JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR

AN ANIMAL MODEL OF THE INTERPERSONALCOMMUNICATION OF INTEROCEPTIVE

(PRIVATE) STATESDAVID LUBINSKI AND TRAVIS THOMPSON

UNIVERSITY OF MINNESOTA

Pigeons were taught to interact communicatively (i.e., exchange discriminative stimuli) based on 1pigeon's internal state, which varied as a function of cocaine, pentobarbital, and saline administration.These performances generalized to untrained pharmacological agents (d-amphetamine and chlordi-azepoxide) and were observed in the absence of aversive stimulation, deprivation, and unconditionedreinforcement. The training procedure used in this study appears similar to that by which humanslearn to report on (tact) their internal environments and may be construed as a rudimentary animalmodel of the interpersonal communication of private events.Key words: private events, tacts, interanimal communication, emotion, pigeons

Roger Schnaitter (1978) has defined privateevents as "[t]hose phenomena of psychologicalinterest taking place 'inside the skin,' at a cov-ert level, observable beyond the first person byindirect means, if at all" (p. 1). The presentexperiment investigated the role of experi-mentally manipulated private events in settingthe occasion for communicative behavior be-tween nonhuman organisms. The aims of thisstudy were to determine whether nonhumananimals could learn to interact communica-tively (i.e., exchange discriminative stimuli witheach other), based on events in their internalmilieu, report on similar internal (private)events that had not been involved in training,and whether performance on these tasks couldbe observed in the absence of a primary es-tablishing operation (i.e., without aversivestimulation or deprivation) and unconditionedreinforcement.That it is possible to teach chimpanzees

(Fouts, 1973; Gardner & Gardner, 1971;Rumbaugh, 1977) and gorillas (Patterson,

This study was based on a dissertation submitted byDavid Lubinski to the University of Minnesota in partialfulfillment of the Doctor of Philosophy degree. The re-search was funded by two grants from the University ofMinnesota Graduate School, a dissertation fellowship anddissertation special grant, and by National Institute onDrug Abuse Training Grant 5RO1-DA02717. We areindebted to James Cleary for the patience and skill hedisplayed in providing technical support during this ex-periment. Reprints may be obtained from Travis Thomp-son, Department of Psychology, Elliott Hall, 75 E. RiverRoad, University of Minnesota, Minneapolis, Minnesota55455.

1978) to interact by exchanging discriminativestimuli that are arbitrarily matched with somefeature of their external environment is clear.Although there is disagreement as to whethersuch exchanges constitute truly linguistic ac-tivity (Brown, 1970; Premack & Woodruff,1978; Terrace, Petitto, Sanders, & Bever, 1979)there is little doubt that basic elements of sym-bolic representation of meaning (Gardner &Gardner, 1984) and rudiments of early humanchild semantics have been taught to nonhumanprimates (Gardner & Gardner, 1975; Savage-Rumbaugh, 1984). For example, Savage-Rumbaugh, Rumbaugh, and Boysen (1978)taught chimpanzees to use pictorial stimuli toreport to other chimpanzees the presence orabsence of the actual objects in a nearby room,to which only one animal had visual access.Moreover, limited elements of such interani-mal exchanges have also been taught to pigeons(Epstein, Lanza, & Skinner, 1980; Lubinski& MacCorquodale, 1984). The present studygrows out of the latter work with substitutionof private interoceptive stimuli for exterocep-tive stimuli as the stimulus source for exchangebetween animals.

This study is based on recent findings re-vealing that laboratory animals can be con-ditioned to emit tacts. A tact is defined as "averbal operant in which a response of a givenform is evoked (or at least strengthened) by aparticular object or event or property of anobject or event" (Skinner, 1957, pp. 81-82).As verbal operants, tacts are not maintainedby particular reinforcers, nor do they covary

1

1987,48, 1-15 NUMBER 1 (J}ULY)

DAVID LUBINSKI and TRAVIS THOMPSON

with the subject's state of deprivation or aver-sive stimulation; they are often maintained bygeneralized conditioned reinforcers (i.e., stimulithat were initially neutral but have accruedreinforcing efficacy through pairing with twoor more unconditioned reinforcers) (cf. Ca-tania, 1984; MacCorquodale, 1969; Segal,1977; Skinner, 1957; Winokur, 1976). Tactsare distinguished from mands (a more primi-tive class of verbal behavior) defined as "averbal operant in which the response is rein-forced by a characteristic consequence and istherefore under the functional control of rel-evant conditions of deprivation or aversivestimulation" (Skinner, 1957, pp. 35-36).

Verbal operants meeting the formal re-quirements of tacts have been conditioned inboth chimpanzees (Savage-Rumbaugh, 1984)and pigeons (Lubinski & MacCorquodale,1984). However, subjects in those studies weretrained to tact exteroceptive stimuli (geometricfigures, pictures, colors, letters, etc.). In thepresent research, nonhuman animals wereconditioned to tact interoceptive stimuli. Theirtacts were reinforced with a generalized con-ditioned reinforcer: a flashing light paired withboth food and water reinforcement (under bothreinforcer-relevant deprivation conditions andwhen satiated).

Skinner (1953, 1957, 1974) and others haveexplored the role private events may play inmodulating overt (verbal and nonverbal) be-havior (cf. Moore, 1980, 1984; Schnaitter,1978). In one of his more important papers,Skinner (1945) provided an interpretation ofthe manner in which humans learn to reporton private events (which are not accessible tothe verbal community for confirmation). How-ever, the systematic laboratory study of inter-oceptive stimuli regulating behavior has itsroots in early work by Pavlov and his col-leagues (Razran, 1961). Subsequently, severalinvestigators demonstrated that animals couldlearn to respond differentially to the internalstate produced by a psychoactive drug versusthe internal milieu associated with a vehicle(usually saline) injection (Schuster & Brady,1964; Thompson & Pickens, 1971).

In typical drug discrimination studies withnonhumans, a food-deprived animal is injectedwith a training drug (e.g., morphine) and isgiven the opportunity to press one of two le-vers, or peck one of two keys, leading to foodreinforcement (the drug cue lever), or a second

lever which produces no reinforcement (thevehicle lever). On intervening days the animalreceives a vehicle injection, in which the op-posite response is defined as correct, and presseson the drug lever are unreinforced. This pro-cedure leads to rapid learning to respond onlyto the drug-correlated lever on days the train-ing drug has been administered and on thealternate lever on days the vehicle has beenadministered. An array of psychoactive drugsproduces differential responding based on in-teroceptive stimulus conditions (Colpaert,1978; Griffiths, Roache, Ator, Lamb, & Lu-kas, 1985; Holtzman, 1985; Overton, 1977).Moreover, by using as the training compounda drug known to occupy a specific neurore-ceptor, these methods have permitted rapid andprecise identification of the types of neuro-chemical receptors occupied by a test drug,which has been confirmed by isolated tissueassays (Woods, Young, & Herling, 1982).

In the present study pigeons were trainedto differentially peck three keys in relation tosaline or one of two drugs, using a proceduresimilar to that described by France and Woods(1985). The birds were then taught to ex-change arbitrary discriminative stimuli cor-related with the agent that had been admin-istered on a given day. In a second phase,generalization to related internal drug states(and the associated exteroceptive discrimina-tive stimuli) was tested. Finally, communica-tive exchanges were evaluated without a pri-mary establishing operation and unconditionedreinforcement.

METHODSubjects

Subjects were 5 experimentally naive femalewhite Carneau pigeons (Columba livia) di-vided into two groups, referred to as "man-ders" (2 birds) and "tacters" (the remaining3). Although our subjects' behavior will notmeet all of the defining features of mands andtacts until later in the experiment, we will referto them as manders and tacters from the outset.All animals were housed in individual cageslocated in a constantly illuminated vivarium,with temperature maintained at 25 ± 10 C.Gravel was constantly available in the homecage, and food and water were available fol-lowing each session according to the schedulediscussed below.

PRIVATE EVENTS

ApparatusThe experimental apparatus was the same

as that used in the Lubinski and Mac-Corquodale (1984) investigation, but wasmodified in several ways. It consisted of twocontiguous experimental chambers, separatedby a 0.64-cm-thick transparent Plexiglas di-vider, each supplied with an individualizedstimulus presentation and response panel (seeFigure 1). The dimensions of each of the twocompartments were 34.3 by 31.6 by 35.6 cm.The aluminum back wall of each chamber con-tained several translucent discs (see Figure 1),which when pecked with a force of at least0.20 N broke the key contacts and was countedas a key peck. Keys were transilluminated withcolor lights (6 W, 110 vac) at indicated times(see below). Stimuli were presented (and allresponses were recorded) by means of electro-mechanical equipment located in a nearbyroom. Extraneous sounds were attenuated bya continuously operating fan, and solenoid-operated delivery mechanisms were used forreinforcement with food and water. Experi-mental sessions were conducted 7 days perweek. The manders were trained in the leftchamber, the tacters in the right. Each subject'splaying of one of two parts (i.e., mander ortacter, shown in Table 1) was first individuallyconditioned according to the following proce-dure.

General ProcedureTraining tacters. The tacters were initially

trained in a complex two component chain:The first component involved a three-key cross-modal matching-to-sample discrimination(composed of interoceptive sample stimulimatched to exteroceptive counterparts). Thesecond component comprised a two-key non-reversible option (Findley, 1962), whereby allpecks on the food and water keys were rein-forced under a continuous reinforcementschedule with their corresponding reinforcers.

In the first step of the tacters' training, thebirds, which were either 28 hr food deprivedand 4 hr water deprived or 28 hr water de-prived and 4 hr food deprived (these two con-ditions alternated in an A-B-A-B fashion), weretrained to peck the food and water keys. Peckson the water key made available 0.4 mL ofwater for 4 s, whereas pecks on the food keyprovided access to mixed grain for 4 s. These

responses were reinforced only when a bluelight was flashing (see Figure 1). This was toestablish the flashing blue light as a general-ized conditioned reinforcer (Lubinski &MacCorquodale, 1984; Savage-Rumbaugh,1984; Segal, 1977; Skinner, 1957). Approxi-mately 6 weeks from the first adaptation ses-sion were required for the flashing blue lightto consistently control subjects' pecking of thefood and water keys.The tacters were then trained to produce

the flashing light by pecking response keyslabeled with letters representing "Depres-sant," "Stimulant," and "No drug," after re-ceiving an intramuscular injection of either adepressant (pentobarbital 8 mg/kg), a stim-ulant (cocaine, 3 mg/kg) or isotonic saline(given in the same volume as the other twodrug solutions), respectively. Saline was alsoused as the vehicle for pentobarbital and co-caine delivery. Cocaine and pentobarbital werechosen as training drugs because of their well-documented discriminative and reinforcingproperties in laboratory animals (Griffiths etal., 1985; Schuster, Fischman, & Johanson,1981) and because of the likelihood for theirabuse in humans (Thompson & Unna, 1977).

Although unitary letters were projected onthe subjects' response keys representing thedrug and drug-class names employed in thisstudy (see Figure 1), we will use the names ofthe drugs and drug classes in describing theprocedure to make this section easier to follow.The specific names presented on the tacters'response keys were chosen for clarity in ex-perimental exposition, not in an attempt toimpart symbolic meaning.

Immediately following injection of one ofthe aforementioned agents, a tacter was placedinto the darkened experimental chamber. Af-ter a 20-min interval, an overhead light wasilluminated and, simultaneously, the three re-sponse keys were transilluminated. Pecksmatching the birds' interoceptive state (i.e.,saline injection = "No drug," pentobarbital ="Depressant," and cocaine = "Stimulant")were followed by presentation of the flashingblue light, and then a response on the food orwater key was reinforced with food or water.Five key pecks were required to produce theflashing light (FR 5); if either of the two otherkeys was pecked five times, the overhead lightand illumination behind the response keys wereterminated for 4 s, after which the houselights

3

DAVID LUBINSKI and TRA VIS THOMPSON

/Sample Key

Lenters

Cocaine Pentobarb

_ FoodDispenser

Tacters' Chamber

*.) FlashingBlue light

nt

No Drug

Response Stimulant

KeysI

Waterc Key

WaterLIDispenser

FoodKey

Foodd--JDiSpL

S= SalineC = CocaineP- PentobarbitalD - DepressantN-No drug

[. =Stimulant

Fig. 1. Adjoining keyboards for the two groups of birds were separated by a Plexiglas divider. The manderswere trained in the left chamber; the tacters were trained in the right chamber. The procedure is given in Table 1.

and response keys were reilluminated and anew trial started. Approximately 7 monthswere required for the tacters to respond reli-ably in this three-key discrimination procedureunder both food and water deprivation con-

ditions; across the final 40 sessions of this7-month training interval, the tacters all per-formed With an overall accuracy of 90%, andwith at least 80% accuracy in each of the six,2(Deprivation) x 3 (Drug), conditions (witha minimum of five observations in each con-

dition). These percentages are based on thefirst FR 5 emitted at the beginning of eachsession.

Manders' training. Four response keys con-fronted the manders on their control panel:three comparison keys labeled "Pentobarbi-tal," "Cocaine," and "Saline" and a samplekey, on which the drug-class names (i.e., De-pressant, Stimulant, and No drug) could beprojected (see Figure 1). The body weights of

both manders were maintained at 85% of theirfree-feeding values throughout the experi-ment. They were first trained (in a conditionaldiscrimination) to match the comparison keyslabeled "Pentobarbital," "Saline," and "Co-caine" to the drug-class names projected ontheir sample keys: "Depressant," "No drug,"and "Stimulant," respectively. Thus, they weretaught to match drug-class names (to specificdrug names) by pecking response keys con-

taining names of the specific drugs.Training consisted of quasirandomly pro-

jecting the names of the three drug classes (De-pressant, No drug, and Stimulant) onto themanders' sample key. Pecking the sample keywhen a drug-class name was projected onto itand then pecking the response key containingthe correct matching response (i.e., Depres-sant = Pentobarbital, Stimulant = Cocaine, andNo drug = Saline) was reinforced with 4-saccess to mixed grain. After the birds became

Manders' Chamber

ComparisonKeys

Saline

0 E_

E

o o

om cm

a

c

N.

o Eo CU.0

'Thank Youa.....(NKey

'How Do You Feel ?'Key . _

PRIVATE EVENTS

90 to 95% proficient at this task, two additionalrequirements were added: The manders couldproduce the drug class names on the samplekey only by (first) pecking an illuminated keylocated at floor level (Figure 1) labeled "Howdo you feel?" and (second) pecking a floor keylabeled "Thank you."The manding performance involved a three-

component chain: The terminal link was anarbitrary matching-to-sample task, whereasthe preceeding two components (first, peckingthe "How do you feel?" key, and second, peck-ing the "Thank you" key) were reinforced un-der a continuous-reinforcement schedule.

At this point during the training sequence,individual subjects in both groups had inde-pendently acquired the necessary componentrepertoire for an interanimal interaction basedon the tacters' interoceptive state. Individualtacters and manders were placed in theirchambers simultaneously and access to eachcomponent of the communicative exchange wasmade contingent on the other bird's behavior(see Table 1). (Before placing pairs of birdsinto the experimental apparatus to perform theinteraction, an adaptation manipulation wasconducted. Pairs of birds, composed of onemander and one tacter, were placed in theirrespective chambers concurrently, with onlythe houselights and the noise-attenuating fanoperating. Each adaptation session lasted 45min and all birds experienced at least four suchsessions.)

PHASE 1: THE INTERANIMALINTERACTION

ProcedureThe interaction sequence began when the

mander pecked its "How do you feel?" key.This response illuminated the drug-class nameson the tacter's three response keys (i.e., "De-pressant," "No drug," and "Stimulant"). Thetacter then pecked the response key corre-sponding to the drug it had received. This re-sponse illuminated the "Thank you" key inthe mander's chamber. When the manderpecked the "Thank you" key, two events en-sued concurrently: The tacter's blue light be-gan to flash and the drug-class name previ-ously pecked by the tacter appeared on themander's sample key. At this point the re-mainder of the response sequence of the 2 birdswere independent of each other. With the blue

Table 1

The complete verbal episode.

Component 1Mander: The mander's "How do you feel?" key is

illuminated and the mander pecks it, whichadvances the sequence to Component 2.

Tacter:Component 2

All three of the tacter's response keys are il-luminated simultaneously. It pecks the re-sponse key correlated with its state (i.e, co-caine state = Stimulant, pentobarbitalstate = Depressant, and saline state = Nodrug). If the bird pecks a noncorrespondingkey, the houselights in its chamber aredimmed for 4 s and the conditions of Com-ponent 1 are reinstated. Pecking the cor-responding key advances the sequence toComponent 3.

Component 3Mander: The "Thank You" key in the mander's cham-

ber is illuminated. The mander pecks the"Thank You" key, which advances the se-quence to Component 4.

Component 4Mander:

Tacter:

The drug name previously pecked by the tac-ter appears on the mander's sample key.

The blue light begins to flash in the tacter'schamber.

Component 5Mander: The mander matches the drug-class name

projected on its sample key (i.e., it pecksthe drug class) and the appropriate drug(Stimulant = Cocaine, Depressant = Pen-tobarbital, or No drug = Saline). This re-sponse is reinforced with mixed grain. Ifthe mander errs, the overhead houselightsin its chamber are dimmed for 4 s and theconditions of Component 1 are reinstated.At the end of this component (i.e., after themander has produced a reinforcer or fin-ishes a timeout), the conditions of Com-ponent 1 are reinstated, and the cycle con-tinues.

Tacter: The tacter receives either mixed grain or waterby pecking the appropriate key (after re-inforcement the blue light stops flashing).On satiated days, it receives only the flash-ing light; the blue light stops flashing whenthe conditions of Component 1 are rein-stated by the mander.

light flashing, the tacter could receive food orwater by pecking the appropriate key, and themander could receive food by correctly match-ing the specific drug (among its comparisonresponse keys) to the drug class (on its samplekey) (see Table 1).

Before each interaction, manders were given

5

DAVID LUBINSKI and TRA VIS THOMPSON

Table 2Percentage of correct responses for each bird and their correspondence on the first interanimalexchange across three phases. Phase 1 consists of four columns of data taken from the first trialacross 40 experimental sessions: The fractions under "Deprivation" represent the proportionof correct responses by the tacters, under each of the six, 2 (food and water deprivation) x 3(cocaine, pentobarbital, and saline), conditions. The denominator is the total number of sessionsfor each condition, and the numerator the total number of correct "tacts" of drug-inducedinternal states. Total % correct for all three conditions summed across both deprivation conditionsis in the next column. The third column represents the manders' accuracy at matching specific-drug discriminative stimuli from drug-class discriminative stimuli (taken from the first trialacross all three conditions on days in which the tacters first matching response was correct).Finally, the last column reflects the percentage of correct correspondence between the two birdsfor each drug state (i.e., the product of the manders' accuracy and the tacters' accuracy for thefirst trial across each condition). Phase 2: Combined performance of the three tacters and thetwo manders (taken from the first trial across 48 experimental sessions). Columns of fractionsand percentages have the same meaning as in Phase 1, but with respect to chlordiazepoxide,d-amphetamine, and saline. Phase 3: The columns in this table have the same meaning as thosedescribing Phase 2. An additional column labeled "Satiation" represents 18 sessions of tacters'performance when satiated (2 satiated sessions, for each of the three conditions, chlordiazepoxide,d-amphetamine, and saline, for all 3 birds).

Phase 1 Manders' Correspond-Deprivation Total % accuracy ence accuracy

Drug Food Water correct (%) (%)

Tacter 1Cocaine 3 mg/kg 5/5 4/5 90 89 80Pentobarbital 8 mg/kg 5/5 5/5 100 90 90Saline 9/10 9/10 90 94 85

Tacter 2

Cocaine 3 mg/kg 5/5 4/5 90 95 86Pentobarbital 8 mg/kg 5/5 5/5 100 100 100Saline 9/10 10/10 95 95 90

Tacter 3Cocaine 3 mg/kg 5/5 5/5 100 70 70Pentobarbital 12 mg/kg 5/5 5/5 100 80 80Saline 10/10 9/10 95 95 90

Phase 2

Tacters' deprivation (n = 3) Manders' Correspond-Total % accuracy ence accuracy

Drug Food Water correct (%) (%)

Amphetamine 2 mg/kg 5/6 6/6 92 91 84Chlordiazepoxide 8 mg/kg 6/6 6/6 100 92 92Saline 10/12 11/12 88 95 84

Phase 3

Correspond-Tacters' deprivation (n =Total % Manders' ence accu-

Drug Food Water Satiation correct (%) (%)

Amphetamine 2 mg/kg 2/3 2/3 5/6 75 95 71Chlordiazepoxide 8 mg/kg 3/3 3/3 6/6 100 92 92Saline 14/15 14/15 6/6 92 97 89

12 warm-up trials of matching drug-classnames to specific drug names before the tacterwas placed in the adjacent chamber. Eachmander performed in an equal number of in-

teractions. Although both manders worked withall of the tacters, one worked primarily withTacter 1, the other primarily with Tacter 3,and Tacter 2 worked 50% of the time with

PRIVATE EVENTS

.

owT, oZa

Conqpwson

ISw do you fe"r key

(A) M e peck howdo you eW" key.

(B) Tac per k o for drug dlescorspnig to its drug stle.

(C) Mander pecks "Thank You" key.

(D) Mee pes duas symbol on) sample key and pecpsiedfx rug

symbol; tacte presnted with fashingbiu k.h

(E) MmKderrewardedwkh food for corctilyco qlsUngttmsy*MbIxchange,and t1se rekvo=d wih wal.

Fig. 2. A 2-pigeon communicative exchange based on the internal state of one of the birds (mander, left; tacter,right). (A) The mander pecks its "How do you feel?" key. (B) The tacter pecks the drug-class letter corresponding toits internal state. (C) The mander pecks the "Thank You" key, which presents the flashing blue light to the tacter;this response also presents to the mander the drug-class letter previously pecked by the tacter. (D) The mander matchesthe drug-class letter (projected on its sample key) by pecking it and then pecking the letter representing the specificdrug that the tacter is currently experiencing; the tacter attends to the flashing blue light. (E) The mander is reinforcedwith food for correctly completing the communicative exchange; the tacter is reinforced with water.

each. For individuals in both groups, the birds'accuracies did not vary as a function of withwhom they were interacting. (Prior to the in-teranimal interaction, Tacter 3 began to showsome signs of tolerance to pentobarbital. To-ward the middle and end of the 40-trial period,it began pecking both the saline and pento-barbital keys and taking some timeouts; there-fore, it was decided to increase its dose of thisagent to 12 mg/kg. We are indebted to SheldonSparber for recommending this modification.After implementing this change, Tacter 3's ac-curacy remained high throughout the 40-trialsessions.)On alternating days the tacters were either

28 hr food deprived and 4 hr water deprived,or 28 hr water deprived and 4 hr food deprived.Administration of the drugs and of saline alonewas quasirandomized: Tacters were given freeaccess to both food and water for 12 hr fol-lowing a cocaine or pentobarbital session, andneither of these agents was given more thantwice in succession (discounting days with sa-

line alone) without the other administered once.On each experimental day, the three tactersreceived different injections. This measure wastaken to prevent the manders from discrimi-nating what drug condition was operating, be-cause on every experimental day, one of themanders would perform twice and the otheronly once (except on the days in which onlytwo tacters were run: on these days each man-der performed once). Each session was con-tinued until the tacter earned 40 reinforcers(either food or water). Finally, the first 2 daysof the interanimal exchange are not reportedin Table 2 (see below); both consisted of salineinjections (one under food deprivation, the otherunder water deprivation) and were viewed aswarm-up days (i.e., a time for the birds toadapt to working interactively).

Results and DiscussionAcross 40 experimental days subjects per-

formed this interaction with a high degree ofaccuracy; all three tacters performed at or above

7

DAVID LUBINSKI and TRAVIS THOMPSON

90% accuracy under all three interoceptivestates (see Table 2A). Random respondingwould generate an accuracy of approximately33% correct. Only the first exchange for eachsession is reported, because although the birdswere required to perform 40 interanimal ex-changes on each day to maintain performance,once the birds received food or water, the dis-criminative stimulus for correct responding wasno longer exclusively the tacter's drug-inducedinternal state. The accuracy of the tacters' per-formance on trials subsequent to the first rangedfrom 95% to 99% across all conditions; stan-dard deviations for these percentages rangedfrom 2.1% to 6.2% (see Appendix 1). The ac-curacy of the manders' performance on eachday's 12 warm-up trials did not differ system-atically from performance in the subsequenttrials in which the tacter was present.The overall correspondence between man-

ders and tacters (i.e., both birds performing acorrect discrimination on the first interanimalexchange) ranged from 70% to 100% accuracy.The probability of a correct correspondencehappening by chance is approximately 11%(the product of the two individual perfor-mances happening by chance, viz., .33 x .33 =.11). Approximately 15 min were required tocomplete 40 interanimal exchanges, across allthree drug conditions (see Figure 2).

Although this may represent a unique in-stance of communicative exchange (via arbi-trary matching tasks) between two nonhumanorganisms based on the interoceptive state ofone of the participants, such exchanges arefamiliar among humans. People often ask oneanother how they feel, and maintain the prob-ability of future reporting of internal states byexpressions of interest, concern, or enthusiasmabout the speaker's self-reported feeling. In thepresent experiment, the manders' key peck re-sponses requested information from the tactersbased on the latter's internal state: tacters rein-forced the manders' requests by tacting (i.e.,pecking) the drug-class name (on illuminatedresponse keys) that corresponded to their cur-rent state; the manders then reinforced thosetacts by emitting a response that produced theflashing blue light and, in turn, matched thetacters' report by pecking a second class ofarbitrary stimuli that corresponded to the spe-cific drugs.

Although the birds' accuracy remained highthroughout the 40 experimental days, some

features of their behavior gradually shiftedduring this phase. (Subjects were observed byclosed-circuit television during all sessions.)Initially, the birds spent nearly all of their timein the area immediately adjacent to the re-sponse keys, and then pecked appropriatelywhen illumination appeared. However, grad-ually, the birds' overall activities seemed tocome under the control of stimuli provided byeach other's behavior as well as of stimuli aris-ing from the response keys. After each birdcompleted a component link in the sequence,it typically oriented toward the adjacent cham-ber. For example, after consuming food orwater, the tacters approached the area adjacentto the manders"'How do you feel?" key. Whenthe "How do you feel?" key became illumi-nated, the tacters often pecked the transparentdivider directly above the key. The mandersthen approached and pecked the "How do youfeel?" key, and then turned toward the "Thankyou" key, standing in that position until thekey became illuminated following the tacters'response. If the tacters were slow in peckingthe drug-class key, the manders tended to peckthe transparent partition in a fashion similarto that described by Epstein et al. (1980) andLubinski and MacCorquodale (1984). Inpractice, however, birds in both groups usuallycompleted successive links within the inter-action quickly, so such partition-pecking wasnot typical; it did occur, however, if either birdpaused.

PHASE 2: GENERALIZATION TORELATED DRUGS

ProcedureThe next objective of this research was to

determine whether the discriminative perfor-mances established would generalize to similarbut somewhat different states (i.e., privateevents induced by pharmacological agents thatthe subjects had not experienced previously).Chlordiazepoxide (Librium*) and d-amphet-amine (Dexedrineg) are commonly used inour culture both therapeutically and recre-ationally (Miller et al., 1983; Szara & Lud-ford, 1981). Although d-amphetamine andchlordiazepoxide differ chemically and phar-macologically from cocaine and pentobarbital,they both share pharmacological propertieswith these agents (Gilman, Goodman, Gil-man, 1980). Moreover, it is well established

PRIVATE EVENTS

that in conventional preparations for two-choice drug versus saline discrimination, theygeneralize to cocaine (Fischman, 1984) andpentobarbital (Ator & Griffiths, 1983), re-spectively.The generalization test required a slight

procedural modification, however. Becausechlordiazepoxide is absorbed and distributedmore slowly than cocaine or pentobarbital, thepresession injection interval was increased to30 min. The first 2 days of the 30 min pre-session injection interval consisted of salineinjections (one under water deprivation, thesecond under food deprivation). Drug gener-alization tests were then conducted in the sameunpredictable fashion with d-amphetamineadministered at 2 mg/kg and chlordiazepoxideat 8 mg/kg intramuscularly in breast muscle.

Results and DiscussionOf 24 trials, six for each test agent under

food and water deprivation, only one erroroccurred (i.e., 95.8% correct). This observa-tions was recorded the second time the subject(Tacter 1) was exposed to d-amphetamine; ithad responded correctly on the first exposure(see Table 2B). The accuracy of the tacters'performance on trials subsequent to the firstranged from 92% to 99% across all conditions;the average standard deviation for these per-centages was 4.6% (see Appendix 2).The tacter pecked the drug-appropriate key,

reporting to its counterpart in the adjacentchamber that the interoceptive stimuli engen-dered by d-amphetamine were more similar toa cocaine state than to either saline- or pen-tobarbital-induced states; similarly, it reportedthat the interoceptive stimuli produced bychlordiazepoxide were more similar to a pen-tobarbital state than to a saline or cocaine state.The correspondence accuracy for this condi-tion ranged from 84% to 92% correct-chancewould be approximately 11%.The tacters' three-key interoceptive discrim-

ination generalized to pharmacological agentsthat are known to produce similar, but pre-sumably somewhat different, internal condi-tions (Ator & Griffiths, 1983; Fischman, 1984).The tacters' performance is similar to that re-ported by France and Woods (1985). Theytrained pigeons to peck three response keyscorresponding to the interoceptive stimuluscontrol of morphine, naltrexone, or salineadministration. After their performance was

firmly established, generalization tests wereconducted (using similar but nevertheless noveltest stimuli, viz., naloxone, ethylketazocine,buprenorphine, and pentazocine). The opiateantagonist naloxone generalized to the nal-trexone response key and the opiate agonistsethylketazocine, buprenorphine, and penta-zocine generalized to the morphine responsekey.

Further, the present generalizations, acrossdepressants and across stimulants, of discrim-inations between cocaine, pentobarbital, andsaline internal states are consistent with earlier(two-choice, drug versus saline) discrimina-tion/generalization laboratory animal studies(Ator & Griffiths, 1983; Brady & Griffiths,1977; Griffiths et al., 1985; Johanson & Schus-ter, 1977). Moreover, these data are in accordwith data from humans who were experiencedwith cocaine; these subjects did not report feel-ing differently when injected intravenouslywith 10 mg of amphetamine versus 16 mg ofcocaine (Fischman & Schuster, 1980). In ad-dition, humans treated with chlordiazepoxidereport experiential effects similar to those pro-duced by pentobarbital (Griffiths et al., 1985).

Although the present findings are consistentwith earlier work, the finding that a nonhu-man animal can learn to perform a discrimi-nation of three internal states maintained bytwo reinforcers, under two distinctive depri-vation conditions, is new. An additional uniquefeature of this performance is that it is con-tained within an interanimal communicativeexchange.

PHASE 3: SATIATION ANDDISCONTINUATION OF

PRIMARY REINFORCEMENTThe type of communicative behavior dis-

played by the tacters qualifies as tacts, that is,verbal responses established with multiplereinforcers and that do not specify a particularreinforcer. In humans, tact relationships areoften said to be maintained solely by gener-alized conditioned reinforcers (Lubinski &MacCorquodale, 1984; Savage-Rumbaugh,1984; Skinner, 1957). In the present experi-ment, the generalized conditioned reinforcerwas the tacters' flashing blue light. To deter-mine whether the tacters would continue toreport on their internal state when satiated*with food and water and without deprivation-relevant reinforcement (but with the flashing

9

DAVID LUBINSKI and TRAVIS THOMPSON

light functioning as a generalized conditionedreinforcer contingent on discriminative re-sponding), the following additional experi-mental probe was conducted.

ProcedureThe same procedure described in Phase 2

for d-amphetamine and chlordiazepoxide wasmaintained; however, on every third or fourthday, the tacters were placed in their experi-mental chamber after receiving 24 hr free ac-cess to both food and water. Their food andwater response keys were inoperative duringthis condition and, because of their satiatedcondition, they were only required to performfive interanimal exchanges rather than theusual 40.

ResultsWhen food and water satiated (Table 2C)

and without consumable rewards, but with theflashing light contingent on correct respond-ing, the tacters continued to respond correctlyto manders' requests by accurately reportingon their internal states 83% to 100% of thetime. Tacter 3 pecked a noncorresponding keyonly once while under amphetamine water de-privation, and Tacter 2 committed the othertwo errors for this condition. The mean ac-curacy of the tacters' performance on trialssubsequent to the first, for satiated sessions,was: amphetamine, 92%; chlordiazepoxide,90%; and saline, 90% (see Appendix 3). Theoverall accuracy of the interanimal correspon-dence for this condition ranged between 71%and 92% (chance performance would give anaccuracy measure of approximately 11 %).

ANCILLARY ASSESSMENTSAfter completing Phase 3 of this study, key-

pecking rate measures were obtained for allthree tacters under administration of each offour pharmacological agents as well as saline.These assessments were accomplished withoutparticipation of the manders, using the train-ing program that had been employed prior tothe interanimal interaction. Three sessions with30-min presession injection intervals (chlor-diazepoxide, saline, d-amphetamine) were fol-lowed by four sessions with 20-min presessioninjection intervals (saline, pentobarbital, sa-line, cocaine). All 7 test days were conductedunder food deprivation with the birds at 85%of their free-feeding weights. The amount of

time it took the birds to earn 40 reinforcers(from the first correct response to final deliveryof the food hopper) was as follows: 3 mg/kgcocaine (382 s), 8 to 12 mg/kg pentobarbital(391 s), 2 mg/kg d-amphetamine (431 s), 8mg/kg chlordiazepoxide (425 s), saline (424s). These times represent the interanimal meanfor each condition; the saline time was com-puted from data collected on all three salinedays (nine data points), and can be viewed asthe birds' base rate.

DISCUSSIONPigeons were conditioned to respond dis-

criminatively to three distinct interoceptivestates; they were also trained to communica-tively exchange arbitrary discriminative stim-uli as a function of these states. Moreover, thetacters' discriminative performances general-ized to similar interoceptive stimulus condi-tions and, ultimately, were observed in the ab-sence of a primary establishing operation andwhen unconditioned reinforcers were no longerdelivered. The pigeons' tendency to interact inthis manner was directly related to the ade-quacy of their experience and the strength oftheir repertoires for reporting on such privateevents.The tacters' behavior involved tacting pri-

vate events (via a matching-to-sample perfor-mance in which the sample stimuli were in-teroceptive). The birds were conditioned todiscriminate three internal states by peckinglettered response keys, and the discriminationwas reinforced with a generalized conditionedreinforcer (a stimulus paired with deliveriesof food and of water). Further, reinforcementof their discriminative behavior did not covarywith the particulars of their state of depriva-tion or aversive stimulation; their discrimi-native responses were reinforced under bothfood and water deprivation conditions andwhen satiated. The manders, on the other hand,were conditioned to emit impure mands (i.e.,verbal responses jointly controlled by a specificstate of deprivation (food) and maintained bythe deprivation-relevant reinforcer, and by dis-criminative stimuli provided by the tacters).One difference between the manner in which

our subjects learned to tact private events andSkinner's (1945, 1984) hypothesis of how hu-mans acquire this skill is the schedule of re-inforcement during conditioning. In the pres-

PRIVATE EVENTS

ent study, the private events experienced byour subjects (i.e., the drug states they wereconditioned to tact) were controlled with vir-tually 100% reliability and validity; our sub-jects' accurate reports on their interoceptivestates were reinforced on a continuous rein-forcement schedule (CRF). The manner inwhich humans typically learn to tact privateevents, according to Skinner, is much less pre-cise, and reinforcement of doing so is inter-mittent.The persistence of generalized discrimina-

tive responding maintained via the behavior ofanother organism suggests that the tendencyfor animals to report on similar, though some-what different, internal experiences can betaught. Technically, the verbal responses emit-ted by the tacters during this phase of theexperiment qualify as extended tacts. "Thereare several ways in which a novel stimulusmay resemble a stimulus previously presentwhen a response was reinforced, and hencethere are several types of what may be called'extended tacts'" (Skinner, 1957, p. 91). Tothe extent that these results adequately depictthe manner in which humans come to reporton related but novel feelings, they are consis-tent with Skinner's (1945, 1984) hypothesis ofhow humans acquire this skill: Skinner sug-gests that people are able to describe novelfeelings because these states often share fea-tures with familiar feelings on which they havelearned to report (i.e., extended tacts based oninteroceptive stimulus generalization).Most of the component behavioral units dis-

played by our subjects have been observed inearlier work but had not been synthesized inthis manner to build an animal model of hu-man communicative behavior. Catania (1983)has discussed the utility of such demonstra-tions: "Behavioral analysis begins with com-plex behavioral relations and breaks them downinto their components. One test of the ade-quacy of such an analysis is an experimentalsynthesis (e.g., Catania, 1972; Catania & Kel-ler, 1981).... Sometimes we begin with con-cepts from human affairs as the bases for pro-ducing novel behavioral relations. Thesynthesis consists of creating within the lab-oratory a performance in some respects anal-ogous to human behavior outside the labora-tory ... once a phenomenon has beendemonstrated by a behavioral synthesis, its de-fining properties and its range of applicability

can be refined by subsequent research. Thesuccess of a synthesis is then judged not onlyon the bases of the empirical results but alsoon the extent to which the refined understand-ing of the phenomenon has implications forthe human nonlaboratory situations fromwhich the analog emerged" (pp. 58-59). (Forrelated discussions, see Lubinski & Thomp-son, 1986; and Thompson & Lubinski, 1986.)Of the four classes of human behavior enu-merated by Epstein (1984) as candidates foranimal simulation research, only covert behav-ior (i.e., thoughts, feelings, and images) haveresisted empirical analysis using animal sim-ulations. The present study demonstrates thatthis class of behavior is also amenable to ob-jective analysis via simulation with nonhumansubjects.The present findings suggest that interocep-

tive stimuli and novel events at the neurore-ceptor level not only are discriminated but cancome to be reported to other organisms. Non-human organisms can be taught to make suchdiscriminations and report them to their neigh-bors, communicatively, through a learningprocess, and will continue to do so even withouta primary establishing operation and uncon-ditioned reinforcement. At the same time, it isunlikely that the performances studied hereconstitute linguistic activity as the term is usu-ally understood (Brown, 1970). There is noreason to suppose any syntactic structure isinherent in the pigeons' response sequencesnor any reason to impart complex intention-ality to the birds' communicative exchanges.Nonetheless, the character of the perfomancesshare features with those seen in very youngchildren or severely handicapped youth (e.g.,children with autism) who are just beginningto respond differentially to and report on theirown internal feelings (Lovaas, 1981). Theyrequire a social community to begin to learnhow to report on their internal states, but theirself-reports often seem rigid and restricted toa very narrow realm of available learning ex-periences.Savage-Rumbaugh (1984) has pointed out

differences between the behavior of the chim-panzees involved in her research and the pi-geons in the Epstein et al. (1980) study, whichshe believes cast doubt on the claim that thebirds in the latter investigation were actuallyengaged in communicative activity. The pi-geons, Savage-Rumbaugh (1984) argues, were

1 1

12 DA VID LUBINSKI and TRA VIS THOMPSON

not trained in a verbal environment, hencecould not properly be said to be engaged inverbal behavior. Moreover, she claims thatsince the contingencies brought to bear on thepigeons' behavior were imposed by electroniccircuitry rather than by another individual, theresult could not reasonably be characterized as"communication." The same concerns wouldpresumably be applicable in the present in-vestigation, because the birds were not trainedby people who communicated like pigeons orpigeons who communicated like people, norwere the exchanges of discriminative stimulimediated without the assistance of electroniccircuitry. The pigeons' experimental verbalcommunity (i.e., the other independentlytrained pigeons) was restricted to an extremelylimited range of behavior conforming to thefunctional form of a dyadic exchange of cues.Thus, the question is, "Given a conspecificaudience prepared to respond discriminativelyto a very limited range of verbal cues, woulda pigeon given the opportunity to report on itsinternal environment do so, without beinghungry or thirsty, or being fed or given waterfor so doing?" The answer seems to be "Yes."We believe the role of external circuitry in

mediating the exchange is less germane to thequestion at hand. There are countless exam-ples of human dyadic exchanges in which por-tions of the contingencies between partici-pants' stimuli and responses are mediated byexternal events between the speaker and thelistener over which they have no control, andwe are still content to refer to such exchangesas "verbal." These contingency manipulationsrange from poor telephone connections be-tween speaker and listener to messages left oncomputer bulletin boards that may be re-sponded to by a computer rather than by aperson. The essential question has to do withthe functional relations between speaker andlistener and the controlling variables, not howthe variables get implemented.

If we have correctly described key featuresof the way in which humans typically learn toreport on internal milieus, we may be in abetter position to begin to understand individ-ual differences in people's ability to adequatelyreport on internal events. In the present ex-periment, the pigeons' ability to report on theirinternal states depended on the adequacy oftheir experiences (specifically, discriminativetraining as a function of internal events ma-

nipulated pharmacologically). If the same holdsfor people, one might expect large individualdifferences in ability to report internal feelingsdepending on the adequacy of their discrimi-native learning histories with respect to inter-oceptive events, often acquired under the tu-telage of parents (or in some instances later inlife, e.g., via counseling or psychotherapy). Inaddition, if one assumes there are substantialconstitutional and/or genetically determinedindividual differences in the availability of typesof receptors upon which neurotransmitterscan act (with their correlated affective events),we would expect significant individual differ-ences in competence for reporting internal ex-periences.The notion that specific internal stimulus

events are critical components of emotion hasbeen widely held since William James (1890),and recent evidence and theoretical suggestionsconcerning the relation of the benzodiazepine-GABA receptor complex to human anxietylends credence to this idea (Gray, 1982; Po-shivalov, 1987). Evidence from the increas-ingly refined animal laboratory drug discrim-ination procedures reveals that animals canreport on events at the neuroreceptor level muchas organisms are able to respond differentiallyto the way rod and cone cells in the retina areactivated. This suggests that it may be possibleto understand more objectively the role of in-ternal affective states by combining the tech-nologies of the animal drug discriminationmodel with those of receptor chemistry. Fur-ther, the interorganism communication of cer-tain qualities of affective states can be assessedby coupling these methods with the domain ofinteranimal communication. As Skinner (1953)remarked over 30 years ago, "The line betweenpublic and private is not fixed. The boundaryshifts with every new discovery of a techniquefor making private events public.... Theproblem of privacy may, therefore, eventuallybe solved by technical advances" (p. 282).

REFERENCES

Ator, N. A., & Griffiths, R. R. (1983). Lorazepam andpentobarbital drug discrimination in baboons: Cross-drug generalization and interaction with Ro 15-1788.Journal ofPharmacology and Experimental Therapeutics,226, 776-782.

Brady, J. V., & Griffiths, R. R. (1977). Drug-main-tained performance procedures and the assessment ofdrug-abuse liability. In T. Thompson & K. R. Unna

PRIVATE EVENTS 13

(Eds.), Predicting dependence liability of stimulant anddepressant drugs (pp. 165-184). Baltimore: UniversityPark Press.

Brown, R. (1970). The first sentences of child and chim-panzee. In R. Brown, Psycholinguistics: Selected papers(pp. 208-231). New York: Free Press.

Catania, A. C. (1972). Concurrent performances: Syn-thesizing rate constancies by manipulating contingen-cies for a single response. Journal of the ExperimentalAnalysis of Behavior, 17, 139-145.

Catania, A. C. (1983). Behavior analysis and behaviorsynthesis in the extrapolation from animal to humanbehavior. In G. C. L. Davey (Ed.), Animal models ofhuman behavior (pp. 51-69). Chichester, England: Wi-ley.

Catania, A. C. (1984). Learning (2nd ed.). EnglewoodCliffs, NJ: Prentice-Hall.

Catania, A. C., & Keller, K. J. (1981). Contingency,contiguity, correlation, and the concept of causation.In P. Harzem & M. D. Zeiler (Eds.), Advances inanalysis of behaviour: Vol. 2. Predictability, correlation,andcontiguity (pp. 125-167). Chichester, England: Wi-ley.

Colpaert, F. C. (1978). Discriminative stimulus prop-erties of narcotic analgesic drugs. Pharmacology Bio-chemistry and Behavior, 9, 863-887.

Epstein, R. (1984). Simulation research in the analysisof behavior. Behaviorism, 12(2), 41-59.

Epstein, R., Lanza, R. P., & Skinner, B. F. (1980).Symbolic communication between two pigeons (Co-lumba livia domestica). Science, 207, 543-545.

Findley, J. D. (1962). An experimental outline forbuilding and exploring multi-operant behavior rep-ertoires. Journal ofthe Experimental Analysis ofBehavior,5, 113-166.

Fischman, M. W. (1984). Behavioral pharmacology ofcocaine in humans. In J. Grabowski (Ed.), Cocaine:Pharmacology, effects, and treatment of abuse (pp. 72-91). National Institute on Drug Abuse ResearchMonograph No. 50 (DHHS Publication No. ADM84-1326). Washington, DC: U.S. Government Print-ing Office.

Fischman, M. W., & Schuster, C. R. (1980). Experi-mental investigations of the actions of cocaine in hu-mans. In F. R. Jeri (Ed.), Cocaine: Proceedings of theinteramerican seminar on medical and sociological aspectsof coca and cocaine (pp. 62-75). Lima, Peru: PacificPress.

Fouts, R. S. (1973). Acquisition and testing of gesturalsigns in four young chimpanzees. Sciences, 180, 978-980.

France, C. P., & Woods, J. H. (1985). Opiate agonist-antagonist interactions: Application of a three-key drugdiscrimination procedure. Journal of Pharmacology andExperimental Therapeutics, 234, 81-89.

Gardner, B. T., & Gardner, R. A. (1971). Two-waycommunication with an infant chimpanzee. In A. M.Schrier & F. Stollnitz (Eds.), Behavior of nonhumanprimates (Vol. 4, pp. 117-184). New York: AcademicPress.

Gardner, B. T., & Gardner, R. A. (1975). Evidence forsentence constituents in the early utterances of childand chimpanzee. Journal of Experimental Psychology:General, 104, 244-267.

Gardner, R. A., & Gardner, B. T. (1984). A vocabulary

test for chimpanzees (Pan troglodytes). Journal of Com-parative Psychology, 98, 381-404.

Gilman, A. G., Goodman, L. S., & Gilman, A. (Eds.)(1980). Goodman and Gilman's The pharmacologicalbasis of therapeutics (6th ed.). New York: Macmillan.

Gray, J. A. (1982). Precis of The neuropsychology ofanxiety: An enquiry into the functions of the septo-hip-pocampal system. Behavioral and Brain Sciences, 5, 469-484.

Griffiths, R. R., Roache, J. D., Ator, N. A., Lamb, R. J.,& Lukas, S. E. (1985). Similarities in reinforcingand discriminative stimulus effects of diazepam, triaz-olam, and pentobarbital in animals and humans. In L.S. Seiden & R. L. Balster (Eds.), Neurology and neu-robiology: Vol. 13. Behavioral pharmacology: The currentstatus (pp. 419-432). New York: Liss.

Holtzman, S. G. (1985). Discriminative stimulus prop-erties of opioids that interact with mu, kappa and PCP/sigma receptors. In L. S. Seiden & R. L. Balster (Eds.),Neurology and neurobiology: Vol. 13. Behavioral phar-macology: The current status (pp. 131-147). New York:Liss.

James, W. (1890). The principles of psychology. NewYork: Holt.

Johanson, C. E., & Schuster, C. R. (1977). Proceduresfor predicting the preclinical assessment of abuse po-tential of psychotropic drugs in animals. In T. Thomp-son & K. R. Unna (Eds.), Predicting dependence liabilityof stimulant and depressant drugs (pp. 203-229). Bal-timore: University Park Press.

Lovaas, 0. I. (1981). Teaching developmentally disabledchildren. Baltimore: University Park Press.

Lubinski, D., & MacCorquodale, K. (1984). "Symboliccommunication" between two pigeons (Columba livia)without unconditioned reinforcement. Journal of Com-parative Psychology, 98, 372-380.

Lubinski, D., & Thompson, T. (1986). Functional unitsof human behavior and their integration: A disposi-tional analysis. In T. Thompson & M. D. Zeiler (Eds.),Analysis and integration of behavioral units (pp. 275-314). Hillsdale, NJ: Erlbaum.

MacCorquodale, K. (1969). B. F. Skinner's Verbal Be-havior: A retrospective appreciation. Journal of the Ex-perimental Analysis of Behavior, 12, 831-841.

Miller, J. D., Cisen, I. H., Gardner-Keaton, H., Hairell,A. V., Wirtz, P. W., Absdar, H. I., & Fishburne, P.M. (1983). National survey on drug abuse: Main find-ings, 1982. National Institute on Drug Abuse (DHHSPublication No. ADM 83-1263). Washington, DC:U.S. Government Printing Office.

Moore, J. (1980). On behaviorism and private events.Psychological Record, 30, 459-475.

Moore, J. (1984). On privacy, causes, and contingencies.Behavior Analyst, 7, 3-16.

Overton, D. A. (1977). Discriminable effects of anti-muscarinics: Dose response and substitution studies.Pharmacology Biochemistry and Behavior, 6, 659-666.

Patterson, F. G. (1978). The gestures of a gorilla: Lan-guage acquisition in another pongid. Brain and Lan-guage, 5, 72-97.

Poshivalov, V. P. (1987). Computerized ethologicalpharmacology: The new synthesis. In T. Thompson,P. B. Dews, & J. Barrett (Eds.), Advances in behavioralpharmacology: Vol. 6. Neurobehavioral pharmacology (pp.193-220). Hillsdale, NJ: Erlbaum.

Premack, D., & Woodruff, G. (1978). Does the chim-

14 DAVID LUBINSKI and TRA VIS THOMPSON

panzee have a theory of mind? Behavioral and BrainSciences, 1, 515-526.

Razran, G. (1961). The observable unconscious and theinferable conscious in current Soviet psychophysiology:Interoceptive conditioning, semantic conditioning, andthe orienting reflex. Psychological Review, 68, 81-147.

Rumbaugh, D. M. (1977). Language learning by a chim-panzee: The Lana project. New York: Academic Press.

Savage-Rumbaugh, E. S. (1984). Verbal behavior at aprocedural level in the chimpanzee. Journal of the Ex-perimental Analysis of Behavior, 41, 223-250.

Savage-Rumbaugh, E. S., Rumbaugh, D. M., & Boysen,S. (1978). Symbolic communication between twochimpanzees (Pan troglodytes). Science, 201, 641-644.

Schnaitter, R. (1978). Private causes. Behaviorism, 6, 1-12.

Schuster, C. R., & Brady, J. V. (1964). The discrimi-native control of operant behavior by interoceptivestimulation (trans. into Russian). Pavlov Journal ofHigher Nervous Activity, 14, 448-458.

Schuster, C. R., Fischman, M. W., & Johanson, C. E.(1981). Internal stimulus control and subjective effectsof drugs. In T. Thompson & C. E. Johanson (Eds.),Behavioral pharmacology ofhuman drug dependence (pp.116-129). National Institute on Drug Abuse ResearchMonograph No. 37 (DHHS Publication No. ADM81-1137). Washington, DC: U.S. Government Print-ing Office.

Segal, E. F. (1977). Toward a coherent psychology oflanguage. In W. K. Honig & J. E. R. Staddon (Eds.),Handbook of operant behavior (pp. 628-653). Engle-wood Cliffs, NJ: Prentice-Hall.

Skinner, B. F. (1945). The operational analysis of psy-chological terms. Psychological Review, 52, 270-277,291-294.

Skinner, B. F. (1953). Science and human behavior. NewYork: Macmillan.

Skinner, B. F. (1957). Verbal behavior. New York: Ap-pleton-Century-Crofts.

Skinner, B. F. (1974). About behaviorism. New York:Knopf.

Skinner, B. F. (1984). The operational analysis of psy-chological terms. Behavioral and Brain Sciences, 7, 547-553.

Szara, S. I., & Ludford, J. P. (1981). Benzodiazepines:A review of research results, 1980. National Institute onDrug Abuse Research Monograph No. 33 (DHHSPublication No. ADM 81-1052). Washington, DC:U.S. Government Printing Office.

Terrace, H. S., Petitto, L. A., Sanders, R. J., & Bever,T. G. (1979). Can an ape create a sentence? Science,206, 891-902.

Thompson, T., & Lubinski, D. (1986). Units of analysisand kinetic structure of behavioral repertoires. Journalof the Experimental Analysis of Behavior, 46, 219-242.

Thompson, T., & Pickens, R. (1971). Stimulus propertiesof drugs. New York: Appleton-Century-Crofts.

Thompson, T., & Unna, K. R. (Eds.). (1977). Predictingdependence liability of stimulant and depressant drugs.Baltimore: University Park Press.

Winokur, S. (1976). A primer of verbal behavior: An op-erant view. Englewood Cliffs, NJ: Prentice-Hall.

Woods, J. H., Young, A. M., & Herling, S. (1982).Classification of narcotics on the basis of their rein-forcing, discriminative, and antagonistic effects in rhe-sus monkeys. Federation Proceedings, 41, 221-227.

Received February 3, 1987Final acceptance April 21, 1987

APPENDIX 1Accuracy of the tacters' discriminative responding in trialssubsequent to the first matching response of Phase 1, withdeprivation conditions collapsed. The first column reportsthe mean % correct across all sessions, and the secondcolumn contains the standard deviation of this percentage.

Mean

cor-rect SD

Tacter 1Cocaine 3 mg/kg (n = 10) 99 2.3Pentobarbital 8 mg/kg (n = 10) 98 3.7Saline (n = 20) 98 2.6

Tacter 2Cocaine 3 mg/kg (n = 10) 95 5Pentobarbital 8 mg/kg (n = 10) 97 4.1Saline (n = 20) 97 5.9

Tacter 3Cocaine 3 mg/kg (n = 10) 99 2.1Pentobarbital 12 mg/kg (n = 10) 99 2.6Saline (n = 20) 98 6.2

APPENDIX 2Accuracy of the tacters' discriminative responding in trialssubsequent to the first matching response of Phase 2, withdeprivations conditions collapsed. The first column reportsthe mean % correct across all sessions, and the secondcolumn contains the standard deviation of this percentage.

Mean

cor-rect SD

Tacter 1Amphetamine 2 mg/kg (n = 4) 97 3.6Chlordiazepoxide 8 mg/kg (n = 4) 95 10Saline (n = 8) 99 4

Tacter 2Amphetamine 2 mg/kg (n = 4) 98 3.6Chlordiazepoxide 8 mg/kg (n = 4) 92 13Saline (n = 8) 97 5

Tacter 3Amphetamine 2 mg/kg (n = 4) 99 1.3Chlordiazepoxide 8 mg/kg (n = 4) 100 0Saline (n = 8) 99 1.1

APPENDIX 3Accuracy of the tacters' responses during the satiated ses-sions of Phase 3 subsequent to the first trial. Each tacterexperienced two sessions for each (amphetamine, chlor-diazepoxide, and saline) condition. The fractions representthe total number of correct responses divided by the totalnumber of matching responses; the percentages representthe tacter's accuracy for this performance.

Tacter 1Amphetamine2 mg/kg

Chlordiazepoxide8 mg/kg

Saline

Tacter 2Amphetamine2 mg/kg

Chlordiazepoxide8 mg/kg

Saline

Tacter 3Amphetamine2 mg/kg

Chlordiazepoxide8 mg/kg

Saline

4/4 = 100%, 4/4 = 100%

4/5 = 80%, 4/4 = 100%4/4 = 100%, 4/7 = 57%

4/6 = 67%, 5/5 = 100%

4/5 = 80%, 4/4 = 100%4/4 = 100%, 4/5 = 80%

4/4 = 100%, 4/5 = 80%

4/4 = 100%, 4/5 = 80%4/4 = 100%, 4/4 = 100%

PRIVATE EVENTS 15