Isoflurane at subhypnotic doses is known to affect cellular and network activity in the auditory pathway, but the behavioral effects of these concentrations of isoflurane on auditory processing have not been tested previously. The authors tested the hypothesis that subhypnotic doses of isoflurane would impair auditory discrimination in rats.
Rats were tested on their ability to discriminate up versus down frequency-modulated sweeps using three different pairs of sweeps ("Long," "Med," "Short"), whose frequency range and duration were varied systematically to make the discrimination more difficult. Rats were tested daily in the absence and presence of isoflurane at 0.2% or 0.4%. The effects of isoflurane (0%, 0.2%, and 0.4%) on performance (= % correct) and efficiency (= time/trial) were assessed using regression analysis.
The effect of isoflurane was stimulus-dependent: performance for the Long stimulus pair was unaffected by isoflurane, performance on the Med stimulus pair was impaired only by 0.4% isoflurane, and performance on the Short stimulus pair was impaired by both 0.2% and 0.4% isoflurane. In contrast, isoflurane decreased efficiency equally for all stimulus pairs at 0.4% and had no effect at 0.2%.
The stimulus dependence of the effect of isoflurane on performance suggests that it is unlikely this effect was secondary to effects on memory, motivation, or motor function. These data indicate that doses of isoflurane known to produce modest effects on neural responses alter cortical sensory processing.
THE effects of general anesthetic agents on neural activity have been extensively tested over the past several decades, and we now have a basic understanding of the molecular targets and cellular effects of a wide array of these agents on spinal cord, brainstem, midbrain, and cortical structures.1–3The functional effects of this modulation of neural activity have been less well studied, and in particular, little is known about the impact of modulatory effects of general anesthetics on the detection and discrimination of sensory stimuli. This study was designed to test the behavioral effects of the volatile anesthetic isoflurane on sensory processing in the auditory system.
At doses near and exceeding those required to cause loss of consciousness, modulation of auditory evoked responses by anesthetic agents is predictive of the level of consciousness in patients4and can be used to monitor anesthetic depth perioperatively.5,6Previous studies have also shown that even subhypnotic doses of general anesthetics modulate neural activity in the auditory system.7–9We are interested in whether there are any functional consequences of this modulation in the absence of loss of consciousness. For example, similar neural effects at subhypnotic doses of anesthetic agents in hippocampus have been linked to impairment of episodic memory formation.10–12These data suggest that low concentrations of anesthetic agents may interfere with processing of acoustic stimuli, but surprisingly, this has never been tested experimentally.
This issue has implications for whether sensory coding by cortical networks is robust to moderate changes in synaptic strength and excitability. In rats, volatile agents modulate cortical evoked responses and inhibitory postsynaptic currents at concentrations less than 0.5 minimum alveolar concentration.9,13,14But the effects are small for these doses of volatile agents and even for much larger concentrations of benzodiazepines.9,15Anesthetic agents are becoming popular tools for probing the neural basis of behavior because they provide a calibrated means to perturb neural activity in an experimental setting. But how do we know when an observed effect of an anesthetic agent on neuronal responses is important behaviorally? Interpreting the functional implications of these changes requires detailed knowledge about the corresponding behavioral effects of anesthetics, which is lacking for sensory systems.
In this study, we tested the hypothesis that subhypnotic doses of isoflurane cause impairment in auditory discrimination. We measured the performance of rats on an auditory discrimination task in control conditions and at two doses of isoflurane (0.2% and 0.4%). We presented pairs of stimuli identical in spectral content but differing in their temporal dynamics: upward- versus downward-going frequency-modulated sweeps. Frequency-modulated sweeps are common features in communication and other sounds produced by rodents and many other species.16–18Neurons in the ascending auditory pathway preferentially respond to sweeps of a specific direction and slope,19,20and lesion studies indicate that specific neural structures are critical for an animal's ability to discriminate between frequency-modulated sweeps.21,22We found that isoflurane caused a significant reduction in task performance and that the effect of isoflurane on performance was correlated with task difficulty. These data demonstrate that doses of isoflurane previously shown to produce modest effects on neural activity have detectable effects on cortical sensory processing.
Materials and Methods
Discrimination Training and Testing
All experimental protocols conformed to American Physiologic Society/National Institutes of Health guidelines and were approved by the University of Wisconsin Research Animal Resources Committee (Madison, Wisconsin).
Seven female Harlan Sprague-Dawley rats (6–8 weeks old) were trained to discriminate between two auditory stimuli using restricted water access as motivation. Electrophysiologic and anatomical properties of cells in the auditory pathway are largely mature by 3–4 weeks,23–25although some maturation of response properties may continue up to 12 weeks.26Because the training phase continued for 12 weeks (see appendix), animals were fully mature before the testing phase of this study. The rats were trained to associate a particular sound with one of two water sources. After training, isoflurane was introduced to determine its effects on performance. Therefore, decreased correct responses as well as increased time for the rat to make the correct response were indicators of poorer performance. A trial consisted of presentation of one of two stimuli and a response by the animal in which they needed to indicate which of the two stimuli were presented. The rats initiated trials and indicated their responses by breaking optical beams with their snouts, and were rewarded with drops of water upon correct responses. All training and testing was performed in the dark.
The task was performed in a soundproof chamber (Industrial Acoustics Company, Inc., Bronx, NY). Inside the soundproof chamber, the training apparatus was contained in a homemade gastight acrylic enclosure (length × width × height = 60 × 30 × 15 cm) that had gas inflow and outflow ports for administering and scavenging isoflurane and a gas sampling port for monitoring the isoflurane concentration using a gas monitor (Multigas Monitor 602; Criticare Systems, Waukesha, WI). The enclosure contained a center nose poke port and left and right water dipper ports (H14-05M; Coulbourn Instruments, Allentown, PA). Three pairs of optical position sensors (EE-SPW311; Omron Electronics, Schaumburg, IL) were placed in the center port and two dipper ports to monitor the position of the animal. A small speaker (TDT-ES1; Tucker Davis Technologies, Alachua, FL) was mounted inside the enclosure, oriented toward the animal. The speaker was calibrated using a microphone (No. 4016; ACO Pacific, Inc., Belmont, CA) placed approximately 4 cm from the speaker, and stimuli presented at approximately 75 dB sound pressure level assuming the animal's head was this distance from the speaker. Because the animal was unrestrained, actual stimulus levels on each trial varied slightly, but it was assumed that this variability was averaged out over many repetitions of the stimulus. Speaker output varied by ±3 dB sound pressure level over the range 10–20 kHz. Free-field stimuli were applied using commercial (RPVDX; Tucker-Davis Technologies) and custom software written in Matlab (Mathworks, Inc., Natick, MA; see appendix). Details of the procedures used for training the animals appear in the appendix.
After achieving criterion performance, rats were tested on discrimination performance using three pairs of stimuli, the original pair (“Long”; see appendix) and two additional pairs: “Med,” with fLow= 12.5 kHz, fHigh= 17.5 kHz, tDur= 125 ms, and “Short,” with fLow= 14.6875 kHz, fHigh= 15.3125 kHz, tDur= 15.625 ms. These three pairs of stimuli share the same slope in the time-frequency domain. The stimulus pairs were chosen to represent three degrees of discrimination difficulty: “easy,”“medium,” and “hard,” with the longest stimuli the easiest to discriminate. To determine the effect of isoflurane on discrimination, rats were tested under three drug conditions: 0%, 0.2%, and 0.4% isoflurane in room air. Before testing, rats were preequilibrated with the drug in a small acrylic enclosure for 15 min before being placed in the training enclosure, which had itself been preequilibrated with isoflurane. It took typically less than 30 s to transfer a rat from the preequilibration chamber to the testing chamber, and the door to the testing chamber typically was open for less than 5 s. Small decreases (< 25%) in the concentration of isoflurane were sometimes observed on the anesthesia monitor after the transfer of the animal, but these always returned in less than 5 min to the desired concentration. These changes were unlikely to affect the data presented here because the first 5 min of each day's testing consisted of warm-up trials that were not included in the data analysis. On any given day during the testing period, a stimulus pair and isoflurane concentration were chosen randomly and the animal tested under those conditions. The animals were tested until they completed 200 trials or the elapsed time reached 45 min, whichever came first. The animals were allowed to recover at least 48 h between exposures to isoflurane. In all cases, the schedule for which isoflurane concentration was tested on which particular day was varied from week to week, so that animals could not anticipate the drug condition. In four animals, stimuli were chosen in succession, i.e. , the same stimulus pair was tested for approximately 30 consecutive training sessions while drug condition was varied daily, and then the next stimulus was selected and tested for approximately 30 consecutive training sessions while drug dose was varied, and so on. In the other three animals, stimuli and drug dose were varied in tandem, i.e. , on a given day a different stimulus pair and drug dose were selected, and this process repeated for 60 training sessions, with the schedule varied from week to week so that the animals could not anticipate the stimulus pair or drug condition. At least six repetitions of each testing condition were obtained for each animal.
Statistical Analysis of Behavioral Data
For each testing session, two dependent variables were measured: performance = % correct response and efficiency = time to complete the session ÷ No. of trials. There were nine test conditions, i.e. , three stimulus pairs (Long, Med, and Short) × three drug concentrations (0%, 0.2%, and 0.4%). The other two independent variables were animal and stimulus repetition number, i.e. , the number of times (+1) the animal had been tested previously on the particular stimulus pair. Each dependent variable was analyzed separately. The effect of stimulus on performance in the absence of isoflurane was analyzed using one-way analysis of variance, with stimulus pair as the within-subjects factor. To compare performance and efficiency at the beginning and end of the testing period in the absence of isoflurane, we performed two-way analysis of variance with week (week 1 vs. week 12 of the testing period) and stimulus as within-subjects factors. In both of these analyses of variance, the dependent variable was averaged over the relevant week for each animal, and the averaged values were used in the analysis of variance.
To determine the effects of isoflurane on performance and efficiency, we used a multiple linear regression analysis (General Linear Model) in SPSS version 12.0 (SPSS, Inc., Chicago, IL). To focus on the effects of isoflurane on performance and its interaction with other dependent variables, data were normalized as follows. For each animal and each stimulus pair, performance was averaged over the final five testing sessions in the control (i.e. , zero isoflurane) condition, and the performance data for that animal and that stimulus pair for all drug conditions and all stimulus repetitions were then normalized to this average performance. This procedure was then repeated for each stimulus pair and each animal. Multiple linear regression analysis was run using this normalized performance as the dependent variable. Some of the independent variables were categorical ([isoflurane], animal, stimulus), and thus, to do regression analysis, it was necessary for the software to recode these variables into dummy variables. Repetition number was treated as a continuous variable. In this analysis, a straight line is fit to the data in the form of y = k +∑ ai* Ai+ b * R + c1* I0.2+ c2* I0.4+ d1* SMed+ d2* SShort, where k is the y-intercept; ai, b, ci, and diare the regression coefficients; Ai, i = 1.6 are the dummy variables representing six of the seven animals; R is repetition number; I0.2and I0.4are the dummy variables representing two of the isoflurane concentrations; and SMedand SShortare the dummy variables representing two of the stimulus pairs. The dummy variables are equal to one for data points collected under those conditions (e.g. , I0.2= 1 when Iso = 0.2%) and are equal to zero otherwise. There is always one less dummy variable than the number of possible values in the category because when no dummy variable is present in the model for that category, it corresponds to the uncoded value of that category. This method is equivalent to fitting multiple lines to the data whose intercepts differ for the different values of the categorical variables. The coefficients for the values of the categorical variables omitted from the model (i.e. , Iso = 0%, Stim = Long, Animal = animal No. 7) are combined into the intercept variable k. A significant coefficient value for one of the dummy variables indicates that the line for the group of data points defined by that value of the categorical variable is significantly different from the Iso = 0%, Stim = Long, Animal = animal No. 7 group of data points, when the data are combined across all values of the other categorical variables (e.g. , a significant value for c1indicates that the data collected under 0.2% isoflurane are significantly different from 0% isoflurane, regardless of the stimulus and animal).
Interactions between pairs of dependent variables were tested separately by including their products as additional dependent variables in the regression model. The model then has the form y = k +∑ai* Ai+ b * R + c1* I0.2+ c2* I0.4+ d1* SMed+ d2* SShort+ e1* I0.2* SMed+ e2* I0.2* SShort+ e3* I0.4* SMed+ e4* I0.4* SShort, where eiare the coefficients for the interaction terms. Again, this is equivalent to fitting different lines to the data, but now there are different lines for each combination of values of the interacted variables. Significant coefficient values for the interaction terms indicate that for that specific combination of Iso and Stim values, the data (combined across animals) are significantly different from the Iso = 0%, Stim = Long case. The coefficient c1now indicates significance for only the Iso = 0.2%, Stim = Long case (regardless of animal) and similarly for c2, d1, and d2.
Effect of Stimulus on Performance
Over a training period of 12 weeks (60 sessions) using the Long stimulus pair, rats improved their performance from 55.2 ± 3.23% correct in the first week (mean ± SD, averaged across animals, with each animals' performance averaged across all repetitions in the given week; range, 48.2–58.9%) to 83.0 ± 5.18% correct in the final week (range, 75.9–90.9%; P < 0.0001, paired Student t test comparing the averages for each animal in week 1 vs. week 12; fig. 1). During the testing phase, which lasted an additional 12 weeks after the initial training period, animals were presented with Long, Med, or Short stimulus pairs under different drug conditions. In the absence of isoflurane, performance depended on stimulus duration and frequency range: Long, 87.0 ± 5.05%; Med, 70.8 ± 5.69%; Short, 63.8 ± 1.59% (mean ± SD, averaged across animals, with each animals' performance averaged across all repetitions; F2,5= 76.2, P < 0.00001, by analysis of variance, with stimulus as within-subjects factor). Bonferroni post hoc tests indicated that performance levels for each stimulus pair were distinct (P << 0.0001 for all comparisons). These data indicate three different levels of difficulty for the task using the three stimulus pairs.
To best illustrate the effects of isoflurane on task performance, we normalized the performance data for each animal and each stimulus to the animal's mean performance in the absence of isoflurane in the last 2 weeks of testing for that particular stimulus. We normalized to the mean over the last 2 weeks of testing because, due to the duration of the testing period, there was significant improvement in performance for each stimulus pair: Mean performance in the absence of isoflurane increased from 85.0 ± 7.20%, 69.4 ± 5.39%, and 63.0 ± 2.18% for Long, Med, and Short stimuli, respectively, in the first week of testing to 88.4 ± 5.00%, 71.4 ± 5.97%, and 64.9 ± 1.83% for the last week of testing. Two-way analysis of variance with stimulus and week (week 1 vs. week 12) as within-subjects factors showed a significant effect of week on performance (F1,6= 15.69, P < 0.01), and Bonferroni post hoc tests indicated significant effects for all stimulus pairs (P < 0.05). In contrast, task efficiency, measured as the average time required per trial in a given testing session, did not change over the testing period (first week: 9.48 ± 3.61, 7.29 ± 2.00, and 8.91 ± 4.00 s for Long, Med, and Short stimuli, respectively; last week: 11.7 ± 4.00, 7.05 ± 2.16, and 7.57 ± 1.81 s for Long, Med, and Short stimuli, respectively; F1,6= 0.121, P = 0.74).
Effect of Isoflurane on Performance and Efficiency
Isoflurane had little effect on task performance when animals were tested with the Long (i.e. , easiest) stimulus pair, but for the two stimulus pairs that were more difficult to discriminate, isoflurane produced a clear deficit in performance. This is apparent in both the time series representation of normalized performance as a function of session number (fig. 2) and normalized performance averaged over the entire test period (figure 3; this figure shows performance for each animal [colored symbols] averaged over all testing sessions for the Long stimulus pair [fig. 3A], Med stimulus pair [fig. 3B], and Short stimulus pair [fig. 3C] as a function of isoflurane concentration, as well as the normalized performance averaged across animals [large black squares]). In addition, the effect of isoflurane on performance with the Short stimulus pair was linear, with twice the effect at 0.4% isoflurane compared with 0.2% isoflurane (fig. 3C). A different pattern was observed with the Med stimulus pair: There was little effect at 0.2% isoflurane and a much larger effect at 0.4% isoflurane (fig. 2B). We used multiple linear regression analysis to assay the statistical significance of the effects illustrated in figures 2 and 3.
In the regression model, stimulus (Stim), isoflurane concentration (Iso), and animal were treated as categorical variables. To incorporate categorical variables into the regression model, one value of the variable was chosen as the reference value (Stim = Long, Iso = 0%, Animal = animal No. 7), and the remaining values were considered relative to this reference. Repetition number (RepNum) was treated as a continuous variable. We first present the results of the model with all independent variables considered independently and then present a more complete model in which interactions between independent variables are considered. Regression parameters are presented in table 1(model without interactions) and table 2(model with interactions).
Regression coefficients for isoflurane at 0.2% and 0.4% were significant in the model without interactions, their negative values indicating that performance was significantly worse in isoflurane than in the absence of isoflurane (table 1). The magnitude of the regression coefficient for 0.4% isoflurane is proportionally much larger than the coefficient for 0.2% isoflurane (−0.12 vs. −0.038), indicating that the higher dose of isoflurane had a proportionally much larger effect on performance, an issue to which we will return below. The coefficient for RepNum was also significant, and its positive value indicates that performance improved the more the animal was exposed to a particular stimulus pair (table 1). The regression coefficient for the Short stimulus pair was significantly different than the Long stimulus pair, indicating that normalized performance differed for the two stimulus pairs (table 1). As will be seen below, this is due to poor performance on the task with the Short stimulus pair in isoflurane.
It is apparent from figures 2 and 3that the effects of isoflurane were dependent on stimulus condition, and this was explored by incorporating into the regression model two-way interactions between Iso and Stim (table 2). In this form of the model, there are terms representing explicit interactions between isoflurane concentration and stimulus. Now the coefficients refer to specific values of Iso and Stim, and thus the stimulus dependence of the effect of isoflurane can be tested directly (see Materials and Methods). Interactions between RepNum and Iso and between RepNum and Stim were not significant (not shown) and were omitted from the model.
The dependence of the effect of isoflurane on stimulus pair is reflected in the significance of regression coefficients for the interaction terms. For the Short stimulus pair, the coefficients representing interactions between Iso and Stim (Iso * Stim) are significant at the P < 0.05 level for both Iso = 0.2% and Iso = 0.4%, whereas for the Med stimulus pair, the coefficient is significant for only Iso = 0.4%. The observation that the coefficients for the noninteracted Iso = 0.2% and Iso = 0.4% terms are not significant indicates that for the Long stimulus, there is no effect of isoflurane (table 2). These results are a quantification of the effects illustrated in figure 3, in which performance was impaired at both concentrations of isoflurane for the Short stimulus pair but only for the highest concentration of isoflurane for the Med stimulus pair. Note also that all of the effect of isoflurane is captured by these interactions between Iso and Stim: The coefficients for the noninteracted Iso parameters are not significant (table 2). In addition, unlike in the model without interactions, the noninteracted Stim = Short coefficient is also not significant, indicating that its significance in the model without interactions was due to the stimulus dependence of the effect of isoflurane.
As an independent measure of the effects of isoflurane on the animals' ability to perform the task, we also measured the time required to complete the testing session, which we then divided by the number of trials (usually 200) to yield a measure of efficiency. In the absence of isoflurane, efficiency values were typically approximately 7 s (fig. 4). The exception to this was for the first group of rats that was tested sequentially (i.e. , first the Long stimulus in varying drug conditions, then the Short stimulus, then the Med stimulus). For these rats, efficiency values of 10–15 s were typically observed for the Long stimulus pair, whereas their efficiency for the Med and Short stimulus pairs was comparable to the second group of rats tested with the stimulus pairs presented in parallel. We used a regression model with efficiency as the dependent variable similar to the model with interactions used for normalized performance (table 3). As suggested by the data in figure 4, isoflurane at 0.2% had no effect on efficiency, whereas isoflurane at 0.4% significantly decreased efficiency (i.e. , increased the time/trial value) compared with efficiency in the absence of isoflurane (table 3). There were no significant interactions between Iso and Stim. These data suggest that isoflurane at 0.4% did cause the animals to take longer to complete their trials in a given session, and more generally suggest that isoflurane at 0.4% affects the animals' motivation, motor skills, or activity level, consistent with the known behavioral effects of this agent. However, there are two important discrepancies between the effects of isoflurane on efficiency and on performance. First, there was no difference between the effects of isoflurane on efficiency observed for the Long stimulus pair versus the Med or Short stimulus pairs, even though the effect of isoflurane on performance was stimulus dependent. Second, there was no significant effect of isoflurane at 0.2% on efficiency, even though there were significant effects of 0.2% isoflurane on performance with the Short stimulus pair. These divergences between effects on performance and efficiency suggest that the observed effects on efficiency, which likely reflect the sedative and motor effects of isoflurane, are not causative of the effects observed on performance.
In rats trained to criterion performance on an auditory discrimination task, isoflurane caused a significant impairment in performance and decreased task efficiency. The effect of isoflurane on performance was unlikely due to an effect on learning, because there was no effect of isoflurane on the trajectory of improved performance over the course of the testing session, and the effect of isoflurane on performance was most pronounced for the most difficult stimulus pair.
Isolating Effects on Sensory Processing
The subhypnotic doses of isoflurane used in this study, 0.2% and 0.4%, may affect the animals' ability to perform the task independent of their effects on sensory processing, for example, due to effects on memory (e.g. , remembering which stimulus just occurred while deciding which dipper to approach, or memory for the association between sweep direction and the specific dipper), motor function, motivation, or attention. Our data indicate that effects on memory, motor function, and motivation are unlikely to underlie the impairment in performance, because there were some stimulus conditions (i.e. , the Long stimulus pair) under which isoflurane had no effect on performance even at 0.4%. In addition, there was no stimulus dependence to the effects of isoflurane on efficiency (fig. 4and table 3) in contrast to its effect on performance (fig. 3and table 2), arguing that although isoflurane affected the animals' ability and/or desire to perform the task, these nonspecific effects cannot explain the effects on performance.
By contrast, it is possible that the observed effects on discrimination are due to reduced capacity to attend to the stimulus in isoflurane. Selective attention enhances sensory processing, and this advantage is more pronounced for sensory stimuli near threshold.27There are no previous studies on the effects of isoflurane or other volatile agents on attention, but other anesthetic agents, such as ketamine, lorazepam, and diazepam, have been shown to impair performance on sensory tasks requiring selective attention,28–30although evidence suggests that this is not the case for nitrous oxide.31,32
There are no previous psychophysical studies of the effect of anesthetic agents on auditory sensory processing, and few studies in other sensory modalities. In addition to the aforementioned studies on ketamine, nitrous oxide, and benzodiazepines, there are two studies on humans that investigated the effect of low doses of isoflurane on performance in visual sensory tasks. In one study, isoflurane at 0.15% significantly decreased contrast sensitivity,33whereas in a second study, the effects of 0.4% isoflurane on accuracy in a visual search task were inconclusive, although reaction times were significantly increased.34There are several studies that have investigated the effects of subhypnotic doses of anesthetic agents on language processing, but these studies were concerned primarily with memory for words presented during the drug condition rather than word recognition per se .35Therefore, the data presented here represent one of the first detailed studies of anesthetic effects on performance on a sensory behavioral task, and the first study to our knowledge in an experimental animal.
We present evidence that isoflurane at subhypnotic doses produces deficits on a sensory discrimination task. Based on the dependence of the impairment on task difficulty and the independence of the effects on efficiency from task difficulty, we conclude that these effects are specific to sensory processing, due to either direct actions of isoflurane on the primary auditory pathway or indirect actions on this pathway secondary to effects on attention. Based on lesion studies showing that primary auditory cortex is necessary for discrimination of complex sounds such as frequency-modulated sweeps but not for sound detection and discrimination of simple sounds such as pure tones,21and based on studies of modulation of auditory evoked responses, which show that brainstem evoked potentials are largely intact at doses of anesthetic agent that alter cortical evoked potentials,13,14,36we speculate that the site of action of isoflurane in these experiments is primary auditory cortex. Future experiments combining electrophysiology and behavior will be able to address the neural locus of the behavioral deficits reported here. Our data do show that detectable impairments in sensory processing can occur at doses of isoflurane previously reported to have modest effects on neural responses.
The authors thank Jane Sekulski, B.S. (Senior Programmer, Department of Anesthesiology, University of Wisconsin, Madison, Wisconsin), Judi S. Bartfeld, Ph.D. (Associate Professor, Department of Consumer Science, University of Wisconsin), and Abigail Schuh, B.S. (Lab Assistant, Department of Anesthesiology, University of Wisconsin), for technical support, and Robert A. Pearce, M.D., Ph.D. (Professor, Department of Anesthesiology, University of Wisconsin), Misha Perouansky, M.D. (Associate Professor, Department of Anesthesiology, University of Wisconsin), Luis C. Populin, Ph.D. (Assistant Professor, Department of Anatomy, University of Wisconsin), and Philip H. Smith, Ph.D. (Professor, Department of Anatomy, University of Wisconsin), for comments on the manuscript.
Appendix: Behavioral Training Procedures
Naive rats were put on restricted water access for 1 week before commencement of training. During this period, the animals' access to water was gradually reduced to 20 min daily. Animals were weighed daily to ensure they received enough water, and if their weight changed by greater than 5%, their water access was adjusted accordingly. Once training commenced, water was provided as a reward during training and at one other period during the day for 3–5 min, depending on how many trials were run and the animal's weight gain trajectory.
Training proceeded in three stages; during the first two stages, animals usually were shaped automatically, with occasional manual assistance (see below). The third stage was fully automated. Stage I was designed to familiarize the rats with the location of the water dippers and to accustom the rats to poking their noses in the dipper ports to receive their reward. This was done by initially reinforcing the rats with a water reward (10 μl) any time they approached either dipper, and then gradually delaying the water reward until the rats were required to place their noses completely inside the dipper port before reinforcement. Stage I was accomplished over 2 days. Stage II was designed to familiarize the animals with the center port, and to associate the center port with initiating a stimulus presentation. On the first day of stage II, one dipper port was hidden, and the animal received a reward when it placed its snout in the center port and went immediately to the unblocked water dipper port. If the rat had difficulty with this stage, manual assistance was provided by the experimenter. The manual program allowed for the experimenter to lure the rat closer and closer to the nose port, while reinforcing with dipper rewards until the rat was repeatedly placing its snout completely in the nose port. At this point, automatic training resumed. On day 2, this procedure was repeated with only the other dipper port unhidden. Day 3 was similar to day 1, but an FM sweep was played when the animal placed its snout in the center port, and on day 4, the same procedure was repeated but with only the other dipper port unhidden. During stage III of the training process, both dipper ports were available, and the training apparatus was controlled by software written in Matlab version 7.2 (Mathworks, Inc., Natick, MA) that interfaced with the training apparatus by programmable hardware devices (System III; Tucker Davis Technologies, Alachua, FL). The animal initiated a trial by placing its snout in the center port and breaking the optical sensor beam, and after a variable hold period of 1,500–2,000 ms, the computer randomly selected one of two stimuli that were played over the single speaker. The stimuli were linear frequency-modulated tone sweeps, either sweeping upward (fLow→fHigh; FMup) or downward (fHigh→fLow; FMdn), with duration = tDurand 5-ms rise/fall times. During the training phase, fLow= 10 kHz, fHigh= 20 kHz, and tDur= 250 ms. The animal was required to indicate which of the two stimuli it heard by moving within the response window of 5,000 ms to either the left (FMup) or right (FMdn) water dipper ports. Correct responses were reinforced with 10-μl drops of water, whereas incorrect responses resulted in a time-out period of 3,000 ms. Rats typically performed 200 trials per training session, one session per day. During stage three, the parameters governing the difficulty of the task, including the mean hold time, response window, time-out period, and reward period were adjusted daily until a criterion response level (five consecutive sessions above 80% correct) was achieved.