An Addiction Science Network Resource

Reprinted from G. Fouriezos and E. Nawiesnaik (1987), A comparison of two methods designed to rapidly estimate thresholds of reward brain stimulation. In M.A. Bozarth (Ed.), Methods of assessing the reinforcing properties of abused drugs (pp. 447-462). New York: Springer-Verlag.
Brain Reward System
Back to book TOC
(Search book contents) (Search entire ASNet)

Chapter 21

A Comparison of Two Methods
Designed to Rapidly Estimate Thresholds
of Rewarding Brain Stimulation

George Fouriezos & Edward Nawiesniak

School of Psychology
University of Ottawa
Ottawa, Ontario, Canada K1N 6N5

Four experiments evaluated the relative merits of two procedures for rapidly estimating thresholds of rewarding brain stimulation. In the first method termed autotitration, rats earned brain stimulation by pressing a lever, and the stimulation became progressively weaker in pulse frequencymulation by pressing a lever, and the stimulation became progressively weaker in pulse frequency with continued responding. The depression of a second lever restored the stimulation to its original strength. In the second protocol or timed method, rats worked through the same descending sequence of frequencies and eventually quit responding but, here, they had no control over resets of stimulation; resets were automatically triggered at timed intervals by the equipment. Experiment 1 showed that autotitration consistently produced higher reset thresholds than the threshold estimates of the timed method. The second experiment showed that autotitration’s reset thresholds climbed when the stimulation sequence began at twice the customary frequency but that thresholds derived from the timed method remained at control levels with this double-frequency test. Reset thresholds from autotitration were found in Experiment 3 to be strongly influenced by changes in only the first frequency of the sequence, but this manipulation had no effect on timed-method estimates. Finally, the last experiment illustrated how autotitration might miss or seriously underestimate the effect of a pharmacological treatment. Together, these data suggest that resets made in autotitration are influenced by the value of the reinforcer made available immediately after the reset but that the alternative, timed method, is free of this problem.



Two somewhat intertwined objectives prompt the study of drug effects on intracranial self-stimulation. One aim is to learn the neurochemical nature of components of the brain’s reward circuitry. By inspecting the effects of compounds with well-documented, relatively specific neurochemical actions, psychopharmacologists hope to identify the neurotransmitters or neuromodulators at the various stages of the circuit. A second concern is to use self-stimulation as a model or an assay for the abuse liability of psychoactive substances. The rationale here is that some compounds may possess abuse potential via their interaction with central reward circuits. Such interactions might be revealed or screened by assessing the ability of drugs to lower the threshold for rewarding brain stimulation.

Threshold estimation is a time-consuming endeavor when conventional procedures are used. It involves measuring the behavioral vigor displayed in earning each of several values of a stimulation variable such as pulse frequency or current amplitude. A threshold estimate is derived by determining either the lowest stimulation value that reliably produces a just noticeable departure from no responding or one that elicits a constant but moderate level of behavior. To assess a drug effect, the procedure is repeated following injection of each of several doses of the compound. The time requireddoses of the compound. The time required to perform these assessments becomes a serious obstacle when drugs with short-lived actions are tested because such drugs cannot be trusted to sustain plateaus of effect for the 15 minutes or so needed to obtain one estimate.

To reduce the time required to estimate self-stimulation thresholds, some investigators have recently re-introduced the Stein and Ray (1960) self-determined "threshold" method. Here, rats are tested in a chamber containing two levers. The depression of one lever triggers a brief train of brain stimulation. Continued responding earns weaker and weaker stimulation; in most paradigms it is the current intensity that drops with successive stimulations. The second lever delivers no stimulation. Its depression instead resets the intensity of the available stimulation to the original suprathreshold value. The rat may then return to the first lever and run through the descending sequence again. The current when the reset occurs is deemed to be the threshold. Each excursion through the sequence takes a fraction of a minute; the speed in amassing data thus allows the assessment of short-acting compounds, provides a finer resolution to time-course studies, and enables testing a greater number of subjects. The reader is referred to Schaefer, Baumgardner, and Michael (1979) for a recent description of control circuitry.

The autotitration method has seen extensive use in psychopharmod has seen extensive use in psychopharmacological experiments. Its earliest (Stein, 1962; Stein & Ray, 1960) and most frequent application involved catecholamine manipulations (Schaefer & Holtzman, 1979; Schaefer & Michael, 1980; Seeger & Gardner, 1979; Zarevics & Setler, 1979), but it has also been used to examine the role of GABA in reward (Nazzaro & Gardner, 1980; Zarevics & Setler, 1981) and to investigate the effects of morphine on self-determined thresholds (Nazzaro, Seeger, & Gardner, 1981).

Despite its recent popularity the autotitration technique has not been validated; indeed, at least six cautionary notes about its use or concerning the interpretation of the self-determined thresholds have appeared. Stein (1961) noted that rats tested in autotitration reset at higher currents than those necessary to sustain responding; thus self-determined estimates of threshold were higher than those obtained by more conventional procedures. A similar, possibly related problem was noted by Gardner (1971) who found that his Rhesus monkeys would reset at higher and higher currents with experience in the task. (Rats, however, were free of this problem.) Gardner’s solution was to impose a time-out of up to 5 seconds after a depression of the reset lever. A task- or effort-dependence of autotitration reset-thresholds was noted by Neill and Justice (1981; Neill, Gaar, Clark, & Britt, 1982) who found that the & Britt, 1982) who found that the distribution of reset currents shifted up when a barrier that partially separated the stimulation and reset levers was removed. Valenstein (1964) raised two concerns. In addition to the fear that the method may lack general applicability—some positive brain sites, because of the low rates or the seizures they engendered, were poor candidates for evaluation by autotitration—Valenstein’s main concern was that the demonstrated ability of a drug, such as amphetamine, to reduce self-determined thresholds might incorrectly be interpreted as a facilitation of reward. The alternative was that amphetamine potentiated perseverative tendencies for reasons independent of any action on reward circuitry, and the resulting persistent pressing of the stimulation lever would artificially drive down the self-determined thresholds. Finally, in his thorough review of techniques designed to distinguish between drug effects on reward and those on general performance, Liebman (1983) noted that chlorpromazine has not always been found to increase reset thresholds (Schaefer & Michael, 1980) and that picrotoxin has been reported to decrease (Nazzaro & Gardner, 1980) and to increase (Zarevics & Setler, 1981) reset thresholds. Sufficient conceptual concerns and empirical discrepancies surround autotitration to justify its re-evaluation as a useful research tool.

Derived from autotitration, a second method was im autotitration, a second method was introduced by Leith and Barrett (1980, 1981; Leith, 1983) which shares autotitration’s speed of data acquisition. Like autotitration this method features a rapid progression through a descending sequence of currents, but unlike autotitration the rats in this second method are not permitted to depress a reset lever nor does their behavior in any way promote a reset. Instead, resets to the initial current occur under the control of programming equipment; in the Leith and Barrett method and in the "timed" method of this report, trials are started (the equipment resets) at fixed intervals of time.

This paper describes an evaluation of these two approaches to rapid estimation of reward thresholds. The first experiment examined long-term stability of the estimates, and the next two challenged the robustness of the estimates with alterations in the sequence of rewarding stimuli. The final experiment tested the prediction derived from the results of earlier tests that autotitration may fail to detect some genuine drug effects.

Experiment 1

This experiment was designed to compare thresholds obtained from the timed and autotitration paradigms. Five measures from each rat were taken using both methods in five consecutive days in the first and last weeks of the 6-week protocol. (Experiments 2, 3, and 4, described below, were conducted in the intervenibed below, were conducted in the intervening weeks.) Thus in addition to a comparison of the two methods, this experiment provided an assessment of within- and between-session reliability with a determination of long-term stability of the measures.



Under sodium pentobarbital anesthesia (60 mg/kg, intraperitoneally) five male Long-Evans rats weighing between 340 and 620 g at surgery were stereotaxically provided with a chronic, monopolar stimulating electrode (0.25 mm diameter.) aimed at the lateral hypothalamus. The heads were fixed in the plane of de Groot (1959) and the stainless steel, Formvar-insulated electrodes were directed with the following coordinate sets: 1.0 mm caudal to bregma (C), 1.5 mm lateral to midline (L), and 8.0 mm below dura (V) (n = 1); 1.0C, 1.0L, 8.0V (n = 1); and 0.4C, 1.7L, 8.0V (n = 3). A wire wrapped around four skull screws served as the current return.


The floor of the wooden test chambers measured 30 x 36 cm and the walls were 38 cm high. Two Lehigh Valley rodent manipulanda were located opposite each other centered on the 30 cm walls. The levers extended 25 mm into the boxes and their top surfaces were 4 cm above the 13 mm grid floor.

A constant-current generator (Mundl, 1980) provided stimulation pulses that were continuously adjustable to 1 mA. Although the pulses were always monophasic, the unit incorses were always monophasic, the unit incorporated a low resistance shunt to ground between pulses to prevent a temporal accumulation of charge at the electrode. The current was adjusted and monitored by reading the voltage drop across a 1 kohm resistor in series with the current return.

The remaining stimulation variables were controlled by a pulse generator that automatically advanced through a programmed sequence of pulse frequencies. This unit controlled both the autotitration and the timed paradigms, and it accumulated the following data: a running total of the delivered stimulations, the number of resets made, the time spent at each frequency, and the number of responses made in earning the eight stimulations at each step.


Figure 1 represents a cumulative record (with the stepper pen traveling down) to illustrate the similarities and differences between the present implementations of the autotitration and timed reset methods. In both methods rats pressed a stimulation lever to deliver a 0.5 second train of 0.1 msec cathodal pulses. All responses were counted but only those occurring after trains were delivered could reinitiate a stimulation train. Current intensities, individually selected in preliminary training sessions for each rat, remained unchanged throughout all testing; in these studies the frequency of pulses in the train decreased with repeated responding. Pulse frequencies followed a desceng. Pulse frequencies followed a descending series with shifts from one frequency to the next occurring after every eight stimulations. The sequence followed approximately a 0.1 log unit progression and the standard or baseline sequence was 100, 80, 63, 50, 40, 32, 25 . . .Hz continuing for a total of 16 steps. The first five steps of this standard sequence are represented by the horizontal bands (labeled 100, 80, 63, et cetera) in Figure 1. The resets to the initial frequency, whether triggered by machine or by rat, were signaled by a brief tone.

The chief difference between the paradigms was in triggering resets to the initial frequency. In the timed method they occurred under timer control at 1 minute intervals without regard to the behavior of the rat. This is shown in the left-hand panel of Figure 1 as the vertical resets of the stepper pen occurring well after responding has ceased which, in turn, is represented by the horizontal trace of the recorder. In the autotitration method the second lever was active; its depression restored the stimulation to the initial frequency (Figure 1; right-hand panel).


Rats were given at least 3 days recovery after surgery before screening for self-stimulation and training in the timed method. Immediately following the machine-produced resets, rats were shaped or primed to self-stimulate. Currents were initially adjusted to obtain self-stimulation performanced to obtain self-stimulation performance over most of each minute-long, inter-reset interval. Shaping and priming were occasionally omitted, and these omissions became more frequent as the session progressed. Once the rat demonstrated proficiency in this paradigm as indicated by repeatedly quitting at about the same frequency and by rapidly returning to the stimulation lever upon tone signaled resets, the current was reduced so that quitting occurred at 50 or 40 Hz (4th or 5th steps). This training session lasted about 2 hours.

Cumulative records comparing timed resets and autotitration
Figure 1: Comparison of methods: representation of a cumulative record with time running left to right and stepper traveling down. In autotitration rats earn eight 0.5 second trains of pulses at 100 Hz, then eight at 80 Hz, then 63 Hz, etc. Stimulation may be reset by the rat to 100 Hz at any time by depression of another lever; resets are indicated by vertical rises of trace to top of 100 Hz band. In the timed method the rat cannot reset. The rat can obtain reinforcements by pressing the stimulation lever or it can quit responding as is shown by the horizontal tracks of the recorder pen. Here, resets are controlled by a timer adjusted to reset at 1 minute intervals.

The rats were then trained in the autotitration paradigm. Initially, shaping was used to draw the rat to the reset lever after it had quit responding on the stimulation lever. Several reset responses were manually reinforced with brain stimulation at the beginning of this training: The number of rewarded resets was reduced to one or two after about 10 shuttles between the levers; rewards for resets were intermittently omitted over the next 30 shuttles or so; and, finally, they were omitted altogether. This training session, too, lasted about 2 hours.

Finally, the rats were given daily training sessions in both paradigms for at least 6 days. These sessions lasted 20 minutes and their order of presentation was reversed daily. Minor re-adjustments in current were made at this stage, but each rat received the full 6-day training at the final current. These currents ranged from 160 to 250 mA across subjects, and they remained unchanged for the remainder of the tests.


Rats were tested for 5 consecutive days in both paradigms during Week 1 and again in Week 6. Both daily sessions lasted 25 minutes; the first 10 minutes constituted a warm up period when no data were collected and the last 15 minutes yielded five threshold estimates. The 3 minutes devoted to each estimate were paes. The 3 minutes devoted to each estimate were partitioned as a 2-minute trial followed by 1 minute used to retrieve data. Although no new data were amassed during retrieval, this process did not interfere with the ongoing testing; that is, from the rat’s perspective, this third minute was no different from the previous 2 minutes. The two methods were presented in opposite orders from day to day and about 2 hours separated them.

Data Analysis

In the majority of autotitration implementations, estimates of threshold are derived by noting the value of the stimulation when the reset is made. In our version we obtained such measures by counting the number of stimulations earned and of resets made in a 2-minute period. Their ratio gave the average number of stimulations per reset, and because each frequency was offered for eight stimulation trains, division of the ratio by 8 yielded the average number of steps per reset. This value was used to interpolate the frequency at which the resets were made; this latter value is termed reset frequency (FRST) in this paper.

In the timed method the rates of responding for each frequency through the sequence were used to interpolate the frequency corresponding to a rate of 0.5 responses per second; we refer to this value as required frequency (FREQ). The theoretical justification of this and of related procedures has been presented elsewhere (Gallistel, Ss been presented elsewhere (Gallistel, Shizgal, & Yeomans, 1981).

Although we wished to use identical analytic methods in the two paradigms, this proved to be impossible. Rats in autotitration rarely drop to rates low enough to permit interpolation, and when tested in the timed method, they do not reset. Moreover, noting the frequency offered when the machine performed the reset would have been misleading because, after having quit at a given frequency, rats in the timed method often emitted anticipatory responses just before the reset. These anticipatory responses usually advanced the pointer to the next frequency. Although unequivocal quitting might have occurred at one frequency, the one available at the machine-produced reset may have been one or two steps lower.

Results and Discussion

The data from these baseline determinations are illustrated in Figure 2. Three features are immediately apparent. First, there seems to be more within-session variability in the timed method than in autotitration. This, we think, is partly an artifact of our fixed trial duration of 2 minutes. In autotitration 2 minutes sufficed for about 6 to 10 resets (see Figure 1). Although individual resets may have been widely scattered, allowing this many passes through the sequence renders stability across trials to the average point of reset calculation. In contrast the 2-minute trial in the timed method permits exact trial in the timed method permits exactly two passes. Based on fewer passes, the individual FREQs are less likely to be anchored near each other. Both methods offer acceptable within-session variability; with five replications per point, standard errors of the mean would be about one third of the magnitude of the shown confidence limits.

Second, both threshold estimates appear to be higher in Week 6 than in Week 1. Day to day fluctuation seems acceptable but the long-term change is substantial. An inspection of baseline data of the experiments performed during the intervening weeks revealed that the climb was gradual. Thus, both paradigms might be trusted in within-subject designs provided that control conditions interpose experimental ones. The comparisons of experimental results to control tests performed days earlier, such as in assessment of irreversible lesions, would not be recommended; between-subject designs would be preferred in evaluating permanent or long-lasting challenges.

Comparison of reset and required frequencies
Figure 2: Comparison of reset frequencies (FRST; autotitration; filled circles) and required frequencies (FREQ; timed method; open circles). Each symbol represents the geommed method; open circles). Each symbol represents the geometric mean and 0.95 confidence interval of five threshold estimates taken daily over the five days of Week 1 and Week 6. With rare exception (Rat E-1, for example) autotitration resets are made at frequencies higher than those where rats quit responding when tested in the timed method.

The third feature is that autotitration gave FRSTs that were consistently higher than the timed method’s FREQs (binomial test, P = Q = 0.5, p < 0.001). Thus these rats reset at stimulation values higher than those values where they quit responding, a finding that replicates Stein’s (1961) observation that rats tested in autotitration do not reset upon arriving at a nonreinforcing stimulation. Instead, they reset at values that would be revealed to be suprathreshold by a more conventional paradigm.

Experiment 2

Methods of threshold estimation should yield stable measures under a variety of manipulations assuming, of course, that the manipulations do not genuinely alter the "true" threshold. One such manipulation we investigated was changing the strength and number of suprathreshold values in the descending progression before the usual point of resetting or quitting. The gradient of descent was psychophysically maintained; each new frequency was still 0.1 log unitly maintained; each new frequency was still 0.1 log unit less than the prior one. However, the frequency to which the rat or timer reset—as was the entire sequence—was double the customary one.


The five rats were given four sessions daily on the 5 days of Week 2. Two of the sessions were control runs, one in each paradigm, that were identical to those described in Experiment 1. That is, we used the sequence 100, 80, 63, 50 . . .Hz. The test sessions contained three additional steps inserted at the beginning thus: 200, 160, 125, 100, 80, 63 . . .Hz. The four combinations of paradigm (autotitration or timed) and condition (double or control) were randomly ordered with the provision that no order was repeated for any particular rat.

Results and Discussion

The four daily sessions generated five estimates each; these measures were averaged and the effect of the double-frequency progression was expressed by the ratio of the test-data geometric mean to control-data geometric mean. Each rat provided two such ratios daily, one from autotitration and the other from the timed method. The geometric means and 95% confidence intervals averaged over the five animals are depicted in Figure 3.

Doubling the frequencies on the sequence had little or no effect on FREQs of the timed method as indicated by the inclusion of the control (1.0) level by the 0.95 confidence intervals. In cvel by the 0.95 confidence intervals. In contrast, the FRSTs of autotitration were strongly affected. Rats reset at frequencies 30% greater, a little more than one full step in our sequence, when the progression was doubled.

Effect of double starting frequency
Figure 3: Effect of double frequency. Geometric means and 0.95 confidence intervals averaged across subjects (n = 5) and replications (r = 5). Data express the effect of starting rats at 200 Hz pulse frequency as percentage of threshold estimate obtained with the usual 100 Hz initial frequency. Both timed and autotitration tests were held on Days 1 to 5 (abscissa) of Week 2; they are shown separately for clarity. Double-frequency starts had little or no effect on threshold estimates made by timed method but the autotitration measures were higher under this condition.

Such an effect on autotitration was not seen, however, by Schaefer and Holtzman (1979) who examined the effect of adding or subtracting 12 to 36 mA to their starting currents which ranged from 103 to 176 mA. The present challenge was relatively greater; that this experiment doubledpresent challenge was relatively greater; that this experiment doubled initial frequencies while the Schaefer and Holtzman starting currents were altered by about 20% may account for the difference. Furthermore, Schaefer and Holtzman (1979) employed pulse durations of 2 msec. From strength-duration data on lateral hypothalamic self-stimulation (Matthews, 1977), 2 msec pulses are roughly 9 times more effective than 0.1 msec pulses, and we suspect that equivalent currents (1 mA for 0.1 msec pulses) probably excite the entire bundle of reward-related neurons; increases in current beyond 1 mA are poorly compensated by reductions in frequency, suggesting that a ceiling of pulse effectiveness is reached by currents this high or their equivalents. Thus, whereas Schaefer and Holtzman (1979) did increase the strength of their pulses, the increase in current might not have sufficiently altered the rewarding value of the initial stimulation. Their conclusion that rats in autotitration are not simply resetting after emitting a constant number of responses is undoubtedly correct; the point made here, however, is that our observed changes in autotitration thresholds may be reconciled with the stability seen by Schaefer and Holtzman (1979) by considering whether the change in stimulation effectively altered the rewarding value of the initial frequency or current.

There are two possible accounts for the shifts seen in autotitration. One is that doubling the frequencies engendered t doubling the frequencies engendered a genuine but subtle contrast effect that was detected only by autotitration. We believe this explanation is incorrect for the following three reasons. First, the manipulation of adding three levels of suprathreshold stimulation at the top of the sequence is unlikely to give rise to a real contrast effect. Contrast might have been produced if the gradient of the progression were steeper, but in this experiment the gradient had been strictly defended. Second, contrast effects are not seen when a conventional paradigm is used to document refractory periods and conduction times for the fibers supporting self-stimulation (Bielajew & Shizgal, 1982). When rats are tested with each stimulation value available for a full minute, the insertion of additional frequencies at the beginning of a determination does not influence the frequency threshold (C. Bielajew, personal communication). Third, if the contrast effect had been genuine, then the timed method would have detected it as well. The timed method is sensitive to pharmacological treatments that affect self-stimulation threshold such as injection of d-amphetamine (see Experiment 4 of this report) and the neuroleptic alpha-flupenthixol (Corbett, Stellar, Stinus, Kelley, & Fouriezos, 1983). Similarly, it readily detects changes in orthogonal stimulation variables such as current intensity and train duration by showing compensatory shifts in Fon by showing compensatory shifts in FREQs (unpublished observations). If a genuine contrast effect had been there to be detected, then both methods should have revealed it.

The alternative explanation for the autotitration shifts is to propose that an immediate history of higher than usual stimulation causes rats to reset earlier than they do with the customary sequence. The results of Experiment 1 showed that autotitration resets occurred at frequencies that were still rewarding. This experiment shows that the point of reset thence rises if higher level stimulation is made available after the reset response. These data lead to the notion that the decision of when to reset might be based upon a relative judgment. Instead of assessing the value of currently earned stimulation for its own merit, rats in autotitration might be comparing the delivered stimulation to that offered immediately after the reset response. For example, 63 Hz may be worth earning when resets restore the stimulation to 100 Hz; 63 Hz might not, however, compare as favorably when the rat can reset the stimulation to 200 Hz. The possibility that this sort of relative assessment occurs in the timed method is precluded because there the rats have no control over resets. The sole decision confronting them is whether to take the offered stimulation. The possibility that autotitration methods suffer this contaminant was tested directly in the following experiested directly in the following experiment.

Experiment 3

The argument to this point is that the reset response is not guided by an absolute decision (Is this stimulation worth earning?) but rather is based on a relative judgment (Is it worth working for given what’s available immediately after a reset?). If a relative judgment does indeed guide resets, then it should be possible to change the point of resetting by altering the standard used for comparison to the present stimulation. The autotitration points of reset did rise, in Experiment 2, in reaction to adding three steps to the top of the sequence, a manipulation which effectively would have doubled the standard if the standard was the first frequency made available after the reset. In addition to altering the initial frequency, however, the manipulation of Experiment 2 suffers the confound of increasing the distance between the initial stimulus and the usual point of reset. Experiment 3 removed the confound; here, the customary sequence was employed except that only the first frequency, normally 100 Hz, was replaced by others. The remainder of the sequence (80, 63, 50, 40 . . .Hz) was left intact.

The prediction was that if relative assessments guided resetting responses and if the standard used was the first value following the reset, then the point of reset would rise when the 100 Hz starting frequency was replaced the 100 Hz starting frequency was replaced by higher initial values (125 or 160 Hz). Moreover, the power of the standard might work in both directions. If the initial frequency was reduced to 80 or 63 Hz (and followed by the normal remainder of the sequence), then the point of reset would drop because the standard would now be lower than normal. In other words, the prediction was that there would be a positive correlation between the initial frequency and the point at which the rats in autotitration reset.


Each rat received four sessions daily for 4 days in Week 3 following the protocol established in Experiments 1 and 2. A test session and control session were held for each paradigm, and these control sessions were identical to the control tests described above. The test sessions featured substitution of the usual starting frequency of 100 Hz with one of these initial frequencies—160, 125, 80, and 63 Hz. The paradigm and condition were randomized with respect to within-day order of presentation, and a unique order of substitutes across days was administered to each rat.

Results and Discussion

Figure 4 shows the effect of altering only the initial frequency on FREQs of the timed method and on FRSTs of autotitration. The method outlined in Experiment 2 of calculating and depicting these data was used. Rats tested in the timed method showed no co tested in the timed method showed no consistent effect of altered initial frequencies. They quit at about the same frequencies regardless of the value of the top frequency. These same rats, however, demonstrated resetting in the autotitration paradigm that closely followed changes in the initial frequency. Not only did they reset above their usual points when offered initial frequencies above 100 Hz, but they also reduced their points of reset when the initial frequency was dropped below the usual value. The correlation between initial frequency and FRST was 0.98. These results provide compelling evidence that rats tested in autotitration make comparative assessments of the stimulation’s value relative to the stimulation available immediately after the reset.

Changes in initial frequency
Figure 4: Changes in initial frequency. Geometric means and 0.95 confidence intervals of five rats’ threshold-estimate averages which, in turn, were based on five determinations. These data are expressed as percentage of control (100 Hz) starting frequency. Abscissa indicates the frequency that replaced the customary 100 Hz and the insets just above the abscissa reflect the progression of pulse frequency. (These insets are not to be read sion of pulse frequency. (These insets are not to be read against this ordinate.) The lines represent best fitting linear regressions. Changes in the starting frequency had no systematic effect on estimates of threshold using the timed method. In contrast, reset frequencies reliably tracked the initial frequency in autotitration tests.

Note that the data shown for the initial frequencies of 63 and 80 Hz in autotitration were based on four instead of five rats. One rat when confronted with the 63 Hz starting frequency ceased to reset altogether. The rat completed only two of the five trials; in both it advanced the frequency to 25 Hz, a frequency that was less than half its usual point of reset. It did complete all five trials when the sequence started at 80 Hz, but its behavior appeared disorganized here as well. On several occasions it quickly turned to the reset lever and without pressing it rapidly returned to the stimulation lever even though it had driven the stimulation well below values that it worked for in other tests. Despite the fact that the rat reliably self-stimulated in autotitration and in the timed method for frequencies of 80 and 63 Hz, its resetting behavior was completely disrupted when these values were used as initial frequencies in autotitration. Apart from providing adequate justification for the exclusion of this rat’s data from the 63 and 80 Hz conditions, this rat’s data from the 63 and 80 Hz conditions, this profound disruption of behavior produced by altering only the initial frequency in the sequence dramatically highlights the importance of that initial stimulus in controlling when the rat resets in the autotitration paradigm.

Experiment 4

Despite the problem that autotitration resets are based upon relative assessments, it may be argued that the paradigm nonetheless retains some usefulness. As long as FRSTs are documented with sequences that are never changed—with standards that remain fixed—then it does not matter that resets are guided by comparisons of the currently obtained stimulation to the reset value; reliable estimates of threshold will emerge. Consider, however, the following. Autotitration was originally developed for drug testing. Without evidence to the contrary, the influence of a drug such as cocaine or amphetamine is thought to affect all values of stimulation equally. For a very wide range of values, there is no reason to believe that the ability of a stimulant to enhance the per-pulse effectiveness differs as a function of pulse frequency. Consider a control run in the autotitration paradigm. As discussed above, frequency of the currently administered stimulation is compared to the initial frequency. At some point the obtained stimulation is deemed to compare poorly to the frequency available immediately after the frequency available immediately after reset and the reset response is made. Now consider a test with the rat injected with an agent that potentiates a downstream synaptic effect of the stimulation; a re-uptake blocker is a good example. Once again a comparison is made; this time it is the currently earned stimulation, with the effect potentiated by the drug, that is compared to the standard, with the effect also potentiated by the drug. If the effects of the standard and the currently earned stimulation have both been potentiated to the same degree, then the relative standing of the frequency where resetting normally occurred will not have changed. Thus, despite the genuine effect, autotitration may fail to detect it. In other words, autotitration may fail in the very domain, psychopharmacology, for which it was originally introduced.

Obviously, an extreme case is presented above. Such extremes are unlikely to occur when one considers the results of Experiment 2. There the entire sequence was doubled but the FRSTs rose by only 30%. The relevance of this result becomes clear if the double frequency is treated as an analogy to a drug experiment; if odd pulses are imagined to be the original ones and if even pulses are imagined to represent an agonist’s contribution, then it must be concluded that autotitration did indeed detect the contribution of the electronic agonist but did not measure it faithfully.

In this experiment we compared the ability of the two methods to detect the threshold-lowering effect of amphetamine. If the argument presented above is correct, that is if the ability of an agonist to potentiate all stimulation by roughly equal degrees poses a problem for autotitration, then the magnitude of amphetamine’s effect in autotitration should consistently be less than that measured by the timed method.


Rats received daily test sessions over the 5 days of Week 4 and Week 5. Each test lasted 33 minutes, not including a warm-up period of about 5 minutes. Three baseline determinations were made in the first 9 minutes, the rat was injected at the start of the next 9-minute period, and five threshold estimates were taken in the last 15 minutes of the session. Amphetamine (d-amphetamine sulphate; 0.5 mg/kg in saline) was injected intraperitoneally on Days 1, 3, and 5 and saline was injected on Days 2 and 4 of each week. Three rats were tested in autotitration on Week 4 followed by timed-method tests in Week 5 while the other two rats had the reverse order of tests.

Results and Discussion

Figure 5 shows in detail the results of these drug tests. (Figure 6 summarizes these details by grouping the data from all subjects without regard to the order of test paradigm.) The data from the three rats which received the autotitration asts which received the autotitration assessments first, followed by the timed evaluations in the next week, are shown in the left half of Figure 5 while the results from Rats E-2 and D-10, which experienced the opposite order, are depicted in the right half. The left member of each pair of linked symbols shows the preinjection estimate of threshold while the second member shows the postinjection result. Filled circles represent amphetamine tests and the open circles are saline tests. Except in the first test for rat E-1, the autotitration method failed to detect a threshold-lowering effect of amphetamine. This is seen clearly in the far right-hand panels for rats E-30 and E-36 and again in the far left panels for Rats E-2 and D-10. In marked contrast to the lack of effect witnessed by autotitration, the timed method revealed strong and consistent drops in threshold after amphetamine administration in the same animals. The results from E-1 are especially instructive; this rat showed a modest drop in reset threshold with the first drug test in autotitration, but the magnitude of the effect was reduced in the second test and altogether absent in the third. This pattern follows the dynamics of tolerance; but tolerance cannot explain the resumption of the drug effect when E-1 was subsequently tested with the timed method in the following week, nor can such an explanation account for the consistent lack of effect measured by autotitration regaffect measured by autotitration regardless of autotitration preceding or following the timed tests.

Effects of amphetamine on thresholds determined by the two methods
Figure 5: Detailed results in d-amphetamine tests. Rats in left half underwent drug tests first using autotitration on Week 4 and then with the timed method on Week 5. Rats E-2 and D-10 on right received opposite order of testing. Each pair of linked symbols shows, first, three preinjection thresholds and, second, five postinjection estimates. Filled symbols represent amphetamine tests (Days 1, 3, and 5) and open symbols depict saline injections (Days 2 and 4). Autotitration tests (left and right extremes) failed to measure any effect of d-amphetamine (0.5 mg/kg, intraperitoneally) in four of the cases but a diminishing effect was seen in Rat E-1. The timed-method tests (inner panels) revealed substantial and consistent drops in threshold after amphetamine injections.

Summary of amphetamine tests comparing the two methods
Figure 6: Summary of the amphetamine tests. The data shown in Figure 5 are collapsed and expressed as geometric means and 0.95 confidence intervals of preinjection baseline tests. Amphetamine (A) tests were administered on Days 1, 3, and 5 while saline (S) was injected on Days 2 and 4. The timed method measured a consistent decrease in required frequency produced by amphetamine injections whereas the autotitration technique appears to have missed the drug effect.


The results of these experiments suggest that rats tested with the autotitration method press the reset lever when the stimulation fails to meet some internal criterion relative to the stimulation available at the beginning of the sequence. This conclusion is consistent with, but not exclusively supported by, the results of Experiment 1 and 2 which demonstrated, first, that autotitration resets occurred at values of stimulation that were still reinforcing and, second, that rats reset at even higher stimulation values when the sequence began at twice its customary value. Better support for the notion was found in Experiment 3 wherein the first frequency of the series—the one that may be viewed as the most immediate consequence of a reset response—was found to strongate consequence of a reset response—was found to strongly influence resetting; there it was shown that reset thresholds rose and fell, faithfully following the first frequency. The fourth experiment provided the strongest support for a relative comparison guiding resetting by showing that autotitration greatly underestimated the magnitude of amphetamine’s threshold-lowering effect. In marked contrast the timed method, which is a simple modification of the autotitration protocol, demonstrated an acceptable stability throughout Experiments 1 to 3 and was sensitive to the amphetamine effect in Experiment 4.

Neill and his colleagues (Neill, Gaar, Clark, & Britt, 1982; Neill & Justice, 1981) have shown that the distribution of reset intensities collected in an autotitration paradigm can be shifted towards higher currents (i.e., rats reset earlier in the sequence) by removing a barrier that partially separates the stimulation and reset levers. They suggested that the decision to reset is influenced by the effort required to make the reset response. With the partition in place, the greater effort required to move to the reset lever biases staying at the stimulation lever; when the barrier is eliminated, the rats leave the stimulation lever earlier in the descending sequence of currents. Obviously, a method that produces shifts in estimates of threshold as a result of altered task requirements is of limited utility in assessing the influence of a dru in assessing the influence of a drug on central reinforcement mechanisms. It denies to the psychopharmacologist the possibility of distinguishing a selective effect on reinforcement from a general debilitation produced by the drug. Together, autotitration’s compliance to task demands and its underestimation of a drug effect seriously mitigate its recommendation for further use. The simple modification of transferring control of resets from the rat to the machine apparently rectifies these shortcomings without sacrificing the speed of data acquisition.


We thank Dr. Catherine Bielajew for critical reading of earlier drafts of this paper. This project was supported by Ontario Mental Health Foundation Grant 808 and Natural Sciences and Engineering Research Council of Canada Grant A7886 to the first author.


Bielajew, C., & Shizgal, P. (1982). Behaviorally derived measures of conduction velocity for rewarding medial forebrain bundle stimulation. Brain Research, 237, 107-119.

Corbett, D., Stellar, J. R., Stinus, L., Kelley, A., & Fouriezos, G. (1983). Time course of alpha-flupenthixol action explains "response artifacts" of neuroleptic action on brain stimulation reward. Science, 222, 1251-1252.

de Groot, J. (1959). The rat brain in stereotaxic coordinates. Amsterdam: N. V. Noord-Hollandsche Uitgevers.

Gallistel, C. R., Shizgal, P., & Yeomans, J. S. (1981). A portrait of the substrate for self-stimulation. Psychological Review, 88, 228-273.

Gardner, E. L. (1971). An improved technique for determining brain reward thresholds in primates. Behavior Research Methods and Instrumentation, 3, 273-274.

Leith, N. J. (1983). Effects of apomorphine on self-stimulation responding: Does the drug mimic the current? Brain Research, 277, 129-136.

Leith, N. J., & Barrett, R. J. (1980). Effects of chronic amphetamine or reserpine on self-stimulation responding: Animal model of depression? Psychopharmacology, 72, 9-15.

Leith, N. J., & Barrett, R. J. (1981). Self-stimulation and amphetamine: Tolerance to d and l isomers and cross tolerance to cocaine and methylphenidate. Psychopharmacology, 74, 23-28.

Liebman, J. M. (1983). Discriminating between reward and performance: A critical review of intracranial self-stimulation methodology. Neuroscience and Biobehavioral Reviews, 7, 45-72.

Mundl, W. J. (1980). A constant current stimulator. Physiology & Behavior, 24, 991-993.

Nazzaro, J., & Gardner, E. L. (1980). GABA antagonism lowers self-stimulation thresholds in the ventral tegmental area. Brain hresholds in the ventral tegmental area. Brain Research, 189, 279-283.

Nazzaro, J. M., Seeger, T. F., & Gardner, E. L. (1981). Morphine differentially affects ventral tegmental and substantia nigra brain reward thresholds. Pharmacology Biochemistry & Behavior, 14, 325-331.

Neill, D. B., Gaar, L. A., Clark, A. S., & Britt, M. D. (1982). "Rate-free" measures of self-stimulation and microinjections: Evidence toward a new concept of dopamine and reward. In B. G. Hoebel & D. Novin (Eds.), The neural basis of feeding and reward (pp. 289-297). Brunswick, ME: Haer Institute.

Neill, D. B., & Justice, J. B., Jr. (1981). An hypothesis for a behavioral function of dopaminergic transmission in nucleus accumbens. In R. B. Chronister & J. F. DeFrance (Eds.), The neurobiology of the nucleus accumbens (pp. 338-347). Brunswick, ME: Haer Institute.

Schaefer, G. J., Baumgardner, D. G., & Michael, R. P. (1979). Constant-current, biphasic titrating stimulator for brain self-stimulation. Physiology & Behavior, 22, 1217-1219.

Schaefer, G. J., & Holtzman, S. G. (1979). Free-operant and auto-titration brain self-stimulation procedures in the rat: A comparison of drug effects. Pharmacology Biochemistry & Behavior, 10, 127-135.

Schaefer, G. J., & Michael, R. P. (1980). Acute effects of neuroleptics on brain self-stimulation thresholds in ics on brain self-stimulation thresholds in rats. Psychopharmacology, 67, 9-15.

Seeger, T. F., & Gardner, E. L. (1979). Enhancement of self-stimulation behavior in rats and monkeys after chronic neuroleptic treatment: Evidence for mesolimbic supersensitivity. Brain Research, 175, 49-57.

Stein, L. (1961). Inhibitory effects of phenothiazine compounds on self-stimulation of the brain. Diseases of the Nervous System, 22(Suppl.), 1-5.

Stein, L. (1962). Effects and interactions of imipramine, chlorpromazine, reserpine, and amphetamine on self-stimulation: Possible neurophysiological basis of depression. In J. Wortis (Ed.), Recent advances in biological psychiatry (pp. 288-308). New York: Plenum Press.

Stein, L., & Ray, O. S. (1960). Brain stimulation reward "thresholds" self-determined in rat. Psychopharmacologia, 1, 251-256.

Valenstein, E. S. (1964). Problems of measurement and interpretation with reinforcing brain stimulation. Psychological Review, 71, 415-437.

Zarevics, P., & Setler, P. E. (1979). Simultaneous rate-independent and rate-dependent assessment of intracranial self-stimulation: Evidence for the direct involvement of dopamine in brain reinforcement mechanisms. Brain Research, 169, 499-512.

Zarevics, P., & Setler, P. E. (1981). Effects of GABAergic drugs on brain stimulation reward as assessed by a ‘ on brain stimulation reward as assessed by a ‘threshold’ method. Brain Research, 215, 201-209.

©1987 Springer-Verlag (printed version)
©2000-2009 Addiction Science Network (web-enhanced version)

ASNet Home

ASNet Profile

Illicit Drug Index

Virtual Lab Tour

Research Reports

Drug Classification

A Primer on Drug Addiction

Experimental Methods

Treatment Resources

Biological Basis of Addiction

Before Prohibition: Early Psychoactive Medicines

Links to Other Websites

Click here to enter the Addiction Science Network Discussion Forum

Brain Reward System©1999-2009 Addiction Science Network
This page was last revised 09 April 2009 08:46 EDT.
Send comments to:
Report technical problems to: