Journal of Comparative Psychology19S4,Vol 98, No 4.172-380

Copyright 1984 by theAmerican Psychological Association, Ipc

"Symbolic Communication" Between Two Pigeons(Columba livid) Without Unconditioned Reinforcement

David LubinskiUniversity of Minnesota

Kenneth MacCorquodaleUniversity of California, San Diego

This study addressed the question: Can nonhuman organisms be conditionedto respond discriminatively to abstract features of their environments withoutunconditioned reinforcement? It demonstrated that two pigeons can be con-ditioned to maintain an interaction in which one pigeon has access to adiscriminative stimulus that the other needs in order to emit a reinforcedresponse, and responds in the absence of deprivation, aversive stimulation,and unconditioned reinforcement. This finding adds to the class of variablesheretofore demonstrated to maintain "symbolic repertoires" and highlights afresh paradigm for conditioning other spontaneous interactions between andwithin nonhuman species. The importance of social stimuli in such inter-changes is demonstrated.

In the early 1970s, Cullen (1972) assertedthat "the term 'animal communication' hasoften been used to refer to the kinds ofsignals which pass to and fro between socialanimals and help to mould each others'behaviour towards some goal which is totheir mutual advantage" (p. 101). These"signals" are typically restricted to species-specific responses that communicate thepresence or location of an unconditionedreinforcer (e.g., food, an unconditioned pos-itive reinforcer; or a predator, an uncondi-tioned negative reinforcer).

Recently, the domain of interanimalcommunication has been expanded to showthat nonhuman organisms can learn to"communicate symbolically" or, more tech-nically, that they can be conditioned tosupply SDs for each other's behavior (socialcommunication) and that the SDs they sup-ply can be only arbitrarily matched withsome aspect of the environment (hence"symbolic," not iconic). Savage-Rumbaugh,Rumbaugh, and Boysen (1978) conditioned

We thank Thomas Kelly for his expert technicalassistance and his patience throughout this experi-ment. We are also indebted to Travis Thompson, forcommenting on an earlier version of this manuscript,and to Howard Rachlin and Evelyn Segal, for provid-ing us with several helpful suggestions throughout theeditorial process.

Requests for reprints should be sent to David Lu-binski, Department of Psychology, Elliott Hall, Uni-versity of Minnesota, 75 East River Road, Minneap-olis, Minnesota 55455.

primates to supply geometric symbols foranother subject's food-reinforced re-sponses, and Epstein, Lanza, and Skinner(1980) conditioned pigeons in a similarfashion to interact by pecking lettered re-sponse keys. In both of these experiments,the correct response to a symbol (letter orgeometric form) was reinforced with food.The question naturally arises: Can thesespecies maintain such interactions in theabsence of unconditioned reinforcement?In the present study, we investigated thispossibility by conditioning such "symboliccommunication" repertoires in nonhumanorganisms without deprivation, aversivestimulation, or unconditioned reinforce-ment.

The procedural and experimental phasesof this experiment were based on Skinner's(1957) formulation of verbal behavior. Ac-cording to Skinner, verbal behavior is be-havior that is reinforced through the me-diation of other organisms. Although Skin-ner identified several classes of verbal be-havior, only two are relevant to the presentinvestigation, mands and tacts. Verbal re-sponses that are controlled by motivationalstates are identified as mands: A mand isdefined as "a verbal operant in which theresponse is reinforced by a characteristicconsequence and is therefore under thefunctional control of relevant conditions ofdeprivation or aversive stimulation" (Skin-ner, 1957, pp. 35-36). Mands are distin-

372

"SYMBOLIC COMMUNICATION" 373

guished from verbal responses controlledby environmental stimuli, which are iden-tified as tacts. A tact is defined as "a verbaloperant in which a response of a given formis evoked (or at least strengthened) by aparticular object or event or property of anobject or event" (pp. 81-82). A "pure" tactshows no influence of any specific motiva-tional state. It is objective and does notcovary with the particulars of the speaker'sdeprivation or aversive stimulation, and itdoes not maintain a one-to-one relationwith a specific reinforcer.

This distinction indicates that the re-sponses conditioned in previous studies(Epstein et al., 1980; Savage-Rumbaugh etal., 1978) were not pure tacts; they werejointly controlled by SDs and deprivationfor the specific reinforcer involved. Ourinvestigation demonstrates that pure tactscan be conditioned in nonhuman organismsand can be maintained in interanimal in-teractions. One way to free the tact fromspecific motivational control is to maintainit by generalized conditioned reinforcers.Generalized conditioned reinforcers arestimuli that were initially neutral but haveacquired reinforcing efficacy through pair-ing with two or more unconditioned rein-forcers. By this process a generalized con-ditioned reinforcer's effectiveness becomesindependent of the behaver's current mo-tivational state (MacCorquodale, 1969; Se-gal, 1977).

Method

Subjects

Subjects were two experimentally naive male Bir-mingham rollers {Columba iwia).

Apparatus

The experimental apparatus is illustrated in Figure1. It consisted of two adjoining experimental cham-bers, separated by a Plexiglas divider; each chamberwas supplied with an individual panel on which thesubjects received discriminative stimuli for choice re-sponses (pecking) that in turn provided an SD on theadjoining subject's panel. For our generalized condi-tioned reinforcer, we used an intermittentenly flashing(one per second) light (notated S0" in Figure 1). Otherdetails of the two chambers differed in ways that aredescribed below.

Procedure

The procedures replicate parts of Epstein et al.'s(1980) procedure but with some significant variations.'We refer to our subjects as the "mander" and "tacter"from the outset, although they met the requirementsfor manding and tacting only later in the proceedings.Each subject was first independently conditioned toplay its role in the interaction presented in Table 1.The pigeon designated the tacter was conditioned inthe right chamber; the mander was conditioned in theleft chamber (see Figure 1). All responses emitted bythe mander technically qualify as impure mands,jointly controlled by deprivation and by the SD cur-rently present. The mander was maintained at 85% ofits free-feeding weight throughout this experiment,and all of its discriminative responses, if correct, werereinforced with food. The tacter, however, emittedboth mands and tacts Its initial responses in eachchain were tacts, discriminatively controlled by SDsand reinforced by an S0", the flashing light. Thiscontingency was maintained whether the bird washungry, thirsty, both, or neither. Thus control wasthrough the SD. The tacter's responses to the food orwater keys were primarily mands; they were reinforcedwith food or water, depending on which key waspecked. Control was, in a sense, through the reinforcer.

1 In contrasting our procedure with that of theEpstein et al. study, two additional variations arenoteworthy. First, the tacter's sample key was in plainview of the mander, rather than being recessed behinda curtain. Second, neither bird was required to lookthrough the Plexiglas to see the letters or colors onthe other bird's panel, because these were insteadprojected onto the keys on their own individual panels.These changes are of no consequence to the mainpurpose of the present study because our central con-cern was with the specific nature of the tacter's be-havior rather than the mander's. We wanted to knowwhether the tacter would continue to arbitrarily matchthe letters (i.e., "tact") when it was satiated and re-ceived no unconditioned reinforcers for matches. Weomitted the curtain because we suspected that it mightbe sufficiently aversive to suppress the tacter's match-ing behavior when the bird was satiated. In any event,the fact that the tacter's sample key was in plain viewof the mander did not enhance the mander's arbitrarymatching accuracy. We know this because in Condi-tions 2 and 3, when an opaque barrier covered thePlexiglas, preventing the mander from seeing the tac-ter's sample key, the mander's matching accuracy wasnot attenuated. Also, direct observation suggested thatalthough the mander did attend to changes in theillumination of the tacter's several sample and com-parison keys, it did not attend to the specific SD

features of these stimuli. In fact, at the end of thisexperiment we scrambled the SDs interchanged by thetacter and mander (in a quasi-random fashion), sothat on 66% of the trials the color that appeared onthe mander's sample key was not the one previouslymatched by the tacter. The mander accuratelymatched the sample on its own key, as before.

374 DAVID LUBINSKI AND KENNETH MACCORQUODALE

Mander's Chamber Tacter's Chamber

CO

Sample KeyRed. Yellow, & White

O ^ Colors

Keys ,~. - ^ - ~

FoodDispenser

O-s G c r

Sample KeyI •o* «y* »yy"

LettersComparison > ' " 'Keys O O O

Red Yellow WhiteColors

Q^WaterKey

^.WateDisp

Q^Water

FoodKeyFood-J 1Disp. ' '

"Thank You"Key '

"What Color?"Key *

O

O

Figure 1. The experimental apparatus. (Adjoining keyboards for the two birds were separated by aPlexiglas barrier. The mander's keyboard is on the left and the tacter's is on the right. The procedureis given in Table 1.)

The tacter's procedure. In the first step of thetacter's procedure, the bird, which was food deprivedon some days, water deprived on others, and occasion-ally both food and water deprived, according to aquasi-random schedule, learned to peck separate keysfor food and water. These responses were reinforcedonly if the S0" light was flashing. In this way, theflashing light was paired with both food and water,two distinct unconditioned reinforcers, and so met therequirements for establishing it as a potential S°".Regardless of the tacter's state of deprivation, whenthe S°™ was flashing, responses to both the food andthe water key were followed by their correspondingreinforcers. Even on days when the tacter was underonly one state of deprivation, it typically respondedfor, received, and consumed both food and water.Pecks on the food key resulted in access to mixedgrain for 4 s; pecks on the water key caused 0.4 ml ofwater to be dispensed.

Once the flashing S0" reliably controlled pecks onthe food and water keys, conditions were changed suchthat the tacter was required to perform successfullyan arbitrary matching-to-sample task to cause the S0"to flash: the letters R, W, and Y were presented quasi-randomly on the sample key, and the tacter was con-ditioned to peck the sample key first and then theappropriately colored comparison key, red, white, oryellow. Correct matching responses were reinforced bythe flashing S0" light, and then a peck on the food or

water key was reinforced with food or water, as before.Hence, the matching-to-sample task became the firstlink in a two-component chain, with pecks on the foodor water key as the second component

The mander's procedure. The mander, on theother hand, first learned to match each color as it waspresented on its sample key to the corresponding letteron a comparison key. A correct matching responsewas directly reinforced with food. After the bird be-came 90% -95% proficient at this task, two additionalrequirements were added: The mander was requiredto "ask for" a sample color by pecking an illuminatedkey located at floor level (Figure 1) labeled "Whatcolor?" This response illuminated a second floor keylabeled "Thank you," and a peck on this key thenprojected a color onto the mander's sample key. Hence,"for the mander, the arbitrary matching-to-sample taskwas the terminal link of a three-component chain inwhich the first link required a peck on the illuminated"What color?" key and the second link required a peckon the illuminated "Thank you" key.

"Both the mander's and tacter's performances wereconditioned through standard fading, shaping, andchaining procedures (Catania, 1979). After each birdhad mastered its individualized chain schedule to90%-95% accuracy, both were placed in their cham-bers concurrently for approximately 1 hr on five oc-casions. During these periods only the house lightsand noise generator were operating. The birds were in

"SYMBOLIC COMMUNICATION" 375

Table 1The Complete Verbal" Episode

Mander: The mander's "What color9"to Component 2.

Component 1

key is illuminated, and the mander pecks it, which advances the chain

Component 2

Tacter: A letter is projected onto the tacter's sample key. The tacter matches a color to the letter (i.e., itpecks the letter projected on its sample key and then the appropriate comparison color). If the tactererrs, the house lights in its chamber are dimmed for 3 s, and the conditions of Component 1 arereinstated. A correct match advances the chain to Component 3.

Component 3

Mander The "Thank you" key in the mander's chamber is illuminated. The mander pecks the "Thank you"key, which advances the chain to Component 4

Component 4

Mander: The color previously pecked by the tacter appears on the mander's key.Tacter The S0" begins to flash in the tacter's chamber.

Component 5

Manden The mander matches a letter to the color projected on its sample key (i.e., it pecks the color and thenthe appropriate comparison letter). This response is reinforced with mixed grain. If the mander errs,the house lights in its chamber are dimmed for 3 s, and the conditions of Component 1 are reinstated.At the end of this component (i.e , after the mander is rewarded or finishes a time-out), the conditionsof Component 1 are reinstated.

Tacter: The tacter receives either mixed grain or water by pecking the appropriate key. On satiated days, itreceives only the S°™, and pecks on the food and water keys are not reinforced.

plain view of each other through the Plexiglas divider.This measure was taken to dissipate emotional re-sponses evoked by the sight of a strange bird.

The interanimal interaction. In the next phase ofthe procedure, the two birds were placed in theirchambers concurrently, and the SD for each link oftheir respective chains was made contingent on theother bird's behavior. The sequence began when themander pecked the illuminated "What color?" key.This response automatically caused a letter to appearon the tacter's sample key. The tacter then matchedto sample (i.e., it pecked the letter projected on itssample key and then the comparison color key thatcorresponded to the sample letter). This response il-luminated the "Thank you" key in the mander's cham-ber. (If the tacter made an error, the house lights inits chamber were dimmed for 3 s after which themander's "What color?" key was again illuminatedand the sequence began from the start.) When themander pecked the "Thank you" key, two things en-sued: The tacter's S°" began to flash, and the com-parison color previously pecked by the tacter appearedon the mander's sample key.

From this point each bird's behavior was independ-ent of the other's; with the S°" flashing, the tactercould receive either food or water by pecking the

appropriate key, and the mander could receive food bycorrectly matching the color on its sample key to thecorresponding letter on its comparison keys. (If themander incorrectly matched to sample, its house lightswere dimmed for 3 s, and its portion of the trial wasaborted without reinforcement.) The S*5** was pro-grammed to flash until the tacter pecked either thefood or the water key; at that point the flashing S0"terminated, and the tacter received an unconditionedreinforcer. (If the tacter did not peck either key, theflashing S0" terminated at the end of the mander'sreinforcement period.)

Both birds learned this interaction with amazingspeed, and in 1 week they were maintaining a seriesof successful interchanges of discriminative stimuli bymeans of arbitrary matching tasks with greater than90% accuracy. The interlocking performances resem-bled a spontaneous conversation. After each bird com-pleted a component link in the sequence, it usuallyattended to the activities in the adjacent chamber; forexample, after consuming either food or water, thetacter typically walked over to the region of the man-der's "What color?" floor key and waited. When the"What color?" key became illuminated, the tacterusually began to peck on the Plexiglas directly abovethe mander's "What color?" key. The mander then

376 DAVID LUBINSKI AND KENNETH MACCORQUODALE

Table 2Percentage of Correct Responses for Each Bird and Their Correspondence

TacterDay condition Tacter* Mander" Correspondence1"

Condition 1

1 D 32/36 = .89 31/32 = .97 .862 S 23/33 = .70 22/23 = .96 .673 D 28/36 = .78 28/28=1.0 .784 S 20/21 = .95 19/20 = .95 .905 D 52/60 = .87 50/52 = .96 .846 S 28/29 = .97 26/28 = .93 .907 D 51/67 = .76 49/51 = .96 .738 S • 22/25 = .88 22/22 = 1.0 .889 D 54/64 = .84 52/54 = .96 .81

10 S 19/25 = .76 17/19 = .89 .68. 1 1 D 50/61 = .82 50/50 = 1.0 .82

12 S 33/38 = .87 32/33 = .97 .8413 D 55/60 = .92 53/55 = .96 .8814 S 25/28 = .89 24/25 = .96 .8515 D 40/42 = .95 39/40 = .98 .9316 S 10/11 = .91 9/10 = .90 .8217 D 53/59 = .90 50/53 = .94 .8518 S 15/16 = .94 11/15 = .73 .6919 D 57/64 = .89 54/57 = .95 .8520 S 20/26 = .77 18/20 = .90 .6921 D 47/55 = .85 45/47 = .96 .8222 S 6/8 =.75 6/6 =1.0 .7523 D 41/45 = .91 41/41 = 1.0 .9124 S 15/23 = .65 15/15 = 1.0 .6525 D 53/58 =.91 50/53 = .94 .8626 S 19/31 = .61 19/19 = 1.0 .6127 D 53/66 = .80 50/53 = .94 .7528 S 16/25 = .64 16/16 = 1.0 .6429 D 67/84 = .80 63/67 = .94 .7530 S 19/34 = .56 19/19 = 1.0 .56

Condition 2

31 D 47/49 = .96 40/47 = .85 .8232 D 35/35 = 1.0 34/35 = .97 .9733 D 43/49 = .88 40/43 = .93 .8234 D 58/62 = .94 57/58 = .98 .9235 D 53/58 = .91 52/53 = .98 .89

Condition 3

36 S 18/20 = .90 18/18=1.0 .9037 D 50/50 = 1.0 48/50 = .96 .9638 S 6/7 =.86 5/6 =.83 .7139 D 48/51 = .94 46/48 = .96 .9040 S 7/7 =1.0 6/7 =.86 .8641 D 56/60 = .93 55/56 = .98 .9142 S 3/5 =.60 2/3 =.67 .4043 D 57/58 = .98 54/57 = .95 .9344 S 12/14 = .86 9/12 = .75 .6545 D 28/32 = .88 26/28 = .93 .8246 S 3/4 =.75 2/3 =.67 .5047 D 45/48 = .94 43/45 = .96 .9048 S 0/249 D 45/50 = .90 41/45 = .91 .8250 S 3/4 =.75 2/3 =.67 .5051 D 45/49 = .92 39/45 = .87 .80

"SYMBOLIC COMMUNICATION" 377

Table 2 (continued)

Day

525354

Tactercondition

SDS

Tacter*

0/159/63 = .94

0/1

Mander*

52/59 = .88

Correspondence*1

.83

5556575859606162636465

DDSDSDSDSDS

Condition 4

54/100 = .5457/65 = .88

2/2 =1.051/54 = .94

5/6 =.8356/58 = .9715/15 = 1.061/62 = .98

2/2 =1.042/46 = .91

7/12 = .58

52/54 = .9654/57 = .95

1/2 =.5038/51 = .75

3/5 =.6042/56 = .7512/15 = .8054/61 = .89

2/2 =1.042/42 = 1.0

7/7 = 1.0

.52

.84

.50

.71

.50

.73

.80

.871.00.91.58

Note All sessions were 45 min in length except for those in Condition 2 which were 1 hr. Before Condition 4began, a 1-day transition session was implemented, which consisted of the tacter's being 23-hr food and waterdeprived (Day 55). S = satiated; D = deprived of food and water.* In the fractions, the denominator is the total number of matching responses, and the numerator is the numberof correct matching responses. Because the opportunity for the mander to engage in matching was contingenton the tacter's accuracy, the denominator for the former equals the numerator for the latter b The product ofthe two fractions represents the percentage of correct correspondence.

typically hurried over and pecked the "What color?"key and then walked over to the "Thank you" floorkey and waited for it to become illuminated by thetacter's matching response. If the tacter was at allsluggish, the mander rapidly pecked the Plexiglas asthough impatient. Similarly, if the mander was theleast bit hesitant about pecking the "Thank you" key,the tacter would peck the Plexiglas directly above the"Thank you" key Typically, however, each bird wasquick in completing its next link, and these "impa-tient" pecks were few in number, but they did regularlyoccur if either bird paused. After each interchange,the "What color?" key was illuminated, and, usuallywithout hesitation, the mander pecked it, manding thetacter to match another letter, and the cycle pro-ceeded. On alternating days the tacter was either 23-hr food deprived or 23-hr water deprived, and on a fewoccasions it was both food and water deprived.

At this point the training was complete. Both birdswere emitting discriminative responses derived froman arbitrary matching-to-sample task under the con-trol of socially supplied discriminative stimuli. Afterobserving this interanimal interaction for approxi-mately 3 weeks during which accuracy remained highon both sides, we decided to test whether the tacter'sperformance could be maintained without deprivationand unconditioned reinforcement in order to ascertainwhether the performance could be maintained withthe tacter emitting only tacts.

For the next 5 days, the tacter was returned to itshome cage and given free access to both food andwater. The birds were then returned to the experimen-tal chambers.

For 30 days the tacter was deprived and satiated onalternate days (A = deprived, B = satiated). Each

day's session lasted 45 min. The 15 odd-numbereddays of the A condition essentially reinstated theconditions of the original experiment, that is, on thosedays the tacter was approximately 22.5-hr food andwater deprived. In the B condition, however, the tacterwas food and water satiated (i.e., it had free access toboth food and water for approximately 23 hr prior toeach B session). Moreover, on the even-numbered Bdays, the food and water keys were disconnected fromtheir respective dispensers so that no unconditioneddeprivation-relevant reinforcers were delivered. All ofthe remaining contingencies were as before; when themander pecked the "Thank you" key, the SGcr flashedin the tacter's chamber, but that was the only conse-quence of a correct match by the tacter. The resultsare summarized in Table 2

Results

Condition 1

Table 2 shows that the tacter did notstop matching colors to the letter samplesin the B condition, even though the birdwas food and water satiated and matcheswere not followed by food or water (seeFigure 2). Thus, the "verbal" behavior emit-ted by the tacter in the B condition in-volved only tacts. In the B condition, thetacter's matches were 79% accurate; in theA condition, 86% accurate. Under either

378 DAVID LUBINSKI AND KENNETH MACCORQUODALE

E•

^^Tm ~j T*iT

9 •

•Hi A two-pigeon exchange involving tacting. (Upper left: The mander [left] asks the tacter

[right] to match [i.e., tact] the letter projected on its sample key. Upper right: The tacter matches acolor to the letter. Middle left: The mander pecks the "Thank you" key, which rewards the tacterwith the flashing SGcr; this response also presents to the mander the color previously pecked by thetacter. Middle right: The mander matches a letter to the color projected on its sample key as thetacter turns toward the SG". Bottom: The mander is rewarded with food for correctly completing the"symbolic" exchange, and the tacter attends to the flashing SGcr.)

condition, the accuracy was significantlygreater than chance (33%). More impor-tant, the difference in accuracy on A versusB days was not significant statistically.What did differ significantly between thetwo conditions, however, was the frequencyof responses emitted by the tacter. In thedeprived condition, the tacter emitted an

average of 48.9 matches per session; in thesatiated condition, an average of only 19.3(p < .001). This resulted because the birdtended to pause before pecking the letteron its sample key when the mander asked,"What color?".

The interaction between the two birdsbecame interestingly elaborated on the tac-

"SYMBOLIC COMMUNICATION" 379

ter's B days. During the long periods of thetacter's pausing, the mander exhibited typ-ical pigeon "aggressive display" behaviorseemingly directed toward the tacter, pac-ing back and forth along the Plexiglas di-vider and pecking at the Plexiglas.

The tacter's behavior was also elaboratedon its B (satiated) days. On some occasions,after receiving the SGcr the bird pecked thefood or water key, even though it was sa-tiated and these keys were inoperative. Atother times it merely attended to the flash-ing light until the light terminated. On stillother occasions, the tacter displayed emo-tional behavior, such as wing flapping, ag-gressive displays directed toward the man-der, and occasionally pecking of the Plex-iglas. This emotional behavior seemed tobe in response to the emotional behavior ofthe mander (which typically became moreintense the longer the tacter paused). Onseveral occasions, however, the tacter'smatching response occurred when the man-der was relatively still and not emitting anynoticeable behavior (sometimes evenstanding with its back toward the tacter).

Because of these dynamic aspects of theinteraction, Travis Thompson (personalcommunication, 1982) suggested that theaudience (i.e., the mere presence of themander, and its impatient behavior) mightbe critical for maintaining the tacter'smatching behavior. To test whether thiswas the case, we isolated the birds fromeach other by covering the Plexiglas barrierwith an opaque divider. The same contin-gencies were left in place at the end ofCondition 1 except now the birds were un-able to see each other.2

Condition 2

Condition 2 was a brief period of adap-tation to the novel barrier. For all 5 daysof this phase, the tacter was approximately22.5-hr food and water deprived. Each ses-sion lasted 60 min. The data in Table 2reveal that the accuracy of responses forthis condition remained high. However, theinteraction lacked the pace observed inCondition 1. In Condition 1, after complet-ing a response, both birds had attended tothe key light changes in the adjoiningchamber as well as to the behavior of the

other bird. With the opaque divider inplace, these stimuli were missing, and eachbird tended to move about in its respectivechamber after completing a component linkin the interaction. Hence, after one of thebirds completed a component in the se-quence, the other bird often did not imme-diately attend to the new stimulus on itsresponse panel, a situation that resulted inlong pauses on the part of both birds. Thetacter, especially, sometimes missed theSGcr completely or received it only after along pause since its last matching response.

Condition 3

Condition 3 was maintained for 19 days.It replicated Condition 1 except that thebirds were unable to see each other. Thetacter was satiated on alternate days, asbefore. As Table 2 reveals, the tacter con-tinued to match for the first 6 satiated days,but on the 7th, 9th, and 10th sessions ofthe satiated condition (i.e., Days 48,52, and54, respectively) the tacter stopped match-ing. In fact, the tacter spent Sessions 9 and10 literally sitting on the floor of its cham-ber, facing away from the response panel.At this point we discontinued Condition 3.

Some topographical observations regard-ing the first 6 satiated days of Condition' 3are noteworthy, however. The emotionalbehavior displayed by the tacter during thesatiated condition in Condition 1 was ab-sent in Condition 3, and responses to theinoperative food and water keys in the pres-ence of the flashing S(>cr were few. Typi-cally, after matching a color to the sample

2 A question might arise at this point as to why wedid not simply discontinue the presentation of the S0"for Condition 2, rather than insert the opaque barrierbetween the two birds. Our reasoning was as follows:At the end of Condition 1, the S*5" was a powerfulreinforcing consequence for the tacter, but so was thepresentation of the mander's "Thank you" floor light,which the tacter could see. In addition, the mander'sresponse directed toward the "Thank you" key was apowerful reinforcer for the tacter. The "Thank you"key and the mander's responses presumably becamereinforcing through temporal pairings with the S0**.Hence, we felt it would be experimentally less clut-tered simply to insert an opaque barrier and thuscontrol for all the stimuli in the adjacent chamber.This left the S0" as the only consequence of thetacter's matching response.

380 DAVID LUBINSKI AND KENNETH MACCORQUODALE

letter, the tacter paused, noticed the SGcr

when it began to flash, and attended to ituntil the flashing light terminated. Thetacts emitted in these sessions, althoughfew in number, satisfy the conditions forpure tacting—they were emitted in the ab-sence of either deprivation or aversivethreatening behavior from the mander.Tacting extinguished in this phase eitherbecause the SGcr lost its effectiveness orbecause stimuli from the mander as anaudience and source of motivation werenecessary for continued tacting (or both).To discover whether tacting would reap-pear if the opaque barrier was removed, wereinstated the conditions of Condition 1 for10 days.

Condition 4

Condition 4 consisted of 10 ABAB ses-sions (with tacter deprived and satiated inalternate sessions). As revealed in Table 2,accurate tacts were again emitted by thetacter in the satiated condition. The visiblepresence of the mander revived the tacter'smatching behavior. As in Condition 1, thetacter's accuracy between A (deprived) andB (satiated) conditions did not differ sig-nificantly by statistical test (A, 94% cor-rect; B, 88% correct). However, the fre-quency of matching responses emitted bythe tacter did differ significantly betweenconditions (mean number of responsesemitted for each session: A, 53.4; B, 6.2; p< .001). These observations substantiateCatania's (1979, p. 237) view that whereasthe audience does not play a part in defin-ing the tact or in determining its topogra-phy, it does play a role in maintaining thetact's strength, although it is not sufficientto maintain the tact indefinitely. It is asupplementary variable that can only en-hance a response that has other primarysources of strength (MacCorquodale, 1969,pp. 838-839; Skinner, 1957, p. 181).

Discussion

This experiment demonstrates that pi-geons can be conditioned to "tact." Thisfinding adds a new feature to the burgeon-ing domain of "interanimal symbolic com-munication" and generates some provoca-tive extensions. It is interesting to specu-

late on whether an SGcr paired with morethan two unconditioned reinforcers mightbe able to maintain a greater frequency ofresponding (i.e., tacting) in the satiatedcondition, especially in the "over thephone" condition of Condition 3; we usedthe weakest SGcr possible (namely, a neutralstimulus paired with only two reinforcers,food and water). Additional reinforcerscould have been paired with the flashinglight to produce an SGcr with possibly amore potent reinforcing effect. If later-evolved organisms were conditioned in thismanner, more durable and more complexperformances might be maintained. Varia-tions of the procedures are possible. Onemight try a more gradual fading or degrad-ing of the tacter's view of the mander todetermine whether tacting could ultimatelybe maintained without an audience in view.One might also test whether the mander,when fully visible, is sufficient to maintaintacting even if the SGcr is omitted entirely.Finally, it would be interesting to ascertainwhether a "new" matching response (e.g.,G for green) could be conditioned in a pi-geon (or, perhaps, a chimpanzee) when theorganism receives only an SGcr for correctresponding and is satiated for the uncon-ditioned reinforcers with which the SGcr wasoriginally paired.

References

Catania, A. C. (1979). Learning. Englewood Cliffs, NJ:Prentice-Hall.

Cullen, J. M. (1972). Some principles of animal com-munication. In R. A. Hinde (Ed)., Non-verbal com-munication (pp. 101-122). Cambridge: CambridgeUniversity Press

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

MacCorquodale, K. (1969). B. F. Skinner's verbal be-havior: A retrospective appreciation. Journal of theExperimental Analysis of Behavior, 12, 831-841.

Savage-Rumbaugh, E. S., Rumbaugh, D. M., & Boy-sen, S. (1978). Intentional symbolic communicationin chimpanzee. Science, 201, 641-644.

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. (1957). Verbal behavior. New York:Appleton-Century-Crofts.

Received November 5,1983Revision received June 15, 1984 •