JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR

REINFORCEMENT MAGNITUDE AND PAUSING ONPROGRESSIVE-RATIO SCHEDULES

ALAN BARON, JEFFREY MIKORSKI, AND MICHAEL SCHLUND

UNIVERSITY OF WISCONSIN-MILWAUKEE

Rats responded under progressive-ratio schedules for sweetened milk reinforcers; each session endedwhen responding ceased for 10 min. Experiment 1 varied the concentration of milk and the durationof postreinforcement timeouts. Postreinforcement pausing increased as a positively accelerated functionof the size of the ratio, and the rate of increase was reduced as a function of concentration and bytimeouts of 10 s or longer. Experiment 2 varied reinforcement magnitude within sessions (number ofdipper operations per reinforcer) in conjunction with stimuli correlated with the upcoming magnitude.In the absence of discriminative stimuli, pausing was longer following a large reinforcer than followinga small one. Pauses were reduced by a stimulus signaling a large upcoming reinforcer, particularlyat the highest ratios, and the animals tended to quit responding when the past reinforcer was largeand the stimulus signaled that the next one would be small. Results of both experiments revealedparallels between responding under progressive-ratio schedules and other schedules containing ratiocontingencies. Relationships between pausing and magnitude suggest that ratio pausing is under thejoint control of inhibitory properties of the past reinforcer and excitatory properties of stimuli correlatedwith the upcoming reinforcer, rather than under the exclusive control of either factor alone.

Key words: postreinforcement pause, reinforcer magnitude, progressive-ratio schedules, ratio sched-ules of reinforcement, timeout from reinforcement, inhibitory effects of reinforcement, lever press, rats

Although research on reinforcement sched-ules has declined over the years, some fun-damental questions remain unanswered(Zeiler, 1984). A case in point is the phenom-enon of postreinforcement pausing on fixed-ratio (FR) schedules and, in particular, therole played by the magnitude of the reinforcingstimulus (Bonem & Crossman, 1988). Oneview is that pausing is controlled by excitatoryproperties of the upcoming reinforcer (e.g.,Shull, 1979). Increases in the magnitude of thereinforcer should, therefore, reduce the dura-tion of pausing. Alternatively, some writershave suggested that pausing reflects inhibitoryaftereffects of the previous reinforcer (Harzem& Harzem, 1981). It follows that increases inthe magnitude of the reinforcer should increasepausing.

Experiments on the pause-magnitude re-lationship have not provided consistent evi-

Some of these data were reported at the 15th and 16thannual conventions of the Association for Behavior Anal-ysis (Milwaukee, Wisconsin, May 1989; Nashville, Ten-nessee, May 1990). We thank Cynthia Mansfield for herhelp in collecting the data and Michael Perone for hiscomments on an earlier version of the manuscript. Addresscorrespondence and reprint requests to Alan Baron, De-partment of Psychology, University of Wisconsin-Mil-waukee, Milwaukee, Wisconsin 53201.

dence for either view. Powell (1969), for ex-ample, found that pausing was inversely relatedto reinforcement magnitude, whereas Lowe,Davey, and Harzem's (1974) results were theopposite; pausing increased as the magnitudewas increased. To complicate matters further,Perone, Perone, and Baron (1987) could notdetect a relationship between pausing andmagnitude, once the minimal magnitudeneeded to maintain responding had beenreached.

In an effort to reconcile these contradictoryresults and interpretations, Perone et al. (1987)proposed that pausing might best be regardedas the product of competing excitatory andinhibitory influences rather than the exclusiveconsequence of either alone. The inverse pause-magnitude relationship in Powell's (1969) ex-periment resulted from a procedure in whichthe different magnitudes were correlated withdifferent colored lights. These discriminativestimuli provided a source of excitatory controlto counteract any inhibitory aftereffects of thelarger magnitudes. By comparison, the positiverelationship in Lowe et al.'s (1974) experi-ment came from a procedure in which differentmagnitudes were varied randomly in the ab-sence of correlated stimuli. Inhibitory factorscould predominate because the upcoming mag-nitudes were unpredictable.

377

1992, 58, 377-388 NUMBER 2 (SEPrEMBER)

ALAN BARON et al.

Finally, the steady-state design of Perone etal.'s (1987) experiment may have been re-sponsible for the failure to observe a pause-magnitude relationship (either positive or neg-ative). Procedures that expose the subject tothe same magnitude for many sessions workagainst dominance by either inhibitory or ex-citatory influences. Although inhibitory after-effects increase with increased magnitudes, ex-citatory effects increase as well, because theparticular magnitude is accurately predictedby stimuli correlated with the schedule itself.Another consequence of extended training isthat control by the schedule reduces control byunique features of the reinforcing event (Kling& Schrier, 1971; Morse & Kelleher, 1977).Perone et al. noted in this regard that exper-iments showing magnitude-pause relation-ships (either positive or negative) more oftenthan not have involved weak degrees of sched-ule control-for example, schedules producingratio strain (Hodos & Kalman, 1963) or sched-ules in which a series of changing magnitudesis contained within the same session (Lowe etal., 1974).The question of magnitude-pause relation-

ships was pursued in the present research inthe context of progressive-ratio (PR) schedulesof reinforcement. The special feature of thistype of schedule is that the response require-ment increases from ratio to ratio (e.g., fiveresponses for the first, 10 for the second, etc.);each session ends when the subject stops re-sponding (the so-called breaking point). Thesetwo considerations-increasing ratios andeventual extinction-make the PR schedule aninteresting vehicle for studying changes in ratiopausing as schedule control weakens. Theoriginal work by Hodos (1961) and Hodos andKalman (1963) showed that the breaking pointincreases as a function of the reinforcementmagnitude (in their study, the concentrationand volume of liquid food). But the picture ofmagnitude-pause relations is incomplete in thisand subsequent experiments (e.g., Keesey &Goldstein, 1968; Thomas, 1974) because theseanalyses relied almost exclusively on thebreaking-point measure-that is, the very lastpause of the session, which is the part of theschedule when schedule control is weakest.Quantitative data on pausing at earlier pointshave not been reported to our knowledge (somestudies have included cumulative records,however).

EXPERIMENT 1Pause data were collected on a ratio-by-ratio

basis, and two variables were manipulatedalong the lines of Perone et al.'s (1987) fixed-ratio experiment: the concentration of thesweetened milk that followed completion ofeach ratio and the duration of a subsequentperiod of timeout from responding. If pausingdepends on inhibitory properties of the rein-forcers, the onset of responding should be de-layed as a function of increasing magnitudes.Inhibitory effects should be attenuated, how-ever, by postreinforcement timeouts, becausesuch effects are expected to dissipate with thepassage of time. If, on the other hand, excit-atory properties of the reinforcers play thedominant role, pausing should decrease as afunction of the magnitude. Moreover, thetimeouts should not influence the inverse mag-nitude-pause relationship insofar as pausingis under the exclusive control of stimuli cor-related with the upcoming reinforcer.

METHODSubjects

Four male albino Sprague-Dawley derivedrats were approximately 10 months old at thestart of the experiment. They had previouslyserved in preliminary experiments with sched-ules and timeout procedures similar to thepresent ones. Each animal was housed in anindividual cage under continuous illuminationwith free access to water. Food deprivation wasaccomplished by scheduling 1-hr feeding pe-riods shortly after the experimental sessions(Hurwitz & Davis, 1983). Sessions were con-ducted 5 to 6 days per week.

ApparatusSingle-lever rodent chambers (Grason-

Stadler, E3125; 29 cm by 24 cm by 19 cm)were enclosed in sound-attenuating ventilatedchests. The lever, which required a minimumforce of 40 g (approximately 0.4 N) to operate,was centered on the front wall, 9.5 cm abovethe grid floor. Directly below was a cylindricalopening into which a 0.05 mL dipper couldbe raised. General illumination was providedby a 3-W lamp mounted behind a translucentwhite screen on the right wall. Extraneoussounds were masked by white noise and thesound of the ventilating fan. Programming andrecording equipment was in an adjacent room.

378

PAUSING ON PR SCHEDULES

ProcedureLever pressing was reinforced according to

a PR schedule with a step size of five. Eachsession began with delivery of a reinforcer uponcompletion of five responses. The second re-inforcer required 10 responses, the third 15,and so on. The schedule continued until 10min had elapsed without a response (thebreaking point). The reinforcer was a solutionof sweetened condensed milk in tap water withmilk concentrations of either 30%, 50%, or70% by volume. During the 3-s reinforcementperiod, the dipper containing the milk wasraised, a tone replaced the background whitenoise, and the lever was deactivated.

Timeouts of different durations separatedthe end of the reinforcement cycle and the re-activation of the lever for the next ratio. Dur-ing the timeouts, the chamber light and mask-ing noise (both normally on during the session)were turned off, and each lever press reset thetimer that controlled the duration; as a con-sequence, responding during the timeout pe-riods was rare. Five timeout durations werestudied: 0, 5, 10, 20, and 30 s. A given durationwas in effect for the entire session; the dura-tions changed from day to day in a haphazardorder with the restriction that each durationappeared once within each 5-day block.The experiment was conducted in three

phases each with a different milk concentra-tion. During the first phase, the concentrationwas set at 50%, the value used during prelim-inary training. After 40 sessions (i.e., eightsessions at each timeout duration), the con-centration was reduced to 30%, and 50 addi-tional sessions were conducted (the two ad-ditional blocks allowed adaptation to the newconcentration). During the final phase, 50 ses-sions were conducted with the 70% concentra-tion.

RESULTSPostreinforcement pauses and run times

were recorded to the nearest second. Pauseswere measured from the end of the timeoutthat followed the reinforcement cycle to thefirst response of the upcoming ratio. Run times,used to calculate running rates, were measuredfrom the first to the last response within a ratio.Analyses were based on the last eight blocksof each concentration condition. Performances,as indicated by daily breaking points, werestable when conditions were changed (in all

cases, the difference between the medians ofBlocks 1 through 4 and 5 through 8 was lessthan 20% of the median for Blocks 1 through8).

Pause data for each rat are summarized inFigure 1. The graphs are organized in termsof the three variables of the experiment: (a)increasing ratios within the schedule, (b) time-out durations, and (c) reinforcer concentra-tions. The values for each ratio are the mediansfor the eight sessions under each timeout-con-centration combination; to simplify the pre-sentation, pairs of ratios were combined (e.g.,FR 5 and FR 10). To accommodate the widerange of latency values (from 1 s to 600 s),performances are plotted on a logarithmic scale.The most general finding is that pausing

increased as a positively accelerated functionof ratio size (Figure 1). Latencies were brieffor a good part of the progression, mainly inthe range from 1 s to 10 s, and subsequentincreases to the final breaking point value wereabrupt, usually spanning no more than the lastfour or five ratios. Also apparent is that pauseduration and reinforcer concentration were in-versely related; that is, latencies increased whenthe concentration was reduced from 50% to30% and decreased when the concentration wasincreased to 70% (the final value of the ex-periment). The only deviation from this pat-tern was R16, whose durations at 50% and70% were about the same.The results also provided evidence that the

postreinforcement timeouts reduced pausing.This relationship may be seen most clearly inthe data from R14 and R15 by comparingperformance at the two extremes (0 s vs. 30s). In the absence of the timeouts (0-s condi-tion), their latencies increased in a more or lesslinear fashion until the final few ratios. Underthe timeout conditions, by comparison, the brieflatencies that characterized the start of theschedule were maintained further into the pro-gression, and the final acceleration to thebreaking point was more abrupt.

Figure 2 summarizes breaking points for thedifferent concentration and timeout conditions.The first panel for each animal shows the con-ventional breaking point measure: the highestratio attained before the session ended (the firstpause exceeding 10 min). The other two panelsshow results when more stringent pause cri-teria are applied: The highest ratio before apause exceeded 10 s (Panel 2) and the highest

379

ALAN BARON et al.

10001

el d~

10 !*=

1-~ ~ R1

1 00=

I~~~~~~~~~~~J

.1 I.~~~~~~~~~~~~

* 10 100 10 100 10 100 10 100 10 100 200

CONSECUTIVE RATIOS (STEP SIZE = 5)

Fig. 1. Experiment 1: Pause durations as a function of ratio size (x axis), timeout duration (the five groups offunctions: 0, 5, 10, 20, and 30 s), and reinforcer concentration (the functions designated by different dotted lines: light= 30%; intermediate = 50%; dark = 70%). Note that latencies are scaled logarithmically and that each set of functionsdepicts a new series of consecutive ratios.

0z00wC]

u1)z

w(ll

zUJ

z

w

m

w0cc0LLzw

CO00-

380

PAUSING ON PR SCHEDULES

0 5 10 20 30 0 5 10 20 30 0 5 10 20 30

TIMEOUT DURATION (SEC)

Fig. 2. Experiment 1: The point in the progressive-ratio schedule when the first pause of a specified durationoccurred (breaking points). Three criteria are shown: 10 min (the duration that ended the session; Panel 1), 10 s

(Panel 2), and 5 s (Panel 3). Values are plotted as a function of reinforcer concentration and timeout duration.

ratio before a pause exceeded 5 s (Panel 3).Taken together, the three criteria provide mea-sures of concentration and timeout effects un-der increasing levels of schedule control. Thus,the 10-min criterion identifies the point in theprogression when control was weakest-justbefore the response extinguished. The 5- and10-s criteria, by comparison, index perfor-mances when schedule control was stronger-the point of transition from maximal control(the initial limbs of the pause-ratio functionsin Figure 1) to progressively weaker control(the rapid acceleration of latencies to the finalbreaking point).

The results in Figure 2 provide evidencethat control by the schedule determined thenature of the interaction between concentra-tion and timeout. Under conditions of weakschedule control (Panel 1; the end of the sched-ule), performances depended exclusively on theconcentration; increased concentration pro-longed responding (delayed the final criterionpause) regardless of timeout duration. Underconditions of increased schedule control (Pan-els 2 and 3), concentration effects were more

likely to depend on timeout duration. For 3 ofthe 4 animals (R16 is the exception), the cri-terion pause was attained earlier when the

200

100

2000

4

.4zIL

z

200

z0

0 100la

200

100

R13 ___1050%

oz-.eDF ( -@j- -0~~~~1- z-zo---

R14_______ xr ====o----

-=-0< .-7 \ @@- --'0

R15 or.

R16

381

ALAN BARON et al.

timeouts were omitted (0 s) or after very brieftimeouts (5 s) than when the timeouts werelonger (10, 20, or 30 s).By comparison with the pause and break-

ing-point measures, running rates were not asconsistently influenced by the experimentalvariables. Figure 3 shows that rates declinedwithin sessions. In most cases, the decline wasgradual, and sustained responding continueduntil the very highest ratios of the schedule.Responding often was erratic at the end, how-ever (note that the session continued until 10min had elapsed without a response); becauseaverage rates might be misleading, rates of lessthan 20 responses per minute are not includedin the functions. There was some indicationthat the higher concentrations were associatedwith higher rates, most clearly in the case ofR14 for all of the timeout conditions and R16for the 30-s timeout. Finally, contrary to out-comes with the other measures, running rateswere more or less independent of the timeouts.

DISCUSSIONThe progressive-ratio schedule has not been

subjected to the same scrutiny as the much-studied fixed-ratio and variable-ratio (VR)schedules. Although all three schedules deliverreinforcers on the basis of response output,they differ in the way that ratios of differentvalues are integrated into the schedule. FR andVR schedules hold the value constant, usuallywithin entire sessions, and effects of ratios ofdifferent sizes must be studied through steady-state comparisons of different values. The pro-gressive schedule, by comparison, imposes thedifferent ratios in an ascending order withineach session, and a steady-state comparison ofperformance under the different ratios mustbe based on combining ratios from a series ofsuch sessions. Given this difference, it is ofinterest to determine whether the variables ofthe present study (ratio size, reinforcer mag-nitude, timeout from responding) have com-mon influences. A review of experiments onFR and VR reinforcement indicates that sim-ilarities are much more common than differ-ences. The characteristic finding has been thatpausing increased as a positively acceleratedfunction of ratio size. Previous investigators ofPR schedules did not attempt quantitativeanalyses; the published cumulative records,however, suggest a similar pattern (Hodos &Kalman, 1963; Thomas, 1974). Positively ac-celerated pause-ratio relationships also have

been found with FR andVR schedules (Blakely& Schlinger, 1988; Felton & Lyon, 1966; Prid-dle-Higson, Lowe, & Harzem, 1976; Schlin-ger, Blakely, & Kaczor, 1990). The rate ofincrease with these other schedules appears tobe logarithmic (e.g., Felton & Lyon, 1966),whereas acceleration within the present PRschedule was considerably greater. This dif-ference, most likely, is a consequence of theincreasing strain as the PR schedule advancesto larger and larger values within each session.Consistent with such an interpretation is thefinding that procedures that trained subjectsfor entire sessions at a given value before in-creasing the ratio produced intermediatepause-ratio functions (Powell, 1968).The further finding was that increases in

the magnitude of the reinforcers reduced paus-ing and attenuated the slopes of the pause-ratio functions. Studies with FR and VRschedules have reported similar results, thatis, steeper pause-ratio slopes with smallermagnitudes (Blakely & Schlinger, 1988; Pow-ell, 1969). Conclusions about the role of mag-nitude must be somewhat tentative insofar asonly one sequence of concentrations was ex-amined in the present study (50%, 30%, 70%).Also, note that not every experiment with ratioschedules has observed inverse pause-magni-tude relationships. Perone et al. (1987) sug-gested that negative results may be linked toprocedures that give subjects extended expo-sure to the ratios and concentrations understudy (steady-state designs). The orderly re-lationships of the present study are contraryto such an interpretation (the subjects had ex-tensive experience with the PR schedule andthe concentrations).The observed effects of the postreinforce-

ment timeouts on pausing have counterpartsin two experiments with FR schedules (Mazur& Hyslop, 1982; Perone et al., 1987). Bothcompared a single timeout duration (30 s) witha control condition (0 s) and found, as in thepresent study, that the timeouts attenuatedpausing. Results from intermediate durations(within the 30-s range) suggest that effectsmay be concentrated in the initial part of thetimeout period.

Finally, some parallels involving runningrates should be noted. Although response ratesdecreased as the ratio size increased, thechanges were not influenced consistently byeither the magnitude or timeout variables. Re-search with FR schedules also revealed that

382

PAUSING ON PR SCHEDULES

HDz

C')wcnz0(nCllw

zzzEl

400-R13

0-sec 5-sec 1 0-sec 20-sec 30-sec300-

'1'''~~ ' ' ' \ * 'i

O- ,.......................................................................400

R14300-

L LA

o- ,...................................................................................................

R15300-

,Ss''VP''';i'' "".,> l''S'\ "S"E < '; '' \\ st ts~~~~~~~~~~~~~3,'l.a}

100- , . . . . ..

0.~~~~~~~~~~~~~~~~~~~~~~~~~~~~~-Ri5~~~~~~~~~~~~~~~~~~~~~~~~~~~i

I~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~I

0 .....~~~~~~~~~'.' ..... ... .......... .... 1.

383

CONSECUTIVE RATIOS (STEP SIZE = 5)

Fig. 3. Experiment 1: Running rates as a function of ratio size (x axis), timeout duration (the five groups offunctions: 0, 5, 10, 20, and 30 s), and reinforcer concentration (the functions designated by different dotted lines: light= 30%; intermediate = 50%; dark = 70%). Note that latencies are scaled logarithmically and that each set of functionsdepicts a new series of consecutive ratios.

ALAN BARON et al.

running rates are inversely related to ratio sizeand are uninfluenced by timeouts (Mazur &Hyslop, 1982). The picture for VR schedulesis less clear, in that the rate-ratio relationshipis not as regular and there is indication thathigher magnitudes may lead to increased rates(Blakely & Schlinger, 1988; Schlinger et al.,1990).

EXPERIMENT 2The interpretation laid out by Perone et al.

(1987) was that the postreinforcement pausereflects the resolution of a competition betweenconditioned excitatory factors that reducepausing and unconditioned inhibitory factorsthat increase pausing. The results of Experi-ment 1 are consistent with such a view, in that(a) increased magnitude led to reduced pausing(the steady-state procedure allowed the pre-vailing magnitude to be signaled by excitatorycues originating in the schedule), (b) postre-inforcement timeouts reduced pausing (thetimeouts allowed inhibitory effects to dissi-pate), and (c) reinforcement-magnitude effectsincreased as the breaking point was ap-proached (features of the reinforcing stimulusassumed greater control as schedule controldecreased).

Experiment 2 pursued further tests of thecompetition account using a novel proceduredeveloped by Perone and Courtney (1992).Their method is designed to pit inhibitory andexcitatory factors against each other by pre-senting different magnitudes in an irregularorder within each session. Under the mixed-schedule condition, the upcoming magnitudesare unsignaled; therefore, only the inhibitoryinfluences of the past reinforcer can be ex-pressed. Under the multiple-schedule condi-tion, by comparison, the upcoming magnitudesare correlated with explicit discriminativestimuli, thereby allowing responding to be in-fluenced by excitatory factors. Performancesunder the multiple-schedule procedure shouldreveal the competition between inhibitory andexcitatory influences. Specifically, when a pre-vious small magnitude (low inhibition) is fol-lowed by a stimulus that predicts a large up-coming magnitude (high excitation), pausingshould be briefer than when a previous largemagnitude (high inhibition) is followed by astimulus predicting an upcoming small one(low excitation).

METHODThree additional rats were trained on a PR

schedule with a step size of two and an ex-tinction criterion of 5 min. The smaller stepsize increased the number of ratios in eachsession (cf. Hodos & Kalman, 1963), therebyfacilitating scheduling of a varied sequence ofsmall and large magnitudes, and the shorterextinction criterion avoided unnecessary pro-longation of the session. The apparatus wasthe same as for Experiment 1.The magnitude of the reinforcers was ma-

nipulated within sessions by varying the num-ber of presentations of the 30% milk concen-tration that followed completion of each ratio.Two magnitudes occurred in an irregular se-quence within each session: either one presen-tation of the dipper for 3 s (small) or threepresentations for 3 s each (large; this increasedthe total duration of the reinforcement periodto 10 s). The sequence was arranged to includeequal numbers of the four orders of past andupcoming reinforcers (i.e., small-small, small-large, large-small, and large-large).

Following preliminary training, observa-tions continued for 16 sessions (mixed condi-tion). For the subsequent 40 to 50 sessions,the upcoming magnitude was signaled by thecue lights to the left and right of the lever(multiple condition). For 2 animals, the leftlight blinked when the next reinforcer waslarge and the right light was on continuouslywhen the next reinforcer was small. Stimulus-magnitude correlations were reversed for the3rd subject.

RESULTSFigure 4 summarizes pausing during the

mixed and multiple phases of Experiment 2.The figure is organized to show latencies atthree points within the progression of ratios:at the start of the session, midway in the ses-sion, and at the end (i.e., the first four ratios,the middle two to five ratios, the number ofwhich depended on the total number of ratioscompleted, and the last four ratios). At eachof these junctures, performances are depictedas a function of the magnitude of the previousreinforcer and the magnitude of the upcomingreinforcer. The points in the figure are themedians of 32 latencies (middle = 20 to 32latencies) collected over the last 8 days of eachphase.

384

PAUSING ON PR SCHEDULES

SL SL SL SL SL SL SL SL SL SL SL SL SL SL SL SL SL SLPAST MAGNITUDE PAST MAGNITUDE PAST MAGNITUDE

Fig. 4. Experiment 2: Pause durations under the mixed and multiple conditions as a function of past and upcomingreinforcer magnitudes. Data are plotted separately for the mixed and multiple conditions. Open squares depict datafrom pauses at the beginning of ratios that ended with the smaller reinforcer; filled triangles depict those from ratiosthat ended with the larger reinforcer.

Results from the mixed condition generallywere similar to those from Experiment 1.Pausing increased as the schedule advanced,and the durations were longer following thelarge magnitude than the small one; this effectbecame more pronounced at the higher ratios.By comparison, effects of the upcoming mag-

nitude on pausing were unsystematic.Introduction of the discriminative stimuli

under the multiple condition produced a char-acteristic interaction that involved the se-

quence of reinforcers and the size of the ratios.At the beginning and middle of the schedule,pauses tended to follow the same pattern as

for the mixed condition: They were longer fol-lowing the large than the small reinforcer, andinfluences of the upcoming reinforcer eitherwere absent (RO1, R03) or relatively small(R04). The picture changed at the end of theschedule. Although pausing continued to bemore prolonged following a large than a smallreinforcer, this effect was considerably smallerwhen the upcoming reinforcer was large thanwhen it was small (compare the slopes of thelight and dark lines in Figure 4). Followingthe procedures of Experiment 1, the resultsare plotted on a logarithmic scale. Therefore,the interaction just described is considerablymore extreme when plotted arithmetically.A noteworthy feature of the results in Figure

4 is the difference between the mixed- andmultiple-schedule conditions. Although excep-tions may be seen, pauses under the multipleschedule tended to be shorter than under the

comparable mixed condition. This effect ispronounced for the data from the end of theschedule, particularly when a small reinforcerhadjust been received. Under the mixed sched-ule, pauses at this point averaged 10 s or more.

By comparison, multiple-schedule pauses werein the range of 1 or 2 s for Subjects R03 andR04 and did not exceed 5 or 6 s for RO1. Alsoapparent is that the upcoming reinforcer playeda minor role at best-pauses tended to be as

brief when the stimulus signaled that the nextreinforcer would be small as when the stimulusindicated that it would be large.

Information about running rates and break-ing points during the terminal eight sessionsis provided in Table 1. Running rates declinedas a function of the size of the ratio, but themagnitude sequence did not have systematicinfluences. The breaking-point data indicatethat although the multiple schedule producedshorter pauses, the mixed schedule tended tomaintain responding to higher ratios. The datain the right part of Table 1 show the magni-tude sequence that was in effect when the an-

imal quit responding (the values give the num-ber of sessions of each type; total = 8).Performances under the mixed condition wereirregular; if magnitude had any effect, it wasin favor of a weak tendency for the extinctioncriterion to be met following a large reinforcer(RO1, R03). Performances under the multiplecondition were more consistent. Along the linesof the pause data, the most frequent combi-nation when responding extinguished was one

C])0z0wC')

10

R01 R03 MULTIPLE R04 MULTIPLE

A

,, ^ s ] ] MIXED

A

A~~~~~~~~~~~~~~~~MIXED MULTIPLE MIXED

1

385

I UV:

ALAN BARON et al.

Table 1

Experiment 2: Median running rates (responses per minute). Values are from the last eightsessions of the mixed and multiple conditions at three points in the schedule (beginning, middle,end). Results are grouped in terms of past and upcoming magnitudes (small-small, large-small,small-large, large-large). Also shown are median breaking points and the number of sessionsthat each magnitude combination was in effect when the animal quit responding (total = 8).

Running rates Breaking Last ratio

Rat Schedule Ratios S-S L-S S-L L-L point S-S L-S S-L L-L

R01 Mixed Beginning 141 149 151 163Middle 106 108 121 111End 116 99 109 101 132 2 1 1 4

Multiple Beginning 150 150 160 155Middle 118 112 123 110End 87 85 87 93 119 1 7

R03 Mixed Beginning 212 210 214 224Middle 168 161 160 168End 125 126 130 131 141 3 3 2

Multiple Beginning 187 201 216 210Middle 167 171 177 179End 145 153 149 147 110 2 6

R04 Mixed Beginning 113 127 127 128Middle 80 81 84 85End 74 70 72 77 101 1 1 3 3

Multiple Beginning 121 133 156 145Middle 94 101 92 94End 87 93 89 97 92 8

in which a large reinforcer had been deliveredand the stimulus signaled that the upcomingreinforcer would be small.

DISCUSSIONResults from the highest ratios of the PR

schedule of Experiment 2 were similar to thosereported by Perone and Courtney (1992) inan experiment with pigeons rather than ratsunder FR rather than PR schedules. In bothstudies, pausing under the mixed condition de-pended exclusively on the previous reinforcer,and magnitude effects were inhibitory in thatpauses were longer following the larger mag-nitude. The multiple condition of both exper-iments provided evidence of the excitatoryproperties of the upcoming reinforcer. If thestimulus signaled that it would be large, thepause was considerably shorter than when thestimulus signaled that it would be small.A finding not reported by Perone and Court-

ney (1992) involves the differences between themixed and multiple schedules. The pausesduring multiple-schedule training not onlywere generally shorter but also tended to bethe same for the small-small and the small-large sequences. Shorter pauses in the latter

case are not unexpected-the multiple-sched-ule stimulus is the one correlated with a largeupcoming reinforcer. The finding of shorterpauses when the stimulus signals a small up-coming reinforcer is more difficult to explain.One possible account is based on the obser-

vation that the procedures included features ofchained schedules. Completion of each ratio inthe PR schedule not only produced the rein-forcer for that ratio (either large or small) butalso allowed access to the next ratio and itsreinforcer. The multiple-schedule stimuli,therefore, had the potential to serve as con-ditioned reinforcers that would reduce the ex-tent of pausing. Consider an animal that pausesin the presence of the stimulus correlated withthe small upcoming reinforcer. Although thediscriminative properties of the stimulus sup-port pausing, responding under these circum-stances bears a 50-50 chance of producing thestimulus correlated with the large reinforcer-a contingency that supports responding. Underthe mixed-schedule condition, by comparison,the absence of the correlated stimuli prohibitssuch an effect.Without data, the above account must be

largely speculative. One test would be to use

386

PAUSING ON PR SCHEDULES

procedures that vary the probabilities of theconcentrations within the schedule. A 25-75small-large distribution, for example, shouldenhance the reinforcing properties of the large-reinforcer stimulus and increase the differencebetween the mixed and multiple schedules. Bythe same token, a 75-25 distribution of largeand small reinforcers should reduce scheduledifferences.

GENERAL DISCUSSIONThe present results, along with those of Pe-

rone et al. (1987) and Perone and Courtney(1992), support an account of pause-magni-tude relations in terms of competing sourcesof control: control by the inhibitory aftereffectsof the past reinforcer that lead to direct rela-tionships and control by the excitatory prop-erties of the upcoming reinforcer that lead toinverse ones. Perone and Courtney explainedhow a competition account can be used to or-ganize the otherwise contradictory literatureon pause-magnitude relations; the interestedreader is referred to their cogent review forthe details.One matter that warrants further discussion

concerns the need for a concept of inhibitionin the analysis. This is certainly not a newissue. For example, Skinner (1938) in an earlydiscussion asked whether extinction can be ac-counted for without recourse to hypotheticalinhibitory processes. He reached the conclu-sion that weakening of performance is betterviewed as no more than the depletion of apreviously established reserve; others (e.g.,Rescorla, 1975) also have viewed the evidencefor an inhibitory process in extinction as lessthan compelling.

In discussing their experiment, Perone andCourtney (1992) also raised the possibility thatpause-magnitude relations might be accountedfor exclusively in excitatory terms. When mag-nitude is varied in the absence of explicit dis-criminative stimuli (the mixed condition of theirexperiment and the present Experiment 2; seealso Harzem & Harzem, 1981), the irregularsequence is intended to exclude stimuli cor-related with the upcoming magnitude. Butwhether this goal was attained may be ques-tioned on the grounds that the reinforcer mag-nitude itself can serve a discriminative functionwithin an irregular sequence of two magni-tudes. When a small reinforcer has been re-

ceived, the average magnitude of the next one,of necessity, must be larger (i.e., [small +large]/2). By the same token, average mag-nitude following a large reinforcer must besmaller. It readily follows that small reinforc-ers would control more prompt responding thanlarge ones because small reinforcers signal alocal increase in reinforcement.

It is apparent that a case can be made fordropping the inhibitory factor in favor of anaccount couched strictly in terms of excitatorycontrol by stimuli correlated with the upcom-ing reinforcer (Shull, 1979), that is, control bystimulus properties of the past reinforcer orcontrol by stimuli included within the sched-ule. The advantages of this approach (parsi-mony, decreased reliance on hypothetical pro-cesses) are balanced by some limitations.Although inferences about inhibition areavoided, assumptions about the subject's sen-sitivity to molar properties of the schedule (theaverage magnitude) introduce a different setof inferences. Moreover, it has become an ac-ceptable, if not desirable, practice since Skin-ner's writings to include an inhibitory com-ponent in the analysis of behavioral processes(e.g., inhibitory stimulus control).The timeout effects also may be problem-

atic. Mazur and Hyslop (1982) treated thetimeout procedure as a multiple FR extinctionschedule, in which case the shortened pausesin the FR component might reflect positivebehavioral contrast rather than dissipation ofinhibitory aftereffects. Behavioral contrast hasnot been the most reliable phenomenon in rats,however (Schwartz & Gamzu, 1977; but seeWilliams', 1983, subsequent review). Anotherdoubt concerns the fit of the pause-timeoutrelationships to what is known about contrast.Williams (1983) concluded that response ratesin the reinforced component of a multipleschedule increase as a direct function of theduration of the extinction component, whereasmost of the change in the present study oc-curred within the first 10 s of the timeoutperiod. On the other hand, the fit of the presentdata to what is known about inhibitory after-effects of reinforcers (Harzem & Harzem,1981) may not be much better.

Finally, the present results are a step towardclarification of the relation of PR schedules toother schedules containing ratio contingencies.As discussed above, the outcomes were re-markably comparable to those from experi-

387

388 ALAN BARON et al.

ments using fixed and variable ratios. Giventhe complex features of the PR schedule, it isnot surprising that a comprehensive accountof the variables controlling performance re-mains to be accomplished. Responding at eachof the increasing series of ratios is subject tocontrol by the preceding sequences, and theanimal's daily behavior runs the gamut fromstrong control by the ratio contingency at thestart to extinction at the end. Nevertheless, theorderliness of the present results recommendsthe PR schedule as an efficient way to ap-proach issues ordinarily studied with ratioschedules, particularly when concern is withinteractions between an experimental variableand variations in the size of the ratio.

REFERENCESBlakely, E., & Schlinger, H. (1988). Determinants of

pausing under variable-ratio schedules: Reinforcermagnitude, ratio size, and schedule configuration. Jour-nal of the Experimental Analysis of Behavior, 50, 65-73.

Bonem, M., & Crossman, E. K. (1988). Elucidating theeffects of reinforcement magnitude. Psychological Bul-letin, 104, 348-362.

Felton, M., & Lyon, D. 0. (1966). The post-reinforce-ment pause. Journal of the Experimental Analysis of Be-havior, 9, 131-134.

Harzem, P., & Harzem, A. L. (1981). Discrimination,inhibition, and simultaneous association of stimulusproperties: A theoretical analysis of reinforcement. InP. Harzem & M. D. Zeiler (Eds.), Advances in analysisof behavior: Vol. 2. Predictability, correlation, and conti-guity (pp. 81-124). New York: Wiley.

Hodos, W. (1961). Progressive ratio as a measure ofreward strength. Science, 34, 943-944.

Hodos, W., & Kalman, G. (1963). Effects of incrementsize and reinforcer volume on progressive ratio per-formance. Journal of the Experimental Analysis of Be-havior, 6, 387-392.

Hurwitz, H. M. B., & Davis, H. (1983). Depriving ratsof food: A reappraisal of two techniques. Journal of theExperimental Analysis of Behavior, 40, 211-213.

Keesey, R. E., & Goldstein, M. D. (1968). Use of pro-gressive fixed-ratio procedures in the assessment ofintracranial reinforcement. Journal of the ExperimentalAnalysis of Behavior, 11, 293-301.

Kling, J. W., & Schrier, A. M. (1971). Positive rein-forcement. In J. A. Kling & L. A. Riggs (Eds.), Ex-perimental psychology (3rd ed., pp. 615-702). New York:Holt, Rinehart and Winston.

Lowe, C. F., Davey, G. C. L., & Harzem, P. (1974).Effects of reinforcement magnitude on interval andratio schedules. Journal of the Experimental Analysis ofBehavior, 22, 553-560. /

Mazur, J. E., & Hyslop, M. E. (1982). Fixed-ratioperformance with and without a postreinforcementtimeout. Journal ofthe Experimental Analysis ofBehavior,38, 143-155.

Morse, W. H., & Kelleher, R. T. (1977). Determinantsof reinforcement and punishment. In W. K. Honig &J. E. R. Staddon (Eds.), Handbook of operant behavior(pp. 174-200). Englewood Cliffs, NJ: Prentice-Hall.

Perone, M., & Courtney, K. (1992). Fixed-ratio paus-ing: Joint effects of past reinforcer magnitude and stim-uli correlated with the upcoming magnitude. Journalof the Experimental Analysis of Behavior, 57, 33-46.

Perone, M., Perone, C. L., & Baron, A. (1987). Inhi-bition by reinforcement: Effects of reinforcer magni-tude and timeout on fixed-ratio pausing. The Psycho-logical Record, 37, 227-238.

Powell, R. W. (1968). The effect of small sequentialchanges in fixed-ratio size upon the post-reinforcementpause. Journal of the Experimental Analysis of Behavior,11, 589-593.

Powell, R. W. (1969). The effect of reinforcement mag-nitude upon responding under fixed-ratio schedules.Journal of the Experimental Analysis of Behavior, 12,605-608.

Priddle-Higson, P. J., Lowe, C. F., & Harzem, P. (1976).Aftereffects of reinforcement on variable-ratio sched-ules. Journal of the Experimental Analysis of Behavior,25, 347-354.

Rescorla, R. A. (1975). Pavlovian excitatory and inhib-itory conditioning. In W. K. Estes (Ed.), Handbook oflearning and cognitive processes (Vol. 2, pp. 7-35). NewYork: Erlbaum.

Schlinger, H., Blakely, E., & Kaczor, T. (1990). Pausingunder variable-ratio schedules: Interaction of rein-forcer magnitude, variable-ratio size, and lowest ratio.Journal of the Experimental Analysis of Behavior, 53,133-139.

Schwartz, B., & Gamzu, E. (1977). Pavlovian condi-tioning of operant behavior. In W. K. Honig & J. E.R. Staddon (Eds.), Handbook of operant behavior (pp.53-97). Englewood Cliffs, NJ: Prentice-Hall.

Shull, R. L. (1979). The post-reinforcement pause: Someimplications for the correlational law of effect. In P.Harzem & M. D. Zeiler (Eds.), Advances in analysisof behavior: Vol. 1. Predictability, correlation, and conti-guity (pp. 193-221). New York: Wiley.

Skinner, B. F. (1938). The behavior of organisms. NewYork: Appleton-Century.

Thomas, J. R. (1974). Changes in progressive-ratio per-formance under increased pressures of air. Journal ofthe Experimental Analysis of Behavior, 21, 553-562.

Williams, B. A. (1983). Another look at contrast inmultiple schedules. Journal ofthe Experimental Analysisof Behavior, 39, 345-384.

Zeiler, M. D. (1984). The sleeping giant: Reinforcementschedules. Journal of the Experimental Analysis of Be-havior, 42, 485-493.

Received September 10, 1991Final acceptance April 15, 1992