Preference reversal and the estimation of indifference points using a fast-adjusting-delay procedure with rats

PREFERENCE REVERSAL AND THE ESTIMATION OF INDIFFERENCE POINTS
USING A FAST-ADJUSTING-DELAY PROCEDURE WITH RATS


















By

SHONNIE M. BENNETT


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE
UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE
REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2002















ACKNOWLEDGMENTS

I thank my committee members (Dr. Marc N. Branch, Dr. Timothy Hackenberg,

Dr. Joanna Peris, Dr. Donald J. Stehouwer, and Dr. Timothy Vollmer) for their assistance

in the completion of this dissertation. I express sincere thanks to Cynthia Pietras, Manish

Vaidya, Stephen Haworth, and Rafael Bejarano for their continuous support and

numerous helpful discussions. Finally, I am greatly indebted to Dr. Frans van Haaren for

his assistance and guidance throughout all phases of the development and preparation of

this dissertation.
















TABLE OF CONTENTS

Page

ACKN OW LED GM EN TS ............. ........... ..... .................. ............ ii

ABSTRACT ........... ........... ........... ...........................

CHAPTER

1 GENERAL INTRODUCTION ........... .............. ............ ......... 1

2 EXPERIMENT 1 .......... .......... ......... ......... .........1..
Introduction .......... ................... ......................... 11
Methods .......... ........... ........... .........................12
Subjects ............ ........... ...................................12
Apparatus ........................ ....... ........... ........ ............ 12
Procedures ................... .................................. ........ 13
Results ........... ........... ........... ....... ..................18
Discussion ........... ....................... ........... ............23

3 EXPERIMENT 2 ......... ... ......... ......... ....................25
Introduction ........................................................ ........ 25
Methods .......... ........... ........... ............................27
Subjects ........... ........................ .......... ............27
Apparatus ...................................................... ............28
Procedures ........... ........... ............ ....... .................28
Results ........................................ ................... .........33
Discussion ....................................... ................. ........ 88

4 GENERAL DISCUSSION ............................... ...........94

REFERENCES .......... ........... ........... ................. 109

BIOGRAPHICAL SKETCH ................................ ................115















Abstract of Dissertation Presented to the Graduate School of the University of Florida in
Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

PREFERENCE REVERSAL AND THE ESTIMATION OF INDIFFERENCE POINTS
USING A FAST-ADJUSTING-DELAY PROCEDURE WITH RATS

By

Shonnie M. Bennett

December 2002

Chairman: Frans van Haaren
Major Department: Department of Psychology

Self-control was defined as choice of a large, delayed reinforcer over a small,

more-immediate one. Rats in Experiment 1 were presented with a choice between a

small, immediate reinforcer (one pellet delayed by 0.1 s) and a large, delayed food

reinforcer (three pellets delayed by 6, 12, 24, or 48 s). The purpose of this experiment

was to determine if preference for the large reinforcer would reverse when the delay to

the large reinforcer (three pellets) was increased across experimental conditions. When

rats were presented with a choice between one pellet after 0.1 s and three pellets after a 6

or 12 s delay, five of the six subjects chose the large, delayed reinforcer. Preference

reversed to almost exclusive choices for one pellet delayed by 0.ls when the delivery of

three pellets was increased to 48 s. The discrete-trial procedure used in Experiment 1

provided data for interpolating the large reinforcer delay producing a preference reversal

as somewhere between 24 and 48 s. The purpose of Experiment 2 was to examine the

use of a fast









adjusting-delay procedure to precisely determine the delay to the larger reinforcer that

would result in preference reversal. This was achieved by adjusting the delay to the large

reinforcer (three pellets) dependent upon the subject's performance across choice trials.

The standard alternative of one pellet followed a fixed delay of 0.1, 2, 4, 8 or 16 s. The

adjusted delay that maintained equal responding between the two alternatives defined the

indifference point and the delay at preference reversal. The indifference points increased

as the delay to the small reinforcer increased suggesting increased preference for the

large, delayed reinforcer as the delay to the small, immediate reinforcer increased. Linear

regression analyses of the resulting discount functions yielded positive slopes that were

very similar to those previously reported using other subjects and procedures. Although

the y-intercepts generated from the current experimental findings were larger than those

previously reported using pigeon subjects. The findings from Experiment 2 were

consistent with those from Experiment 1. Consistent with previous findings, the current

findings provide support for a hyperbolic equation of choice.















CHAPTER 1
GENERAL INTRODUCTION

Self-control choice has been defined as choice of a large, delayed reinforcer over

a small, immediate one; impulsive choice was defined as the opposite (see review by

Logue 1988). The current experiments investigated the influence of increased large,

reinforcer delays on self-control choice of rats. Previous self-control research with

human and non-human subjects has produced a number of consistent experimental

findings. First, pigeons have been shown to make impulsive choices for food reinforcers

when the small reinforcer is immediately available. That is, pigeons will choose 2 s of

grain access now over 6 s of grain access after a 6 s delay (Ainslie 1974; Logue 1988;

Logue and Mazur 1981; Logue, Rodriguez, Pefia-Correal, and Mauro 1984; Mazur and

Logue 1978; Rachlin and Green 1972). Rats have also been shown to make impulsive

choice under similar choice conditions (Eisenberger and Masterson 1987; Eisenberger,

Masterson, and Lowman 1982; Evenden and Ryan 1999; Tobin, Chelonis, and Logue

1993). While self-control choice patterns are generally reported with humans and

sometimes with rat subjects (King and Logue 1987; Logue, King, Chavarro, and Volpe

1990; Logue, Pefia-Correal, Rodriquez and Kabela 1986; Takahashi and Fujihara 1995;

van Haaren, van Hest, and van de Poll 1988). Further investigation determined that

several procedural variables may have contributed to the higher percentage of self-control

choices observed with human subjects (Flora, Schieferecke, and Bremenkamp 1992;









Forzano, Porter, and Mitchell 1997; Green, Fry, and Myerson 1994; King and Logue

1987; Logue 1988; Logue et al. 1990; Logue et al. 1986; Millar and Navarick 1984;

Navarick 1986; Takahashi and Fujihara 1995). Most research examining human self-

control choice has included the presentation of choices between hypothetical amounts of

money or points exchangeable for money. Points that are earned within a session are

typically exchanged for money at the end of an experimental session or following some

post-session delay. Pigeons made more self-control choices when this type of second-

order arrangement was utilized. Jackson and Hackenberg (1996) presented pigeons with

self-control choices for tokens (the presentation of light emitting diodes) that could be

exchanged for food after the completion of several or all of the choice trials in a session.

The data suggested that higher self-control responding was observed in this choice

arrangement due to the similarity of delays to food for choosing either alternative.

Furthermore, the use of consumable or primary reinforcers such as access to a video

game, pictures, or loud noise presentation have been shown to result in more impulsive

choice by humans (Flora et al. 1992; Forzano et al. 1997; Millar and Navarick 1984;

Navarick 1986). Collectively, these data suggest that the use of conditioned or

specifically token reinforcers that are only exchangeable at the end of a session may

result in more self-control choice than when primary reinforcers are used. Notably, a

study reported by van Haaren et al. (1988) demonstrated high percentages of self-control

choices in rats utilizing consumable reinforcers. In Experiment 1, van Haaren et al.

(1988) showed that rats made self-control choices when presented with a choice between

3 pellets available after a 6 s delay and one pellet available after a 0.1 s delay. However,

these subjects had previously been exposed to choices between equally delayed (at 6 s)









but unequal amounts of food (1 versus 3 pellets) before exposure to the self-control

choice arrangement. This history may have biased preferences. Even if the initial

exposure to equal delays was a variable that contributed to the observed preference of

these rats, this finding is still inconsistent with findings reported with pigeons. They

require an extensive history of gradual changes in the delay to maintain even a small

percentage of self-control choice when the small reinforcer is immediately available

(Logue and Mazur 1981; Logue et al. 1984; Mazur and Logue 1978). The delays were

not gradually faded in Experiment 1 of van Haaren et al. (1988) but instead decreased

from six seconds to a 0.1 s delay to one pellet. Tobin et al. (1993) showed that a similar

history of equally delayed but unequal amounts of reinforcement was not sufficient to

maintain self-control choice in rats. Rats were initially presented with a choice between

two seconds versus six seconds of access to sweetened condensed milk before the delay

to the small reinforcer was decreased to zero seconds. Less than thirty percent of the

choices were for the large, delayed reinforcer. Thus, the possibility remains that rats

would make self-control choices even without this brief history of delayed reinforcement

in a choice arrangement similar to that used by van Haaren et al. (1988).

Another defining feature of self-control choice with humans and nonhumans is

preference reversal from the small to the large reinforcer with increased delays to both

alternatives (Ainslie 1974; Fantino 1977; Ito and Asaki 1982; Logue 1988; Navarick and

Fantino 1976; Rachlin and Green 1972) or from the large to the small reinforcer with

increased large reinforcer delays (Green, Fristoe, and Myerson 1994; Logan 1965; Logue

1988; van Haaren et al. 1988). Preference reversals have been examined using two types

of experimental procedures. In the fixed procedure, subjects are presented with a choice









between a small, more-immediate reinforcer versus a large, delayed one and the delay to

one or both alternative is manipulated across conditions. The delay producing a switch in

preference is then interpolated across conditions. An example of increasing the delay to

both alternatives comes from an experiment reported by Rachlin and Green (1972).

Pigeons chose a small, immediate food reinforcer over a large, delayed one until an

opportunity to make a commitment response was provided at a point before the choice

opportunity. The pigeons could make a commitment response by completing twenty-five

responses (FR 25) on one of two response keys prior to the presentation of a choice

between an immediate but small reinforcer and a large but delayed reinforcer.

Completion of the ratio on one key resulted in the presentation of only the large, delayed

reinforcer alternative during the choice component while completing the ratio on the

other key resulted in the presentation of both choice options. The addition of this

commitment opportunity resulted in increased self-control choice.

A second type of fixed procedure was demonstrated in Experiment 2 by van

Haaren et al (1988). Rats were presented with a choice between one pellet delayed by 6 s

and three pellets following a delay ranging from 9 to 36 s. The delay to three pellets

increased across conditions and the subjects did not make impulsive choices until the

delay to three pellets was 36 s. The data showed that a preference reversal would require

a large, reinforcer delay in excess of 36 s.

The second type procedure used to predict preference reversals by determining

the indifference point is an adjusting or titrating procedure. In this arrangement, the

delay or amount of reinforcement changes/adjusts due to the subject's preference (Bickel,

Odum, and Madden 1999; Mazur 1987; 1988; Richards, Mitchell, de Wit, and Seiden








5
1997; Richards, Zhang, Mitchell, and De Wit 1999). Mazur (1987) introduced the use of

an adjusting delay procedure to examine changes in the value of delayed reinforcers in a

self-control choice arrangement. In this procedure, pigeons chose between a standard or

adjusting alternative. Choice of the standard alternative resulted in a small reinforcer, 2 s

of access to grain, after a fixed delay (0, 1, 2, 6, 10, 14, or 20 s). Choice of the adjusting

alternative resulted in a large reinforcer, 6 s of grain access, after a delay that changed

based on the subject's choices during a session. A block of trials started with exposure to

each alternative followed by two choice trials. Based on the choices made during that

block, the delay to the large reinforcer changed during the next four-trial block. If the

subject chose the large reinforcer during the two choice trials, then the delay to that

reinforcer increased by 1 s during the next block of trials. However, if the subject chose

the small reinforcer during the two choice trials, the delay to the large reinforcer was

decreased by 1 s during the next block of trials. Finally, if the subject made one choice

for each alternative, the delay to the large reinforcer did not change during the next trial

block. The point of indifference was defined as the delay at which the subject responded

equally often on both choice alternatives.

Examining changes in the indifference point with changing delays to the small

reinforcer allowed experimenters to fit equations to these data which quantify changes in

reinforcer value with increased delays to their receipt (the discount function).

Indifference curves, plots indifference points as a function of increased standard delays,

are then described with the parameters associated with the best fitting lines (the discount

function). One major advantage of the adjusting procedure is that the delay at preference

reversal can be determined more rapidly, within an experimental condition as opposed to









across several experimental conditions. A second advantage is the sensitivity of the

procedure to individual (across) subject variability and to within subject changes in

preference. The performance of each subject produces changes in the delay to the

adjusting alternative. If a subject is sensitive to reinforcer delays, the adjusting delay

decreases accordingly until responding maintains for each choice alternative. The

opposite pattern of adjustment would be observed with a subject that is not sensitive to

reinforcer delays. Finally, the delay producing a reversal of preference can be narrowly

defined instead of interpolated across conditions. Richards et al. (1997) provided a

modification of this procedure where the amount of immediate reinforcement adjusted

across choice trials instead of the delay. The procedure included several other

modifications designed to allow for a much more rapid determination of indifferences

points than Mazur's adjusting-delay procedure. Currently, the procedure proposed by

Richards et al. (1997) has not been systematically replicated to determine if stable

indifference points could be rapidly determined when the procedure included an adjusting

delay instead of the adjusting amount. Experiment 2 was designed to systematically

replicate the procedures proposed by Richards et al. (1997) when the delay adjusts.

Mazur (1987) proposed a hyperbolic function to describe how the value of a

choice alternative is influenced by reinforcement parameters. This equation has been

used to successfully describe self-control choice behavior across species and fixed or

adjusting procedures (Ainslie 1974; Bickel et al. 1999; Green, Fristoe, and Myerson

1994; Green, Fry, and Myerson 1999;Green, Myerson, and McFadden 1994; Ho, Wogar,

Bradshaw, and Szabadi 1997; Mazur 1987; 1988; Mazur et al. 1987; Rachlin and Green








7
1972; Richards et al. 1997; Richards et al. 1999; Tobin, Chelonis, and Logue 1993). The

equation is:

V = bA/(1+KD) (1-1)

describing changes in the value of a reinforcer as a function of reinforcement parameters

of amount and delay. In this equation, A is a parameter that is monotonically related to

the amount of reinforcement, D is the delay to its receipt and K is a discount parameter

that is empirically determined for each subject. The equation describes value/preference

for a given alternative to be a direct function of the amount of reinforcement and

inversely related to the delay. The equation sometimes included a bias parameter, b, to

describe consistent preference for one alternative irrespective of reinforcement

parameters. Furthermore, the adjusted delay at indifference can be predicted by setting

the equations for each alternative equal to each other such as:

AL / (1+KDL) = AR / (1+KDR) (1-2)

The subscripts of L and R designate the parameters associated with the left and right

response alternatives, respectively. For illustration, the bias term is set equal to 1.0,

assuming no response bias.

Several predictions regarding choice behavior follow from this equation. First,

the equation predicts a preference reversal. Second, the equation predicts a linear relation

between the adjusting delay at indifference as a function of increased standard delays

with a y-intercept greater than one and a positive slope. Experimental evidence

supported the predictions of this equation (Ainslie 1974; Green, Fristoe, and Myerson

1994; Green, Myerson, and McFadden 1997; Ho, Mobini, Chiang, Bradshaw, and

Szabadi 1999; Logan 1965; Logue 1988; Mazur 1987; 1988; Mazur et al. 1987; Rachlin









and Green 1972; Richards et al. 1997; Richards et al. 1999). Although, the difference in

choice behavior of pigeons and the findings with rats reported by van Haaren et al. (1988)

may suggest that the influence of delays on the reinforcer value is qualitatively different

for rats and pigeons, similarities in the observed discount functions suggest qualitative

similarity but instead leave the possibility that the differences are quantitative in nature.

An evaluation of discount functions would provide data to address this issue. If the

observed differences in self-control choice across species was in fact due to some

qualitative differences in how delays for reinforcement influence reinforcer value, then

the discount functions would be expected to have different forms and be inconsistent with

the predictions of the hyperbolic equation. Alternatively, if the differences were

quantitative in nature, the functions would be expected to look similar in form but differ

in the absolute parameters associated with the function.

The adjusting-delay procedure proposed by Mazur produced data describing how

delays to reinforcer delivery influenced reinforcer value or preference. Experimental

research has demonstrated that as the delay to the small reinforcer increased, the mean

adjusting delay to the large reinforcer increased at the point of indifferent responding

between the two alternatives. One equation that has accurately described this relationship

was a hyperbolic discounting equation. Although this procedure has been successfully

used to examine discount functions across species, one limitation of the procedure is the

lengthiness of producing the indifference curve. Richards et al. (1997) proposed a

procedure to examine the influence of changing reinforcer amounts on indifference points

with a procedure that rapidly determined the indifference curve. The indifference curves

reported by Richards et a. (1997) described how changes in reinforcer amount influenced









reinforcer value instead of how reinforcer delay did as in the Mazur (1987) procedure.

Extending the procedures of Richards et al. (1997) to an adjusting delay procedure may

allow rapid determination of indifference curves with rats and provide data illuminating

similarities and/or differences in such said function between rats and pigeons. As

previously noted, rats have been reported to make self-control choices in conditions

where pigeons make impulsive ones suggesting a potential qualitative difference in the

delay discount functions from rats to pigeons. The purpose of the current experiments

was not to examine species differences per se although the data illuminate whether the

reported differences between the performance of rats and pigeons is the result of

quantitative or qualitative differences in the delay discount functions. Examination of the

delay discount functions with rat subjects would provide evidence regarding this possible

account.

The systematic replication of the procedures proposed by Richards et al. (1997)

provide delay discount functions with rats that have not previously been reported in the

self-control choice literature (see review by Logue 1988). Furthermore, the data will

extend previous reports of delay discount functions reported with pigeon (Mazur 1987;

1988) and human subjects (Rachlin, Raineiri, and Cross 1991; Richards et al. 1999).

In summary, comparisons of self-control choice between rats and pigeons leave

several questions to be addressed: First, was the history provided to subjects in

Experiment 1 of van Haaren et al. (1988) responsible for the observed self-control with

rats that has not typically been observed with pigeon subjects? Secondly, if the subjects

show near exclusive self-control choice without such a history, what are the large,

reinforcer delays at which rats make impulsive choices? Thirdly, will discount functions








10
obtained using rats in a modified fast adjusting-delay procedure proposed by Richards et

al. (1997) be qualitatively similar to those observed with pigeons using Mazur's (1987)

adjusting- delay procedure? Finally, will the modified procedure of Richards et al.

(1997) provide delay discount functions that are similar to those observed using other

subjects and using the standard adjusting delay procedure?















CHAPTER 2
EXPERIMENT 1
Introduction

As previously noted, pigeons typically make impulsive choices when presented

with a choice between a small, immediate reinforcer and a large, delayed one unless they

have been provided with an extensive history of gradually changing delays (Ainslie,

1974; Logue and Mazur, 1981; Logue 1988; Logue and Mazur 1978; Logue et al. 1984).

The findings reported from Experiment 1 by van Haaren et al. (1988) suggest that rats

may not require a history of gradually changing delays to maintain preference for the

large, delayed reinforcer when a small one is immediately available. Other researchers

have reported impulsive choice patterns using rats (Eisenberger and Masterson 1987;

Eisenberger, Masterson, and Lowman 1982; Evenden and Ryan 1999; Tobin, Chelonis,

and Logue 1993). Tobin et al. (1993) reported impulsive choice with rats even after

being provided with a history similar to that provided by van Haaren et al. (1988). These

findings suggest that a history with delayed reinforcers may not be sufficient to obtain the

exclusive self-control choice that were observed with rats by van Haaren et al. (1988).

The inconsistent experimental findings are not attributable to different durations of

history as subjects in both experiments were provided with similar durations of exposure

to the choice arrangement in terms of number of sessions and the number of trials per

session. The first question addressed in the present experiment was whether the history

provided the rats in Experiment 1 of van Haaren(1988) was necessary to obtain the self-

control choice preferences observed. The subjects in the current experiment were not

11









provided with a history of delayed reinforcement before self-control choices were

examined. The second goal was to determine the large, reinforcer delay that would be

required to obtain a switch in preference from the large reinforcer to the small reinforcer

as determined using a fixed procedure.

Methods

Subjects

The subjects were three male and three female experimentally naive Wistar rats.

Female rats weighed 334 grams and male rats weighed 464 grams at the start of the

experiment. The subjects were housed in same-sex groups of three per cage. Water was

continuously available but access to food was restricted to a daily total of 36 grams for

female subjects (12 grams per female) and 45 grams for male subjects (15 grams per

male), presented immediately after the experimental session or at approximately that time

of day on weekends. This daily feeding regimen allowed approximately 22 hours of food

deprivation at the start of each experimental session. Rats were housed under a reversed

light cycle (lights on at 7:00 p.m. and off at 7:00 a.m.); experimental sessions were

conducted during the dark cycle between 11:00 a.m. and 1:00 p.m. on Monday through

Friday.

Apparatus

All experiments were conducted in three identical standard rodent operant

chambers that were 30 cm long, 25 cm wide, and 29 cm high (Coulbourn Instruments,

Allentown, PA). The floor of each chamber consisted of 16 metal rods, spaced 1.75 cm

apart. The two sidewalls were made of Plexiglas while the front experimental panel and

back wall were made of stainless steel. The experimental panel consisted of two









nonretractable response levers, stimulus lights above each lever, and a centrally located

food dispenser and houselight. The houselight was located 3 cm from the chamber

ceiling and provided diffuse illumination. The pellet tray was located 1.7 cm from the

floor and was illuminated during pellet presentation (Noyes, 45-mg rodent formula P).

The response levers were located 12.5 cm apart and 6.3 cm above the grid floor. Each

lever required a force exceeding 0.20 N to close a microswitch. Each chamber was

enclosed in a sound-attenuating and ventilated cubicle. Experimental events were

controlled and data collected using a PDP 11-23 microcomputer (Digital Equipment

Corporation) and SKED-11 programming software (Snapper and Inglis, 1981).

Procedures

Preliminary training. Each session started with a dark period of 10 minutes

during which lever presses had no programmed consequences. Once the houselight and

stimulus lights above each lever were illuminated, pellets were delivered according to a

conjoint fixed ratio (FR 1) of one response (responding on either lever could satisfy this

contingency) and a variable time (VT) 60 s schedule of delivery (van Haaren, 1992).

Sessions lasted for 45 minutes or until 40 reinforcers had been presented, whichever

occurred first. Once subjects reliably responded on both levers (as determined by visual

inspection of the data), the VT schedule was eliminated and lever presses were reinforced

on a FR 3 schedule. During this part of training, the fixed-ratio schedule was correlated

with only one lever at a time and was signaled by illumination of the stimulus lights

above that lever. The "active" lever was alternated after five reinforcers were delivered.

The strict alternation between the two levers ensured equal responding and equivalent

histories of reinforcement for responding on both levers.








14
General experimental procedures. Each session started with a dark period often

minutes as used in preliminary training. Each session consisted of two trial types: forced

exposure trials and choice trials. A session started with forced exposure to both choice

options but the order of their presentation was randomly determined at the beginning of

each session. The remaining forced exposure trials for the large reinforcer (three pellets)

were then presented on trials 5, 8, 12, and 20 while forced exposure trials for the small

reinforcer (one pellet) was presented on trials 6, 10, 16, and 24. Figure 1 shows the

arrangement of stimulus lights and the sequence of stimulus changes during a given trial.

A trial was initiated with the illumination of the houselight and stimulus lights above the

lever(s). One set of stimulus lights was illuminated continuously and the other set of

stimulus lights blinked (on/off cycle of 0.75 s/0.1 s). Completion of three consecutive

responses (FR3) on the lever associated with blinking stimulus lights extinguished all

stimulus lights and initiated a delay to pellet presentation that was signaled by a blinking

houselight (HL)(on/off cycle of 0.75 s/0.1 s). Completion of the FR3 on the lever

associated with continuous stimulus lights extinguished all stimulus lights and initiated a

delay to pellet presentation signaled by continuous illumination of the houselight (HL).

A pellet was always presented simultaneously with the illumination of the food

receptacle. For the large amount of food, two additional pellets were delivered at 1 s

intervals. The food receptacle was illuminated for 2 s for both reinforcers. Completion

of pellet delivery initiated the intertrial interval (ITI), during which all chamber lights

were extinguished. Trials were scheduled to start every 60 s if at least 10 s had elapsed

since the beginning of the intertrial interval. If the 10 s minimum requirement was not

met, the next trial began after another 60 s had elapsed. Each session lasted 45 minutes














Lever


Flashing


0.1 s


6s


1 Pellet

HL




Blackout for
3 ITI
Pellets

Figure 1. Diagram of events during a trial. Sequence of stimulus presentations in a
choice trial. The sequence begins at the top of the diagram showing the delay
between a choice and food delivery followed by a blackout until the next trial
begins.


Lever








16
or until completion of 34 trials (10 forced and 24 free-choice trials), whichever occurred

first.

Delay conditions. The delay to the delivery of one pellet was scheduled at 0.1 s

while the delay to the delivery of three pellets was systematically manipulated across

experimental conditions. Table 1 shows the order of experimental conditions as well as

the number of sessions that subjects were exposed to each delay condition. For five out

of six subjects the delay to three pellets was 6, 12, 24, 0.1, 24, 0.1, and then 48 s across

conditions. The delay to three pellets was 6, 3, 6, 0.1, 24, 3, and then 6 s for subject

132F.

All of the subjects were initially presented with a choice between one pellet

delayed by 0.1 s and three pellets delayed by 6 s. The contingencies associated with the

two levers were reversed as each subject's performance reached the stability criteria. For

five subjects, the delay to three pellets was then increased to 12 s and the delay to one

pellet remained at 0.1 s. As each subject's performance met the stability criteria, the

contingencies associated with each lever were reversed. The stimulus lights remained the

same but the lever associated with each alternative switched during a contingency

reversal. In other words, three pellets may be presented following responses on the right

lever with a flashing stimulus array during one condition but follow responses for the left

lever with fixed stimulus lights in the contingency reversal condition. The delay to three

pellets was then increased to 24 s while the delay to one pellet remained at 0.1 s for one

pellet. Again, the contingencies were reversed as performance became stable. The delay

to three pellets was then decreased to 0.1 s making the delay equal to that for one pellet.

The contingencies were reversed as each subject's performance became stable before the








17
delay to three pellets was again increased to 24 s. The contingencies associated with each

lever were reversed as each subject's behavior met the stability criteria. The delay to

both alternatives was then made equal at a 0.1 s delay and the contingencies associated

with each lever was also reversed at the same time. This is the first condition in which

two changes were made in the same condition; the delay and contingency associated with

each lever were both changed. The delay to three pellets was then increased to 48 s.

During this condition, five subjects were choosing between three pellets delayed by 48 s

and one pellet delayed by 0.1 s. The delay to three pellets was then decreased to 24 s

before the contingencies associated with each lever were reversed.

Subject 132F was provided with a different history. When subject 132F was

presented with a choice between three pellets delayed by 6 s and one pellet delayed by

0.1 s, the percentage of choices for three pellets was almost zero. The delay to three

pellets was then decreased to 3 s until the stability criteria were met and the contingencies

associated with each lever were reversed. The delay to three pellets was then increased to

6 s and the contingencies associated with each lever were reversed after performance

became stable. The delays were both 0.1 s during the next condition before the

contingencies were again reversed. Due to a programming error, the delay to three

pellets was then increased to 24 s. The delays were then made equal at 0.1 s during the

next condition before the delay to three pellets was increased to 3 s. The delay to three

pellets was then increased to 6 s and the contingencies were again reversed when the

performance met criteria.

Stability criteria. Several stability criteria were met before experimental

conditions were changed. The percentage of choice trials in which subjects chose the









large, delayed reinforcer over the small, immediate reinforcer was used to determine

stability. First, a minimum of 10 sessions was required in each experimental condition.

Table 1. Delay Conditions in Experiment 1

Delay 6 s/R 12s/R 24s/R 0.1 s/R 24 s/R 0.1 s 48 s 24s R
to 3 &R
Pellets
136F 30/50 40/20 11/24 17/14 25/30 73 43 26 27

133F 30/48 39/35 24/76 12/26 66/35 18 52 32 37

170M 32/48 37/43 12/72 11/40 120/32 38 18 19

171M 30/47 37/31 23/69 11/23 104/36 32 53 23 17

172M 31/45 36/32 23/14 70/27 24/45 66/23 36 53 22

Delay 6 s/R 3 s/R 6 s/R 0.1 s/R 24 s 0.1 s 6 s 3 s R
to 3
Pellets
132F 30/48 61/30 30/55 14/41 55/34 25 42 36 43

The order of conditions for each subject is shown in the double outlined rows of the table
(from left to right). Each row shows the number of sessions subjects were exposed to
each condition and the subsequent contingency reversal. The delay to one pellet
remained 0.1 s in each condition.

Second, an individual's performance could not show any consistent downward or upward

trend upon visual inspection of the data from the last 10 sessions. Finally, the mean

percentage of self-control choices for the last five sessions could not differ by more than

10% from the mean of the preceding five sessions. Also, note that preference is

considered impulsive when the session percentage of large, reinforcer choices dropped

below 50%.

Results

Figure 2 shows the mean percentage of large reinforcer choices over the last ten

sessions of each condition and the last ten sessions of the contingency reversal. The








19
percentage of large-reinforcer choices was determined by dividing the number of choices

for three pellets by the total number of choice trials completed during the session. The

dotted horizontal line on each graph references the preference reversal criterion of less

than 50 percent self-control choices. All contingency reversals yielded choice

percentages almost identical to those before the reversals suggesting that no position bias

developed during the course of the experiment. Although it is not indicated in the figure,

the percentage of self-control choices was zero during the sessions immediately

following a contingency reversal and then recovered suggesting that performance was

sensitive to the contingencies associated with responding on each lever. Choice latencies

were omitted as the data did not demonstrate consistent trends across trial types.

All subjects were originally exposed to a choice between three pellets delayed by 6 s and

one pellet after a programmed delay of 0.1 s. Under these conditions, all subjects, except

132F, almost exclusively chose three pellets delayed by 6 s. For the five subjects

showing exclusive choices for the large reinforcer, the delay to three pellets was then

increased to 12 s and then 24 s during which large reinforcer preference was maintained.

The delay associated with the large reinforcer was then decreased to 0.1 s before a delay

of 24 s was reintroduced to determine if the gradual change in the delays was responsible

for the observed preference for the large, delayed reinforcer. When the 24 s delay was

reintroduced, four subjects maintained preference for the large, delayed reinforcer. The

choice patterns of subject 170M differed from the other four subjects showing a slight

decrease in preference for the large, delayed reinforcer with the first introduction of a

delay of 24 s to three pellets and an even larger decrease in self-control choice when the

contingencies were reversed (<80% large, delayed reinforcer choices). The delay to three




















































Figure 2. Each graph shows the percentage of large reinforcer choices for an
individual subject across delay conditions. Each bar plot shows the mean
and 99% confidence interval based on the last ten sessions of each
condition. Means during contingency reversals are indicated by an axis
label of R and the large reinforcer delay is the same as that indicated by
the bar to the left of that reversal.











100
80
60
40
20

100
80
60
40
20
0
100
80
60
40
20
0
100
80
60
40
20
0


c)
^6I Ai nfl


( uu -
= 80-
(U
_. 60-
CO 40-
L 20-
L- 20
O


0
100 -
O
-C 80 -
60
S40 -
S20-
0


- 136F






133F






71 M

- i I IT




172M
-b~ \Q; Q/ ^ |11
^ 0>


I I I I I I


Delay to 3 Pellets (sec)
Delay to 3 Pellets (sec)


170M


I I I I I I I









pellets was again decreased to 0.1 s and the contingencies reversed prior to the

introduction of a 48 s delay. All five subjects (133F, 136F, 170M, 171M, and 172M)

showed exclusive large, reinforcer preference during equal 0.1 s delays. Preference

reversed for the first time for subject's 136F, 133F, 170M, 171M, and 172M to the small,

immediate reinforcer when the large reinforcer was delayed by 48 s.

Five subjects (136F, 133F, 170M, 171M and 172M) continued to choose the

larger, delayed reinforcer on fewer than 20% of the trials when the delay was decreased

from 48 s to 24 s. Preference was maintained for the small reinforcer when the

contingencies were reversed at a large reinforcer delay of 24 s for subjects 136F, 133F,

171M, 170M, and 172M.

Subject 132F was the only subject to choose one pellet delayed by 0.1 s when the

delay to three pellets was 6 s. This subject chose three pellets on almost 100% of the

choice trials when the delay was decreased to 3 s. The percentage of large-reinforcer

choices remained near 100% or exclusive when the delay to three pellets was then

increased to 6 s. This subject, as well as the other five subjects, was sensitive to the

amount differences across the two choice alternatives as illustrated by preference for

three pellets when both were equally delayed by 0.1 s. Preference reversed to the small

reinforcer when the delay to three pellets was increased to 24 s for subject 132F.

Reintroduction of a 6 s delay to three pellets resulted in an 80% preference for the large

reinforcer that increased to above 90% of large-reinforcer choices when the delay was

decreased to 3 s and the contingencies were reversed.











Discussion


The results of this experiment illustrate that when rats are given an opportunity to

choose between one pellet that was available almost immediately and three pellets

presented following a 6 s delay, most of the subjects will choose the three pellets and will

continue to prefer this alternative until that delay is increased to more than 24 s. Some

individual differences were observed in that subject 132F chose one pellet available

immediately when three pellets were delayed by 6 s. However, most of the subjects

continued to choose three pellets on most trials until the delay was increased to 48 s

(136F, 133F, 170M, 171M, and 172M). The current findings are inconsistent with those

found with pigeons as subjects where preference for the more immediate reinforcer is

observed when the delays to the larger reinforcer of 6 s (see review by Logue 1988) and

those reported using rats (Eisenberger and Masterson 1987; Eisenberger, Masterson, and

Lowman 1982; Evenden and Ryan 1999; Tobin et al. 1993). Tobin et al. (1993)

presented rats with a choice between 2 s versus 6 s access to sweetened condensed milk

after a delay of 2 s versus 6 s. The current findings are consistent with those reported by

van Haaren et al., (1988) who showed that male and female rats chose three pellets

delayed by 6 s over one pellet available immediately following a brief history of exposure

to both reinforcers equally delayed at 6 s. The current experiment illustrated that such a

history was not necessary to obtain self-control preferences in rats.

Choice behavior appeared to depend on the specific sequence of experimental

conditions consistent with findings reported with pigeons (Mazur and Logue 1978; Logue

and Mazur 1981, Logue et al. 1986). Logue et al. (1986) showed that pigeons would









prefer the delayed but larger reinforcer more often following a history of gradually

decreasing the delay to the small reinforcer of 2 s grain access. The current experiment

illustrated a brief exposure to each increase of the large reinforcer delay would maintain

preference for that reinforcer but also that a brief exposure to a large, reinforcer delay that

could produce a preference reversal modified the influence of shorter delays. For

instance, when the delay to three pellets changed from 6 s to 12 s and finally to 24 s, five

of the subjects continued to prefer three pellets. Those subjects showed that the effects of

gradually increasing the delay was maintained even after a condition of equal, short

delays to both three pellets and one pellet, when the delay increased from 0.1 s to 24 s.

However, all five subjects (136F, 133F, 170M, 171M, and 172M) choose the small

reinforcer when exposed to a delay of 48 s for three pellets when it was preceded with a

0.1 s delay condition. This impulsive choice pattern was maintained when a delay of 24 s

was presented again. This suggests that a subjects delay history can have significant

influences on the self-control choice patterns in the current delay conditions. In other

words, if a delay of 48 s to three pellets were the first large, reinforcer delay presented in

this experiment, all of the subjects would have made impulsive choices for the small,

more immediate reinforcer.















CHAPTER 3
EXPERIMENT 2
Introduction

Experiment 1 showed that rats would choose a large delayed reinforcer when

presented with a choice between one pellet available almost immediately and three pellets

following a delay of 6, 12, or 24 s (five out of the six rats). When the delay to three

pellets increased to 48 s, preference reversed and subjects chose three pellets very

infrequently if at all. While these results are informative as they confirm and extend

previous findings, the procedure produced exclusive preference. The changes in

preference from exclusive choices of the large, delayed reinforcer to exclusive choices of

the small, immediate reinforcer made it difficult to precisely estimate the delay at which

preference began to change. The delay at which preference began to shift from three

pellets to one pellet was at a delay between 24 and 48 s, a range of 24 s. The current

experiment investigated the point of preference reversal without interpolation by utilizing

a fast adjusting-delay procedure, a modification of the procedure proposed by Richards,

Mitchell, de Wit, and Seiden (1997).

Richards et al. (1997) presented rats with a choice between a standard alternative

of 100 il of delayed water delivery and an adjusting amount of immediate water delivery.

The amount of water immediately delivered changed by 10% after each choice trial

depending on the subject's performance. The amount of water increased following a

choice for the standard alternative and decreased following a choice for the adjusting








26
alternative. Different standard delays were presented each day of the week (0, 2, 4, 8 or

16 s) over the course of a five-week cycle of standard delay presentation. Each session

began with forced exposure to both choice alternatives followed by choice trials. Forced

exposure trials were only presented following two consecutive choices for one alternative

and it was forced exposure to the non-chosen alternative. Sessions included sixty choice

trials and indifference points were determined during the final thirty choice trials within a

session.

The rapid adjusting procedure proposed by Richards et al. (1997) includes several

procedural differences from that used by Mazur (1987; 1988). First, the adjusting

alternative changed following each choice trial instead of following the completion of a

block of several trials. Second, which alternative adjusted differed across the two

procedures. In Mazur's procedure the delay for the large, delayed reinforcer changed and

the amount of small, immediate reinforcement changed in the Richards et al. (1997)

procedure. Third, the adjusting parameter changed by a percentage of the current amount

of water instead of by a fixed adjustment of 1 s to the delay. Finally, Richards et al.

(1997) presented a new standard alternative each day of the week such that the

development of indifference was assessed for many standard alternatives during the same

condition of the experiment.

Regardless of these procedural differences, the two procedures have been used to

produce comparable findings (Ho et al. 1997; Mazur 1987; 1988 Mazur et al. 1987;

Mobini et al. 2000; Tobin et al. 1993; Wade et al. 2000; Wogar et al. 1992; 1993). The

discount functions, the mean adjusting delay to the large reinforcer as a function of the

standard reinforcer, produced by the pigeons in Mazur's (1987) study showed that as the









standard delay to a small reinforcer increased the adjusting delay to the large reinforcer

also increased. The discount functions produced by rats in Richards et al. (1997)

adjusting amount procedure produced a similar function in that as the delay to the larger

reinforcer increased the adjusted amount of immediate (water) reinforcement at

indifference decreased. The results from both experiments can be summarized as

showing that as some delay to reinforcer receipt is added, the value of that reinforcer or

preference for that reinforcer is decreased. The hyperbolic equation (Mazur, 1987) has

reliably quantified this relationship between reinforcer value and reinforcement

parameters of amount and delay.

The current experiment was designed to examine indifference points of rats using

a modified version of Richard's et al. (1997) rapid adjusting procedure; the delay

associated with the large reinforcer was manipulated instead of the amount of immediate

reinforcement. The specific goals of this experiment were threefold: First, to examine the

utility of using this modified version of the Richards et al. (1997) rapid adjusting

procedure to produce indifference points within an individual session; Second, to

examine the utility of using this modified rapid adjusting procedure to produce data

consistent with findings reported using the standard adjusting procedure; Third, to

determine if predictions made based on the observed indifference points from Experiment

II are consistent with the findings from Experiment I.

Methods

Subjects
The subjects were three female and four male experimentally naive Long Evans

rats. At the beginning of the experiment, the mean body weight for the female subjects

was 324 grams and the mean body weight for male subjects was 387 grams. The subjects









were individually housed with continuous availability of water but restricted access to

food. Female subjects received 12 grams and males received 15 grams of food each day,

presented immediately following the experimental sessions or at approximately that time

of day on weekend days when no experimental session was conducted. This daily

feeding regimen allowed approximately 22 hours of food deprivation at the start of each

experimental session. A reversed light cycle consisted of lights off at 7:00 a.m. and on at

7:00 p.m. with all experimental sessions conducted during the dark cycle between 11:00

a.m. and 6:00 p.m. on Sunday through Friday.

Apparatus
The apparatus was the same as that used in Experiment 1.

Procedures

Preliminary training. Same as in Experiment 1.

General experimental procedures. A session consisted of two trial types-forced

exposure trials and choice trials. A session started with forced exposure to both choice

options but the order of their presentation was randomly determined at the beginning of

each session. The remaining forced exposure trials were presented based on a subject's

making two consecutive choices for one alternative producing a forced exposure trial for

the non-selected alternative. For instance, if a subject chose the adjusting alternative on

trials 10 and 11 then the following trial (trial 12) would be a forced exposure trial for the

standard alternative. This arrangement ensured exposure to both alternatives during the

session but on some undetermined number of trials.

A trial was initiated with the illumination of the houselight and stimulus lights

above the lever(s). One set of stimulus lights was illuminated continuously and the other

set of stimulus lights blinked (on/off cycle of 0.75/0.1 s). Completion of three








29
consecutive responses (fixed ratio or FR 3) on the lever associated with blinking stimulus

lights extinguished all stimulus lights and initiated a delay to pellet presentation that was

signaled by a blinking houselight. Completion of the FR 3 on the lever associated with

continuous stimulus lights extinguished all stimulus lights and initiated a delay to pellet

presentation signaled by continuous illumination of the houselight. The pellet receptacle

was illuminated for 2-s during food delivery. A pellet was always presented

simultaneously with illumination of the receptacle. For the large reinforcer, two

additional pellets were delivered at 1 s intervals. Completion of the reinforcer delivery

initiated the intertrial interval (ITI), during which all chamber lights were extinguished.

Trials were scheduled to start 30 s from a choice response if at least 10 s had elapsed

between food presentation and the scheduled start of the next trial. If the 10 s minimum

requirement was not met, the next trial started after another 30 s had elapsed.

The standard alternative of one pellet was presented following a fixed delay and

the adjusting alternative was three pellets presented following an adjusting delay. The

standard delay to one pellet was 0.1, 2, or 8 s for three female and two male subjects in

this experiment (1125F, 1126F, 1127F, 1131M, and 1132M). The standard delay was

0.1, 4, or 16 s for two additional male subjects in this experiment (1134M and 1136M).

Each standard delay was presented twice a week. The adjusting delay was either 6 s

(Low) or 20 s (High) at the start of an experimental session. When the possible standard

delays and either a high or low start value of the adjusting delay are combined, subjects

were exposed to a five-week regimen of parameter presentations that are indicated in

Table 2. This regimen was recycled for the duration of the experiment. The adjusting

delay changed by 10% or 1 s following each choice trial. If 10% of the adjusting delay









was less than 1 s, then the delay on the following trial changed by 1 s but otherwise the

adjusting delay changed by 10%. Although the adjusting delay could never decrease

below the standard delay, no upper limit was placed on the adjusting alternative. The

number of trials that could be completed in a session was the only upper limit on the

adjusting delay.

Experimental conditions. There were three experimental conditions in

Experiment 2 and subjects were exposed to the conditions in numerical order. In the first

two conditions of the experiment, one of three different standard delays was presented

each day of the week. As indicated in Table 2, each delay was presented twice a week,

once with a high (20 s) starting value and once with a low starting value (6 s).

Stability criteria. Subjects were exposed to the five-week regimens shown in

Table 2 at least three times in Conditions 1 and 2 and at least once in Condition 3.

Stability was defined in terms of within session changes in the adjusting delay (within

session criteria) and by the number of sessions including stable performance during a

five-week regimen (across session criteria). The main dependent measure across all three

experimental conditions was the adjusting delay in seconds. Within-session stability was

assessed over a block of ten choice trials and was considered stable when several criteria

were met. First, no consistent upward or downward trend in the adjusting delay was

observed during the ten trial block. Second, the mean adjusting delay for the last five

choice trials of that block could not differ from the preceding five-trial mean by more

than 10%. Stability was assessed during choice trials after completion of trial 30. The

first ten-trial block to be assessed for stability was across trials 31 through 40. If the

within session stability criteria were not met in that ten-trial block, stability was then








31
assessed following the completion of five more choice trials, i.e., on trials 36 through 45.

This recursive process of stability assessment was used for the remaining trials of the

session. The across session criteria included the following requirements: A minimum of

three exposures to the entire five-week regimen; Completion of all 60 choice trials in 24

out of 30 sessions included in a five-week cycle (only in Condition 1); Meeting the

within session stability criteria in a minimum of 80 percent of the sessions in a single

five-week series, 24 out of 30 (Conditions 2 and 3).

Condition 1. Sessions lasted for a total of 120 minutes or until 60 choice trials

were completed. This condition lasted for a total of 20 weeks or four times through the

five-week series shown in Table 2. Performance was considered stable and the

condition terminated when analysis of within-session data during a five-week period

included the development of stable indifference points (defined as within-session criteria)

during 80% of the sessions, i.e., 24 out of 30 sessions. Tables 4 and 5 indicate the mean

adjusting delay and the trials over which stability was met for individual sessions in this

condition.

Condition 2. This condition lasted for another twenty weeks of four exposures to

the same five-week regimen as in Condition 1. This condition differed from Condition 1

in that sessions terminated as soon as stability was met within each session. As

previously described, within-session stability was assessed over a block of ten choice

trials

beginning after completion of choice trial 30. If within session stability was met, the

session ended. If performance did not meet the within session stability criteria,

another five choice trials were completed and stability assessed over the last ten trials.









This change was included to determine if stable indifference points could be obtained

during short sessions. If within-session stability was not reached, the session terminated

following 120 minutes or completion of 60 choice trials as in Condition 1.


Table 2. The order of standard delay presentation in experimental Conditions 1 and 2.


Day


Week 1 Week 2


Week 3


Week 4


Week 5


Sunday

Monday

Tuesday

Wednesday

Thursday

Friday


4.


8 s (Low)

0.1 s (Low)

2 s (High)

0.1 s (High)

8 s (High)

2 s (Low)


0.1 s (Low)

2 s (Low)

8 s (High)

8 s (Low)

2 s (High)

0.1 s (High)


0.1 s (Low)

8 s (High)

0.1 s (High)

2 s (Low)

8 s (Low)

2 s (High)


2 s (Low)

0.1 s (High)

8 s (Low)

2 s (High)

0.1 s (Low)

8 s (High)


8 s (High)

2 s (Low)

2 s (High)

0.1 s (Low)

0.1 s (High)

8 s (Low)


Day Week 1 Week 2 Week 3 Week 4 Week 5


Sunday

Monday

Tuesday

Wednesday

Thursday

Friday


16 s (Low)

0.1 s (Low)

4 s (High)

0.1 s (High)

16 s (High)


0.1 s (Low)

4 s (Low)

16 s (High)

16 s (Low)

4 s (High)


4 s (Low) 0.1 s (High)


0.1 s (Low)

16 s (High)

0.1 s (High)

4 s (Low)

16 s (Low)


4 s (Low)

0.1 s (High)

16 s (Low)

4 s (High)

0.1 s (Low)


4 s (High) 16 s (High)


16 s (High)

4 s (Low)

4 s (High)

0.1 s (Low)

0.1 s (High)

16 s (Low)


The top portion of the table shows the weekly regimen for subject's 1125F, 1126F,
1127F, 1131M, and 1132M in experimental Conditions 1 and 2. The lower portion of the
table shows the weekly regimen for subject's 1134M and 1136M in experimental
Conditions 1 and 2. A high in parenthesis are sessions when the adjusting delay was 20 s
at the start of the session. A low in parenthesis are sessions when the adjusting delay was
6 s at the start of the session. This five-week regimen was recycled a minimum of three
times.










Condition 3. The final condition of the experiment included an assessment of

indifference points when the same standard delay was presented across all sessions for an

entire five-week period. Table 3 shows the order of standard delay presentations for all

subjects in this condition of Experiment 2. This condition was designed to investigate

if different indifference points would be observed when the standard delay remained

fixed instead of changing each day of the week as in Conditions 1 and 2. This condition

only included six of the original seven subjects as one subject died during Condition 3

and those data were excluded.


Table 3. The order of standard delay presentation during experimental Condition 3.

1126F 1127F 1131M 1132M 1134M 1136M

First 0.1 s 8 s 8 s 0.1 s 0.1 s 16 s

Second 2s 2s 2s 2s 2s 2s

Third 8s 0.1 s 0.1 s 8s 4s 4s

Fourth 16s 0.1 s

Each standard delay was presented for a single five-week period with the same order of
high and low start values of the adjusting delays as shown in Table 2.


Results

Condition 1. Figures 3 through 9 show changes in the adjusting delay across

trials during the final five weeks of the first experimental Condition. Each figure

includes data for one subject. The vertical axis of each plot shows the adjusting delay in

seconds ranging from 0 to 120 s or 0 to 70 s. Each graph includes two sessions with the

same standard delay, indicated by the same symbols, but different starting values for the









adjusting delay of either 6 s (open symbol) or 20 s (filled symbol). The dotted vertical

line in each graph references trial thirty after which within-session stability was

determined. The legend to the right of each graph indicates the order of standard delay

presentation for that week starting from Sunday through Friday. Table 4 shows the mean

adjusting delay when stability was reached during the session for 1125F, 1126F, and

1127F. Table 5 shows the mean adjusting delay when stability was reached during the

session for subjects 1131M, 1132M, 1134M, and 1136M. Both tables show the mean

adjusting delay across the first ten-trial block after trial 30 that met the within session

stability criteria and an asterisk indicates that those criteria were not met within that

session. The within-session stability criteria were not met during some sessions as the

subjects completed fewer than 40 choice trials during the session as noted in parenthesis.

Subject 1125F developed stable indifference points during 27 out of 30 sessions

in the final five-weeks of the first experimental condition shown in Figure 3. Table 4 has

asterisks to indicate the three sessions during which stability was not reached within the

session. Figure 3 shows that some sessions included the development of a stable

adjusting delay over a ten trial block that was then followed by an unstable adjusting

delay where subject 1125F repeatedly chose one alternative. For instance, in week one,

the adjusting delay fulfilled the within session stability criteria across trials 41 through 50

with a mean adjusting delay of 47 s that was followed by persistent choices for the

standard alternative of one pellet delayed by 8 s producing a consistent decrease in the

adjusting delay and unstable performance. As indicated in Table 4, the start value of the

adjusting delay sometimes influenced the mean adjusting delay at indifference.

Sometimes the indifference point was higher when the adjusting delay started low rather









than high. Both Table 4 and Figure 3 show that the range of mean adjusting delays at

indifference overlapped across standard delays, i.e., 4 s to 33 s, 7 s to 14 s, and 12 s to 47

s with standard delays of 0.1 s, 2 s, and 8, respectively. Regardless of this similarity

across standard delays, stable indifference was observed much earlier in the session when

the standard delay was 0.1 s than when it was either 2 s or 8 s.

Figure 4 shows data for subject 1126F during the final five-week period of the

first condition. All sessions included the establishment of indifference points (stable

adjusting delays) within the session. The development of indifference was

followed by two different within-session patterns. During some sessions, stable

indifference was followed by unstable performance during the final trials of the session.

This pattern is demonstrated when the standard delay was 8 s in the second week. The

adjusting delay was increased from 6 s to 68 s over trials 1 to 45 and performance met

stability criteria during trials 31 through 40. This was followed by a brief increase and

then a consistent decrease across the remaining trials of that session with the session

terminating on unstable within-session performance. A second within-session pattern

that was observed included the development of a stable adjusting delay at two different

times during the same session. This pattern is shown in the fifth week when the standard

delay was 8 s. The adjusting delay increased from 20 to 115 s over trials 1 through 30

and then decreased to approximately 70 s during trials 30 through 45, meeting the

stability criterion, and then decreased to a stable mean adjusting delay of 48 s during

trials 51 through 60. Two different stable indifference points were obtained during that

single session, 70 s and 48 s. This within-session pattern was seen with the high standard

delay than the two shorter standard delays. Table 5 shows that the mean adjusting delays








36
ranged from 12 s to 38 s, 3 s to 45 s, and 21 s to 73 s when the standard delays were 0.1,

2, or 8 s, respectively. The range of indifference points was more similar when the

standard delay was 0.1 s or 2 s but extended up through larger delays when the standard

delay was 8 s.

The data for subject 1127F in Figure 5 show the development of stable

indifference within 28 out of 30 sessions during the final five-weeks of Condition 1.

Stability was not met during these two sessions because less than 40 choice trials were

completed during those two sessions. Several sessions included the development of

stable indifference at some point in the session followed by a consistent decrease in the

adjusting delay. This is seen in the fifth week during trials 31 to 40 when the standard

delay was 2 s and the mean adjusting delay for three pellets was 61 s. Although

performance was stable at trial 40, subject 1127F consistently chose the standard

alternative for the remainder of the session producing a steady decrease in the adjusting

delay. Data from the second week shows a similar pattern of stability followed by a

decline in the adjusting delay during both sessions when the standard delay was 8 s.

Figure 5 also shows that, although some overlap was observed, the adjusting delay was

consistently higher with the higher standard delays. This is supported by the range of

indifference points observed which were 1 s to 11 s, 2 s to 14 s, and 11 s to 38 s with

standard delays of 0.1 s, 2 s, or 8 s, respectively.

Figure 6 shows that subject 1131M developed a stable indifference point in 28 out

of 30 sessions during the final five-week period of the first condition. Subject 1131M's

general pattern within a session was to increase or decrease the adjusting delay until the

subject was indifferent between the two alternatives and then to alternate between the two









Table 4. The Mean Adjusting Delay at Indifference in Condition 1.


1125F Week 1 Week 2 Week 3 Week 4 Week 5
0.1 s (Low) 14(41-50) 6(31-40) 4(31-40) 5(31-40)
0.1 s (High) 5 (31-40) 33 (41-50) 33 (41-50) 7(31-40) 6(31-40)
2 s (Low) 9 (41-50) 13 (36-45) 12 (36-45) 13 (36-45) 14 (46-55)
2 s (High) 13 (36-45) 7(31-40) 11(31-40) 10(41-50)
8 s (Low) 47 (41-50) 34 (46-55) 42(51-60) 16 (41-50) 19(36-45)
8 s (High) 12(46-55) 16(36-45) 24(31-40) 23(31-40)

1126F Week 1 Week 2 Week 3 Week 4 Week 5
0.1 s (Low) 21(31-40) 18(51-60) 12(36-45) 21(31-40) 38(36-45)
0.1 s (High) 27 (36-45) 38 (41-50) 38 (41-50) 29 (36-45) 25 (41-50)
2 s (Low) 23 (51-60) 45 (31-40) 42(46-55) 23 (31-40) 3 (36-45)
2s (High) 35 (41-50) 23 (36-45) 31(36-45) 41(31-40) 35(46-55)
8 s (Low) 34(41-50) 68(31-40) 46(46-55) 34(31-40) 39(36-45)
8 s (High) 53 (31-40) 73 (41-50) 21(46-55) 55 (31-40) 70(36-45)

1127F Week 1 Week 2 Week 3 Week 4 Week 5
0.1 s (Low) 1 (36-45) 1 (31-40) 1 (31-40) 1 (31-40) 6(31-40)
0.1 s (High) 1(31-40) 1(31-40) 11(31-40) (Not finish) 1(31-40)
2 s (Low) 7 (31-40) 4(31-40) 11(31-40) 4(31-40) 61(31-40)
2 s (High) 2 (31-40) 6 (31-40) 4 (31-40) 5 (31-40) 4 (31-40)
8 s (Low) 15 (31-40) 39(46-55) 14(36-45) 12(36-45) 11(31-40)
8 s (High) 37 (31-40) 47 (41-50) 38 (36-45) (Not finish) 12 (31-40)
Mean adjusting delay, to the nearest second, when stability was reached within a session
for subject's 1125F, 1126F, and 1127F. The choice trials comprising this mean are
shown in parentheses.


alternatives for the remainder of the session. Only a few sessions terminated with a

decreasing (when standard delay was 8 s in the second week) or increasing (when the

standard delay was 8 s in week 1) trend in the adjusting delay. Table 5 indicates that the

adjusting delays at indifference were very similar across standard delays with all of the

mean adjusting delays ranging from 6 s (only one session) to 55 s while the central

tendency ranging from 25 s to 42 s across all standard delays. Nonetheless, the highest

indifference points were correlated with the highest standard delays.








38
Data for subject 1132M are shown in Figure 7. The data from the final five-week

period for this subject included three sessions during which the subject continually

preferred the large reinforcer (adjusting alternative), which resulted in steady increases of

the adjusting delay until the session terminated. Two of these sessions occurred when the

standard delay was 0.1 s (terminal adjusting delay of 142 s and 200 s) and one when the

standard delay was 2 s with a terminal adjusting delay of 170 s. One session in the fifth

week did not include the development of stable indifference due to consistent preference

for the standard alternative. The remaining sessions included less exclusive responding

and 26 out of 30 sessions included the development of stable indifference at some point

during the session. The range of stable indifference points were 11 s to 28 s, 12 s to 36 s,

and 16 s to 46 s when the standard delay was 0.1 s, 2 s, or 8 s, respectively.

Figure 8 and 9 include data for subjects 1134M and 1136M, respectively. Unlike

the other five subjects in this experiment, these two subjects were exposed to standard

delays of 0.1 s, 4 s, and 16 s. Figure 8 shows that the within-session performance

of subject 1134M became stable or indifferent during 25 out of 30 of the final sessions in

Condition 1. Performance during one session did not meet the stability criteria as less

than 40 choice trials were completed during the session. Although many sessions

included the development of a stable adjusting delay, the performance did not always

remain stable for the remainder of the sessions. Subject 1134M consistently chose the

small reinforcer following either a 4 s or a 16 s standard delay during the final trials of

sessions in the first and second weeks producing consistent decreases in the adjusting

delay. The mean adjusting delays at stability were closely approximated across sessions

containing the same standard delay. For instance, the mean adjusting delay during the









Table 5. The Mean Adjusting Delay at Indifference Condition 1.


1131M Week 1 Week 2 Week 3 Week 4 Week 5
0.1 s (Low) 21(36-45) 6(31-40) 32(31-40) 34(41-50) 23 (36-45)
0.1 s (High) 42 (31-40) 38 (36-45) 24 (36-45) 25 (31-40)
2 s (Low) 38(31-40) 24(41-50) 34(36-45) 36(46-55) 43 (36-45)
2 s (High) 16 (46-55) 48 (31-40) 48 (31-40) 33 (31-40)
8 s (Low) 26(41-50) 31(41-50) 41(31-40) 38(41-50) 37(51-60)
8 s (High) 24 (31-40) 47 (36-45) 41(31-40) 55 (31-40) 42 (31-40)
1132M
Week 1 Week 2 Week 3 Week 4 Week 5
0.1 s (Low) 12(46-55) 28(46-55) 19(31-40)
0.1 s (High) 15(36-45) 11(36-45) 13(36-45) 15(36-45) 12(41-50)
2 s (Low) 25 (36-45) 12 (36-45) 30 (46-55) 14 (51-60)
2 s (High) 13 (31-40) 23(31-40) 19(31-40) 36(46-55) 12(31-40)
8 s (Low) 16(31-40) 16(41-50) 46(46-55) 20(31-40) 32(36-45)
8 s (High) 21 (36-45) 23 (36-45) 18 (46-55) 32 (41-50) *
1134F
Week 1 Week 2 Week 3 Week 4 Week 5
(Not
0.1 s (Low) 21(31-40) finish) 29 (36-45) 10(41-50) 20 (36-45)
0.1 s (High) 12 (36-45) 25 (36-45) 31(31-40) 30 (31-40) 24(31-40)
4 s (Low) 58(31-40) 48(36-45) 36(46-55) 52(31-40) 26(51-60)
4 s (High) 25 (36-45) 39(31-40) 38(51-60) 37(31-40) *
16 s (Low) 63 (31-40) 52 (31-40) 52 (31-40)
16 s (High) 40(41-50) 40(31-40) 55(31-40) 42(31-40)
1136F
Week 1 Week 2 Week 3 Week 4 Week 5
0.1 s (Low) 24(31-40) 27(31-40) 35 (31-40) 10(41-50) 20(36-45)
0.1 s (High) 30(31-40) 30(31-40) 41(36-45) 30(31-40) 24(31-40)
4 s (Low) 36(31-40) 29(31-40) 51 (36-45) 52(31-40) 26(51-60)
(Not
4 s (High) 32 (31-40) finish) 37 (41-50) *
16 s (Low) 25 (41-50) 52 (36-45) 52(31-40)
16 s (High) 51(31-40) 50(31-40) 60(31-40) 42(31-40)
Mean adjusting delay, to the nearest second, when stability was reached within a session
for subject 1131M, 1132M, 1134M, and 1136M during Condition 1. The choice trials
comprising this mean are shown in parentheses.


fifth week was 52 s and 42 s when the standard delay was 16 s but 20 s and 24 s when the

standard delay was 0.1 s. Indifference points increased with increased standard delays;








40
they ranged from 10 s to 31 s, 25 s to 58 s, and 40 s to 63 s with standard delays of 0.1 s,

4 s, and 16 s.

The performance of subject 1136M showed the development of stable

indifference within the session for 24 out of 30 of the sessions included in Figure 9. Two

of the sessions that did not include the development of stable performance, included

persistent responding for the adjusting alternative or large reinforcer (three pellets) when

the standard delay to one pellet was 4 s or 16 s and both of these performances were

observed during the 5th week. The mean adjusting delay was very similar across standard

delays with ranges of 10 s to 41 s, 29 s to 53 s, and 25 s to 52 s with standard delays of

0.1 s, 4 s, and 16 s, respectively. Similar indifference points were observed when the

standard delay was either 4 s or 16 s but the lowest indifference points were shown with

the 0.1 s standard delay.

Several general choice patterns emerge from the data in Figures 3 through 9. All

of the subjects were beginning to show stable adjusting delays at some point during the

final portion of individual sessions, i.e., trials 30 through 60, during at least 80% of the

30 sessions from the final 5 weeks of Condition 1. Secondly, subjects completed almost

all 60 possible choice trials during the 2-hour experimental session. Although the

adjusting delay often became stable during the session and then unstable during the final

trials of the session. Fourth, the indifference point tended to be higher with higher

standard delays although many indifference points were nominally similar across

standard delays especially during sessions including the standard delays of 0.1, 2, or 4 s.

Finally, sessions when the standard delay was highest, 8 or 16 s, included more within-









session variability which is mostly likely due to how the adjusting delay was

programmed to change.

Figure 10 is a summary plot of the data in Condition 1. Each graph shows the

mean adjusting delay at indifference when the within session stability criteria were met as

indicated in Tables 4 and 5. Each mean adjusting delays was determined across the final

10 trials of the five individual sessions shown in Figures 3 through 9. The filled symbols

indicate the means when the adjusting delay was 20 s at the beginning of each session

while the open symbols indicate sessions when the adjusting delay was 6 s. Indifference

points were not consistently modified by the adjusting delay at the beginning of the

session but instead the indifference points were similar whether the adjusting delay was

20 s or 6 s at the beginning of the session. The summary plot demonstrates some across

session variability in the indifference point in sessions with the same standard delay but

an increasing indifference point with increased standard delays to one pellet.

Condition 2. Figures 11 through 17 show choice performance during the final

five-week period of the second experimental condition when session termination was

based on meeting the within-session stability criteria. Each figure includes five graphs

corresponding to the final five-weeks of this condition. Each graph includes within-

session data for six sessions completed during a week. A legend is provided to the right

of each graph indicating the order of standard delay presentations from Sunday through

Friday of that week. A dotted reference line is included at trial 30 after which within-

session stability was determined (vertical line on each graph). Table 6 shows the mean

adjusted delay during the final 10 choice trials of each session for subject's 1125F,




















































Figure 3. Condition 1 for subject 1125F. This figure shows within-session data during
the final five weeks of the first condition. Each graph plots the adjusting
delay in seconds across trials within the session for all six sessions of each
week. The graphs include two sessions with the same standard delay (same
symbols) when the adjusting delay started at either 6 s (open symbols) or 20 s
(filled symbols).






























































0 10 20 30 40 50
Trials Within a Session


El

e.







p.FF

A


e.





E3




e






E3
C.


(Low)
s (Low)
(High)
s (High)
(High)
(Low)


s (Low)
(Low)
(High)
(Low)
(High)
s (High)



s (Low)
(High)
s (High)
(Low)
(Low)
(High)


(Low)
s (High)
(Low)
(High)
s (Low)
(High)


(High)
(Low)
(High)
s (Low)
s (High)
(Low)





















































Figure 4. Condition 1 for subject 1126F. This figure shows within-session data during
the final five weeks of the first condition. Each graph plots the adjusting delay
in seconds across trials within the session for all six sessions of each week. The
graphs include two sessions with the same standard delay (same symbols) when
the adjusting delay started at either 6 s (open symbols) or 20 s (filled symbols).










120
100 Week 1 1126F
S8 s (Low)
80 0.1 s (Low)
60 1 2 s (High)
40 0.1 s (High)
20 *8 s (High)
0 ,- 2 s (Low)
120
100 -Week 2
100- Week 0.1 s (Low)
80 2 s (Low)
60 8 s (High)
40 B.. .A ..t -B- 8 s (Low)
0 SR A 2 s (High)

20
Ti als0.1 s (High)
(,) 120
0loo Week 3
0.1 s (Low)
" 80 -,- _- 8 s (High)
(,
-0 60 0.1 s (High)
40 2 s (Low)
S2- 8 s (Low)
20 0
__ 2s(High)
120
10oo Week 4
2 s (Low)
280 -,- 0.1 s (High)
601 3 8 s (Low)
40 '2 s (High)
20 A 0.1 s (Low)
0 8 s (High)
120
loo Week 5
..8 s (High)
80 .* .2 s (Low)
60 .+* .2 s (High)
40 0.1 s (Low)
20 -- 0.1 s (High)
8 s (Low)
0 10 20 30 40 50 60
Trials Within a Session




















































Figure 5. Condition 1 for subject 1127F. This figure shows within-session data during
the final five weeks of the first condition. Each graph plots the adjusting
delay in seconds across trials within the session for all six sessions of each
week. The graphs include two sessions with the same standard delay (same
symbols) when the adjusting delay started at either 6 s (open symbols) or 20 s
(filled symbols).










70
60 Week 1 1127F
50 8 s (Low)
40 e.- 0.1 s (Low)
30 2 s (High)
20 -. 0.1 s (High)
10 --1, 8 s (High)
o................ ...... 2 s (Low )
70

60 Week 2
50 -e- 0.1 s (Low)
.2 s (Low)
40
8 s (High)
30
20 *B 8 s (Low)
10 2 s (High)
10 -
0--- 0.1 s High
7 0 ,
,) 70
60 Week 3
> *0.1 s (Low)
Cu 50
Z- 40 8 s (High)
-0 0.1 s (High)
30
0 --2 s (Low)
C 20 -
0 s. 8 s (Low)
S 10 --.0. (Lw
= 0 -- s (High)
0 70
60 Week 4
50 2 s (Low)
40 0.1 s (High)
30 8 s (Low)
20 12 s (High)
a0.1 s (Low)
10 -
.. .08. s (High)
0
70 -
60 Week 5
60 .I ,/ ,,. i 8 s (High)
-..40 -. 8sgiE 2 s (Low)
4030 2 s (High)
S0.1 s (Low)
200.1 s (High)
10
8 s (Low)
0 10 20 30 40 50 60
Trials Within a Session



















































Figure 6. Condition 1 for subject 1131M. This figure shows within-session data during
the final five weeks of the first condition. Each graph plots the adjusting
delay in seconds across trials within the session for all six sessions of each
week. The graphs include two sessions with the same standard delay (same
symbols) when the adjusting delay started at either 6 s (open symbols) or 20 s
(filled symbols).










70
60 Week 1 1131M -
0 -El (Low)
40 0.1 s (Low)
30 .' *'' *** ...*.2 s (High)
20 0.1 s (High)
10 -- 8 s (High)
50 -- 0. 2 s (Low)
70
60 -Week 2
20.1 s (Low)
0 --- 2 s (Low)
40
18-- 2 s (High)
30
60 W- -3 8 s (Low)
20
1. 2 s (High)
10 -
-- 0.1 s (High)
0 0_
(, 70
60 Week 3
S>50 0.1 s (Low)
650
4- 8 s (High)
40 -A--- 2** (Low)
0' .0.1 s (High)
30
S202 s (Low)
20
3 8 s (Low)
Sn,2 s (High)
60
70 -
60 Week 4
50 1 22 s (Low)
40 T 0.1 s (High)
30 P. 8 s (Low)
20 _- .- ..... 2 s (High)
10 0.1 s (Low)
0 ----- 8 s (High)
70

50 8 s (High)
40 2 s (Low)
.* 2 s (High)
20 0.1 s (Low)
t 0.1 s (High)
10
8 s (Low)
0 10 20 30 40 50 60
Trials Within a Session


















































Figure 7. Condition 1 for subject 1132M. This figure shows within-session data during
the final five weeks of the first condition. Each graph plots the adjusting
delay in seconds across trials within the session for all six sessions of each
week. The graphs include two sessions with the same standard delay (same
symbols) when the adjusting delay started at either 6 s (open symbols) or 20 s
(filled symbols).












8 s (Low)
0.1 s (Low)
2 s (High)
0.1 s (High)
8 s (High)
2 s (Low)


70
60
50
40
30
20
10
0
70
60
50
40
30
20
10
7 0
(D 70
60
cu 50
(D 40
30
C 20
n 10
3 0
20
< 70
60
50
40
30
20
10
0
70
60
50
40
30
20
10
0


s (Low)
(Low)
(High)
(Low)
(High)
s (High)


0.1 s (Low)
8 s (High)
0.1 s (High)
2 s (Low)
8 s (Low)
2 s (High)


(Low)
s (High)
(Low)
(High)
s (Low)
(High)


(High)
(Low)
(High)
s (Low)
s (High)
(Low)


0 10 20 30
Trials Within a


40 50
Session


-e-- 0.1
-- 2s
--- 8s
--8 s
S2s
0.1


El
e


e




El




A

E3
4
e









E3


















































Figure 8. Condition 1 for subject 1134M. This figure shows within-session data during
the final five weeks of the first condition. Each graph plots the adjusting
delay in seconds across trials within the session for all six sessions of each
week. The graphs include two sessions with the same standard delay (same
symbols) when the adjusting delay started at either 6 s (open symbols)
or 20 s (filled symbols).










120
1134M
100 Week 1 16 s (Low)
80 0.1 s (Low)
60 4 s (High)
40 0.1 s (High)
20 -+fFf ** 'C' '16 s (High)
20
0 --4 s (Low)
120 -
10 -Week 2
1k 2 0.1 s (Low)
80 -- 4 s (Low)
60 16 s (High)

.0.4 s (High)
20
) 0.1 s (High)
o
S120
10 Week 3
100 0.1 s (Low)
80 16 s (High)
-0 60 0.1 s (High)
40 4 s (Low)
4..0 .16 s (Low)
20 -- 4 s (High)
-0
120
0lo Week 4
S- 4 s (Low)
80 0.1 s (High)
60 E3 16 s (Low)
40 --* 4 s (High)
.20 ..* ...... G 0.1 s (Low)
20 -----16 s (High)
120
loo Week 5
16 s (High)
80 A 4 s (Low)
60 A 4 s (High)
40 0* *** *0.1 s (Low)
20 .0.1 s (High)
0 16 s (Low)
0 10 20 30 40 50 60
Trials Within a Session




















































Figure 9. Condition 1 for subject 1136M. This figure shows within-session data during
the final five weeks of the first condition. Each graph plots the adjusting
delay in seconds across trials within the session for all six sessions of each
week. The graphs include two sessions with the same standard delay (same
symbols) when the adjusting delay started at either 6 s (open symbols) or 20 s
(filled symbols).










120
1136M
oo Week 1 1136M
16 s (Low)
80 -- 0.1 s (Low)
60 ... 4 s (High)
40 0.1 s (High)
20 16 s (High)
20 4 s (Low)
120
o100o- Wek 0.1 s (Low)
80 1- 4 s (Low)
60 016 s (High)
40 'n^ ^ ^ ^ ^ *^ 0 16 s (Low)
2.0 -. 4 s (High)
0.1 s (High)
o
( 120
100 Week 3
> 100 0.1 s (Low)
80 -- -- 16 s (High)
(D 60 0.1 s (High)
C 40 -' -u *. :.4 s (Low)
E3 16 s (Low)
C/) 20
2-- 4 s (High)
120
100 Week 4
4 s (Low)
80 0.1 s (High)
60 -- 16 s (Low)
40 -- 4 s (High)
20 *+ 0.1 s (Low)
16 s (High)
120
loo Week 5
: 16 s (High)
80 B 4 s (Low)
60 .4 s (High)
40 0.1 s (Low)
0.1 s (High)
016 s (Low)
0 10 20 30 40 50 60
Trials Within a Session














80

60

40

20

0
80

60

40
>,
-r 20
U,
S0
.E 80

:' 60
<
- 40
,U


1134M


1131M
60

40

20

0


Standard Delay (sec)


0.1 4 16

Standard Delay (sec)


Figure 10. Condition 1 Summary Plot. This figure shows mean adjusting delay and 95%
confidence intervals of the data in the second experimental condition over the
last ten trials of the session as a function of the standard delay. The open
symbols show data from sessions during which the adjusting delay started at 6
s and the filled symbols plot data from when the adjusting delay started at 20


6 s Start
- 20 s Start


1127F








57
1126F, and 1127F and Table 7 shows the mean adjusting delay during the final 10 choice

trials of each session for subjects, 1131M, 1132M, 1134M, and 1136M.

Figure 11 shows data for Subject 1125F which included the development of stable

indifference and early termination of 29 out of 30 experimental sessions during the final

five week period of Condition 2. Indifference points were highly correlated with the

standard delay with higher standard delays resulting in higher adjusting delays. For

instance, the mean adjusting delays from the fourth week were 3 s, 14 s, and 56 s when

the standard delays were 0.1 s, 2 s, or 8 s, respectively and the start value of the adjusting

delay was 6 s. Furthermore, the mean adjusting delays were 3 s, 12 s, and 40 s when the

standard delays were 0.1 s, 2 s, and 8 s, respectively and the start value of the adjusting

delay was 20 s. The mean adjusting delay tended to be more similar when the standard

delay was either 0.1 or 2 s than when the standard delay was 8 s during all five weeks

shown in Figure 11. The mean adjusting delays were less than 10 s in 3 out of the four

indifference points in that second week when the standard delay was 0.1 s or 2 s while the

indifference points were much higher (43 s and 118 s) when the standard delay was 8 s.

The mean adjusting delays observed with a given standard delay during the fourth week

were almost identical regardless of the start value of the adjusting delay.

The performance of subject 1126F met the within session stability criteria, shown

in Figure 12, in 28 out of 30 of the sessions during the final five-weeks of Condition 2.

Early session termination was typically within the first fifty trials with most sessions

terminating after trial 45. Stability was not reached within the session for two sessions

when the standard delay was 8 s and the adjusting alternative started at 20 s. Table 6

shows the indifference points ranged from 6 s to 55 s, 2 s to 54 s, and 19 to 95 s with








58
standard delays of 0.1 s, 2 s, and 8 s. Indifference points were virtually indistinguishable

when the standard delay was 0.1 s or 2 s.

Subject 1127F developed stable indifference within the session for 27 out of 30 of

the sessions. Overall, the mean adjusting delay at indifference was consistently lower

than observed for the other subjects. The adjusting delay at indifference was consistently

2 seconds when the standard delay was 0.1 s, with one exception, while the adjusting

delay only ranged from 8 s to 20 s when the standard delay was 8 s. Although there were

some similarities in the indifference points across standard delays, the indifference points

tended to increase as the standard delay increased. For instance, the indifference points

during the fifth week were 2 s and 2 s, 10 s and 6 s, or 10 s and 20 s when the standard

delays were 0.1 s, 2 s, or 8 s and the adjusting delay started at 6 s and 20 s, respectively.

Figure 13 shows that sessions typically terminated following completion of only 40

choice trials, as performance was stable. As a result, sessions only lasted approximately

twenty-five to thirty minutes.

Subject 113 IM's data are shown in Figure 14 and this subject showed stable

indifference points during 29 of 30 of the final sessions of the second experimental

condition. The only session that did not include stable performance was because fewer

than 40 choice trials being completed during that session. Again, the mean adjusting

delays tended to increase with increasing standard delays for this subject as seen with

previous subjects. The performance of the first week shows indifference points of 2 s or

12 s, 13 s or 23 s, and 61 s or 33 s for standard delays of 0.1 s, 2 s, or 8 s and low or high

adjusting delays at the start of the session, respectively. Similarly, the performance of the








59
second week shows indifference points of 2 s or 21 s, 11 s, and 26 s or 29 s for standard

delays of 0.1 s, 2 s, or 8 s, and low or high beginning adjusting delays, respectively.


Table 6. The Mean Adjusting Delay at Indifference in Condition 2.


1125F Week 1 Week 2 Week 3 Week 4 Week 5
1 s (Low) 2 5 8 3 7
0.1 s (High 2 4 42 3 3
2 s (Low) 6 27 13 14 45
2 s (High) 21 2 16 12 13
8 s (Low) 80 43 56 29
8s (High) 41 118 94 40 58

1126F Week 1 Week 2 Week 3 Week 4 Week 5
0.1 s (Low) 23 10 36 31 34
0.1 s (High 9 6 53 28 55
2 s (Low) 2 34 34 29 54
2 s (High) 19 10 33 25 42
8 s (Low) 19 52. 56 41 62
8 s (High) 95 49 45

1127F Week 1 Week 2 Week 3 Week 4 Week 5
0.1 s(Low) 2 2 2 2 2
0.1 s (High 2 2 19 2 2
2 s (Low) 12 7 18 10
2 s (High) 2 3 4 2 6
8 s (Low) 11 8 14 21 10
8 s (High) 18 16 20
Mean adjusting delay, to the nearest second, when stability was reached within a session
in Condition 2 for subject's 1125F, 1126F, and 1127F.


The data for subject 1132M are shown in Figure 15 and include the development

of stable adjusting delays for 28 of the 30 sessions (indicated in Table 7). The two

sessions when the subject did not show the development of indifference was with a

standard delay of 8 s. Subject 1132M established similar indifference points when the

standard delay was either 0.1 s or 2 s such as seen in the second week with means of 13 s

(low start value) or 5 s (high start value) when the standard delay was 0.1 s and 21 s or 13








60
s when the standard delay was 2 s. Regardless of this similarity, indifference points were

highest when the standard delay was 8 s at 42 s and 22 s in week 2.

Figures 16 and 17 show data for subjects 1134M and 1136M, respectively, who were

exposed to 0.1 s, 4 s, and 16 s standard delays to one pellet. Both of these subjects also

showed the development of stable indifference within the session for most of the sessions

shown. Subject 1134M was unable to meet the within session stability criteria in 6 out of

the final 30 sessions of Condition 2. Subject 1134M did not develop stable performance

during the 1st and 2nd week with a 4 s standard delays and the 4th week for a 0.1 s standard

delay. Some sessions during the second week did not meet the within session stability

criteria because the subject did not complete at least thirty choice trials before the session

terminated. During sessions when stability criteria were met, the performance of 1134M

showed increasing adjusting delays with increased standard delays. For example, the

mean adjusting delay ranged from 5 s to 22 s when the standard delay was 0.1 s, ranged

from 17 s to 33 s when the standard delay was 4 s, and ranged from 12 s to 89 s when the

standard delay was 16 s. This example also illustrates some of the similarities of

indifference points across standard delays as the ranges overlap. An overlap of

indifference points across standard delays was also seen in the performance of subject

1136M. The mean adjusting delays ranged from 6 s to 32 s when the standard delay was

0.1 s, ranged from 13 s to 27 s when the standard delay was 4 s, and ranged from 16 s

(the lowest possible value) to 49 s when the standard delay was 16 s. The mean adjusting

delays were virtually indistinguishable for sessions when the standard delay was either

0.1 s or 4 s and irrespective of the start value of the adjusting delay. The performance of

subject 1136M met the within session stability criteria in all 30 sessions shown in









Table 7. The Mean Adjusting Delay at Indifference in Condition 2.


1131M Week 1 Week 2 Week 3 Week 4 Week 5
0.1 s(Low) 2 21 5 9 13
0.1 S (High) 12 2 33 23 22
2 s (Low) 13 11 33 27 29
*(Not
2 s (High) 23 finish) 28 40 23
8 s (Low) 61 26 29 33 39
8 s (High) 33 29 26 33 34

1132M Week 1 Week 2 Week 3 Week 4 Week 5
0.1 s(Low) 4 13 9 6 9
0.1 s (High) 2 5 13 10 3
2 s (Low) 8 21 11 13 22
2 s (High) 12 13 12 12 11
8 s (Low) 16 42 29 21 13
8 s (High) 32 22 32 *

1134M Week 1 Week 2 Week 3 Week 4 Week 5
0.1 s(Low) 5 22 19 15 12
0.1 s (High) 13 12 22 18
4 s (Low) 17 27 21 23 23
(Not
4 s (High) finish) 23 33 28
16 s (Low) 12 39 43 48
(Not
16 s (High) 23 finish) 20 89

1136M Week 1 Week 2 Week 3 Week 4 Week 5
0.1 s(Low) 12 22 9 17 6
0.1 s (High) 32 13 21 13 12
4 s (Low) 17 20 16 23 13
4 s (High) 23 18 27 22 19
16 s (Low) 48 23 27 22 16
16 s (High) 49 29 24 13 22
Mean adjusting delay, to the nearest second, when stability was reached within a session
in Condition 2 for subjects 1131M, 1132M, 1134M, and 1136M.

Figure 17. However, stability was not reached until the final choice trials for two

sessions (see Week 1 and Week 4). Despite the sometimes similar indifference









points, the mean adjusting delay was highest when the standard delay was 16 s even

though only slightly higher during some sessions.

Some general trends were observed in the performance of all subjects in this

experimental condition. First, indifference points were established for all subjects across

standard delays (also seen in many sessions of Condition 1). Second, the mean adjusting

delay at stability was similar when the starting value of the adjusting delay was high (20

s) or low (6 s). Third, the highest indifference points were observed when the standard

delay was longest at either 8 s or 16 s. However, the mean adjusting delays at stability

were somewhat similar across short standard delays of 0.1 s, 2 s, or 4 s. Finally, at least

80% of the sessions during this final five-week period (24 out of 30 sessions) included

the development of stable indifference and hence early session termination.

Figure 18 provides summary plots of mean indifference points across standard

delays during the final five-weeks of Condition 2. Each mean adjusting delays was

determined across the final 10 trials of the five individual sessions shown in Figures 11

through 17. The open symbols show data when the adjusting delay started at a low value

of 6 s; the closed symbols show data when the adjusting delay started at 20 s. The mean

adjusting delays at indifference for several subjects (1125F, 1126F, 1127F, and 1132F)

were almost the same when the standard delay to one pellet was either 0.1 s or 2 s

suggesting that a change in the delay from 0.1 s to 2 s does not alter the subjective value

of one pellet by much. The mean indifference points for subjects 1134M and 1136M

were closely approximated when the delay to one pellet was 0.1 s or 4 s. As previously



















































Figure 11. Condition 2 for subject 1125F. This figure shows within-session data for
the final five weeks of the second condition. Each graph plots the adjusting
delay in seconds across trials within the session for all six sessions of each
week. The graphs include two sessions with the same standard delay (same
symbols) when the adjusting delay started at either 6 s (open symbols) or 20 s
(filled symbols). The vertical dotted line at trial 30 indicates the trial after
which stability was determined within the session.










120
100 -Week 1 1125F
80 -18 8 s (Low)
0 0.1 s (Low)
60 --- 2 s (High)
40 0.1 s (High)
620 8 s (High)
o0 s 2 s (Low)
140
120 -Week 2 .1 s
S- 0.1 s (Low)
100
S* 20. s (Low )
o W k 5-O
8 s (High)
60
S- -8 s (Low)
40 -
20 -- 2 s (High)
20
U. ~- 0.1 s (High)
0 120
o10 Week 3 S
0.1 s (Low)
" 80 -- 8 s (High)
-0 60 .-- 0.1 s (High)
40 2 s (Low)
0 E3.. 8 s (Low)
o3 2 s (High)
0 0
120
100oo Week 4
2 s (Low)
80 -
0.1 s (High)
60 -- 8 s (Low)
40 -- 2 s (High)
20 0.1 s (Low)
100 --C'- 8 s (High)
120
loo Week 5
8 s (High)
0 ,i- 2 s (Low)
60 2 s (High)
40 0.1 s (Low)
20 0.1 s (High)
0 8 s (Low)
0 10 20 30 40 50 60
Trials Within a Session



















































Figure 12. Condition 2 for subject 1126F. This figure shows within-session data for the
final five weeks of the second condition. Each graph plots the adjusting delay
in seconds across trials within the session for all six sessions of each week.
The graphs include two sessions with the same standard delay (same symbols)
when the adjusting delay started at either 6 s (open symbols) or 20 s (filled
symbols). The vertical dotted line at trial 30 indicates the trial after which
stability was determined within the session.














El

e.


0 10 20 30 40 50
Trials Within a Session


8 s (Low)
0.1 s (Low)
2 s (High)
0.1 s (High)
8 s (High)
2 s (Low)


0.1 s (Low)
2 s (Low)
8 s (High)
8 s (Low)
2 s (High)
0.1 s (High)


0.1 s (Low)
8 s (High)
0.1 s (High)
2 s (Low)
8 s (Low)
2 s (High)



2 s (Low)
0.1 s (High)
8 s (Low)
4 s (High)
0.1 s (Low)
8 s (High)


8 s (High)
2 s (Low)
2 s (High)
0.1 s (Low)
0.1 s (High)
16 s (Low)


e


A.




E3





E3




e




E)



















































Figure 13. Condition 2 for subject 1127F. This figure shows within-session data for the
final five weeks of the second condition. Each graph plots the adjusting delay
in seconds across trials within the session for all six sessions of each week.
The graphs include two sessions with the same standard delay (same symbols)
when the adjusting delay started at either 6 s (open symbols) or 20 s (filled
symbols). The vertical dotted line at trial 30 indicates the trial after which
stability was determined within the session.












Week 1 1127F








Week 2









Week 3









SWeek 4


Week 5








0 10 20 30 40 50 60
Trials Within a Session


El

e.






e



A.


E3



e.




C.FF


A




E3



E3


8 s (Low)
0.1 s (Low)
2 s (High)
0.1 s (High)
8 s (High)
4 s (Low)


0.1 s (Low)
2 s (Low)
8 s (High)
8 s (Low)
2 s (High)
0.1 s (High)


0.1 s (Low)
8 s (High)
0.1 s (High)
2 s (Low)
8 s (Low)
2 s (High)



2 s (Low)
0.1 s (High)
8 s (Low)
2 s (High)
0.1 s (Low)
8 s (High)


8 s (High)
2 s (Low)
2 s (High)
0.1 s (Low)
0.1 s (High)
8 s (Low)




















































Figure 14. Condition 2 for subject 1131M. This figure shows within-session data for the
final five weeks of the second condition. Each graph plots the adjusting delay
in seconds across trials within the session for all six sessions of each week.
The graphs include two sessions with the same standard delay (same symbols)
when the adjusting delay started at either 6 s (open symbols) or 20 s (filled
symbols). The vertical dotted line at trial 30 indicates the trial after which
stability was determined within the session.






























































0 10 20 30 40 50
Trials Within a Session


E3

p.









E.



E3



E)





A.

E3







E3
a,


(Low)
s (Low)
(High)
s (High)
(High)
(Low)


s (Low)
(Low)
(High)
(Low)
(High)
s (High)


s (Low)
(High)
s (High)
(Low)
(Low)
(High)



(Low)
s (Low)
(Low)
(High)
s (Low)
(High)


(High)
(High)
(High)
s (Low)
s (High)
(Low)



















































Figure 15. Condition 2 for subject 1132M. This figure shows within-session data for the
final five weeks of the second condition. Each graph plots the adjusting delay
in seconds across trials within the session for all six sessions of each week.
The graphs include two sessions with the same standard delay (same symbols)
when the adjusting delay started at either 6 s (open symbols) or 20 s (filled
symbols). The vertical dotted line at trial 30 indicates the trial after which
stability was determined within the session.












8 s (Low)
0.1 s (Low)
2 s (High)
0.1 s (High)
8 s (High)
2 s (Low)


70
60
50
40
30
20
10

78
60
50
40
30
20
10
0
70
60
50
40
30
20
10
0
70
60
50
40
30
20
10
0
120
100
80
60
40
20
0


- -
- --


Week .
-


- -



Week 2i


El

e*






A



E3























A
p
e.


0 10 20 30 40 50
Trials Within a Session


s (Low)
(Low)
(High)
(Low)
(High)
s (High)


Week 3


(Low)
s (High)
(Low)
(High)
s (Low)
(High)


(High)
(Low)
(High)
s (Low)
s (High)
(Low)


0.1 s (Low)
8 s (High)
0.1 s (High)
2 s (Low)
8 s (Low)
2 s (High)



















































Figure 16. Condition 2 for subject 1134M. This figure shows within-session data for the
final five weeks of the second condition. Each graph plots the adjusting delay
in seconds across trials within the session for all six sessions of each week.
The graphs include two sessions with the same standard delay (same symbols)
when the adjusting delay started at either 6 s (open symbols) or 20 s (filled
symbols). The vertical dotted line at trial 30 indicates the trial after which
stability was determined within the session.










120
0 Week 1 1134M
100 -
-- 16 s (Low)
80 -0.1 s (Low)
60 --4 s (High)
40 -- 0.1 s (High)
l~20 *16 s (High)
120 -- 4 s (Low)



80 -- 4 s (Low)
60 A -- 16 s (High)
40 16 s (Low)
40 -
... 4 s (High)
20
20 --- 0.1 s (High)
(D 128


80 16 s (High)
( 60 0.1 s (High)
40 -- 4 s (Low)
.. 16 s (Low)
---= 4s (High)
-0
S 120
100 Week 4
4 s (Low)
8o 0.1 s (High)
60 16 s (Low)
40 4 s (High)
20 -. 0.1 s (Low)
~t 16 s (High)
120
loo Week 5
6l 0t a-- 416 s (High)
80 4 s (Low)
60 4 s (High)
40 0.1 s (Low)
20 +- 0.1 s (High)
0 16 s (Low)
0 10 20 30 40 50 60
Trials Within a Session


















































Figure 17. Condition 2 for subject 1136M. This figure shows within-session data for the
final five weeks of the second condition. Each graph plots the adjusting delay
in seconds across trials within the session for all six sessions of each week.
The graphs include two sessions with the same standard delay (same symbols)
when the adjusting delay started at either 6 s (open symbols) or 20 s (filled
symbols). The vertical dotted line at trial 30 indicates the trial after which
stability was determined within the session.










70 -
60 Week 1 1136M
50 -- 16 s (Low)
0-e-- 0.1 s (Low)
40 -
30 -...4 s (High)
30
S....... 0.1 s (High)
20
10 --- 16 s (High)
10 -
0 4 s (Low)
70
60 Week 2
S0.1 s (Low)
50 -
--- 4 s (Low)

30 -9-- 16 s (Low)
20- 4 s (High)
S10 -- 0.1 s (High)
S0
S70
60 Week 3
>, 0.1 s (Low)
S50 -- 16 s (High)
(, 40- .
-0. 40.. 0.1 s (High)
30 --- 4 s (Low)
C 20- E3** *" 16 s (Low)
10 2- 4 s (High)
T0
" 70
60 Week 4
50 4 s (Low)
40 A. 0.1 s (High)
S16 s (Low)
20 .. ..
0.1 s (Low)
0'- 16 s (High)
70

60 Week 5
50 1s (Hig h)
40 4 s (Low)
30 4 s (High)
20 0.1 s (Low)
20
.10 0.1 s (High)
OEl 16 s (Low)
0 10 20 30 40 50 60
Trials Within a Session















80

60

40

20

0
80

a, 60
U,
> 40
co
0 20
CO
S80
80
- 60

S40
(,
- )n


1132M


1125F







1126F



E,^


-6 s Start
20 s Start


1127F


1134M


1136M


1131M


40

20

0
0.1 2 8
Standard Delay (sec)


Standard Delay (sec)


Figure 18. Condition 2 summary plot. This figure shows mean adjusting delay and 99%
confidence intervals of the data in the second experimental condition over the
last ten trials as a function of the standard delay. The open symbols show data
from sessions during which the adjusting delay started at 6 s and the filled
symbols plot data from when the adjusting delay started at 20 s.



noted, all of the subjects maintained the highest indifference points with the highest

standard delay of 8 s or 16 s. The figure also demonstrates the similarity of indifference

points with the same standard delay regardless of the starting value of the adjusting delay.


i


n

h~-"---i


I









This summary figure also demonstrates the linear changes in indifference points as the

standard delay increased.

Figure 19 summarizes the data shown in Figure 18 including a single mean

indifference point across each standard delay. Each mean adjusting delay was

determined by combining the data from the final 10 trials reaching stability when the

adjusting delay began the session at 6 s or 20 s. A line was then fitted to the data

describing the changes in the indifference point as a function of the standard delay to one

pellet. Figure 18 are indifference curves of the obtained mean adjusting delays at

indifference. Figure 19 are discount curves of best fitting lines. All of the individual

performance analyses produced positive y-intercepts and slopes. The y-intercepts were

approximately 20 s for four of the seven subjects (1126F, 1131M, 1134M, and 1136M)

with a range of 5.31 to 23.74. These findings suggest that when the delay to one pellet

was near zero, a delay between 5 and 24 s to three pellets was required to make them

equally preferred or chosen with equal frequency by the subjects. Analysis of individual

subject performance show slopes ranging from 1 to 6. The slope summarizes how a

change in the delay to the delivery of one pellet influences the adjusting delay to three

pellets that will produce equal preference or responding between the two alternatives. If

the slope is high, a change in the delay to one reinforcer alternative will then require a

large change in the adjusting delay to the other alternative before both are equally

preferred. If the slope is low, a change in the delay to one reinforcer alternative will then

require a small change in the adjusting delay to the other alternative before both are

equally preferred. The slopes were lower for the two subjects exposed to standard delays

of 0.1 s, 4 s, or 16 s (1.1 and 1.49) than for subjects exposed to standard delays of 0.1 s, 2




















































Figure 19. Linear Regressions for Condition 2. This figure shows scatter plots of the
mean adjusting delay as a function of the standard delay during the final
sessions of the second condition. The mean adjusting delay during the final
ten choice trials of the session when the adjusting delay began at either 6 or 20
s contribute to calculations of the mean (n=20 trials). A linear regression is
fitted to each subject's data and a group analyses is in the bottom right plot.












80
1125F
60

40

20

0
0 2 4
80
1126F
60

40

20

0
0 2 4
80
1127F
60

40

20

0
0 2 4 6
80
1131M
60

40

20 e

0
024


1132M


y=5.8x+12.21
r2=0.996

6 8 10 12 14 16 18








y=3.2x+25.78
r2=0.983

8 10 12 14 16 18


y=1.67x+5.31
r2=0.992

8 10 12 14 16 18


20 y=2.68x+8.23
r2=0.99
0
0 2 4 6 8 10 12 14 16 18
80
1134M
60

40

20 y=1.49x+17.95
r2=0.99
0
0 2 4 6 8 10 12 14 16 18
80
1136M
60


40

20

0
0 2 4 6
80 M
Mean


40
^ ^ r- ^


^_-

y=1.1x+23.74
r2=0.99

8 10 12 14 16 18


**


y=2x+21.45 zu y=1.64x+18.9
r2=0.8773 r2=0.858
0
6 8 10 12 14 16 18 0 2 4 6 8 10 12 14 16 18

Standard Delay (sec)








81
s, or 8 s (5.8, 3.2, 1.67, 2, and 2.68). The variance accounted for by the fitted line ranged

from 87% to 99% for individual subject data. The bottom right plot shows a linear

regression analysis for the mean across all of the subjects in this experiment. The

analysis accounted for 86% of the variability with a slope of 1.64 and an intercept of

18.9. The largest discrepancy from the fitted line was observed at the 8 s standard delay.

The group analysis produced slope and y-intercept parameters very similar to those found

with individual performances of the subjects.

Figure 20 shows linear regression analyses for the group data from the final

sessions of the first and second experimental conditions. Both scatter plots show the

mean adjusting delay for all subjects during the ten-trial block that reached the within

session stability criteria. The linear regression analyses showed positive y-intercepts and

slopes during both experimental conditions although both the slope and intercept

decreased across conditions. For instance, the y-intercept was 19.84 s at the end of the


- 80 80
( Condition 1 Condition 2
60 60

Q 40 40
0)
u 20 y=2.29x+19.84 20 y=1.64x+18.9
*y=1.64x+18.9
0 r2=0.823 r2=0.858
< 0 0
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16

Standard Delay (sec)

Figure 20. Linear regression comparisons for Conditions 1 and 2. This figure shows
summary data for all of the subjects in the first and second experimental
conditions. A group mean adjusting delay, seconds, is shown as a function of
the standard delay are shown in the scatter plot along with a linear regression
analysis.








82
first but 18.9 s at the end of the second experimental condition. The amount of variability

accounted for by the linear regression was higher for the second experimental condition.

Condition 3. Figures 21 and 22 show data from sessions in the second and third

experimental conditions. The mean adjusting delay during the final 10 choice trials of

the sessions is plotted as a function of sessions. The left column shows data when the

standard delay changed each day of the week (second experimental condition) and the

right column of each plot shows data when the standard delay remained the same (third

experimental condition). Open symbols are mean adjusting delays when the delay started

the session at a low value of 6 s while the filled symbols are means when the delay

started at a high value of 20 s. Data for six of the original seven subjects are shown in

Figures 21 and 22. One subject (1125F) was lost during the early sessions of the third

experimental condition thus those data were omitted. Across-session variability was seen

in the adjusting delays across sessions in both experimental conditions and similar

amounts of variability were detected in both conditions.

The two top rows of Figure 21 include data for subject's 1126F and 1127F. The

indifference points for subject 1126F are in the top row of Figure 21. As in the other

experimental conditions, the lowest indifference points were observed when the standard

delay was 0.1 s than when it was 2 s or 8 s. This pattern became more pronounced in

Condition 3 as evidenced by less nominal similarity of indifference points across standard

delays. This pattern was consistent when the adjusting delay began a session at high or

low value. The second row of graphs in Figure 21 is data for subject 1127F. The mean

adjusting delay varied between 5 s and 40 s across standard delays in both conditions

although most indifference points were near or below 20 s. The indifference points









varied across sessions although similar levels of variability were detected across both

conditions. In Condition 3, the mean adjusting delays were similar when the standard

delay was 0.1 s or 2 s with ranges of 2 s to 10 s and 3 s to 14 s, respectively. The range

of indifference points was a little higher when the standard delay was 8 s with a range of

8 s to 40 s. The mean adjusting delays were also similar across standard delays in

Condition 2.

The performance of subjects 1131M and 1132M is shown in the bottom portion of

Figure 21. Mean adjusting delays observed in Condition 2 for subject 1131M ranged

from 2 s to 37 s, 4 s to 39 s, and 27 s to 49 s when the standard delays were 0.1 s, 2 s, or 8

s, respectively. Such ranges in indifference points clearly include a large deal of overlap

when the standard delays were 0.1 s or 2 s. Similarly for subject 1131M in Condition 3,

the indifference points ranged from 9 s to 29 s, 5 s to 35 s, and 15 s to 57 s when the

standard delays were 0.1, 2, or 8 s, respectively. Again, the indifference points

overlapped when the standard delays were either 0.1 s or 2 s but were higher when the

standard delay was 8 s. The performance of subject 1132M as shown in the bottom row

of Figure 20. The indifference points ranged from 2 s to 13 s, 9 s to 26 s, and 16 s to 43 s

when the standard delay was 0.1 s, 2 s, or 8 s in Condition 2. Similarly, the indifference

points ranged from 2 s to 12 s, 2 s to 18 s, and 10 s to 98 s when the standard delay was

0.1 s, 2 s, or 8 s in Condition 3. The upper range of indifference points increased with

increased standard delays in both experimental conditions despite the overlap observed

across standard delays. The performance of this subject showed less across session

variability in Condition 3 than observed in Condition 2.









Figure 22 shows data for subjects 1134M and 1136M. The data for subject

1134M are shown in the top row of Figure 21. Data from two sessions are missing from

Condition 3 for this subject. Note that the symbols are the same for the same standard

delays shown in Figure 21 but that subjects 1134M and 1136M were exposed to a new

standard delay during this final condition of the experiment. These two subjects were

originally presented with 0.1 s, 4 s, or 16 s standard delays and the 2 s standard delay was

added during the final condition. As would be expected from the findings in the previous

two experimental conditions, the mean adjusting delays with a 2 s and 4 s standard were

very similar showing minor differences in the influence of these two delays on the

indifference points for subject 1134M. During Condition 2, the indifference points

ranged from 6 s to 69 s, 6 s to 55 s, and 29 s to 103 s when the standard delay was 0.1 s, 4

s, or 16 s. Interestingly, the indifference points only ranged from 15 s to 50 s across all of

the standard delays (0.1, 2, 4, or 16 s) in the third experimental condition. In other

words, this subject showed less across session variability during Condition 3 than was

observed in Condition 2. The mean adjusting delays shown by subject 1136M were more

variable during the fixed standard delay condition than those observed when the standard

delay alternated daily (Condition2). For instance, the mean adjusting delays ranged from

26 s to 50 s during the final week of the third experimental condition when the standard

delay was 16 s but only ranged from 40 s to 50 s in the second experimental condition.

The indifference point was highest when the standard delay was 16 s and lowest when the

standard delay was 0.1 s across both conditions for subject 1136M. Performance in

Condition 2 appeared to include a decreasing trend in indifference points across standard

delays suggesting that performance was not yet stable at the termination of Condition 2.

















































Figure 21. Mean indifference points across sessions in Conditions 2 and 3. The mean
adjusting delay during last ten trials of a session across sessions for subject's
1126F, 1127F, 1131M, and 1132M are shown in the individual plots. The
left column includes data from the second experimental condition when
a different standard delay was presented daily and the right column includes
data from the third experimental condition when the standard delay was
constant for an entire five-week regimen. Filled symbols are sessions when
the adjusting delay began the session at 6 s and the open symbols are sessions
when the adjusting delay began at 20 s.









Changing Standard


150

120

D 90

60
( 30
c 0
120

< 90

I 60

30

0


0.1 s
12 15-- 2s
-- 2s
--- 8 s
--- 8 s


12 15


Fixed Standard
Condition 3 120

90

60 "

30
(D
0
1202
27F
90

60 c

30

0
3 6 9 12 15


3 6 9
Sessions


120

90 (

60 c_
(D
Co
30 o)
0
120"

90 g

60

30

0
12 15


-- 0.1 s


1127F


120

90

60

30

0
120

90

60

30

0


3 6 9
Sessions















Changing Standard


120
100
80
8 60
V)
40
D 20

128
-) 100
< 80
m 60
S40
20
0


3 6 9 12 15 3 6 9
Session Session


120
100
80
60
40
20

920
100
80
60
40
20
S 0
12 15


-- 0.1 s
0.1 s
-- 4s
4s
S16s
16s
2s
S2s
Figure 23. Mean adjusting delay in Conditions 2 and 3 for subjects 1134M and 1136M.
The mean adjusting delay during last ten trials of a session across sessions
for subjects 1134M and 1136M are plotted. The left column includes data
from the second experimental condition when a different standard delay was
presented daily and the right column includes data from the third
experimental condition when the standard delay was constant for a three
five-week regimens.


In summary, subjects were able to meet the within-session stability criteria

although indifference points varied from session to session with the same standard delay.


Fixed Standard








88
This variability was observed whether the standard delay changed each day (Condition 2)

or remained fixed (Condition 3). The performance of subjects 1131M, 1132M, 1134M,

and 1136M showed slightly less variability when the standard delay remained fixed from

session-to-session.

Figure 23 is a summary plot of the data in Condition 1. Each graph shows the

mean adjusting delay at indifference when the within session stability criteria were met.

Each indifference point was determined as the mean adjusting delay of the within session

mean adjusting delay. The filled symbols indicate the means when the adjusting delay

was 20 s at the beginning of each session while the open symbols indicate sessions when

the adjusting delay was 6 s. A 95% confidence interval is indicated by the error bars but

indicate much less variability than was observed in Condition 1 or Condition 2

performances. Indifference points were not consistently modified by the adjusting delay

at the beginning of the session but instead the indifference points were similar whether

the adjusting delay was 20 s or 6 s at the beginning of the session. The summary plot

demonstrates some across session variability in the indifference point in sessions with the

same standard delay but an increasing indifference point with increased standard delays

to one pellet. Note that the indifference curve includes the standard delay of 2 s for

subjects 1134M and 1136M during Condition 3 only. The mean adjusting delay at

indifference was similar when the standard delay was 2 or 4 s for these subjects.

Discussion

The findings from Experiment 2 demonstrate the use of a fast adjusting-delay procedure

to obtain indifference points and delay discount functions with rats. All subjects

developed stable indifference points within an experimental session even in conditions










when a different standard delay was presented each day of the week. Indifference points

increased with increased standard delays. Beginning a session with a low and high start


1126F


1127F


0 40



S0
S 80


.,
3
"0
0)
<,
t-


- 6 s Start
20 s Start


1134M


1131M
60

40

20

0
80 3
1132M


60
60 c




40 )
20 >
C
80 .)
80

60

40 :


1136M


0.1 2 8
Standard Delay (sec)


0.1 2 4 1
Standard Delay (sec)


Figure 23. Condition 3 mean indifference points. Each graph includes the mean
adjusting delay across sessions with a 95% confidence interval. The open
symbols indicate sessions when the standard delay was 6 s at the start of the
session while filled symbols indicate a 20 s adjusting delay at the start of the
session.









adjusting delay value did not produce consistently different indifference points. The

current findings extend the procedure proposed by Richards et al. (1997) from a rapid

adjusting procedure during which the amount of immediate reinforcement adjusted across

trials to a procedure where the delay to the larger, delayed reinforcer adjusted.

Furthermore, the data from Condition 2 showed that indifference points could be

established using a rapid adjusting delay procedure in sessions lasting only 40 choice

trials or approximately 35 minutes. One limitation of the procedure was the observed

across-session variability of the mean adjusting delay at indifference. This variability

was similar when a different standard delay was presented each day (Conditions 1 and 2)

and when the same standard delay was presented each day (Condition 3) suggesting that

daily changes in the standard delay was not the variable responsible for the observed

variability. This level of variability in individual subject performances would make

detection of the influence of other independent variables on performance difficult. It is

unclear which variable(s) of the current procedure is responsible for producing this

variability. The data reported by Richards et al. (1997) also included variability across

subjects most clearly indicated from the fitted parameter ranges of 0.07 to 0.32 for k (the

discount parameter) and bias parameters ranging from 0.59 to 1.52.

A general pattern demonstrated by within-session changes in the adjusting delay

across trials during Condition 1 was the establishment of stable indifference followed by

unstable performance for the remainder of the session or followed by the development of

a second indifference point. One possible reason for this performance may have been

that subjects became satiated during the course of the session. For instance, subject

1134M showed the development of indifference followed by consistent preference for the









standard alternative for the remainder of the session (Week 2 when the standard delay

was 16 s). Several lines of evidence argue against such an account. First, this subject

was able to develop stable indifference during the other sessions that week. So why

would satiation be selectively observed during only one session? Second, research

examining the influence of deprivation levels on self-control choice have provided

inconsistent affects on choice. Several researchers have reported no effect (Logue,

Chavarro, Rachlin, and Reeder 1988; Logue and Pefia-Correal 1985), some reporting

decreased self-control with increased deprivation (Eisenberger and Masterson 1987;

Eisenberger et al. 1982; Snyderman 1983), and others reporting increased self-control

(mean adjusting delays at indifference) with increased deprivation (Bradshaw and

Szabadi 1992; 1993; Christensen-Szalanski, Goldberg, Anderson and Mitchell 1980; Ho

et al. 1997; Wogar et al. 1992).

One of the potential benefits of the fast adjusting procedure is the rapid

development of indifference points across standard delays. One methodological question

of interest was whether stable indifference points could be obtained in comparable time

with a fast adjusting-delay procedure as compared to that reported by Richards et al.

(1993) with a fast adjusting-amount procedure. The total time of exposure to the current

procedure in Experiment 2 was approximately 55 weeks or 11 months. This is longer

than the time taken to establish indifference that was reported by Richards et al. (1997)

where it took only 15 weeks. Part of the reason for the added time in the current

experiment was to conduct some manipulations that were not included in Experiment 1 of

Richards et al. (1997). The first experimental condition is the most comparable to the

findings from Experiment 1 of Richards et al. (1997). Figures 5 through 11 are data from








92
the 20th week showing the development of indifference within the session for many of the

standard delays. The data shown in these figures suggest that a fast-adjusting delay

procedure can be used to successfully establish indifference points in approximately the

same time required to establish indifference points with an adjusting-amount procedure.

The current experimental findings shared several other similarities with those

reported by Richards et al. (1997). One similarity was that the value of reinforcers

decreased with increased delays to their receipt. The data from Richards et al. (1997)

showed that as the delay to the larger amount of water was increased, the adjusted

amount of immediate water at indifference decreased. The rats preferred the smaller,

more immediate amount of water as the delay to the larger amount was increased. The

current experimental findings showed increasing adjusting delays with increased delays

to the standard alternative. In other words, the rats in the current experiment preferred

the larger, more delayed reinforcer when the delay to the smaller reinforcer was

increased. Another similarity was that both experiments provide support for a hyperbolic

function to describe the influence of reinforcement parameters on self-control choice.

Comparison of the data from Condition 1 and Condition 2 as shown in Figure 19

showed that both the y-intercept and discount parameter, k, decreased. It is unclear why

this decrease was observed. It is possible that the history with the standard delays

modified the influence of other delays in a manner similar to that observed in Experiment

1. This is not supported by a consistently higher or lower indifference point as a function

of the standard delay presented during the preceding session. Another possibility is that

performance was more stable by the end of Condition 2 than it was at the end of

Condition 1. This contention is supported by the higher number of sessions where the








93
within-session stability criteria were met in Condition 2 than in Condition 1 (see Tables

4, 5, 6, and 7).

The within-session changes in the adjusting delay showed substantial fluctuations

in the indifference point within the same session, as evidenced by sessions where more

than one indifference point was obtained within the same experimental session in

Condition 1. Variability was also demonstrated across sessions with the same standard

delay as shown in Figures 21 and 22. It is unclear from previous experimental reports

whether this level of across session variability has also been detected as most of the

reports in the literature include mean indifference curves and/or discount functions

(Mazur 1987; 1988; Ho et al. 1997; Wogar et al. 1993; 1992).















CHAPTER 4
GENERAL DISCUSSION

The current experiments were designed to investigate the choice behavior of rats

when choosing between a large, delayed reinforcer and a small, more immediate one.

Experiment 1 demonstrated that most (5 out of 6) rats exclusively chose three pellets

delayed by 6, 12, or 24 s over one pellet presented immediately. Rats reversed their

preference and chose one pellet presented immediately when the delay to three pellets

was increased to 48 s. Previous research shows that pigeons are likely to make impulsive

choices, i.e., choose the small, more immediate reinforcer under similar conditions (see

review by Logue 1988). One question that remained following the completion of

Experiment 1 was whether the observed differences in self-control choice patterns of rats

and pigeons were the result of a quantitative or qualitative difference in self-control

choice patterns as a function of increased delays. A quantitative difference would be

evident by similar forms in the function relating changes in preference to delay increases.

Different forms of the function would be indicative of a qualitative difference across

species. Experiment 2 was designed to address this question by utilizing an adjusting-

delay procedure. In an adjusting-delay procedure, the delay to the large reinforcer

changed across choice trials as a function of a subject's choice patterns. The delay that

produced equal responding for both alternatives (the small, more immediate

versus the large, delayed option) defined the indifference point. Summary plots of

indifference points as a function of increased delays to one pellet were used to calculate









linear regressions. The linear regressions demonstrated similar functions with positive

slope and y-intercepts across subjects in Experiment 2. These findings are consistent

with findings previously reported with pigeons (Mazur 1987; 1988; Logue et al. 1984),

with rats in a standard adjusting delay procedure (Ho et al. 1993; Mazur et al. 1987;

Richards et al. 1997; Tobin et al. 1993; Wogar et al. 1992; 1993), and humans (Bickel et

al. 1999; Green, Fristoe, and Myerson 1994; Green, Fry, and Myerson 1994; Green et al.

1997; Rachlin et al. 1991; Richards et al. 1999). Thus, the data suggest qualitative

similarity across species.

The indifference points were variable in conditions where the standard delays

changed daily (Experiment 2, Conditions 1 and 2) or when the standard delay was fixed

(Experiment 2, Condition 3). Changing the standard delay daily did not result in a

contrast effect on the indifference points. Comparisons of the data shown in Figures 21

and 22 show no systematic influence of the standard delay from the previous session. A

contrast effect would be a consistently higher or lower indifference point as a function of

the standard delay during the previous session being low (0.1 s) or high (8 s or 16 s).

Session-to-session variability was observed across sessions with the same standard delay

in both Conditions 2 and 3. Subjects 1132M, 1134M, and 1136M were the only subjects

demonstrating less variable performance when only one standard delay was presented

each day of the week suggesting that a daily change in the standard delay may have

increased variability for some subjects. Future research is needed to determine which

variable(s) were responsible for the observed variability.

The data from Experiment 1 suggest that the manner of delay introduction may have

attenuated the effects of delay on self-control choice. When subjects were given a history