Cognition is currently a much-employed term in psycho-pharmacology, referring to a collection of higher-order processes that intervene between sensory processing and motor output to produce behavior. Cognition is thus not a unitary construct and has to be carefully decomposed into its constituents, which can be modeled in terms of well-designed procedures that provide objective measures with good test-retest reliability. These constituents are derived from theories that provide operational definitions of constructs such as perception, attention, working memory, associative learning, and executive control. Further requirements are for tests that are validated in terms of their presumed psychological processes, neural basis, and sensitivity to drug effects. The ultimate requirement is to find procedures that can predict cognitive enhancing effects in humans with neurological or neuropsychiatric disorders. A secondary consideration is for the procedures to be sensitive to detrimental effects of certain drugs, neurotoxins, or other manipulations to provide models that can be remediated by appropriate drug treatments. The final, and perhaps key, consideration is that the tests have some translational validity for humans (McArthur and Borsini 2008a,b). The last is a contentious issue, as it is of course unclear to what extent cognitive functions in rodents might map onto functionally homologous processes in humans. Nevertheless, there has been an excellent degree of translatability thus far in certain domains, which bodes well for the further development of this discipline.
Although cognition can be segregated into its different aspects, it is important in any evaluation to measure a range of functions to be best able to define the functional selectivity of any drug effect; for example, is an effect on a memory or learning test in fact dependent on a perceptual or attentional effect of the drug? Moreover, one has to be certain that basic sensory, motor, motivational, or sedative actions are not in fact responsible for any behavioral change. This can be achieved by a battery approach, in which performance on several tests with different requirements are compared, or by the more elegant and economic method of incorporating "control" tests of such factors as motivation and sensorimotor capacity within a test of a cognitive construct such as attention or working memory.
Several examples of this will be provided later. The possibility of preclinical cognitive test "batteries" are currently popular in the wake of such approaches to measure cognition in clinical trials and experimental studies in humans, embodied, for example, by the ► MATRICS battery for schizophrenia and more generally by the
This is generally best tested by means of discrimination learning or performance, whereby responding in the presence of one stimulus is reinforced, whereas the other is not. The stimuli have to be presented randomly across two locations to avoid a confound by spatial factors. The test can be made more sensitive by varying the degree of similarity of the stimuli, and also using a titration method to determine the limits of discrimination (by which the stimuli are made more similar following a correct response and less so by an incorrect one). By this means, it may be feasible to determine a psychological ► threshold for detection. Of course, one has to rule out other factors such as motivation that might produce possible changes (often deficits) in perceptual function as a result of a drug effect. If a deficit, for example, occurs regardless of level or type of motivation, then a perceptual impairment is more likely. Comparison of different types of discrimination performance, for example, in the visual, auditory, or olfactory modalities will also serve to test the specificity of a perceptual explanation. Similarly, learning factors are also ruled out if the deficit occurs following training to asymptotic performance. It is quite difficult however to distinguish between an effect on perception or attention, unless special manipulations are used to influence the latter. A useful quantitative technique for separating sensory/perceptual from motivational or other response biasing factors is ► signal detection theory, which has been applied with success to determine whether drugs affect primarily perceptual/sensory factors (d') or response bias (6) (Appel and Dykstra 1977). Finally, operant psychologists tend to avoid using such terms as "perception" in such tasks and prefer the more theoretically neutral term, "stimulus control." However, a loss of stimulus control is not necessarily due to perceptual factors.
There are many forms of human attention, including
► selective attention (focusing on one input or feature, while ignoring the rest), ► sustained attention (maintaining attention over a long period), vigilance (detecting rare inputs) and ► divided attention (maintaining attention to more than one input or task). Most, if not all, of these can be measured in experimental animals. Continuous performance tests measure the capacity to sustain attention and generally reveal impairments in disorders such as ► schizophrenia or ► attention deficit hyperactivity disorder. A simple analogue of this in experimental animals is the ► five-choice serial reaction time task (Robbins 2002), based on a paradigm once used to assess attention in human volunteers in a variety of experimental situations, including stress, distracting white noise, and following drug treatment. The five-choice task (Fig. 1) measures the accuracy (errors of commission) and latency of detecting visual targets, as well as errors of omission and impulsive responding (i.e., responding prior to target onset). The latency to collect food pellets provides a control measure of motivation. The difficulty of the task can be enhanced in various ways, including shortening of the duration of the visual target, varying its rate of presentation and temporal predictability, and also the occurrence of defined distractors, such as burst of white noise interpolated into the inter-trial interval. This task has now been widely used in rats, and more recently, mice to measure effects of drug, regional brain lesions, and manipulations of the central neurotransmitters or genetic mutations. Its major uses have been to reveal beneficial effects on response accuracy of some putative "cognitive enhancing'' drugs such as dopamine D1 agonists, and also to characterize the neuropharmacology of impulsive behavior, which has also been shown to predict escalation of cocaine self-administration. There are several variants of this standard task, the main one of which requires the rat to make an observing response into a central location to detect a peripheral target, and has also been used to quantify the "attentional neglect'' that can occur after unilateral manipulations of cortico-striatal brain regions. A rather different form of test requires the cross-modal integration of auditory and visual stimuli.
Some tests of rodent selective attention appear to mimic strongly specific tests in humans that are sensitive to frontal lobe damage, involving the formation and shifting of attentional "sets" and the capacity to avoid a prepotent response to one aspect of a stimulus in order to respond to another (Birrell and Brown 2000). The attentional set-shifting task is based on the use of compound visual stimuli (i.e., that vary in at least two perceptual dimensions). Humans or nonhuman primates are trained to attend to one dimension on the basis of reinforcement and to ignore the other one, including tests of reversal (where the two stimuli within a dimension have their reinforcement contingencies reversed), and intra-dimen-sional shifting (where novel stimuli are introduced, but the
Rodent Tests of Cognition. Fig. 1. Schematic of the rodent five-choice task (See Robbins 2002; Figure provided by courtesy of Dr. Christelle Baunez).
same dimension is reinforced). Finally, an extra-dimensional shift is arranged in which novel stimuli are again introduced, but now the previously irrelevant dimension is reinforced. This latter stage is analogous to the category shift on the ► Wisconsin Card Sort Test, which is much used to assess cognitive flexibility in human patient populations, especially those with presumed damage to the prefrontal cortex. In the rodent version (being available for mice as well as rats), the test is implemented using olfactory cues and texture in a "digging for food'' test paradigm. Performance across the various stages is qualitatively comparable to that seen in primates; the extra-dimensional shift is the most sensitive stage to drug effects, performance at other stages usually being employed as internal controls. There are now various versions of these tests of "cognitive flexibility," which use similar logic for shifts between, for example, responding according to body turns or to space on a cross-maze, or alternatively attending to discrete (e.g., visual) cues versus contextual cues on a maze. These tests do not use different stimuli at each test and so are also confounded by any response to interference. However, they do exemplify the general requirements of tests of executive function (see below).
Another major source of tests of attention comes from animal learning theory, in which the repeated non-reinforced presentation of stimuli retards subsequent learning about them, a process called ► latent inhibition. One theory of latent inhibition ascribes the loss of salience of the preexposed stimuli to a loss of attention, although it is possible that an associative account might suffice. However, a particular advantage of latent inhibition paradigms, which are very sensitive to dopamine D2 receptor antagonists (i.e., ► antipsychotic drugs) is that impairments are indexed subsequently by successful learning, in drug-free state, thus controlling for many other explanations of impairment.
Learning (see Gluck et al. 2007)
Both ► Pavlovian and ► instrumental conditioning are now considered to have cognitive aspects, given that the basis for the former is prediction and expectancy, and for the latter, cognitive control over environmental contingencies. The detection of instrumental contingency, in particular, can be thought of as a higher-order cognitive process, which plays an important component of our ability to make voluntary actions that form part of goal-directed behavior. Disruptions of aspects of Pavlovian and instrumental learning almost certainly underlie all of the major forms of neuropsychiatric disorder, including drug addiction. However, these forms of conditioning will be considered elsewhere in more detail under specific entries.
Learning is generally measured in simplified chambers or operant settings (especially for rats); however, maze learning is often employed when spatial cognition is the main subject of study, being especially compatible with the well-developed foraging tendencies of rodents.
Memory (see Gluck et al. 2007; Morris 2007)
This section serves to overview the many distinct forms of memory paradigms used in rodents, which are considered in greater detail under individual entries. Memory can be divided in many ways; a basic distinction is between relatively transient short-term memory and long-term or more permanent memory. Memory "traces" are thus hypothesized to be "consolidated" into long-term memory. Another distinction that has been made in human long-term memory by Tulving, ► "episodic" (generally autobiographical, the "what, where and when'' of memory), versus ► semantic memory (memory for meaning) has not been exploited so far to any great extent in rodents.
Perhaps the greatest contribution made to the study of memory from rodent studies has been the post-trial or post-training paradigm popularized by McGaugh (McGaugh and Roozendaal 2009). Here, what is generally a single trial or training session is followed immediately by a drug treatment that can either be amnestic (e.g., protein synthesis inhibitors) or promnestic (e.g., amphetamine). ► Retention is tested on a subsequent trial, perhaps 24 h or 3d later. The post-training manipulation is thus designed to influence ► consolidation, either beneficially or adversely. The procedure most often used is of aversive memory; the rodent is punished for stepping down from a platform or through a door by presentation of electric foot-shock. Memory is expressed on the retention trial by a longer response latency to step down or through the door. The great advantage of this design is that the drug cannot be said to have affected memory indirectly by its actions on perceptual, attentional, or motivational mechanisms, as it is administered at a time when these no longer impinge on learning. It is necessary however to perform controls with longer post-trial treatments to check that the drug effects are not affecting retention proactively (i.e., by being active at the time of retention and affecting memory retrieval). Studies of the consolidation of appetitive memory are also feasible, but are used less often because of the unreliability of one trial appetitive learning. The post-trial paradigm has been employed for example to demonstrate the contribution of norad-renergic, opioidergic, and GABAergic mechanisms to emotional memories laid down in the ► amygdala.
Reference memory is a form of long-term memory, which refers to rodent task requirements that stay constant from trial to trial. This definition was originally applied by Olton to rats remembering the constant location of the food-baited arms in an eight-arm ► radial maze. However, it can also be applied to the ► Morris water maze, a notable assay of hippocampal function, in which rodents are required over a number of learning trials to learn the location of a hidden platform in order to escape from a vat of water (D'Hooge and De Deyn 2001; Morris 2007). The rodent is allowed to swim the maze beginning from different vantage points, and so successful learning depends on the construction of a "cognitive map'' to navigate the environment. In contrast, the term ► "working memory'' in the Olton maze refers to the requirements of another memory test procedure in which rodents are required to visit each of the eight arms once and once only in order to retrieve a maximum of eight pellets. So, the animals have to remember only where they have recently been, and this memory is irrelevant to performance on subsequent test days. It can be argued that this form of "working memory'' is not quite the same as that defined by human memory theorists such as Bad-deley, where there is a coordination of different, modality specific short-term memory buffers for use in various tasks such as planning, linguistic discourse, and logical reasoning. However, it does seem to overlap the human form of working memory in some important respects (see also Ko and Evenden 2009).
Olton's working memory tasks are strongly reminiscent of the tests of ► spatial delayed response and delayed alternation that have been used to establish the role of the primate ► prefrontal cortex in working memory. Delayed alternation in rodents is easily implemented in a maze or operant chamber, where it is often referred to as "delayed non-matching to position.'' Nonmatching is an easier task for rodents than matching because of their preexisting foraging tendency to alternate spatial choices. The operant versions of the task allow the systematic variation of delay intervals, which can extend from 0-60s. A "delay-dependent'' effect in such a task is generally taken as evidence of a specific memory effect, independent for example, of attention. However, for that inference to be valid it is necessary for performance on the task at zero seconds to be shown not to be similarly susceptible when the perceptual difficulty of the task is enhanced. An additional artefact that is difficult to surmount in the operant task is that of mediating responses, by which the rodent adopts postures or positions that minimize the memory requirement of the task. One way of overcoming this problem is to use sensitive touch screens to record responding and which can more precisely vary the spatial requirements of the memory tasks, as in the ► CANTAB battery for humans and nonhuman primates.
Recognition memory tasks have a superficial resemblance to those employed for working memory. A commonly used variant is that of object recognition (devised by Aggleton and others based on the paradigm of Delacour) in which a rodent explores a novel object during a sample trial, and is then given a choice between this familiar object and a novel object, in terms of the amount of time it allocates to exploring both objects. Lesser exploration of one object indicates greater familiarity and hence recognition of it (in some, restricted sense). The test can also be adapted to measure ► "social recognition'' by using experimental animals as the "object." Recognition memory is generally manifested over long delays, up to 24 h, although it can be tested at much shorter intervals also and has been shown to depend on structures such as the rodent perirhinal cortex. What distinguishes such tasks from those used to test working memory, apart from the precise test material, is that recognition memory tasks generally employ stimuli only once, so that the test is "trial unique.'' If the same set of objects were to be used over many trials (as occurs in the spatial delayed alternation or delayed response task), this would produce considerable proactive interference, and the test therefore becomes one of recency memory (how recently the stimulus has been experienced) rather than one of recognition memory. In that case, the test also becomes one more of frontal rather than temporal lobe function.
Recognition memory is a less sensitive test of memory than either cued or free recall, in which the memory has to be generated from long-term memory store. Unlike recall, recognition is not particularly sensitive to hippocampal damage, and nor is it the earliest manifestation of ► Alzheimer's disease, where amnesia for episodic memories is more evident. Some human and primate data indicate that the hippocampus is implicated in forms of associative memory, particularly in animals involving space, for example, remembering the location of objects. Recent advances have begun to focus on these forms of associative memory, building on the classical Morris water maze. For example, recent touch-screen tests have been developed for mice and rats of their ability to remember the location of objects, and also to form associative memories of tastes with specific locations (Bussey et al. 2008).
Recognition and recall correspond to what Squire has denoted for humans as ► "declarative memory'' as distinct from "procedural memory'' (memory for "how," or for "skill"). We have not discussed procedural memory in any detail in this article, but it may readily be tested in rodents in motor-learning situations such as the rotor-rod test, or as memory for "habits," being part of the process of instrumental learning.
In human cognitive neuropsychology, this term refers to that set of cognitive processes that serve to optimize performance. The term embraces a number of functions already described earlier, including cognitive flexibility (e.g., attentional set shifting), response inhibition, working memory, planning, ► decision making, and aspects of social cognition, some of which are difficult to test in rodents.
A recent development has been a rodent analogue of Logan's ► "stop-signal" task, which is much used in the testing of patients with ► Attention Deficit Hyperactivity Disorder, for assessing their impulsivity. The stop-signal task requires the cancellation or termination of a motor response that has already been initiated by a "Go" cue, by a less frequently occurring "Stop" cue. It is feasible to measure a stop signal reaction time, as well as a conventional (go) reaction time. The rodent stop signal task (Fig. 2), so far has shown a similar pharmacology to that of the human task, both being sensitive, for example, to the beneficial effects on stopping of the selective nor-adrenaline uptake blocker ► atomoxetine (Eagle et al. 2008). Intriguingly, the classical ► Go/No Go task, which has formal similarities with the Stop task, appears to be differentially sensitive to certain drug effects in a manner suggesting the mechanisms of stopping a motor response to be distinct from those of selecting it (Eagle et al. 2008).
Until recently, it was considered difficult to implement a version of the human Stroop task for rodents. The human Stroop requires subjects to resist responding to the dominant aspect of a colored word stimulus (the word itself) and to report the color. A similar conflict has been achieved in rats by training them on two different conditional discriminations with different rules for responding to each ► discriminative stimulus, and then placing the rat into conflict by presenting combinations of the discriminative stimuli that are either congruent (i.e., both stimuli indicate the same response or incongruent, both stimuli indicating opposing options) (Haddon and Killcross 2006).
Decision-making cognition has attracted considerable recent interest, in the context of both neuropsychological studies of human patients and in neuroeconomics. This
Rodent Tests of Cognition. Fig. 2. Schematic of the rodent stop-signal reaction-time task (See Eagle et al 2008; Figure provided by courtesy of Dr. Dawn Eagle).
trend is followed in rodent studies by the provision of tests of important component processes, including the discounting of rewards over time, as well as time perception itself. Recently, rat analogues of the ► Iowa Gambling Task, as used to show deficits in patients with lesions of the ventromedial prefrontal cortex, have begun to be implemented, but there are, as yet, few pharmacological studies.
Was this article helpful?