7.2.1. Introduction

In 1983, Robbins and colleagues [3] developed a test for assessing attention in rats based on a test for human subjects that was originally ascribed to Leonard [4], and is still in use [5]. The rat method was called the “5-choice serial reaction time test” (5-CSRTT) and has since been widely applied for exploring the neurobiology of normal attentional processes and dysfunctions associated with disease states. In the prototypical application, a rat or mouse faces five openings (or ports) in a horizontal array along one curved wall of a test chamber, and a food cup with a clear plastic door is located on the wall behind it. The animal initiates a trial by opening the food cup door. After a short delay, a visual signal is presented, consisting of a brief illumination of one of the five ports. If the animal then breaks a photobeam at the opening of the port, a food pellet (or fixed volume of liquid reinforcer) is delivered into the food cup.

A 3-choice variant of this type of task has been developed by Strupp and colleagues [6–11]. This task is similar in concept to the 5-CSRTT and taps similar functions but uses a slightly different apparatus. The most important difference is that the rat is not required to turn around and obtain the reinforcer at the back wall; the pellet dispenser delivers the reinforcer under the center response port.

In these tasks, the animal must maintain attention to the array of ports in order to detect the signal and respond correctly. Accurate responding thus requires attention in both the temporal and spatial domains. In addition, because responses prior to cue presentation (premature responses) are tallied as errors and terminate the trial, the task also places demands on inhibitory control, which permits inferences about effects on impulsivity. The location, duration, and timing (pre-cue delay) of the visual cue can be varied across trials, enabling independent assessments of sustained attention as well as impulsivity. Varying the duration of cue illumination allows one to parametrically manipulate the demands on sustained attention. Selective attention may also be tapped by presenting distracting stimuli on some trials during the interval between trial onset and cue presentation.

An impressive accumulation of studies over the past 25 years using the 5-CSRTT has substantially increased understanding of the neural substrates underlying sustained and selective attention, as well as inhibitory control [12,13]. These studies have generally used selective lesions or pharmacological manipulations of ascending monoaminergic systems. In general, accuracy of responding on the basic task appears to depend upon cortical acetylcholine, and speed of responding is mediated by mesolimbic dopamine. Auditory distractors are particularly disruptive to rats with loss of ceruleocortical norepinephrine, and adequate forebrain serotonin appears to be necessary to suppress premature responding. Further work with both methods has illuminated conditions known—or suspected—to cause deficits in attention and inhibitory control, such as attention deficit hyperactivity disorder [14], prenatal cocaine exposure [7,11,15], and early childhood lead exposure [8,10].

Research into the genetic underpinnings of attention has been facilitated by new techniques to manipulate the mouse genome, which has stimulated the development of behavioral methods for assessing attention in mice. Humby et al. [16] first showed that mice could be trained to perform the 5-CSRTT, and demonstrated the sensitivity of two mouse strains to parametric manipulations and the muscarinic cholinergic antagonist scopolamine. Since that time, a number of studies have employed genetic manipulations to examine the influence of affective states [17,18] and neurochemical pathways [19–21] on sustained attention. This task has also proven to be a valuable tool for studying murine models of genetic disorders in which attentional dysfunction is prominent; examples include attention deficit hyperactivity disorder [20,22], fragile X syndrome [23], and Down syndrome [24].

In addition to its use in its original form for assessing visuospatial sustained attention, variations on the method have been used for a number of interesting purposes. For example, true “serial reaction time”—that is, the accuracy and speed of responding to sequentially-presented stimuli—has also been modeled in rats using illuminated nosepoke ports. This method focuses on the analysis of sequential behaviors per se, rather than of control of behavior via attention to temporally unpredictable stimuli [25–27].

To probe attention in terms of the Pearce–Hall model of attention [28], Holland’s group [29,30] modified the 5-CSRTT method to dissociate effects of the information value of the cue, using continuous reinforcement for responses to two ports and partial reinforcement for responding to two other ports in the five-choice apparatus. Responses to the fifth port were never reinforced. Trials were paced by the experimenter, not the rat, to maintain an appropriate balance of trial types. Asymptotic performance was more accurate to continuously reinforced ports, but new learning (involving discriminative auditory cues) was more rapid to partially reinforced ports. Lesion studies using this behavioral method showed that cholinergic projections from the nucleus basalis magnocellularis to the amygdala central nucleus, medial prefrontal cortex, and posterior parietal cortex support performance of the task.

7.2.2. Materials and Methods

The following section describes the apparatus and methods commonly used for the 5-CSRTT for rats. Some modifications that pertain to the 3-choice variant for rats (discussed above), as well as those for the mouse version of the 5-CSRTT are noted briefly at the end of this section.

7.2.3. Preparation of the Subjects

1. Ensure that the animals are motivated (hungry). One method is to determine their free-feeding body weight, and then reduce that weight by about 15%. Do not deprive the animals completely of food, but reduce their daily allotment such that target body weights are achieved within 5–10 days. If the animal is fully grown, this target weight can be maintained for the remainder of the experiment. On the other hand, if the animal is still growing, allow it to grow in parallel with free-fed animals to a maximum level (e.g., 350 g for an adult male Long-Evans rat and 250 g for an adult female L-E rat).
2. Adapt the animals to the handling procedures [31] and to the food pellets or liquid reinforcers that will be used to reinforce responding in the chambers. Commercially available precision 45-mg pellets are appropriate for adult male rats; use 25- or 12-mg pellets for small rats or mice. This latter step can be accomplished by offering the animals the new food each day in their home cages or in a holding cage for several days prior to beginning training. This adaptation will obviate possible bait-shyness that may accompany introduction of a novel food. The following procedure describes methods for using food pellets.

7.2.4. Training Steps

1. Cover the response openings. Place the hungry animal in the chamber and turn on the house light. Provide food pellets in the food cup. On the first day, begin with 10–20 pellets in the cup, and allow the animal 30 min to explore the chamber and collect the food. On the three following days, deliver the pellets singly at 30-sec intervals and ensure that the animal retrieves the pellets and consumes them.
2. Remove the covers from the openings. Turn on the house light and deliver a single food pellet to start the session and begin the first trial when the animal retrieves the pellet. A trial involves illuminating the signal light in one opening (selected at random) after an inter-trial interval (ITI) of 2 sec and recording the nosepokes made by the animal into the five openings. When the animal pokes its nose into the illuminated opening, turn off the signal light and deliver a single reinforcer. (An auditory cue may be helpful in informing the rat that food has been delivered, if the action of the dispenser is quiet.) If the animal pokes its nose into a dark opening, turn off the house light for a 2-sec timeout period and do not deliver food. Reset the timeout each time the animal pokes its nose into a dark opening. Begin a new trial when the animal pokes its nose into the food cup after either timeout or food delivery. Present signals in each opening an equal number of times in each session, and select at random the opening to illuminate on each trial. Criterion: 100 reinforced nosepoke responses in a 30-min test session.
3. Repeat step 2, but place a time limit on the signal light and the response period (called a “limited hold”). Allow the animal to initiate each trial as above, and use a 2-sec delay (pre-cue delay) before illuminating a signal. Illuminate each opening for 60 sec and set a 60-sec limited hold after the signal period. If the animal pokes its nose into the illuminated opening during this 2-min period, deliver a food pellet and count a correct response. If the animal pokes its nose into an unlit opening during this period, turn off the signal light and house light and count an error of commission. If the animal fails to make a nosepoke response in this period, turn off the signal light and the house light and count an error of omission. If the animal makes a nosepoke into any opening during the pre-cue delay, turn off the house light, count a premature response, and restart the same trial. Sessions for rats commonly terminate after either 100 correct responses in a 30-min test session, or a 100-trial session with 80% correct responding (see below).
4. Repeat step 3, progressively shortening the signal duration and limited hold, ending with a signal duration of 0.5 sec and a limited hold of 5 sec. Lengthen the pre-cue delay to 5 sec and the timeout period to 3 sec during these steps. A stable baseline of about 80% correct responding with about 15% omissions should be achieved in about 30 training sessions.
5. After the basic rules have been learned, it is useful to vary the duration of the pre-cue delay across trials within each session (e.g., 0, 3, 5, and 9 sec for rats; 0, 2, and 4 sec for mice). Similarly it is useful to vary the duration of the visual cue across trials within the session (as discussed below).

7.2.5. Alternative Methods

1. The apparatus may be modified so that the food is dispensed on the same side of the apparatus as the response ports [17]. This arrangement requires more hardware but facilitates training. Another advantage of this setup, as noted above for the 3-choice variant, is that the animal does not have to turn around and traverse the chamber to obtain the reinforcer. The animal can thereby maintain attentional focus on the response ports throughout the testing session, rendering it more similar to human tests of attention.
2. Trials may be paced by the experimenter rather than the animal [29,30,33,34] by starting a new trial at some fixed or variable time after the animal’s choice response on the previous trial, rather than at the time that the animal retrieves the reinforcer (as is done in the standard procedure). This procedure gives the experimenter control over the timing of events in the test, and removes any potential influence that the animal might exert on the trial pacing by its speed of retrieving reinforcers.

7.2.6. Testing Mice in the 5-CSRTT

DeBruin et al. [35] provide a systematic, step-by-step approach to training mice in this task, based on their previous work with rats [23].

7.2.7. Apparatus and Methodology for the 3-Choice Variant

The 3-choice serial reaction time task used by Strupp and colleagues is identical in concept to the 5-CSRTT, but uses a 3-choice chamber originally developed by Eichenbaum for olfactory discrimination tasks [37]. Briefly, the apparatus is comprised of a small testing alcove containing three response ports, separated from a larger waiting area by a metal guillotine-type door that closes at the end of each trial, and then opens to initiate the next trial following an inter-trial interval. Light emitting diodes (LEDs), one positioned above each of the three response ports, provide the target cue. A correct response (a nosepoke into the port under the illuminated LED) is rewarded with a 45-mg Noyes pellet, delivered onto the floor of the chamber via an opening beneath the center port. As noted above, this feature obviates the need for the animal to turn around to retrieve the reward as required in the 5-CSRTT. This feature of the 3-choice variant makes the task more similar to vigilance tasks used for humans and nonhuman primates and, consequently, facilitates extrapolations from the animal findings to the target human population [38]. This task has provided important information on the lasting cognitive and affective changes produced by prenatal cocaine exposure [7,9,11,39–42], early postnatal lead exposure [8,10,43–47], prenatal or postnatal hyperphenylalaninemia (models of maternal and classic phenylketonuria) [48,49], as well as the attentional role of the ceruleocortical noradrenergic system [6].

7.2.8. Data Analysis and Notes

7.3.1. Introduction

The ability of human subjects to report the occurrence of rare and unpredictable signal events over prolonged periods of time has been extensively characterized [52,53]. Accurate detection of such signals is assumed to depend upon maintaining attention to the task over time, and many of the factors that affect performance of humans on these tasks have been systematized. Sustained attention tasks comprise an important and sensitive component of neurobehavioral test batteries used for assessing the effects of drugs in humans, e.g., benzodiazepines [54], stimulants [55], and ethanol [56].

A major problem for tests of sustained attention involves quantifying and minimizing the false alarm rate. That is, a subject can successfully report many signals simply by responding frequently—though doing so will generate a large number of erroneous reports that a signal had occurred (false alarms). Human subjects can be instructed not to respond in this manner and will normally withhold most false alarms; animals can be trained to do so as well. However, manipulations that increase or decrease overall “responsivity” or response rate are difficult to interpret if no independent measure of the false alarm rate is obtained.

Better estimates of the false alarm rate can be obtained by counting responses to specified non-signal events (blank trials). This approach has been used with both fixed and retractable response levers. The task described below employs a discrete-trial, two-lever approach that requires rats to report the occurrence or nonoccurrence of a single, brief, centrally located signal. Thus, two retractable levers are inserted into the test chamber after a variable period of time to “ask” the rat to report whether a brief signal was presented during that period. If a signal was presented (“signal trial”), a press on one lever produces food and a press on the other lever produces a short timeout period without food. If no signal has occurred (“blank trial”), the converse contingencies apply. Because the levers are retracted between trials, no presses can occur during the inter-trial interval (ITI). Because an explicit response is required on each trial, the proportions of hits and false alarms (P[hit] and P[fa]) can be calculated in relation to the total number of completed signal and blank trials, respectively.

Validation of this method includes both studies of the effects of parameters known to affect human sustained attention, and pharmacological and neurobiological manipulations. Parametric studies include observations that signal intensity, signal rate, and the type of task all affect response accuracy [57–59], as predicted from studies of vigilance in humans [52]. Thus the parameters that affect the behavior of rats in this task closely parallel those that affect sustained attention in humans. In addition, three of the variables that control the behavior of rats in this task (signal intensity, trial presentation rate, and whether detection of a single stimulus or discrimination between two stimulus classes is required) have been shown experimentally to control the behavior of humans in this task.

Pharmacological studies have shown dose-related impairment of signal detection in this task after a variety of nicotinic drugs [59,61–68], d-amphetamine [59], the muscarinic drugs pilocarpine and scopolamine, and the α₂-adrenergic compounds clonidine and idazoxan [65]. Further, the influence of cholinergic projections from the basal forebrain to the cortical mantle in sustained attention has been described in a series of elegant studies [69–74]. This work has led to advances in understanding the neurobiology of sustained attention [75,76] and hypotheses regarding the role of attention in addictive behavior [63]. The method has also been used to characterize the acute effects of organic solvents [58,78–80] and other neurotoxic chemicals [81–83].

This method has also been enhanced by systematic manipulation of the post-signal interval to engage working memory as well as attention [84]. Using this hybrid task, the effects of scopolamine and mecamylamine, drugs often presumed to impair working memory, were shown to affect attention. Martin et al. [85] trained wild type and lurcher mice to perform this task, and determined that the effects on performance in the lurchers were due to motoric rather than attentional deficits.

7.3.2. Materials

7.3.3. Preparation of the Subjects

Same as above for the 5-CSRTT.

7.3.4. Training Steps

1. Shape the rats to press the signal lever for food by autoshaping [86,87], by long (e.g., overnight) sessions with a continuous reinforcement schedule in effect, or by hand shaping. If an autoshaping procedure is used, turn on the signal lamp whenever the lever is extended into the chamber, and turn it off when the lever retracts (either when the rat presses it, or after 15 sec without a press). If overnight sessions are used, be sure to provide adequate water. Criterion: one session of 50 reinforced responses on this lever.
2. Shape the rats to press the other (“blank”) lever by the same means. However, do not turn on the signal lamp when shaping responses on this lever. Criterion: one session of 50 reinforced responses on this lever.
3. Begin training using trials in which both levers are extended into the chamber simultaneously on each trial. Light the signal lamp in half the trials (“signal” trials) and not in the other half (“blank” trials). Retract both levers as soon as one is pressed. Deliver a food pellet after a press on the signal lever in a signal trial and after a press on the blank lever in a blank trial. Turn off all lights for 3 sec after a press on the signal lever in a blank trial and after a press on the blank lever in a signal trial. In signal trials, turn on the signal light 2 sec before extending the levers, and leave it on until the rat presses a lever. Use correction trials to reduce the likelihood of position habits: repeat the conditions presented in each trial that terminates in an incorrect response, up to a maximum of three such correction trials. If the rat makes three consecutive errors, extend only the other (correct) lever in the fourth trial to force a correct response. Criterion: two 100-trial sessions with overall accuracy of 80% or better.
4. Remove the correction trials and increase the total number of trials to 120. Criterion: one 120-trial session with overall accuracy of 80% or better.
5. Turn off the signal when the levers extend (rather than when the rat makes a response) and increase the total number of trials to 150. Criterion: one 150-trial session with overall accuracy of 80% or better.
6. Reduce the duration of the signal from 2 sec to 0.3 sec in gradual steps (e.g., 1.5 sec, 1.0 sec, 0.7 sec, 0.5 sec, and 0.3 sec). The onset of the signal should occur 2 sec before insertion of the levers in all cases, leaving an empty period between offset of the signal and insertion of the levers. Increase the total number of trials in stages to 240. Criterion: one session with overall accuracy of 80% or better at each signal duration.
7. Move the signal lamp from its position above the signal lever to a position at the top of the panel, centered between the levers (above the food cup). Retrain to criterion accuracy.
8. Make the interval after the signal offset variable (select values of 2, 3, or 4 sec randomly on each trial). Maintain accuracy at the 80% criterion.
9. Make the interval before the signal onset variable. (Begin with a list of relatively short and homogeneous values, and work up to a list of values ranging from less than 1 sec to about 25 sec, selected randomly on each trial. A constant-probability list, such as that provided by Fleschler and Hoffman [88] is recommended for the final stage.) Maintain accuracy at the 80% criterion.
10. Vary the strength of the signal. Either the intensity or duration may be varied. Varying the intensity is preferred, but requires digital control of the voltage provided to the signal lamp (through a digital-to-analog converter). At least three signal strengths should be used, preferably more (up to seven). Administer at least 20 trials at each signal strength (10 signal and 10 blank). Maintain accuracy at the 80% criterion.

7.3.5. Data Analysis and Notes

The proportion of correct detections of the signal (P[hit]) should increase with increasing signal strength. The signal strength should be adjusted so that the weakest signal produces a P(hit) about equal to the guessing rate, and the strongest signal produces a P(hit) of about 1.0. The guessing rate is given by the proportion of errors on blank trials, or false alarms (P[fa]). P(fa) should be independent of signal strength and range from about 0.10–0.20.

A wide range of signal strength values improves the consistency of the baseline from day to day, and allows one to differentiate between changes in attention and visual function [65]. That is, poor attending to the signals should cause an increase in P(fa) and a decrease in P(hit) at all signal strengths where P(hit) exceeds P(fa). In other words, the P(hit) by signal strength gradient should shift downward. In contrast, a change in the ability of the rat to see the signal should produce a horizontal shift in the P(hit) by signal strength gradient, so that P(hit) is altered only for signals of intermediate intensity; in addition, P(fa) should not change.

P(hit) and P(fa) can be used to calculate signal detection indices of sensitivity and bias by any of a number of methods [88–91]. However, interpretation of these derived measures depends upon the particular assumptions upon which their calculation is based, and explanation of their meaning invariably requires reference to the values of P(hit) and P(fa) from which they were derived. Thus the advantages of deriving signal detection indices—which involves trading one pair of measures (P[hit] and P[fa]) for another pair of more derived measures (sensitivity and bias)—are generally outweighed by the effort required to calculate and explain these derived measures.

Response time may also be measured as the latency between insertion of the lever and the rat’s response. This variable provides an index of motor function similar to a simple reaction time, because rats typically choose which lever to press during the time interval after the signal by positioning themselves in front of one of the levers and pressing it during its insertion into the chamber. Response time typically does not vary with signal intensity, but does tend to be shorter for hits and false alarms than for misses and correct rejections [58].

7.4.1. Introduction

Another aspect of attentional function frequently assessed in human neuropsychological testing batteries is attentional set-shifting, commonly indexed by the Wisconsin Card Sorting Task (WCST), the primary clinical index for frontal lobe dysfunction. This function can also be tested by the extradimensional shift (EDS) task which is part of the Cambridge Neuropsychological Test Automated Battery (CANTAB), a testing battery originally developed for the assessment of cognitive function in elderly and dementing patients [93], but now also widely used to test patients with Alzheimer’s disease and other forms of dementia, basal ganglia disorders including Parkinson’s disease, Korsakoff syndrome, depression, and schizophrenia, as well as children with learning difficulties or autism (see [94]). Notably, versions of the EDS paradigm have been developed for nonhuman primates and rodents (described below).

In this paradigm, which includes a series of tasks, the subject is first trained to respond to one stimulus dimension (e.g., odor) of a multidimensional compound stimulus and is then required to respond instead to a previously irrelevant dimension (e.g., texture). This shift from one stimulus dimension to another defines an EDS. Insight into the nature of the dysfunction is provided by comparing the rate of mastering the EDS to the rate of mastering an intradimensional shift (IDS), in which two novel stimuli are presented, but the predictive dimension is the same as in the original discrimination. If the subject has formed an attentional set, the mastery of the IDS is more rapid than for the original discrimination, and mastery of the EDS is slower than for the IDS. The EDS phase requires cognitive flexibility (to shift attention from the previously predictive dimension to the newly predictive dimension), and associative ability (to figure out the new contingencies), as well as selective attention (to attend selectively to the new predictive dimension while ignoring the previously predictive dimension). In a typical study, reversals of the correct and incorrect cues within a dimension are also commonly introduced following both the IDS and the EDS, to determine the extent to which behavior is controlled by the dimension as opposed to the specific exemplars of the dimension. (Further discussion of methodological issues can be found in [95,96].)

As discussed by Chudasama and Robbins [94], the ED/ID set-shifting test can serve several functions. First, it provides a sensitive index of frontal lobe dysfunction, based not only on recent empirical evidence from lesion studies [96,97], but also on the fact that it taps the primary function required for successful performance on the WCST, the primary clinical index for frontal lobe dysfunction [92]. Second, it allows one to distinguish between two levels of cognitive flexibility: perseveration to a specific exemplar (tapped by the reversal learning task in this series) versus inflexibility with respect to shifting attention from one perceptual domain to another (i.e., attentional set-shifting). An attractive feature of this task series is that parallel versions have been devised for testing monkeys [98], rats (e.g., see [41,45,96]), and mice [99], thereby providing an opportunity to integrate clinical findings with human subjects (e.g., from the CANTAB) with information concerning the neural and neurochemical systems that underlie specific aspects of task performance.

Both operant and sand-digging versions of the set-shifting paradigm have been described for rodents. In the following sections, we discuss these two versions and outline the key advantages and disadvantages of each, to aid the reader in deciding which task would be preferable for achieving specific goals while considering temporal and fiscal constraints. We describe the sand-digging version of the paradigm in preference to the operant method because of its relative efficiency in time and equipment.

The sand-digging EDS method developed by Brown and colleagues [96] consists of a series of seven tasks, including a simple discrimination [50], a compound discrimination, a reversal of the CD (R1), an IDS, and an EDS, and a reversal of the EDS (R3). Two sessions are required: an initial training session, in which the animals are trained to dig for bits of food hidden in bowls of digging medium, followed by a second test session in which seven discrimination, reversal, and shift tasks are given in sequence. The task is detailed below.

There are several attractive features of this task. First, it does not require expensive operant equipment and can be set up quickly. Second, the entire series of seven tasks can be completed in a single session, representing a considerable savings in time relative to the operant EDS task version. Third, this task series includes novel stimuli at each stage, i.e., a “total change design.” This feature, which aids in interpretation of results (discussed in [96,97]) is more difficult to implement in operant setups. Finally, the task has been validated as an index of frontal lobe dysfunction for rats based on the fact that, in both rats and primates, medial prefrontal lesions impair EDS but not reversal learning, and orbitofrontal cortex lesions impair reversal learning but not EDS learning [96,97,100].

Despite these assets, there are two drawbacks to the sand-digging EDS task relative to the operant version. First, it cannot be automated because it requires hands-on, trial-by-trial administration by an experimenter. Second, as presently configured, it does not enable one to determine the basis of an observed alteration in EDS learning rate. That is, if the EDS task is the only task in the series that is impaired by the experimental manipulation (i.e., no group differences are observed in original learning, IDS, or reversals), it is possible to conclude that the manipulation being tested (e.g., a lesion or a drug) specifically impaired the rate at which the rat mastered an EDS, but the nature of the cognitive change remains ambiguous. This is because there are at least two possible reasons for a selective EDS impairment: (1) an impaired ability to shift attention from the previously predictive cues (inflexibility with regard to attentional set); and (2) impaired selective attention (an inability to filter out the previously predictive cues.

In contrast, because acquisition of the operant EDS task is more prolonged, it is possible to demarcate different phases of learning based on the subject’s patterns of responding. These phases include a perseverative phase, characterized by repetitive responding to the previously correct stimulus; a subsequent chance phase, reflecting trial-and-error responding, with inconsistent patterns of correct responding; and a post-chance phase, in which accuracy exceeds chance and increases steadily toward criterion. Changes in the durations of the specific phases can shed light on the nature of the impairment. For example, an experimental group that exhibits a significant lengthening of the initial perseverative phase, with later learning phases of normal length, is likely to suffer from cognitive inflexibility. In contrast, a group that exhibits an elongated post-chance phase, combined with normal IDS performance, is likely to be impaired in selective attention, unable to focus selectively on the new predictive dimension as a result of being distracted by the previously predictive cues.

Previous research involving a rodent model of prenatal cocaine exposure illustrates how phase analysis of this type of task series can shed light on the integrity of these specific cognitive functions. In this study, the animals were administered a series of olfactory discrimination and reversal tasks followed by two EDS tasks [41]. The first and third EDS tasks required a shift from the olfactory to the spatial dimension; the second EDS task required a shift from spatial to olfactory cues. Analyses of learning rate (errors to criterion) demonstrated that the cocaine-exposed (COC) animals were significantly impaired in the two spatial EDS tasks, but not in the olfactory EDS task. The fact that the COC animals exhibited slower learning than controls only late in the task (after the subjects had made eight consecutive correct responses) suggests that the deficiency was not related to being inflexible in shifting attentional set, as they did not show perseverative responding to the dimension that was previously predictive. Rather, the increased error rate (relative to controls) seen later in the task suggests impairment of selective attention (i.e., difficulty ignoring the irrelevant cues).

7.4.2. Sand-Digging Task: Materials and Methods

Below we describe the sand-digging EDS task series developed for [96] and also adapted for mice [99]. This method entails easily mastered tasks that do not require expensive equipment, and therefore may be accessible to investigators on a tight budget in terms of time or money. However, the interpretive limitations associated with the rapid learning, in terms of being able to identify the specific nature of the dysfunction, should also be kept in mind.

7.4.3. Preparation of the Subjects

Rats require habituation to the apparatus and to digging in bowls for food (e.g., a small piece of sweetened cereal as bait). During a single 60-min session, the rat first learns to dig for bait, which is replaced in the bowl every 5 min as the animal retrieves it. Next, the rat is given three simple discrimination tasks, in which bait is placed in one of two bowls, which differ along one of the three stimulus dimensions (texture, odor, or medium). The exemplars used in this session are not used further. Each rat is trained to a criterion of six consecutive correct choices, where a choice is defined as the first bowl that the rat digs in.

7.4.4. Training Steps

In the second session, the rats are given a series of seven discriminations using three different pairs of stimulus exemplars of each dimension: a simple discrimination (SD), a compound discrimination (CD), a reversal of the CD (R1), an IDS, a second reversal (R2), an EDS, and a reversal of the EDS (R3). Each rat receives the tests in the same order, and is tested with one relevant stimulus dimension and one irrelevant dimension. In the first five tests, the relevant dimension is always the same; it changes at test 6, the EDS.

In each test, the bowls are baited and placed on either side of the permanent barrier at the end of the cage. The removable barrier is set in place, and the rat is placed in the test box on the side of the removable barrier opposite the bowls. A trial begins with removal of the barrier and ends when the rat obtains the bait or makes an error. The first four trials of each test are “discovery trials” in which the rat is permitted to dig in each bowl to find the bait. Beginning on the fifth trial, the rat is permitted to eat the bait if it chooses the baited bowl first. If it chooses the unbaited bowl first, an error is scored and the trial is terminated. The rat is trained to a criterion of six consecutive correct choices.

7.4.5. Data Analysis and Notes

The stimulus dimensions used and the correct and incorrect exemplars must be distributed systematically across individual subjects so that, when averaged across animals, learning scores are not biased by the relative difficulty of the specific discriminations used in each test. In contrast, treatment groups should be matched for stimulus conditions at each stage of testing, to prevent confounding of differences in stimulus discriminability or dimensional salience with the treatment. See Birrell and Brown [96] for further details of the design.

For control animals, the rate of learning (errors or trials to criterion) should be faster for the ID than for the ED. Note that the rat is presented with two multidimensional stimuli in both of these tasks; if the animal had not formed an attentional set (predisposing it to focus on one perceptual dimension over others), then these two tasks would be solved at exactly the same rate and, indeed in this method, a particular set of cues must be used for the ED for some animals and the ID for others, ruling out the possibility that a particular set of cues is easier to master than others.

The inclusion of the five task types in the series (simple and compound discrimination, IDS, EDS, and reversals) aids in interpreting the nature of observed group differences. For example, if the simple and compound discrimination tasks do not reveal group differences, but the experimental group is impaired on the IDS, the reversals, and/or the EDS tasks, then one can exclude the possibility that the observed impairment on one or more of these latter tasks is a result of decreased motivation, impaired ability to perform the motoric demands of the task, or impaired sensory acuity. If the experimental group is unimpaired on all tasks except for the reversals (as seen following lesions of the orbitofrontal cortex [100]), then one can conclude that the experimental group has a specific deficit in the ability to inhibit responses to a previously rewarded object within a given perceptual domain. Finally, if the deficit is limited to the EDS phase, one can conclude that the manipulation specifically altered the ability to shift attentional set; the intact IDS phase allows one to rule out formation of the attentional set as the locus of the impairment. However, as noted above, an impaired ability to shift attentional set could be due to either inflexibility in shifting attentional set or impaired selective attention.

Behavioral Test Systems

Campden Instruments Ltd. http://www.campdeninstruments.com/home.htm, (Europe), Loughborough, LE127XT, England, Tel: 0150-981-14790, moc.stnemurtsninedpmac@selasKU, (Worldwide), Lafayette, IN USA, Tel: 765-423-1505, moc.stnemurtsninedpmac@selasSU

Coulbourn Instruments, LLC, 7462 Penn Drive, Allentown, PA 18106 USA, Tel: 610-395-3771, www.coulbourninst.com

Med Associates Inc. PO Box 319, St. Albans, VT 05478 USA, Tel: 802-527-2343, www.med-associates.com

Panlab, SL, C/Energia,112, 08940 Cornellà (Barcelona), Spain, Tel: +34-934-750-697 (Int’l Sales), http://www.panlab.com/

TSE Systems, http://www.tse-systems.com/, (USA/Canada/Mexico), TSE Systems, Inc. 784 S. Poseyville Road, Midland, MI 48640 USA, Tel: 989-698-3067 (Worldwide), TSE Systems GmbH, Siemensstr. 21, 61352 Bad Homburg, Germany, Tel: +49-(0)6172-789-0

Calibration: Audiometric

Brüel & Kjær Instruments, Inc. 185 Forest St. Marlborough, MA 10752 USA

Calibration: Photometric

EG&G Gamma Scientific, 8581 Aero Drive, San Diego, CA 92123 USA

Food Pellets

Bio-Serv, One 8th Street, Suite 1, Frenchtown, NJ 08825 USA, www.bio-serv.com

P.J. Noyes Company, Inc. PO Box 381, Bridge St. Lancaster, NH 03584 USA