Sustained attention in mice: expanding the translational utility of the SAT by incorporating the Michigan Controlled Access Response Port (MICARP)

Abstract

Advances in mouse genetic technology have spurred increasing interest in the development of cognitive tasks for mice. Here, we describe and discuss the modifications necessary to adapt a task for the assessment of sustained attention performance for use in mice, including for taxing the top-down control of such performance. The validity of the Sustained Attention Task (SAT), including the distractor version (dSAT), have previously been demonstrated in rats and humans. This task requires moveable or retractable operanda; insertion of operanda into the operant chambers cues animals to respond to a prior signal or non-signal event, reporting either a hit or a miss, or a correct rejection or false alarm, respectively. Retractable levers did not support sufficiently high and stable levels of performance in mice. Given the widespread use of static nose-poke devices for testing operant performance in mice, we therefore designed and fabricated a retractable nose-poke device. As this device extends into chambers, a hole for nose-poking is slowly opened and closed again as the device retracts (termed the “Michigan Controlled Access Response Port”; MICARP). Results describe the effects of variation of signal duration and event rate, trial outcome and trial type probability, effects of mice deprivation levels, and the reliability of SAT and dSAT performance. Mice perform the SAT and dSAT at levels comparable to those observed in rats. This task will be of assistance in expanding the translational usefulness of the SAT and dSAT.

1. Introduction

Behavioral technologies need to be developed in order to foster the modeling of human behavior in nonhuman animals and to study theories of behavior in humans using laboratory settings [38]. The quest for “translational research” has clarified the utility of behavioral tasks that validly measure cognitive processes and capacities in animals and humans. These tasks assist in generating corresponding brain-behavior relationships in patients, healthy humans, and animal models [e.g., 46].

Efforts to develop treatments for the cognitive symptoms of schizophrenia have exemplified the essential role of such translational tasks [e.g., 4,27,45,58]. Impairments in sustained attentional performance, including the deficient (top-down) control of such performance, are major cognitive impairments in schizophrenia [e.g., 26]. Cognitive control in this context concerns the mental mechanisms that support sustained task compliance, particularly in interaction with performance challenges, and the recruitment of mechanisms contributing the stabilization and recovery of attentional performance in response to such challenges [35,52].

In our prior research, we validated a Sustained Attention Task (SAT), including a distractor condition (dSAT) that serves to tax top-down control mechanisms, for use in humans, including schizophrenic patients, and rats [12,37,42,46]. This task consists of a random sequence of signal and non-signal events, with the subject reporting the prior presence or absence of a signal by manipulating one of two operanda. Correct responses are hits and correct rejections, and incorrect responses are misses and false alarms, respectively. The operanda are extended into the chamber following a signal or non-signal event, thereby avoiding the need for an additional visual or auditory cue to signal the onset of the response period (typically 4 s, beginning 1 s after the event). Early experiments associated with the refinement of the rat version of the task indicated that such an additional cue causes high levels of false alarms (claims for signals in non-signal trials). In contrast, extending levers into the chambers does not cause such disruptive effects. As guided by theories of sustained attention [47,49], important additional characteristics of the task, such as variable intertrial intervals (ITI) and variation of signal intensity or duration, were implemented in order to further increase the cognitive load imposed by this task [12,42].

While rats and humans perform comparably in this task, there are also important differences, specifically concerning their differential propensity to adopt a more liberal versus a more conservative response criterion during performance challenges [detailed in 12]. The demonstration and analyses of such differences are important for constraining the interpretation of effects on the performance of animals, specifically in the context of drug development efforts [e.g., 51].

Neurobiological research demonstrated the necessity of the cortical cholinergic input system in performing the SAT and dSAT, assisted in revealing the cortical circuitry involved in the enhancement of dSAT performance by agonists at subtypes of nicotinic acetylcholine receptors (nAChRs), the role of mesolimbic-cholinergic interactions in dSAT performance and, in humans, confirmed that right prefrontal regions mediate the initiation of top-down control mechanisms required to sustain dSAT performance [e.g., 6,14,29,30,43,56].

The evidence from studies in rats and humans has converged to indicate that the sequence of signal and non-signal trials, and the discrete implementation of all four response types (hits, misses, false alarms, correct rejections) are critical for revealing major aspects of the cholinergic mediation of SAT performance. Specifically, trial sequences that begin with a factual or perceived non-signal trial (correct rejection or miss), and are followed by a signal, require a transient increase in right prefrontal cholinergic activity in order for this signal to be detected. Thus, the conceptualization of the SAT has evolved from being viewed as a pure sustained attention/vigilance task to involving more complex cognitive mechanisms, including attentional re-orientation and shifts from endogenous to exogenous attention. Impairments in the ability to engage with external stimuli and re-orient attention contribute essentially to cognitive disorders [for review see 29,54].

Importantly, schizophrenic outpatients are specifically and robustly impaired in performing the dSAT version of this task [13]. Likewise, rats modeling aspects of schizophrenia are impaired in performing the SAT/dSAT and exhibit dysregulated performance-associated increases in acetylcholine release [34,41,53].

Prior attempts to adapt the SAT for use in mice have remained rare [for review see 9] and revealed important challenges. Martin and colleagues used a straight forward analogue of the rat version of the task to assess effects of loss of cerebellar Purkinje cells [40]. However, less than half of the mice acquired this task. Furthermore, inspection of their data in Table 1 suggests that animals’ performance remained close to chance (defined as 50% hits and 50% correct rejections). Consistent with this view, the distractor produced significant yet relatively limited impairments in performance (less than 7% decrease in correct rejections and 1.5% decrease in hits in wild type mice). Our own attempts to use retractable levers produced comparable levels of performance and limited distractor effects. Moreover, within- and between-session performance was prohibitively variable. We speculated that, in interaction with demands on cognitive processes, the use of retractable levers in mice does not support adequate levels of SAT performance (see Methods for more explanation of our rationale and additional references). As nose-poke devices have been used frequently in mice and appear to offer a response type closer to their spontaneous behavioral repertoire [e.g., 39; see Methods for additional references], we therefore designed and fabricated a moveable/retractable nose-poke device, as will be described in Methods. As a result of implementing this device, mice perform the SAT/dSAT at levels statistically similar to those observed in rats. The SAT/dSAT is therefore available for translational research in mice, rats and humans.

Table I.

Stages of task training and major training conditions

Stage	Correction Trials	Signal Durations	Houselights	Days to Criterion or on Training (M; SEM)
Stage 1	n/a	n/a	Off	5.83 ± 0.31
Stage 2	Yes	5 s	Off	14.33 ± 1.17
Stage 3	Yes	1 s	Off	11.67 ±1.52
Stage 4	No	0.5, 0.05, 0.025 s	Off	7.33 ± 1.20
Stage 5 – Final	No	0.5, 0.05, 0.025 s	ON	12.17 ± 3.20
Distractor (dSAT)	No	0.5, 0.05, 0.025 s	On-off at 0.5 Hz during dSAT blocks	4.5 ± 0.56

2. Materials and Methods

2.1. Subjects

MIce, water deprivation schedule, body weights

Mice, originating from our breeding program, of both sexes were used (n= 38). These mice originated from breeding mice hemizygous for the high-affinity choline transporter [CHT; 17,18,19,20]. CHT hemizygous (CHT+/−) mice, obtained from Vanderbilt University (R.D. Blakely’s laboratory) were bred on a C57BL/6 genetic background at the University of Michigan. Mice were genotyped at weaning by a commercial vendor (Transnetyx, Cordova, TN). For this study, only wild-type mice (CHT+/+; congenic on a C57 background) were used. Animals were individually housed in a temperature (23°C) and humidity controlled (45%) environment with a 12:12 light/dark cycle (lights on at 7 a.m.).

Mice were water-deprived by restricting access to water to a 4-min period following each operant training or practice session (see Results for effects of varying deprivation levels). During the first week of deprivation, access to water is reduced in steps, from 24 hours to 12, 6, 4, 2, 1 hour, and eventually to 4 min. Their average intake of water during this free period was 1.92 mL. Correct responses during the task sessions was rewarded using sweetened water [0.2% saccharin; 6 μL per reward; total average session delivery: 0.45 mL; 2]. On days not tested, water access in their home cages was increased to 20 min. Food (Rodent Chow, Harlan Teklad, Madison, WI) was available ad libitum.

Mice began experiments at 12 weeks of age when they weighed 20–30 g. Mice initially lost 4–15% of their ad libitum body weights. Thereafter their body weights remained stable at 90–100% of their ad libitum weights. Animals were weighed twice weekly.

Experiments occurred at the same time each day, 5–6 days a week. Rats and mice were tested between 8 a.m. and noon. Similar to rats [25], mice practicing the SAT during the lights-on period acquire a diurnal activity pattern (Paolone et al., unpublished evidence).

Rats

Adult male Wistar rats (Harlan Laboratories, Indianapolis, IN; n=5), aged three to five months and weighing between 250 and 300 g at the beginning of the experiments, were used. Animals were individually housed in a temperature (23°C) and humidity controlled (45%) environment with a 12:12 light/dark cycle (lights on at 7 a.m.). Animals were water deprived by restricting water access to a 10-min period following each operant training or practice session. Water was also provided as a reward during task performance (30 μL/reward). On days not tested, water access was increased to a total duration of 30 min. Food (Rodent Chow, Harlan Teklad, Madison, WI) was available ad libitum. SAT shaping and dSAT testing in rats have been previously described in detail [e.g., 34].

All procedures were conducted in adherence with protocols approved by the University Committee on Use and Care of Animals at the University of Michigan and in AAALAC (Association for Assessment and Accreditation of Laboratory Animal Care)-accredited laboratories.

2.2. Operant chamber configuration

Behavioral training and testing took place using twelve operant chambers (Med Associates, Inc., St. Albans, VT) located inside sound attenuating chambers fabricated at University of Michigan (Ann Arbor, MI). The control of stimuli and recording of responses were managed by a SmartCtrl Package 8-In/16-Out with additional interfacing by MED-PC for Windows (Med Associates Inc., St. Albans, VT) using custom programming.

The chambers were custom built by Med Associates, modifying the wide modular mouse chamber to the exterior dimensions of: 24.10 cm L x 20.00 cm W x 29.50 cm H. Each chamber was equipped with an intelligence panel consisting of two panel lights (2.8 W), a liquid dispenser, and two MICARPs (below) that were fabricated at University of Michigan. Each attenuating chamber was also equipped with a ventilation fan, a video camera, and 4 red LED’s (JameCo P/N 333489; Jameco Valuepro, Belmont, CA) for background illumination and for video recording. This intelligence panel is comprised of three columns. The MICARPs were located in the lateral columns of the intelligence panel and the liquid dispenser in the center column (see Figure 1). The two panel lights were placed immediately above the liquid dispenser in the middle column, at 8.25 cm and 12.50 cm from the base of the operant chamber. Finally, a house light (2.8 W) was located at the top of the middle column on the rear wall (21.51 cm from the base of the operant chamber).

Figure 1.

Technical drawing of the MICARP. (A) shows the individual components of the MICARP and (B) per column, is the bird’s eye, side, and frontal views are shown with the MICARP retracted (top panel) and extended (bottom panel). 1) face plate; 2) end cap; 3) moveable cylinder; 4) stationary core; 5) key way; 6) key; 7) photocell port; 8) wire channel; 9) side plate; 10) support rod; 11) photocell cut-out switch; 12) motor; 13) crank arm; 14) link rod; 15) crank-end pivot assembly; 16) cylinder-end pivot assembly; 17) motor mount.

The ambient light in the operant chamber generated by the various lights was measured with a UDT Instruments Model 351C photometer (UTD Instruments, San Diego, CA). The receiver was positioned vertically in the center of the chamber floor, with the sensing element approximately 5 cm above floor level. A baseline of 0.01 lux was established with all lights off. The red LEDs provided 3.72 lux of background illumination. The house light generated 12.33 lix of ambient light, the signal light 9.12 lux, and the distracter light 10.04 lux. Because the signal light and the distracter panel light (below) employed the same device, the measured difference is likely due to the distracter light being situated above the signal light (outer distance: 41.30 mm), providing a better angle toward the sensing element. The combination of the red LEDs and the signal light generated 12.75 lux and the combination of the house light and the signal light produced 21.30 lux. It should be noted that the red LEDs were turned on during Acquisition Steps 1–4 (below) but not in the final task when the houselights were on throughout the session. In the dSAT mode the house light and the distracter light flashed asynchronously, with light level fluctuating between 12.33 and 10.04 lux.

2.3. MICARP

Rationale

Initial attempts to train mice to SAT criterion performance (defined below), using retractable levers similar to the rat version of the task [1,12,42,43,56], failed to generate stable and useful levels of performance. Force requirements for pressing levers, species-specific sensorimotor repertoires and response preferences collectively appear to account for substantial strain-specific differences in the rate of acquisition of lever pressing responses by mice and high variability of performance of operant schedules of tasks using this operandum [11,21,44,55]. However, not all studies support this conclusion [28], suggesting that the utility of lever pressing depends on complex, task- and strain-related variables.

The use of retractable or moving levers, which is a necessity for the SAT (below), further increases the complexity of this issue. In the standard rat version of the SAT, levers are extended into the chamber on average every 12th second, prompting a response to indicate the prior presence or absence of a signal (below). A lever press within 4 seconds is counted as a response and triggers withdrawal of the levers. The lack of a response after 4 seconds is counted as an omission and withdrawal of the levers. Mice were able to acquire the initial discrimination but failed to reach above-chance performance in the actual SAT (below). Variation of task parameters, including signal duration, intertrial interval, and lever extension time, did not resolve this issue. Martin et al. [40] also attempted to train mice in a version of the SAT that included retractable levers. Less than half of the animals reached the final testing stage and reported performance levels remained close to chance (about 50% hits across signal durations and 60–70% correct rejections; their Table 1). Therefore, it seems that in interaction with demands on attentional processes (defined below), moveable levers do not serve to generate and maintain informative response rates and levels of response accuracy.

Operant procedures in mice have employed nose-poke devices as operanda. Operant performance based on such devices seems to be more rapidly acquired and remain more stable over sessions than when using levers [e.g., 5,36,39]. We therefore set out to develop a moveable nose-poke device which, in addition, would prevent exploration while withdrawn, and prevent mice from getting stuck in the opening, particularly as the device is retracted from the chamber. Furthermore, moving speed and other aspects of the device were adjusted to ensure that mice readily detect the extension of the device while minimizing startle responses. This design and construction of the device, termed the Michigan Controlled Access Response Port (MICARD) will be described next.

Design and initial testing

The MICARP is a hole (diameters below) with an infrared photo beam interruption circuit, contained within a retractable cylinder with a stationary core. The end of the stationary core acts as a plug, controlling access to the port. The retractable cylinder is manipulated by a DC motor through a link rod and a bell crank arrangement. The motor is driven through a MedAssociates interface with extension and retraction points controlled by limit switches. The retractable cylinder travels on a stationary core, extending and retracting into the operant arena in the same manner as a retractable operant lever, albeit at a slower speed. Commercially available retractable mouse levers (MedAssociates) extend at a rate of 0.063 m/s (in comparison to rat levers, at 0.05 m/s). The MICARP extends at 0.021 m/s.

Several prototypes were attempted before the final version, yielding informative insights. The first prototype had a 22.2 mm Ø response port and a horizontal IR photo beam midway between the top and bottom of the port, inset 7.00 mm from the entrance. This port proved to be too large and mice could easily insert their snouts into the port under or over the photo beam, resulting in a large number of unregistered responses. These unregistered responses, especially in the early learning phase, interfered dramatically with the training regimen.

The next prototype consisted of a 22.2 mm × 12.7 mm oval shaped port. This port shape reduced the animal’s tendency to poke under or over the photo beam and improved the response success rate. However, the oval shaped port, and the corresponding end on the stationary core, proved difficult to make and was deemed impractical for production.

The third prototype was a 19.0 mm round port that was equipped with a second photo beam. The two photo beams were positioned perpendicular to one another, one horizontal and one vertical. The photo beam circuit is a standard MedAssociates ENV-254, with an additional photocell set. The IR emitters are wired in parallel, and the phototransistors are wired in series. Interrupting either one, or both, of the photo beams registered a response. This cross-beam configuration for the two photo beams further improved the response success rate. Still, analysis of video footage showed that not all responses were being registered.

The fourth and final prototype is identical to the previous version, except with a 16.0 mm diameter face plate (Figs. 1 and 2). After making the inside of the movable cylinder the same diameter as the hole in the face plate, and the outside diameter of the stationary core a consistent diameter, eliminating the step on the plug end, all issues were resolved. This assembly is our current version of the MICARP that has been in daily use since August 2010.

Figure 2.

Photographs showing the MICARP extended (a), half-way (b) and fully (c) retracted.

2.4. SAT in mice: acquisition and task parameters

The SAT consists of signal and non-signal trials and rewards the detection of signals (hits) as well as correct rejections during non-signal events (CR). Misses and false alarms, respectively, have no scheduled consequences other than the retraction of the operanda. Therefore, this task requires two operanda; extension of operanda into the chambers signals the response period. During non-signal trials, the operanda are extended into the chambers after the intertrial interval (ITI). The validity of measures of performance, in rats and humans, in terms of indicating sustained attention performance, as well as psychometric characteristics of SAT performance, have been described previously [12,46]. The distractor version of this task (dSAT) is of particular importance as this manipulation taxes the control of attention, via activation of cortico-mesolimbic pathways that in turn enhance cholinergic attention systems [14,46,56]. Typically, the distractor is presented in the middle 2–3 blocks of trials @ 8 min, out of a total 5–8 blocks per session. Major parameters used during the different stages of acquisition training, and the number of days requires to reach criterion for each stage are detailed in Table 1.

Magazine training

Procedures used to train and maintain SAT performance in mice correspond with those previously detailed for rats [e.g., 12,42,56], except for several modifications concerning the training of the task. Mice were first acclimated to the operant chambers and practiced retrieval of rewards. With MICARPS extended for one or two sessions (M; SEM: 1.4± 0.3 sessions), all nose-pokes resulted in a the delivery of reward (6 μL of 0.2% saccharin water) until at least 100 rewards were delivered within a 40-min period.

Acquisition Stage 1

Mice were habituated to moving MICARPs. Both MICARPs were extended until a nose-poke triggered retraction. Animals were required to retrieve the reward before MICARPs were extended again into chambers. To move to the next stage, animals were required to generate 90 rewards/session and retrieve at least 80% of these rewards within 1.5 s after the nose poke for two consecutive sessions/days.

Acquisition Stage 2

Mice were then trained to discriminate signal from non-signal events. Half of the mice were trained to report a hit on the left and the other half on the right MICARP. The session began by the random selection of a signal or non-signal event (MICARPs retracted). During signal trials, the signal light was turned on and both MICARPs extended 1 s after signal onset. During this stage of training, the signal light remained turned on for 4 s or until the animal nose-poked within the 4-sec MICARP extension period. During non-signal trials, the MICARPs were extended and likewise remained active for 4 sec. Failure to respond resulted in MICARP retraction (ITI: 12± 3 s) and this was counted as an error of omission. During this stage of training, incorrect responses triggered the usual retraction of the MICARPs but also repetition of this particular trial, for up to three times. If the 4th trial again resulted in an incorrect response, a forced-choice trial extended only the correct MICARP for 90 s. This transient procedure serves to abolish the manifestation of side - or single MICARP-biases. After three consecutive days of stable performance, defined as at least 60% hits and 60% CR, and less than 33% omissions, animals moved on to the next training stage.

Acquisition Stage 3

During this stage, signal duration was shortened to 1 s and the delay between signal/non-signal onset and MICARP extension was reduced to 0.5 s. Correction trials continued. Criterion performance was identical to Stage 2.

Acquisition Stage 4

Correction trials were removed and variable signal durations were implemented (0.5, 0.05, and 0.025 s). Note that the implementation of variable signal durations serves to prevent the development of fixed detection threshold and thus to increase the processing load imposed by the task [32,47–50]. Signal duration-dependent hit rates per se are not sufficient to deduce task validity in terms of sustained attention, or attention in general, as has been frequently suggested in the literature. After five consecutive days of stable performance, defined as at least 60% hits and 60% CR, and less than 33% omissions, animals moved on to the final stage of the task.

Acquisition Stage 5

The final stage of the acquisition training was similar to Stage 4 except that the house light in the back panel is illuminated throughout the session. This final step represents the primary manipulation that requires that animals sustained their attentional focus toward the signal source. As this manipulation results in a drastic, transient loss of performance, researchers occasionally removed this step, reducing the task largely to a mere discrimination task. Prior to Stage 5, animals appear to be able to detect signals even if oriented away from the intelligence panel and while engaged in competitive behaviors, such as grooming and exploration. As a result of this final, major step, such competitive behaviors decrease in frequency as animals are required to continuously monitor the signal source. Mice require approximately 10–12 days of practice in this final stage of acquisition to reach performance levels comparable with those observed prior to illuminating the houselights. Animals were required to reach at least three consecutive days of stable performance, defined as at least 60% hits and 60% CR, and less than 33% omissions, before being introduced to the distractor version of this task (dSAT) or other experimental manipulations.

dSAT

The distractor version of the Sustained Attention Task (dSAT) has emerged as a major translational tool in animal and human research on the cognitive symptoms of schizophrenia [12–14,37,46,54]. Generally, the distractor is hypothesized to tax top-down control of attentional performance; neuronal mechanisms involved in such top-down control include prefrontal regions and activation of mesolimbic-cholinergic interactions [6,14,56]. During the dSAT, the house light and panel light above the signal light flashed asynchronously at 0.5 Hz in the middle portion (minutes 8–24) of the 40 minute task (the upper panel light is off during SAT periods). Following a dSAT session, animals were required to regain SAT criteria performance for three consecutive days before being eligible for another dSAT session.

2.5. Measures of performance and performance analyses

Responses were recorded as hits, misses, correct rejections, false alarms, and omissions. The relative number of hits (hits/hits+misses) was calculated for each signal length, and the relative number of correct rejections (correct rejections/correct rejections+false alarms) and omissions (number of trials omitted / total number of trials) were also calculated.

In addition to the analysis of these primary measures of performance, we also calculated an overall measure of accuracy that collapses signal and non-signal trial accuracy into a single score (SAT/dSAT score) and the bias measure B″_D [15] that is derived from signal detection theory and indicates shifts in the subject’s propensity to report more or less signals under conditions of high “noise” (more liberal versus more conservative bias, respectively). Accordingly, changes in bias are thought to reflect alterations in top-down control of performance.

The SAT/dSAT score is derived from the Sensitivity Index [22] except that the SAT/dSAT score is based on the relative number of hits and false alarms, as opposed to the probabilities for hits and false alarms, and thus is not confounded by errors of omission. The SAT / dSAT score is calculated using the formula: SAT / dSAT score = (hits – false alarms) / [2(hits + false alarms) – (hits + false alarms)²]. This score ranges from +1.0 to −1.0, with +1.0 indicating that all recorded responses were hits or correct rejections and – 1.0 indicating all recorded responses were misses or false alarms. A score of zero indicates complete loss of discrimination between signal and non-signal events, or random response selection. Scores were calculated for each signal duration (SAT_500,50,25) and, where appropriate, averaged over all durations (SAT/dSAT).

B″_D was calculated as follows: B″_D = [(1 – hits)(1 – false alarms)] – (hits * false alarms)] / [(1 – hits)(1 – false alarms)] + (hits * false alarms)]. No bias is indicated by a value of zero. Negative numbers represent a liberal bias towards the signal-associated MICARP, positive numbers represent a conservative bias, and the maximum in either direction is 1.0. All measures were calculated for the entire 40 min task as well as for each of five task blocks (8 min each).

ANOVAs were primarily used (mixed design and repeated measures) to determine main effects and interactions; T-tests were used as appropriate. All statistical analyses are explicitly described in each section. The Greenhouse-Geisser sphericity correction was applied as needed. Alpha was set at 0.05. Corrected F and p values are reported. Least significant difference pair-wise comparisons were conducted for post hoc analyses unless otherwise described. As was suggested [24], exact p-values are reported. Statistical analyses were performed using SPSS Version 14.0 for Windows (SPSS Inc, Chicago, IL).

3. Results

3.1 Baseline SAT performance in mice

Of the 38 mice which began SAT training, 36 (94.74%) reached final SAT criterion. The other two mice were excluded from the study after not progressing past Stage 2 (see Methods), after being trained on this stage for over 6 weeks (see Table 1 for sessions/days to criterion for each acquisition stage). The experiments described below were conducted using separate subgroups of animals (see df in ANOVA results) in order to generate evidence devoid of confounds resulting from multiple manipulations of task parameters.

Initially we assessed performance using signals lasting 5, 3, and 1 s, and 1, 0.5, and 0.1 s, neither of which yielded duration-dependent hit rates (Fig. 3; see Methods concerning the significance of duration-dependent hit rates). Signal durations that matched those used routinely for SAT testing in rats, 0.5, 0.05, and 0.025 s, generated significant duration-dependent hit rates (Fig. 3c; F(2,14)=13.54; p=0.001). Pairwise comparisons revealed that longest signals generated significantly more hits than medium and shortest signals (both p<0.004). Animals correctly rejected 67.73±3.15% of non-signal events. Neither hits nor correct rejections varied across the 5 blocks of trials (8 min each; both F<2.17; both p>0.09). The baseline SAT score, averaged over all signal durations, was 0.28±0.05 (SAT₅₀₀: 0.53±0.04; SAT₅₀: 0.29±0.04; SAT₂₅: 0.18±0.10) and did not vary over blocks of trials (F(4,36)=1.91; p=0.13; see also Fig. 3d). Finally, animals omitted 17.97±5.37% of all trials and this again did not vary across the session (main effect of block: F(4,36)=0.72; p=0.58). Likewise, omissions did not differ by trial type (signal versus non-signal trials; F(1,7)=1.34; p=0.28).

Figure 3.

Relative number of hits to 3 groups of signal durations, 5-1 s (a), 1-0.1 s (b), and 0.5-0.025 s (c). Signal duration-dependent hit rates (see Methods for significance) were found only for the shortest set of signal durations. Coincidently, this set corresponds with the durations used routinely to test SAT performance in rats (*; p<0.05 based on a significant ANOVA and post-hoc comparisons; see Results). (d) shows the SAT scores, calculated for each signal duration and over blocks of trials.

3.2. SAT performance reliability

We selected performance data obtained from the first criterion session and from sessions that occurred 5 and 10 sessions/days later, to determine performance reliability. ANOVA of SAT scores indicated the absence of an effect of session and of interactions between the effects of session and block (all F<1.31; p>0.28; SAT scores averaged over all animals and signal duration and sessions: session 1: 0.25±0.03; session 2: 0.29±0.043; session 3: 0.26±0.025). The reliability of mice performing this task is also indicated by Cronbach’s alpha, 0.84 (calculated over averaged SAT scores from the three sessions). Thus, the reliability of the performance of mice corresponds well with that of rats (0.83), both performing slightly less reliable than humans (0.93), as was reported earlier [46].

3.3. Increasing water deprivation levels

The relationships between levels of deprivation and performance in operant tasks are complex and by no means are linear. In rats performing the SAT, Echevarria and colleagues showed that longer pre-session water access periods robustly increased errors of omissions but had relatively little effects on SAT response accuracy [16]. We assessed the performance effects of variable post-session periods of water access (4, 8 and 15 min) on SAT performance in mice. Performance during the subsequent session (22–23 hours later) was analyzed.

Post-task water access time (4, 8, 15 min) did not affect performance accuracy (main effect on SAT score: F(2,7)=0.27; p=0.78). As always, SAT scores varied by signal duration [F(2,14)=4.92; p=0.024] but the degree of water deprivation did not interact with this effect (F(4,14)=1.53; p=0.25). The relative number of omissions increased as a function of water deprivation (F(9)=36.71; p<.001), from 10.28±0.77% (4 min), to 37.64±8.40% (8 min) and 79.52±7.68% (15 min; average number of trials per session: 160.30±3.18 trials). Regardless of the duration of water access, errors of omission increased during the last blocks of trials compared with the initial 3 blocks of trials (F(4,28)=4.66; p=0.03; block 1: 31.78±1.781%; block 5: 51.63±4.10%; p=0.04 for block 5 against 1,2 and 3). There was no interaction between the effects of water access duration and block (F(8,28)=2.15; p=0.07), indicating that the increases in omissions that resulted from longer water access periods did not interact with the demands for continuous attentional performance.

3.4. Performance decrement resulting from increased event rate

A key feature of the construct of sustained attention concerns the demonstration of performance decrements over time. “Attentional fatigue” is thought to reflect the exhaustion of top-down efforts to sustain task compliance, including the suppression of competitive behavioral and cognitive abilities and to maintain the task rules in working memory [e.g., 3,48]. As the SAT involves rules governing the responding in two separate trial types, and as the dSAT version imposes particularly significant demands on top-down control, performance readily declines over time (or across blocks of trials) as a result of task-parametric, behavioral, or neurobiological manipulations. Increasing the rate of events (signal or non-signal in the SAT) represents a “classic” method to foster decrements in performance and thus assists in validating the SAT [e.g., 8,10,12,16,42].

Animals (n=10) were tested in sessions with ITIs of 9±3 s (high event rate) and 15±3 s (low event rate). The order of exposure to different event rates was counterbalanced and separated by at least three SAT sessions using the standard event rate (12±3 s). Performance effects were analyzed using the overall performance measure, the SAT score (see Methods).

A repeated measures ANOVA using the factors of event rate (high, low) x block (1–5) revealed main effects of event rate (F(1,9)=7.69; p=0.02) and block (F(4,36)=4.18; p=0.007), and an interaction between both factors (F(4,36)=2.91; p=0.035). Post-hoc ANOVAs indicated the absence of an effect of block for the low-event rate condition (F(4,36)=0.32; p=0.86) and an effect of block for the high event rate condition (F(4,36)=9.68; p<0.001). Multiple LSD pairwise comparisons indicated that compared with blocks 1–4 of the high event rate condition, SAT scores were significantly lower during the 5th block of trials (all p<0.04; Fig. 4). The errors of omission did not differ across event rate conditions and block (main effects and the interaction: all F<3.93; all p>0.08).

Figure 4.

Performance decrement as a result of a high rate of signal and non-signal events. Event rate was manipulated by modifying the intertrial interval (high event rate: 9±3 s; low: 15±3 s or 4 events/min). As predicted by theory and consistent with data from humans and rats, the performance of mice in the high event rate version of the SAT declined toward the end of the session (*, p<0.05 based on significant ANOVA and multiple comparisons; see Results).

3.5. Performance changes as a result of trial-type probability and outcome manipulations

SAT performance by rats and humans is sensitive to changes in the probability of particular trial type (equal probability for a signal and a non-signal event in the standard SAT) and in the amount of reward obtained for detecting a signal versus rejecting a non-signal event [12]. As, for example, the probability for signals decreases and the outcome for a hit decreases, the subjects’ performance would be expected to become biased toward reporting a correct rejection. Such a bias would not be expected to emerge if performance was considered to be based strictly on bottom-up mechanisms, meaning that the mere presence or absence of the sensory signal controls response accuracy. If such biases emerge they reflect changes in the subject’s response criterion that result from monitoring trial probabilities and response outcomes values. Thus this experiment was designed to test the sensitivity of SAT performance of mice to such manipulations of the top-down control of performance. Following a regular SAT session (session #1), the probability for a non-signal event was increased from p=0.50 to p=0.67 and the probability for a signal event decreased from p=0.50 to p=0.33. At the same time the reward amount for a correct rejection was increased from 6 μL to 8.04 μL, while hits earned 3.96 μL. The effects of this compound manipulation was assessed for two sessions (#2 and 3).

This manipulation decreased the relative number of hits (F(2,10)=17.22; p=0.006; averaged over durations: session 1: 68.46±0.95%; 2: 46.35±2.21%; 3: 47.42±5.81%) and increased the relative number of correct rejections (F(2,10)=15.08; p=0.001; session 1: 67.41±1.64%; 2: 82.09±2.61%; 3: 84.84±2.00%). This effects indicate that SAT performance by mice, similar to rats and humans, is readily modified as a result of changing event probabilities and trial outcomes.

3.6. dSAT performance in mice

dSAT performance was assessed by flashing both the houselight and the panel light that was placed above the signal source asynchronously on/off at 0.5 Hz during the 2nd and 3rd of a total of 5 8-min blocks of trials. In the rat version, only the houselight was used as a distractor. Pilot studies indicated that, in mice, a single distractor source produced less robust and less reliable effects on performance. The analysis of SAT/dSAT scores indicated main effects of block (F(4,40)=11.65; p<0.001), signal duration (F(2,20)=12.52; p<0.001) and an interaction between the two factors (F(8,80)=2.14; p<0.04; Fig. 5a). Post-hoc comparisons indicated that overall performance during the pre-distractor block 1 was higher than during any other block of trials, and that the distractor impaired performance in block 2 when compared with block 1, and in block 3 when compared with all other blocks (all p<0.04). Thus, performance during post-distractor blocks recovered, but not completely when compared with the pre-distractor level of performance. The interaction between block and signal duration reflected the absence of distractor effects for performance involving shortest signals (“floor” effect), while scores from the distractor blocks were lower when dSAT scores were calculated based on hits to longer signals and compared with block 1 (p<0.03; Fig. 5a). The results of the analysis of hits (Fig. 5b) corresponded precisely with the analysis of SAT/dSAT scores, indicating that the distractor suppressed the detection of the two longer signals only (all main effects and interactions: p<0.04). Likewise, the distractor impaired the mice’ performance during non-signal trials (F(4,40)=12.41; p<0.001; see Fig. 5c for multiple comparisons). Importantly, the distractor did not increase the errors of omission (F(4,40)=1.18; p=0.34; 16.56±3.58%).

Figure 5.

dSAT performance of mice ((a): SAT/dSAT; (b): hits; (c): correct rejections; cr). The period during which the distractor was presented (2nd and 3rd block of trials) is illustrated by the transparent (pink) rectangle. The distractor suppressed the detection of longer but not shorter signals (“floor” effect) and correct rejections. Post-distractor performance largely but incompletely recovered during post-distractor periods (see Results for ANOVAs and multiple comparisons for SAT/dSAT scores and hits; results of multiple comparisons of correct rejections are illustrated in (c); *, p<0.05). (d): Relative number of hits, averaged over all signal durations, of rats and mice performing the dSAT. The presentation of the distractor in mice and rats produced statistically similar effects as indicated by the absence of an effect of species and of interactions between the effects of species and block with respect to SAT/dSAT scores, hits, and correct rejections (see Results). (e): The analysis of the bias measure indicated that both species adopted a more liberal criterion in response to longest signals when compared to shorter durations. The distractor did not differentially affect the biases of the two species (not shown; see Results).

3.7. Species comparisons: mice versus rats

SAT performance did not differ between mice and rats (signal duration: 0.5, 0.05, and 0.25 s; main effect of species on SAT score: F(1,11)=0.59; p=0.46; species x signal duration: F(2,22)=0.21; p=0.81). Analysis of hits and misses likewise did not indicate an effect of species (all F<1.56; all p >0.23). However, mice omitted more trials than rats (F(1,11)=6.58; p=0.03; mice: 16.21±3.58%; rats: 1.39±4.53%).

To compare dSAT performance of rats and mice, a mixed design ANOVA for species (rat, mouse) and block (5 levels) indicated a significant effect of block, reflecting the distractor effects as described above (F(4,56)=12.32; p<0.001) but no effects of species (F(1,14)=1.42; p=0.09) and no interactions involving species as a factor (all F<1.21; all p>0.09). The analysis of the individual measures of performance, hits (Fig. 5d) and correct rejections, also failed to indicate an effect of species. However, mice again omitted more trials than rats (main effect of species: F(1,14)=9.03; p=0.009) but this effect did not interact with the effects of the distractor (F(4,56)=0.48; p=0.75).

Bias scores were analyzed to reveal potential species-specific strategies during SAT performance and to respond to the distractor challenge. In the absence of a distractor (SAT), signal duration affected bias in both species (main effect: F(2,22)=14.83; p<0.001; Fig. 5e). Both mice and rats adopted a more liberal bias in responding to longest signals when compared with the two shorter durations (Fig. 5b). However, this effect was not modulated by species (main effect: F(1,11)=0.88; p=0.37; interaction: F(2,22)=0.36; p=0.70). The presence of the distractor likewise did not reveal species-specific biases (species: F(1,11)=2.878; p=0.12; signal duration: F(2,22)=12.42; p<0.001; interaction: F(2,22)=0.11; p=0.90).

Finally, the reliability of the distractor effect in mice was determined by comparing performance during 3 dSAT sessions that were separated by at least 3 SAT sessions. Repeated dSAT exposure did not yield different levels of performance or distractor effects (main effect of sessions and all interactions involving sessions: F<2.18; p>0.79).

4. Discussion

The evidence described above collectively indicates that mice can be readily trained in, and then perform stably, the SAT. Manipulations of event rate, trial type probability, and trial outcome resulted in significant changes in performance, reflecting the sensitivity of mice to manipulations that act top-down to alter SAT performance. Furthermore, the distractor version of the task generates robust impairments in performance that are followed by recovery of performance during the post-distractor period. We also found that SAT and dSAT performance levels of mice correspond with those observed in rats, and that the two species perform comparably using largely corresponding task parameters, including signal durations.

These findings challenge the often implicit assumptions that the behavior of mice is dominated by internal programs and that an almost compulsive propensity for exploration and grooming behavior interferes with the ability to sustain cognitive performance. We need to note, however, that in order to generate corresponding levels of performance of rats and mice, the standard event rate for mice is lower than in rats (approximately 1.3 events/min less in mice), presumably reflecting that rats are more effective in suppressing competitive behaviors than mice and thus can tolerate faster event rates. Furthermore, the distractor stimulus required for mice involves two sources flashing asynchronously (see above); we may speculate that mice require a stronger distractor because of their lower propensity to process unconditioned, non-threatening stimuli. These modifications are not in conflict with the conclusion that mice are capable of sustaining attention [see also 31,57] and deploy top-down control mechanisms for that purpose.

In contrast to our prior difficulties to generate robust and stable levels of performance using retractable levers [see also 40], the introduction of the MICARP was sufficient to yield a high acquisition success rate in mice (with over (>90% reaching the final performance criterion), within about 2 months of training, and to generate high and stable levels of performance. The main properties of the MICARP, such as relatively slow extension speed, the diameter of the opening and the speed and timing of the opening and closing of the nose-poke access (see Methods) all may contribute to the efficacy of this device in terms of serving as a cue for indicating the on- and offset of the response period, while not interfering with demands for the sustained monitoring of signal status.

The final signal durations used in mice to generate duration-dependent hit rates correspond with those employed in rats. To reiterate, duration-dependent hit rates per se are not a sufficient marker of the validity of performance in terms of reflecting sustained attention (see also Methods). Although we can only speculate about the cognitive-perceptual mechanisms yielding lower hit rates to shorter stimuli, it is unlikely that they are a primary result of sensory processing limitations. First, given the simple on-off characteristics of the signals used in this task, these signals may preferably be processed via collicular rather than geniculo-striate pathways. As a result, it seems unlikely that shorter signals are missed because their sensory processing is compromised by early filtering [e.g., 7,23]. Second, as indicated in Figure 5e, the animals’ decisions to score a hit were associated by sizable duration-dependent shifts in bias, consistent with the general view that cognitive-decisional processes dominate the animals’ response selection. With shorter signal duration, an increasingly conservative bias may cause an increasing propensity for reporting misses.

Extended post-task access to water increased the errors of omissions during next day’s session. However, residual performance levels were unaffected. This finding suggests that motivational saturation reduces task compliance, presumably by relaxing the top-down mechanisms that sustain prolonged task engagement in less saturated animals. However, it is important to note that the reverse may not be necessarily true: if animals omit more trials in response to task challenges or neuronal manipulations, this may reflect capacity limitations in sustaining attention that are not necessarily confounded by a loss of motivation. The interpretation of effects on errors of omissions is rarely straight forward. It is in part for this reason that the preferred calculation of SAT/dSAT scores is not contaminated by errors of omission. It is also noteworthy that our collective evidence from mice (this report) and rats [e.g., 33,34,43,56] indicates that the impairments or enhancements of SAT/dSAT performance that result from neurobiological manipulations, including disease modeling, are rarely dominated by, and often not even associated with, changes in the number of errors of omissions.

Task validation is an ongoing process which reveals the complex cognitive mechanisms underlying SAT and dSAT performance (see Introduction). The effects of the high event rate (Fig. 4) confirms just one aspect of the task’s validity in mice, that is, the role of sustained attention. Higher event rates tax the ability to continue monitoring the signal source over extended periods of time [47–49]. In mice, performance during a session with a relatively high rate of signal and non-signal events was characterized by a steep decline in performance in the late block of trials, eventually resulting in random response selection (SAT scores near zero). This finding is consistent with the hypothesis that the SAT assesses sustained attention performance in mice, as it does in rats and humans [12,16,42].

In conclusion, the present evidence indicates that, by using the MICARP as an operandum in mice, the SAT/dSAT, already in use for testing humans and rats, can be readily employed for research using mice. Performance levels in mice are highly comparable to those observed in rats. The reliability of performance over days and weeks, including performance during repeated distractor exposure, is high. This task is hoped to assist the field in expanding research on the neuronal mechanisms of attention and drug finding research to mouse models of neuropsychiatric and neurodegenerative disorders.

Highlights.

• Design and validation of a task for assessing sustained attention task (SAT) in mice.

• Design of retractable nose poke device (MICARP) as an operandum.

• The distractor condition of the SAT (dSAT) taxes top-down control mechanisms.

• Performance levels, including distractor effects, are comparable to rats.

• These methods and results expand the translational usefulness of the SAT/dSAT.