Measuring state changes in human delay discounting: an experiential discounting task

Abstract

A new experiential discounting task (EDT) is presented. Unlike existing question-based measures of delay discounting, which rely on imagined consequences during task completion, this EDT requires that participants experience choice consequences (i.e. delays and pseudo consumatory responses) during the measurement period. As a preliminary examination of this task’s sensitivity to variability in discounting, 12 participants (six females) completed a timing test (production and reproduction), a question-based measure of delay discounting, and the EDT during non-sleep-deprived (awake 7 h) and sleep-deprived (awake 21 h) conditions. Based on evidence that sleep deprivation increases impulsive behavior, it was hypothesized that participants would underrepresent time intervals in both production and reproduction procedures and discount significantly more with the discounting procedures while sleep deprived. Unfortunately, data from the question-based discounting procedure could not be reported due to invalid task completion. However, as hypothesized, certain production and reproduction intervals were underrepresented on the timing test, and discounting was significantly steeper on the EDT when participants were sleep deprived. Also, rate of discounting on the EDT was better characterized by a hyperbolic function than by exponential function, which is consistent with previous delay-discounting research. These preliminary results suggest this EDT may be a useful measure for assessing state changes in discounting processes.

Introduction

The primary goal of this article is to introduce and present preliminary data for an experiential discounting task (EDT). This EDT is different from the more typically reported hypothetical, or even “real-reward” procedures (reinforcing one choice response at the end of the procedure, e.g. Johnson and Bickel, 2002, Reynolds et al., 2004), in how participants experience choice consequences. This task is called an “experiential” task because participants more directly experience certain choice consequences (i.e. delays and pseudo consumatory responses) during the measurement procedure. Therefore, participants receive consequence feedback while completing the task in a manner we believe will make it sensitive to moment-to-moment intra-individual variability in discounting.

Delay discounting is typically considered an index of impulsive behavior. It is an assessment of the degree to which subjective value of a commodity decreases as a function of a delay to its delivery (e.g. Rachlin, 2000). By this definition, greater discounting indicates greater impulsivity (Daruna and Barnes, 1993, Oas, 1985, Reynolds and Schiffbauer, in press–a, Richards et al., 1999). In most delay-discounting procedures, participants make choices between rewards that are smaller but immediate versus rewards that are larger but delayed. For example, Schweitzer and Sulzer-Azaroff (1995) examined delay discounting in children with attention deficit hyperactivity disorder (ADHD). Participants were required to make choices between three nickels to be received after a 16 s delay versus one nickel with no delay. They received the nickels following each choice. Children with ADHD made significantly more choices for the immediate reward (one nickel) than did matched non-ADHD controls. In other delay-discounting procedures the delay periods are varied, thus yielding more parametric data (e.g. Green et al., 1994, Rachlin, 2000, Richards et al., 1999). In these procedures, participants typically answer a series of questions about choice preferences between larger delayed and smaller immediate amounts of money, e.g. Would you prefer US$ 10 180 days from now, or would you prefer US$ 3 now? Over successive questions, the smaller non-delayed amount is adjusted until the subjective value of US$ 10 in 180 days can be determined. The smallest amount an individual chooses to receive immediately in lieu of the delayed larger amount is called an indifference point, which is the point when the immediate and delayed amounts are of equal subjective value.

Indifference points can be plotted to form a discount function. Discounting data are well characterized by the hyperbolic model (Mazur, 1987), notated as follows:

$$\text{Value=}\text{A}\text{(1+kD)}\text{,}$$

where Value represents the value of the delayed reinforcer, and A and D are the amount of reinforcer and length of delay to its delivery, respectively. The k is a free parameter and indicates the steepness of the discount curve. Higher k-values indicate more rapid discounting, which has been defined as more impulsive (e.g. Mazur, 1987, Rachlin, 2000, Richards et al., 1999). A number of studies have shown that patterns of discounting by delay are better characterized (i.e. are fit better) by a hyperbolic function than by exponential function (e.g. Kirby and Herrnstein, 1995, Myerson and Green, 1995, Rachlin et al., 1991, Richards et al., 1999), which is notated as follows:

$$\text{Value=A}\text{e}{}^{\mathrm{-kD}}\text{,}$$

where again A is amount of reinforcer, D the delay to receiving the reinforcer, and k the free parameter indicating steepness of the discount function. However, it is this exponential model that has historically been the standard of rational choice in the field of economics (e.g. Samuelson, 1937, Loewenstein, 1992); therefore, findings of better fits with a hyperbolic, non-rational model take on importance in the study of impulsive behavior.

Question-based delay-discounting procedures are effective for assessing stable discounting patterns related to behaviors generally considered impulsive. For example, chronic cigarette smoking (e.g. Bickel et al., 1999, Mitchell, 1999, Reynolds et al., 2004), excessive alcohol consumption (e.g. Vuchinich and Simpson, 1998), opioid dependent drug abuse (e.g. Madden et al., 1997), addictive gambling (Petry and Casarella, 1999), and the clinical diagnoses of substance abuse and dependence, borderline personality disorder, and bipolar disorder (Allen et al., 1998, Crean et al., 2000) all have been related to higher rates of discounting by delay. A comparatively small number of studies exist to suggest these measures detect state changes in discounting (de Wit et al., 2002, Giordano et al., 2002, Hinson et al., 2003); however, other findings are less clear. For example, several drugs that might be expected to affect impulsive behavior, such as alcohol (Ortner et al., 2003, Richards et al., 1999), diazepam (Reynolds et al., in press-b) and Δ⁹-tetrahydrocannabinol (McDonald et al., 2003) have had no significant effects on delay discounting. One possible reason for such null findings is that these manipulations do not, in fact, increase discounting. However, this possibility is inconsistent with some non-human animal discounting research showing alcohol and diazepam effects (e.g. Evneden and Ryan, 1996, Poulos et al., 1998, Tomie et al., 1998). Conversely, it is possible that the question-based procedure used in human research may itself lack robustness in detecting moment-to-moment variability in discounting. As already suggested, such insensitivity to state-change effects may result from the time frame of choice consequences. That is, participants do not experience direct choice consequences during the drug state with such procedures. A recent review also suggested that operant procedures more generally might be more sensitive than question-based procedures due to reinforcement and consumption processes (Navarick, 2004). More than with question-based tasks, this EDT is a delay-discounting procedure that consequates choices during the actual measurement timeframe.

To examine this EDT’s sensitivity to state-like variability in discounting, participants completed the EDT while sleep deprived (awake 21 h) and not sleep deprived. Sleep deprivations of approximately 21 h produce a circadian nadir (low-point) in performance on various tasks and measures (e.g. Babkoff et al., 1991, Williamson and Feyer, 2000). Some of these measures involve processes that may be implicated in delay discounting, such as temporal memory. For example, sleep deprivation reduces accuracy of the timing (recency) of prior events (Harrison and Horne, 2000, Morris et al., 1960). There also is evidence that an impaired ability to make accurate judgments about the passage of time is related to high impulsivity, as determined by impulsivity questionnaires (van den Broek et al., 1992). Also, sleep deprivation and fatigue increase risk-taking and behavioral disinhibition. For example, Brown et al. (1970) found that drivers became increasingly willing to take risks with increasing fatigue by engaging in specific hazardous “overtaking maneuvers” (e.g. passing in low visibility, forcing other drivers to adjust speed to permit them to pass). The same participants, however, did not adopt other overly cautious (non-risky) bad driving behaviors that would indicate a more general deterioration in driving skills. Harrison and Horne (1998) also found that sleep-deprived participants were likely to take risks, to their own detriment, in a complex strategic task that required them to draw cards from different stacks of cards, with each having different payoff-to-penalty ratios. In sum, substantial evidence exists to suggest that sleep deprivation generally increases impulsive behavior, and it also specifically affects some timing processes that may be implicated in delay discounting.

In the current study, participants completed a timing test (Barkley et al., 2001b) to assess the effects of sleep deprivation on timing processes, and they also in the same session completed both a question-based measure of delay discounting and the EDT. However, as is sometimes unfortunately the case with this particular question-based measure, not all participants completed the procedure in a manner that could be fitted to Eq. (1) (only four in this case; see also Reynolds et al., 2003, Richards et al., 1999). While it was initially intended that these data be included in the current report, due to unreliable performance on this task, these data will not be presented nor discussed further.

Related to our hypotheses, several researchers have hypothesized linkages between timing processes and choice behavior related to delayed options (e.g. Gibbon, 1977, Killeen and Fetterman, 1988). Also, research in the area of ADHD suggests an association between poor behavioral inhibition (similar to that observed in the sleep-deprived) and time-interval underreproduction (e.g. Gerbing et al., 1987, Seigman, 1961). It has been suggested that such timing deficits may lead delays to appear longer if actually experienced for such individuals, thus leading to greater impulsivity in contexts involving delays or waiting (Barkley et al., 2001b). Extending from such evidence, it was hypothesized that time intervals would be underrepresented and discounting on the EDT would increase (i.e. higher k-values) when participants were sleep deprived. Also, it was expected that the pattern of discounting on the EDT would be better characterized by the hyperbolic function (Eq. (1)) than by the exponential function (Eq. (2)).