 2024). In our analysis, two of the three studies have indicated a positive correlation between SAD and DD. Both Hurlemann et al. (2019) and Steinglass et al. (2017) identified a positive correlation between a SAD diagnosis and a higher discounting rate. This indicates that individuals with this diagnosis discount delayed reinforcement more strongly, thereby demonstrating a preference for immediate and smaller rewards. Conversely, Jenks and Lawyer (2015) employed an experimental design in which anxiety was induced through a public speaking task and observed no variation in discounting rates in either the experimental group with this diagnosis or the control group. A synthesis of these findings suggests that the increase in the discount rate may not be a consequence of anxiety, but rather a potential causal factor. While correlational studies offer valuable insights into a phenomenon, experimental designs are essential for elucidating the underlying mechanisms of observed correlations.

The interaction between state and trait anxiety (Zhao et al., 2015) offers an intriguing perspective on the role of DD in anxiety experience. Therefore, individuals with high trait anxiety may be more susceptible to exhibiting a risk perception bias for delayed rewards during periods of elevated anxiety compared to periods of low anxiety. Consequently, they may opt for immediate, albeit modest rewards due to the fear of missing out on larger, albeit more delayed rewards. Those with low trait anxiety are less susceptible to this bias, demonstrating superior performance even in high-anxiety contexts (Kim et al., 2020; Notebaert et al., 2016). Haines et al. (2020) propose that there may be an interaction between individuals’ experienced anxiety and their impulsivity-trait, with the latter influencing the speed at which subjects discount rewards. This implies that the devaluation of reinforcement shown by individuals with high state anxiety could produce more random responses in the DD paradigm, which would limit the validity of the results.

Conversely, it is conceivable that individuals with elevated anxiety levels may exhibit heightened discounting tendencies, not directly, but through a more distanced representation of the future that mediates the experience of anxiety and DD (Robertson et al., 2023). The role of temporal representation as a mediator requires further investigation in future studies. The direction of causality in this relationship, namely whether anxiety produces this temporal distortion or whether it is the temporal distortion that influences the development and experience of anxiety, remains unclear. Furthermore, this relationship has been studied through the lens of language. While the relationship between the use of verbal forms and decision-making is assumed (Banerjee and Urminsky, 2022), it would be beneficial to explore other paradigms.

An alternative and more recent approach to the delay discounting paradigm is the Drift Diffusion Model (DDM), which integrates a greater number of variables than the traditional DD paradigm. The traditional DD paradigm employs the parameter k (discount rate) to evaluate this process. However, this parameter does not adequately reflect the complexity of the trade-off between delay and magnitude of reinforcement. In a study conducted by Gui et al. (2023) the influence of smartphone addiction and anxiety on impulsive decision-making was investigated through the lens of the DDM. The findings revealed a significant correlation between monetary differences in decision weight and anxiety, but not between anxiety and the difference in reward delay. These findings provide further insight into the cognitive mechanisms underlying the intertemporal decision process, indicating substantial variability between individuals with high and low state anxiety with respect to the balance they consider optimal for making a decision between immediate and small versus delayed and larger rewards. That is, the temporal delay of these rewards is of lesser importance than the magnitude of the delayed reward, such that a larger magnitude increase in reinforcement is required to compensate for the delay.

In the course of this review, a multitude of methodologies for measuring DD have been employed, all of which are predicated on the dichotomy between immediate, albeit modest, rewards and those that are delayed and of a greater magnitude. However, the specific characteristics of the test appear to be relevant to the resulting outcome. The evidence suggests that individuals’ performance on tasks varies depending on whether they are behavioural or hypothetical tasks (Patt et al., 2021). The studies included in this review employ both types of tasks; however, the majority are hypothetical. This could indicate that participants do not demonstrate their true delay of gratification ability in hypothetical tasks due to a lack of real motivational incentives that they otherwise encounter in experiential paradigms or in their own daily lives. Other possibilities could be the overestimation of one’s own capabilities in hypothetical situations or the emotional influence of real situations on decision-making.