Upload
lilithiranda
View
17
Download
0
Embed Size (px)
Citation preview
Behavioral Neuroscience1990, Vol. 104, No. 2, 298-319
Copyright 1990 by the American Psychological Association, Inc.0735-7044/90/S00.75
Learning of Physiological Responses:I. Habituation, Sensitization, and Classical Conditioning
Barry R. Dworkin and Susan DworkinDepartment of Behavioral Science
Pennsylvania State University College of Medicine
Rats with chronic neuromuscular block (NMB) maintained by continuous infusion of a-bun-
garotoxin were classically conditioned. All rats showed reliable discriminative-conditioned tibial
nerve firing, hind limb vasoconstriction, hypertension, bradycardia, and electroencephalograph-
ic (EEG) desynchronization. A regression analysis indicated that the conditioned vasoconstric-
tion was neither centrally mediated by, nor inextricably linked to, skeletal (tibial) nerve firing.
Throughout the experiment there were normal blood gases, pH, Na, serum protein, hematocrit,
blood pressure, heart rate, vasomotor tone, and tibial nerve activity. The vital signs, EEG
spectra, and cortical evoked potentials reflected regular sleep-wakefulness cycles and respon-
siveness to mild stimuli. The NMB rat preparation with its stable physiological state and fully
intact central nervous system may be a useful model for a variety of physiological, medical,
and neurobehavioral studies.
We report on the properties of a long-term neuromuscular-blocked (NMB) rat preparation, which has the behavioral
and physiological integrity necessary to study learning. Theexperiments described in this article are concerned only withclassical conditioning, but are part of a broader program todetermine whether the procedures of instrumental learningcan modify the autonomic nervous system (ANS). The over-all logic and technical requirements guiding these develop-ments have been reported elsewhere (Dworkin & Miller,1986).
The Preparation
For periods as long as 30 to 90 days, carefully maintainedcontinuously paralyzed rats have normal and stable phys-iological parameters and sleep-wakefulness cycles, retaincentral modulation of skeletal and autonomic function (Fig-ure 1), and have sufficiently intact sensory function to reliablydiscriminate between 2- and 4-kFfz equi-intensity tones. In
response to very slight disturbance by low-intensity noise orlight touch, the NMB rat exhibits increased firing of skeletalnerves, autonomic arousal, and desynchronization of the
This research was supported in part by National Institutes of Health
Grant NIH HL 40837 and by grants from the Kettering Foundation
and the John D. and Catherine T. MacArthur Foundation.
Neal E. Miller, Michael L. Brines, and W. Andrew Kofke made
essential intellectual contributions to this work. Others having sig-
nificant roles were postdoctoral fellows Ethel Eissenberg and RobertJ. Filwitch, technician Jose DaCosta, and programmers Michael
Eisenberg and Mark Silverman. Marshall B. Jones advised us on
some of the statistical procedures.This article honors Neal E. Miller on the occasion of his 80th
birthday. His insight, imagination, and integrity continue to guide
and motivate our work.Correspondence concerning this article should be addressed to
Barry R. Dworkin, Department of Behavioral Science, Pennsylvania
State University College of Medicine, Hershey, Pennsylvania 17033.
electroencephalographic (EEG) output; but less than 30 minlater the baseline state of these parameters resumes. Thisapparently normal responsiveness to subtle stimuli implies a lackof stress or discomfort in the undisturbed state (Figure 2).
Because temporary narcotic analgesia or deep anesthesiacan be induced repeatedly as needed during paralysis to tem-porarily reduce or eliminate all types of responsiveness, var-ious manipulations can be performed without disrupting thedepth of neuromuscular block, disrupting the constancy ofthe physiological state, or causing pain or distress. Dependingon the agent, induction and recovery may be achieved cithergradually or quite rapidly. Limited primarily by its smallsize, the paralyzed rat offers many opportunities for unin-terrupted observation of physiological function. For exam-ple, continuous arterial access can be maintained reliablythroughout the duration of paralysis, permitting high-fidelity
(0-10 Hz) recording of blood pressure (BP), reliable arterialblood sampling, and behaviorally unobtrusive intravascularinjection. Other transducers, implanted before paralysis un-der deep anesthesia, allow continuous monitoring of corticalEEG and evoked potentials, heart rate (HR), core tempera-ture, the activity of selected skeletal nerves, and peripheralvasoconstriction.
Classical Conditioning and Paralysis
Complex skeletal motor patterns associated with "auto-nomic" classical conditioning procedures were first observedby Pavlov (1927, pp. 13-14 and 29-30, 1932), and becausethey have been a sufficiently ubiquitous finding over the years,Smith (1954,1964) eventually proposed that all conditioningof autonomic responses was skeletally mediated. The strong-est form of Smith's hypothesis is that actual movement of askeletal muscle either directly affects the measurement trans-ducer or stimulates the receptive field of an autonomic reflexand that the appearance of an autonomic response, correlatedwith the conditioned stimulus, is either a simple measure -
298
LEARNING OF PHYSIOLOGICAL RESPONSES 299
ment artifact or a secondary consequence of skeletal learning.In either case, the key point is that the site of neural plasticityis in the skeletal, not the autonomic, pathway. The strongform of Smith's assertion is strong, because it can be testedexperimentally; a rigorous rejection depends only on con-ditioning in a totally paralyzed animal.
The first conditioning studies with curare were done morethan 60 years ago (Harlow & Stanger, 1933), but the resultswere not conclusive. Until the mid-1940s experiments withcurare were complicated by erratic contamination of the rel-atively crude, botanically derived drug with a hypnotic prin-ciple (see Solomon & Turner, 1962, for a more detailed ac-count of the history). For the past 25 years, a variety ofincreasingly specific nicotinic cholinergic blocking agents,lacking significant central action, have been available (Bar-nard & Dolly, 1982). Beyond elimination of general centralnervous system (CNS) effects, a further concern in the studyof visceral conditioning has been that some receptors in theautonomic ganglia can be significantly affected by some nico-tinic blocking agents (Oilman, Goodman, Rail, & Murad,1985).
For any particular drug, the ratio of skeletal to autonomicpotency is highly species-dependent. For example, for deca-methonium, this ratio is approximately 1:10 in the rat and9:1 in the cat. These differences are due to both receptorstructure—although similar, the nicotinie receptors in theneuromuscular junction and ganglia are not identical—andaccessibility of the circulating ligand to the receptors. Thesetwo factors determine the dissociation constants and the timecourse of the blocking action in each tissue, respectively. Forligands not destroyed in the target tissue, permeability-de-pendent first-order kinetic effects reach asymptote within afew hours, and thus, for a chronic preparation the ratio ofdissociation constants is the necessary (Chiappinelli, Cohen,& Zigmond, 1981; Loring & Zigmond, 1988; Zigmond &Loring, 1988), if not always sufficient (Brown & Fumagalli,1977), factor.
In the rat, most paralytic drugs have clear autonomic ef-fects at doses required to reduce junctional transmission to1 % or less (Gilman et al., 1985); however, the alpha fractionof the venom of the eastern Asian snake Bungarus mutti-
cinclus is two orders of magnitude more effective at the neu-romuscular junction than at the superior cervical ganglion.Underlying this selectivity is an extraordinarily high absoluteaffinity for the end plate; thus, with approximately 1000 timesthe potency of succinylcholine, an intravenous a-bungaro-toxin (a-BTX) infusion of 1 nmole per kilogram body weightper hour will paralyze a rat completely, without any detect-able effect on autonomic function.
NMB is not an inclusive control for skeletal mediation: Iteliminates muscle contraction and any contraction-depen-dent afferent feedback mechanisms, but it does not precludeevery kind of skeletal-autonomic connection. For example,a vasomotor response in a paralyzed limb could be drivenby a central collateral from a conditioned limb flexion. Thisinsight led Obrist, Sutterer, and Howard (1972) to generalizeSmith's hypothesis and led others (Black, 1974; Brener, 1974;Roberts, 1974) to express similar views. Ignoring some ofthe diversity and subtlety of the individual statements, their
arguments came down to the following: (a) Behavior occursin organized units that include both autonomic and skeletalresponse patterns (fixed action patterns); and (b) in auto-nomic conditioning under paralysis, although the musclecontraction is suppressed, the entire fixed action pattern isalways what is conditioned. Brener (1987, pp. 272-274) hasexplicitly outlined the major neurophysiological and energyeconomy models implied by the central mediation hypoth-esis. Because of the generality of this hypothesis, a criticaltest is difficult, but at a behavioral level, central mediationpredicts that, al least in an NMB preparation, the magnitudesof conditioned skeletal nerve activity and of related condi-tioned vascular responses should be correlated.
In rats chronically paralyzed by a-BTX, we classically con-ditioned autonomic and skeletal responses. Because paralysiseliminates muscle contraction, the skeletal responses wereobserved by recording directly from the tibial nerves. Theunconditioned stimulus was a shock to the tail and the con-ditioned stimulus (CS+) was an auditory tone. Another tone(CS-) was specifically unpaired with the unconditioned stim-ulus (UCS). The rats were initially habituated to both con-ditioned stimuli and then tested for sensitization using un-paired presentations of the conditioned and unconditionedstimuli. Finally, a series of conditioning trials was presented.
Method
The general description given here is supplemented with a more
comprehensive and detailed account in the Appendix. The numbers
that appear in brackets in this section and elsewhere in this article
refer to specific subsections of the Appendix.
Background
The present preparation originated with an effort to replicate(Dworkin & Miller, 1986) the experiments on visceral instrumental
learning in acutely curarized rats by Miller and his associates (Miller,
1969). During the replication attempt, we learned that adequate
maintenance of the curarized rat for even brief experiments was far
more critical and demanding than our predecessors had appreciated
(DiCara, 1974). Furthermore, we began to better grasp the impli-
cations for medicine and basic regulatory physiology of long-term
operant ANS learning (Dworkin, 1986). Thus, in 1974 we began
developing a chronic paralyzed rat preparation. Because of the pre-
vious history of replication problems, we set the additional goal of
establishing a preparation and procedures that involved minimal
elective judgment and intervention. Whenever possible, we have
used standardized materials and explicit protocols with the goal of
eventually enabling other laboratories to easily and reliably repro-
duce our methods.
The work proceeded through many stages of technical refinement.
Beginning with the discovery that the then-available commercial rat
ventilators were unsatisfactory (Dworkin, 1973; Dworkin & Miller,
1977), we developed and evaluated a new design. We also studied
the effects of paralysis and positive pressure ventilation on pulmo-nary dynamics and the regulation of intravascular volume, devel-
oped methods for accurate control of depth of paralysis, and estab-lished procedures for maintaining over several days (and eventually
weeks and months) proper core temperature, adequate nutrition,
electrolyte balance, and freedom from pain or discomfort. Because
components requiring frequent manual maintenance or calibration
could interrupt the experimental protocol and provide opportunities
300 BARRY R. DWORKIN AND SUSAN DWORKIN
LEARNING OF PHYSIOLOGICAL RESPONSES 301
for inadvertently influencing the experimental outcome, the patency
of cannulas and catheters, the integrity of recording and stimulating
electrodes, and the stability of transducers presented additional unique
problems. Eventually, we incorporated a computer system [1.1] for
the control of most of the maintenance and all of the behavioral
procedures. The programming language is a commercially available
multitasking laboratory BASIC [1.2].
Anesthesia, Set-up, NMB, and Life Support
The key technical developments that make the long-term para-
lyzed rat preparation possible include a specially designed respirator
system [4.3] and tracheal cannula. The respirator is a constant mi-
nute-volume, rotary-valve, inspiratory-jet device. It uses standard
gas mixtures, can maintain a preset tidal volume to within 0.2% for
several months, and because it has only one moving part, almost
never fails. The rotary valve, which switches the gas streams with
the respiratory' phase, was specially constructed and is unusuallysimple and reliable. In somewhat less demanding applications, more
readily available electronic and pneumatic components could be
successfully substituted. The tracheal cannula is constructed of thin-
walled Teflon tubes and has a coaxial design: The inspiratory flow
is carried at high (jet) pressure through a small (24 gauge) central
lumen to a point approximately 0.2 cm beyond the tip of the sur-
rounding 12-gauge expiratory lumen. Because the cannula is without
any dead volume, flow is unidirectional for the entire length of the
large expiratory tube; unlike conventional cannulas that eventually
accumulate mucous obstructions, with this cannula tracheal-bron-
chial secretions are continuously removed by the expiratory gas
stream. The respirator has a gated port for measuring end-expiratory
Pco, [4.3.6] and pneumatic valves for regulating the positive end-
expiratory pressure (PEEP) and delivering periodic controlled hy-
perinflations to discourage atelectasis formation [4.3.4]. Humidifi-
cation and collection of secretions [4.3.5] are automatic, and minute
volume and rate are regulated by a commercial electronic mass
flowmeter [4.3.3.1]. Again in less critical applications, this flowmeter
could be replaced with a simpler floating-ball rotameter. Details of
the apparatus and the arrangement of the major components are
shown in Figures 3A and 3B.
Surgery
All surgery is performed with carefully monitored deep anesthesia
using Nembutal (sodium pentobarbital) prior to tracheal cannulation
Figure 2. Grass direct-current integrator trace of RMS 3-12-Hz
electroencephalograph (EEG) activity. (At arrow marker, the undis-
turbed rat was touched gently and the EEG desynchronized; how-
ever, within 30 min after this stimulation, the sleep pattern began
to resume normal density.)
and then Forane (isoflurane) [2.1]. A 100-/*g bolus of a-BTX induces
NMB during anesthesia before connecting the respirator [2.3]. In
addition to the tracheal cannula, surgery includes implantation of
the following (see Figures 3A and 3B): EEG electrodes [2.2.8], a
trans-urethral bladder cannula [2.2.1], subcutaneous electrocardio-
graph (EKG) electrodes [2.2.2], a carotid artery cannula [2.2.4], a
core-temperature probe [2.2.5], silicone-imbedded recording-stim-
ulating electrodes on the tibial nerves [2.2.6], bipolar electromy-
Figure 1. Typical 2-hr polygraph record of physiological variables from an NMB rat undisturbed for the previous 12 hr. (From top to
bottom: left and right tibial nerve firing rate [log,0 impulses per second]; left and right vasomotor activity [vasoconstriction is up]; heart rate
[increased rate is up]; mean arterial pressure [MAP]; electroencephalograph (EEG) RMS power in the 3-12-Hz band; and inspiratory pressure
[the pressure in the expiratory channel of the tracheal cannula]. The "spikes" in the inspiratory-pressure record are intentionally produced
by the respirator system to prevent the formation of atelectases in the pulmonary alveoli. The average MAP for the entire period was 123
mmHg with a range of 95-160 mmHg; the heart rate mean was 393 beats per minute [bpm] with a range of 352-454 bpm. Data from our
laboratory [Dworkin, Filewich, Miller, Craigmyle, & Pickering, 1979] using a similar blood pressure measurement technique and from Iwata
and LeDoux [1988] give comparable values for both heart rate and MAP in freely moving rats. The EEG trace shows a typical regular pattern
of sleep cycles. This sleep pattern is present during rest hours, and the rats are desynchronized for most of the sensitization and classical
conditioning [CC] hours; however, late in the CC procedure some animals have periods of synchronization during the mtertrial intervals.
The sleep cycles have a typical duration of 13-15 min; this can be observed by inspection of the record and was verified by application ofthe Cooley-Tukey algorithm [Bracewell, 1978] to the complete 14-hr record after removal of the linear component. The fast Fourier transform
revealed a major peak at 15 min and a secondary peak at 27 min; the latter is probably either a subharmonic or a reflection of occasionally
missed cycles. The sleep-cycle characteristics correspond reasonably well with established data for nonparalyzed rats [Campbell & Tobler,1984; Friedman, Bergmann, & Rechtschaffen, 1979; Mistlberger, Bergmann, Waldenar, & Rechtschaffen, 1983; Van Twyver, Webb, Dube,
& Zackheim, 1973]. The relationship of autonomic activity patterns to the EEG are complex; however, in general EEG synchronization is
associated with reduced blood pressure, heart rate, vasomotor tone, and tibial nerve firing rate. The general impression of a positive relationship
between heart rate and blood pressure in the sample record was confirmed by correlational analysis [see text].)
302 BARRY R. DWORKJN AND SUSAN DWORKIN
temperaturecontrol plate
[4-5-3]
cannulahold down plate
ntra-gaslric \tracheal
"carotid artery feeding tube /- cannulacannula [2-2-14] / |2-2-3]
|2-2-4| \ /
expiratorychannel
Figure 3A. Arrangement of the major components of the preparation. (The hold-down plate maintains
the alignment of the tracheal cannula and the trachea and isolates the rat's tissues from movement,
which would cause damage or discomfort. See the legend of Figure 3B for additional details.)
ograph (EMG) electrodes in the gastrocnemius muscles [2.2.7], cu-
taneous tail-stimulation electrodes [2.2.12], and a chronic oral-gastric
feeding tube [2.2.14]. Also under deep anesthesia, the following de-
vices were positioned externally: precision optical vasomotor trans-
ducers [2.2.13], the microphone of an electronic stethoscope [2.2.9],and earphones for auditory stimulation [2. 2.15].
The surgery usually requires 3 hr, but analgesic therapy is always
continued for at least 24 hr, and until no distress is evident [3.1].
Forane is used for the first 20 hr and then followed by narcotic
analgesia until regular sleep-wakefulness cycles become evident in
the EEG and autonomic measures. This latter stabilization phase
usually requires an additional 10-20 hr, and during this time blood
Po,, Pro,, pH Na% and K* are gradually centered [2.3.2; 4.9] in thenormal range by adjusting the minute volume of the ventilator and
the electrolyte composition of a continuous inlra-artcrial infusion
[3.2]. Also, to determine the dose of a-BTX required to maintain
1% synaptic transmission, the depth of the NMB is assessed by
measuring the evoked gastrocnemius EMG response to a 30-^ssingle-pulse stimulation of the tibial nerve [3.4.1.1]. The core tem-
perature is regulated at 38 °C by a proportional controller [4.5] ref-
erenced to a vaginal probe. Intragastric dextrose is administered until
the stabilization phase is complete, and then a denned liquid diet is
begun I imragastnc High Nitrogen Vivonex supplemented with iron
and vitamin K) [3.3]. Low-level light and 65-dB background masking
noise are continuously present [2.2.16].
Basic physiological data are recorded on a regular schedule. In
addition to continuous (5-s resolution) digital recording of HR [2.2.2;
4.2], BP [2,2,4; 4.4], EEG [2.2.8; 4.7], peripheral temperature [2.2.5;4.5], urine output [2.2.1; 4.1], hind paw vasoconstriction [2.2.13;
4.8], and tibial nerve activity [2.2.6; 4.6], measurements are made
LEARNING OF PHYSIOLOGICAL RESPONSES 303
irans-urethral
bladder cannula
[2-2-1J
Figure 3B. Arrangement of the major components of the preparation (continued). (During actual
conduct of the experiment, the rat's head is enclosed in a hemispherical dome that rests on the substrate;see [2.2.16]. The bladder cannula and temperature probe are held in place by a wire frame [not shown].
The paws are held to the vasomotor transducers by a layer of collodion between the dorsal surface
and the photo diode; adequate distance is maintained for expansion and contraction of the tissue.
Pan of the unconditioned-stimulus electrode can be seen at the top of the tail. See Appendix for more
detailed descriptions of the various transducers and cannulae [referenced numbers in brackets].)
every 24 hr of plasma and urine osmolarity; blood P0l, PCOl, and
pH; hematocril [4.9]; visual-evoked potentials [4.7.5]; and depth of
NMB [3.4]. During experimental protocols additional data are re-
corded on digital audiotape (I Hz to 20 kHz) and on a digital os-
cilloscope.An automated alarm system monitors the preparation continu-
ously [5J. An alarm in the laboratory is triggered immediately by
abnormal excursions of HR, BP. temperature, urine output, respi-
ratory pressure, EMG, or EEG sleep-cycle activity. The experiment
can be controlled from a remote terminal, and a bolus of Nembutalcan be injected intraperitoneally by a special emergency code or
automatically by an alarm system failure, but in practice alarm con-
ditions are rare. In the most recent 100 experimental days, therewere no alarms posted, and in the same period no experiments were
terminated because of mechanical or electronic failure of the life
support system.
Behavioral Procedures
Stimuli. The presentation of stimuli and collection of data were
controlled by a Macsym 351 industrial process computer [1.1]. TheCSs were 2- and 4-kHz square-wave, analog-generated auditory tones
presented through a pair of dynamic earphones [2.2.15], directed
along the intra-aural line and rigidly mounted 3 cm from the mid-
sagittal suture. The sound level on the C scale of a General Radio1555A meter was adjusted to be 80 dB at 1.5 cm from the diaphragm;
this is 15 dB above the background [2.2.16.2]. The CS duration was
always 60 s, and the tail-shock UCS [2.2.12] was a 500-ms train of
I mA RMS series-regulated 60-Hz sinusoids.Schedules. The interstimulus interval had a range of 3-12 min
and a mean of 8.5 min (see Figure 4), resulting in 3—4 trials of each
type per hour. For classical conditioning (CC) the shock always
cotcrminated with the tone CS+ and was always separated by at
least 4 min from the CS-; thus, the CS- was specifically unpaired
with the UCS and not a true "neutral" stimulus (i.e., the CS- sig-
naled no shock). During habitualion (HAB) the same schedule of
tones was presented without shock, and during sensilization (SENS)
the shock was specifically unpaired with either tone (always separated
by 3 min or more).
One-hour presentations of HAB, SENS, and CC trials were begun
at noon and followed by 3 hr of rest. The hourly trial sequence was
repeated for a total of six sessions per 24 hr. Each behavioral pro-
cedure continued until both the time and asymptotic performance
criteria were satisfied, which are as follows: The HAB sequence
continued for at least 12 hourly trial sequences and until the HR
response (mean of the 120-s pre-CS interval subtracted from the
mean of the CS interval) to both tones was less than three times the
pre-CS standard deviation (three Z units) for 2 successive hr, the
SENS procedure continued for 3 hourly trial sequences, and the CC
procedure continued for at least 15 hr and until the change in themagnitude of the HR response to the CS+ over 4 successive trial
sequences was less than three Z units.
Results
Factors Affecting Survival and Physiological
Condition of the Preparation
Since 1973 we have gradually improved the longevity andquality of the preparation. Figure 5 shows the survival timeof recent preparations as a function of starting date. The
304 BARRY R. DWORKJN AND SUSAN DWORKIN
,500msecrShock
K60 sec-HIH 8.5 minI VI
CS-
Figure 4. Schematic representation of the time relationships in the
classical conditioning (CC) paradigm. (The pseudo-random distri-
bution of interstimulus intervals ranged from 3 to 12 min with a
mean of 8.5 min, resulting in three or four trials of each type per
hour. The shock coterminated with the CS+ and was always sepa-
rated by at least 4 min from the CS- . One-hour presentations of
habituation [HAB], sensitation [SENS] and CC trials were begun at
noon and followed by 3 hr rest. The hourly trial sequence was re-
peated for a tolal of six sessions per 24 hr.)
median survival time for the last 10 preparations is 26 dayswith a continuing trend toward longer survival. The largestsingle improvement occurred with the exclusive use of femalerats. In most respects the females proved to be similar tomales, bul placement and maintenance of a bladder cannulawas vastly simpler and less traumatic. The bladder was amajor difficulty with males. Many males died soon after beingplaced in the apparatus of hemorrhage associated with thesurgical placement of a trans-abdominal cannula, and othersdied during the 2nd or 3rd week because of retrograde ejacu-lation and formation of an obstructing seminal plug in thebladder. Placement of the females' cannulae did not requiresurgery, and if obstruction occurred after several weeks, as
sometimes it did with accumulation of a crystalline precip-itate, a new cannula was easily substituted with minimaldisturbance.
Another major factor that increased longevity was an im-proved understanding of the variables affecting intravascularvolume. In a paralyzed animal, two mechanisms of volumeregulation are comprised: The hydrostatic effects of positive-pressure ventilation inhibit venous return, and the lack ofskeletal muscle contractions disables the venous pumps(Cowley & Trump, 1982; Frankel & Mathias, 1976; Kofke& Levy, 1986). The effects of both are in the same direction,but for an animal as small as the rat, mechanical ventilationis the far greater problem. Unlike natural respiration, in whichduring inspiration active expansion of the thorax draws airinto the lungs under reduced pressure, positive-pressure ven-tilation forces expansion by increasing intrathoracic pressure.Higher pressure around the large veins opposes, rather thanassists, the return of the peripheral venous affluent, whichinitially decreases venous return to the right heart, but even-tually, with compensation, elevates the systemic venous pres-sure and causes fluid transudalion into the interstitial space.If unchecked, the net result is an accumulating loss of plasmaprotein, a reduction of the colloid oncotic pressure, and ul-timately, collapse of the circulation. Early on, rats sometimesdied in the first 24 hr of a syndrome that resembled surgicalshock: low urine output, edema, hypotension, and tachycar-dia. We documented some of the underlying pathophysiologyby measuring the central venous pressure and the plasmatotal protein and osmotic pressure. Using these data, andsome concepts from the surgical intensive care literature, weevolved a more effective protocol. For approximately 75%
of animals, presurgical attention to adequate hydration [2.1.6]and administration of a single relatively large dose of hy-drocortisone is sufficient to entirely prevent initial hypovo-lemia. In the remainder of the rats, the condition occurs to
some degree; but for most, conservative approaches to fluidand electrolyte management accomplish correction. For ex-ample, volume can be expanded temporarily by infusion ofnormal saline, or if protein is required, human plasma al-bumin can be infused (dextran-type plasma expanders arenot tolerated by rats). By following the defined protocol [2.1.4—2.1.6.3] in the first 24-36 hr we were successful in achievingsatisfactory stabilization in 90% of the rats, but we do notfully understand the pharmacology of hydrocortisone (Ham-merschmidt. White, Craddock, & Jacob, 1979) or the phys-
iological sequelae that eventually lead to appropriate self-regulation of intravascular volume.
Although many animals would stabilize without any in-tervention, we conclude that some initial intervention is usu-ally advisable. Significant episodes of hypotension in the im-mediate postsurgical period could compromise the behavioralintegrity of a preparation days or weeks later. On the otherhand, carefully selected prophylactic interventions are un-likely to have effects beyond 24-48 hr or influence subsequentexperimental protocols. Thus, we recorded vital signs fromthe initiation of anesthesia and infused saline to correct hy-potension (mean arterial pressure [MAP] < 70 mmHG),
tachycardia (HR > 450 beats per minute [bpm]) or sustainedanuria (<0.5 ml/hr). In additional observations on subjects(not included in the present behavioral data), intentionalreduction of MAP to 50 mmHg for 20 min was withoutdetectable effect on cortical visual- or auditory-evoked po-tentials, behavioral responsiveness, or the composition of
the EEC, so we believe that our criteria for intervention wereadequately stringent.
The first 24-36 hr can be viewed as constituting postsurgi-cal intensive care. Once stabilization is complete, usually
100 -L
3/83 11/84
Preparation Starting Date
Figure 5. Survival time of recent preparations as a function of
starting date. (The median for the last 10 preparations was 26 days
with a continuing trend toward longer survival. The largest single
improvement occurred with the exclusive use of female rats. Another
major factor that increased longevity was an improved understand-
ing of the variables affecting intravascular volume [see text]. Tri-
angles = males; circles = females.)
LEARNING OF PHYSIOLOGICAL RESPONSES 305
left tibial n.
03+
left tibial n.
right tibial n.
cs+
right tibial n.
Figure 6. Average responses of the left and right tibial nerves to the CS+ and CS- for 5 animals.
(The dots are the time-locked averages of the last 30 trials for each rat. The solid line is the "grand"
average for all 5 subjects [150 trials]. The abscissa is in seconds, and the ordinate is in Z units based
on the pre-CS baseline mean and variance. Tail shock was delivered at the arrow marker. The difference
between CS+ and CS- magnitudes is reliable for each response measure [Wilcoxon rankp < .005
for all responses]. Although the detailed response topography varied among subjects, the magnitudes
were in the same direction for all responses in every rat.)
after 48-96 hr, and most obviously signaled by the com-mencement of regular cycles of EEC slow-wave activity andcorrelated autonomic variability (Figure 2), maintenance ofthe basal physiological state requires little if any intervention.The ventilation parameters, composition of the intravascularinfusate (including the concentration of a-BTX), intragastricfeeding, and temperature regulation can remain set for theduration of the experiment. Routine maintenance of thepreparation becomes largely a matter of removing feces andurine, changing the ophthalmic ointment, filling syringes,checking gas supplies, and monitoring vital signs and evokedpotentials for evidence of deterioration.
Figure 1 is a 2-hr polygraph record from an undisturbedrat. It is typical of approximately 3000 hr of baseline datacollected on the 5 CC subjects. The bottom trace is the pres-sure in the expiratory channel of the tracheal cannula [4.3.4.3];the "spikes" in the record are produced by the respiratorsystem [4.3.4.2] and help to prevent the formation of atelecta-
ses in the pulmonary alveoli. During the first 24-36 hr of
paralysis, these artificial positive-pressure "sighs" sometimescause brief arterial hypotension; however, this effect gradu-ally attenuates and disappears by the end of the stabilizationphase.
The EEG trace is the integrated 3-12-Hz activity and shows
a typical regular pattern of sleep cycles. Figure 2 is a Grassintegrator trace of a similar measure; the arrow indicates thetime at which the undisturbed rat is touched gently and theEEG desynchronizes. Within 30 min the sleep pattern is againreestablished. The sleep pattern is ordinarily present duringrest hours, and the rats are desynchronized for most of theSENS and CC hours. However, late in the CC procedure, asthe discrimination becomes well established, some animalshave periods of synchronization during the intertrial inter-vals. The relationship of autonomic activity patterns to theEEG are complex. In general, cortical synchronization is as-sociated with reduced BP, HR, and vasomotor tone. Tibial
306 BARRY R. DWORKIN AND SUSAN DWORKIN
right voso
cs+
right vaso
Figure 7. Average vasoconstriction responses of the left and right hind paws to the CS+ and CS-
for 5 animals. (Dots are the time-locked averages of the last 30 trials for each rat. The solid line isthe "grand" average for all 5 subjects [150 trials]. The abscissa is in seconds, and the ordiante is inZ units based on the pre-CS baseline mean and variance. Tail shock was delivered at the arrow marker.
The difference between CS+ and CS— magnitudes is reliable for each response measure [Wilcoxonrank p < .005 for all responses]. Although the detailed response topography varied among subjects,the magnitudes were in the same direction for all responses in every rat.)
nerve firing rate is also reduced, but the proportion of vari-
ance due to EEG is generally less for the skeletal than auto-
nomic measures. The general impression of a positive rela-
tionship between HR and BP in the sample record is confirmed
by more formal measures: for 5-s (~35 beats) data samples
during the rest hours, the range of the correlation is .50-.85.
Weight
Most rats maintained a relatively constant weight during
paralysis. The median weight change was —28 g. The mean
pre-NMB starting weight for the group of 5 animals was 324.4
± 44 g; at the termination it was 295.8 ± 59.9 g; the difference
of -8.5% was nonsignificant (paired t = -1.34; p > .20).
Termination of the Experiment
Experiments were terminated by intra-arterial injection of
150 mg/kg Nembutal, followed at least 30 min later by inter
ruption of ventilation.
Classical Conditioning
Figures 6-9 show the averaged responses to the CS+ and
CS- during the last 5 hr of CC. The ordinal units are Z
scores. For each response and each rat, the mean and stan-
dard deviation were computed for the 120-s pre-CS period,
the mean was subtracted from each point, and the result was
divided by the standard deviation. Finally, the means of the
values at each time point (5-s bin width) were computed.
The result is similar to the average evoked response proce-
dure used in neurophysiology. Z scores obscure the common
physical units of measurement but permit meaningfully scaled
comparisons among the different responses. In Figures 6-9,
the dots are the averages within each individual animal, and
the solid curves are the averages over the 5 rats. The differ-
ence between the CS+ and CS— magnitudes (the sums over
the CS-onset to shock-onset points) is reliable for each of the
seven response measures (Wilcoxon rank p < .005 for all
responses except HR; for HR, p = .02). Although the detailed
LEARNING OF PHYSIOLOGICAL RESPONSES 307
bloodpressure
cs+
heart rate
15
10
3
o
-10
-13
-is120 300
Figure S. Average blood pressure and heart rate responses to the CS+ and CS- for 5 animals. (The
dots are the time-locked averages of the last 30 trials for each rat. The solid line is the "grand" averagefor all 5 subjects [150 trials]. The abscissa is in seconds, and the ordinate is in Z units based on thepre-CS baseline mean and variance. Tail shock was delivered at the arrow marker. The difference
between CS+ and CS- magnitudes is reliable for each of the response measures [Wilcoxon rank p <.005 for blood pressure; for heart rate, p < .02]. Although the detailed response topography variedamong subjects, the magnitudes were in the same direction for all responses in every rat, with the
exception of 1 rat that increased rather decreased heart rate.)
response topography varied among subjects, the final mag-
nitudes were in the same direction for all responses in every
animal with the exception of an increased rather than de-
creased HR in 1 rat.
Figures 10 and 11 are in physical units of measurement,
rather than Z scores, and show for the 5 rats the mean of the
magnitude of each CS+ response over the three behavioral
procedures. The left and right tibial nerve activity and left
and right vasomotor responses were very similar and were
combined for simplicity. HAB1 and HAB2 are the first and
last 5 hr of the habituation, respectively; SENS is the entire
3 hr of sensitization; and CC1, CC2, and CC3 are the first,
middle, and last 5 hr of classical conditioning, respectively.
Sensitization is an appropriate control for classical condi-
tioning because it includes all features of the CC schedule,
except the specific CS+ - shock contiguity (Oleson, Ashe, &Weinberger, 1975; Oleson, Westenberg, & Weinberger, 1972).
The standardized slopes and reliability of each individual
rat's acquisition curves for both the CS+ and CS- are in
Table 1; they were computed using the individual 5-s data
points from SENS, CC1, CC2, and CC3. Excepting HR, there
are 30 CS+ regression coefficients; 27 are highly reliable (p
< .001) for each response and are in the direction of the final
response magnitude (Figures 6-9). Some of the CS- slopes
are also reliable, but for 1 rat they are in the opposite direction
to the CS+, and in every case, again excepting HR, the CS+
regression coefficient is of greater magnitude in the predicted
direction than the CS-. This result is shown in Figure 12;
the Wilcoxon rank reliability (p = .02) is the maximum pos-
sible for n = 5; the paired t test results given in Table 1 aresimilar.
Although the just-mentioned results directly address theindependence of autonomic from skeletal classical condi-
tioning at a peripheral level, they do not exclude the possi-
bility that some type of mediation occurs central to the neu-
romuscular junction; however, several interresponse
308 BARRY R. DWORKIN AND SUSAN DWORK1N
7-15 Hz EEC
Figure 9. Average 7-15-Hz EEG responses to Ihe CS+ and CS- for 5 animals. (The dots are the
time-locked averages of the last 30 trials for each rat. The solid line is the "grand" average for all 5
subjects [150 trials]. The abscissa is in seconds, and the ordinate is in Z units based on the pre-CS
baseline mean and variance. Tail shock was delivered at the arrow marker. The difference between
CS+ and CS- magnitudes is reliable for each response measure [Wilcoxon rank p < .005 for all
responses]. Although the detailed response topography varied among subjects, the magnitudes werein the same direction for all responses in every rat.)
correlations bear on this point. If the vasomotor response iscentrally mediated by an anatomically associated skeletalnerve outflow, then on a given trial the ipsilateral responsemagnitudes should be related. There is in fact some weakevidence in our data for generalized response lateralization.Using the final mean (CC3) response magnitudes for eachrat, the normalized difference ratios, (left - right)/(left +right), for vasoconstriction and nerve activity are concordantfor 4 of the 5 rats. Although this result appears to suggest adegree of lateral dominance that encompasses both the auto-nomic and skeletal responses, a trial-by-trial analysis of re-sponse covariation fails to support the idea that the classicallyconditioned vasoconstriction response is mediated by, orclosely associated with, central activation of the ipsilateralskeletal nerve.
To analyze the relationship between nerve firing and va-soconstriction, the response magnitudes for the final 5 hr of
acquisition (CC3) were separately Z scored for each rat to
remove between-subject variance, and the following corre-lations were computed: left and right tibial nerve firing rate(LN/RN), left and right vasoconstriction (LV/RV), and bothipsilateral nerve firing rate and vasoconstriction (LN/LV and
RN/RV). The results are given in Table 2 for CS+ and CS-responses. The general pattern was the same for all rats andindividually reliable for 3 of the 5 rats. Rank statistics gaveapproximately the same results as the product-moment cor-relation, and the differences discussed later would in fact havebeen somewhat larger without within-subject normalization.
If a skeletal response, as reflected in the nerve activitymeasures, caused or mediated the autonomic response, asreflected in the vasomotor measures, then the LN/RN cor-relation would have to have been at least as large as the LV/
RV correlation; or, if both responses were being generatedin a common central pathway, diverging from the CS input,
Table 1Coefficients of Linear Regression and Reliability (p) Over the Trials for the Responses of Each Rat to the Paired (CS+) and
Unpaired (CS-) Stimili
Tibial nerve
EEG
Subject
12345
MI
P
CS +
-.24-.32-.32-.14-.24-.25
P
.000
.000
.000
.000
.000
4.64.005
CS-
-.15-.10-.23
.13'
.02-.07
P
.000
.007
.000
.001
.642
CS+
.57
.29
.01
.57
.32
.35
Right
P
.000
.000
.710
.000
.000
-2.38.038
CS-
.39
.08-.03-.17*
.07
.07
P
.000
.050
.422
.000
.106
CS+
.41
.41
.30
.33
.35
.36
Left
P
.000
.000
.000
.000
.000
-2.14.049
CS-
.34
.10
.29-.31*
.18
.12
P
.000
.012
.000
.000
.000
Asterisks indicate that slopes are reliable in the opposite direction, CS+ = tail-shock conditioned stimulus paired with an auditory
LEARNING OF PHYSIOLOGICAL RESPONSES 309
-15habl hab2 sens cc1 cc2 cc3
Figure 10. Development of blood pressure, Vasoconstriction, and
heart rate responses to the CS+ during the several phases of the
experiment. (The ordinate is the mean magnitude of the responsein physical units. HAB1 and HAB2 refer to the first and last 5 hr of
habituation, respectively; SENS refers to the entire 3 hr of sensiti-
zation; and CC1, CC2, and CC3 refer to the first, middle, and last
5 hr of classical conditioning, respectively. During HAB both tones
were presented without shock, whereas during SENS shock was pre-
sented but was not contingent on termination of the CS+ tone.)
0.5
xc 0.4
0.3
0.0
0.0
-0-5"
-1.0
I
-1.5
-2.0
Tibial nerve activity
habl hab2 sens cd cc2
EEC (3-12 Hz)
habl hab2 sens cc1 cc2 cc3
Figure 11. Development of tibial nerve activity and 3-12-Hz EEG
responses to the CS+ during the several phases of the experiment.
(The ordinate is the mean magnitude of the response in physical
units. The bilateral tibial nerve and vasomotor responses were sim-
ilar and, thus, combined. HAB1 and HAB2 refer to the first and last
5 hr of habituation, respectively; SENS refers to the entire 3 hr of
sensitization; and CC1, CC2, and CC3 refer to the first, middle, and
last 5 hr of classical conditioning, respectively. During HAB both
tones were presented without shock, whereas during SENS shock
was presented but was not contingent on termination of the CS+
tone.)
then the LV/LN and RV/RN correlations should have been
of the order of the product of the square roots (McNemar,
1962, Chap. 10) of the LN/RN and LV/RV correlations (ap-
proximately .53). Neither assertion is substantiated in Table
2: the CS+ correlation between vasomotor responses is re-
liably greater (p < .05), not smaller, than the nerve responses,
Table 1
Continued
Vasoconstriction
Right
CS+
.54
.42
.36
.25
.31
.38
P
.000
.000
.000
.000
.000
-3.31.015
CS-
.35
.05
.35-.24*
.02
.11
P
.000
.202
.000
.000
.620
CS+
.59
.34
.42
.10
.04
.30
Left
P
.000
.000
.000
.007
.199
-3.38.014
cs-.39.05.41
-.17*-.07
.12
P
.000
.233
.000
.000
.095
CS+
.19
.17
.31
.54
.25
.29
BP
P
.000
.000
.000
.000
.000
-2.01.057
CS-
.18
.08
.25-.01-.04
.09
P
.000
.057
.000
.777
.315
CS+
.04-.24
.11-.07-.35-.10
HR
P
.200
.000
.001
.044
.000
1.34.125
CS-
.05
.18
.06-.09-.09
.02
P
.239
.000
.151
.029
.035
tone; CS— = auditory tone specifically unpaired with the tail shock; EEG = electroencephalogram; BP = blood pressure; HR = heart rate.
310 BARRY R. DWORKIN AND SUSAN DWORKIN
«>. 0.3
I
c .i!D 3cl) cr
0.1 --
0)
8 -0.2 +
-0.3-1-
• cs-ES CS+
EEG
J 111
HeartRate
Right Left Right Left BloodNerve Nerve Vaso Vaso Pressure
* Wilcoxon Rank P= .02N= 5
Figure 12. Mean CS+ and CS- regression coefficients for each response. (Standardized slopes werecomputed using 5-s data points from SENS, CC1, CC2, and CC3 for each rat and each response; 30of the 35 regression coefficients for the CS+ are reliable in the expected direction al p < .01, and the
mean CS- regression coefficient for the combined 5 rats is reliable \p < .001] for all responses. Forthe Wilcoxon rank test, p = .02 is the highest reliability that can be achieved with an n = 5. [SeeTable 1 for the data on individual subjects.])
and the ipsilateral vasomotor and tibial nerve responses are
uncorrelated. The CS+ correlations taken alone argue strong-
ly against central mediation of the autonomic response by
the skeletal response, with possibly one caveat: If the nerve
measure was inherently inaccurate, all three correlations in-
volving nerves could have been spuriously attenuated; how-
ever, the acquisition-curve data (Table 1; Figure 12) and the
CS— data confirm that the nerve measures are in fact quite
good. Of the CS+ nerve regression coefficients, 9 are indi-
vidually reliable (p < .0001), and the reliably larger (p <
.01) CS- LN/LV and RN/RV correlations (Table 2) show
more specifically that the low CS+ correlations are not due
to measurement error in that the error should have been
proportionally at least as great for CS— responses. Finally,
extreme truncation of a measure by ceiling or floor effects
can spuriously attenuate correlations; however, the smaller
CS + correlations are also not an artifact of restricted variance
because the standard deviations of the CS+ nerve responses
are actually larger than the standard deviation of the CS—
nerve responses.
Because the magnitude of the final CS+ nerve response is
greater than the magnitude of the final CS— nerve response
(Figures 6-9), and because the nonstimulus-related variance
is probably similar, the variance specific to learning is prob-
ably larger; yet, the CS+ diminished the correlation between
tibial nerve activity and ipsilateral vasoconstriction, and this
is inconsistent with the autonomic response either being
mediated by or inextricably linked to the learned skeletal
response.
Table 2
Correlations Among Individual Trials in the Final 5-Hr
Acquisition Period (CC3) for the Response Magnitudes to
the Paired (CS+) and Unpaired (CS-) Stimuli
Response pairs
Stimulus LN/RN
.32
.43
LV/RV
.86
.65
LV/LN
.32-.02
RV/RN
.34
.09
Note. CS+ = tail-shock conditioned stimulus paired with an au-ditory tone; CS— = auditory tone specifically unpaired with the tailshock; LN/RN = left and right tibial nerve firing rate; LV/RV = leftand right vasoconstriction; LN/LV and RN/RV both refer to ipsi-lateral nerve activation and vasoconstriction; N = 75.p.™, = .30 Pjas = A9.
Discussion and Conclusions
Discrimination and Rate of Acquisition
The CC paradigm used 2- and 4-kHz tones that were equat-
ed for intensity; no attempt was made to optimize the rateof acquisition or test the limits of the rat's discrimination
ability. The use of extensive HAB and SENS trials and a
relatively long CS duration probably slowed acquisition;
nevertheless, clear evidence of discriminative classical con-
ditioning was obtained for all response systems within 30
trials. Blackwell and Schlosberg (1943) studied the discrim-
ination ability of freely moving rats using a shock-avoidance
procedure. They found that the limit of discrimination was
a 1-2 kHz difference at 10 kHz and that asymptotic perfor-
mance required at least 200 trials. Ray and Stein (1959),
using a conditioned suppression procedure for 4 trials per
LEARNING OF PHYSIOLOGICAL RESPONSES 311
day, obtained reliable discrimination between a 200-Hz toneand an 1800-Hz tone in approximately 19 days. Jamison(1951), using ammonia gas as a UCS to elicit 300-bpm bra-dycardia and a 4-kHz tone CS, obtained a stable CR and ageneralization decrement in approximately 25 trials. All ofthese studies used short CS durations and none precededconditioning with either habituation or sensitization se-quences; although direct comparison is difficult, the perfor-mance of the NMB rats was at least not conspicuously worse.
Kelly and Masterton (1977) determined the audiogram forSprague-Dawley albino rats: they found that the averagethreshold of 4 kHz is approximately 7 dB SPL lower thanthe threshold at 2 kHz; however, among their 3 subjects therewas overlap of the thresholds at these particular frequencies,and the rank order reversed several times over the full fre-quency spectrum. Jamison (1951) similarly found a 5-dBdifference. Given these results, and that our stimuli were wellabove threshold, the discrimination was probably based onfrequency, but not having randomized intensity, a loudnessdifference of as great as 7 dB could have been an additionalcue.
Response System Characteristics
Among the response systems studied, vasoconstriction andtibial nerve firing rate had the most consistently monotoniclearning curves; HR was least consistent across subjects. Al-though the average HR response in CCS was a reliable netdecrease, it, unlike the other six measures, developed in anerratic manner: The average HR response during HAB wasan increase (opposite to the conditioned response), severalanimals showed net increases in the early stages of condi-tioning, and 1 of the 5 rats persisted in a substantial increasein CC3.
HR is determined in large part by two potentially antag-onistic neural systems. If these systems condition indepen-dently (although both acquisition curves may be monotonic),then as learning progresses the balance between them mayshift, changing the magnitude or even the direction of thenet response. Pharmacological manipulations or lesions canhelp to isolate individual components and, thus, infer theirinteraction in various conditioning paradigms. Iwata andLedoux (1988) reported that in freely moving intact rats,conditioned BP responses were readily elaborated, but con-ditioned bradycardia emerged only after propranolol blockof the cardiac sympathetic input. They also found in pairedversus nonpaired subjects that parasympathetic block byatropine differentially enhanced cardioacceleration.
Because HR and BP share hydraulic and neural inter-dependence, conditioned BP responses can add yet anothercomplicating factor to the HR learning curve. Because of itscomplexity, HR may not be an ideal model response systemfor studies of the neurophysiology of classical conditioning(Cohen, 1980). Its appeal has been ease of measurement andthe accumulated base of knowledge of the neural circuity;but even the latter advantage may be outweighed by usingresponses that are freer of homeostatic constraints, have moresimple final common pathways, and have more consistentlymonotonic learning curves: The vasomotor and tibial nerveresponses probably meet these criteria, and in our prepara-
tion the vasomotor measurement is especially stable andtechnically straightforward.
Mechanism of Conditioned Vasoconstriction
The central mediation hypothesis has at its root the ideathat the ANS lacks the independent capacity for modificationby learning and that autonomic responses are in some sense"slaved" to functionally related modifiable skeletal responses(Obrist, 1968). If the hypothesis were true, then sets of skel-etal and visceral responses would be linked in fixed patterns,and it would be impossible to condition a visceral responsewithout also simultaneously atfecting motor responses. Thisarrangement would greatly limit the adaptive homeostaticpotential of all types of visceral learning (Dworkin, 1986).
The correlations in Table 2 are inconsistent with a closelinkage between tibial nerve activity and lower leg vasocon-striction; thus, they appear to contradict central mediationwith respect to these particular responses. The responses wereanatomically adjacent to one another and sampled a largeproportion of the total regional distribution of skeletal effer-ent and vasomotor activity, respectively. The statistical ques-tions about this result were already addressed, but it is atleast arguable that an entirely different skeletal nerve mighthave been correlated with vasoconstriction, or that if prop-erly selected, subparts of the vasomotor field and tibial dis-tribution would have shown strong, but complementary, cor-relations that cancelled in the aggregate measure. Becausethe central mediation hypothesis is intrinsically weak (Elliot,1974, pp. 527-537), excluding every conceivable scenario ofthis type is nearly impossible; however, the substantial co-variance of the ipsilateral CS— responses shows that, in ad-dition to their obvious physical proximity, the particularmeasures chosen can, and at times do, share common sourcesof variance; the diminished correlation of the CS+ responseswould not have been expected if the vascular and skeletalcomponents were inextricably linked in a fixed pattern (Co-hen & Obrist, 1975).
Replicability
The NMB rat is clearly a complicated preparation, and itis reasonable to question whether it is practical for others touse it. The median survival time for the last 10 preparationswas 26 days, and there has been a trend toward longer sur-vival times (Figure 5). Although this longevity is probablymore than adequate for most experimental paradigms, ratshave survived in excellent condition for as long as 90 days.With adequate care, routine survival of 2 months or morecan be obtained if necessary. The longevity of the preparationcreates an unusual problem in maintaining the required skillsfor setting up. Thus, for our own consistency and to assurethat our methods are transferrable to other laboratories, wehave used only surgical techniques that require no more thanaverage competence, prepared step-by-step documentationfor each procedure, used only standard components and re-agents (with a few exceptions), and arranged that the actualcontrol of the preparation and collection of data is automatedand explicitly defined in an easily accessible programminglanguage [1].
312 BARRY R. DWORK1N AND SUSAN DWORKIN
Application to Other Problems
The NMB rat preparation has applications to a variety ofphysiological, medical, and neurobehavioral problems. Bycombining an uncommon degree of control and precision
with a fully intact CNS, the preparation provides a realisticmodel for a variety of studies in surgical intensive care, chronic
immobilization, and neonatology. Because the stabilizedpreparation can be repeatedly anesthetized or narcotized,transducers and electrodes can be placed and manipulatedwithout causing distress or interrupting the constancy of thephysiological state. Immobilization actually reduces post-surgical pain and speeds wound healing. Differences betweensubjects in stereotaxic coordinates and the specific synap-tology of neural networks contributes to error variance inmany experimental designs. By using the NMB rat, multi-stage designs that often include replication can be completedwithin a single subject. By reducing the between-subject vari-ance, the number of animals that are required for statisticallyreliable results can be reduced, which has positive implica-
tions for both laboratory economics and animal utilization.Noninvasive techniques such as nuclear-magnetic resonance,positron emission tomography, and cardiovascular ultra-sound are possible on the fully functioning CNS of an un-stressed small mammal. The rat's phylogenetic position and
small size create limitations for certain procedures, but forother procedures there are substantial advantages, particu-larly those in which homogeneous magnetic, acoustic, or op-tical fields or expensive reagents are required. It is also pos-sible to combine necessarily extended procedures, such asmultitrial learning paradigms or endocrine manipulations,with sensitive stereotaxic manipulations, such as simulta-neous intracellular recording, in vivo microdialysis, and mi-cro-iontophoresis (e.g., see Greuel, Luhmann, & Singer, 1988;Oleson, Ashe, & Weinberger, 1975). Higher CNS sensorystudies, such as auditory threshold or cutaneous discrimi-nations, could possibly benefit from greatly improved sta-bility and accuracy of transducer-sensor geometry.
Humane Treatment and AnimalUse Considerations
Preliminary measurements in several NMB rats of bothbrain regional CJ4-gmcose metabolism and blood catechol-amine levels anticipate that these measurements will be with-in the limits usually found for freely moving subjects. Al-though further studies will be required to fully characterizethe neurochemical state, our extensive physiological and be-havioral data show that the NMB rats develop and maintainnormal levels of HR and BP, have normal responsiveness tomild stimuli, and have well-organized regular sleep patterns(see Figures 1 and 2). None of these findings are consistentwith the usual picture of an organism in distress. Further-more, several scientists who, as experimental subjects, ac-tually experienced pharmacological neuromuscular block didnot in fact find it either painful or frightening (Matin et al.,1982; L. Matin, personal communication, July 1987).
In summary, we believe that our data (Figures 5-12) showthat the chronic NMB rat is a practical preparation for the
study of the physiological substrates of behavior and thehigher CNS control of autonomic function.
References
Barnard, E. A., & Dolly, J. O. (1982). Peripheral and central nicotmic
ACh receptors—how similar are they? Trends in Neuroscience,5(2), 325-327.
Black, A. H. (1974). Operant autonomic conditioning: The analysis
of response mechanisms. In P. A. Obrist, A. H. Black, J. Brener,
& L. V. DiCara (Eds.), Cardiovascularpsychophysiohgy (pp. 229-
250). Chicago: Aldine.
Blackwell, H. R., & Schlosberg, H. (1943). Octave generalization,
pitch discrimination, and loudness thresholds in the white rat.
Journal of Experimental Psychology, 33, 407-419.
Bracewell, R. N. (1978). The Fourier transform and its applications.New York: McGraw-Hill.
Brener, J. (1974). A general model of voluntary control applied to
the phenomena of learned cardiovascular change. In P. A. Obrist,
A. H. Black, J. Brener, & L. V. DiCara (Eds.), Cardiovascular
psychophysiohgy (pp. 365-391). Chicago: Aldine.
Brener, J. (1987). Behavioural energetics: Some effects of uncertainty
on the mobilization and distribution of energy. Psychobiology, 24,
499-512.
Brown, D. A., & Fumagalli, D. (1977). Dissociation of n-bungaro-
toxin and receptor block in the rat superior cervical ganglion. BrainResearch, 129, 165-168.
Campbell, S. S., & Tobler, I. (1984). Animal sleep: A review of sleep
duration across phytogeny. Neuroscience and Biobehavioral Re-
views, 8, 268-300.
Chiappinelli, V. A., Cohen, J. B., & Zigmond, R. (1981). The effects
of a- and /3-neurotoxins from the venoms of various snakes on
transmission in autonomic ganglia. Brain Research, 211, 107-126.
Cohen, D. H. (1980). The functional neuroanatomy of a conditioned
response. In R. F. Thompson, L. H. Hicks, & V. B. Shuyrkov
(Eds.), Neural mechanisms of goal-directed behavior and learning
(pp. 283-302). New York: Academic Press.
Cohen, D. H., & Obrist, P. A. (1975). Interactions between behavior
and the cardiovascular system. Circulation Research. 37, 693-706.
Cowley, R. A., & Trump, B. F. (1982). Pathophysiology of shock,
anoxia, and ischemia. Baltimore, MD: Williams & Wilkins.
DiCara, L. V. (1974). Some critical methodological variables in-
volved in visceral learning. In P. A. Obrist, A. H. Black, J. Brener,
& L. V. DiCara (Eds.), Cardiovascular psychophysiohgy (pp. 276-294). Chicago: Aldine.
Dworkin, B. R. (1973). An effort to replicate visceral learning in the
curarized rat. Unpublished doctoral dissertation, Rockefeller Uni-
versity.
Dworkin, B. R. (1986). Learning and long-term physiological reg-
ulation. In D. Shapiro, G. Schwartz, & R. Davidson (Eds.), Con-
sciousness and self-regulation /K(pp. 163-182). New York: Ple-
num Press.
Dworkin, B. R., Filewich, R. J., Miller, N. E., Craigmyle. N., &
Pickering, T. G. (1979). Baroreceptor activation reduces reactivity
to noxious stimulation: Implications for hypertension. Science.
205, 1299-1301.Dworkin, B. R., & Miller, N. E. (1977). Visceral learning in the
curarized rat. In G. Schwartz & J. Beatly (Eds.). Biofeedback:Theory and research (pp. 221-242). New York: Academic Press.
Dworkin, B. R., & Miller, N. E. (1986). Failure to replicate viscerallearning in the acute curarized rat preparation. Behavioral Neu-
roscience, 100, 299-314.Elliot, R. (1974). The motivational significance of heart rate. In P.
A. Obrist, A. H. Black, J. Brener, & L. V. DiCara (Eds.), Cardiovas-
cular psychophysiohgy (pp. 505-537). Chicago: Aldine.
LEARNING OF PHYSIOLOGICAL RESPONSES 313
Frankel, H., & Mathias, C. (1976). Cardiovascular systems in tetra-
plegia and paraplegia. In P. J. Vinken & G. W. Bruyn (Eds.),
Handbook of clinical neurology: Volume 26 (pp. 313-333). Am-
sterdam: North Holland.
Friedman, L., Bergmann, M., & Rechtschaffen, A. (1979). Effects of
sleep deprivation on sleepiness, sleep intensity, and subsequent
sleep in the rat. Sleep. I, 369-391.
Oilman, A. G., Goodman, L. S., Rail, T. W., & Murad, F. (Eds.).
(1985). Goodman and Oilman's the pharmacological basis of ther-
apeutics (7th ed.). New York: MacMillan.
Greuel, J. M., Luhmann, J. H., &Singer, W. (1988). Pharmacological
induction of use-dependent receptive field modifications in the
visual cortex. Science, 242, 74—77.
Hammerschmidt, D. E., White, J. G., Craddock, P. R., & Jacob, H.
S. (1979). Corticosteroids inhibit complement-induced granulo-
cyte aggregation: A possible mechanism for their efficacy in shock
states. Journal of Clinical Investigation, 63, 798-803.
Harlow, H. F.,&Stanger, R.(1933). Effect of complete striate muscle
paralysis upon the learning process. Journal of Experimental Psy-
chology, 16, 283-294.
Iwata, J., & Ledoux, J. E. (1988). Dissociation of associative and
nonassociative concomitants of classical fear conditioning in the
freely behaving rat. Behavioral Neuroscience, 102, 66-76.
Jamison, J. H. (1951). Measurement of auditory intensity thresholds
in the rat by conditioning of an autonomic response. Journal of
Comparative and Physiological Psychology, 44, } 18-125.
Kelly, J. B.,&Masterton, B. (1977). Auditory sensitivity of the albino
ral. Journal of Comparative and Physiological Psychology, 91, 930-
936.
K.ofke, W. A., & Levy, J. H. (1986). Postoperative critical care pro-
cedures of the Massachusetts General Hospital. Boston: Little Brown
and Company.
Loring, R. H., & Zigmond, R. E. (1988). Characterization of neuronal
nicotinic receptors by snake venom neurotoxins. Trends in Neu-
rosciences, 11, 73-78.
Matin, L., Picout, E., Stevens, J. K.., Edwards, M. W., Young, D.,& MacArthur, R. (1982). Oculoparalytic illusion: Visual-field de-
pendent spatial mislocalizations by humans partially paralyzed
with curare. Science, 216, 198-201.
McNemar, Q. (1962). Psychological statistics. New York: Wiley.
Miller, N. E. (1969). Learning of visceral and glandular responses.
Science, 163, 434-445.Mistlberger, R. E., Bergmann, B. M., Waldenar, W., & Rechtschaf-
fen, A. (1983). Recovery sleep following sleep deprivation in intact
and suprachiasmatic nuclei-lesioned rats. Sleep, 6, 217-233.
Obrist, P. A. (1968). Heart rate and somatic-motor coupling during
classical aversive conditioning in humans. Journal of Experimen-
tal Psychology. 77, 180-193.
Obrist, P. A., Sutterer, J. R., & Howard, J. L. (1972). Preparatory
cardiac changes: A psychobiological approach. In A. H. Black &
W. F. Prokasy (Eds.), Classical conditioning 11: Current researchand theory (pp. 312-340). New York: Appleton-Century-Crofts.
Olcson, T. D., Ashe, J. H., & Weinberger, N. M. (1975). Modification
of auditory and somatosensory system activity during pupillary
conditioning in the paralyzed cat. Journal ofNeurophysiology, 38,
1114-1139.
Oleson, T. D., Westenberg, I. S., & Weinberger, N. M. (1972). Char-
acteristics of the pupillary dilation response during Pavlovian con-
ditioning in paralyzed cats. Behavioral Biology, 7, 829-840.
Pavlov, I. P. (1927). Conditioned reflexes. London: Oxford Univer-
sity Press.Pavlov, I. P. (1932). The reply of a physiologist to psychologists.
Psychological Review, 39, 91-127.
Ray, O. S.,& Stein, L. (1959). Generalization of conditioned suppres-
sion. Journal of the Experimental Analysis of Behavior, 2, 357-361.
Roberts, L. E. (1974). Comparative psychophysiology of the elec-
trodermal and cardiac control systems. In P. A. Obrist, A. H.
Black, J. Brener, & L. V. DiCara (Eds.), Cardiovascular psycho-
physiology (pp. 163-189). Chicago: Aldine.Smith, K. (1954). Conditioning as an artifact. Psychological Review,
61, 217-225.Smith, K. (1964). Curare drugs and total paralysis. Psychological
Review, 71. 77-79.
Solomon, R. L., & Turner, L. H. (1962). Discriminative classical
conditioning in dogs paralyzed by curare can later control dis-
criminative avoidance responses in the normal state. Psychological
Review, 69, 202-219.
Van Twyver, H., Webb, W., Dube, M., & Zackheim, M. (1973).
Effects of environmental and strain differences on EEG and be-
havioral measurement of sleep. Behavioral Biology, 9, 105-110.
Zigmond, R. E., & Loring, R. H. (1988). Characterization and lo-
calization of ganglionic nicotinic receptors using neuronal bun-
garotoxin. In F. Clementi (Ed.), Nicotinic acetylcholine receptors
in the nervous system (NATO AS1 Series, Vol. H25, pp. 31-39).
Berlin: Springer-Verlag.
Appendix
Technical Notes
1. Computer and Software
1.1 The computer system is a Macsym 351 (Analog Devices, Inc.,
Norwood, MA), which consists of a 8086/8087-based general pur-
pose unit and a mated bidirectional analog-digital interface. The
interface operates with some independence and communicates
with the main processor via dedicated 350kb RS422 interface.
The system executes a compiled basic (MACBASIC) at approximately
2000 lines per second. MACBASIC contains a variety of built-in
commands for I/O operations but lacks the vector operations found
in some special technical BASIC dialects. A full-function version
of MACBASIC is available for PC/XT class, but not for 80286- or
80386-based machines. XT bus cards are available to execute most
of the MACBASIC I/O commands. MACBASIC is multitasking and
runs only under CCP/M, which is also supplied by Analog Devices.
It closely resembles a superset of Microsoft BASIC; conversion is
straightforward, except for the I/O functions, which would require
integration of assembly language extensions.
1.2 The software is composed of a series of interrupt-driven, in-
dividual-task modules that are prioritized and run concurrently.
The real-time system is functionally similar to RT/11. The tasks
provide data collection and archiving, calibration of all trans-
ducers and electrodes, monitoring of the physiological and be-
havioral status of the preparation, automatic alerting to out of
range conditions, programmed stimulus presentation for classical
conditioning, and some limited on-line analytical capability, such
as, the generation of histograms and time-averaged responses to
various stimuli. Additional modules for shaping of instrumental
responses and delivering response contingent stimuli were not used
in the present study.
2. Preparation
2.1 Anesthesia and Preoperative Preparation
2.1.1 Methoxyflurane vapor anesthesia:
2.1.1.1 The rat is removed from the transfer cage and placed in a20-L glass container at 25 °C. One milliliter of methoxyflurane is
sprayed into the chamber, and 3-5 min later the rat is lightlyanesthetized.
314 BARRY R. DWORKIN AND SUSAN DWORKJN
2.1.1.2 The preanesthetic reduces excitement and stress of Nem-
butal (sodium pentobarbital) administration and reduces the re-
quired Nembutal dose. It also improves the accuracy of subsequent
intraperitoneal injections.
2.1.2 Nembutal (40 mg/kg ip) the principal anesthetic prior to
tracheal intubation (see 2.1.5), has a pronounced effect on myo-
cardial contractility, and the minimum dose consistent with ade-
quate anesthesia is used. Supplementary doses are given if any
evidence of surgical responsiveness is observed.
2.1.3 Atropine sulfate (0.32 mg/kg sc) reduces secretion prior to
tracheal intubation and minimizes the bradycardia associated with
manipulation of the trachea and carotid vessels.
2.1.3.1 The duration of the peak effect of atropine in the rat is lessthan 1 hr.
2.1.3.2 Atropine must be administered after the bladder cannula
(see 2.2.1) is in place because of its action on the sphincter.
2.1.4 Hydrocortisone sulfate (16 mg/kg sc) is administered. The
mechanism of action of this agent is not fully understood; however,
it prevents accumulation of fluid in the third space at initiation
of positive-pressure ventilation. The action is probably either
through a relatively well-documented effect on renal plasma flow
or by some less known action on the permeability of the capillary
membrane.
2.1.5 Instead of Nembutal, subsequent to insertion of the tracheal
cannula (see 2.2.3), and before induction of paralysis, Forane (iso-
flurane) is administered as required.
2.1.5.1 Prior to paralysis surgical responsiveness is the criterion
for controlling administration.
2.1.5.2 Subsequent to paralysis, mean arterial pressure (MAP) is
maintained at below 90 mmHg and lowered to 75 mmHg before
an incision or major manipulation.
2.1.5.3 Administration of Forane is accomplished with an appa-
ratus which heats the material to an equilibrium pressure of 15
psi (10'1 N/m:) and precision-meters the vapor into the inspiratory
gas stream.
2.1.6 Hydration status is a critical variable:
2.1.6.1 Frequently rats will fail to drink for 12-24 hr after being
moved from the colony.2.1.6.2 To prevent dehydration prior to the induction of paralysis,
the rat is brought from the group cage in the animal quarters to
the laboratory the morning of the experiment.
2.1.6.3 Hypovolemia, subsequent to induction of paralysis and
artificial ventilation, can usually be corrected by the injection of
5-10 ml of Ringers solution. Rats may occasionally require ad-
ditional hydrocorlisone or small amounts (0.2-0.5 ml) of 25%
human plasma to increase oncotic pressure. This requirement is
ordinarily indicated by a total plasma protein of less than 2 g/dl.
Failure to respond to conservative treatment indicates that the rat
should be discarded. Rats do not tolerate dextran-type plasma
expanders.
2.2 Surgery and Transducers
2.2. \ Urinary bladder catheter:
2.2.1.1 For female rats a simple transurethral catheter is satisfac-
tory. The catheter is constructed of silicone tubing (0.6 x 0.12
mm) with a bevel tip. If atropine has not been administered and
the bladder is relatively full, insertion is straightforward.
2.2.1.2 The catheter is replaced every 5 days under fentanyl an-
algesia (sec also 4.1).
2.2.1.3 Using an isotonic filling procedure the bladder is flushed
daily with a solution of Mycostatin (nystatin) and Rcnicidin
(Guardian Chemical, Hauppauge, NY) in saline; these agents in-
hibit yeast growth and prevent mineral stone deposits from ob-structing the catheter.
2.2.2 Electrocardiograph (EKG) electrodes:
2.2.2.1 Electrodes are constructed of silver wire, spiral wound on
18-gaugc hypodermic needles.
2.2.2.2 Placement is accomplished by inserting the needle subcu-
taneously and then withdrawing the needle, leaving the coil in
place. The wire is secured to the skin at the entry point with
cyanoacrylate adhesive.
2.2.2.2.1 EKG electrodes are placed on the ventral thorax near the
right axillary region and 1 cm above the lower thoracic margin
and 1 cm left of the midline.
2.2.2.2.2 Reference electrodes are placed 1 cm left and right of the
midline and 1 cm above the pubis.
2.2.2.3 The EKG amplifier is a dual field effect transistor (FET),
ultralow-bias current device. During long-term continuous re-
cording, significant bias currents corrode the electrodes and de-
posit silver in the tissue.
2.2.3 Tracheotomy and tracheal cannula:
2.2.3.1 The cannula is constructed entirely of transparent thin-
walled Teflon shrinkable tubing, and the design is coaxial; seeMethods section.
2.2.3.2 The cannula is inserted into a transverse tracheal incision
after gradual dilatation using a set of tapered Teflon tubes. Anelastic silicone rubber suture provides a patent seal.
2.2.4 Carotid arterial cannula:
2.2.4.1 Constructed of 30-gauge Teflon tubing with a fire polished
tip. A solvent-shrunk silicone sheath is applied to prevent kinking.
2.2.4.2 Frequency response with a Statham P23db transducer ex-ceeds 10 HA
2.2.4.3 The cannula is placed with the tip protruding 1 mm into
the aorta. Accurate positioning usually can be accomplished by
observing the arterial pulse wave as the cannula tip is advanced.
2.2.5 Vaginal temperature probe:
2.2.5.1 Vinyl-covered thermistor probe is covered with solvent-
shrunk silicone.
2.2.5.2 The probe is coated with Mycostatin ointment and intro-
duced 12 mm into the vagina.
2.2.5.3 The probe temperature correlates well with a similar rectal
probe: however, rectal probes cause intestinal obstruction if used
for more than several days.
2.2.5.4 The probe temperature is controlled by a direct-current
servo temperature controller, which heats the substrate. The max-
imum substrate temperature is limited to 40 °C to prevent dis-
comfort. The substrate temperature is continuously recorded.
2.2.6 Tibial nerve recording and stimulating electrodes:
2.2.6.1 ElectrodcsareofTeflon-insulatedNo. 30multistrandsilver
wire formed into 1 mm hooks.
2.2.6.2 Exposure of the nerve is by a ventral skin incision and
blunt dissection. Care is taken to minimize tissue damage andencourage rapid healing.
2.2.6.3 The hooks are placed 1 cm apart on each nerve, and the
wires are twisted to minimize electromagnetic induction. The en-
LEARNING OF PHYSIOLOGICAL RESPONSES 315
tire nerve with electrodes is embedded in RTV silicone using
stanous octoate catalyst for rapid setting (approximately 10 min)
2.2.6.4 Input amplifiers have specially selected 2N5525 ultralow-
bias JFETs. IgFETs have unsatisfactorily high noise levels for
whole-nerve recording. The amplifiers have isolated relay-switched
front ends for stimulation.
7.2.7 Gastrocnemius electromyograph (EMG) electrodes:
2.2.7.1 Basmajian-type intramuscular electrodes of 80-/*m stainless
wire are used.
2.2.7.2 Amplifier configuration is the same as 2.2.2.3
2.2.7.3 Output to a triggered digital-recording oscilloscope is used
for measuring the evoked potential to a single 30-;ts, 20-jiA stim-
ulus applied to the tibial recording electrodes.
2.2.7.4 The magnitude of the evoked response is used to verify the
adequacy of neuromuscular block.
2.2.7.4.1 Fentanyl analgesia is used prior to the test.
2.2.7.4.2 The standard level is 1% of the preparalysis evoked re-
sponse.
2.2.8 Electroencephalograph (EEG) electrodes:
2.2.8.1 Four 0-80 stainless screws are threaded into the scull at
bregma and lambda, 5mm lateral to the midsagittal suture. Silver-
plated wire is wrapped around the screws on the right side and
embedded in acrylic.
2.2.8.2 Amplifier characteristics are the same as 2.2.2.3.
2.2.8.3 The frontal-occipital lead pairs are band-pass filtered to
measure the ratio of 3-12 Hz to total power. The raw signal is
recorded on a digital audio recorder, displayed on an oscilloscope,
and a dedicated signal-averaging computer, which can be triggered
by a Grass strobe (Grass Instrument Co., Quincy, MA).
2.2.9 Electronic stethoscope:
2.2.9.1 A silicone ceramic crystal microphone is placed in contact
with the skin at approximately 5 mm below the level of the bron-
chial bifurcation.
2.2.9.2 The signal is amplified with a Grass audio monitor and
continuously made audible in the laboratory. Both respiratory and
heart sounds are audible (no routine recording).
2.2.10 Head stabilization fixture:
2.2.10.1 A 4-40 flathead screw is cemented to the skull with acrylic.
2.2.10.2 The screw is suspended from a frame that maintains the
head position to provide a normal head posture.
2.2.11 Protection of the cornea:
2.2.11.1 The eyelids are loosely sutured (1-mm gap) with 5-0 silk.
The suture is necessary to relieve strain, caused by paralysis of
intraocular muscles, which would otherwise damage the optic nerve
and cause pain.
2.2.11.2 The eyes are closed with a thick coat of Mycostatin oint-
ment, which prevents drying and fungus infection. The ointment
is changed every 72 hr under fentanyl narcotic analgesia.
2.2.12 Tail stimulation electrodes:
2.2.12.1 Four stainless surgical staples are placed on the tail 1 cm
apart, beginning 4 cm from the anus. Electrical contact is through
standard microclips.
2.2.12.2 The stimulator is a series-regulated constant-current de-
vice with relay isolation of the secondary circuit and triac controlof the primary. The computer independnetly controls the two
circuits to insure safe operation, and the pulse duration and fre-
quency are hardware-limited to 1 s and 1 pulse per 30 s, respec-
tively; by a simple, reliable resistor-capacitor network. Current is
monitored with an isolated peak-reading digital volt meter.
2.2.13 Optical vasomotor transducers:
2.2.13.1 A stabilized dual-bundle, single quartz-lamp, fiber-optic
light source is located on the dorsal surface of each hind paw. A
filtered silicon low-noise photodetector is aligned axially on the
plantar surface. Initial positioning is accomplished with a me-
chanical fixture; when alignment is established, the foot and trans-
ducers are coated with collodion and covered with black felt to
exlude ambient light.
2.2.13.2 The photo current is read with Analog Devices 310J elec-
trometers configured as current-to-voltage converters. The con-
verter outputs are directly applied to a 16-bit, line-averaging, mul-
tiplexed A/D converter on the computer. The converter is
automatically calibrated before each reading. System stability (0-
I Hz) is better than 0.1% over 30 days.
2.2.14 Per os gastric feeding tube:
2.2.14.1 A standard 5 Fr vinyl infant feeding tube is used. Ordi-
narily the tube is not moved or changed during an experiment.
2.2.14.2 The tip of the tube is passed per os 15 cm beyond the
lower incisors. A standard Harvard infusion pump is operated
continuously to deliver the diet.
2.2.15 Auditory transducer:
2.2.15.1 Standard 75-S! hearing aid-type earphones are used.
2.2.15.2 The earphones are held with the head-support frame, and
positioned on the interaural line, 3 cm either side of the midsagittal
suture.
2.2.16 Environmental control of illumination and noise:
2.2.16.1 The rat's head is enclosed in a 15-cm diameter hemi-
spherical white translucent plastic dome. The external surface of
the dome is illuminated by a cold source at a uniform intensity
of approximately 500 Ix.
2.2.16.2 The ambient noise level in the dome, provided primarily
by continuous operation of the suction system, is approximately
65 dB.
2.3 Induction of Neuromuscular Blockade
2.3.1 While the rat is deeply anesthetized with Nembutal after
placement of the tracheal cannula, but before connection to the
respirator, the neuromuscular blocking agent, a-bungarotoxin(a-BTX) is injected through the arterial cannula. The initial dose
is 100 tig per rat.
2.3.1.1 There is no evidence that a-BTX either crosses the blood-
brain barrier or blocks autonomic transmission.
2.3.1.2 a-BTX in the dose used requires approximately 20 min toparalyze the respiratory muscles.
2.3.1.2.1 Reduction of EKG noise (see 4.2.2.2) indicates that the
block is beginning to take effect.
2.3.1.2.2 Connection to the respirator is attempted and the inspi-
ratory pressure is monitored for negative pressure deflections, which
indicate continued muscle activity; if evident, the respirator is
disconnected and reconnected again several minutes later.
2.3.2 When the respiratory pattern is being smoothly controlled
by the ventilator, the tidal volume is adjusted to an initial value
of 2.2 ml. Twenty minutes later, a 75-^1 blood sample is drawn,
and the Pa0,, PCO;, and pH are measured with a modified Radi-ometer BMS-3 apparatus. The tidal volume is readjusted if nec-
essary to P0, > 80 mmHg and Pco, = 35^»5 mmHg.
316 BARRY R. DWORKIN AND SUSAN DWORK1N
3. Stabilization of the Preparation
3.1 Maintenance of Surgical Anesthesia
3.1.1 Blood pressure (BP) and EEG arc used as indices of adequate
anesthesia.
3.1.2 Forane is used during the initial 24 hr:
3.1.2.1 Highly controllable.3.1.2.2 Does not depress myocardium.
3.1.2.3 Vasodilatation reduces afterload.
3.1.3 Narcotic analgesia is ordinarily adequate in second 24 hr:
3.1.3.1 Rats require relatively high dose levels and develop tol-
erance rapidly:
3.1.3.1.1 Morphine approximately 5 mg/kg iv every 4 hr.
3.1.3.1.2 Fentanyl approximately 3 M&/g iv every 4 hr.
3.1.3.2 Lack of movement promotes healing and receptor adap-tation; thus, postsurgical discomfort is of limited duration, and
there is usually no need for analgesic therapy to extend for more
than 72 hr.3.1.3.3 Heart rate (HR), BP levels, and EEG characteristics indicate
lack of serious distress.3.1.3.4 Regular sleep cycles and responsiveness to mild stimuli
develop in the undisturbed preparation within 72 hr.
3.1.4 The preparation is attended continuously either directly or
via a dedicated communications system (see 5.0). The response
time is always less than 10 min.
3.2 Stabilization of Fluid and ElectrolyteMetabolism
3.2.1 A central venous catheter is not used routinely:
3.2.1.1 Earlier in the development of the preparation, central ve-
nous pressure (CVP) was monitored to study changes in inlra-vascular volume with paralysis and positive-pressure ventilation.
The constantly infused, relatively large caliber, venous catheter
was prone to form clots after several weeks.3.2.1.2 Continuous measurement of the CVP has been unnecessary
on a routine basis because the QRS amplitude and HR provide
sufficiently accurate indications of central volume for management
of the preparation:
3.2.1.2.1 Steady-state levels of HR, variability, and QRS peak-to-
peak amplitude can be established for a given subject.3.2.1.2.2 Dynamic responses to intravenous challenges of 1-2-ml
boluses of saline are more accurate indicators and can be inter-
preted across subjects.
3.2.2 Ordinarily, after the initial 96 hr, Na* is well regulated andspecial intervention is not required. The parenteral mixture con-
tains no added Na*, except that associated with the various drugs.The intragastric feeding mixture contains various minerals.
3.2.3 The K concentration of the parenteral mixture is occasion-
ally varied to maintain serum levels within normal limits.
3.2.4 The standard parenteral mixture contains, per 100 ml, the
following:
5.0 g dextrose0.15 mg Synkavite (vitamin K)
50 mg sulfisoxazole8.0 mg gentamicin200 mg oxacillin
200 1U heparin
800 ^g a-BTX (see 3.4.2)
3.2.5 The normal infusion rate into the arterial catheter is 1.15 ml/hr.
3.3 Nutrition(see 2.2.14 for feeding tube description)
3.3.1 In the first 48 hr, feeding is restricted to 1.15 ml/hr of 5%dextrose in water.
3.3.2 Subsequently, the following mixture (per 100 ml of water) is
infused intragastrically at a rate of 1.15 ml/hr:
65 g High Nitrogen Vivonex10'UMycostatin
18 mg Fer-In-Sol iron drops (iron supplement)
3.3.3 This regimen has been found to approximately maintain the300-g initial weight of the female rats for as long as several months.
3.4 Maintenance of NeuromuscularKlockade (NMB)
3.4.1 Measurement of degree of NMB:
3.4.1.1 The classical method of measurement is stimulation of askeletal nerve and measurement of the muscle evoked potential.Comparison to the unblocked response permits an accurate assess-
ment of the remaining fraction of normal junctional transmission.This can be accomplished under fentanyl analgesia by stimulating
the tibial nerve through the electrode (2.2.6) and measuring theevoked response with gastrocnemius EMG electrodes. The necessarydose of a-BTX to maintain less than 1% transmission is established
by direct titration.3.4.1.2 Alternatively, high-frequency (1-3 kHz) noise present on
the EK.G electrodes reflects activation of the underlying intercostalmuscles (2.2.2). The noise increases from less than I ^V (peak-to-peak) at maximal block (i.e., when transmission through the junctionis undetectable with averaging) to as much as 40 i/V (peak-to-peak)
when transmission reaches 5-10%. Thus, monitoring of the EK.Gnoise permits a continuous passive, but less accurate, estimation of
the depth of block.
3.4.2 a-BTX is added directly to the vascular infusion (see 4.4.1.2).There is some variation between subjects in required dose; the range
is 150-250 /ug/day. There is a substantial therapeutic ratio, becausedoses of 2,000 /ig/day do not appear to produce obvious autonomicor central effects.
4. Monitoring and Controlling of the
Physiological State
4.1 Urine
4.1.1 Flow is continuously monitored and recorded with a digitalscale of 10-mg (0.01 -ml) accuracy. An artificial (hydrostatic) sphinc-ter arrangement maintains the normal tone of the detrusor and cyclicfilling and emptying of the bladder. The normal rapid flush producedby the natural contraction reduces urine stagnation and the chance
of infection.
*4.1.1.1 Urinary obstruction is detected rapidly from abrupt changesin the output pattern. (Asterisks preceding subsection numbers refer
LEARNING OF PHYSIOLOGICAL RESPONSES 317
(b)
INSPIRATION SECTION
OF BUBBLE-TIGHT
ROTARY CYCLE
CONTROL VALVE
HUMIDIFIER
WATER PUMP
Ic!
INSPIRATION
PRESSURE
TRANSDUCER
HUMIDIFIER
REGULATORS
ELECTRONIC SERVO
GAS FLOWMETER
AND CONTROLLER
500 cc
RESERVOIR
(d)
•(g)
Figure AI. Diagrammatical characterization of the artificial ventilation system. (Letters in paren-
theses are cross-referenced within section 4.3 of the Appendix.)
to conditions that are continuously monitored by an alarm system;
see 5.1.)
4.1.1.2 Diuretics are occasionally administered to correct inade-
quate urine flow of renal origin. Positive-pressure ventilation can
cause hydrostatic fluid retention.
4.1.2 Specific gravity is measured daily by optical refraction; pH,
protein, and glucose are determined with standard clinical test strips.
4.1.3 Daily microscopic examination using phase optics is per-
formed on sediment for detection of abnormal numbers of eryth-
rocytes, yeast, or bacteria.
4.1.3.1 Bacterial infection rarely occurs, but if evident, a culture is
prepared to choose appropriate antibiotics.
4.1.3.2 Yeast infections occur occasionally, but are usually con-
trolled with a concentrated Mycostatin flush.
4.2 EKG
4.2.1 The amplified filtered signal activates a Schmitt trigger, which
produces a logic pulse for the digital counting circuits and for a Grasstachograph.
*4.2.1.1 The interbeat interval (IBI) is measured by a precision
clock to an accuracy of 10 MS.
4.2.1.2 IBI is also converted to HR by the tachograph and displayed
continuously on a polygraph channel. This display is independentof the computer system.
4.2.2 The amplified signal is displayed on a digital oscilloscope
with storage and disk recording and data analysis capabilities.
4.2.2.1 The amplitude and R-wave slope are evaluated as rough
indices of myocardial function and adequacy of venous return to the
heart.
4.2.2.2 High-frequency baseline noise is reliably related to activity
in the intercostal muscles and serves as a continuous secondarymeasure of neuromuscular blockade.
4.2.3 Extended (2 hr) detailed (1 Hz to 20 kHz) records of the EKG
topography are made on a two-channel digital audiotape recorder.
The recorder includes marker channels for correlation with experi-
mental procedures.
4.3 Characteristics of the ArtificialVentilation System
Letters in brackets refer to Figure AI labels.
4.3.1 Tracheal cannula [h, i] (see Method section):
4.3.2 The inspiratory gas mixture is supplied from tanks and is
composed of 20-40% O,, with the balance N,. The Oj fraction is
varied to maintain a Pa,,, of 100-140 mmHg and Pa^, of 35^15
mmHg. In practice the Pat 0. is established by setting the tidal vol-
ume, and the percentage of O: is manipulated to bring the Pa^ to
within the specified range. During stabilization most rats display
some degree of ventilation-perfusion imbalance due to atelectasis
and slight pulmonary edema. This is ordinarily self-corrected within48 hr.
4.3.3 The respirator is a custom-designed, custom-built rotary-
valve device incorporating only a single moving part [b, fj. Inspi-
318 BARRY R. DWORKIN AND SUSAN DWORKIN
ratory:expiratory ratio is fixed at 1:2, and the rate is ordinarily fixed
at 0.83 Hz but is potentially variable. A constant minute volume isdelivered, but a safety valve limits the peak inspiralory pressure to
20 cmH,O. Because the usual operating range is less than 10 cmH,O,
the pressure limit is ordinarily only exceeded during programmed
hyperinflations (see 4.3.4). The instrument has high accuracy, sta-
bility, and reliability.
4.3.3.1 The minute volume can be specified to within 1% [g].
4.3.3.2 The minute volume remains within 0.1% of the set value
for more than 30 days.
4.3.3.3 The system is operated from an uninterruptible battery-
backed power supply and has a mean time before failure of greater
than 5 years (continuous operation).
4.3.4 Control of respiratory cycle pressures:
4.3.4.1 The positive end-expiratory pressure (PEEP) is controlled
by a pneumatically biased diaphragm valve located in the expiratory
circuit between the cannula and the respirator [i]. The PEEP is or-
dinarily maintained at 2—4 cmH,O throughout the experiment; how-
ever, if during the stabilization period there is evidence of incipient
pulmonary edema, then the PEEP may be increased as high as 10
cmH2O for several hours.
4.3.4.2 A periodic hyperinflation is programmed at approximately
10-min intervals. The peak pressure is controlled at 20 cmH:O, and
the duration is 1.5 s [m].
*4.3.4.3 The pressure in the expiratory channel of the cannula [i]
is continuously measured by a strain-gauge transducer [k], displayed
on an analog polygraph channel, and monitored by the computer.
The analog display is independent of the computer.
4.3.5 Humidification and mucous flow:
4.3.5.1 Humidification is accomplished by the addition of vapor-
ized distilled water to the inspiratory circuit. The water is introduced
with a high-pressure metering pump at a rate of 1.5 ml/hr through
a heated (40 °C) low-volume vaporizer [c].
4.3.5.2 Because of the coaxial design of the catheter [h, i], mucous,
which is kept liquid by adequate humidification, is continuously
discharged through the expiratory channel. To prevent obstruction
of the respirator orifices, the mucous is collected in serial ultralow-
volume traps [j], which are continuously pumped clear.
4.3.6 A synchronous 5% end-expiratory sampling valve [d] can be
connected to a standard CO2-analysis device [e] for continuous re-cording of end-expiratory CO,. Approximately five samples are re-
quired to fill the chamber, and the time constant for the Beckman
LB-2 (Beckman Instruments, Inc., Shiller Park, IL) operated in this
mode is approximately 30 s.
4.4 Measurement and Recording of
Arterial Blood Pressure
4.4.1 The arterial cannula and transducer (see 2.2.4):
4.4.1.1 Calibration of the transducer is done daily with a pressure
manometer.4.4.1.2 A continuous infusion of Ringers and dextrose containing
a-BTX, antibiotics, vitamin K, and heparin flows from a Harvard
infusion pump into the cannula at a rate of 1.15 ml/hr. The exact
proportions depend on the specific experimental protocol, clotting
time, serum electrolytes, and so on (see 3.2.4 for standard values).
4.4.1.3 The cannula has a special ultralow-volume valve system
for blood sampling. A 70-nl sample is ordinarily required for a
complete analysis. The cannula is highly reliable, and usually re-mains clear for the entire experiment (i.e., 2-12 weeks, without
intervention). However, each cannula has a precision-cut nylon
obdurator that can be used to dislodge small clots.
*4.4.2 The output of the arterial pressure transducer is displayed
on a polygraph channel, which is independent of the computer. The
signal is also applied to the computer analog-to-digital converter for
digital recording of MAP and pulse amplitude. Systolic/diastolic
values are determined by their relationship to the EKLG rather than
by peak detection.
4.5 Temperature Regulation and Recording
*4.5.1 Core temperature probe (see 2.2.5).
*4.5.2 Interface temperature probe:
4.5.2.1 Measures the temperature between the ventral surface of
the rat and the heated substrate.
4.5.2.2 A high-accuracy Analog Devices 540 semiconductor de-
tector is connected directly to a multiplexer on the computer. Cal-
ibration is against a National Bureau of Standards-certified ther-mometer. All other thermometry is referenced to this device.
4.5.3 Substrate and substrate temperature probe:
4.5.3.1 The substrate is a 1-cm-thick sheet of nickel-plated copper
to which a 16-gauge sheet of stainless steel has been bonded with
thermally conductive silicone adhesive. Silicone-embedded heating
mesh is bonded to the underside of the copper.
4.5.3.2 A thin thermistor probe is bonded to the top surface of the
copper and measures the substrate temperature.
4.5.3.3 The temperature uniformity of the substrate is better than
±0.5 °C. The substrate temperature is limited to 40 °C by the reg-
4.5.4 Regulation of temperature is accomplished with a dual-loop,
direct-current servo-regulator. The principal reference is the vaginal
temperature probe. When a discrepancy from the set point is de-
tected, a current is passed through the heating mesh that is sufficient
to bring the interface temperature to a value calculated to be optimal
for correcting the discrepancy without causing significant overshoot;
thus, the interface temperature, rather than heater power, is regulated
as a function of core temperature deviation. An independent-refrig-
eration cooling system is available in case of failure of the central
air system or the need of supplementary cooling.
4.5.5 Core temperature is ordinarily clamped by this regulator at
38 ±0 .1 °C; however, it may be varied by computer control.
4.6 Skeletal Nerve Recording and Stimulation
4.6.1 The nerve signals are amplified (see 2.2.6.4) and filtered with
a band-pass of 0.3-3 kHz. A level detector produces a pulse for each
spike at or above the detection limit. The output pulses are applied
to a hardware counter, which is part of the computer system. The
resulting counts are log,,, transformed, and the transformed values
are assigned to the nerve activity variables.
4.6.2 The level detectors are set 72 hr after the induction of pa-
ralysis to produce a rate of 300-700 impulses per second. The
correlation between the left and right nerves is continuously com-
puted and displayed. The relative levels of the two detectors are
adjusted to approximately maximize the correlation.
4.7 Measurement oj'EEG andSleep-Wakeful ness Cycles
4.7.1 Electrodes and preamplifier configuration (see 2.2.8).
LEARNING OF PHYSIOLOGICAL RESPONSES 319
4.7.2 The regularity and topography of the sleep-wakefulness cycles
are important indices of the physiological and behavioral status. EEC
cycles ordinarily have a period of 20 ± 10 min. The cycles have a
complex form but are relatively well characterized by variations in
the ratio between the 3-12-Hz band and the total energy; ±6 dBvariations are typical.
4.7.3 The log of the energy ratio is recorded on a polygraph channel
and recorded digitally in computer-readable form. In addition, fast
Fourier transform analysis is performed on the raw data to char-
acterize the frequency content of the signal at several different energy
ratios.
4.7.4 The raw EEG is sampled periodically and recorded in VHS
format using a two-channel digital audiotape recorder (1 Hz to 20
kHz)
4.7.5 A flash-evoked potential measurement from frontal-occipital
cortical electrodes is performed every 96 hr. These data are a useful
index of gross cortical function, and attenuation of the signal fre-
quently anticipates behavioral evidence of deterioration in central
nervous system function.
4.7.6 Autonomic and skeletal outflow is correlated with the EEG
cycles, and examination of the polygraph record is a further quali-
tative verification of the rat's overall condition.
4.7.7 Desynchronization of the EEG, tachycardia, hypertension,
vasoconstriction, and tibial nerve firing arc ordinarily observed with
mild disturbance such as noise or light touching. The duration of
these responses depends on the strength of stimulation and degree
of habituation from previous stimuli.
4.8 Measurement and Recording of Peripheral
Vasomotor Activity
4.8.1 Vasomotor transducers are of the optical transmission type
(see also 2.2.13).
4.8.1.1 The electronics and optics are of a very high-stability de-sign, and the characteristics of these instruments are different from
those of commercially available devices. Ultralow-noise silicon pho-
todiodes are used in the current-output mode, which results in linear
response over 9-10 decades of light intensity. The Analog Devices
3IOJ varactor diode electrometer amplifiers, configured as current
to voltage converters, drift less than 1 part in 10' per month.
4.8.1.2 The light source uses a 0.1% stabilized power supply, and
the quartz-iodine lamp is run at 60% of normal voltage to reduce
aging effects. In addition, because of the use of a split random glass-
fiber bundle for transmitting the light from the lamp to the two
detectors, the left and right transducers "see" the same light source.
An intentionally long fiber-optic path and a dichroic filter eliminate
virtually all infrared and, consequently, local heating. Heating of the
tissue is a significant problem with most commercial transducers.
4.8.2 Vasomotor activity is displayed on the polygraph as Z-trans-
formed data and recorded in computer-readable digital format. Ab-
solute light intensity is also recorded.
4.8.2.1 Greater light transmission occurs with increased vasocon-
striction. The photo-current can be shown to be a nonlinear mono-
tonic function of phenylephrine dose and responds appropriately to
standard physiological challenges such as hemorrhage or rapid vol-ume expansion.
4.8.2.2 When the right and left transducers are applied, the absolute
percentage transmission and variability with standard manipulations
of vasoconstriction are carefully matched; then, the transducers arecemented to the feet with collodion. Along with a simple mechanical
support, the collodion accurately maintains the transducer orien-
tation for as long as several months.
4.9 Measurement of Blood Pa0,, Pai0j,pH,
Na*, K', Hematocrit, Protein, and Glucose
4.9.1 A volume of blood (70-100 n\, depending on the hematocrit)
is drawn from the arterial catheter every 24 hr. Special connections
are used. The sample is introduced into the gas-measurement cham-
ber of a modified Radiometer BMS-3 analysis system. Pa0] and Paco,
are determined.
4.9.2 The sample is anaerobically withdrawn into a precalibrated
capillary tube from which 30 fil is drawn into the pH electrode and
read.
4.9.3 The remainder of the sample is centrifuged in a clinical cap-
illary head for 3 min, and the hemalocrit is determined.
4.9.4 The tube is cut to provide a 20-^1 sample for Na* and K*
determination using a standard micro-flame photometer.
4.9.5 The tube is cut again and the remainder of the sample applied
to a refractometer for measurement of approximate protein (a vapor
osmometer is used alternately).
4.9.6 The plasma is absorbed off from the refractometer cell with
an enzymatic glucose strip.
4.9.7 The residual cells (approximately 80% of the original sample)
can be resuspended and returned to the preparation. However, most •
preparations maintain a normal hematocrit (35-45%) without rein-jection.
5. Automated Vital Signs and Central-StateAlarm System
5.1 The alarm system continuously monitors HR, BP, urine out-
put, and body temperature. If any variable deviates outside the preset
range, then an alarm condition is established. The conditions mon-
itored are marked in Section 4 with an asterisk. The alarm system
is computer based. The computer is operated from an emergency-
backed power system. Auditory and visual signaling occur in thelaboratory with automated problem identification on a CRT.
5.2 An investigator's home is connected to the laboratory computer
system through a dedicated 2400-baud circuit. A terminal monitors
the condition of the preparation every 6 min, with detailed updates
of all variables described in Section 4. In addition the alarm system
provides instantaneous notification of any out-of-range condition by
activating an 80-dB signal, which is clearly audible throughout the
house. The time to respond to an alarm is ordinarily less than 10
min from initial activation to arrival in the laboratory. By activating
a special code from the remote terminal, the rat can be given a 75-
mg/kg ip injection of Nembutal.
5.2.1 Failure of the alarm system:
5.2.2 In more than 500 actual experimental days, the alarm and
remote terminal system have not failed.
5.2.3 Failure to respond to an alarm within 20 min automatically
activates the emergency Nembutal injection system.
Received April 14, 1989Revision received July 5, 1989
Accepted July 5, 1989 •