|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
1 TIMING BEHAVIOR UNDER PEAK-INTERVAL SCHEDULES: THEORETICAL AND METHODOLOGICAL CONSIDERATIONS By KATHRYN A. SAULSGIVER A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2008
2 2008 Kathryn A. Saulsgiver
3 To Jackson and Megan, my nephe w and niece, for inspiring me.
4 ACKNOWLEDGMENTS I thank m y parents, Bonnie and Richard, for th eir love and support, without which none of this would be possible. I thank Erin McClure, the most amazing lab-mate and friend, for helping in every aspect of this project. I thank C live Wynne, my advisor, for his guidance and instruction. I thank Jesse Dallery for acting as a second advisor and helping direct my future aims in research. I thank Anthony Defulio, Beth any Raiff, Julie Marusi ch, and Jon Pinkston for their support and encouragement and for shaping my research along the way. I thank my various committee members (Marc Branch, Tim Hacke nberg, Jane Brockman, Neil Rowland, David Smith, and Alan Spector), each of whom has infl uenced my behavior as a scientist. I thank Harley, the best dog in the wo rld, for being a source of joy.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ..........7 LIST OF FIGURES.........................................................................................................................8 ABSTRACT...................................................................................................................................10 CHAP TER 1 GENERAL INTRODUCTION.............................................................................................. 12 Timing.....................................................................................................................................12 Theories on Timing............................................................................................................. ....13 Scalar Expectancy Theory............................................................................................... 14 Behavioral Theory of Timing.......................................................................................... 17 Rate Dependency and Timing.........................................................................................20 Stimulus Control and Timing.......................................................................................... 22 Procedural Variations.......................................................................................................... ...23 Level of Analysis............................................................................................................. 25 Session Length/Dosing Delay......................................................................................... 29 Present Experiments............................................................................................................ ...30 2 EXPERIMENT 1....................................................................................................................34 Method....................................................................................................................................36 Subjects............................................................................................................................36 Apparatus.........................................................................................................................37 Procedure.........................................................................................................................37 Training.................................................................................................................... 37 Drug dosing..............................................................................................................38 Data analysis............................................................................................................38 Results.....................................................................................................................................41 Peak Interval Distributions..............................................................................................41 Gaussian Derived Parameters and Esti m ates on Session Average Peak Trials............... 43 Parameters Derived from Single Trials ...........................................................................44 Individual Subjects vs. Group Average Data.................................................................. 45 Ziggurat Analysis............................................................................................................ 45 Rate Dependency.............................................................................................................47 Discussion...............................................................................................................................48
6 3 EXPERIMENT 2....................................................................................................................66 Method....................................................................................................................................67 Subjects............................................................................................................................67 Apparatus.........................................................................................................................67 Procedure.........................................................................................................................67 Training.................................................................................................................... 67 Drug administration.................................................................................................. 69 Data analysis............................................................................................................70 Results.....................................................................................................................................71 Peak Interval Distributions..............................................................................................71 Gaussian Derived Parameters and Esti m ates on Session Average Peak Trials............... 73 Additional Parameters..................................................................................................... 76 Rate Dependency.............................................................................................................80 Discussion...............................................................................................................................81 4 GENERAL DISCUSSION..................................................................................................... 94 Timing Theories................................................................................................................ ......95 Procedural Variations.......................................................................................................... .102 Summary...............................................................................................................................107 LIST OF REFERENCES.............................................................................................................109 BIOGRAPHICAL SKETCH.......................................................................................................115
7 LIST OF TABLES Table page 2-1 Post-hoc Scheffe results by bin..........................................................................................54 2-2 Comparison of drug effect on parameters derived from a session average and a individual trial analysis for the cu rrent and two previous experiments............................. 55 3-1 Delay significant main effects........................................................................................... 84 3-2 Dose significant main effects............................................................................................. 85 3-3 Significant interaction between delay and dose ................................................................. 86
8 LIST OF FIGURES Figure page 1-1 Two simulations of possible effects of am pheta mine on rates of responding in a peak trial on PI 15.......................................................................................................................32 1-2 PI distributions adapted from previous publications......................................................... 33 2-1 Example individual trial da ta across various doses of d-am phetamine.............................56 2-2 The condition average response rates per 3s bin during peak trials for each subject under each d ose of d-amphetamine.................................................................................... 57 2-3 Response rates as a percent of saline are shown for each 18-s bin.................................... 58 2-4 Peak Time and standard deviat ion shown as a percent of saline ....................................... 59 2-5 Start time, stop time, middle, run lengt h, run rate, and num ber of responses between start and stop times shown as a percent of saline............................................................... 60 2-6 Average drug effect on derived parameters shown as a percent of saline......................... 61 2-7 Example individual trial da ta across various doses of d-am phetamine.............................62 2-8 Number of individual trials best accounted for by a oneand two-step function across all doses for all subjects ..................................................................................................... 63 2-9 Log session average response rates unde r drug conditions as a function of log response rates averaged acr oss all baselin e sessions .........................................................64 2-10 Slope and intercept derived from the scatter plots showing log session average response rates under drug conditions as a function of log response rates averaged across all base line sessions ................................................................................................65 3-1 Average response rates per 3-s bin during peak trials for each subject under each dose of d-amphetamine across each delay......................................................................... 87 3-2 Average response rates per 3-s bin during p eak trials for the average of all subjects under each d ose of d-amphetamine across each delay....................................................... 88 3-3 Percent saline changes for the average of all subjects across each delay for each param eter...................................................................................................................... ......89 3-4 Peak Time, standard deviation, and wait tim e shown as a percent of saline..................... 90 3-5 Number of responses made under the RR schedule, PI schedule, total num ber of switches made between schedules, and the duration spent on the RR and PI schedules are shown as a percent of saline........................................................................ 91
9 3-6 Log session average response rates unde r drug conditions as a function of log response rates averaged across all b aselin e sessions for the entire trial for each subject under each delay.................................................................................................... 92 3-7 Slope and intercept derived from the scatter plots showing log session average response rates under drug conditions as a function of log response rates averaged across all base line sessions ................................................................................................93
10 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy TIMING BEHAVIOR UNDER PEAK-INTERVAL SCHEDULES: THEORETICAL AND METHODOLOGICAL CONSIDERATIONS By Kathryn A Saulsgiver December 2008 Chair: Clive D. L. Wynne Major: Psychology Pigeons were exposed to peak interval (PI) schedules. These schedules consist of fixed intervals (FI) and longer peak or extinction interv als that allow for the examination of temporal regulation in behavior. Several th eories of timing postulate that an internal pacemaker controls timing. One theory, Scalar Expectancy Theory (SET), predicts that dopamine agonists directly alter the internal pacemaker by speeding it up and thus lead to an underestimation of time, or subjects to respond as if time is passing faster than it is. This unde restimation in time is represented by a decrease in peak time, a measure of temporal accuracy, under PI schedules. SET also predicts that variance in temporal accuracy should not increase but rather decrease proportionally under these conditions. In Experi ment 1, pigeons were exposed to a PI 45-s schedule where FI trials were 45 s long and PI trials lasted 135 s. Subjects were administered damphetamine and resulting changes in behavior were examined. Changes in measures of temporal control under drug did not support SE T. Peak time did decrease under drug as predicted by SET but the variance of temporal control increased as well, which violates implications of scalar timing. Rate dependency and other accounts indicating a loss of stimulus control better characterized and explained the data under d-amphetamine. The form or structure of individual trials under drug was al so assessed. It was found that a peak-like form exists at the
11 individual trial level but this impression breaks down with increas ing drug dose. In Experiment 2, pigeons were exposed to a conc urrent PI 30-s random ratio (RR) schedule. The effects of damphetamine and pre-session feeding were examined across several delays (0-, 30-, 60-, and 90min) between administration and behavioral testi ng. Results indicate that as behavior under a concurrent task breaks down, temporal accuracy under a PI schedule also suffers. Results also show that for individual subjects, delay between behavioral testing a nd administration of drug can impact obtained results. Over all, these results indicate that SET does not accurately predict changes in timing behavior under PI schedules. Rather results point towards an explanation that implicates the loss of stimulus control as the main source of disruption in timing behavior.
12 CHAPTER 1 GENERAL INTRODUCTION Timing Tim e is of the essence, but what is the essence of time? --Karan Varsheni As humans we use the passage of time ever y day to guide our behavior and schedules. Whether it is working for certain durations, sche duling meetings at certain times, or waiting for time to pass to engage in some activity, time plays an integral part in our behavior. Despite this ubiquity, researchers have only re cently started to understand the dynamics of timing behavior. The ability to accurately discriminate or produce a relatively short interval of time is known as interval timing. The ability to discriminate time at short intervals or to behave with respect to time has been implicated in many other behavioral processes both in humans and in non-humans. Whether timing is implicated in choice behavior, in Pavlovian processes, in contributing to the efficacy of consequences, or involved in other beha viors such as foraging it is a feature of the world that deserves more recognition. Isolating time as a variable in the natural en vironment has been difficult but not impossible (Boisvert and Sherry, 2006; Henders on et al., 2006). In order to be tter investigate this variable researchers have turned to the laboratory to assess temporal c ontrol of behavior. Through this research investigators have tr ied to understand the essence of timing behavior and how it is controlled. Many different methods exist for asse ssing behavior under control of the passage of time in the laboratory. One method used to investigate drug and environmental effects on behavior controlled by time is a Peak Interval (PI) procedure. Th is procedure is trained by first exposing subjects to a discrete tr ial Fixed Interval (FI) schedule. After some exposure to this schedule, PIs are intermixed with FI trials during the same sessi on. PI trials are longer than scheduled FI trials and end without food presentati on. These trials are indistinguishable from the
13 accompanying FI trials except that they last longer. These longer extinction trials allow for an examination of responding in all temporal periods surrounding the time of scheduled reinforcement on FI trials (Catan ia, 1970). The ratio of FI to PI trials used varies across experiments but PI trials never outnumber FI tr ials (1(FI):1(PI) ratio: Cheng et al., 2007; MacDonald and Meck, 2005; Maricq et al., 1981 ; Matell et al., 2004, 2006; Taylor et al., 2007; 3:2 ratio: Bayley et al., 1998; Eckerman et al., 1987; Frederick and Allen, 1996; 4:1 ratio: Saulsgiver et al., 2006; 7:1 ratio: Kraemer et al., 1997). The experiments included here employ PI procedures in order to understand timing behavior and how it is controlled. Theories on Timing Several theories of how the passage of tim e comes to control behavior have been proposed over the last few decades. These theories are aimed at identifying the variables that control timing behavior. Many of these theories empha size hidden variables, su ch as an internal pacemaker, as mechanisms of action (Gibbon an d Church, 1984; Killeen and Fetterman, 1988). Two theories, Scalar expectan cy theory (SET: Gibbon and Chur ch, 1984) and the Behavioral theory of timing (BET: Killeen and Fetterman, 1988), have been examined in great detail and have contributed to the understanding of this be havior. SET proposes that an internal pacemaker is the mechanism that controls timing while BET suggests that mediating behavior comes to aid in timing through its correlation with an internal pacemaker. Corroboration for these theories has been shown in laboratory settings (SET: Gibbon, 1977; 1992; Gibbon et al., 1997; Meck, 1983; 1996; Maricq et al., 1981; BET: Fetterman et al., 1998; Killeen et al., 1999). These theories will be discussed more fully. SET proposes that the workings of the internal pacemaker are altered by drugs that modify dopamine levels (Meck, 1996). Thus much of the research using drugs to investigate timing has used SET as an explanatory framework. Several st udies of this type have presented data that
14 support the notion that an internal pacemaker controls tem poral regulation (SET: Gibbon, 1977; 1992; Gibbon et al., 1997; Kraemer et al., 1997 ; Meck, 1983; 1996; Maricq et al., 1981). Alternatively, other research has supported the theory that amphe tamine and similar drugs cause a general disruption in behavior or loss of stimulus control (Bay ley et al., 1998; McClure et al., 2005; Odum et al., 2002; Saulsgiver et al., 2006). BET on the othe r hand has not typically been investigated with the use of drugs. This theory holds that varying rates of arousal and reinforcement will alter timing behavior and thus research investigating BET has manipulated these variables instead (Ludvig et al., 2007; MacEwen and Killeen, 1991). The following experiments arrange variables with the aim of evaluating each of these explanations of timing and how environmental manipulations influence it. Scalar Expectancy Theory Scalar Expectancy Theo ry (SET) has dominate d research on timing for the last 30 years. This theory, developed by Gibbon and Church (1978, 1984), proposes that timing behavior is primarily controlled by a pacemaker that operate s in a similar manner to a ticking clock. SET proposes that the passage of time under an FI schedule comes to control behavior in the following way: pulses are released from a pacemaker at some regular interval and cumulate in a construct called an integrator. As time passes more pulses accumulate here. At some point the subject will make a response. If the response was premature and th e required interval of time has not passed, the pacemaker will continue to tick and pulses will continue to be collected in the integrator. If enough time has passed to satisfy the FI schedule, reinforcement will occur and the number of pulses present in the integrator will be transferred into reference memory for later use. On subsequent trials the process of the p acemaker ticking and pulses accumulating in the integrator will repeat. As these new pulses accu mulate the number of pulses present will be compared to the number of pulses stored in refere nce memory. If the ratio between pulses in the
15 integrator and pulses in reference memory meets some criterion according to a relative proximity rule a response will be made, if the ratio does not meet the requirement no response will occur. If the FI schedule has elapsed reinforcement w ill occur and this new value of pulses in the integrator will be transferred to reference memory. According to SET, the pacemaker is the primary controlling mechanism behind timing behavior (Gibbon and Church, 1984). The pacemaker is continuously running and emitting pulses. When there is a stimulus to be timed, a gate or switch is activated and pulses accrue into an accumulator (Matell and Meck, 2000). The rate of the pacemaker under SET is not affected by the rate of reinforcement as other theories predict (Killeen and Fette rman, 1988). Pulses are emitted at some constant rate, as long as there is a stimulus to be time d, and collected for later comparison. If the rate of reinforcement drops to zero the number of pulses emitted by the pacemaker will not change. Environmental manipul ations that directly alter the rate of the ticking pacemaker will subsequently cause changes in timing behavior. One primary focus of SET is how dopaminergic drugs affect the p acemaker (Meck, 1996). It is hypothesized that dopamine agonists, such as amphetamine, will speed up the rate of pulse emissions from the pacemaker causing an overestimation in time. Pulses accumulate faster in the integrator causing the ratio of pulses in the integrator to the pulse s in the reference memory to meet some critical value causing subjects to respond as if the interval had elapsed. In other words amphetamine would lead a subject to respond as if more time has passed than actually has. Dopamine antagonists, such as haloperidol, will slow dow n the pacemaker resulting in an underestimation of time, or subjects to respond as if time is pa ssing slower than it is. Under these conditions the pacemaker would emit pulses at a slower rate. Th e ratio between pulses in the integrator and the reference memory would take longer to reach the critical value thus delaying responses. One
16 focus of the present research was to investigate whether SET made accurate predictions about how timing behavior would change under d-amphetamine and to compare these predictions to other theories. As mentioned above, according to SET, manipulations that directly alte r the pacemaker will result in altered timing behavior. Researchers have hypothesized that the discrimination or perception of time involves dopamine-releasing neurons (Meck, 1996). Pharmacological studies have supported this belief by showing that drugs that interact with the dopamine system alter timing accuracy in a predictable manner (Mar icq and Church, 1983). Specifically, dopamine agonists like amphetamine increase the release of dopamine which speeds up the rate of the internal pacemaker. This increase in rate re sults in subjects responding as if time is passing faster (Maricq and Chur ch, 1983; Meck, 1986). SET also predicts that the variance or error in accuracy will be proporti onal to the mean of the interval to be timed (Scalar timi ng: Gibbon, 1977; 1992; Gibbon et al., 1997). This requirement is Webers law in the temporal domai n and is illustrated in Figure 1-1 (Figure 1: Saulsgiver et al., 2006). The existence of We bers law in time perception forms the foundation of SET (Bizo et al., 2006). This figure shows h ypothetical data and the re sulting changes in the PI distribution predicted by SET and rate dependency (discussed later). SET states that when a subject is administered amphetamine the pacemaker will speed up resulting in a leftward shift in the entire distribution (dotted line) (Buhusi and Meck, 2002; Cheng et al., 2007; Matell et al., 2006). As variance in timing is to remain a consta nt proportion of the interval to be timed this distribution should not widen, it sh ould decrease proportionally with the mean interval to be timed (Buhusi and Meck, 2002; Gibbon, 1977; 1992; Gibbon et al., 1997). Specifically, if a dopamine agonist speeds up the pacemaker, pul ses will accumulate faster and the relative
17 proximity rule to initiate re sponding will be satisfied sooner. Similarly, the criterion for terminating responding will also be encountered ear lier thus maintaining scalar timing (Buhusi and Meck, 2002; Matell et al., 2006). Published research measuring both the mean (o r Peak Time, Catania, 1970; Maricq et al., 1981) and variance (or width, standa rd deviation [s.d.], Matell et al ., 2006) of temporal accuracy has failed to entirely support th is theory. Research using PI procedures and dopamine agonists has shown decreases in peak time (Buhusi and M eck, 2002; Cheng et al., 2007; Eckerman et al., 1987; Frederick and Allen, 1996; Hinton and Meck, 1996; Maricq et al., 1981; Matell et al., 2006; Meck, 1996; Meck and Williams, 1997; Taylor et al., 2007). Many of these same studies however, also contradict predictions made by SET by also showing an increase in the width or s.d. of these PI distributions (Bayley et al., 1998; Cheng et al ., 2007; Eckerman et al., 1987; Kraemer et al., 1997; Matell et al ., 2006; Saulsgiver et al., 2006; Taylor et al., 2007 ). Figure 1-2 shows that in many cases the PI distributions wi den under drug whether or not there is a clear decrease in peak time (Bayley et al., 1998; Eckerman et al., 1987; Maricq et al., 1981; Meck, 1996). Additional research that examined the width of the distribution directly supports this impression (Meck, 1996; Saulsgiver et al., 2006; Taylor et al., 2007). This widening of the distribution, in the presence or abse nce of a leftward shift, is inc onsistent with SETs account of timing since it would contradict scalar timing (Figur e 1-1). These predictions will be explored in Experiment 1 and 2. Behavioral Theory of Timing Killeen and Fetterm ans (1988) Behavioral Th eory of Timing (BET) includes a pacemaker as well as a behavioral component The behavioral component suggests that mediating behaviors are elicited by signals of reinforcement and th at these mediating behaviors may assist in discriminating the passage of tim e. Mediating behaviors do not necessarily take place because
18 they are influential in aiding timing behavior, rather the occurrence of these behaviors may become useful in accurately discriminating the passage of time. Under an FI schedule subjects might engage in these mediating behaviors a nd after some experience come to associate a particular behavior in a chain of mediating behavior with the passage of a certain amount of time. If this chain of behavior is followed by responding to sa tisfy the FI schedule, and this responding is reinforced, the occu rrence of a particular behavior in a chain could serve as a discriminative stimulus that indicates the FI in terval has expired. These behaviors, then, may come to mediate the passage of time and serve as cues indicating the passage of time. These behaviors may allow for conditional discriminati ons of time passing to be made (Killeen and Fetterman, 1988). The pacemaker in BET varies from that in SET. In this theory the rate of the clock or pacemaker is not constant but rather varies with the rate of reinforcement or the arousal of the subject (Killeen et al., 1999). When the rate of reinforcement increases or decreases the rate of the pacemaker also increases or decreases respectiv ely. The higher the rate of reinforcement, the more behavioral states are thought to exist. The more behavi oral states that the animal experiences the more accurate timing behavior w ill be (Killeen et al., 1999). After some exposure to a timing schedule, BET proposes that subjects will come to asso ciate some behavioral state with the passage of time. The pacemaker in this case emits pulses that drive the organism through these behavioral states. The more aroused a s ubject is the faster the pacemaker will emit pulses and subsequently the faster the organism will go through the behavioral states associated with that particular number of pulses. If the ra te of reinforcement drops to zero as it does with extinction, the rate of the pacem aker will drop and timing accuracy will suffer. Under SET the
19 rate of the pacemaker does not vary with the rate of reinforcement (Killeen and Fetterman, 1988). BET holds that if the rate of reinforcem ent is altered, whethe r through the animals exposure to drugs, extinction, pr e-session feeding or other en vironmental changes, timing behavior will also be affected. While the authors of BET have not expanded their predictions to cover drug effects per se, predictions of possible drug effects on timing behavior can be made by considering the mechanisms of ac tion in this theory along with research on how drugs affect the value of reinforcement. Research has shown th at when administered amphetamine animals will complete fewer fixed ratios to obtain food (Wh ite et al., 2007), engage in less food taking behavior (Foltin, 2001), and c onsume less food overall (Foltin and Haney, 2007). Amphetamine has also been shown to increase responding on a continuous reinforcement schedule while decreasing break points on a progres sive ratio schedule (Caul and Brindle, 2001). Increases in break points under a progressive ratio schedule under low doses and decreases in break points under high doses of amphetamine have also been observed (Mayorga et al., 2000). Animals in some of these studies, when administered am phetamine, were less likely to respond for food indicating that food in th ese instances was not as valuable a reinforcer as it was under non-drug conditions. If amphetamine decreas es the value of reinforcement in some cases then one might assume that it would decrease the value of reinfo rcement in timing experiments. According to BET, timing accuracy will decrease when arousal or the rate of reinforcement decreases. Thus one can assume that amphetamine administration w ould lead to a decrease in accuracy in timing behavior. Killeen and Fetterman (1988) hypothesized that changes in timing behavior should be accompanied by changes in mediating behavior. The problem up to now is that researchers may
20 not have access to all of the mediating behavior or that their measurement of this behavior may not be adequate to observe changes. The identificat ion of each behavioral state could be difficult. These states could be similar in form and vary ju st by rate or location, th ey could be the absence of behavior, or could even be ex tremely subtle movements. Using a temporal discrimination task, Fetterman and colleagues (1998) were able to id entify and measure possible collateral behavior in rats and pigeons that may have been useful in di scriminating the passage of time. They found that collateral behaviors, when identified, bette r predicted temporal judgments. This sort of analysis has yet to be done using a PI procedure. One reason for this is that the identification of the collateral behavior is not always possible. One way around this problem may be to use a PI procedure with a concurrent task. Sanabria and colleagues (2005) showed that behavior on a PI schedule was better controlled by the interval in place when a concurrent task was available. The width or s.d. of the PI distributions was smaller for subjects where a concurrent task was used. Experiment 2 of this manuscript exposes subjects to a concurrent PI random ratio (RR) schedule in an attempt to replicate this finding and to examine whether changes in a collateral task predict changes in a timing task. Adding an additional task to complete, that does not require interval timing, allows for measur able mediating behavior that may assist in timing under the PI schedule. If me diating behavior assists in timing intervals then changes observed in responding on th e concurrent task should pr edict changes seen in timing behavior under PI schedules. If changes in medi ating behavior are inconsistent with changes in timing behavior BET would not be supported. Rate Dependency and Timing The rate dependency hypothesis (Dews, 1958, 1970) states that changes observed in responding under drug conditions depend on the rate ob tained under baseline conditions. The direction and magnitude of change s in responding is predictable based on the rate of responding
21 under control conditions. When em ploying psychomotor stimulant drugs the result is that low response rates increase and high re sponse decrease or change to a lesser degree (Dews, 1958). Researchers investigating timing be havior with drugs have reporte d data consistent with ratedependent findings resulting in an increase in lo w response rates across the interval (Beecher and Jackson, 1976; Byrd, 1979; Dews, 1958; Flores and Pellon, 1995; Goudie, 1985; Knealing and Schaal, 2002; McMillan and Healey, 1976; Odum et al., 2002; Saulsgiver et al., 2006; Wuttke, 1970). When logged response rates obtained under drug conditions are plotted as a function of logged control response rates, the re sulting distribution can be fit with a straight line of which the slope and intercept can be assessed for cha nges. Increases in intercept accompanied by decreases in slope indicate that low response ra tes under control sessions have increased with drug while high response rates have decreased or been left unchange d. This pattern of change is evident in previously published studies using th e PI procedure as well (s ee Saulsgiver et al., 2006). While order is apparent in these data as well as in other published reports, these characterizations do not offer an explanation for the controllin g mechanisms and do not account for the changes observed at the level of the individual trial (Branch and Gollub, 1974; Knealing and Schaal, 2002). Rate dependency may lack a mechanism of ac tion but it has been shown to account for changes seen in timing behavior better than SE T (Knealing and Schaal, 2002; Odum et al., 2002; Saulsgiver et al., 2006). Figure 1-1 shows rate de pendent characterizations of a stimulants effect on responding under a PI distribution (dashed line). In this demonstration response rates have been altered in a rate-dependent manner, low rates of responding increase under drug and high rates of responding are less altered. This prediction is suffi cient enough to account for leftward shifts and increases in the width of the distribution. This effect can be se en in previously published PI
22 distributions (Figure 1-2). In each case the administration of drug resulted in low response rates resulting in an increase in width of the PI. Experiments 1 and 2 will examine these rate dependent predictions. Stimulus Control and Timing The observed increase in width of the PI distributions in m any experiments and the interpretation that the width of the interval is a reflection of the control that the time interval has on behavior, has led some researchers to conclu de that the main effect of stimulants on temporally controlled behavior is to disrupt stimulus cont rol (Knealing and Schaal, 2002; McClure et al., 2005; Odum et al ., 2002; Saulsgiver et al., 2006) Whether this disruption is specific to temporal aspects, control exerted on behavior due to the interval length, or to control exerted by all schedule aspects, such as reinforcer value, attention, motori c activity, and temporal components, needs to be explored. Is it that ch anges in temporal accura cy are secondary to the overall disruption of behavior rather than demonstrating a se lective effect on timing? Previous results indicate that amphetamine ma y alter overall stimulus control at least under high doses of the drug (Katz, 1988; McClure et al., in preparatio n; Saulsgiver et al., 2006). Using a PI procedure Saulsgiv er et al. (2006) found that d -amphetamine disrupted measures that signify stimulus control, while measures of timing were not affected to the same degree. In this experiment increases in the width of the distri bution were more apparent than changes in the location of peak time. Using a temporal bisec tion task McClure et al (in preparation) found similar results. The point of subjective equality was not consistently altered by amphetamine in this experiment but stimulus control measures re liably changed in a direction indicating loss of overall stimulus control. Ka tz (1988) found that under higher dos es of amphetamine, stimulus control was diminished as responding decreased under a multiple FI schedule. Responding on a red key in this experiment was reinforced when a houselight was illuminated and responding on
23 an amber key was reinforced when the houselight was terminated. When high doses of amphetamine were administered, subjects showed a decrease in accuracy on this conditional discrimination task. These effects on stimul us control caused by drug could be reduced by modifying aspects of the schedul e that increase stimul us control. Laties (1972, 1975) showed that adding an external cue to timing experi ments also decreased the overall effect of amphetamine on stimulus control. He hypothesized that timing in the absence of an additional cue that indicated the passage of time resulted in less control of th at time interval over behavior. These results taken together could indicate that when the stimulus is not salient, as the passage of time with no external cue may be, the main effect of amphetamine is to disrupt overall control. Whether or not loss of stimulus control exerted by all aspects of the schedule is enough to explain disruptions in temporal control unde r drug will be examined in the following experiments. Procedural Variations There has been considerable variability in re sults obtained while inve stiga ting the effects of dopamine agonists on timing behavior. Many of these discrepancies can be traced back to variations in the procedure used. McClure et al. (in pr eparation) have shown that the nature of the choice alternatives used in a matching-to-sample of durati on (MTSD) procedure can greatly impact the observed changes in the psychophysical function under d-amphetamine. McClure et al. (in preparation) exposed groups of subjects to different MTSD tasks and tested the effect of damphetamine on this behavior. One group was expos ed to a color-matching task. In this task subjects were first trained to respond to one co lor alternative (red) afte r a 2-sec duration and to another color alternativ e (green) following an 8sec duration. These choi ce alternatives could alternate between two choice keys. Following tr aining, subjects were exposed to intervening time intervals intermixed with 2and 8-sec dur ations. Choices following these intermediate
24 values were never reinforced. A second group was exposed to a side-tracking task. This task was identical to the color-matching task except that the choice alternatives were the location of the key rather than the color. D-ampheta mine was then administered. Following drug administration the psychophysical functions for subjects exposed to the co lor-matching task were flattened, indicating a loss in stimulus control under drug. Accu racy for these subjects at the two training durations was greatly decreased by d-amphetamine. This was, however, not the case for the side-tracking group. Psychophysical functions obtained under drug for this group showed less disruption in accuracy for the trained durations, rather a shift in the PSE (point of subjective equality) was observed. Results of this study indicate that these slight methodological differences can lead to very di fferent results. In 2005 McClure et al. reviewed the literature using both procedures and found that these two methodologies c onsistently led to different results. Moreover they also found that only data utilizing a side-tracking task had been used to support SET. There are many procedural differences used in implementing PI schedules as well. These differences in procedure include session length (Bayley et al., 19981 hour; Frederick and Allen, 1996; Matell et al., 2006; Saulsgiv er et al., 20062 hours; Penney et al., 19962 hours 50 min; Maricq et al., 19814 hours; Others use a required number of tr ials with no concrete session length), training history (see Cheng et al., 2007 for review; <70 sessions: Abner et al., 2001; Bayley et al., 1998; Hinton and Meck, 1996; Ma cDonald and Meck, 2005; Maricq et al., 1981; Matell et al., 2004, 2006; Penney et al., 1996; > 70 sessions: Eckerman et al., 1987; Odum et al., 2002; Santi et al., 1995; Saulsgiv er et al., 2006), delay betw een drug dosing and behavioral testing (20 and 100 min after methamphetamine ad ministration: Cevik, 2003), and ratio of FI trials to PI trials (1:1 ratio : Cheng et al., 2007; MacDonald and Meck, 2005; Maricq et al., 1981;
25 Matell et al., 2004, 2006; Taylor et al., 2007; 3: 2 ratio: Bayley et al., 1998; Eckerman et al., 1987; Frederick and Allen, 1996; 4:1 ratio: Saulsg iver et al., 2006; 7:1 ratio: Kraemer et al., 1997). There are also differences in how th e data are assessed. Some researchers assess behavioral change at the level of the session average (Bayley et al., 1998; Eckerman et al., 1987; Frederick and Allen, 1996; Maricq et al., 1981; Meck, 1996; Saulsgiver et al., 2006) while others examine individual trials (Che ng and Westwood, 1993; Matell et al., 2004, 2006; Taylor et al., 2007). Whether data are assessed for individual subjects (Saulsgiver et al., 2006) or on the average of multiple subjects (Bay ley et al., 1998; Eckerman et al., 1987; Frederick and Allen, 1996; Maricq et al., 1981; Meck, 1996) also va ries. These slight differences in methodology used with the PI procedure could contribute to the observed inconsistencies in the literature. A few of these differences will be explored in Experiment 1 (data assessment techniques/level of analysis) and Experiment 2 (session length). Level of Analysis Many prior studies of the PI procedure have relied on the aver aging together of responding in all peak trials and then f itting a function to this distributi on (Bayley et al., 1998; Cheng and Roberts, 1991; Cheng et al., 2007; Eckerman et al., 1987; Frederick and Allen, 1996; Kraemer et al., 1997; Maricq et al., 1981; M eck, 1983, 1996; Penney et al., 1996; Saulsgiver et al., 2006). This type of analysis is sometimes conducte d on the average response rates across multiple sessions and even subjects (Bayley et al., 1998; Cheng et al., 2007; Kraemer et al. 1997; Matell and Meck, 1999; Matell and Portugal, 2007). More recently a break-and-run analysis devised by Cheng and Westwood (1993) has been implemented. Th is analysis is applied to single trials and assumes that on each trial responding starts off low or near zero, increases abruptly to a higher rate and then subsequently decreases to the near zero rate. For this analysis single trials are fit with an algorithm that identifies th e start of the run (start time: 1st of 2 consecutive bins in which
26 responding occurred), the point at which responding decreased again (stop time: 1st of 2 consecutive bins following start time in which no responding occurred ), and the number of responses between start and stop time. Two add itional parameters can also be derived. Run length, or the spread of responding, is calculated by subtracting the start time from the stop time on each trial. The middle parameter, taken as a measure of temporal accuracy similar to peak time (Taylor et al., 2007), is defined as the mi dpoint between start and stop time. Similar analyses, differing slightly in the number of bins required to identify st art and stop time, have been applied to PI schedules, and it was found that there was order in meas ures identif ied by this method (Grace et al., 2006; Matell et al., 2006; Taylor et al., 2007) Grace and colleagues (2006) found that start time, stop time, middle, and run length, were all longer fo r the longer of two PI schedules, showing control by the temporal contingency. Other st udies using the PI procedure have employed this analysis when assessing the e ffects of amphetamine, but results have varied and researchers have failed to s how individual data (Matell et al ., 2006; Taylor et al., 2007) or to test a wide range of doses (Tay lor et al., 2007). These studies ha ve shown that results at the single trial level have been inc onsistent with changes seen at the session average level and conclusions derived from each analysis failed to be consistent (Matell et al., 2006; Taylor et al., 2007). For example, Matell and colleagues showed a decrease in peak time when using the session-average analysis but di d not show the same decrease in the middle parameter. Experiment 1 here employs both the session average and single trial analyses in order to compare these methods within subjects and across a wide vari ety of doses. If there is consistency between both analyses within subjects then one might argue that both methods are adequate for assessing changes in behavior. If there is a lack of congruity across methods of assessment then a decision on which method is best should be made.
27 The level at which changes in behavior are assessed can impact the conclusions drawn from data. Important theories have been limited based on results from a more detailed analysis. For example, Branch and Gollub (1974) assessed the effects of d-amphetamine on responding under individual trials of a discrete-trial Fixed-In terval (FI) procedure. Their analysis differed from prior studies examining rate-dependent e ffects which had assessed changes in behavior over multiple trials averaged together. When examined at the individual trial level, the data showed response rates obtained by averaging FI trials together were not representative of interval-by-interval performance in portions of e ach trial. Branch and Gollub (1974) showed that while rate dependency can account for the effect of drugs on the average of multiple FI trials it did not accurately describe the e ffects of drug on individual trials because on any typical FI trial subjects responded in some time bins but not in ot hers. Individual trials we re better described as consisting of states of responding and no respondi ng, rather than having the gradual increase in response rates observed when multiple trials are averaged together. Rate-dependency failed to account for data from individual trials, thus limiting its scope a nd usefulness as an explanation for drug effects on FI schedules (Branch, 1984). Branch and Gollub (1974) showed that while there is an increase in responding towards the end of a FI interval, there are still many points in the interval after resp onding has started where no responding occurs when responding is examined on a fine scale, a pattern that is typically described as a two-state or break-and-run pattern (Branch an d Gollub, 1974; Grace et al., 2006; Schneider, 1969). A similar pattern of behavior for individual trials has been reported for behavior under PI schedules (Cheng and Westwood, 1993). Individual trials need to be assessed in order to determine if each trial does contain a two-state pattern of responding as assumed by the single-trial analysis, or if th ere is no indication of a peak-like fo rm under individual trials. If
28 behavior is temporally organized at the level of individual trials then responding in a peak trial should gradually increases from a low rate of responding to some higher rate of responding at about the time of reinforcement then decrease as the interval elapses. If behavior is not temporally organized under individual trials then rates should abruptly increase from near zero response rate and then decreases again. To date an assessment of the shape of each individual peak trial has not been conducted. Rather researchers have excl uded trials where the assumptions of these analyses are not met (Cheng and Westwood, 1993; Matell et al., 2006; Taylor et al., 2007). In order to identify the structure of i ndividual trials an assessment was created that fit two functions base d on different sets of assumptions to each individual trial. The one-step function assumes behavior lacks temporal organization in each peak trial. Responding starts off at a rate of zero, increases and maintain s some rate for a period of time, then decreases again to a rate of zero. In this function ther e are not multiple rates of responding, rather the distribution is flat. The two-st ep function assumes that a two-state pattern of responding occurs in each peak trial, resulting in a peak-like dist ribution. This analysis is referred to as the Ziggurat analysis (Ziggu rat: A temple tower of the ancien t Assyrians and Babylonians, having the form of a terraced pyramid of successively r eceding stories). This analysis is aimed at identifying whether a peak-like pattern does exist at the individual trial level and whether this impression is maintained across doses of amphetamine. There is a great deal of individual subject variability in studies examining drug effects on timing behavior (Cheng et al., 2007 ; Saulsgiver et al., 2006). Ye t researchers opt to either conduct analyses on the average of multiple subjects, or to averag e the assessments of individual subjects together rather than presenting changes for individuals separately (Bayley et al., 1998; Cheng and Roberts, 1991; Cheng et al., 2007; Ec kerman et al., 1987; Frederick and Allen, 1996;
29 Kraemer et al., 1997; Maricq et al., 1981; Matell et al., 2006; Meck, 1983, 1996; Penney et al., 1996; Taylor et al., 2007). This practice would be justified if the average were representative of each subject used, however in a study where individual subjects have been reported this does not appear to be the case (Saulsgiver et al., 2006). E xperiment 1 will present data for both individual subjects and group averages in or der to determine if i ndividual subject variability and differences is captured in a group analysis. Session Length/Dosing Delay Recent research has sho wn that the time from dand methamphetamine administration to behavioral testing can impact the observed behavioral changes (C evik, 2003; Pinkston, 2006, 2008). The time between methamphetamine admini stration and testing wa s manipulated across a temporal discrimination experiment with rats by Cevik. This experiment examined changes in temporal discrimination on a side-tracking MTSD procedure while administering 0.5 mg/kg of methamphetamine 20 minutes prior to an 180-mi nute testing session. Data were assessed separately for behavior during the first 80 minutes (Experiment 1 20-100 min) and for the last 80 min (Experiment 2 100-180 min). Results from Experiment 1 showed a flattening of the psychophysical function typical of those seen in temporal discri mination procedures that employ a color-matching procedure (see McClure et al 2005 for review). When an analysis was performed on the second half of the experime ntal session, a clear diffe rence in the degree of flattening was observed. When methamphetami ne was administered 100 minutes prior to experimental testing, results showed no flattening of the psyc hophysical function but did show a rightward shift in the function (overestimation of time)-results typical of a side-tracking procedure (see McClure et al, 2005 fo r review). When testing occurred closer in time to drug administration, the main effect observed was a disr uption of overall stimulus control. However, when testing occurred farther in time from admi nistration, it appears that the main effect of
30 amphetamine was to disrupt temporal discriminati on. These results indicate that time between methamphetamine administration and behavioral te sting is an important variable to consider when investigating temp oral discrimination. Pinkston (2006) showed preliminary data on locomotor activity in rats when administered a 5.0 mg/kg dose of d-amphetamine. In this experiment lo comotor activity increased to a steady rate throughout the first hour. Du ring this time all responding for food was eliminated. Early in the second hour of testing the increased locomotor activity star ted to wane and responding for food began to increase. These preliminary results with d-amphetamine indicate that interference due to increases in locomotor activity during behavioral testing, when te sting occurs close in time to drug administration, may alter the resu lts and interpretations of experimentation. The impact of a drug on behavior can vary ac ross time since the drug was administered as seen in the aforementioned experiments (Cevi k, 2003; Pinkston, 2006). Wh en session length is either allowed to vary, or programmed to last multiple hours, the selective effects of a drug on behavior may change. Experiment 2 will explor e whether averaging behavior under drug across an extended session masks other selec tive effects of the drug on behavior. Present Experiments The purpose of Experiment 1 was to evaluate th e acute effects of a wide range of doses of d-am phetamine on various dependent variables unde r the control of a PI schedule. Data from Experiment 1 were assessed with both the sessio n average and single trial analysis. Data were also examined for individual subjects as well as for the group average. Following previous reports we expect that obtained data from th e session average analysis will not necessarily correspond to changes seen under the single trial analysis. In additi on an analysis of the form of individual trials, or Ziggurat analysis, will be conducted to determine if the assumptions of a
31 peaked pattern are met. Assessments of group aver age data will also be ex amined to see if these results reflect changes observed for individual subjects. Experiment 2 will examine behavior under a concurrent PI 45 sec RR 60 with pigeons while varying systematically the time between behavioral test ing and drug administration. A pre-session feeding condition was incl uded in this experiment in or der to manipulate the value of reinforcement or arousal of the subject. BET predicts that when the value of reinforcement decreases timing behavior will suffer. This pre-session feeding condition was compared to results obtained under drug at each delay in order to see if any similarities in disruption occurred. Effects of d-amphetamine on this concurrent task may shed light on what features of behavior are altered by d-amphetamine and how these effects change across administration times. Specifically, changes in the concurrent RR task will be compared to changes in temporally controlled behavior to see if th ere is a correlation. If decr eases in RR responding correspond to decreases in accuracy on a timing task an argument for the existence of mediating behavior and its assistance in timing can be made. A rate depe ndent analysis was also conducted on all data in order to determine whether typical rate dependent changes effects occurred.
32 0.00 5.00 10.00 15.00 20.00 25.00 051015202530 time in peak trialresponse rate baseline rate dependent SET-10% Figure 1-1. Two simulations of possible effects of amphetamine on rates of responding in a peak trial on PI 15 s. The solid line shows typical responding on an averaged PI trial, represented by a Gaussian function with m ean 15 s and a maximum response rate of 20 responses per time unit. The dotted lin e shows how a rate-dependent effect on responding in peak trials can move the time of peak responding to the left. Response times on unreinforced trials were increm ented by an amount that decreased as response rate increased. Response rates af ter food time were unc hanged. The function for the section of the rate-dependent curve before reinforcement time was given by ccacy )( where c is the baseline rate, is the maximum baseline rate and a is a constant (.07). The dashed line shows the prediction of SET assuming only that the rate of the pacemaker in that model is increased by 10%. Note that in the rate dependent model, but not SET, the width of the distribution is incr eased by the drug.
33 Time in Peak Trial 0246810121416 0 10 20 30 40 Maricq et al. Time in Peak Trial 024681012141618Response Rate 0 10 20 30 40 Kraemer Exp 1 Time in Peak Trial 05101520253035Response Rate 0 5 10 15 20 25 30 Time in Peak Trial 051015202530 0.0 0.2 0.4 0.6 0.8 Kraemer Exp 2 Time in Peak Trial 0510152025 0 5 10 15 20 25 30 Bayley et al. Eckerman et al. Figure 1-2. PI distributions ad apted from previous publications. Each panel is a recreation of the PI distributions obtained under baseline (circles) and drug (triangles) conditions for each experiment. Response rate is shown on the y-axis and time in trial is shown on the x-axis.
34 CHAPTER 2 EXPERIMENT 1 The use of drugs as a tools to investigate beha vior has a long history in behavioral science (Branch, 2006) and tim ing beha vior is no different (Sidman, 1955). SET has postulated the existence of an internal pacemaker that c ontrols temporal regulation (Gibbon, 1977; 1992; Gibbon and Church, 1984; Gibbon et al., 1997; Meck, 1996). This pacemaker is said to be modifiable by dopamine levels, su ch that its operation under dopamine agonists can be predicted. Many studies employing dopamine agonists have suppor ted this theory by showing a decrease in peak time (Buhusi and Meck, 2002; Cheng et al., 2007; Eckerman et al., 1987; Frederick and Allen, 1996; Hinton and Meck 1996; Maricq et al., 1981; Matell et al., 2006; Meck, 1996; Meck and Williams, 1997; Taylor et al., 2007). Increases in s.d. or the width of the distribution with decreases in peak time however violate predictions of scalar timing made by SET. If the pacemaker speed is increased by drug the critic al ratio between pulses in the accumulator to pulses in the pulses in memory will be met sooner leading subjects to resp ond earlier. Likewise, the critical value to terminate responding w ill also be reached earlier and subjects will subsequently stop responding at an earlier time (Buhusi and M eck, 2002; Matell et al., 2006). Studies showing a decrease in peak time have al so shown an increase in s.d. (Bayley et al., 1998; Cheng et al., 2007; Eckerman et al., 1987; Kraemer et al., 1997; Matell et al., 2006; Saulsgiver et al., 2006; Taylor et al., 2007). These combined effects have been taken to support the hypothesis that amphetamine-like drugs cause a disruption in overall stimulus control (Odum et al., 2002; McClure et al., 2005; Saulsgiver et al., 2006). Experiment 1 aims to evaluate the application of both these theories to observed d -amphetamine effects on beha vior under a PI schedule in pigeons.
35 Two common methods of analysis exist fo r evaluating behavior under PI schedules, a session-average analysis (Cheng et al., 2007; Hinton and Meck, 1996; Maricq et al., 1981; Matell et al., 2006; Meck, 1996; Meck and Williams, 1997; Saulsgiver et al., 2006; Taylor et al., 2007) and a single-trial analysis (C heng and Westwood, 1993; Matell et al., 2006; Taylor et al., 2007). The session-average analysis first pools all PI trials within or across sessions and subjects. The average response rate as a function of time is th en fit with a function to derive estimates of temporal accuracy, peak time, and the variance of temporal estimates, s.d. A single trial analysis assumes that the gradual increase in res ponse rates observed for average response rate distributions is not necessarily present for single trials rather this gradual increase is a byproduct of averaging multiple trials together. Single trials are said to consis t of responding that is organized in a break-and-run pattern where responding abruptly increases after some time then decreases again. A step function has been used to describe the topography of single PI trials. This analysis uses an algorithm fit to single trials to derive measures rela ted to timing behavior. These measures are then averaged for each session for presentation. Both analyses were employed for this experiment and measures de rived from each analysis that are taken to represent similar indices of timing are compared (Matell et al., 2006; Ta ylor et al., 2007). An additional analysis of behavior was develope d to assess behavior at the individual trial level, the Ziggurat analysis. This consisted of fitting a one-step and two-step function to each individual trial. Least squared residuals were used as a metric to determine which model best represented each trial. This statistic determin ed the best fit by controlling for the different degrees of freedom across each fit. This assessment identified whether the topography of individual trials in this experime nt were peak-like in form (two-step model) or if the data were better described as lacking temporal organizatio n (one-step model). If a peak-like form is
36 present at the individual trial level then one c ould argue that a session -average analysis is appropriate to use in assessing be havioral change. If the one-s tep model better describes more individual trials then there woul d be evidence that temporal organization is not apparent at the individual trial level. Differences in response rates observed thr oughout the PI distributions were assessed visually by examining changes across time as a percent of saline. This method of assessment highlighted portions of the distri bution that were most changed by drug. Typically drug effects on PI schedules are only assessed on parameters derived from session average and individual trial analysis rather than on obt ained response rates under the dist ribution. This novel analysis was conducted to better quantify exactly wh ere response rates are most altered. Previous research examining individual subj ects has indicated that the group average data may not reflect the subject-to-subject differen ces in drug effect (Saulsgiver et al., 2006). Individual subject assessment s were also compared to a ssessments conducted on the entire group. This was done to determine if group aver ages are representative of individual subject variability in this experiment. In addition the cu rrent experiment examined the acute effects of a wide range of doses of d -amphetamine (0.1-3.0 mg/kg) on be havior. This range of doses of d amphetamine is not novel for experiments of th is type but examining the entire range on each subject is. Method Subjects Five White Carneau pigeons with previous experience on PI schedul es (Subject 662) and on MTSD procedures (Subjects 95, 302, 451, 593) served as subjects. A ll subjects had been exposed to 1.5 and 3.0 mg/kg doses of d -amphetamine chronically for 20 days. At least 60 days separated this previous experien ce from the start of this experiment. Subjects were individually
37 housed in a humidityand temperaturecontroll ed colony room on a 16:8 hour light: dark cycle. Water and health grit were continually available in the home cages. Subjects were maintained at their 83% free-feeding weights by post-session feedings as needed. Apparatus Five operant test chambers (Med Associates Model ENV-007) with internal dimensions of 30.5 x 24.1 x 29.2 cm served as the apparatus fo r this experiment. The front intelligence panel contained two circular response keys 2.5 cm in diameter which could be transilluminated with red, green or white lights. Each key was placed 2. 5 cm from the side walls and 6.5 cm from the top of the chamber. The keys were 14 cm apart and required between 0.12 and 0.15 N for effective depression. Reinforcement was mixed gr ains dispensed into an illuminated 5.5 by 6.5 cm rectangular hopper opening centered 20 cm from the top of the chamber. The chamber was illuminated by a 100-mA houselight and contained in a sound attenuating chamber. Experiments were programmed in Med-PC IV software. Procedure Training Due to previous experience subjects required no magazine or key-peck training. Subjects were first placed on a discrete-trial FI 45-s schedu le. At the start of a trial the houselight came on and the center key was illuminated white. Res ponses on the center key prior to 45 s elapsing had no programmed effect on the schedule. The firs t peck after 45 s resulted in the center key extinguishing and access to mixed grain for 3 s. The houselight was then turned off and an average 40-s inter-trial interval (ITI) commenced (range 5 to 120 s). E ach session consisted of 50 trials. Experimental sessions were conducted at approximately the same time seven days a week.
38 After response rates stabilized, PI trials were added. Stability on FI trials was determined by examining the pattern of response rates in 3-s bins. PI trials we re three times as long as the scheduled FI duration (i.e., 135 s) and terminated without reinforcement. In each block of five trials one, randomly chosen, was th e PI trial resulting in 10 PI trials per session. Subjects remained on this procedure for a minimum of 50 sessions before drug was administered (Range 50-77 sessions). Drug dosing Once average response rate distributions for PI trials appeared stable across 10 days acute effects of d -amphetamine were determined. Five doses of d -amphetamine (0.1, 0.3, 1.0, 1.7, and 3.0 mg/kg) and saline injections were given twice a week in des cending order (saline, 3.0, 1.7, 1.0. 0.3, 0.1 mg/kg) 20 min prior to experimental conditions. This dosing cycle was repeated at least twice. Supplemental doses were given as needed to re-determi ne the dose effect for inconsistent effects of the drug. Data analysis Session average parameters. Total response obtained across each PI trial were averaged across the session and plotted in 3-s bins for each subject. P eak time of responding (point of maximal response rate) and the width (s.d.) of the distribution of response rates was estimated for each session by a Gaussian functi on with an added linear component. ( 2-1) For Equation 2-1 a is a constant, tO is the peak time or timing accuracy, and b is the s.d. (Matell et al., 2004; Saulsgiver et al., 2006). Days on which the coefficient of determination of the model fit was less than .6 or the session contai ned fewer than five peak trials were excluded
39 from this analysis. This occu rred for all administrations of the 3.0 mg/kg dose for Subject 302 and for one administration of this dose for Subject 593. Single trial parameters. An algorithm described by Cheng and Westwood (1993) and Grace et al. (2006) was used to derive measures that describe single trial behavior. These parameters were calculated for each trial then av eraged across each session. The parameters derived were start time, stop time, middle, run length, run rate, and number of responses between start and stop time. Start time is defined as the mean over individual trials of the first bin of the first two consecutive bins where responding occurs for each single peak trial. Stop time is defined as the mean over individual trials of the first bin of the first two consecutive bins where no responding occurs following the start time. Middl e is defined as the av erage of start time and stop time for each PI trial. Run length is the time between start time and stop time. Run rate is defined as the average response ra te for the time period between th e start and stop time. Finally the number of responses between start and stop time was calculated for each PI trial and averaged across trials. Single PI trials where fewer than 20 responses occurred were excluded from this analysis. Across subjects this resulted in the exclusion of fewer than 2% of all PI trials. It should be noted that these trials al l occurred under th e higher doses of d -amphetamine. Ziggurat analysis. Individual trials for each session a nd subject were evaluated in order to determine if each trial was best described by a model that only assu mes states of responding and no or low responding (one-ste p model) or a model that has two non-zero rates of responding and appears more peak-like (two-step model) The one-step model assumes that when responding occurs it does so at a constant ra te, whereas the two-step model assumes that responding increases and then decreases througho ut the individual trial. This was done by plotting responses per second for individual trials and then adjusting the parameters for each
40 model to minimize the sum-squared difference betw een the predicted and the obtained functions. Individual trials were truncated at 90 s for this analysis. The better fitting function overall was then determined by examining a reduced chi-squared statistic. This statistic allowed for a direct comparison of the two models desp ite the difference in parameters. It was calculated by dividing the total sum squared difference for each model by the degrees of freedom for that model. The fit with the lower reduced chi-squared statistic was the better fit for the obtained data. The parameters assessed for the one-step function incl uded start time, stop time, and response rate. The start time indicated the point in the trial when responding consistently increased from a rate of zero (see Figure 2-1 for an illu stration of all parameters from this assessment). Stop time designated when responding returned to zero. Re sponse rate identified th e average response rate obtained between start and stop time in each tria l. The two-step function included two start times, two stop times, and two response rates. Th e first start time indicated the first increase in responding from zero to the first average respons e rate. The second start time indicated when responding increased again to a second, higher, average response rate. The first stop time specified when responding decreased from the second higher response rate back to the first average response rate. The final stop time was obtained when res ponding decreased from the first response rate back to a rate of zero. Rate dependency. Log response rates in each succe ssive 3-s time bin under drug were plotted as a function of log response rates for eac h successive time bin averaged over all baseline sessions. These resulting scatter plots were then fit with a straight line of which slope ( a ) and intercept ( b ) were derived. Log (rate under drug) = a log (baseline rate) + b (2-2)
41 A one-way ANOVA with within subjects fact or dose was run on the entire group for each parameter derived from the session average analysis, individual trial analys is, and rate dependent analysis. Separate statistical tests were used wher e indicated. An alpha level of .05 was used in all statistical tests. Results Peak Interval Distributions The left column of Figure 2-2 shows the se ssion-average response rates per second during peak trials for each subject under each dose. The effects of d -amphetamine varied across dose and subject. In some cases high doses of d -amphetamine increased responding during all portions of the distribution (e.g., Subjects 302, 45 1 and 593). In other cases (Subject 95) there was a decrease in responding under higher doses. An increase in the width or spread of the distributions occurred at higher doses for Subjects 451, 593, and 662. Apparent shifts in peak time were variable across subjects and doses. PI distributions were transformed to reflect changes throughout th e distribution under drug as a percent of saline for the right column of Figure 2-2. This data transformation allows for a better assessment of changes occurring at various portions of the distribution under drug. There was an increase in responding under drug observed for most portions of the distribution for most subjects. For most subjects large increases in re sponse rate can be seen at the beginning of the interval and following where responding tended to decelerate after the time of reinforcement. A decrease in response rate at the typical time reinforcement is deliv ered on FI trials can also be observed at higher doses for Subject s 302 and 662 while there is little change in this portion of the distribution for Subjects 95 and 593. Subject 451 showed increases in this portion of the distribution. With the exception of the increas e in rate observed for Subject 451 at the typical time of reinforcement, changes under drug co rrespond to a widening and flattening of the
42 distribution. These observed changes in res ponse rate were investig ated with a two-way ANOVA with repeated subjects factors dose (there were five dos es per replication) and block (PI distributions were divided into eight 18s blocks). This assessment was conducted on individual subjects and on the enti re group. An alpha level of 05 was used to judge significance for each test. A significant main effect of dose was f ound for Subjects 451, 593, and for the group (lowest F = 7.77, individual subject d.f. = 5, 10 gr oup d.f. = 5, 55, p < .05). The right column of Figure 2-2 shows that for Subjects 451 and 593 th e magnitude of change in response rates across the distribution increased with increasing doses of d -amphetamine. This pattern of effects can be seen to a smaller degree in the remaining subjec ts with the exception of Subject 95 who showed the smallest change at any portion of the distribution. There wa s a main effect of block for Subjects 593, 662, and the group (lowest F = 2.92, indi vidual subject d.f. = 7, 14, group d.f. = 7, 77, p < .05). Response rates in blocks corres ponding to the first 18 s and seconds 90-108 were increased the most by drug for both these subjects as well as for the group as a whole. These portions of the distribution contained the lo west observed response rates during baseline conditions. A significant intera ction between dose and block wa s found for all subjects, except Subjects 95 and 302, as well as th e combined group (lowest F = 1.62, individual subject d.f. = 35, 70, group d.f. = 35, 385, p < .05). In general the increasing effect of d -amphetamine on response rates were observed during the first 18 s and seconds 90-108 of the distribution. In order to further understand where the la rgest changes were occurring with this interaction a post-hoc Scheffe analysis was conducted on these subjects as well as on the group. In Figure 2-3 error bars show the minimum change required for a difference to be significant at each time bin. If the value at a particular dose for a particular bin falls outside the range of the
43 error bar then there is a significant change at that dose for that bin. Ta ble 2-1 summarizes where these changes were observed for Subjects 451, 593, 662 and the group. The areas of the distribution that were most cha nged by the drug were those that had the lowest response rates during baseline conditions. The response rates du ring the first 18 s of the distribution (bin 1) increased with increasing drug dose for these subj ects and for the group. Response rates in bins corresponding to 90-125 s (bins 5-6) were signi ficantly altered by higher drug doses for Subject 451 as well. Gaussian Derived Parameters and Estimates on Session Average Peak Trials To assess the effects of d -amphetamine on particular aspect s of the pigeons behavior a Gaussian distribution with an added linear component (Cheng and We stwood, 1993; Matell et al., 2004; see methods above) was fit to the respons e rates of the daily average distributions for each subject. The two parameters derived fr om the curve fitting procedure on the session average distribution, peak time and standard devia tion (s.d.), are shown as a percent of saline in Figure 2-4. Various effects of drug on peak time were seen across subjects. There was no change in peak time across doses for Subject 95 while some increases in peak time were observed for Subject 302. Decreases with incr easing drug dose can be seen for Subjects 593 and 662. The largest of these decreases occurred und er the 3.0 mg/kg dose where peak time fell to about 50% of baseline levels for Subject 593. Su bject 451 showed a large decrease in peak time under the 1.7 mg/kg dose. A significant main e ffect of dose was found for the group for peak time (F = 2.82, d.f. = 5, 25, p < .05). As can be seen in Figure 2-4 the decrease in peak time was large for Subjects 593 and 662 while th ere is little decrease in this parameter for other subjects. It is most likely the effect of drug for these two subjects that led to th is significant effect of dose. An increase in s.d. with in creasing drug dose can be seen for Subjects 95, 593, and 662. There was no change in s.d. across all doses fo r Subject 302. The only in crease in s.d. observed
44 for Subject 451 was under the 1.7 mg/kg dose. A significant main effect of dose was found for s.d. (F = 6.03, d.f. = 5, 25, p < .01). S.d. increased relative to baseline by over 100% at the higher doses of d -amphetamine for three subjects and under the 1.7 mg/kg dose for Subject 451. Parameters Derived from Single Trials An assessment of individual peak trials with in each session was also conducted according to the methods of Grace et al. (2006; see met hods above). Parameters derived from assessing changes in individual trials are shown in Figure 2-5. Start tim e decreased with increasing dose for Subjects 95, 451, 593, and 662 (1st row, Figure 2-5). The largest of these decreases can be seen under the highest doses of d -amphetamine. Subject 302s start time initially decreased slightly under low doses but then returned to sa line levels under high dos es. A significant main effect of dose was found for start time for the gr oup (F = 3.33, d.f. = 5, 55, p < .05). There was little effect of drug on stop time for Subjects 95, 302, and 662 with some exceptions at the highest dose for S ubject 662 where stop time decreased (2nd row, Figure 2-5). There was an increase in stop time across doses for Subjects 451 and 593 however, no significant main effects were found (F = 0.75). Small variable changes were observed fo r the middle paramete r across subjects (3rd row, Figure 2-5). Little effect was s een for Subjects 95 and 302 and increases were seen for Subjects 451 and 593. Subject 662 showed a consistent d ecrease in middle across doses. There were no significant effects found for this para meter either (F = 1.07). The effect of d -amphetamine on run length varied across subjects and doses (4th row, Figure 2-5). There was little effect of the drug on run length for Subjects 95 and 302. An increase in run length can be seen under most doses for Subjects 451, 593 and 662. Sometimes there was more than a 100% increase detected. There were no significant main effects found for run length (F = 1.42).
45 Run rate (5th row, Figure 2-5) was unaffected by d -amphetamine for Subjects 95, 302, 593, and 662. An increase in run rate with increasi ng doses was evident for Subject 451. There were no significant main effects found (F = 0.25). Substantial increases in the number of respons es between start and st op time can be seen for Subjects 451, 593, and 662 across some doses of d -amphetamine (6th row, Figure 2-5). There was relatively little change for the other two subj ects. There were no significant main effects found for the group (largest F = 1.07). Individual Subjects vs. Group Average Data Figure 2-6 shows the average effect of drug on parameters derived from the session average analysis and individual trial analysis. When data are averaged across subjects a clear decrease in peak time is observed. A 10-15% decrease in peak time was observed at the higher doses. S.d. increases for the average of all s ubjects with increasing drug dose. A 20% increase was seen at the 3.0 mg/kg dose. Start time decreases and stop time increased with increasing drug dose. No change in the middle parameter wa s observed for the group average. The largest increases for the groups were observed in the numb er of responses between start and stop time. An increase of 40% was seen at the highest dose. Ziggurat Analysis Figure 2-7 provides examples of obtained individual trial re sponse rates (solid line) for Subject 95 across various doses of d -amphetamine. The one-step function is represented by the dashed line and the two-step f unction is shown by the dotted line. These examples are typical of the remaining subjects. As can be seen in thes e examples there does appear to be a peak-like nature to some individual trials while other tria ls are better described by a single-step function. Individual trials obtain ed under the 1.0, 1.7 and 3.0 mg/kg doses are examples where a one-step function better described the data. While occasional bins contain rates higher than the average
46 rate predicted by a one-step function these distri butions overall were better described by the onestep model. When responding occurred, response rates were relatively uniform resulting in a flatter distribution. The other gr aphs in this figure provide ex amples where a two-step function better described the data. In thes e examples a clear increase in responding from a rate of zero is followed by a second increase in response giving the impression of a peak-like distribution. To better quantify whether a peak-like dist ribution was observed under individual trials, Figure 2-8 plots the number of trials for each su bject that were best described by a one-or twostep model or by neither model as determined by the reduced chi-squared statistic. When fewer than 20 responses occurred on a particular trial neith er model was a satisfactory to fit to the data. In general for all subjects a two-step function described more individual trials under saline and low doses of d -amphetamine. As the dose of d -amphetamine increased the two-step function accounted for fewer of the individual trials. This increase in the number of trials accounted for by a one-step function, and decrea se in number of trials accounted for by a two-step function with increasing drug dose can be seen across all subjects. In addition for some subjects the number of trials that was not well described by either function b ecause it contained less than 20 responses increased. This impression was s ubstantiated by a repeat ed measures two-way ANOVA with within-subjects factors Model (one or two-step) and Dose (saline 3.0 mg/kg doses). A significant main effect for model was found (F = 43.14, d.f. = 1, 4, p < .01). The twostep function accounted for more tr ials overall. There was also a significant interaction between model and dose (F = 4.73, d.f. = 5, 20, p < .01). This interaction can be seen in the increase in the number of trials best described by the one-s tep model and decrease for the two-step model with increasing drug dose. No main effect of dose was found (F = 0.87).
47 Rate Dependency Rate dependency has been used to describe results obtained in tim ing experiments using drugs (Knealing and Schaal, 2000; Odum, 2002; Saulsgiver et al., 2006, 2007). In these experiments there is an increase in lower re sponse rates while higher response rates either decrease or are altered to a lesser degree by the drug. To ev aluate whether rate dependent changes occurred in this experiment, response ra tes in each successive 3-s bin of the sessionaveraged PI trials under drug were plotted against baseline average response rates in the same 3-s bins on log axes. Figure 2-9 shows scatter plots of the average for each drug dose against saline on log axes for the entire PI distribution. Each plot was fit with a separate linear function (Equation 2) from which the slope and intercept we re derived. The solid line in each plot has a slope equal to one and intercept of zero which would imply that the response rates are unchanged from baseline. An increase in the intercept al one would indicate that response rates increased equally for the entire distribu tion whereas a decrease in inter cept would indicate that response rates had decreased. A decrease in slope occurs when low response rates increase and/or high response rates decrease. An increase in slope could suggest that low response rates have decreased and/or high response ra tes have increased. Rate dependent changes for amphetamine are typically characterized by decreases in slope a nd increases in intercept. Visual inspection of the linear fits in Figure 2-9 show that low response rates increased under d -amphetamine while higher response rates decreased, or were left unchanged especially at the higher doses for all subjects. Figure 2-10 shows the resulting intercepts and slopes for each subject at each dose of amphetamine. Increases in intercept can be seen for all subjects especially at the highest doses of d -amphetamine. Lower doses of drug tended to decrease or had no effect on intercept across
48 subjects. This increasing trend in intercept with increasing dose resulted in a significant main effect of dose (F = 18.32, d.f. = 5, 50, p < 0.05). The effect of d -amphetamine on slope varied across s ubjects. Slope decreased especially at the highest doses for Subjects 95 and 662. There was little change or sm all increases in slope observed for subjects 451 and 593. Subject 302 showed decreases under th e lower doses and an increase under the 1.7 mg/kg dose. Despite th ese inconsistencies there was a significant main effect of dose (F = 3.54, d.f. = 5, 50, p < 0. 05). This significant effect of dose is most pronounced for Subjects 95, 302, and 662. Discussion SET postulates that subjects exposed to the PI procedure and administered dopamine agonists should show a decrease in peak time relative to baselin e (Gibbon, 1977; 1992; Gibbon and Church, 1984; Gibbon et al., 1997). This theo ry also states that the variance in time estimates should not increase with decreases in peak time; rath er the variance should remain proportional to the mean interval to be timed. Th e data presented for peak time and s.d. together do not support these predictions. When subjects were examined separately only two (593 and 662, Figure 2-4) showed a clear decrease in peak time with incr easing drug dose. When the average peak time was examined (Figure 2-6) howe ver, there was a clear decrease in peak time for the group. Drug effects on s.d. for indivi duals as compared to the group average were similar. When subjects were examined separately only three showed an increase in s.d. (95, 593, and 662, Figure 2-4). When subjects were averaged together there was a in crease in s.d. (Figure 2-6). Regardless of which data presentation is us ed to represent the data, SET is not supported. If the average is used then the required decrea se in peak time is achieved but it is accompanied by an average increase in s.d. which violates S ET predictions. On the other hand if individual subjects are assessed then a decrease in peak time is not seen for most subjects. It should be
49 noted however, that this lack of decrease in peak time seen for some subjects in this experiment may be similar in other experiments, however since only one other study reported individual peak times across subjects (Saulsgiver et al., 2006) it difficult to determine. It is also possible that the absence of a decrease in peak time here is due to other variations in procedures used here in comparison to other studies (i.e ., ratio of FI to PI trials, se ssion length, training history). Furthermore in cases were a decrease in peak time is observed (Subjects 593 and 662) there is also a clear increase in s.d. for these subjects, again violating SET predictions. Data from the single trial analysis contradi cts SET to a greater degree. There was no consistent change observed for the middle paramete r regardless of data presentation (Figure 2-5 and 2-6). The middle parameter has been taken to be synonymous with peak time in that it indicates temporal accuracy (Taylor et al., 2007). A decrease in this parameter is needed to support SET at the level of the indivi dual trial. Run length is similar to s.d. in that it is taken to be a measure of the variance in time estimates from trial to trial (Taylor et al., 2007). Run length increased for two subjects (451 and 593, Figure 2-5) and there was little change for the other three when subjects were examined independently When the average was examined (Figure 26) however, there was an increase observed. As stated before, SET predicts that temporal perception should decrease and the width of the distribution should decrease proportionally. This was not the case for these data. Increases in the variance of time estimates under drug have been taken to indicate a loss of stimulus control as exerted by the schedule in pl ace (Saulsgiver et al., 2005). Increases in this width could result from rate-dep endent like changes in behavior (Odum et al., 2002; Saulsgiver et al., 2007). This characterization states that under stimulants low response rates should increase and/or high response rates should decr ease or be left unaffected (Dews, 1958, 1970).
50 This would be reflected by an incr ease in the intercept and a decr ease in the slope parameter. When logged drug response rates were plotted as a function of logged baseline response rates and fit with a straight line, intercept increased with increasing drug dose fo r all subjects (Figure 2-10). Slope decreased with incr easing drug dose for three subjects. The rema ining two subjects (451 and 593) showed a variable effect of drug on slope. When the scatter plot s for these two subjects are examined more cl osely (Figure 2-9) it can be seen that slope was relatively unchanged due to an increase in responding observed for higher response rates under drug. These rate dependent changes resulted in an increase in the width of the distribution by increasing low rates before and after the typical time of reinforcement. These impressions are further supported by the analysis conducted on the PI distributi ons (Figure 2-2 and 2-3). The portions of the distribution that were most altered by drug was the segment of the distribution that had lower response rates under baseline cond itions. This widening of the distribution caused by increases in responding early and late in the tr ial conforms to an account that implicates the loss of stimulus control as the main effect of stimulant drugs on temporal ly organized behavior (Knealing and Schaal, 2002; McClure et al., 20 05; Odum et al., 2002; Sa ulsgiver et al., 2007). These results show that changes in behavi or found at the sessi on average level do not wholly reflect changes observed at the individual trial level. Peak time, derived from the session average analysis, and middle, taken from the singl e trial analysis, have been both been postulated to represent temporal accuracy (Taylor et al. 2007). In this experiment however there was an inconsistent effect of d -amphetamine on these two parameters There was a significant main effect of dose found for peak time while no main effects were found for middle across subjects. A closer look at Figures 2-4 a nd 2-5 shows the effect seen for each variable for each specific subject does not always match. There was an increase in peak time for Subject 302 across all
51 doses of d -amphetamine but a decrease in middle was observed for the lower three doses. The only effect of d -amphetamine on peak time for Subject 451 was a decrease seen at the 1.7 mg/kg dose. Contrary to this d -amphetamine increased the middle parameter across all doses for this subject. The decrease in peak time observed for Subject 593 was not observed for the middle parameter, rather an increase was observed. The effect of d -amphetamine on Subjects 95 and 662s peak time and middle was similar across anal yses however variations in the magnitude of the effect can be observed. A similar pattern of differences can be seen for s.d. and run length parameters both of which represent the spr ead of responding for th e session average and individual trials respective ly (Matell et al., 2006). When the average effect is examined for thes e parameters across analyses there is still a discrepancy in effect seen between peak time and middle. The average peak time decreases while there is no change in the average middle parameter. These differences across analyses have been reported before (Matell et al., 2006; Ta ylor et al., 2007). Tabl e 2-2 shows the general effect of drug on each of these para meters for this and previous experiments. Matell et al. (2006) and this experiment showed a lack of consistenc y across the peak time and the middle parameter. Taylor et al. (2007) and Matell et al. (2006) showed divergen t results for the s.d. and the run length parameter. Taylor et al. (2007) found no change in s.d. under drug but found an increase in run length. An increase in s.d. was repor ted under drug by Matell et al. (2006) while run length decreased. If these parameters are meant to illustrate the same aspect of behavior just at different levels of analysis one would expect similar effects of drug to be shown within an experiment. If there are differences in the resu lts obtained from each analysis which is the best one to use?
52 In an attempt to answer this question a nove l assessment was used. Each PI trial was fit with two different models, one based on the assumpti ons of an individual tr ial analysis (one-step model) and one based on the session average an alysis assumptions (two-step model). This analysis revealed that the majority of trials under saline and low doses of d -amphetamine are similar in appearance to the session average distribu tions typically used. As drug dose increased however, these individual trials became less and le ss peak-like in form. Individual peak trials appeared flatter and were not necessarily represen ted by the session average form. This impression of individual trials validates using a session averag e analysis to assess chan ges under low doses of d -amphetamine but calls into question the use of this analysis wh en a peak-like form is no longer present under higher doses. Figure 2-6 shows the average effect of d -amphetamine on parameters derived from the session average analysis and single trial analysis When these panels are compared to the corresponding panels in Figures 2-4 and 2-5 it is clear that individua l variability is lost. If data were assessed at this level, one would have to conclude that there was a decrease in peak time with increasing drug dose, however when subjects were examined individually this effect was only seen for two subjects. A similar diffe rence between individua l subjects and the group average can be seen for s.d.. A cl ear increase in s.d. is seen for the average but this effect is only apparent for two subjects. This pattern of eff ects across parameters holds for the single trial analysis. The data presented here are c onsistent with other published re ports that failed to support and supported SET (Bayley et al., 1998; Cheng et al., 2007; Eckerman et al., 1987; Kraemer et al., 1997; Matell et al., 2 006; Saulsgiver et al., 2006; Taylor et al., 2007 ). It may be that differences in procedure could account for the inc onsistencies in reported data seen here and
53 across other experiments. These differences could include training ti me, the length of the session, and the delay between drug dosing and beha vioral testing. Some of these procedural differences and how they contri bute to obtained results will be explored in Experiment 2.
54 Table 2-1. Post-hoc Scheffe results by bin assessing where significant departures from baseline rates of responding occurred 451 593 662 All Subjects Bin 1 1.0, 1.7, and 3.0 3.0 1.7and 3.0 1.7 and 3.0 Bin 2 Bin 3 Bin 4 Bin 5 1.7 and 3.0 Bin 6 3.0 Bin 7 Bin 8
55 Table 2-2. A comparison of dr ug effect on parameters derived from a session average and a individual trial analysis for the cu rrent and two previous experiments. Saulsgiver et al. Taylor et al. (2007) Matell et al. (2006) Peak time Some (Avg: ) Middle Variable (Avg: No change) N o change SD Some (Avg: ) N o change Run length *Note: This parameter was not reported for Taylor et al. (2007). Start time and stop time were however reported so run length was derived from those two meas ures for the purpose of this report.
56 0 1 2 3 4 1-sec bins 0102030405060708090Response rate (sec) 0 1 2 3 4 One-step model Two-step model Start time Stop time Start 1 Start 2 Stop 1 Stop 2 Figure 2-1. Example individual tr ial data across various doses of d -amphetamine. A solid line represents the actual obtained data for a part icular trial. Long-dashes and dotted lines represent the fitted oneand twostep function in the top and bottom panel respectively. Circles and arrows show ex amples of the derived parameters for each model.
57 0 50 100 150 200 250 300 0 50 100 150 200 250 300 Response rate per second (% saline) 0 50 100 150 200 250 300 0 50 100 150 200 250 300 0 50 100 150 200 250 300 302saline1 302-0.1-1 302-0.3-1 302-1.0-1 302-1.7-1 302-3.0-1 302saline1 302-0.1-1 302-0.3-1 302-1.0-1 302-1.7-1 302-3.0-1 302 0.0 0.5 1.0 1.5 2.0 2.5 3.0 302saline1 302-0.1-1 302-0.3-1 302-1.0-1 302-1.7-1 302-3.0-1 662 3-sec bin 0.0 0.5 1.0 1.5 2.0 2.5 3.0 593 0.0 0.5 1.0 1.5 2.0 2.5 3.0 451 Response rate 0.0 0.5 1.0 1.5 2.0 2.5 3.0 95 0.0 0.5 1.0 1.5 2.0 2.5 3.0 DRC 1 saline 0.1 mg/kg 0.3 mg/kg 1.0 mg/kg 1.7 mg/kg 3.0 mg/kg 0 15 30 45 60 75 90 105 120 135 0 15 30 45 60 75 90 105 120 135 Figure 2-2. The left column shows the conditi on average response rates per 3-s bin during peak trials for each subject under each dose of d -amphetamine. The right column shows the changes throughout the dist ribution under drug as a percent of saline. The solid line represents the average distribution unde r saline, the long-dashed line shows the average distribution under the 0.1 mg/kg dose, the short-dashed line shows the average distribution under the 0.3 mg/kg dose, the dotted line shows the average distribution under the 1.0 mg/kg dose, th e long-dash, single dot line shows the average distribution under the 1.7 mg/kg dose, and the long-dash, double dotted line shows the average distribution under the 3.0 mg/kg dose.
58 5930 100 200 300 400 500 600 1836547290108126144% salin e 0.1 mg/kg 0.3 mg/kg 1 mg/kg 1.7 mg/kg 3.0 mg/kg 6620 200 400 600 800 1000 1836547290108126144 4510 100 200 300 400 500 600 1836547290108126144 Elapsed time (seconds)% salin e all0 100 200 300 400 500 600 1836547290108126144 Elapsed time (seconds) Figure 2-3. Response rates as a percent of sa line are shown for each 18-s bin. The minimum change required for a significant differen ce to be found at each bin is represented by the error bars. If the value at a particular dose for a particular bin falls outside the range of the error bar then there is a significan t change at that dose for that bin. Note the y-axis scale difference for Subject 622.
59 662 V0.10.311.73s.d. 0 50 100 150 200 250 300 350 Peak time 0 20 40 60 80 100 120 140 95 V0.10.311.73 302 V0.10.311.73 451 V0.10.311.73 593 dose (mg/kg) V0.10.311.73 % saline Figure 2-4. Peak Time and sta ndard deviation derived from the curve fitting methods described in Methods for the session average distribu tions shown as a percent of saline. The solid line represents th e mean of all values.
60 451 V0.10.311.73# Responses between 0 50 100 150 200 250 300 Stoptime 0 50 100 150 200 250 Run length 0 50 100 150 200 250 300 Run rate 0 50 100 150 200 250 300 Middle 0 50 100 150 200 V0.10.311.73 V0.10.311.73 V0.10.311.73 302 593 662 95 starttime 0 50 100 150 200 Dose (mg/kg) V0.10.311.73 % saline Figure 2-5. Start time, stop time, middle, run le ngth, run rate, and numbe r of responses between start and stop times derived from the single trial analysis are s hown as a percent of saline. All other features of the figure are identical to Figure 2-4.
61 Peak time 60 80 100 120 140 SD 60 80 100 120 140 Start time 60 80 100 120 140 Stop time 60 80 100 120 140 dose (mg/kg) V0.10.311.73Run length 60 80 100 120 140 160 180 V0.10.311.73Middle 60 80 100 120 140 % saline Figure 2-6. The group average dr ug effect on parameters from Figure 2-4 and 2-5 are shown as a percent of saline. Standard error is shown by the error bars.
62 Responses per sec 0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 observed 1-step 2-step 95 1-sec bins 0102030405060708090 0 1 2 3 4 0102030405060708090 Saline 0.1 mg/kg 1.0 mg/kg 1.7 mg/kg 1.7 mg/kg 3.0 mg/kg 0.1 mg/kg Saline Figure 2-7. Example individual tr ial data across various doses of d -amphetamine. A solid line represents the actual obtained data for a part icular trial. Long-dashes and dotted lines represent the fitted oneand twostep function respectively.
63 95 0 10 20 30 40 Neither 1-step 2-step 302 0 10 20 30 40 451 # of trials 0 10 20 30 40 593 0 10 20 30 40 662 Dose S a lin e 0 1 mg / k g 0 3 mg / k g 1 0 mg / k g 1 7 mg / k g 3 0 m g / k g 0 10 20 30 40 Figure 2-8. The number of individual trials be st accounted for by a one and two-step function across all doses for all subjects. A gray ba r represents the number of trials that no function would work for, a black bar show the number of trials best fit by a one-step function and the white bar shows the number of trials best fit by a two-step function.
64 95 0.01 0.1 1 10 0.1 mg/kg 0.3 mg/kg 1.0 mg/kg 1.7 mg/kg 3.0 mg/kg No change 302 0.01 0.1 1 10 451 Comparison Response Rates 0.01 0.1 1 10 593 0.01 0.1 1 10 662 Average control response rates 0.010.1110 0.01 0.1 1 10 Figure 2-9. Log session averag e response rates under drug cond itions as a function of log response rates averaged across all baselin e sessions for the entire trial for each subject. A line of unit slope indicates no change in response rates from baseline.
65 Intercept -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 S0.10.311.73 dose (mg/kg) S0.10.311.73 S0.10.311.73 S0.10.311.73 S0.10.311.73Slope 0.0 0.5 1.0 1.5 2.0 2.5 95 302 451 593 662 Figure 2-10. Slope and intercept derived from the scatter plots showing log session average response rates under drug conditions as a function of log response rates averaged across all baseline sessions for the entire inte rval. All other features of the figure are identical to Figure 2-4.
66 CHAPTER 3 EXPERIMENT 2 Killeen and Fetterm ans (1988) Behavioral Theory of Timing (BET) states that mediating behaviors are elicited by signals of reinforcement and that these mediating behaviors may assist in discriminating the passage of time. Identify ing and measuring these mediating behaviors has been difficult. One possible solution to this probl em is to program a concurrent task and allow for the adjunctive or collateral behavior to develop on this alternative task. Sanabria and colleagues (2005) showed that behavior was tem porally regulated under a PI schedule when a concurrent task was available. It is possible that be havior that occurs on this concurrent task could come to assist in timing. If mediating behavior develops on this concurrent task while subjects learn to time, and this collateral behavior assists in mediating the passage of time, then changes in the alternative task may predict change s in timing. BET predicts that as mediating behavior declines in rate timing accuracy will also decrease. The current experiment aims to test predictions of BET, SET and rate dependency on a concurrent PI RR schedule. The degree to which damphetamine disrupts stimulus control under these schedules will also be explored. Recent research has shown that the interval from d and meth-amphetamine administration to behavioral testing can imp act observed behavioral changes (Cevik, 2003; Pinkston, 2006). The time between metham phetamine administration and testing was manipulated within a temporal di scrimination task with rats by Cevik. When an analysis was performed on both the first and second halves of the experimental sessions, differences in the degree of flattening of the psyc hophysical functions were observed. There was more flattening or loss of accuracy at the extreme values of the psychophysical function observed under drug during the first 80 minutes of each session than du ring minutes 100-180 of experimental testing. These results indicate that time between metham phetamine administration and behavioral testing
67 is an important variable to c onsider when investigating tempor al discrimination. The current experiment explores whether time differences between drug administration and behavioral testing result in similar differences under a PI procedure. This experiment will test the impact of a range of delays between drug administration and behavioral testing while controlling session length to investigate the impact delay to behavioral testing and session length have on obtained results. Ward and Odum (2005) investigated the impact of non-pharmacological disruptors on timing behavior. One method used was to pre-feed subjects prior to e xperimentation. This manipulation was shown to disrupt temporal disc rimination under a MTSD procedure. The acute effects of pre-session feeding on behavior under a PI schedule wi ll also be examined in the current experiment. BET holds that the pacemaker is tied to the rate or value of reinforcement and that as that rate or value decreases, eith er through extinction or satiation, timing accuracy will also diminish. A comparison between this disruptor and d -amphetamine will also be included. Method Subjects Five experimentally nave White Carneau pigeon s served as subjects. Care and housing of these subjects was identical to that used in Experiment 1. Apparatus Identical to Experiment 1. Procedure Training Subjects were magazine trained by reinforc ing successive approximation to approaching the hopper when the hopper light was on prior to experimental sessions. Auto-shaping (Brown
68 and Jenkins, 1968) was used to establish key pe cking on both the left and right keys. When responding was established on both ke ys, subjects were placed on a fixed-ratio (FR) 2 schedule. The FR 2 schedule was in effect on the left key when the red key light was illuminated and on the right key when the green key light was illuminated. The active key alternated every five minutes. The ratio incremented across sessi ons to an FR 15 schedule with 40 scheduled reinforcer deliveries (20 on each key). Upon comple tion of these sessions, subjects were placed on a concurrent discrete trial FI 10-sec RR 10 schedul e. The FI schedule was in effect on the left key under the red stimulus light and the RR sche dule was in effect on the right key under the green stimulus light. The FI schedule was trained by first implementing a FI 10-sec schedule on the left key. This interval value was increased by 5 sec across subsequent sessions to a terminal value of 30 sec. Inter-trial intervals (ITI: av erage 20 sec) where the keylights and houselight were terminated separated FI trials. Food deliveries were a 3-sec hopper presentation for both schedules. When a food delivery and an ITI comme nced all key lights were extinguished and the RR schedule was not in effect. The RR sche dule was shaped by first programming a RR 10 schedule on the right key. This value was increm ented by steps of 10 to a terminal value of 60 across sessions. There was no programmed ITI for this schedule. When food was delivered for responding on the RR schedule the scheduled interval was still in effect on the FI schedule. A change-over-delay (COD) of 2 sec was used to prevent adventiti ous reinforcement of switching keys. The resulting schedule was a concurrent FI 30-sec RR 60 schedule. The houselight was illuminated throughout the experimental session a nd terminated only during the ITI. Sessions lasted a minimum of 60 minutes and a maximum of 80 minutes. If 50 food deliveries had been earned on the FI schedule and 60 minutes had passed the session terminated.
69 After responding stabilized on bot h the FI and the RR schedules PI trials were added. One out of every five FI trials, random ly chosen, became an extinction trial that lasted for 90 sec. Only sessions in which subjects completed at least 40 FI trials and 10 PI trials were analyzed from this point forward in training. The RR sc hedule value was adjusted for some subjects so that an assessable level of re sponding on the PI schedule occurred. These manipulations resulted in Subjects 453, 739, and 930 being on a concu rrent PI 30-sec RR 60 schedule and Subjects 551 and 739 being on a concurrent PI 30-sec RR 90 schedule. Drug administration D-amphetamine was administered in a decreasing acute fashion using saline, 0.3, 1.0, 1.7, 2.25 and 3.0 mg/kg doses twice weekly. Subjects were first administered saline then received the 3.0 mg/kg dose on the subsequent acute day. Other doses followed in a decreasing order then the order was repeated. Subjects also experi enced a pre-feeding session on the acute day following the lowest dose of drug. Subjects ha d 15 min access to 60 grams of food prior to the start of an experimental se ssion or programmed delay for th ese sessions. At least two administrations of each dose and pre-session f eeding were examined for each delay. All intervening non-drug/pre-session feeding days served as control. Supplemental dosing for inconsistent drug effects was given as needed. This acute dosing regimen was repeated across four different delays between drug administration and behavioral tes ting. The delays examined were 0, 30, 60, and 90 minutes. The orders of exposure to delays were counterbalanced across subj ects. Subjects 453 and 930 were exposed to the delays in a de scending order and Subjects 135, 551, and 739 experienced the delays in ascending order. The delay was pr ogrammed into the experimental session so that subjects were injected or fed and then placed directly into the experime ntal apparatus. The specified delay then counted down and the sessio n began once the delay had expired. Subjects
70 were only exposed to a new delay if all acute dr ug and pre-feed administrations for the current delay were complete. Subjects were then expos ed to the new delay for a minimum of 10 days without drug or pre-session feed ing. Stability was reassessed before a new series of acute administrations took place. Data analysis Session average parameters. Response rates across each PI trial were averaged across the session and plotted in 3-s bins for each subject. A Gaussian function with an added linear component was then fit to the remaining data (see methods Experiment 1). Peak time of responding (point of maximal response rate) a nd the width (standard deviation) of the distribution of response rates wa s estimated for each session usi ng these fits. Sessions that contained fewer than 10 PI trials were excluded from this analysis. Wait time, the time between the start of a trial and the first response was also recorded for each trial. Rate dependency. The same rate-dependent analysis that was carried out on data from Experiment 1 was used for this experiment. Additional parameters. The number of responses, excluding those during the COD, for each schedule was recorded. The durations spent on both the PI and RR schedules were documented, as was the number of switches from one schedule to the other. Inter-response times on the RR schedule were also recorded. A two-way ANOVA with within subjects factors dose (pre-s ession feeding was included here) and delay was run on the entire group and fo r individual subjects for each parameter. An alpha level of .05 was used in all statistical tests.
71 Results Peak Interval Distributions The most consistent finding across all s ubjects, drug doses, pre-session feeding and delays was that d -amphetamine tended to increase low rate s of responding early and later in the trial and decreased high rates. This effect resu lted in response-rate di stributions becoming wider and flatter. Figure 3-1 shows the average response ra tes per second during peak trials for all subjects for each dose at each delay. When no de lay was in place (top panel) the average effect was to increase the low rates of responding that occur early and late in the interval under higher doses of d -amphetamine and pre-session feeding, resul ting in wider distribu tions. Low doses of drug had less of an effect. A leftward shift in the distribution can be seen under the 1.7 and 2.25 mg/kg dose. Following a 30-min delay low response rates increased under all doses of drug and pre-session feeding. A leftward shift in the location of the peak rate of responding can be seen for the 1.0, 1.7, and 2.25 mg/kg doses. There is less of an increase in low rates prior to the typical time of reinforcement under the 60-min de lay condition across most doses. Increases in low rates after the typical type of reinforcement can be seen for mo st doses of drug. Pre-session feeding had less of an effect ove rall at this delay. Following a 90-min delay, the 1.7 and 2.25 mg/kg doses of d -amphetamine had the largest impact on th e PI distribution. Low response rates throughout the distribution were increased under these doses resulting in a widening of the distribution. The location of the peak time of responding appears to have shifted leftwards under the 2.25 mg/kg dose. Other doses of drug increas ed low rates of responding to a lesser degree and decreased high rates of respondi ng at about the time of typical reinforcement. Similar rate decreasing effects can be seen across all delays under pre-sess ion feeding and the 3.0 mg/kg dose.
72 Figure 3-2 shows the average re sponse rates per second during peak trials for individual subjects for each dose under each delay. Effects of d -amphetamine on these distributions varied by dose and delay. After the no delay condition (top row), Subject 135s PI distributions (1st column) showed very little change from saline. There is some depression in response rate with increasing d -amphetamine and there is evidence that lo w response rates early in the interval are increasing under drug. A wideni ng and flattening of the PI distributions are apparent for Subjects 453 (2nd column), 593 (3rd column), and 930 (5th column) under this delay. Low response rates have increased throughout the distribution under higher do ses of drug and presession feeding for Subject 739 (4th column). This resulted in the PI distributions becoming wider and flatter throughout the interval. A leftward shift in the peak time of responding is apparent for this subject under most dose of d -amphetamine and pre-session feeding. Following the 30-min delay (2nd row) d -amphetamine decreased response rates around the location of peak responding for Subj ect 135, resulting in a flattening of the curve. Low rates of responding throughout the interval increased un der drug and pre-session feeding for Subjects 453, 551, and 739 resulting in a widening of the dist ribution. Leftward shif ts in the location of peak time can be seen under the 3.0 mg/kg dose for Subject 551 and under all doses and presession feeding for Subject 739. Th ere is little disruption of S ubject 930s PI distribution under most doses of drug. The 3.0 mg/kg dose and presession feeding depressed response rate for this subject. An increase in responding throughout the interval, fo llowing the 60-min delay (3rd row), relative to saline can be seen under most doses of drug for Subject 135. There are decreases in the response rate at th e peak time of responding for all doses of drug and pre-session feeding for Subjects 453 and 551, under low doses of drug and pre-session feeding for Subject 739 and
73 under higher doses of drug and pre-session feeding for Subject 930. A rightwards shift in the location of peak time can be seen under pre-sessio n feeding for Subject 551. Leftward shifts in the distribution under most doses of drug can be seen for Subjects 739 and 930. Under the 90-min delay condition (bottom row), there is little effect of drug on the PI distribution for Subjects 135 a nd 930. Pre-session feeding however decreased responding throughout the distribution. Ther e was a widening in the PI di stribution under most doses of drug for Subject 453. Only the highest dose and pre-session feeding resulte d in a depression of rates around the location of peak time for this subject. Increases in response rates under drug and pre-session feeding can be seen for Subjects 551 and 739, resulting in a wider distribution. Gaussian Derived Parameters and Estimates on Session Average Peak Trials Figure 3-3 shows the average change in each Gaussian derived parameter (peak time and s.d.) and wait time as a percent of saline across all doses of d -amphetamine and pre-session feeding. Parameters derived from fitting a Gau ssian distribution with an added linear component to the PI distributions will be discussed first. Figure 3-4 shows these parameters across delays for individual subjects. Tables 3-1, 3-2, and 3-3 present the F-st atistic and p-value for significant effects across the group average a nd individual subjects for factor s of delay, dose and delay by dose interaction respectively. If no value is listed then no significant effects were found for the group or individual subjects. As seen in Table 3-2 there was a significan t main effect on all measure of dose for the group average. A decreasing trend in peak time with increasing drug dose when subjects were averaged together can be seen in Figure 3-3 acro ss all delays. There was no main effect of delay or delay by dose interaction for the group. A significant main effect of delay (Table 31) was found for Subject 135 but no main effect of dose (Table 3-2) or a delay by dose interaction (Table 3-3). There was no change in peak time
74 across doses of drug, pre-session feeding, and de lay for this subject following all delays except the 0-min delay. There was an increasing trend in peak time with increasing drug doses (excluding the 2.2.5 mg/kg dose) when no delay was in place. There was little effect of dose on peak time for Subject 453 across most delays. There were observed decreases in peak time under hi gher doses of drug under the 30-min delay. A significant main effect of delay was found for peak time. No other significant effects were found. Pre-session feeding had no effect. A significant main effect of delay and dos e was found for peak time for Subject 739. There was a separation of drug effects be tween delays seen at lower doses of d -amphetamine and pre-session feeding. Peak time was unaffected at low doses following the 30and 90-min delay but decreased at higher doses under both delays. Peak time decreased un der all doses of drug following the 0and 60-min delay. Pre-session feeding decreased peak time following the 0-, 30-, and 60-min delays but had no effect following the 90-min delay. No significant main effects on peak time were found for S ubjects 551 and 930. There were variable effects on peak time observed for these subjects across delays and doses. A seen in Figure 3-3, there was an increase in s.d. with increasing drug dose and presession feeding across all delays resulting in a significant main effect of dose for the group average. A significant main effect of dose on s.d. for Subject 551 was found. An increase in s.d. with increasing drug dose under all delays can be s een in Figure 3-4. The magnitude of these increases varied across dose for each delay. Pre-session feeding increased s.d. for this subject under the 30and 60-min delays. In general as dose increased the width of the distribution increased for this subject.
75 There was an increasing trend with increasing drug dose in s.d. observed for Subject 739 at all delays except the 60-min delay. These increases were greatest during the no delay condition. There was no change in s.d. observed under the 60min delay. Pre-session feeding increased s.d. under the 0and 90-min delay and decreased s. d. under the 30-min delay. A significant interaction between delay and dos e was found for Subject 739. Low doses of d -amphetamine decreased s.d. during the 0and 90-min delay condition for Subject 930. As dose increased s.d. also increased from saline unde r all delays exce pt at the 3.0 mg/kg dose. At this dose s.d. decreased followi ng the 30-, 60and 90-min delays relative to the 2.2 mg/kg dose. A significant main effect of delay and dose and a si gnificant interaction between delay and dose on s.d. were found for Subject 930. No significant main effects or interactions on s.d. were found for Subjects 135 and 453. A decrease in wait time with increasing drug dose for the average of all subjects can be seen in Figure 3-3. This corresponds to subjec ts initiating responding earli er in the trial than under saline conditions. Under all delays wait time decreased on average for the group. There was a significant main effect of delay and dose on wait time for the group average. There was some separation of drug effects at the higher doses of d -amphetamine with the effect following the 90-min delay being the greatest (50% decr ease under the 3.0 mg/kg dos e). There was little effect of pre-session feeding on wait time fo r the average of the group across delays. There was a decreasing trend in wait time with increasing drug dose observed across delays for Subjects 135 and 739. Increases in wait time were observed following pre-session feeding under the no delay condition for Subject Pre-session feeding decreased wait time following all delays for Subject 739. A signifi cant main effect of dose was found for these subjects.
76 Wait time decreased with increasing dose up to the 2.25 mg/kg dose for Subject 930 under the 30-, 60-, and 90-min delays. At the 3.0 mg/k g dose wait time increased to saline levels following the 60-min delay. Pre-session feeding had little effect on wait time. A significant main effect of dose was found (F = 4.33, d.f. = 6, 6, p < .05). No significant main effects or interactions on wait time were found for Subjects 453 and 551. Additional Parameters Additional parameters derived from this e xperiment for the group average (Figure 3-3) and for individual subjects (Figure 3-5) are pr esented as a percent of saline. There was a decrease in the number of responses made on th e RR schedule with drug and pre-session feeding on average across subjects and dela ys. A significant main effect of dose (Table 3-2) was found for the group. An overall decrease in the num ber of responses under the RR schedule was seen across delays for Subjects 453, 739, and 930. Higher doses of d -amphetamine completely eliminated responding on this schedule after the 60and 90-min delay a nd pre-session feeding reduced responding under this schedule across delays for Subject 453. This resulted in a significant main effect of delay (Table 3-1) for Subject 453. A significant main effect of dose was found for Subjects 453, 739, and 930. This is most likely due to the decreasing effect pre-session feeding had on behavior across all delays for these subjects. A significan t interaction between delay and dose (Table 3-3) was found for Subjects 453 and 930. The interaction was most likely caused by the discrepancy in drug effect observed under th e some delay condition relative to others. D-amphetamine did not alter the number of responses made on the RR schedule for subject 551 across the three shortest delays. There was a decrease in the number of responses following
77 the 90-min delay at the higher doses of drug. Pre-session feeding decreased the number of responses across delays. There was a significant main effect of dose found for this subject. No significant main effects on the number of responses observed under the RR schedule were found for Subject 135. The number of responses observed under th e PI schedule increased on average with increasing drug dose across delays indicating an increased preference for this schedule. Presession feeding had less of an eff ect. There was a significant main effect of dose for the group. A decreasing trend was observed for the num ber of responses on the PI schedule for Subject 135 under the 0and 30-min delays. Th e opposite was observed for the other delays. Pre-session feeding decreased this parameter rega rdless of delay. A significant main effect of delay and a significant interaction between de lay and dose was found for this subject. Subject 453s data showed an increase in the number of responses made under the PI schedule following the 0and 90-min delays. Th ere was no change in this parameter following the 30and 60-min delay. Pre-session feeding d ecreased this parameter after the 60and 90-min delays and increased it after the 0and 30-min delay. A significant main effect of delay and dose and significant interaction betw een delay and dose was found for this parameter for Subject 453. Subject 739s data showed an increasing tre nd in the number of responses made under the PI schedule with increasing drug dose across delays. Pre-session feeding increased this parameter under the 0and 30-min delays. The ma gnitude of these increases varied across delay which resulted in a significant main effect of delay. A significant main effect of dose was also found due to the overall increasing trend observed. There was also a si gnificant interaction between delay and dose.
78 Drug increased the number of responses made under the PI schedul e for Subject 930 under the 30-, 60-, and 90-min delays except under the 3. 0 mg/kg dose. Pre-session feeding had little effect on this parameter across delay. The overa ll increase in the number of responses on the PI schedule with increasing drug dose led to a signif icant main effect of dose. A significant interaction between delay a nd dose was also found. No significant main effects on the number of responses observed under the PI schedule were found for Subject 551. The total number of switches made between the two schedules is shown in the third row of Figure 3-5. Pre-session feeding and drug sign ificantly decreased switching for the average of all subjects across all delays. The effect of drug observed under the no delay condition was not as dramatic as that seen for the other delays which resulted in a significant main effect of delay. There was also a significant inte raction between delay and dose. There was a decrease in total switches made across drug and pre-session feeding under all delays for Subjects 453, 739, and 930. A signifi cant main effect of delay found for Subject 453 reflected that switching was completely eliminat ed for doses above 1.0 mg/kg under the 60and 90-min delay but not under the other delays. A significant main effect of dose was found for Subjects 453, 739, and 930. These effects were due to the overall decrease observed in total switching across drug, pre-session feeding and delays. A signifi cant interaction between delay and dose was found for Subjects 453 and 930. Pre-session feeding and d -amphetamine decreased switchi ng behavior for Subject 551 across all delays. This decrease was largest following the 90-min delay. A significant main effect of dose was found for switchi ng between the two schedules. No significant main effects on switching were found for Subject 135.
79 There was an increase in the duration spen t on the PI schedule with increasing drug dose and pre-session feeding for the average of all subj ects. There was a more potent effect of the 1.7 and 2.25 mg/kg dose following the 90-min delay than for the other delays. A main effect of dose and a significant interact ion between delay and dose was found for the group. Pre-session feeding and drug administration resulted in an increase in the duration of total time spent on the PI schedule across delays fo r Subjects 453, 551, 739, and 930. The increases in this parameter across all doses resulted in a significant main effect of dose for Subjects 453, 551, 739 and 930. A significant main effect of delay was found for Subject 453. There was a larger increasing effect of drug and pre-session feeding following the 60and 90-min delays for this subject. Similar separation between drug e ffects across delays led to a significant main effect of delay for Subject 739. A significant interaction between delay and dose was found for Subject 930. There were different magnitudes of drug effect across each delay that contributed to this effect for this subject. There were no significant effects found for Subject 135. Duration of time spent on the RR schedule decr eased with increasing drug doses and presession feeding for the average of all subjects ac ross all delays resulting in a significant main effect of dose. Differences in magnitude of eff ects seen across delays resulted in a significant main effect of delay. A significant intera ction between delay and dose was also found. Subjects 453, 551, 739 and 930 spent less tim e on the RR schedule following pre-session feeding and drug administration ac ross all delays. A significant main effect of delay was found for Subject 453. Behavior on the RR schedule wa s completely eliminated for the 1.0 mg/kg dose and all higher doses of d -amphetamine following the 60and 90-min delay conditions for this subject. Overall decreases with increasing drug doses across delays led to a significant main effect of dose for Subjects 453, 551, 739, and 930. A significant interaction between delay and
80 dose was found for Subjects 453 and 930. This is mostly likely due to the separation of drug effects observed across some delays. There was little effect of drug and pre-session feeding on Subject 135s data across all dela ys. No significant main effects were found for this subject. Rate Dependency Figure 3-6 shows scatter plots of the average for each drug dose against saline on log axes for the entire PI distribution for each delay. Ea ch plot was fit with a separate linear function (Equation 2) from which the slope and intercept we re derived (Figure 3-7). As can be seen in Figure 3-6 low rates were increas ed under drug and pre-session feed ing across all delays for all subjects. High rates were left unchanged mostly with some exceptions were decreases in high rates can be seen (Subjects 551 and 930). Figure 3-7 plots the intercepts and slopes derive d from this analysis. Intercept increased with increasing drug dose for the average of all subjects and for each subject individually across all delays. Pre-session feeding also resulted in an increase in inte rcept. A significant main effect of dose was found for the group and for all subjects This was a result of the overall increasing effects drug and pre-session feed ing had on intercept across dela ys. Some separation of drug effects on intercept seen across delays led to a significant main effect of delay for Subject 739. A significant interaction between dose and delay was found for intercept for Subject 453. Slope decreased with increasing doses of d -amphetamine and pre-session feeding across all delays for both the average and for individual subjects. This ove rall increase following drug and pre-session feeding resulted in a significant main effect of dose for th e group and all subjects except Subject 135. A significant main effect of delay was found for Subject 135. This was due to differences in drug effect on slope seen acro ss delays. A significant in teraction between delay and dose was found for Subject 551.
81 Discussion BET (Killeen and Fetterman, 1988) stated that mediating behaviors may develop in the course of a subject learning to time. As ou tlined in the general introduction these mediating behaviors do not control timing behavior but may assist in timing. In a previous study Fetterman and colleagues (1998) found that changes in these mediating behaviors predicted changes in timing. The present experiment concurrently prog rammed an alternative task in order to see if a reliable, measurable collateral behavior would de velop and if changes in this behavior would predict changes in timing. Decr eases in peak time and wait time measures of temporal accuracy, and increases in s.d., a measure of temporal control, were accompani ed by decreases in responding on the RR schedule as well as decrea ses in the time spent on this schedule for the average of all subjects (Figure 3-3). The largest decreases in RR responding were found under the same conditions where the largest disrupt ions in peak time, s.d., and wa it time were obser ved. Similar changes can be seen at the individu al subject level as well. This is the pattern of results predicted if responding on the concurrent schedule assists the temporal control on the PI schedule. Subjects responded less on the RR schedule unde r drug and pre-session feeding and thus started responding sooner and longer on the PI schedule. A decline in accuracy on the timing task corresponded to decreases in responding under the RR schedule. One possible contradiction to this theory is that even when behavior was completely eliminated under the RR schedule, timing behavior remained somewhat intact under the PI schedule. It is possible that some other behavior not measured here came to assist in timing throughout experiment and that this behavior was not completely eliminated. According to BET, changes in arousal or rein forcement value directly alter the pacemaker. If the reinforcement rate is altered through extincti on, or the value of an upcoming reinforcer is altered through pre-session feeding, then timing behavior should become less precise. In this
82 experiment subjects were pre-fed mixed grain prior to sessions under each delay. This manipulation had little effect on peak time, increased s.d. for some subjects, and had a variable effect on wait time across subject s. Pre-session feeding affected th e other parameters in the same direction as d -amphetamine. The effect of pre-session fe eding on temporally controlled behavior was minimal with the exception of s.d. but seemed to alter non-temporally controlled aspects of behavior more. Rate dependency (Dews, 1958, 1970) has been used to characterize the effects of amphetamine on timing behavior (Knealing and Sc haal, 2002; Odum et al., 2002; Saulsgiver et al., 2006). The same analysis was applied to data in this experiment. Pre-session feeding and increasing doses of d -amphetamine increased intercept and decreased slope across subjects and delays. This corresponded to increases in low rate s of responding and decreases in high rates of responding. Thus subjects we re initiating responding on the timin g task earlier under drug and presession feeding and responding l onger across the interval as co mpared to saline and non-presession feeding conditions. The delay between behavioral testing and dr ug or pre-session feeding altered wait time, switching behavior, and the duration spent on th e RR schedule differently. The evidence that drug and pre-session feeding disrup ted behavior differently across de lays is an important finding. Cevik (2003) found that the delay between drug in jection and behavioral testing impacted the types of changes observed in a temporal bisec tion task. There is evidence in the present experiment that the drug alters the initiation to res pond differently across delays. It was shown in Figure 1-1 that when low rate s of responding early in the interv al increase, as is seen when wait time decreases, this change al one is enough to cause shifts in peak time. If the drug affects timing behavior in significantly different ways across a time period then averaging the data
83 together could lead to conclusions that are not en tirely accurate, i. e. a large decrease in wait time during the first half hour but not the last could sh ift the distribution across a 1-hour session. The delay between behavioral testi ng and drug or pre-session feedi ng did not lead to significant effects for the group on other parameters. Howeve r, a difference between delay conditions is evident when individual subjects were examined Many times there were differences in the size of the drug effect (i.e., Peak time for Subject 551; s.d. for Subjects 551 and 930; wait time for Subjects 551 and 930; Responding under the RR schedule for all subjects; Responding under the PI schedule for Subjects 551, 739, and 930; tota l switching for Subjects 453 and 930; duration under both schedule for Subjects 453, 739, and 930) Different delays sometimes resulted in opposite directional effects (i.e., peak time for S ubject 135; s.d. for Subjects 135; wait time for Subjects 551; Responding under the PI schedul e for Subjects 135 and 453). Many of these differences across delay were suppo rted by a significant main effect of delay for each subject. While the overall effect of delay on behavior was not as robust as was expected, the data supports the notion that delay to be havioral testing or a variable session length can contribute to various differences in obtained results (Cevik, 2003). More specifically this manipulation identifies that drug effects on parameters iden tifying timing under a PI schedule can vary based on time since injection.
84 Table 3-1. Two-way ANOVA Significant main effects for delay Avg (d.f. = 3,27) 135 (d.f. = 3,3) 453 (d.f. = 3,3) 551 (d.f. = 3,3) 739 (d.f. = 3,3) 930 (d.f. = 3,3) Peak time F = 26.17; p < .05 F = 20.97; p < .05 s.d. F = 15.86; p < .05 Wait time F = 2.73; p < .05 F = 58.45; p < .01 RR # resp F = 15.72; p < .05 PI # resp F = 10.46; p < .05 F = 23.70; p < .05 F = 9.48; p < .05 Switch F = 3.13; p < .05 F = 22.09; p < .05 Duration PI F = 9.55; p < .05 F = 9.69; p < .05 Duration RR F = 3.33; p < .05 F = 14.48; p < .05 RR IRT F = 30.51; p < .05 Intercept F = 11.39; p < .05 Slope F = 21.45; p < .05
85 Table 3-2. Two-wa y ANOVA Significant main effects for dose Avg (d.f. = 6,42) 135 (d.f. = 6,6) 453 (d.f. = 6,6) 551 (d.f. = 6,6) 739 (d.f. = 6,6) 930 (d.f. = 6,6) Peak time F = 5.84; p < .01 F = 7.39; p < .05 F = 18.36; p < .01 s.d. F = 5.23; p < .01 F = 9.61; p < .01 F = 10.55; p < .05 Wait time F = 5.91; p < .01 F = 12.88; p < .01 F = 9.22; p < .01 F = 5.63; p < .05 RR # resp F = 16.44; p < .01 F = 29.29; p < .01 F = 15.20; p < .01 F = 30.85; p < .01 F = 30.40; p < .01 PI # resp F = 4.05; p < .01 F = 4.74; p < .05 F = 15.33; p < .01 F = 20.03; p < .01 Switch F = 16.94; p < .01 F = 40.93; p < .01 F = 8.63; p < .01 F = 49.04; p < .01 F = 31.89; p < .01 Duration PI F = 15.79; p < .01 F = 15.20; p < .01 F = 8.05; p < .05 F = 95.08; p < .01 F = 6.47; p < .05 Duration RR F = 16.61; p < .01 F = 51.43; p < .01 F = 8.91; p < .01 F = 35.75; p < .01 F = 39.82; p < .01 RR IRT F = 7.72; p < .01 F = 9.08; p < .01 F = 20.96; p < .01 F = 6.89; p < .05 Intercept F = 16.06; p < .01 F = 14.41; p < .01 F = 89.13; p < .01 F = 20.18; p < .01 F = 45.02; p < .01 F = 9.14; p < .05 Slope F = 25.18; p < .01 F = 11.65; p < .01 F = 6.01; p < .05 F = 17.50; p < .01 F = 34.69; p < .01
86 Table 3-3. Two-way A NOVA Significant interacti on between delay and dose Avg(d.f. = 18,126) 135 (d.f. = 18,18) 453 (d.f. = 18,18) 551 (d.f. = 18,18) 739 (d.f. = 18,18) 930 (d.f. = 18,18) Peak time F = 2.51; p < .05 s.d. F = 2.67; p < .05 F = 4.76; p < .01 Wait time RR # resp F = 3.36; p < .01 F = 3.99; p < .01 PI # resp F = 2.71; p < .05 F = 6.27; p < .01 F = 2.91; p < .05 F = 4.88; p < .01 Switch F = 2.05; p < .01 F = 5.55; p < .01 F = 6.12; p < .01 Duration PI F = 1.88; p < .05 F = 2.42; p < .05 Duration RR F = 1.74; p < .05 F = 4.42; p < .01 F = 5.59; p < .01 RR IRT F = 2.41; p < .01 F = 14.91; p < .01 Intercept F = 4.67; p < .01 Slope F = 2.45; p < .05
87 0-min delay 0.0 0.5 1.0 1.5 2.0 90-min delay 3-sec bins 051015202530 0.0 0.5 1.0 1.5 2.0 30-min delay 0.0 0.5 1.0 1.5 2.0 60-min delay Response rate (sec) 0.0 0.5 1.0 1.5 2.0 Saline 0.3 mg/kg 1.0 mg/kg 1.7 mg/kg 2.25 mg/kg 3.0 mg/kg Pre-feed Average Figure 3-1. The average response rates per 3-s bin during peak trials for the average of all subjects under each dose of damphetamine across each delay. The average across all subjects and delays is shown in the bottom panel. The 0-min delay is shown in the top row, the 30-min delay in the second row, the 60-min delay in the third row, and the 90-min delay shown in the bottom row. The solid line represents the average distribution under saline, the long-dashed line shows th e average distribution under the 0.3 mg/kg dose, the medium-dashed lin e shows the average distribution under the 1.0 mg/kg dose, the short-dashed line show s the average distribution under the 1.7 mg/kg dose, the dotted line shows the averag e distribution under the 2.25 mg/kg dose, the long-dash, single dot line shows the average distribution under the 3.0 mg/kg dose, and the long-dash, double dotted line shows the average di stribution under presession feeding.
88 Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Plot 7 Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Plot 7 Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Plot 7 Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Plot 7 0.0 0.5 1.0 1.5 2.0 2.5 135 3-sec bins 051015202530Response Rate (sec) 0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 2.5 0-min delay 30-min delay 60-min delay 90-min delay Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Plot 7 Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Plot 7 Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Plot 7 Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Plot 7 453 051015202530 Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Plot 7 Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Plot 7 Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Plot 7 Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Plot 7 551 051015202530 Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Plot 7 Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Plot 7 Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Plot 7 Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Plot 7 739 051015202530 Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Plot 7 Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Plot 7 Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Plot 7 Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Plot 7 930 051015202530 Saline 0.3 mg/kg 1.0 mg/kg 1.7 mg/kg 2.25 mg/kg 3.0 mg/kg Pre-feed Figure 3-2. The average response rates per 3-s bi n during peak trials for each subject under each dose of damphetamine across each delay. S ubjects are separated by columns. All other features of the figure are identical to Figure 3-1.
89 SD 50 100 150 200 250 300 Peak Time 50 100 150 Wait time 0 50 100 150 % salineAverage dose (mg/kg) V PF0.311.72.253RR # of responses 0 50 100 150 Total switch 0 50 100 150 PI # of responses 50 100 150 200 V PF0.311.72.253Duration RR 0 50 100 150 Duration PI 50 100 150 200 250 300 0-min delay 30-min delay 60-min delay 90-min delay Figure 3-3. The percent saline changes for the av erage of all subjects across each delay for each parameter. Standard error is s hown by the error bars at each dose.
90 Peak Time 0 30 60 90 120 150 SD 0 100 200 300 400 500 V PF0.311.72.253Wait time 0 50 100 150 200 250 135 dose (mg/kg) V PF0.311.72.253 453 V PF0.311.72.253 551 V PF0.311.72.253 739 930% saline 0-min delay 30-min delay 60-min delay 90-min delay V PF0.311.72.253 Figure 3-4. Peak Time and sta ndard deviation derived from the curve fitting methods described in Methods for the session average distribu tions shown as a percent of saline. Wait time is shown in the bottom row. Standard error is shown by the error bars at each dose.
91 RR # of responses 0 50 100 150 200 250 Total switch 0 30 60 90 120 150 135 453 551 739 930 0-min delay 30-min delay 60-min delay 90-min delay % saline PI # of responses 0 50 100 150 200 250 V PF0.311.72.253Duration RR 0 30 60 90 120 150 Duration PI 0 100 200 300 Dose (mg/kg) V PF0.311.72.253 V PF0.311.72.253 V PF0.311.72.253 V PF0.311.72.253 Figure 3-5. The number of res ponses made under the RR schedule, PI schedule, total number of switches made between schedules, and the duration spent on the RR and PI schedules are shown as a percent of saline. All othe r features of the figure are identical to Figure 3-3.
92 60-min delay 0.01 0.1 1 10 0-min delay 0.01 0.1 1 10 135 90-min delay 0.010.1110 0.01 0.1 1 10 30-min delay Comparison response rates 0.01 0.1 1 10 453 0.010.1110 551 Average control response rates 0.010.1110 739 0.010.1110 930 0.010.1110 0.3 mg/kg 1.0 mg/kg 1.7 mg/kg 2.25 mg/kg 3.0 mg/kg Pre feed Figure 3-6. Log session averag e response rates under drug cond itions as a function of log response rates averaged across all baseline se ssions for the entire trial for each subject under each delay. A line of unit slope indi cates no change in response rates from baseline. The solid line represents the av erage distribution under the 0.3 mg/kg dose, the long-dashed line shows the average distribution under the 1.0 mg/kg dose, the medium-dashed line shows the average di stribution under the 1.7 mg/kg dose, the short-dashed line shows the average distribution under the 2.25 mg/kg dose, the dotted line shows the averag e distribution under the 3.0 mg/kg dose, the long-dash, single dot line shows the average distributi on under pre-session f eeding. All other features of the figure are identical to Figure 3-1.
93 Intercept 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 0-min delay 30-min delay 60-min delay 90-min delay V PF0.311.72.253Slope 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 V PF0.311.72.253 dose (mg/kg) V PF0.311.72.253 V PF0.311.72.253 V PF0.311.72.253 135 453 551 739 930 Figure 3-7. Slope and intercept derived from the scatter plots showing log session average response rates under drug conditions as a function of log response rates averaged across all baseline sessions for the entire interval. Each column shows a different subject. The solid line represents the aver age distribution under the 0-min delay, the long-dashed line shows the average distribution under the 30-min delay, the longdashed line shows the average distribution under the 60-min delay, the dotted line shows the average distribution under 90-min delay.
94 CHAPTER 4 GENERAL DISCUSSION In Experim ents 1 and 2, d -amphetamine altered pigeons be havior under the control of PI schedules. These changes were observed for severa l measures of temporal behavior at the group average level. A decrease in peak time with increasing dr ug dose was observed across both experiments. An increase in s.d. with increasi ng dose of drug was also observed. Subjects also began responding much earlier in each trial when amphetamine was administered. This decrease reflected in start time (Experiment 1) a nd wait time (Experiment 2) corresponded to rate dependent characterizations of stimulant drug effects across indi vidual subjects. In Experiment 1, changes in temporal be havior were assessed by two common methods employed in the literature examining PI schedules, a session-average and single-trial analysis. Parameters derived from each method were compared for directional effects. Table 2-2 presents these comparisons for Experiment 1 and two pub lications that reported the same measures. These analyses did not necessarily reflect the same changes across analysis. The form of individual trials was also examined. It was found that individual trials were peak-like in form under saline and low dose d -amphetamine conditions, suggesting that average PI distributions are reflective of single trial performance. Res ponse distributions became flatter at higher doses reflecting a loss in temporal control. In Experiment 2, various delays between behavioral testing a nd drug administration revealed differences in magnitude and directional effects for indi vidual subjects. When subjects were averaged together these differences across delay vanished. Decreas es in responding on the concurrent task were accompanied by declines in temporal accuracy across subjects. As responding on the RR schedule decreased s.d. in creased and peak time on the PI schedule decreased. Pre-session feeding led to increased variability in time estimation but did not always
95 decrease temporal accuracy across delays. The effects of pre-sess ion feeding on the shape of PI distributions were similar to d -amphetamine effects. Timing Theories SET predicts that administration of dopami ne agonists to subjects responding on a PI schedule will decrease peak time, or temporal accuracy, by 10-15% (Buhusi and Meck, 2002; Meck, 1983, 1986). This decrease is a result of the internal pacemaker speeding up under drug, resulting in the subject res ponding as if more time has pa ssed than actually has, an overestimation in time. In the current experiments this effect was observed. When decreases in peak time were observed for individual subjects or the group average the largest decreases were around 20% relative to saline conditions. SET also predicts th at the ratio of the width of the distribution of peak response rates to the mean interval being timed will be constant (Scalar timing: Buhusi and Meck, 2002; Gibbon, 1977; 1992; Gibbon et al., 1997). If amphetamine increases the clock speed then one would expect a complete leftwa rd shift of the distribution and a proportional decrease in the spread of responding (Buhusi and Meck, 2002; Meck, 1996; Saulsgiver et al., 2006; Taylor et al., 2007). Thus any observed increases in the width of the distribution along with a decrease in peak time contradict this pr ediction of SET. Results from Experiments 1 and 2 showed that the width of the distribution (s .d.) increased under d amphetamine when peak time decreased. This e ffect is true for both the group average (Exp 1: Figure 2-6; Exp 2: 3-3) and for individual subjects (Exp 1: Figure 2-4; Exp 2: Figure 3-4) that showed decreases in peak time. In some cases s.d. increased in the absence of a decrease in peak time. This is not the first experiment to s how this combined effect of dopamine agonists on behavior under a PI schedule, Figure 1-2 illustrates a few of these previous studies (Bayley et al., 1998; Eckerman, et al., 1987; Kraemer, 1997; Maricq et al., 1981; Matell et al., 2006; Saulsgiver et al., 2006; Taylor et al., 2007).
96 When this increase in width is inspected by ex amining the PI distribut ions (Exp 1: Figure 2-1; Exp 2: Figures 3-1, 3-2) it can be seen that this incr ease in width is a result of subjects responding earlier in the trial as well as sustaining responding longer under d -amphetamine than under saline. These observations are supported by the observed d ecreases in start time (Exp 1) and wait time (Exp 2). Experiment 1 showed an increase in stop time on average across subjects (Figure 2-4, 2-5) indicating th at responding continued longer afte r the FI interval had elapsed under drug in comparison to saline conditions. This type of widening is also consistent with rate dependent predictions of stimulant drugs on inte rval responding (Dews, 1958). As was shown in Figures 2-7, 2-8, 3-6, and 3-7 low rates of responding increased under drug across the interval. Rate dependent characterizations have been a pplied to the effects of drug on FI and timing behavior in previous research (Beecher and Jackson, 1976; By rd, 1979; Dews, 1958; Flores and Pellon, 1995; Goudie, 1985; Knealing and Schaal 2002; McMillan and H ealey, 1976; Odum et al., 2002; Saulsgiver et al., 2006; Wuttke, 1970). These reports have found that when psychomotor stimulants are administered to subj ects exposed to FI schedules low response rates tend to increase and high respons e rates are decreased or left una ffected resulting in a flattening of the distribution of responding (Dews, 1958). These rate dependent effects could lead to behavior that is temporally uns tructured under PI schedules by increasing the width of the PI distribution. The predictions of stimulant effects on timing behavior made by rate dependency are more accurate than SET predictions. SET predicts that the rati o of the width of the distribution of peak response rates to the mean time to food will be constant (Buhusi and Meck, 2002; Gibbon, 1977; 1992; Gibbon et al., 1997). If am phetamine increases the clock speed as stated by SET then a complete shift of the distri bution should occur, resu lting in a proportional decrease in the spread of responding (Buhusi and Meck, 2002; Meck, 1996 ; Saulsgiver et al.,
97 2006; Taylor et al., 2007). An examination of the PI distributions under dr ug shows that this is not the case (Exp 1: Figure 21; Exp 2: Figures 3-1, 3-2). It has been suggested that at tentional processes can be aff ected by amphetamine and that this effect of d -amphetamine can result in the loss of stimulus control (Buhusi and Meck, 2002; Odum and Ward, 2007). Using a PI procedure wher e peak trials were oc casionally interrupted by a gap in the stimulus to be timed, Buhusi and Meck (2002) isolated methamphetamines effects on attentional and clock processes in rats. In this series of experiments subjects experienced 30-32 FI 30-s trials, 6-8 uninterrupted PI trials, and 24 gap trials where a PI was interrupted at some predetermined time. Me thamphetamine was found to decrease peak time under normal PI trials but increas ed peak time during gap trials, indicating that that time was resetting. Data from these experiments indicated that the shift in peak time was not proportional to the gap in the interval, as was predicted if the drug only altered th e clock rate. It was concluded that methamphetamine was possibly a ffecting attentional processes such as when timing of an interval began. Experiments in wh ich gap-associated stimuli were made similar to ITI-associated stimuli in the absence of the dr ug showed that subjects behavior reflected a resetting of the clock. Subjects in this situation started timing from the end of the gap rather than from the beginning of the interval to be tim ed (Experiment 2: Buhusi and Meck, 2002). The effect of methamphetamine on the gap in these experiments was similar to procedures that increase the similarity between the gap and the ITI suggesting th at timing was reset based on its attentional or stimulus-controlled effects rather than selectively affecting the pacemaker (Buhusi and Meck, 2002). From these results it has been suggested that the eff ect of methamphetamine seen in Buhusi and Mecks experiment might be a sign of a change in the salience of the gap (Buhusi and Meck, 2002; Gray et al., 1997). A sim ilar attentional explanation has been applied
98 to results investigating d -amphetamines effects on a temporal bisection task (Odum and Ward, 2007). Odum and Ward (2007) administered d -amphetamine to pigeons exposed to two different temporal bisection tasks. D-amphetamine flattened the obtained psychophysical function indicating a loss in stim ulus control. If d -amphetamine disrupts attention to the stimulus then subjects may not attend to the stimulus on some trials resulting in chance performance on those non-attended trials. This disruption could account for the observed flattening of the psychophysical function (Odum and Ward, 2007). Buhusi and Mecks results could also be explained by the flickering switch hypothesis (Lejeune, 1998). This theory postulates that a gate that either blocks pulses from entering or al lows pulses to pass through to the accumulator is altered by dopamine agonists. This gate or sw itch can be opened or closed when stimuli are interrupted or at the beginning of an interval to be timed. Change s in the operation of this gate could reflect changes in attention to the stimul us. If methamphetamine increases the rate at which the attentional switch flickers it could re sult in a resetting of the clock in these gap procedures (Buhusi and Meck, 2002; Lejeune, 1998). These explanations could apply to the current experiments. If d -amphetamine administration decreased attention to the stimulus on some trials on a PI schedule subjects could respond randomly in time resulting in a widening of the distribution. If d -amphetamine made the stimulus to be timed less salient, subjects ma y begin timing during the ITI and responding would start earlier due to an early st art in the clock on some trials. Similarly, responding could have been prolonged by a delay in the start of the cloc k on some trials. The attentional altering effects of d -amphetamine could have resulted in an increase d flickering rate of the gate where the gate either allowed pulses to accumulate earlier or postponed the accumulation of pulses differently across trials. The averaging of PI trials or data derived from single trials separately under either
99 explanation could also account for the widening of the distribution under d -amphetamine. What these explanation do not address however, is the f act that some studies have shown a selective effect on timing using dopamine agonists (Cheng et al., 2007; Eckerman et al., 1987; Frederick and Allen, 1996; Hinton and Meck, 1996; Matell et al., 2006; Meck, 1996; Meck and Williams, 1997) and other data indicate that drug disrupts behavior more cons istently with an attentional account (Bayley et al., 1998; Kraemer, 1997; Maricq et al., 1981; Odum and Ward, 2002, 2007; Saulsgiver et al., 2006). Future research should be aimed at understanding under what conditions dopamine agonists alter timing behavior and under what conditions attention is affected. BET offers an alternative to SETs accounts of temporal disruption. BET has postulated that mediating behavior may devel op as a subject learns to time a nd that this behavior may come to assist in timing (Killeen and Fetterman, 1988). As an interval passes a subject will engage in other behavior and if arousal is sufficient then conditioning will take place. Each behavior the subject engages in will be conditioned to the passage of a certain amount of time. These behaviors as they occur will then come to serve as discriminative stimuli for the passage of time (Fetterman et al., 1998; Killeen and Fetterman, 1988). The controlling mechanism behind timing in this theory is a pacemaker where the rate of em ission of pulses is directly tied to the rate of reinforcement. The pulses from this pacemaker drive the subject thr ough different behavioral states indicating the passage of time. As the rate or the value of reinforcement drops the pacemaker will also slow, emitting fewer pulses acro ss some interval of time. Under these low states of arousal the subject would also engage in less mediating behavior. When fewer pulses and mediating behavior occur, there is less of an opportunity for conditioning to take place.
100 Consequently, the behavior is less likely to serve as a discriminative stimulus for the passage of time (Killeen and Fetterman, 1988). Experiment 2 evaluated changes in timing accuracy in relation to changes in a concurrent task while exposed to d -amphetamine and pre-session feedin g. Under drug conditions it was found that a reduction in temporal accuracy (p eak time) accompanied decreases in responding on the RR schedule. The largest reduc tions in peak time occurred when there were large reductions in RR responding across subjects. Even when responding on the RR schedule was completely eliminated however, there was still evidence of te mporal organization in th e data. It is possible that in these cases other unmeasured mediati ng behavior was present. While this is not conclusive evidence that mediati ng behaviors assist in timing this data does show that when drug reduces behavior that is occurring concurrently with a timing task timing accuracy decreases as well. Pre-session feeding was used in Experiment 2 to directly affect the value of reinforcement. Several methods of reducing arousal or the value of reinforcement during timing studies have been used including pre-session feeding, extinction, and intertrial interval feeding (Killeen et al, 1999; Ward and Odum, 2007). Extinction or rem oval of food reward was used by Killeen and colleagues (1999) in assessing predictions made by BET. When pigeons were exposed to periods of extinction, timing accuracy diminished in a way predicted by BET. It was found that when the rate of reinforcement was reduced through extinction, the ra te of the pacemaker decreased, consistent with BET. Mediating beha viors however, were not ex amined in this study. In Experiment 2, pre-session feeding did decr ease responding on the RR schedule across subjects and reduced temporal accuracy on the PI schedule as predicted by BET. Pre-session feeding was more likely to result in an increase in s.d. than in a reduction of peak time for this experiment. PI
101 distributions were altered simila rly under pre-session feeding and d -amphetamine conditions. Low response rates across the in terval increased under pre-se ssion feeding and high response rates were reduced, resulting in a widening and fl attening of the distribution. These combined effects of an increase in s.d. and a disruption of the shape of PI distributi ons are more consistent with an account that implicates the loss of stimul us control as the main effect of pre-session feeding on timing behavior. Ward and Odum (2007) came to a similar conclusion when examining the effects of pre-se ssion feeding, extinction, and in tertrial interval feeding on a temporal bisection task. These manipulations were shown to reduce tim ing accuracy by causing a choose-short effect. These results were interpreted as indicating an ov erall loss of stimulus control rather than selective effect on timing. Results from Experiment 2 support this account. One proposed alternative to SET and BET is that dopamine agonists disrupt stimulus control in these experiments rather than havi ng a selective effect on timing (Odum and Ward, 2002, 2007). This proposed explanation is that dopamine agonists like amphetamine disrupt behavior in a way that is not specific to tem poral control. Instead these drugs disrupt all behavior associated with any schedule. A gustn-Pavn and colleagues (2007) showed that unconditioned preference for the male sex pheromones in female mice was reduced by amphetamine. Other choice experiments have found that preference for the more immediate reinforcer was also reduced by amphetamine (W ei-Min et al., 2008). Amphetamine has also been shown to reduce responding accuracy and increase response omission under a serial reaction time task (Fletcher et al., 2007). Laties (1972, 1975; et al., 1981) found that d amphetamine disrupted accuracy when there was no external discriminative stimulus present. Accuracy in responding under a fixed consecuti ve number schedule (a specified number of responses had to be made on one key before reinforcement became available on a second key)
102 was disrupted when no external cue signaling the co mpletion of the first component was present. If amphetamine has been shown to disrupt control of behavior re gardless of the schedule in place then it can be postulated that the same e ffect is occurring under timing schedules. S.d. or the width of the distribution has been taken as a measure of overall stimulus control exerted by the PI schedule (Saulsgiver et al, 2006). As was seen fo r the average of all subjects in both the present experiments this parameter increas ed with increasing drug dose. When subjects were examined individually it was found that de creases in peak time were often accompanied by increases in s.d. The loss of overall stimulus control is further supported by the ziggurat analysis conducted on individual tr ials for Experiment 1. Figure 2-6 showed that the form of individual trials was less peak-like under increasing doses of drug. Response distributions for individual trials became flatter and responding wa s more uniform under higher doses of damphetamine. Pretreatment with saline and low doses of d -amphetamine left behavior less disrupted and gave the impression that behavior was temporally organized. Higher doses of drug resulted in responding that occurred randomly in time rather than in some relation to the passage of time. A similar widening of the average PI distributio ns under high doses of drug for Experiment 2 supports this impression as well. These results to gether indicate that th e contingencies in place were not exerting as much control over behavior when drug was administered. Procedural Variations Behavior should be examined at the level in which order is f ound (Skinner, 1935) but when order is found at multiple levels which should be used? One may argue that if a molecular analysis provides a more complete description of behavior while mainta ining order then that analysis is superior (Branch and Gollub, 1974). Two common methods exist for examining behavior under a PI schedule. A session average analysis averages data across multiple peak trials and then fits a Gaussian function to deri ve estimates of temporal accuracy (peak time) and
103 error (s.d.) (Eckerman et al., 1987; Kraemer et al ., 1997; Matell et al., 2006; Maricq et al., 1981; Meck, 1983, 1996; Taylor et al., 2007). Use of th ese aggregates and curve fitting techniques to derive particular parameters ha s been based on the no tion that the average is representative of single trials; however fluctuations in the averag e of many trials may not reflect changes observed for single trials. This has le d to the single-trial analysis developed by Cheng and Westwood (1993). This analysis fits an algorithm to single trials and derives measures that indicate when responding starts (start time) and stops (stop tim e) for each trial. Parameters of temporal accuracy (middle) and the spread of responding (run length) are then calculated from these measures. Parameters are then averaged across a session. Both methods involve some averaging of data but the order in which that is done varies. Assessments done with both analyses are not always consistent within an expe riment (Matell et al., 2006; Taylor et al., 2007). In Experiment 1, differences between parameters derived by both analyses that are taken to represent similar features of timing were inconsistent. Peak tim e and middle are both taken to represent temporal accuracy (Taylor et al., 2007). Amphetamine, howev er, affected these parameters differently. Figures 2-3, 2-4, and 2-5 show that drug decreased peak time across increasing drug dose but middle was unaffected. This is not the first report to show inconsistent effects across parameters from each analysis. Table 2-2 shows a summary of results of both the single trial analysis and session average analysis from Experiment 1 a nd two previous studies using amphetamine to examine timing behavior under a PI schedule. As seen in Table 2-2 the effect of amphetamine on peak time and middle is not consistent for E xperiment 1 and that of Matell and colleagues (2006). Using a PI 30-sec procedure (PI trials 90110 sec) with rats Matell et al. examined the effects of methamphetamine on parameters deri ved from a single trial analysis and a session average analysis. A significant decrease in pe ak time was observed at some higher doses but
104 there were no significant effects observed for the middle parameter. If both the peak time and middle are to represent the subjective perception of time one would expect a similar effect to be observed across these parameters (Taylor et al., 2007). In another experiment by Taylor and colleagues (2007) the effect of damphetamine on peak time and middle for rats exposed to a PI 24-sec procedure (PI trials 96 sec) was consistent across analysis. Run length from the single trial analysis and s.d. from the aggregate session analysis are both taken to represent the spread of responding on individual trials and for the session average. If the average of peak trials ac ross a session or across subjects is re presentative of what occurs at the level of the individual trial then one woul d expect a consistent effect of drug on these parameters. As seen in Table 22 this was not always the case. Matell et al. (2006) showed an increase in s.d. but a decrease in the run length under drug. Taylor et al (2007) reported no change in s.d. under drug and failed to report run length. However, with a reported decrease in start time and no change in stop time one can assume that run length increased (because run length = stop time start time). The effect of d -amphetamine on s.d. and run length was consistent across analyses for Experiment 1. If these inconsistencies in similar parameters across analyses are present what conclusions should be drawn? Which analysis is the right one to use? The information gained from each analysis differs. Measures of temporal accuracy and error are derived directly with the sessionaverage analysis while similar m easures are based on other derive d measures in the single-trial analysis. Peak time, as a temporal estimate, indicates the time of maximal responding for session-average distributions (Catania, 1970; Me ck, 1996). Middle, on th e other hand, does not indicate where the peak rate of responding occurred. It is not obtained dir ectly from distributions of responding, rather it is calculated by averag ing start and stop time t ogether (Grace et al.,
105 2006). Changes in middle under drug are changes in the average betwee n start and stop time rather than being a measure of temporal accuracy since this measure does not indicate where the peak in responding occurred. When the middle pa rameter is viewed this way there is no reason to assume that drug effects on peak time shoul d be similar to drug effect on middle. The dependency of the middle and run length parameters on other derived parameters then indicates that these measures are not reliable independent measures of accuracy and variance and thus one would not necessarily expect simila r drug effects across parameters. The single trial analysis however does provide information about when subjects begin and stop responding which the session average analysis does not. Ex aminations of when subjects start and stop responding under dopamine agonist s have led researcher s to conclude that separate, non-correlated, change s in the starting and stopping of responding cause the apparent shift in peak time and middle (Sau lsgiver et al., 2006; Taylor et al., 2007). Figure 1-1 simulates how shifts in when subjects begin to respond alone can lead to decreases in peak time. Previous experiments have shown a decrease in peak tim e observed along with a decrease in start time alone (Taylor et al., 2007) or a de crease in wait time alone (S aulsgiver et al., 2006). If this information was not available and on ly the decrease in peak time had been reported, these reports would have been taken to support the notion that amphetamine aff ects the speed of the internal pacemaker thus altering time perception. While th is discussion does not conclude which analysis is more appropriate it does draw attention to the fact that each analysis reveals different aspects of behavior and that each feat ure is important in the analys is of drug effect on timing. Differences in procedures have also been im plicated in causing disp arity between reported data. Variability in the use of PI procedures may contribute to the lack of consistent effects seen across experiments. Cheng et al (2007) investigated how an extensive tr aining history altered
106 drug effects using a PI procedure and showed that differences in length of training resulted in diverse drug effects on behavior under a PI schedule. Dissimilarities in the ratio of PI to FI trials are also common in this area of research (1(FI): 1(PI) ratio: Cheng et al., 2007; Matell and Meck, 1999; Matell et al., 2006; Matell and Portugal, 2007; 4:1: Bayl ey et al., 1998; Cheng and Westwood, 1993; Kraemer et al. 1997; Saulsgiver et al., 2006). Kaiser (2008) showed that acquisition of responding under PI trials was more rapid when the ratio favored FI trials. The ratio of FI to PI trials may modulate drug effects on behavior ma intained by these different PI schedules as well. Further research is needed to explore this possibility. Cevik (2003) demonstrated how differences in the time between injection and behavioral testing can alter drug effects on timing as well. This manipulation was examined for behavior under a PI schedule in Experiment 2. Th e time between pre-session feeding and drug administration and behavioral tes ting had a larger impact on individual subjects data than on the average across all subjects. These differences in effect seen across de lays were sometimes a difference in magnitude of drug effect. For example the 1.7 and 2.25 mg/kg dose of damphetamine increased s.d. after the 0and 90min delays more than under the 30and 60-min delays for Subject 930. Divergences in direction of effects were also seen. For example, peak time decreased under higher doses of drug following the 0-, 60-, and 90-min delays but increased following the 30-min delay. In addition effect s of pre-session feeding and drug were only observed under certain delays and not others such as that seen for Subject 135s peak time. There was an increase in peak time following the 0-min delay for this subject but no changes in peak time were observed following other delays. This multitude of variation in effect on behavior following different dela ys could impact obtained results. Studies have differed in terms of session length when examining drug effects on behavior under a PI sc hedule (Bayley et al.,
107 19981 hour; Frederick and Allen, 1996; Matell et al., 2006; Saulsgiver et al., 20062 hours; Penney et al., 19962 hours 50 min; Maricq et al., 19814 hours; Ot hers use a required number of trials with no concrete session length). When an extended session length is used and data is averaged across the entire session differential e ffects of drug on behavior could be lost. The differences in session length across PI experi ments examining drug eff ects on behavior could contribute to the differences in obtained data. Future research should focus on these differences and investigate how each variab le not only contributes to th e overall occurrence of timing behavior but also how drug affects th is behavior across these variations. Summary Amphetamines effects on timing behavior in the current experiments are best accounted for by explanations that implicate a loss of stim ulus control or a direct effect on attentional processes rather than se lectively effecting timing behavior. Strict interpretations of SET and BET are not well supported by these data. Increases in the width of the PI distributions under d amphetamine across both experiments violate scalar timing and indicate that behavior is less temporally organized under drug. The ziggurat analysis conducted on th e form of individual trials in Experiment 1 also s uggests that under high doses of drug temporal organization is lost. This research adds to the literature that implicates a loss of stimulus control as the main effect of dopamine agonists on behavior under temporal schedules (Kneali ng and Schaal, 2002; Odum et al., 2002; Saulsgiver et al., 2006). A discrepancy in the literatur e examining dopamine agonists effects on timing behavior still needs explaini ng however. Some published reports have shown selective timing effects under dopamine agonists (Cheng et al., 2007; Eckerman et al., 1987; Frederick and Allen, 199 6; Hinton and Meck, 1996; Matell et al., 2006; Meck, 1996; Meck and Williams, 1997) while other data implicates the disr uption of stimulus control or other attentional processes under drug (Bayle y et al., 1998; Kraemer, 1997; Odum and Ward, 2002, 2007;
108 Saulsgiver et al., 2006). The conditions under which these effects occur should be examined in future research. Experiment 2 systematically manipulated the time between behavioral testing and d amphetamine administration. Results showed that different delays can imp act the observed drug effect on temporal and nontemporal behavior. If d -amphetamines effects are averaged over an extended period of time then differences in how be havior is disrupted ma y be masked. However it is not clear that particular effects of dopami ne agonists on behavior under PI schedules should be expected under longer or shorte r sessions. Future research is needed to understand the extent to which session length contribu tes to observed drug effects. Session length and the delay between drug ad ministration and behavioral testing are not the only differences that exist in terms of how PI schedules are implemented. Published experiments vary in the amount of training subjects receive (see Cheng et al., 2007 for review) and the ratio of FI trials to PI trials (1:1 ratio: Cheng et al., 2007; MacDonald and Meck, 2005; Maricq et al., 1981; Matell et al ., 2004, 2006; Taylor et al., 2007; 3:2 ratio: Bayley et al., 1998; Eckerman et al., 1987; Frederick and Allen, 1996; 4:1 ratio: Saulsgiver et al., 2006; 7:1 ratio: Kraemer et al., 1997). The impact of these met hodological differences needs to be examined in order to determine how they later impact changes in timing behavior under drug. Future research should be aimed at exploring these a nd other procedural differences.
109 LIST OF REFERENCES Abner, R. T., Edwards, T., Douglas, A., Brunner, D. 2001. Pharm acology of temporal cognition in two mouse strains. Inte rnational Journal of Compar ative Psychology. 14, 189-210. Bayley, P. J., Bentley, G. D., Dawson, G. R. 1998. The effects of selected antidepressant drugs on timing behaviour in rats. Psychopharmacology, 136, 114-122. Beecher, M. D., Jackson, D. E. 1976. Rate-dependent effect of amphetamine in rats: Extension to between-subjects effect. Psychopharmacologia. 46, 307-309. Bizo, L. A., Chu, J. Y. M., Sanabria, F. 2006. The failure of Weber's law in time perception and production. Behavioural Processes. 71, 201-210. Boisvert, M. J., Sherry, D. F. 2006. Interval timing by an invertebrate, the bumble bee Bombus impatiens. Current Biology. 16, 1636-1640. Branch, M. N., Gollub, L. R. 1974. A detailed analysis of the effects of d -amphetamine on behavior under fixed-interval schedules. Journal of the Experimental Analysis of Behavior. 21, 519-539. Branch, M. N. 1984. Rate dependency, behavior al mechanisms, and behavioral pharmacology. Journal of the Experimental Anal ysis of Behavior. 42, 511-522. Branch, M. N. 2006. How research in behavior al pharmacology informs behavioral science. Journal of the Experimental Anal ysis of Behavior. 85, 407-423. Brown, P. L., Jenkins, H. M. 1968. Auto Shaping of the pigeon's key peck. Journal of the Experimental Analysis of Behavior. 11, 1-8. Buhusi, C. V., Meck, W. H. 2002. Differential effects of methamphetamine and haloperidol on the control of an internal clock. Behavioral Neuroscience. 116, 291-297. Byrd, L. D. 1979. The behavioral effects of co caine: Rate dependency or rate constancy. European Journal of Pharmacology. 56, 355-362. Catania, A. C. 1970. Reinforcement schedules and psychophysical judgm ents: A study of some temporal properties of behavior. In W. N. Schoenfeld Ed., The theory of reinforcement schedules p. 1-42 pp. 1-42. New York: Appleton Century Crofts. Caul, W. F., Brindle, N. A. 2001. Schedule-dependent effects of haloperidol and amphetamine: Multiple-schedule task shows within-subject e ffects. Pharmacology, Biochemistry and Behavior. 68, 53-63. Cevik, M. 2003. Neurogenetics of interval timing. Meck, Warren H. Ed. Functional and neural mechanisms of interval timing. pp. 297-316. Boca Raton, FL, US: CRC Press 2003 xli, 551 pp.
110 Cheng, K., Roberts, W. A. 1991. Three psychophysical principles of timing in pigeons. Learning and Motivation. 22, 112-128. Cheng, R., Hakak, O. L., Meck, W. H. 2007. Habit formation and the loss of control of an internal clock: Inverse relationship between the level of baseline training and the clockspeed enhancing effects of methamphetam ine. Psychopharmacology. 193, 351-362. Cheng, K., Westwood, R. 1993. Analysis of single trials in pigeons' timing performance. Journal of Experimental Psychology: Anim al Behavior Processes. 19, 56-67. Church, R. M., Meck, W. H., Gibbon, J. 1994. Appli cation of scalar timing theory to individual trials. Journal of Experimental Psychology: Animal Behavior Processes, 20, 135-155. Dews, P. B. 1958. Studies on behavior. IV. Stimul ant actions of methamphetamine. Journal of Pharmacology and Experimental Therapeutics, 122, 137-147. Dews, P. B. 1970. The theory of fixed-interval responding. In W. N. Schoenfeld Ed., The Theory of Reinforcement Schedules pp. 43-61. New York, NY: Appleton Century Crofts. Eckerman, D. A., Segbefia, D., Manning, S., Brees e, G. S. 1987. Effects of methylphenidate and d -amphetamine on timing in the rat. Pharm acology, Biochemistry and Behavior, 27, 513515. Fetterman, J. G., Killeen, P. R., Ha ll, S. 1998. Watching the clock. Behavioural Processes. 44, 211-224. Flores, P., Pellon, R. 1995. Rate-dependency hypothesis and the rate-d ecreasing effects of d amphetamine on schedule-induced drinking. Behavioural Pharmacology, 6, 16-23. Foltin, R. W. 2001. Effects of amphetamine, dexfenfluramine, diazepam, and other pharmacological and dietary manipulations on food "seeking" and "taking" behavior in non-human primates. Psychopharmacology. 158, 28-38. Foltin, R. W., Haney, M. 2007. Effects of the cannabinoid antagonist SR141716 rimonabant and d-amphetamine on palatable food and f ood pellet intake in non-human primates. Pharmacology, Biochemistry and Behavior. 86, 766-773. Frederick, D. L., Allen, J. D. 1996. Effects of selective dopamine D-s ub-1 and D-sub-2-agonists and antagonists on timing performance in rats. Pharmacology Biochemistry and Behavior, 53, 759-764. Gibbon, J. 1977. Scalar expectancy theory and Weber's law in animal timing. Psychological Review, 84, 279-325. Gibbon, J. 1992. Ubiquity of scalar timing with a Poisson clock. Journal of Mathematical Psychology. 36, 283-293. Gibbon, J., Church, R. M. 1990. Representation of time. Cognition. 37, 23-54.
111 Gibbon, J. 1991. Origins of scalar timi ng. Learning and Motivation, 22, 3-38. Gibbon, J., Church, R. M. 1984. Sources of varian ce in an information processing theory of timing. In H. L. Roitblat, T. G. Bever, and H. S. Terrace Eds., Animal Cognition. Hillsdale, N.J.: Lawrence Erlbaum Associates. Gibbon, J., Malapani, C., Dale, C. L., Gallistel, C. R. 1997. Toward a neurobiology of temporal cognition: Advances and challenges. Curre nt Opinion in Neurobiology. 7, 170-184. Goudie, A. J. 1985. Comparative effects of cathi none and amphetamine on fixed-interval operant responding: A rate-dependency analysis. Pharmacology Biochemistry and Behavior, 23, 355-365. Grace, R. C., Berg, M. E., Kyonka, E. G. 2006. Choice and timing in concurrent chains: Effects of initial-link duration. Behavi oural Processes. 71, 188-200. Gray, J. A., Buhusi, C. V., Schmajuk, N. 1997. The transition from automatic to controlled processing. Neural Networks. 10, 1257-1268. Henderson, J., Hurly, T. A., Bateson, M., Heal y, S. D. 2006. Timing in free-living rufous hummingbirds, Selasphorus rufus. Current Biology. 16, 512-515. Hinton, S. C., Meck, W. H. 1996. Increasing th e speed of an internal clock: The effects of nicotine on interval timing. Drug Development Research. 38, 204-211. Kaiser, D. H. 2008. The proportion of fixed interv al trials to probe tria ls affects acquisition of the peak procedure fixed interval timing ta sk. Behavioural Processes. 77, 100-108. Katz, J. L. 1988. Effects of drugs on stimulus c ontrol of behavior. III. Analysis of effects of pentobarbital and d-amphetamine. The Journal of Pharmacology and Experimental Therapeutics. 246, 76-83. Killeen, P. R., Fetterman, J. G. 1988. A behavi oral theory of timing. Psychological Review. 95, 274-295. Killeen, P. R., Hall, S., Bizo, L. A. 1999. A clock not wound runs down. Behavioural Processes. 45, 129-139. Knealing, T. W., Schaal, D. W. 2002. Disruption of temporally organize d behavior by morphine. Journal of the Experimental Analysis of Behavior, 77, 157-169. Kraemer, P. J., Randall, C. K., Dose, J. M., Brown, R. W. 1997. Impact of d -amphetamine on temporal estimation in pigeons tested w ith a production procedure. Pharmacology Biochemistry and Behavior, 58, 323-327. Laties, V. G. 1972. The modification of drug effects on behavior by external discriminative stimuli. Journal of Pharmacology and Experimental Therapeutics. 183, 1-13.
112 Laties, V. G. 1975. The role of discriminativ e stimuli in modulating drug action. Federation Proceedings. 34, 1880-1888. Lejeune, H. 1998. Switching or gating? The attentional cha llenge in cognitive models of psychological time. Behaviour al Processes. 44, 127-145. Ludvig, E. A., Conover, K., Shizgal, P. 2007. The effects of reinforcer magnitude on timing in rats. Journal of the Experimental Analysis of Behavior. 87, 201-218. MacDonald, C. J., Meck, W. H. 2005. Differential effects of clozapine and haloperidol on interval timing in the supraseconds ra nge. Psychopharmacology. 182, 232-244. MacEwen, D., Killeen, P. R. 1991. The effect s of rate and amount of reinforcement on the speed of the pacemaker in pigeons' timing be havior. Animal Learning & Behavior. 19, 164-170. Maricq, A., Roberts, S., Church, R. 1981. Meth amphetamine and time estimation. Journal of Experimental Psychology: Animal Behavior Processes, 7, 18-30. Maricq, A. V., Church, R. M. 1983. The differential-effects of haloperidol and methamphetamine on time-estimation in the rat. Psychopharmacology. 79, 10-15. Matell, M. S., Bateson, M., Meck, W. H. 2006. Si ngle-trials analyses demonstrate that increases in clock speed contribute to the methamphe tamine-induced horizontal shifts in peakinterval timing functions. Psychopharmacology. 188, 201-212. Matell, M. S., King, G. R., Meck, W. H. 2004. Differential modulation of clock speed by the administration of intermittent versus continuous cocaine. Beha vioral Neuroscience, 118, 150-156. Matell, M. S., Meck, W. H. 1999. Reinforcementinduced within-trial rese tting of an internal clock. Behavioural Processes. 45, 159-171. Matell, M. S., Meck, W. H. 2000. Neuropsychological mechanisms of interval timing behavior. Bioessays. 22, 94-103. Matell, M. S., Portugal, G. S. 2007. Impulsive responding on the peak-interval procedure. Behavioural Processes. 74, 198-208. Mayorga, A. J., Popke, E. J., Fogle, C. M., Paule, M. G. 2000. Similar effects of amphetamine and methylphenidate on the performance of co mplex operant tasks in rats. Behavioural Brain Research. 109, 59-68. McAuley, F., Leslie, J. C. 1986. Molecular analyses of the effects of d -amphetamine on fixedinterval schedule performance of rats. Journa l of the Experimental Analysis of Behavior, 45, 207-219.
113 McClure, E. A., Saulsgiver, K. A ., Wynne, C. D. L. 2005. Effects of d -amphetamine on temporal discrimination in pigeons. Behavioral Pharmacology, 16, 193-208. McClure, E. A., Saulsgiver, K. A., Wynne, C. D. L. In preparation. Effects of d -amphetamine on two temporal bisection tasks. McMillan, D. E., Healey, M. L. 1976. Some effects of d -amphetamine and pentobarbital on performance under a long fixed-interval schedu le. Journal of the Experimental Analysis of Behavior, 25, 389-399. Meck, W. H. 1983. Selective adjust ment of the speed of internal clock and memory processes. Journal of Experimental Psychology: Anim al Behavior Processes, 9, 171-201. Meck, W. H. 1986. Affinity for the dopamine D2 receptor predicts neuroleptic potency in decreasing the speed of an internal clock. Pharmacology Biochemistry and Behavior. 25, 1185. Meck, W. H. 1996. Neuropharmacology of timing and time perception. Cognitive Brain Research, 3, 227-242. Meck, W. H., Williams, C. L. 1997. Characteriza tion of the facilitative effects of perinatal choline supplementation on timing and temporal memory. Neuroreport: An International Journal for the Rapid Communication of Re search in Neuroscience. 8, 2831-2835. Odum, A. L. 2002. Behavior al pharmacology and timing. Beha vioural Processes. 57, 107-120. Odum, A. L., Lieving, L. M., Sc haal, D. W. 2002. Effects of d -amphetamine in a temporal discrimination procedure: Selec tive changes in timing or rate dependency? Journal of the Experimental Analysis of Behavior, 78, 195-214. Odum, A. L., Ward, R. D. 2007. Characteriz ing the effects of d-amphetamine on temporal discrimination. Behavioural Processes. 75, 156-166. Penney, T.B., Holder, M.D., Meck, W.H. 1996. Clonidine-induced antagonism of norepinephrine modulates the attentional proc esses involved in peak-interval timing. Experimental and Clinical Psychopharmacology. 4, 82-92. Pinkston, J., Fowler, S. 2006. Performan ce Following the Recovery from AmphetamineInduced Focused Stereotypy. Southeastern Association for Behavior Analysis, Greenville, SC. Pinkston, J., Fowler, S. 2008. Evaluation of IRT > t Performance Following the Recovery from Amphetamine-Induced Focused Stereotypy. Association for Behavior Analysis, Chicago, Il. Sanabria, F., Sitomer, M. T., Bizo, L. A., Killeen, P. R. 2005. The peak procedure: A question half-answered. Annual Meeting of the Society for the Quantitative Analysis of Behavior, Chicago, IL.
114 Saulsgiver, K. A., McClure, E. A., Wynne, C. D. L. 2006. Effects of d -amphetamine on the behavior of pigeons exposed to the peak pr ocedure. Behavioural Processes, 71, pp. 268285. Schneider, B. A. 1969. A two-state analysis of fixed-interval responding in the pigeon. Journal of the Experimental Analysis of Behavior. 12, 677-687. Sidman, M. 1955. Technique for assessing the eff ects of drugs on timing behavior. Science. 122, 925Skinner, B. F. 1935. The generic nature of the concepts of stimulus and response. Journal of General Psychology. 12, 40-65. Staddon, J. E., Higa, J. J. 1999. Time a nd memory: Towards a pacemaker-free theory of interval timing. Journal of the Experimental Analysis of Behavior, 71, pp. 215-251. Taylor, K. M., Horvitz, J. C., Balsam, P. D. 20 07. Amphetamine affects the start of responding in the peak interval timing task. Behavioural Processes. 74, 168-175. Ward, R. D., Odum, A. L. 2005. Effects of morphine on temporal discrimination and color matching: General disruption of stimulus cont rol or selective effect s on timing? Journal of the Experimental Analysis of Behavior. 84, 401-415. White, W., Sherrill, L. K., White, I. M. 2007. Time-dependent effects of amphetamine on feeding in rats. Brain Research. 1171, 75-82. Wuttke, W. 1970. The effects of d -amphetamine on schedule-contro lled water licking in the Squirrel monkey. Psychopharmacologia, 17, 70-82.
115 BIOGRAPHICAL SKETCH The interest I had in pursuing psycho logy as a major stemmed from my interest in cetacean behavior. When I was young I was fortunate enough to have had interactions with dolphins and killer whales and these experiences left me fascin ated about these animals. Discouraged with the field of marine biology I had ab andoned my goal to pursue a care er working with these animals until I was able to meet with a dolphin trainer at S ea World. It was here that I first learned of the applicability of psychology to animal behavior. Following this interac tion I returned to the University of Florida and changed my major to psychology. During the next two years I expl ored different areas of psychol ogy. Initially this path led me to social psychology where I assisted in conducting research for almost two years under the supervision of Dr. Dolores Albarracin. Here I implemented questionnaires and conducted experimental research, exploring topics relevant in social psychology including selective exposure, attitudes and social c ognition. While I was interested in these areas of psychology I was unsatisfied with the treatment and interpretati on of the data. The flaw s in the experimental design were also troublesome. These reservations I had about social psychology were brought on by topics discussed in the principles of behavior analysis class I attended in th e spring of my junior year. This class overview ed a field that was quite different from the other psychologies. Following this class I began to participate in re search in both Dr. Marc Branch and Dr. Timothy Hackenbergs laboratories. During this time I was offered an internship conducting research with dolphins at The Living Seas in EPCOT. This internship allowed me to gain first hand experience with the field of animal cognition. I found a sim ilar dissatisfaction with the interpretation of results that I had with interpretations made in social psychology. In many cases these interpretations went beyond
116 the behaviors observed in these dolphins. In addition the way experimental design was approached and altered when problems occurred was also inadequate. One example of this came from problem-solving research that I participated in. During this task dolphins were placed in a situation wher e a barrier, one that they had never experienced, had to be removed in order to ob tain food attached to a towel with a cloth handle. Dolphins had been previously trained to pull th is towel out of a clear plastic box to gain access to food. At the time I was young and had relatively little experien ce in conducting resear ch but I still found it odd that the researchers in charge of this project did not try to account for the previous experience of these animals. The barrier used in this project was a rectangul ar plastic barrier that could slide through the clear plastic box to block access to food. The objective was to see if the dolphins would spontaneously remove this barrier with no training prior to removing the towel. Essentially researchers were investigating in sight behavior. The dolphins used in this experiment had no previous experience with this barrier. Furt hermore there was no resemblance between this barrier and any othe r object found in their environment. When I inquired about the dolphins history with such objec ts trainers could not identify when any object similar to the barrier had been used. Prior to this internship I had the opportunity to watch Epstein and Skinners Columbian simulations that explored the so-called insight learning cases with primates. In this demonstration they illustrate d the importance of a hist ory of experience in the future expression of behavior. While my hypothe sis might be naive I st ill feel that if the dolphins were given experience in so me other context with these barriers or if the barriers were made to resemble familiar objects that they woul d have been successful in this project. This example, coupled with many more, reaffirmed my commitment to behavior analysis for its
117 experimental design and interpreta tions of behavior but also because of the fields commitment to taking a history of experience into account when studying behavior. After this internship I returned to UF and con tinued my work in behavi or analysis. It was then that I made my decision to apply to gradua te school in the behavior analysis field. I was fortunate enough to be accepted to the UF departme nt by Dr. Clive Wynne. In this department my commitment to behavior analysis was born and was developed further.