Behavioural Brain Research 175 (2006) 62–74

Research report

Interaction of raclopride and preparatory interval effects on simple reaction time performance Christopher J. MacDonald, Warren H. Meck ∗ Department of Psychology and Neuroscience, Genome Sciences Research Building II, 3rd Floor, 572 Research Drive, Box 91050, Duke University, Durham, NC 27708, United States Received 12 June 2005; received in revised form 1 August 2006; accepted 2 August 2006 Available online 7 September 2006

Abstract In a series of three experiments, simple reaction time (RT) was characterized with respect to a variable preparatory interval (PI) in order to investigate the relationship between interval timing and RT. In Experiment 1, it was shown that RT decreases as a function of PI and that this effect varies with amount of training. In Experiment 2, RT was shown to increase during probe trials that used a novel 6.25 s PI, suggesting that the specific durations of the PIs encoded during initial training contribute to the PI effect on RT. In Experiment 3, 100 ␮g/kg i.p. of raclopride proportionally slowed RT as a function of the PI. These results are discussed within the context of neuropsychological models of interval timing and support an underlying role for cortico-striatal dopaminergic function in temporal processing and simple RT measurements. © 2006 Elsevier B.V. All rights reserved. Keywords: Dopamine; Raclopride; Cortico-striatal circuits; Preparatory interval; Reaction time; Timing and time perception

1. Introduction The measurement of reaction time (RT) has a long and illustrious history in experimental psychology where it is assumed that information processing takes place during the latency leading up to response initiation [42,52,79,96,100]. In this way, RT is observed to change predictably depending on how the cognitive demands of the task are manipulated, such as attention or memory load [82,94,95]. It was realized early on that anticipatory processes influence RT [39,83,93,100]. Indeed, the stimulus to which the participant responds – the imperative stimulus – always turns-on after a delay period, which is termed the preparatory interval (PI). It stands to reason that an uncertainty arises during a simple RT experiment concerning when the imperative stimulus will be presented [47]. The appreciation of this temporal uncertainty has motivated the design of numerous experiments that sought to characterize RT with respect to the PI [45,47,77,78]. One approach to manipulating temporal uncertainty is to use a procedural design composed of several different PIs, each of which has an equiprobable chance of being

∗

Corresponding author. Tel.: +1 919 660 5765; fax: +1 919 660 5798. E-mail address: [email protected] (W.H. Meck).

0166-4328/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.bbr.2006.08.004

used on a given trial. During these variable PI conditions, a common finding is that RT speeds-up as the PI increases—hereafter referred to as a PI effect. Under this design, the uncertainty of the imperative stimulus’ presentation decreases as time passes, which accounts for the faster RTs [79]. Many theoretical accounts of RT measurements have incorporated some aspect of time perception to explain such systematic regularities during human RT tasks [48–50,81]. In the human literature, RTs are often conceptualized as depending on a diffusion process that takes place during each trial [52,87]. The onset of the imperative stimulus is thought to effect an accumulation of information towards a hypothetical decision boundary. The RT that is recorded reflects the time it takes for the boundary to be crossed after the onset of the imperative stimulus. One may obtain faster RTs if the accumulation process is set in motion prior to the imperative stimulus’ onset—this process is called premature perceptual sampling [49,50]. In this way, when the imperative stimulus is presented, information will have already accumulated toward the decision boundary and the RT will be faster. The onset of perceptual sampling depends on the uncertainty concerning when the imperative stimulus will be presented, which is presumed to be proportional to the interval being timed (i.e., Weber’s law). Therefore, perceptual sampling would take place at relatively earlier times for longer PIs. This

C.J. MacDonald, W.H. Meck / Behavioural Brain Research 175 (2006) 62–74

arrangement might explain why RTs are faster for longer PIs in a set [49,50]. The cognitive processes that mediate the PI effect can be characterized further by extending the analysis to include how it is influenced by neurological dysfunction. RT tasks have been developed for rats and other animals, in part to allow dopamine (DA)-related deficits to be characterized [19,20]. Examples of such deficits include an impairment in acquiring appetitive and aversive responses to a predictive signal [24,28] or the emergence of a contralateral neglect-like effect [60,61]. These RT procedures have also revealed that the PI effect is influenced by DA-manipulations. Systemic injections of DA agonists such as methamphetamine speeds-up RT overall, but enhances the PI effect. Following methamphetamine administration, the degree of RT speed-up observed for the longest PI is relatively greater in comparison to that observed for the shortest PI used in the set [13]. In contrast, systemic administration of raclopride, a relatively specific DA D2 receptor antagonist, has been reported to lengthen RT overall and increases the number of “delayed responses”, the latter of which is used to characterize RTs longer than a given duration [4,7,59]. In addition, unilateral DA lesions of the striatum abolish the PI effect (i.e., flatten RT as a function of PI) for responses initiated contralateral to the side of the lesion, while preserving the PI effect for responses initiated ipsilateral to the side of the lesion [12]. These findings suggest that DA levels contribute to the PI effect, perhaps by altering the animal’s sense for discriminating uncertainty, which is a function of the temporal parameters used in the procedure. In support of this idea, a strong relationship has been established between interval timing and dopaminergic systems [18,35,69–73]. For example, DA antagonists that preferably bind to DA D2 receptors are strongly correlated with slowing down the rate at which subjective time is perceived to pass [17,25,53,56,58,67–69,91]. On the other hand, DA agonists, such as methamphetamine, exert the opposite effect and speed up the internal clock [17,21,58,64–69,86]. Because DA agonists and antagonists influence the speed of an internal clock, there is a proportional relationship between the size of the effect produced by these drugs and the duration of the interval being timed. The proportional characteristic of a clock-speed effect is analogous to that which is observed during simple RT tasks after methamphetamine is systemically injected (i.e., enhancement of the PI effect). This observation suggests that the PI effect is influenced by the speed of the clock—i.e., perhaps the onset of perceptual sampling depends on the output from this clock which accumulates during a trial [55]. However, the opposite effect (i.e., slower RT during trials that use the longest PI with respect to trials that use the shortest PI), which might be expected to occur following DA antagonist administration, has not been reported in the RT literature. It is common within the animal RT literature to find a distinction drawn between RT and movement time (MT) measurements [9]. MT is considered to be the latency from the time at which the animal leaves the central fixation point to the time at which it makes a response to a lateralized operandum. The MT measurement tends not to vary as a function of the PI. Interestingly, while unilateral striatal DA depletion influences RT

63

measurements, it does not change the MT measurement. This dissociation supports a functional distinction between RT and MT [12,37]. However, whether RT and MT should be regarded as outcomes of different cognitive processes is a contentious issue. In fact, whether brain manipulations influence RT, MT, or both may depend on specific aspects of the behavioral task, and any observed dissociation may not readily generalize across conditions [11]. Within the animal RT literature there has been little effort put forth to understanding the PI effect within the context of contemporary theories of interval timing [31,32,46,62,63]. However, the commonalities shared between these cognitive processes, with regard to the underlying neuropharmacology and feedback contingencies, are compelling insofar that they may be subserved by similar, if not the same, neural substrate [55,85]. Consequently, the primary objective of this series of experiments is to shed some light on a proposed connection between interval timing and RT performance in rats as a function of raclopride administration. 2. Materials and methods 2.1. Subjects The experiment used seven male Sprague-Dawley rats weighing between 350 and 400 g (Charles-River Laboratories, Raleigh, NC). The rats were approximately 10 months of age at the beginning of the experiment and were housed individually under a 12:12 light:dark cycle with lights on from 07:00 to 19:00 h. Rats were given free access to water and maintained at 85% of their free-feeding weight by giving them a daily ration of Lab Diet (#5001) rat chow shortly after the experimental session ended.

2.2. Apparatus Seven standard lever boxes were used (Coulbourn Instruments, Allentown, PA). Their dimensions were 29 cm long and 24 cm wide. There were 16 parallel stainless steel bars that crossed the floor. The front and back of the chamber were aluminum and the side-walls were made-up of a transparent acrylic. A pellet dispenser (Coulbourn Instruments, Allentown, PA) dispensed 45 mg precision food pellets (BioServe, Frenchtown, NJ) into a food cup that was recessed near the floor in the center of the front wall. The food cup was 4 cm high, 3 cm wide, and the photobeams (Coulbourn Instruments, Allentown, PA) were positioned on each side of the food cup 1 cm above the base and 2 cm deep. Two retractable response levers (Coulbourn Instruments, Allentown, PA) were positioned 6 cm on each side and 13 cm above the bottom food-cup. The rat was unable to reach the levers with its paws while placing its nose in the food cup. A 2.5 cm Sonalert (P.R. Mallory, Indianapolis, IN) was located 19 cm directly above the food cup and was calibrated to present a sound intensity of 85 dB above background. A 6 W houselight was located on the ceiling in the center of the chamber. Each operant chamber was housed in a sound and light attenuating wooden box equipped with a ventilation fan. An IBM-PC compatible computer was connected to an interface that permitted control of the operant chambers and allowed for data acquisition. The computer sampled events at 18.2 Hz.

2.3. Procedure 2.3.1. Training: constant-PI each session (Sessions 1–45) The rats first received magazine and lever training using a modified autoshaping procedure [17]. The next stage of pre-training was similar in design to the final task; it differed such that it contained autoshaping components. The training procedure began with the onset of a 15 s houselight stimulus. The houselight terminated with the delivery of a food pellet and cued an intertrial interval (ITI) that lasted 90 ± 30 s. Because the houselight was consistently paired with the

64

C.J. MacDonald, W.H. Meck / Behavioural Brain Research 175 (2006) 62–74

delivery of a food pellet, it came to elicit food cup behavior. If the rat inserted its nose into the food cup (a nosepoke) longer than 0.5 s (the PI criterion) during the houselight presentation, a sound stimulus was presented along with a single lever extension–retraction sequence. The side of the lever chosen for each rat was arbitrarily assigned before the experiment with even-numbered rats trained with the left-lever and odd-numbered rats trained with the right lever. If the rat withdrew its nose before the PI criterion was reached, the houselight was turned-off, no reinforcement was delivered, and the ITI started. The sound remained turned on for 8 s. Also during this period, an extension–retraction sequence was initiated for the lever only after 2 s in order to signal to the rat the availability of a reinforced lever press. If the rat pressed the lever at any point while the sound was on, the sound and houselight turned off, and a food pellet was delivered to the food cup, followed by the ITI. On the condition that the rat did not press the lever during this 8 s period, the sound and houselight turned-off and a food pellet was delivered, followed by the onset of the ITI. Once the tone and lever extension–retraction sequence was cued following a nosepoke that was longer than the PI criterion, the rat did not have to sustain the nosepoke in order to obtain reinforcement. It should be noted that the term “constant PI” designates that only a single PI criterion was used within a given session. Constant PI training began with a PI criterion of 0.5 s (e.g., Session 1) and the PI criterion was gradually lengthened from session to session with 0.5 s increments until a final PI criterion of 4 s was acquired. The criterion was increased by 0.5 s when a rat managed to earn at least 40 food pellets during a session with each pre-training session lasting 3 h. In general, if a rat’s correct response fraction (see Section 2.3.6) declined below 0.6 upon increasing the PI criterion, it was returned to the previous PI for two sessions. The longest PI criterion used during this training was 4 s and Experiment 1 (introduction of variable PIs) began once the rat’s correct response fraction exceeded 0.6 for three sessions at a constant PI of 4 s. 2.3.2. Experiment 1: variable-PIs (Sessions 46–60 and 64–74) During the variable-PI training phase, rats reliably pressed the lever during the sound’s presence as illustrated in Fig. 1. This procedure was the same as the pre-training procedure with three exceptions: (1) The houselight’s offset was not accompanied by the delivery of a food pellet. (2) If the lever was not pressed during the extension–retraction sequence and sound, no food pellet was delivered. (3) Within a session, the PI for each trial, i.e., the length of time that the rat was required to keep its nose in the food cup before it was presented with a lever and sound, was chosen randomly to be 1, 2, 3, or 4 s in duration.

2.3.3. Experiment 2: variable-PIs with 6.25 s PI probes (Sessions 61–63) Experiment 2 used the same variable-PI procedure as in Experiment 1 with one exception. Across the three daily sessions, 4% of the trials used a PI that was 6.25 s long. With the remaining percentage of trials, a PI of 1, 2, 3, or 4 s was chosen with equal probability. All trials that ended with a lever press were reinforced. 2.3.4. Experiment 3: raclopride testing with variable-PIs (Sessions 75–80) Raclopride injection sessions were pseudo-randomly intermixed with vehicle injection sessions for a total of three drug and three vehicle injection sessions. The same RT procedure as described in Experiment 1 was used during these test sessions. 2.3.5. Drugs Raclopride (Sigma, St. Louis, MO, R-121) was dissolved in water first and then diluted with saline (Sigma, S8776) 10:1 to a final concentration of 400 ␮g/ml. Approximately 20 min before each session, rats were injected intraperitoneally (i.p.) with 0.1 ml of raclopride solution containing a concentration of 100 ␮g/kg or an equivalent volume of vehicle solution (i.e., dilute saline, 10:1). 2.3.6. Performance measures There were five different event times that were recorded during the variablePI procedure. They were: (1) Time of trial onset—the time of trial onset during the session was recorded. (2) Time of food cup entry—the time at which the rat first placed its nose in the food cup after trial onset. (3) Time of the imperative signal’s onset—this was the time at which the sound turned-on relative to the start of the trial. (4) Time of food cup exit—the time that the rat removed its nose from the food cup during a trial. (5) Time of a lever press—the time at which the rat pressed the lever during a trial. We focused on time-stamped events that were recorded during trials (i.e., while the houselight was on). Each event also had the chosen PI and trial number recorded for the rat. Given this arrangement, we were able to extract a series of different datum of interest from the variable-PI procedure: (1) Reaction time (RT): The length of time that passed between the onset of the sound (the imperative signal) and the rat removing its nose from the food cup as illustrated in Fig. 1.

Fig. 1. Diagram of the structure of the simple reaction-time (RT) procedures used in Experiments 1–3. A houselight turns-on, which signals the rat to insert its nose into the food-cup. The rat sustains this response in the food-cup for a variable preparatory interval (PI) after which the imperative sound stimulus is presented to cue the rat to remove its nose from the food-cup. The rat then proceeds to press the lever to obtain food reinforcement, which turns off the houselight and sound and initiates the intertrial interval (ITI).

C.J. MacDonald, W.H. Meck / Behavioural Brain Research 175 (2006) 62–74 (2) RT per opportunity (RT/Opp): In Experiments 1 and 3, we characterized RT/Opp in the manner similar to that described in previous work [5]. Because Experiment 2 contained such a low percentage of probe trials within a session, we did not collect sufficient RT counts using the “long” probe PI to adequately assess the RT/Opp index. Briefly, RTs recorded during trials were first summarized as a frequency distribution using 100 ms time bins and organized depending on the PI used in the trial. Regardless of the PI used in the trial, the first time-bin corresponded to the subinterval 1.0–1.09 s after the beginning of the PI (e.g., a 0.5 s RT recorded during a 1 s PI trial was assigned to the 6th time-bin and a 0.5 s RT recorded during a 2 s PI trial was assigned to the 16th time-bin). This yields a single 1 × Z vector, with each element in the vector representing a specific 100 ms sub-interval and the total RT counts assigned to that subinterval; Z is an integer corresponding to the highest index used for the analysis. Next, the total number of RT counts (N) were summed across all Z elements in the vector. In order to obtain RT/Opp, for each element in vector RT(1, x) , the RT count in that element was divided by the difference between N and the sum of the RT counts across vector elements starting from RT(1, 1) and up to the immediately preceding element RT(1, x−1) . Used in this way, RT/Opp is a conditional probability that expresses all of the RTs for correct responses recorded in a session with a duration of X s, given the correct response RT was at least X s long. In addition the distribution was smoothed with a three-bin box-car filter in which each time-bin was 100 ms wide. In this manner, analyzing RT/Opp distributions can provide a more sensitive measure of performance than unfiltered RT mean values as they are less influenced by outliers [52]. This calculation method makes it possible that very slow responses during one PI trial type can fall into time bins that are more frequently occupied by responses during trial types associated with longer PIs. However, given that the overwhelming majority of RTs across all experiments were less than 1 s (>98%), this possibility does not conflate the interpretation of the response distributions. More importantly, by considering all responses together irrespective of when the imperative signal turns on, we believe that this display-type better reflects the temporal control of responding with respect to an internal cue (e.g., clock reading) rather than the external cue (i.e., the imperative signal). (3) Movement time (MT): The length of time that passed from when the rat removed its nose from the food cup to the time that it pressed the lever. (4) Vincentizing RT distributions: In Experiments 1 and 3, the RT distributions were constructed with respect to each PI trial type. Because Experiment 2 contained such a low percentage of probe trials in a session, we did not collect sufficient RT counts using the long probe PI to adequately assess the Vincentized distribution. Moreover, the distributions were Vincentized so that the RT distribution of each rat was broken into 20% quantiles and each quantile was averaged across rats [97]. This transformation preserves the shape of each individual function and provides a better description of the mean RT distribution for the group [88]. (5) Correct response fraction–correct trials are considered trials in which the rat withheld removing its head from the foodcup until the onset of the imperative signal and pressed the lever for reinforcement. Anticipatory errors describe trials during which the animal withdrew its head from the food cup before the onset of the imperative signal and no reinforcement was collected. The correct response fraction for each PI trial type describes the percentage of correct trials with regard to total number of trials (i.e., including anticipatory errors). 2.3.7. Data analysis In Experiment 1, we pooled data across 3-session blocks, and calculated statistics using data from Sessions 47–49 for variable-PI training (early training) and Sessions 72–74 sessions of variable-PI training (late training). It should be noted that whether late sessions were considered to occur before Experiment 2 (Sessions 58–60) or after Experiment 2 (Sessions 72–74) did not affect our interpretation—i.e., statistical significance was met whether one used the mean RT and MT before Experiment 2 or after Experiment 2. Therefore, it is unlikely that the introduction of probe trials during Experiment 2, which comprised a low percentage of total trials anyway, influenced our results. We conducted a repeated-measures analysis of variance (ANOVA) using the within-subject factors of training stage (i.e., early or late) and PI in order to make statistical comparisons between RTs and MTs recorded as a function of PI. A

65

repeated-measures ANOVA was also performed on Vincentized RT distributions using the factors training, PI, and quantile. In Experiment 2, the RTs recorded with respect to each PI were pooled across probe sessions. A repeated-measures ANOVA using the within-subject factor PI was performed on the RT index. We used a planned comparison (paired t-test) to compare the mean RT during a probe trial to the RT recorded during the longest PI (4 s). The RT distributions for each PI were not compared due to the small number of probe trials. However, we did construct the cumulative distribution of anticipatory errors emitted during probe trials across test sessions. The anticipatory errors were pooled across rats and placed into 100 ms bins with respect to the beginning of the PI. In Experiment 3, a repeated-measures ANOVA was performed on both RT and MT measures using the factors drug and PI. The data from the three raclopride sessions/dose were pooled and compared to the data pooled across vehicle sessions (see below). A repeated-measures ANOVA was also performed on Vincentized RT distributions using the factors drug, PI, and quantile for the behavioral data obtained from the 100 ␮g/kg raclopride and vehicle injection sessions (see below). In all the three experiments, statistical significance was set at p < 0.05.

3. Results 3.1. Experiment 1: effect of variable preparatory intervals on reaction time (RT) The goal of this experiment was to characterize RT throughout an extended period of training with a simple RT task. In doing so, the aim was to replicate the basic PI effect [12,20]. While it is true that RT can be shown to decrease as a function of PI in rats, nevertheless this property is not universal. Other variable PI procedures have found equal RTs across PIs [3]. The lack of a precise description as to what procedural features facilitate the appearance of the PI effect is surprising given the importance placed on the PI effect for evaluating the role of frontal–striatal circuits in cognition [7,12,38,90]. Moreover, we intend to compare the RT and MT measurements in order to determine whether there is any evidence of a functional dissociation. Mean RT and MT measures during early and late sessions are plotted as a function of the PI in the top and bottom panels of Fig. 2, respectively. A repeated-measures ANOVA conducted on RT measures with factors training and PI revealed a significant main effect of PI, F(3, 18) = 7.93, but not of training, F(1, 6) = 1.60. However, there was a significant training × PI interaction, F(3, 18) = 3.30. Therefore, the same analysis using the within-subject factor PI was performed on the early and late session functions separately. A significant main effect of PI was obtained for early sessions, F(3, 18) = 5.90. Moreover, a reliable effect of PI was observed during early sessions, as indicated by a significant linear regression component, F(1, 6) = 6.13. There was no effect of PI on RT late in training, F(3, 18) = 1.81. Interestingly, the MT measure showed a different pattern from the RT measure as illustrated in Fig. 2. There was a significant main effect of PI, F(3, 18) = 9.71, but not of training, F(1, 6) = 0.07, nor for the training × PI interaction, F(3, 18) = 0.06. A significant linear regression component was observed for the MT measure as a function of PI, F(1, 6) = 16.35. The degree of change in RT over time depended on the PI such that the shortest PI of 1 s was influenced to a greater degree than the longest PI of 4 s. In order to further illustrate the training effect, we constructed the RT distribution for each rat with

66

C.J. MacDonald, W.H. Meck / Behavioural Brain Research 175 (2006) 62–74

Fig. 3. Mean reaction time (RT) for each of five quantiles that compose the RT distribution, which were measured both early (squares) and late (triangles) in training during trials with preparatory intervals (PI) of 1 s (solid line) and 4 s (broken line). A relatively greater change in the RT distribution across training for trials with a 1 s PI takes place in comparison to trials with a 4 s PI.

Fig. 2. Mean reaction time (RT: top panel) and movement time (MT: bottom panel) as a function of the preparatory interval (PI) at early and late stages of training. The top panel displays the attenuation of the PI effect after extensive training. The late RT function (triangles) appears flatter than the early RT function (squares). The bottom panel shows MT as a function of PI during comparable stages of training. In contrast to RT, the late MT function (triangles) does not appear different from the early MT function (squares).

respect to (1) the two PI extremes (i.e., 1 or 4 s) and (2) whether the RTs were measured early or late in training. This strategy was undertaken because the shape of the RT distribution can provide additional insight concerning the nature of the cognitive processes that underlie the RT measurement [52]. An accurate interpretation of the mean RT distribution might be problematic because averaging across rats would obscure the effects of PI and training on the RT distribution. Indeed, there were individual differences observed in the RT distributions among rats. Therefore, the RT distributions were Vincentized (see Section 2) and the mean RT across rats for 20% quantiles were determined. The results of this transformation are displayed in Fig. 3. A two-way ANOVA with repeated-measures on factors training and PI was applied to the mean RT for each 20% quantile. There was no reliable effect of training, F(1, 6) = 2.92, but a significant effect of quantile, F(4, 24) = 25.32, was observed. There was a significant PI × quantile interaction, F(4, 24) = 11.61, and training × PI × quantile interaction, F(4, 24) = 2.89. The latter result prompted a separate analysis of each RT distribution with respect to PI. A two-way ANOVA was conducted on the RT distributions recorded for 1 and 4 s PIs separately, with repeated measures on factors training and quantile. With respect to the 4 s PI, there was no reliable effect of training, F(1, 6) = 0.17, but as expected, a significant main effect of quantile was observed, F(4, 24) = 31.63,. There was no reliable training × quantile interaction, F(4, 24) = 0.34. The same analysis applied to the 1 s PI RT distributions revealed no reliable main effect of training, F(1, 6) = 5.53, but a significant effect of quantile, F(4, 24) = 21.21.

More importantly, there was a significant training × quantile interaction, F(4, 24) = 3.62. Interestingly, the decrease in mean RT that took place across training occurred in each quantile, but a relatively greater change took place with higher quantiles—this was confirmed by a significant linear trend with respect to the difference in RT as a function of training in each quantile, F(1, 6) = 25.04. Therefore, the effect of training was mainly confined to RTs recorded during trials using 1 s PIs. Moreover, the longer RTs composing the RT distribution were observed to change to a greater extent than the shorter RTs that composed the RT distribution. The RT/Opp index during the early and late stages of training is illustrated in Fig. 4. The form and horizontal placement of the RT/Opp distribution provides a sensitive indicator for the degree to which duration discrimination is controlling performance. It deals with the “change in opportunity” for a given RT. Because a shorter RT that is emitted precludes the opportunity for a longer RT, the RT/Opp measurement ensures that these shorter RTs do

Fig. 4. Mean reaction time/opportunity (RT/Opp) distributions for correct responses early and late in training are displayed. The conditional probability of a RT response is shifted rightward during trials that use short PIs early in training compared to RTs occurring late in training.

C.J. MacDonald, W.H. Meck / Behavioural Brain Research 175 (2006) 62–74

Fig. 5. Mean correct response fraction (CRF) as a function of preparatory interval (PI) during early sessions (black bars) and late sessions (white bars) of training. While the CRF differs as function of PI, the relative difference of the CRF with reference to the PI does not change across training.

not disproportionately influence one’s interpretation of the RT distribution. By definition, this graph displays the conditional probability that a RT was emitted at a given time following the onset of the imperative stimulus (i.e., correct responses). Gaussian distributions were fit to each rat’s RT/Opp response function for each PI value during early and late stages of training. Comparison of the differences in the modes of these Gaussian distributions indicated an non-significant effect of training stage, F(1, 6) < 1.0 accompanied by a significant effect of PI value, F(3, 18) = 1797.11 and a significant training stage × PI value interaction, F(3, 18) = 3.89. We considered the possibility that relative differences in the rate of reinforcement for each PI across training might account for the change in the PI effect. There is some evidence that reinforcement can influence RT measurements in a way that is consistent with a diffusion process [10]. In this fashion, the rate of information accumulation is influenced by the reinforcement percentage. The correct response fraction (see Section 2.3.6) for each PI trial-type is displayed in Fig. 5. In this case, lower percentages for a particular PI trial-type indicate relatively more anticipatory errors and hence less reinforcement for that trial-type. A two-way ANOVA using within-subject factors of training and PI indicated a significant main effect of PI, F(3, 18) = 22.46, However, there was no reliable effect of training on the correct response fraction, F(1, 6) = 2.56, nor for the PI × training interaction, F(3, 18) = 2.52. In addition, a significant linear component was confirmed for the fraction of correct responses as a function of PI, F(1, 6) = 52.65. Therefore, there is a difference in the relative amount of reinforcement with respect to PIs – i.e., responses for shorter PIs are reinforced relatively more than longer PIs – however, this difference does not change on account of training. For that reason, it is difficult to understand how the differences in the PI’s relative reinforcement rate can account for the PI effect’s disappearance—i.e., assuming that relative rate of reinforcement is controlling performance and not reinforcement per PI opportunity ([5], see Section 4). The main finding of this experiment is that after prolonged training under variable-PI conditions, the PI effect disappeared. The MT measurement varied as a function of PI in a way that seemed similar to the PI effect. In general, the MT measurement

67

was considerably more variable both within and between animals and did not change on a time course that was similar to RT, which suggests a functional distinction between RT and MT. To our knowledge, this finding has not been reported in other animal RT procedures and supports the idea that results obtained from different RT procedures may not generalize easily across one another [11]. In any case, we assume that an attenuation of the PI effect over training does not occasion the loss of a relationship between interval timing and RT processes. The PIs used in the current experiment were much longer than those traditionally used during animal RT preparations. In most animal RT experiments, the longest PI used in a set is never more than 1.5 s and the spacing between PIs is much shorter than 1 s—but see [92]. We believe that the variable-PI conditions give rise to a situation during which multiple time-intervals are used to optimize RT performance. The change in the PI effect is a result of “preparedness” taking place earlier on account of the responses for shorter PIs that are reinforced. Indeed, this is supported by the RT/Opp distribution (Fig. 4). The RT/Opp distribution around shorter PIs early in training appears rightward shifted in comparison to the late training distribution. Conversely, the RT/Opp distribution around longer PIs early in training appears similar if not leftward-shifted relative to the distribution late in training. This is consistent with the idea that longer PIs are influencing RT generation relatively more than shorter PIs early in training. Up until this point, 4 s is the shortest PI that is being rewarded. After extensive training however, they become less “stimulus bound” with reference to the longest PIs because all PIs are regularly being reinforced. The greater PI range used in our experiment may have allowed us to characterize the change in the PI effect at a higher resolution because the relatively greater spacing between PIs and longer durations may have facilitated their simultaneous and independent use, perhaps preventing stimulus generalization. It is also possible that the change in the PI effect reflects changes in the associative representations that accompany extended training [1,2]. Interestingly, DA pathways play a role in modulating the associative framework that takes place during learning [26] and, in some cases, its role depends on the amount of training [22,25]. The change in the PI effect occurred gradually over the course of ∼25 sessions, during which the rats performed well over 75% in terms of the correct response fraction. Finally, our RT task is atypical—e.g., we required a lever press in order to obtain reinforcement and the sessions were broken into discrete trials signaled by the houselight, as opposed to being self-paced [12,21]. 3.2. Experiment 2: effect of novel preparatory intervals on reaction time (RT) In this experiment, we characterize RT during probe trials that use PIs longer than those associated with reinforcement in our previous experiments. The goal is to investigate the degree of temporal control over RT. We used variable-PI conditions but also included 6.25 s probe trials. The idea was to determine if faster RTs are dependent on encoding specific distributions of PIs during training. If this were the case, then one might expect

68

C.J. MacDonald, W.H. Meck / Behavioural Brain Research 175 (2006) 62–74

Fig. 6. Mean reaction time (RT: top panel) and movement time (MT: bottom panel) as a function of PI and during the 6.25 s probe trials. The top panel shows RT speeding up as a function of PI. However, the RT during probe trials increased in comparison to the RT recorded during trials using 4 s PIs (denoted by the asterisk). Conversely, the bottom panel shows how MT does not differ as a function of PI and does not change during probe trials.

RT to increase during longer probe trials that are not contained within the originally trained PI distribution. The mean RT and MT measures as a function of PI are shown in the top and bottom panels of Fig. 6, respectively, as a function of PI. A repeated-measures ANOVA using the within-subject factor PI found a significant main effect of PI, F(3, 18) = 9.50. A planned comparison between the probe trial RT and the RT recorded during a 4 s PI trial confirmed a significant change in RT, F(1, 6) = 7.75. Conversely, there was no reliable main effect of PI on MT, F(3, 18) = 1.04, nor of the probe trial on MT in comparison to the MT measured during trials with 4 s PIs, F(1, 6) = 0.15. The mean correct response fraction with respect to PI is shown in the top panel of Fig. 7. The relatively higher number of anticipatory errors made by rats during the 6.25 s probe trials suggests that the rats were sensitive to the passage of time during the PI. A repeated-measures ANOVA was applied to the correct response fraction index using the within-subject factor PI. There was a significant effect of PI, F(3, 18) = 6.84, and the decrease in this index was further characterized by a polynomial trend analysis, which confirmed a significant linear component, F(1, 6) = 24.88. A significant difference was also observed between the correct response fraction for probe trials and that which corresponded to 4 s PI trials, F(1, 6) = 11.75. The cumulative distribution of anticipatory errors during probe trials with respect

Fig. 7. Mean correct response fraction (CRF) as a function of preparatory interval (PI) and during the 6.25 s probe trials and the cumulative distribution of anticipatory errors during probe trials (top panel). The mean CRF differs with respect to the PI that was used in a trial. The bottom panel shows the mean anticipatory errors across rats, which increases at a constant rate up until approximately 4 s, after which the distribution increases substantially.

to the onset of the PI is plotted as a function of 100 ms time bins in the bottom panel of Fig. 7. As a group, the cumulative distribution resembles a fixed-interval (FI) scallop that is typical of FI schedules of reinforcement [27]. The cumulative distribution remains low early on in the trial and begins to increase substantially after 1 s, with substantial positive acceleration beginning after 4 s. However, one should be cautious while interpreting this graph given the individual differences among the rats with respect to both the total number of anticipatory errors and when the anticipatory responses were made. In any event, the shape of the anticipatory error cumulative distribution suggests that anticipatory errors reflect a sensitivity to the PI [7,90]. Taken together, these results provide further support for the idea that RT is influenced by the distribution of PIs that were encoded during initial training. In effect, the introduction of a novel, “unexpected” PI that was much longer than those encoded during training induced an increase in RT, but no change in MT.

C.J. MacDonald, W.H. Meck / Behavioural Brain Research 175 (2006) 62–74

69

3.3. Experiment 3: effect of raclopride on reaction time (RT) as a function of the preparatory interval (PI) While the development of RT procedures have refined our understanding of the basal ganglia’s role with respect to behavior, there is less emphasis placed on the connection between DA-function and RT in the context of interval timing. The main goal of this experiment is to further explore the idea of interval timing being relevant to performance in a variable-PI task. Consequently, we decided to test the effects of raclopride, a potent inhibitor of D2 receptors, on performance during the variable-PI, simple RT task [51,76,80]. Earlier studies using standard interval-timing procedures have found that the receptor binding affinity of D2 antagonists predicts the degree to which the internal clock is slowed [68,69]. If interval-timing processes are contributing to RT performance, then raclopride would be expected to effect a differential change in RT as a function of PI. Previous research has indicated a relatively narrow dose/response window for performance of operant tasks under the influence of raclopride [51]. In the present study, all seven rats reliably performed the simple RT task under the influence of 100 ␮g/kg raclopride, but only three rats responded on more than 25% of the PI trials under the influence of 200 ␮g/kg raclopride in a subsequent test. As a result of this disruption in behavior, only the data for the 100 ␮g/kg dose of raclopride are reported here. The mean RT and MT response measures from vehicle and 100 ␮g/kg raclopride injection sessions are shown in the top and bottom panels of Fig. 8, respectively. A two-way ANOVA conducted on RT measures using the within-subject factors drug and PI revealed significant main effects of drug, F(1, 6) = 14.93, and PI, F(3, 18) = 6.59. More importantly, there was a significant drug × PI interaction, F(3, 18) = 4.82, which is evident from the mean RT function in the top panel of Fig. 8. A separate twoway ANOVA with repeated measures on factors drug and PI was performed on the vehicle function and there was no reliable effect of PI, F(3, 18) = 2.95. However, the same analysis on the raclopride RT function found a significant main effect of PI, F(3, 18) = 7.00. Post hoc comparisons (Fisher LSD, α = 0.05) indicated that mean RTs recorded during all PI trial types differed between vehicle and raclopride treatments. The effects of 100 ␮g/kg raclopride on behavior were not restricted to the RT measure. The same analysis was conducted on the MT measurement. There was a significant main effect of raclopride on MT, F(1, 6) = 8.00, but no reliable effect of PI, F(3, 18) = 1.75, nor the interaction between raclopride and PI, F(3, 18) = 2.36. The Vincentized analysis of the RT distributions recorded during 1 and 4 s PI trials after 100 ␮g/kg raclopride or vehicle administration are shown in Fig. 9. A repeated-measures ANOVA conducted on RT measures using within-subject factors PI, drug, and quantile revealed significant main effects of drug, F(1, 6) = 7.02 and quantile, F(4, 24) = 43.5, but no reliable effect of PI, F(1, 6) = 3.59. All interactions were significant—the PI × drug × quantile interaction was of particular interest, F(4, 24) = 4.93. This result suggested a differential effect of raclopride on quantile, and that this effect also depended on whether a 1 or 4 s PI was used on the trial. The same analysis using

Fig. 8. Mean reaction time (top panel) and movement time (bottom panel) as a function of the preparatory interval for rats trained under vehicle (squares) or 100 ␮g/kg raclopride (triangles).

the within-subject factors drug and quantile was applied to each Vincentized distribution for a PI separately. Concerning the 4 s PI, there a significant main effect of quantile, F(1, 6) = 34.65, on RT but no reliable effect of drug, F(1, 6) = 1.22, nor the drug × quantile interaction, F(4, 24) = 0.83. Conversely, there was a significant main effect of drug, F(1, 6) = 12.19, and quantile, F(4, 24) = 34.1, on RT for the 1 s PI distribution, as well as a significant drug × quantile interaction, F(4, 24) = 9.58.

Fig. 9. Mean reaction time (RT) for each of five quantiles that compose the RT distribution, which were measured after vehicle (squares) or raclopride (triangles) administration during trials with preparatory intervals (PI) of 1 s (solid line) and 4 s (broken line).

70

C.J. MacDonald, W.H. Meck / Behavioural Brain Research 175 (2006) 62–74

Fig. 10. Mean reaction time/opportunity (RT/Opp) distributions for correct responses following vehicle (open circles: broken line) or 100 ␮g/kg raclopride (closed circles: solid line) administration. The conditional probability of a RT response is shifted rightward following raclopride injections in a proportional manner.

The effect of 100 ␮g/kg raclopride compared to vehicle injections using the RT/Opp index for RT is illustrated in Fig. 10. Gaussian distributions were fit to each rat’s RT/Opp response function for each PI value under vehicle and raclopride. Comparison of the absolute and relative differences in the modes of these Gaussian distributions for the vehicle and raclopride conditions indicated that raclopride shifted the RT/Opp response functions rightward in a proportional manner. This observation was confirmed by a significant increase in the effect of raclopride as a function of the absolute (but not relative) shifts at each PI value, F(3, 18) = 10.64 and F(3, 18) < 1.0, respectively. We also characterized the effects of raclopride administration on the mean correct response fraction as a function of PI. A repeated-measures ANOVA using the within-subject factors of drug and PI found no reliable main effect of drug on the correct response fraction, F(1, 6) = 0.96, nor the drug × PI interaction, F(3, 18) = 0.41. The mean correct response fraction was collapsed across raclopride and vehicle conditions and this is displayed in the top panel of Fig. 11. There was a significant main effect of PI, F(3, 18) = 7.51, which was substantiated by a significant linear component, F(1, 6) = 8.70. The total number of trials, defined as the sum of anticipatory errors and correct responses, was dramatically affected by 100 ␮g/kg raclopride administration as illustrated in the bottom panel of Fig. 11. A repeated-measures ANOVA using drug and PI as factors indicated a significant main effect of drug, F(1, 6) = 52.52. There was no reliable effect of PI, F(3, 18) = 2.11, nor the drug × PI interaction, F(3, 18) = 1.34. This result is consistent with earlier studies that made use of systemic raclopride administration [59]. In these experiments, however, the decrease in total number of trials was observed with larger doses of raclopride (e.g., 200+ ␮g/kg) compared to the 100 ␮g/kg dose used in the current experiment. The robust effect of raclopride on the total number of trials may be explained by recent results showing raclopride to reduce the number of spontaneous food cup entries in the absence of a strong eliciting stimulus [40].

Fig. 11. Mean correct response fractions (CRF) as a function of preparatory interval (PI) collapsed across both vehicle and 100 ␮g/kg raclopride sessions are shown in the top panel. Mean total number of trials with reference to each PI during vehicle (black bars) or 100 ␮g/kg raclopride (RAC) sessions (white bars) are shown in the bottom panel.

The primary effect of raclopride administration (100 ␮g/kg) was to produce a proportional rightward shift in the RT/Opp distribution that was a function of the PI used in the trial. This suggests that the D2 -specific antagonist influenced imperative stimulus processing, but the effect depended on the time at which the stimulus was turned-on. Experiment 3 also provided additional evidence for a functional distinction between MT and RT. Indeed, while the effects of raclopride on RT depended on the PI that was used during a trial, this was not the case for MT. Raclopride increased MT overall, with no evidence for a reliable interaction with PI. Given the established relationship between DA and interval timing, Experiment 3 provides additional support for a proposed relationship between interval timing and RT. At this point, our results are not inconsistent with our hypothesis regarding the cognitive mechanisms underlying RT performance. DA antagonists, specifically those that operate through D2 receptors, are hypothesized to slow the accumulation of subjective time [56,58,64,67–69]. In light of RT processes, one might suggest that faster RTs are a result of the organism being in a prepared state. We use the term, “prepared state”, as an umbrella term to account for a condition after which behavior changes with reference to a forthcoming event. It might be argued that diffusion models of RT characterize a prepared state by information accumulating toward the decision boundary before the imperative stimulus turns on—i.e., premature perceptual sampling. If the internal clock mediates the onset of premature perceptual sampling at some level, then this relationship may account for raclopride’s effect on RT. Because the onset of premature perceptual sampling depends on a running accumu-

C.J. MacDonald, W.H. Meck / Behavioural Brain Research 175 (2006) 62–74

lation of subjective time (i.e., the internal clock) starting from the beginning of the PI, it follows that trials using longer PIs will be relatively more influenced by a slowed internal clock, hence the relatively greater rightward shift in RT for longer PIs. However, it should be noted that the mean RT appears to be particularly influenced by raclopride during trials that use a 1 s PI, which contrasts with the effect observed using the mode of the RT/Opp distribution. This suggests that raclopride might increase the latency to start the internal clock (i.e., switch latency) on some trials, which would lead to longer RTs. The nature of this latency effect is normally interpreted as attentional [69], and is one in which the 1 s PI trials might be uniquely susceptible. Alternatively, the decision boundary corresponding to the 1 s PI might be set so that relatively more information is required to cross it in comparison to the other PIs [55]. 4. Discussion The overall objective of these experiments was to investigate the role of interval timing during a simple RT task. To this end, we have presented evidence to support the proposal that interval timing is fundamental to simple RT performance. In Experiment 1, it is clear that prior training with a fixed PI determined the PI effect during the early stages of variable-PI training. Moreover, continued exposure to variable-PI conditions flattened the RT function. MT was also observed to vary with respect to PI, which has not been reported previously. Experiment 2 incorporated longer probe trials during which RT, but not MT, increased in comparison to trials with 4 s PIs—again providing evidence for a duration discrimination process. Experiment 3 showed that the effect of raclopride on RT was to lengthen RT in a manner that depended on the PI used during the trial. This finding is attributed to raclopride decreasing the speed of the internal clock, whose output is integrated during the PI, so that rats are relatively less likely to be in a “prepared state” for longer PIs in comparison to shorter PIs. Interestingly, raclopride increased MT but this increase did not depend on when the imperative stimulus was presented. This latter finding provides additional support for a functional dissociation between RT and MT, as well as the active timing of the PI contributing to the RT. It is unclear why no other studies have reported a disappearance of the PI effect during the course of training, as observed in Experiment 1. We are hesitant to claim that extensive training abolishes any influence of the PI on RT given the relatively small number of animals in our experiment and the suggestive trends we observed. Regarding the latter point, it should be noted that in 5/7 rats late in training, there was a decrease in RT as PI increased from 1 to 2 s while RT changes among other PIs for each rat were less systematic. However, the significance of Experiment 1 is not the PI effect’s disappearance per se but that the PI effect is dynamic and changes over the course of training. This finding highlights the importance of how one trains the animal to perform the variable-PI RT procedure and which behavioral criteria are used to determine whether the animal has sufficiently “learned” the task. Following neural perturbations, perhaps one may expect different effects on RT as a function of PI that depend on training. We considered the possibility that a change in the

71

shorter PI’s relative rate of reinforcement played a role in the disappearance of the PI effect, but there were no differences in the relative rate of reinforcement among PIs across training. Because the relative rate of reinforcement among PIs did not change as a function of training, it is unlikely that this variable contributed to the change in RT we observed. However, it is possible that the change in the PI effect is not mediated by relative changes in PI reinforcement density, but by the absolute number of times they are associated with reinforcement across sessions. In this way, longer PIs are paired with reinforced responses less frequently. On the other hand, one may consider longer PIs as being paired with reinforcement more frequently when considering PIs in the context of reinforcement/opportunity [5]. When considering PIs in this way, it is evident that 4 s PIs are reinforced far more often that shorter PI values. This is because, for example, during 4 s PI criterions, PIs < 4 s necessarily must not be reinforced in order to reach the criterion. Future studies should aim to address these possibilities, perhaps by varying the reinforcement percentage as a function of PI. In the RT literature, there is less discussion about premature perceptual sampling in the case where the imperative stimulus is not presented at the expected time. The results from the 6.25 s probe trials in Experiment 2 suggest that the prepared state is influenced by the encoded PI values, which were learned during initial training. A similar experiment was conducted to test the relationship between interval timing and RT in humans, although the experimenters used a choice RT procedure and the PI was held constant during training [36]. Following training, probe trials were introduced in which the imperative stimuli were presented either earlier or later than expected. RT was observed to increase reliably during “early” trials compared to constant conditions. “Late” trials shifted the RT distribution only in comparison to the previous trial, which used a constant PI. As indicated by Grosjean et al. [36], both of these results are amenable to a diffusion-based RT model that incorporates premature perceptual sampling. Interestingly, the distribution of anticipatory errors during probe trials in Experiment 2 also suggested that rats were sensitive to PIs associated with reinforcement insofar that anticipatory errors increased to a greater degree after 4 s, which was the longest PI associated with reinforcement during training. With this in mind, another interesting proposal would be to include probe trials during which the imperative stimulus is never presented. In this way, one could characterize when the rats “give-up”, so to speak, and emit an anticipatory error. This approach is very similar to the method used to study optimal foraging within the context of interval timing [14–16,33,44]. Research in that domain regularly focus on how long an animal is willing to remain in a foraging site (i.e., a patch) that is not paying-off until the decision to leave is made. Expectedly, the “giving-up time” depends on the distribution of interreinforcement-intervals, with the assumption that memories for each time of reinforcement are inter-mixed and ultimately influence behavior. As predicted, raclopride administration in Experiment 3 produced a multiplicative increase in RT as a function of the PI duration. Such a proportional effect would be expected if D2 -

72

C.J. MacDonald, W.H. Meck / Behavioural Brain Research 175 (2006) 62–74

specific antagonists selectively attenuated the rate at which subjective time accumulates during the trial [55,56,68,69]. This idea has been used to explain the effects of Parkinson’s disease (PD) on RT [8,43]. One of the hallmarks of PD is neurodegeneration of DA-producing neurons that originate in the substantia nigra pars compacta [6]. Indeed, PD patients tested off of their medication do not exhibit a PI effect, which is consistent with a slower accumulation of subjective time such that at any given time during a trial’s PI, PD patients are less likely than normal patients to have an accumulated clock value close to the “remembered” temporal criterion [43]. This supports the idea that the onset of “preparedness” might be a function of an internal clock’s operation.One might be inclined to consider the effect on RT we observe as a non-specific motor effect. PD patients may be impaired in action preparation, but the impairment appears to be contextual such that the RT increase depends on the specific temporal contingencies of the task. Because temporal processing is scalar, it may take a relative (rather than an absolute) amount of time to prepare for a response [31,32,34]. Such comparisons are made relative to the intervals comprising the “preparatory” set and can be shown to scale across different sets of intervals when plotted as a function of the longest PI in the set. This type of “time-scale invariance” or “superimposition” is illustrated by replotting the human RT data from [78] as a function of the maximum PI within a set as shown in Fig. 12 [29,30,98]. In summary, our experiments are an attempt to refine the nature of motor preparedness as described by Brown and Robbins [12]. In this respect, we propose that temporal expectations, derived from the reinforcement contingencies of the task, are intrinsic to preparedness [75]. Diffusion models have proven to be a useful heuristic to describe RT measurements [36,55]. Interestingly, neurons in the superior colliculi are observed to modulate their activity during an imperative stimulus in a way that is best described by a diffusion-like process [89]. Similarly, early information-processing approaches provided the basis for

Fig. 12. Mean reaction time (RT) data from four simple RT experiments using human participants [78] are presented with the original PIs re-plotted on a relative time scale. In each case, RT is plotted as a proportion of the longest PI used in a set. The absolute PI durations were selected from a range of 100–900 ms in a manner that allowed for the relative duration of a given PI to vary with respect to the other PIs within the set. Set 1 = 100, 300, 500, 700, and 900 ms; Set 2 = 100, 200, 300, 400, and 500 ms; Set 3 = 300, 400, 500, 600, and 700 ms; Set 4 = 500, 600, 700, and 900 ms.

the striatal beat frequency (SBF) model of interval timing, which is a neurobiologically realistic description of interval timing [62,63]. In the SBF model medium spiny neurons in the dorsal striatum can act as a filter for biologically relevant, temporal patterns of cortical activity [40,41,54,99]. Indeed, these neurons have been shown to modulate their activity with respect to the temporal criteria employed during variations of the peak-interval procedure, which is commonly used to characterize interval timing in humans and other animals [23,57,66,84]. We recently presented a conceptual framework to explain the PI effect that integrates the scalar property of interval timing with diffusion models of RT [56]. Our proposal is that premature perceptual sampling is mediated by a response-rule, similar to those espoused by the scalar expectancy theory of interval timing [31–35,74], in which the current trial’s clock reading is compared to memory distributions for the PI set. We specifically proposed that the output of dorsal striatal neurons influence downstream brain circuits that make use of diffusion-like processes. Assuming DA antagonists influence dorsal striatal neurons by slowing their detection of temporally relevant patterns of cortical activity in real time, then their output to downstream brain circuits that regulate diffusion-like processes would be delayed as well. This interaction would provide a biological substrate for the PI effect and its perturbation by dopaminergic manipulations as demonstrated in this report. References [1] Adams CD, Dickinson A. Actions and habits: variations in associative representations during instrumental learning. In: Spear NE, Miller RR, editors. Information processing in animals: memory mechanisms. Hillsdale, NJ: Lawrence Erlbaum Publishers; 1981. p. 143–66. [2] Adams CD. Variation in the sensitivity of instrumental responding to reinforcer devaluation. Q J Exp Psychol B 1982;34:77–98. [3] Amalric M, Koob GF. Depletion of dopamine in the caudate nucleus but not in nucleus accumbens impairs reaction-time performance in rats. J Neurosci 1987;7:2129–34. [4] Amalric M, Berhow M, Polis I, Koob GF. Selective effects of low-dose D2 dopamine receptor antagonism in a reaction-time task in rats. Neuropsychopharm 1993;8:195–200. [5] Anger D. The dependence of interresponse times upon the relative reinforcement of different interresponse times. J Exp Anal Behav 1956;52:145–61. [6] Barzilai A, Melamed E. Molecular mechanisms of selective dopaminergic neuronal death in Parkinson’s disease. Trends Mol Med 2003;9:126–32. [7] Baunez C, Nieoullon A, Amalric M. Dopamine and complex sensorimotor integration: further studies in a conditioned motor task in the rat. Neuroscience 1995;65:375–84. [8] Bherer L, Belleville S, Gilbert B. Temporal preparation strategy may inflate RT deficit in patients with Parkinson’s disease. J Clin Exp Neuropsychol 2003;25:1079–89. [9] Blokland A. Reaction time responding in rats. Neurosci Biobehav Rev 1998;22:847–64. [10] Blough DS. Reaction time signatures of discriminative processes: differential effects of stimulus similarity and incentive. Learn Behav 2004;32:157–72. [11] Brasted PJ, D¨obr¨ossy MD, Robbins TW, Dunnett SB. Striatal lesions produce distinctive impairments in reaction time performance in two different operant chambers. Brain Res Bull 1998;46:487–93. [12] Brown VJ, Robbins TW. Simple and choice reaction time performance following unilateral striatal dopamine depletion in the rat. Impaired motor readiness but preserved response preparation. Brain 1991;114:513–25.

C.J. MacDonald, W.H. Meck / Behavioural Brain Research 175 (2006) 62–74 [13] Brown VJ, Brasted PJ, Bowman EM. The effect of systemic damphetamine on motor versus motivational processes in the rat. Psychopharmacology (Berl) 1996;128:171–80. [14] Brunner D, Fairhurst S, Stolovitzky G, Gibbon J. Mnemonics for variability: remembering food delay. J Exp Psychol Anim Behav Process 1997;23:68–83. [15] Brunner D, Kacelnik A, Gibbon J. Optimal foraging and timing processes in the starling Sturnus vulgaris: effect of intercapture interval. Anim Behav 1992;44:597–613. [16] Brunner D, Kacelnik A, Gibbon J. Memory for inter-reinforcement interval variability and patch departure decisions in the Starling, Sturnus vulgaris. Anim Behav 1996;51:1025–45. [17] Buhusi CV, Meck WH. Differential effects of methamphetamine and haloperidol on the control of an internal clock. Behav Neurosci 2002;116:291–7. [18] Buhusi CV, Meck WH. What makes us tick? Functional and neural mechanisms of interval timing. Nat Rev Neurosci 2005;6:755–65. [19] Carli M, Evenden JL, Robbins TW. Depletion of unilateral striatal dopamine impairs initiation of contralateral actions and not sensory attention. Nature 1985;313:679–82. [20] Carli M, Jones GH, Robbins TW. Effects of unilateral dorsal and ventral striatal dopamine depletion on visual neglect in the rat: a neural and behavioural analysis. Neuroscience 1989;29:309–27. [21] Cevik MO. Effects of methamphetamine on duration discrimination. Behav Neurosci 2003;117:774–84. [22] Choi WY, Balsam PD, Horvitz JC. Extended habit training reduces dopamine mediation of appetitive response expression. J Neurosci 2005;25:6729–33. [23] Church RM, Meck WH, Gibbon J. Application of scalar timing theory to individual trials. J Exp Psychol Anim Behav Process 1994;20:135–55. [24] Cooper BR, Breese GR, Grant LB, Howard JL. Effects of 6-hydroxydopamine treatment on active avoidance responding: evidence for involvement of brain dopamine. J Pharmacol Exp Ther 1973;185:358–70. [25] Drew MR, Fairhurst S, Malapani C, Horvitz JC, Balsam PD. Effects of dopamine antagonists on the timing of two intervals. Pharmacol Biochem Behav 2003;75:9–15. [26] Faure A, Haberland U, Conde F, El Massioui N. Lesion to the nigrostriatal dopamine system disrupts stimulus-response habit formation. J Neurosci 2005;25:2771–80. [27] Ferster CB, Skinner BF. Schedules of reinforcement. New York: Appleton-Century-Crofts; 1957. [28] Fibiger HC, Phillips AG, Zis AP. Deficits in instrumental responding after 6-hydroxydopamine lesions of the nigro-neostriatal dopaminergic projection. Pharmacol Biochem Behav 1974;2:87–96. [29] Gallistel CR. The modular structure of learning. In: Gazzaniga M, Altman JS, editors. Brain and mind: evolutionary perspectives, vol. 5. Strausbourg: Human Frontiers Science Program; 1998. p. 56–68. [30] Gallistel CR, Gibbon J. Computational versus associative models of simple conditioning. Curr Dir Psychol Sci 2002;10:146–50. [31] Gibbon J. Scalar expectancy theory and Weber’s law in animal timing. Psychol Rev 1977;84:279–325. [32] Gibbon J, Church RM. Representation of time. Cognition 1990;37:23– 54. [33] Gibbon J, Church RM, Fairhurst S, Kacelnik A. Scalar expectancy theory and choice between delayed rewards. Psychol Rev 1988;95:102–14. [34] Gibbon J, Church RM, Meck WH. Scalar timing in memory. Ann NY Acad Sci 1984;423:52–77. [35] Gibbon J, Malapani C, Dale CL, Gallistel CR. Toward a neurobiology of temporal cognition: advances and challenges. Curr Opin Neurobiol 1997;7:170–84. [36] Grosjean M, Rosenbaum DA, Elsinger C. Timing and reaction time. J Exp Psychol Gen 2001;130:256–72. [37] Hauber W, Bubser M, Schmidt WJ. 6-Hydroxydopamine lesion of the rat prefrontal cortex impairs motor initiation but not motor execution. Exp Brain Res 1994;99:524–8. [38] Hauber W, Giertler C, Bohn I. NMDA and dopamine D2 receptors in the caudate-putamen are not involved in control of motor readiness in rats. Psychopharmacology (Berl) 2001;155:43–51.

73

[39] Helson H. Design of equipment and optimal human operation. Am J Psychol 1949;62:473–97. [40] Horvitz JC, Eyny YS. Dopamine D2 receptor blockade reduces response likelihood but does not affect latency to emit a learned sensorymotor response: implications for Parkinson’s disease. Behav Neurosci 2001;114:934–9. [41] Houk JC. Information processing in modular circuits linking basal ganglia and cerebral cortex. In: Houk JC, Davis JL, Beiser DG, editors. Models of information processing in the basal ganglia. Cambridge, MA: MIT Press; 1995. p. 3–10. [42] Johnson HM. Reaction time measurements. Psychol Bull 1923;20: 562–89. [43] Jurkowski AJ, Stepp E, Hackley SA. Variable foreperiod deficits in Parkinson’s disease: dissociation across reflexive and voluntary behaviors. Brain Cogn 2005;58:49–61. [44] Kacelnik A, Brunner D, Gibbon J. Timing mechanisms in optimal foraging: some applications of scalar expectancy theory. In: Hughes RN, editor. Behavioural mechanisms of food selection. Berlin: Springer-Verlag; 1992. p. 61–82. [45] Karlin L. Reaction time as a function of foreperiod duration and variability. J Exp Psychol 1959;58:185–91. [46] Killeen PR, Fetterman JG. A behavioral theory of timing. Psychol Rev 1988;95:274–95. [47] Klemmer ET. Time uncertainty in simple reaction time. J Exp Psychol 1956;3:179–84. [48] Kornblum S. Simple reaction time as a race between signal detection and time estimation: a paradigm and model. Percept Psychophys 1973;13:108–12. [49] Laming D. Choice reaction time following an error. Acta Psychol (Amst) 1979;43:199–224. [50] Laming D. Autocorrelation of choice-reaction times. Acta Psychol (Amst) 1979;43:381–412. [51] Ljungberg T. Attenuation of water intake and operant responding by dopamine D2 antagonists: raclopride provides important cues for understanding the functional mechanism of action. Pharmacol Toxicol 1989;65:9–12. [52] Luce RD. Response times: their role in inferring elementary mental organization. New York: Oxford University Press; 1986. [53] Lustig C, Meck WH. Chronic treatment with haloperidol induces working memory deficits in feedback effects of interval timing. Brain Cogn 2005;58:9–16. [54] MacDonald CJ, Meck WH. Cortico-striatal mechanisms of interval timing and bird song: time flies and may also sing. In: Meck WH, editor. Functional and neural mechanisms of interval timing. Boca Raton, FL: CRC Press; 2003. p. 393–418. [55] MacDonald CJ, Meck WH. Systems-level integration of interval timing and reaction time. Neurosci Biobehav Rev 2004;28:747–69. [56] MacDonald CJ, Meck WH. Differential effects of clozapine and haloperidol on interval timing in the supra-seconds range. Psychopharmacology 2005;182:232–44. [57] Malapani C, Rakitin B, Meck WH, Deweer B, Dubois B, Gibbon J. Coupled temporal memories in Parkinson’s disease: a dopamine-related dysfunction. J Cogn Neurosci 1998;10:316–31. [58] Maricq AV, Roberts S, Church RM. Methamphetamine and time estimation. J Exp Psychol Anim Behav Process 1981;7:18–30. [59] Marrow L, Overton P, Clark D. Disruption of conditioned reaction time performance by dopamine receptor antagonists in the rat. Behav Pharmacol 1993;4:15–28. [60] Marshall JF, Richardson BH, Teitelbaum P. Sensory neglect produced by lateral hypothalamic damage. Science 1971;174:523–5. [61] Marshall JF, Teitelbaum P. Further analysis of sensory inattention following lateral hypothalamic damage in rats. J Comp Physiol Psychol 1974;86:375–95. [62] Matell MS, Meck WH. Neuropsychological mechanisms of interval timing behaviour. Bioessays 2000;22:94–103. [63] Matell MS, Meck WH. Cortico-striatal circuits and interval timing: coincidence-detection of oscillatory processes. Cogn Brain Res 2004;21:139–70.

74

C.J. MacDonald, W.H. Meck / Behavioural Brain Research 175 (2006) 62–74

[64] Matell MS, Bateson M, Meck WH. Single-trials analyses demonstrate that increases in clock speed contribute to the methamphetamine-induced horizontal shifts in peak-interval timing functions. Psychopharmacology 2006;188:201–12. [65] Matell MS, King GR, Meck WH. Differential adjustment of interval timing by the chronic administration of intermittent or continuous cocaine. Behav Neurosci 2004;118:150–6. [66] Matell MS, Meck WH, Nicolelis MAL. Interval timing and the encoding of signal duration by ensembles of cortical and striatal neurons. Behav Neurosci 2003;117:760–73. [67] Meck WH. Selective adjustment of the speed of internal clock and memory processes. J Exp Psychol Anim Behav Process 1983;9:171–201. [68] Meck WH. Affinity for the dopamine D2 receptor predicts neuroleptic potency in decreasing the speed of the internal clock. Pharmacol Biochem Behav 1986;25:1185–9. [69] Meck WH. Neuropharmacology of timing and time perception. Cogn Brain Res 1996;3:227–42. [70] Meck WH. Neuropsychology of timing and time perception. Brain Cogn 2005;58:1–8. [71] Meck WH. Frontal cortex lesions eliminate the clock speed effect of dopaminergic drugs on interval timing. Brain Res 2006;1108:157–67. [72] Meck WH. Neuroanatomical localization of an internal clock: a functional link between mesolimbic, nigrostriatal, and mesocortical dopaminergic systems. Brain Res 2006;1109:93–107. [73] Meck WH, Benson AM. Dissecting the brain’s internal clock: how frontal–striatal circuitry keeps time and shifts attention. Brain Cogn 2002;48:195–211. [74] Meck WH, Church RM, Wenk GL, Olton DS. Nucleus basalis magnocellularis and medial septal area lesions differentially impair temporal memory. J Neurosci 1987;7:3505–11. [75] M¨uller-Gethmann H, Ulrich R, Rinkenauer G. Locus of the effect of temporal preparation: evidence from the lateralized readiness potential. Psychophysiology 2003;40:597–611. [76] Nakajima S, Baker JD. Effects of D2 dopamine receptor blockade with raclopride on intracranial self-stimulation and food-reinforced operant behaviour. Psychopharmacology (Berl) 1989;98:330–3. [77] N¨aa¨ t¨anen R. The diminishing time-uncertainty with the lapse of time after the warning-signal in reaction time with varying fore-periods. Acta Psychol (Amst) 1970;34:399–419. [78] Nickerson RS. Response time to the second of two successive signals as a function of absolute and relative duration of the intersignal interval. Percept Mot Skills 1965;21:3–10. [79] Niemi P, N¨aa¨ t¨anen R. Foreperiod and simple reaction time. Psychol Bull 1981;89:133–62. [80] Ogren SO, Hall H, Kohler C, Magnusson O, Sjostrand SE. The selective dopamine D2 receptor antagonist raclopride discriminates between dopamine-mediated motor functions. Psychopharmacology (Berl) 1986;90:287–94.

[81] Ollman RT, Billington MJ. The deadline model for simple reaction times. Cogn Psychol 1972;3:311–36. [82] Posner MI, Boies SJ. Components of attention. Psychol Rev 1971;78: 391–408. [83] Poulsson EH. Perceptual anticipation and reaction time. Q J Exp Psychol 1950;2:99–112. [84] Rakitin BC, Gibbon J, Penney TB, Malapani C, Hinton SC, Meck WH. Scalar expectancy theory and peak-interval timing in humans. J Exp Psychol Anim Behav Process 1998;24:15–33. [85] Rammsayer TH. Are there dissociable roles of the mesostriatal and mesolimbocortical dopamine systems on temporal information processing in humans? Neuropsychobiology 1997;35:36–45. [86] Rammsayer TH. Neuropharmacological evidence for different timing mechanisms in humans. Q J Exp Psychol 1999;52:273–86. [87] Ratcliff R. A theory of memory retrieval. Psychol Rev 1978;85:59–108. [88] Ratcliff R. Group reaction time distributions and an analysis of distribution statistics. Psychol Bull 1979;86:446–61. [89] Ratcliff R, Cherian A, Segraves M. A comparison of macaque behavior and superior colliculus neuronal activity to predictions from models of two-choice decisions. J Neurophysiol 2003;90:1392–407. [90] Risterucci C, Terramorsi D, Nieoullon A, Amalric M. Excitotoxic lesions of the prelimbic–infralimbic areas of the rodent prefrontal cortex disrupt motor preparatory processes. Eur J Neurosci 2003;17:1498–508. [91] Roitman MF, van Dijk G, Thiele TE, Bernstein IL. Dopamine mediation of the feeding response to violations of spatial and temporal expectancies. Behav Brain Res 2001;122:193–9. [92] Sakurai Y, Shibata R, Kurachi M. Effect of 6-hydroxydopamine lesions of the medial prefrontal cortex of the rat in a reaction time task. Eur Neuropsychopharmacol 1995;5:69–73. [93] Schmidt RA. Anticipation and timing in human motor performance. Psychol Bull 1968;70:631–46. [94] Sternberg S. High-speed scanning in human memory. Science 1966;153:652–4. [95] Sternberg S. Memory-scanning: mental processes revealed by reactiontime experiments. Am Sci 1969;57:421–57. [96] Teichner WH. Recent studies of simple reaction time. Psychol Bull 1954;51:128–49. [97] Vincent SB. The function of the vibrissae in the behavior of the white rat. Behav Monogr 1912;1:1–81. [98] Wearden JH, Bray S. Scalar timing without reference memory? Episodic temporal generalization and bisection in humans. Q J Exp Psychol B 2001;54:289–309. [99] Wilson CJ. The contribution of cortical neurons to the firing patterns of spiny striatal neurons. In: Houk JC, Davis JL, Beiser DG, editors. Models of information processing in the basal ganglia. Cambridge, MA: MIT Press; 1995. p. 29–50. [100] Woodworth RS. Experimental psychology. New York: Henry Holt and Company; 1938.