Reaction Time Responding in Rats

Abstract

BLOKLAND, A. Reaction time responding in rats. NEUROSCI BIOBEHAV REV 22(6) 847–864, 1998.—The use of reaction time has a great tradition in the field of human information processing research. In animal research the use of reaction time test paradigms is mainly limited to two research fields: the role of the striatum in movement initiation; and aging. It was discussed that reaction time responding can be regarded as “single behavior”, this term was used to indicate that only one behavioral category is measured, allowing a better analysis of brain–behavior relationships. Reaction time studies investigating the role of the striatum in motor functions revealed that the initiation of a behavioral response is dependent on the interaction of different neurotransmitters (viz. dopamine, glutamate, GABA). Studies in which lesions were made in different brain structures suggested that motor initiation is dependent on defined brain structures (e.g. medialldorsal striatum, prefrontal cortex). It was concluded that the use of reaction time measures can indeed be a powerful tool in studying brain–behavior relationships. However, there are some methodological constraints with respect to the assessment of reaction time in rats, as was tried to exemplify by the experiments described in the present paper. On the one hand one should try to control for behavioral characteristics of rats that may affect the validity of the parameter reaction time. On the other hand, the mean value of reaction time should be in the range of what has been reported in man. Although these criteria were not always met in several studies, it was concluded that reaction time can be validly assessed in rats. Finally, it was discussed that the use of reaction time may go beyond studies that investigate the role of the basal ganglia in motor output. Since response latency is a direct measure of information processing this parameter may provide insight into basic elements of cognition. Based on the significance of reaction times in human studies the use of this dependent variable in rats may provide a fruitful approach in studying brain–behavior relationships in cognitive functions.

Introduction

Various types of dependent variables can be used to assess the cognitive performance in subjects, e.g. errors, response characteristics, time/latency. These measures should predictively vary as a consequence of (1) the degree of complexity of the task; (2) introducing independent variables that affect cognitive processes; and (3) conditions that indirectly affect the cognitive performance (e.g. motivation, stress). The dependent variable is assumed to reflect the(those) cognitive process(es) that underlies the performance in a task. The use of time measures in cognitive tasks has a long tradition in experimental psychology. Among others, the use of reaction time, c.q. response latency, as a dependent variable has contributed significantly to the present knowledge of cognitive information processing.

The use of reaction time as a tool for the study of mental processes was first described by the Dutch physiologist Donders in the late nineteenth century [29]. It was his idea that mental events were possibly measurable in terms of time. In one of his first experiments he observed that the reaction times were different when subjects (including himself) had to repeat a letter under conditions where either the letter was known or the letter was not known beforehand. He reasoned that the difference in reaction time had to be related with the number of psychological processes involved in both conditions. Although it could be argued that reaction time cannot be strictly regarded as a measure of a cognitive function, it cannot be refuted that this measure reflects basic elements of more complex cognitive functions.

Since the latency to respond in a simple reaction time task is minimal (usually within 1 s), it can be expected that this may reflect the shortest route of information processing (from stimulus perception to behavioral response). This allows to define the neuronal substrate underlying this response. Assuming that the shortest route of information processing takes place from the onset of the stimulus and the overt response, reaction time response could be regarded as a “single behavior”. This terminology is used to compare this response with “multiple behavior” which involves more/different information processing stages (e.g. swimming in a water basin and finding a platform using spatial cues).

It has been suggested that brain–behavior relationships can best be studied when a single-behavior–multiple-brain-systems approach is used (cf. 59, 82). Such an approach may be more powerful in studying brain–behavior relationships than a multiple-behavior–multiple-brain-systems approach. This is based on the logical argument that the number of possible explanations for the effects of experimental manipulations are reduced when the single-behavior–multiple-brain-systems approach is used. For example, changes in the parameter reaction time after manipulating the striatal dopaminergic neurotransmission implies that striatal dopamine is involved in reaction time response. On the other hand, changes in spatial discrimination learning performance after manipulations of the hippocampal cholinergic neurotransmission may not necessarily imply that the hippocampal cholinergic system is involved in place learning since spatial discrimination performance is also dependent on other behaviors. Thus, because reaction time response is highly correlated with a definable neuronal substrate this measure can be a powerful tool in studying brain–behavior relationships.

Although the use of response latency as the main dependent variable has a long tradition in human cognitive experimental psychology, its use in animal cognition research is rather scarce. Two main research fields can be distinguished in which reaction time has been used as the main dependent variable in animal studies. One of these research fields is related with movement disorders, e.g. Parkinson's disease and Huntingdon's chorea, in which the role of the basal ganglia and frontal cortex in reaction time response have been investigated 4, 5, 10, 20, 28, 40, 49, 51, 61. The rationale for studying reaction time is that this is a sensitive measure for movement initiation. In human studies it has been shown that reaction time, as a measure of movement initiation, is affected in the disorders described above, giving a logical reason for the use of this measure in animal studies. However, it should be mentioned that the slowing of reaction time in Parkinson's disease is not exclusively due to an impaired motor performance but appears also to be related with a cognitive dysfunction 25, 43, 56.

Aging is associated with cognitive slowing which has been explained in terms of a general decrease in psychomotor speed (e.g. 24, 62). Although the notion of cognitive slowing in humans is well accepted, only a very limited number of studies have been conducted in which the effect of age on psychomotor speed performance has been investigated in rats 21, 22, 47, 54, 65. All animal studies showed an age-related increase in response times which is in good agreement with findings from human studies (e.g. 42, 81). However, the reactive capacity of old rats appears not always to be different from that of young rats 47, 50.

Different test paradigms have been introduced in which reaction time was assessed in rats. A first test to measure reaction time in rats was described by Spirduso et al. [65]. In this test rats were trained to press a lever to avoid a mild foot shock and release the lever as quickly as possible when a conditioned stimulus (CS; light/buzzer) was presented, again to avoid a mild foot shock. They assumed that this task measured the “reactive capacity” of rats. The variables they used were reaction time and percent correct avoidances (percent responses before a foot shock was given when the rats did not release the lever within the inter-stimulus interval).

A second test paradigm in which reaction time was the main dependent variable was developed by Amalric and Koob [3]. In this task the rats were first trained to hold down a lever. After a variable interval a light stimulus was presented and the rats were required to release the lever within a predetermined time period in order to obtain a food reward (for different studies time windows between 500 and 1000 ms were used). In these studies several dependent variables were used to describe the performance of the rats in this task. Next to reaction time the measure “late responses” (i.e. responses that were made after the predetermined time window of e.g. 500 ms) was used. Also, distributions of the reaction times and the reaction, times per interval were used to exemplify the manner in which the rats responded in this task and how treatments affected these measures 9, 49.

A technical sophisticated test for measuring reaction time in rodents was developed by Hauber [36]. In this test the movement initiation of rats was registered by force sensors, which enabled the researchers to detect small movements of rats. In this task the reaction time and the number of incorrect runs (runs in which the reaction times were faster than 100 ms or slower than 1000 ms) were used to analyze behavior. In addition, it was measured when the rat left the start box. This latency was taken as the movement time. The use of force sensors also allowed to examine the effects of treatment on the force of the reaction time. This test also has been found to be sensitive to manipulations of the striatal dopaminergic system 37, 39.

It is obvious that the above mentioned paradigms were designed to measure simple reaction time and that these tasks did not allow one to measure choice reaction time. The use of choice reaction time versus simple reaction time can reveal information about the manner in which the motor responses are selected and that the effects of treatments on the level of information processing can be differentiated. Thus, in addition to the processes which are involved in a simple reaction time, choice reaction time involves additional stages of information processing (viz. stimulus evaluation and response selection) 29, 46. If both simple and reaction times can be measured this would provide a better insight into the effects of treatments on information processing.

Two tasks have been developed that allow the assessment of simple and/or choice reaction time. The first task that allowed the assessment of both simple and choice reaction time was introduced by Brown and Robbins 19, 20. In this task the rats were tested in a conditioning chamber in which there were three response holes on one side [see Ref. [23]]. A rat had to make a nose poke in the central hole and, after a tone signalled the end of a variable interval, had to respond to either the left or the right hole in order to obtain a food reward. In the simple reaction time condition the side to which the rats had to respond was signaled before the response had to be made, whereas in the choice reaction time task the rat had to decide which response to make upon the presentation of a stimulus that was presented directly after the end of the variable interval. As expected, the mean reaction time for the simple reaction time task was shorter than in the choice reaction time condition [20]. In this test different measures of responding were evaluated. The reaction time was analyzed per duration of the variable interval and the percent correct responses were evaluated.

Based on the above-mentioned task, a choice reaction time task was developed in a Skinner box [28]. In this task a rat had to push a hinged panel that gave access to the food tray, positioned between two levers. After a variable interval a light on either the right or the left lever signalled which lever had to be pressed in order to obtain a food reward. The behavior in this test was carefully examined allowing a detailed analysis of behavior whereby non-specific effects of treatment on reaction time responding could be detected. The behavioral parameters included mean reaction time, reaction time per hold duration, number of anticipatory responses, and correct trials.

Several other tasks have been developed in which the parameter reaction time has been used as a dependent variable 6, 30, 31, 53, 54, 70. However, the parameters that were used in these tasks were based on behaviors that, in addition to the reaction time response, also included the motor response. For example, in one task rats had to detect a visual stimulus and respond to the lever on the same side [30]. The time between the onset of the stimulus and the lever press was taken as the latency to respond. Accordingly, these latencies were in the range of about 1–2 s indicating that this also included motor response time. A further important aspect of these type of tasks is that the rats are not necessarily engaged during a trial (e.g. hold down a lever, or push back a panel). This may increase the likelihood that a rat is not attending to a possible stimulus.

In Table 1 an overview is given of the studies in which the effects of different treatments on reaction time performance were investigated. It can be seen that a great emphasis has been placed on the role of the striatal dopaminergic neurotransmission. In general, stimulation of (striatal) dopamine receptors leads to a decrease in reaction time and a change in lever press responding (e.g. decrease of correct lever presses, increase of anticipatory responses; 5, 9, 18, 28, 51, 75). On the other hand, blockade of dopamine receptors (D₁ and/or D₂ receptors) appear to have an opposite effect on these behavioral measures 2, 9, 1, 2, 3, 4, 5, 37, 39, 49, 52, 76, 78. However, it should be noted that these effects were not always consistent within or between tests. It is likely that a part of these differences can be attributed to the doses applied. A further important point that could explain some inconsistencies between the results is that the parameters of reaction time responding differed between the tests used.

These pharmacological data of dopaminergic drugs are in good agreement with the finding that there is a negative correlation between striatal dopaminergic function and reaction time, i.e. the higher the dopaminergic parameters the faster the response times 22, 47, 66, 67, 79. This correlation was found in untreated young and aged rats. As mentioned before, aging was found to increase the reaction time in rats. Further, dopaminergic lesions of the striatum were found to impair reaction time performance in a manner that was similar to that observed after treatment with dopaminergic antagonists 3, 9, 20, 28, 67, 74. Taken together, these data provide strong evidence that the striatal dopamine modulates the executionlinitiation of responses.

Recent studies have shown that the motor output of the basal ganglia is dependent on different neurotransmitters (e.g. dopamine, glutamate, GABA, and acetylcholine) which are intimately connected in a striatal network 1, 27, 33, 35. The functional role of the different striatal neurotransmitters have also been investigated using reaction time tasks. NMDA antagonists appear to increase the reaction time (at doses lower than 0.04 mg/kg, i.p.; 37, 39, 75) and is likely to be mediated by the NMDA receptors in the substantia nigra [7]. An additional effect of NMDA antagonists was that the number of anticipatory responses increased 5, 8, 75, which appeared to be mediated via different brain regions [7]. Finally, the effects of dopaminergic antagonists can (partially) be reversed after administration of an NMDA antagonist 8, 37, 39, 48.

Besides these interactions between glutamate and dopamine in the striatum, reaction time studies provided evidence that muscimol, a GABAergic agonist, decreases the response time when injected into the substantia nigra, but had no effect when infused into the globus pallidus [4]. The number of anticipatory responses was increased after muscimol infusions in the substantia nigra, whereas this parameter was unaffected after infusions in the globus pallidus. From these findings it was concluded that the dorsal pallidum has an important role in the striatal motor output. Another study showed that infusions of scopolamine, a muscarinic antagonist, into the caudate putamen region did not affect the reaction time of rats [12]. However, scopolamine infusions led to a decrease in the number of anticipatory responses and the number of trials completed per session. Thus, blockade of muscarinic and dopaminergic receptors both lead to a decrease in the number of anticipatory responses, but appear to have different effects on the speed of response.

Lesion studies have provided evidence that reaction time response is dependent on discrete regions of the caudate putamen complex in rats. In a simple reaction time task it was found that the motor initiation time was increased after lesioning the dorsomedial part of the caudate putamen complex 10, 19, 40. In contrast, lesions of the (dorso)lateral part or the nucleus accumbens did not affect the reaction time performance 3, 16, 19, 40.

Two studies examined the effects of 6-OHDA lesions in the prefrontal cortex. These two studies revealed opposite effects. One of the studies reported a decrease in reaction time after prefrontal cortex lesions [61], whereas another study found an increase in reaction time after the rats were lesioned [38]. Probably these contradictory effects were related to the test used and the associated differences in the value of reaction times measured in control rats [80 ms [61]versus 200 ms [38]]. Since the value of a simple reaction time should be in the range of 180 ms, it is likely that the performance as assessed by Sakurai et al. [61]may not reflect true reaction time and that these data should be treated with caution.

A recent study,using a choice reaction paradigm in which rats were required to respond to a hole that was either near or far from the central hole (on both sides; [14]). In this experiment it was investigated whether unilateral striatal lesions affected the responding to near and far holes on the ipsilateral and contralateral side in a different manner. The data provided evidence that the nature of the unilateral striatal lesion-induced response deficit was related to the spatial organization of an egocentric response.

Lesions of the subthalamic nucleus lead to faster reaction times and an increase in the number of anticipatory responses [10]. In the same study it was found that these lesions reversed the impairments in reaction time response induced by striatal lesions. In contrast, the subthalamic nucleus lesion-induced increase in anticipatory response was not reversed after combined lesions.

Unilateral aspirative lesions of the medial agranular cortex have been found to increase the choice reaction time independent of the side of the response [17]. The increase in reaction time on both sides was interpreted as evidence for a strong interhemispheric connection of the medial agranular cortex regions. Although the response latency was increased independently of the side of the response there was a strong ipsilateral response bias, which resulted in an increase in incorrect responses. Thus, lesions of these cortex regions, which were assumed to be the homologue of the primate secondary motor cortex, did impair the initiation of responses to the ipsilateral side, but affected the initiation time to both sides.

Summarizing, pharmacological and lesion studies using reaction time as a dependent variable have shown that dopaminergic, gluatmatergic, and GABAergic neurotransmission have a pivotal role in the execution/initiation of motor responses. Although the primary site for these functions can be located within the striatum, it has been shown that other brain regions may (in)directly affect aspects of the motor output. The pharmacological studies have provided evidence that the motor output is dependent on a complex network in which different neurotransmitters are involved and that this network is not strictly located within the striatum. Although some differences between experiments were observed, it appears that this single-behavior–multiple-brain-systems approach may indeed be a powerful tool to investigate brain–behavior relationships.

When measuring reaction time in rats it is unclear how fast rats can respond to such a task. This is primarily due to the fact that animals cannot be instructed to respond as quickly as possible. Hence, a reaction time task should be constructed in such a way that it can be assumed that the rats (are forced to) respond as quickly as possible. For example, this could be achieved by giving a positive reinforcement if the rat does respond within a specific time window [e.g. Ref. [3]], or to give a negative reinforcement when a rat does not respond within a specific time window [e.g. Ref. [65]]. Although this may ensure that rats respond as quickly as possible, one should control for possible behavioral artifacts that could affect the validity of the parameter response time. For example, in one task the response time has been defined as the time between the onset of the stimulus and the release of a lever. The reaction time will be affected by the likelihood by which a rat releases the lever. In the case of a behavioral stimulant the number of premature lever releases will increase and may artificially lead to an increase in fast reaction times, since the likelihood of an unintended lever release after an interval increases. Therefore, attention should be given to possible confounding factors in this type of tasks.

Another approach that can be used to validate reaction time performance in rats is to compare it with the performance as observed in humans. In this respect it can be mentioned that, when using an simple auditory stimulus, the mean value of a simple reaction time is about 150–180 ms in humans [72]. When using the same stimulus in a choice reaction time paradigm the mean response latency is in the range of 260–300 ms. On the basis of analogy it can be anticipated that the value of the reaction time in rats should be in the range of the values that have been reported for humans. It should be mentioned that deviations of only 100 ms (which correspond to a 30% increase) could implicate that other (evaluation) processes are involved. Consequently, in case the mean value of response latencies exceeds the above-mentioned values they may not reflect “true” reaction time.

Human reaction time studies have shown that within a test session there is a variation in the values of the response latencies [46]. This variation has been attributed to various factors, e.g. arousal [58]. In general, the reaction times distribution of individual subjects have been found to have a characteristic shape. Thus, a response distribution should be skewed, i.e. a sharp increase in the number of responses after the onset of a stimulus and a less sharp decrease after the mode of the distribution curve. The specific characteristics of the reaction time distribution also has consequences for the manner in which the data should be analyzed statistically. These reaction time distribution analyses have been developed in human studies and will be discussed at a later point.

Summarizing, in a reaction time task the response latency of rats should be in the range of what has been reported for humans. Further, the reaction time data should have a characteristic distribution, which require appropriate analyses. Since the manner in which rats respond in a reaction time task remains open to any possible interpretations one should control for possible confounding factors that may affect the validity of the main dependent variable.

The experiments described in this paper were designed to assess reaction time responding in rats in an automated Skinner box. The aim of the presentation of these data was to demonstrate that there are possible pitfalls in the assessment of reaction time in rats. Further, these data may give an impression which behavioral aspects of rats should be analyzed to increase the validity of the measure reaction time. Finally, these data were used to demonstrate how reaction time data can be ananlyzed. It was attempted to assess simple and choice reaction time in the same rats (cf. [20]). The task that is presented here was very similar to the task described by Döbrössy and Dunnett [28]with only minor procedural deviations. Because the mean reaction time in the simple reaction time condition was higher than may have been expected, the motivation of the rats was increased by applying a partial reinforcement schedule (i.e. 50% of the trials rewarded) in both the simple and choice reaction time task. Aside from the main dependent variable reaction time, different parameters were introduced to detect possible confounding factors that may affect the validity of the measure reaction time. Finally, the effects of systemically applied amphetamine were examined in one of the tasks to demonstrate how treatment effects can be detected in this task.