COVERT VERBAL MEDIATION IN ARBITRARY MATCHING TO SAMPLE

CARL T. SUNDBERG1, MARK L. SUNDBERG

1, AND JACK MICHAEL2

1BEHAVIOR ANALYSIS CENTER FOR AUTISM2WESTERN MICHIGAN UNIVERSITY

Covert verbal mediation was examined in an arbitrary matching-to-sample (MTS) preparation with ahigh-verbal group (college students) and a low-verbal group (adults with intellectual disabilities). Arbi-trary relations were established between nonsense words, visual symbols, objects, and hand signs. Taskdifficulty was balanced for the groups based on errors during acquisition. All participants experienced ahand sign condition, and three MTS conditions each with a unique configuration of the comparisonarray: fixed location, random location, and all symbols the same. The same symbol condition wasdesigned to impede a participant’s ability to label individual symbols. The results showed that disruptinglabeling adversely affected MTS performance for high-verbal participants, but not for low-verbal partici-pants. The data suggest that high-verbal participants depended on mediating verbal behavior and jointcontrol to assist them in finding the correct comparison stimulus. Low-verbal participants could notbenefit from verbal mediating variables and likely relied on unmediated contingencies, or some formof nonverbal mediation. For the high-verbal group, 19 different putative emergent relations were identi-fied as occurring at various stages of acquisition between the sample stimulus and the selectionresponse. These emergent relations likely provided supplementary sources of stimulus control that par-ticipated in evoking MTS selection behavior.Key words: covert verbal mediation, emergent stimulus–stimulus relations, joint control, matching to

sample, multiple control, verbal behavior

Sidman’s (1971) original work on stimulusequivalence provided a conceptual foundationand research methodology for the empiricalinvestigation of complex behaviors. The studyof equivalence and resulting emergent rela-tions has since produced a rich body of dataand remains at the forefront of behavior analy-sis. However, despite a sizable collection ofresearch, a consensus on the variables respon-sible for the emergence of new behavioralrelations remains elusive (Dougher, Twohig, &Madden, 2014).Since the pioneering work of Sidman, the

study of stimulus equivalence and emergentrelations has produced four similar, but dis-tinct, conceptual frameworks that strive to pro-vide an account of emergent stimulus–stimulusrelations. The four frameworks include:

(1) Sidman’s (1994) equivalence theory,(2) Hayes, Barnes-Holmes, and Roche’s (2001)relational frame theory (RFT), (3) Horne andLowe’s (1996) naming theory, and (4) Lowenk-ron’s (1998) joint control account. A major dis-tinction between these theories involves therole of a participant’s mediating verbal behav-ior as a causal variable in emergent stimulus–stimulus relations.

Both Sidman (e.g., Sidman, 1994; Sidman &Tailby, 1982) and Hayes (e.g., Hayes et al.,2001; Quinones & Hayes, 2014) generallymaintain that verbal mediation is not relevantto the emergence of new relations or rela-tional frames. These researchers tend to notidentify, discuss, or account for potential overtor covert mediating verbal behavior in theirmatching-to-sample (MTS) research with ver-bal participants. For example, in the study byQuinones and Hayes (2014) on ambiguousrelational networks, the results obtained withcollege students were attributed to theexperimenter-manipulated contingencies.With the exception of a brief mention of thepossible role of rule-governed behavior, partic-ipant verbal behavior between the samplestimulus and the selection response was dis-missed as a relevant source of stimuluscontrol.

We thank Jessica Linneweber, Dave Palmer, Bob Ryan,Hank Schlinger, Cindy Sundberg, and Devon Sundbergfor their contributions to this study. Portions of this paperwere based on a dissertation submitted to Western Michi-gan University in partial fulfillment of the requirementsfor a doctoral degree by the first author under the supervi-sion of the third author.Address correspondence to: Carl T. Sundberg, Behavior

Analysis Center for Autism, 11902 Lakeside Drive, Fishers,IN 46038 (email: ctsundberg@aol.com)doi: 10.1002/jeab.434

JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR 2018, 109, 600–623 NUMBER 3 (MAY)

© 2018 Society for the Experimental Analysis of Behavior600

By contrast, Horne and Lowe (1996) andLowenkron (1998), who base their positionson Skinner’s (1957) analysis of verbal behav-ior, argue that overt and covert verbal andnonverbal mediating behaviors play an impor-tant role in the development of equivalenceclasses, relational frames, and emergent rela-tions. In support of their view, several studieshave demonstrated a participant’s self-verbalbehavior can affect performance on variouscomplex tasks (e.g., DeGraff & Schlinger,2012; Greer & Longano, 2010; Horne, Lowe, &Randle, 2004; Lowenkron, 1991, 2006; Miguel,Petursdottir, Carr, & Michael, 2008; Randell &Remington, 1999; Sidener & Michael, 2006;Wulfert, Dougher, & Greenway, 1991).Stewart, McElwee, and Ming (2013)

objected to Horne and Lowe’s (1996) focuson mediation, stating “the fact that namingand joint control both require an additionalmediational process to explain derived stimu-lus relations in comparison to RFT can beseen as a weakness of the former” (p. 143).Similarly, Stromer (1996) expressed concernabout Horne and Lowe’s verbal mediationaccount. The objection to mediating verbalbehavior is typically due to the private natureof these variables. Nonetheless, if they play acausal role they arguably should be taken intoaccount (Skinner, 1974).Covert verbal mediation must be at least

indirectly quantified to identify its role in com-plex behavior. Direction can be found in theearly research on overt nonverbal mediatingbehavior. Ferster and Skinner (1957) initiallydefined mediating behavior as “Behavioroccurring between two instances of a responsebeing studied (or between some other eventand such instance) which is used by the organ-ism as a controlling stimulus in subsequentbehavior” (p. 729). For example, Sidman(1960) noted “The subject will often adopt aposture in which his whole body or part of itmaintains a constant position relative to thecorrect container. Such mediating behaviormay enable the subject to select the correctcontainer even after the lapse of a consider-able amount of time” (p. 375).Several studies have been conducted on var-

ious types of overt nonverbal mediating behav-ior with pigeons, monkeys, and rats(e.g., Blough, 1959; Eckerman, 1970; Hodos,Ross, & Brady, 1962; Laties, Weiss, Clark, &Reynolds, 1965; Laties, Weiss, & Weiss, 1969;

Schoenfeld & Cumming, 1963; Shimp & Mof-fitt, 1977; Wilson & Keller, 1953). For exam-ple, in Blough (1959), pigeons that emitteddifferential stereotypic behavior during differ-ent delay conditions demonstrated more cor-rect responding than pigeons that did notengage in such behaviors. In addition, ifpigeons emitted the wrong mediating behav-ior, they typically emitted the wrong response.Blough concluded “behavior during the delayinterval seemed to determine the matchingresponse” (p. 156). This line of research isimportant to the current study because it dem-onstrates that overt nonverbal mediatingbehaviors can provide discriminative stimulifor matching responses.

Parsons and colleagues (Parsons & Ferraro,1977; Parsons, Taylor, & Joyce, 1981; Polson &Parsons, 1994) extended the animal researchon mediating behavior to humans. For exam-ple, in Parsons et al. (1981), kindergarten chil-dren were trained on a delayed MTS taskunder three conditions. In the differentialcondition, children were required to emit aspecific collateral behavior during the delayinterval based on each sample stimulus(a bright light or a dim light). In the commoncondition, the same response was required forboth sample stimuli, and in the nondifferen-tial condition, either response was reinforced,regardless of the sample stimulus. The resultsshowed that “subjects who reliably engaged insample-specific collateral behavior…rapidlylearned the conditional discrimination. Sub-jects who engaged in identical or nonspecificcollateral behavior…either failed to acquirethe discrimination or did so incrementallyover a number of sessions” (pp. 259-260).These findings, along with the animal data,demonstrate that overt nonverbal mediatingbehavior can function as causal variablesaffecting subsequent behaviors in a MTSpreparation.

Lowenkron (1984, 1988, 1989) furtherextended this line of research in a series ofexperiments in which participants were taughtovert responses designed to mediate general-ized identity matching (e.g., rotating an arrow,hand signs, use of a compass). For example, ina study examining the role of overt verbal medi-ation, Lowenkron (1988) taught four childrenwith intellectual disabilities to use a hand signto tact (label) a sample shape. Participantswere then taught to maintain that hand sign

601COVERT VERBAL MEDIATION

over a delay and to tact a matching compari-son stimulus with the same hand sign. In ageneralization test in which participants didnot receive hand sign tact training, they onlyreliably matched the novel stimuli when thecomparisons were present; if the sample wasremoved with a delay before the comparisonselections appeared, participants matched atchance level (25%). However, as soon as theparticipants learned hand signs to tact thenovel shapes, generalized matching appearedimmediately. Lowenkron (1988) concludedthat generalized delayed matching was depen-dent on these overt verbal mediatingresponses because the responses preserved theidentity relation during generalization tests.Sundberg and Sundberg (1990) also noted

a mediating effect of hand signs while examin-ing the distinction between sign languagetraining and icon selection training with adultswith intellectual disabilities. In that experi-ment, various arbitrary verbal relations wereestablished following what Michael (1985)defined as topography-based verbal behaviorand selection-based verbal behavior. Intopography-based verbal behavior, a speakeremits a different response form for each refer-ent (e.g., spoken words, hand signs), whereasin selection-based verbal behavior, the speakeremits the same response form (e.g., pointing,exchanging) to indicate a specific stimulus(e.g., icon selection apps, PECS). The partici-pants in Sundberg and Sundberg were trainedon both topography-based and selection-basedverbal behavior, and then tested for emergenttransitivity. The results showed thattopography-based training with hand signsproduced some transitivity in the equivalencetests, whereas selection-based training pro-duced none. These results were replicated andextended by Wraikat, Sundberg, and Michael(1991), who also showed that low-verbal partic-ipants demonstrated some transitivity in thetopography-based conditions, but none inselection-based conditions.The Sundberg and Sundberg (1990) and

Wraikat et al. (1991) data support Lowenk-ron’s (1988) findings that topography-basedverbal behavior can produce response prod-ucts that function as causal variables that mayparticipate in evoking other forms of verbalbehavior, including MTS stimulus selection intransitivity tests. Reflecting on the Sundbergand Sundberg data and his own, Lowenkron

(1991) suggested that even for the low-verbalparticipants the selection-based response inarbitrary MTS “depended on mediation by atopography-based behavior” (p. 126).

Lowenkron’s (1998) analysis of joint controlmay help to identify some of the behavioralprocesses involved when topography-based ver-bal behavior functions as a causal variable in aselection-based task. Joint control occurs whentwo separate antecedents that evoke the sameresponse topography arise simultaneously. Thecombination of these variables generates anemerged SD that can immediately evokebehavior. For example, when looking for adeparture gate at an airport, one might firstlook at the departure board and find the cor-rect gate number (e.g., B85). A self-echoicprompt allows the traveler to retain the termi-nal and gate number over time, distance, anddistractions (if she fails to do so, she may findherself rechecking a departure board). As thetraveler passes by gates, she may overtly orcovertly tact the numbers (e.g., “there’s B81”).Joint control occurs when her self-echoicmatches the response form produced by thetact of the numbers. As a result of the conflu-ence of these variables, a new discriminativestimulus (SD) emerges that evokes selectionbehavior (going to that gate) followed by rein-forcement (successful boarding).

Lowenkron (2006) extended the analysis ofjoint control to covert verbal mediation with twoexperiments in which at least one of the vari-ables involved a private event. The first experi-ment attempted to establish generalizedmatching behavior without the tact compo-nent. The results showed that without the tact,novel generalized responding failed toemerge. However, once the tact was estab-lished, novel generalized responding occurredimmediately. In the second experiment theself-echoic aspect of joint control was dis-rupted. Participants were required to engagein an incompatible vocal distractor task (read-ing numbers out loud) following the presenta-tion of the sample stimulus. The resultsshowed that the distractor task impaired per-formance on tasks that involved a relativelylong delay between the sample and compari-son stimuli. These data not only replicatedLowenkron’s (1984,1988,1989) overt media-tion data and those of Parsons et al. (1981),but they also demonstrated how indirect mea-sures could be used to quantify private events

CARL T. SUNDBERG et al.602

(Palmer, 2011; Skinner, 1953). Subsequentstudies have further extended the research ondisruption of covert verbal mediators(e.g., Clough, Meyer, & Miguel, 2016;DeGraaf & Schlinger, 2012; Gutierrez, 2006).The purpose of the current study was to fur-

ther examine verbal mediation and its effectson the performance of high-verbal (collegestudents) and low-verbal (adults with intellec-tual disabilities) participants across four sepa-rate tasks. One task involved topography-basedverbal behavior (hand signs), and three tasksinvolved selection-based verbal behavior(MTS). In the topography-based condition,arbitrary visual and auditory stimuli were usedto establish various relations with arbitraryhand signs. In selection-based training, all par-ticipants experienced three conditions. In thefixed-location condition, the comparison stim-uli were always in the same position, and inthe random-location condition, the position ofthe comparison stimuli changed after eachcorrect response. In the same-symbol condi-tion, all the symbols in the comparison arraylooked exactly alike. The goal of this conditionwas to make it harder for participants to givespecific names to each symbol, thereby dis-rupting covert tacting as a source of control

for selection behavior. Exit interviews and talk-aloud procedures were conducted at the con-clusion of the study (Ericsson & Simon, 1993;Potter, Huber, & Michael, 1997).

Method

ParticipantsEight participants were recruited for the

study. Four were adults with intellectual dis-abilities between the ages of 24 and 52 years,and four were college students between theages of 21 and 25 years. The participants withintellectual disabilities were identified as thelow-verbal group and the college participantswere identified as the high-verbal group. Allparticipants are referred to by pseudonyms.

Low-verbal group. The participants in thelow-verbal group functioned in the severe tomoderate range of intellectual disability andall resided in the same group home. Two werefemale and two were male. In addition, theparticipants (1) had moderate to severe defi-cits in language skills, (2) possessed the man-ual dexterity for the formation of signs,(3) demonstrated the ability to imitate,(4) demonstrated the ability to follow simple

Table 1

VB-MAPP levels and skills for the participants in the low-verbal group

AgeGeneral

VB-MAPP Level Mand Tact Listener Imitation Intraverbal

Molly 52 Level 1: Lessthan18 monthslanguage level

No mands:Nonvocal, nosigns, noselectionsystem

No tacts Can follow manyin-contextdirections andselect at least10 items oncommand

Early VB-MAPPLevel 2: Canimitate at least20 actions

None

Jim 38 Level 1: Lessthan18 monthslanguage level

Can emit at least4 vocal mands.No signs, noselectionsystem

Can tactat least6 items

Can follow manyin-contextdirections andselect at least10 items oncommand

Mid VB-MAPP Level2: Can imitate atleast 30 actions

None

Eric 31 Level 2: Lessthan30 monthslanguage level

Over 20 vocalmands

Over50 vocaltacts

Can follow manydirections andselect at least50 items oncommand

Can imitate at least20 three-stepactions

Can emit atleast25 fill-ins

Debra 24 Level 2: Lessthan30 monthslanguage level

Over 20 vocalmands

Over50 vocaltacts

Can follow manydirections andselect at least50 items oncommand

Can imitate at least20 three-stepactions

Can emit atleast25 fill-ins

603COVERT VERBAL MEDIATION

instructions, and (5) had no prior experiencewith either sign language or an icon communi-cation system.Table 1 contains the age and approximate

verbal skills for each of the participants in thelow-verbal group at the outset of the study. Interms of the Verbal Behavior Milestones Assess-ment and Placement Program (Sundberg, 2014),all participants were classified as Level 1 ormid-Level 2 responders, which approximatelycorresponds with the linguistic abilities of atypically developing 1- to 2-year-old child.High-verbal group. All the participants in

the high-verbal group were female students atWestern Michigan University (WMU). Allstated that they had sufficient time for thestudy and that they had no history of sign lan-guage or an icon communication system.

SettingAll sessions were conducted in a room con-

taining a table, chairs, and filing cabinets/stor-age shelves. There was no foot traffic duringthe sessions and noise was at a minimum. Ses-sions were conducted 5 or 6 days a week for4 to 6 weeks.Low-verbal group. Sessions for the partici-

pants in the low-verbal group were conductedin a room of their residential facility. Sessionslasted 10 to 20 min between the hours of3:00 p.m. and 9:00 p.m.High-verbal group. Sessions for participants

in the high-verbal group were conducted in aroom at WMU. Sessions lasted approximately30 min.

MaterialsSymbol boards were created for selection-

based responding for both groups andcontained a variety of arbitrary symbols invarious positions. Each symbol was 4 in. ×4 in. in size, printed on paper, and pastedon 18 in. × 24 in. poster boards placedapproximately 2 in. apart (see Appendix Afor symbol board samples). Arbitrary handsigns were chosen for topography-basedresponding (see Appendix B for a descrip-tion of the hand signs). For two of the par-ticipants in the low-verbal group, a variety ofobjects were used as target nonverbal stimulifor tacting (e.g., LEGO block configurations,wooden cylinders).

Interobserver AgreementA trained observer collected reliability data

on 30% of the trials for the low-verbal groupand 28% of the trials for the high-verbalgroup. Interobserver agreement was calculatedby dividing the number of agreements, on atrial-by-trial basis, by the number of agree-ments plus disagreements, then converting theresult to a percentage. Mean agreement forboth groups was 98% (range 97% to 98%).

DesignAn alternating treatments design (Barlow &

Hayes, 1979) was used to compare four condi-tions for each participant in an A-B-C-D or A-B-C-D-A fashion. In addition, a repeated acqui-sition design (Boren & Devine, 1968) was usedto equate the difficulty of the MTS tasks acrossthe two groups and to mitigate possiblesequence effects. Several sessions were runprior to the formal experiment to identifytasks that would be equally difficult for the twogroups. The complexity of the tasks, whichinvolved various nonsense words, symbols,objects, hand signs, and array configurations,were manipulated to identify combinationsthat produced a similar number of errors forboth groups.

Each condition of the study was divided intoseveral rounds. A round represented the num-ber of trials until criterion was met. Duringeach round, a specific word or object corre-sponded to a symbol or hand sign. Trials wereconducted until the participant mastered therelation, or until 60 stimulus blocks were pre-sented without meeting the criterion. Whenthe criterion was met, a new round was con-ducted in the same way, except the samewords and objects now corresponded to differ-ent hand signs and symbols. This sequencewas repeated until it was determined by visualanalysis of the data (number of errors perround) that the participants had stoppedimproving (i.e., there was no downward trendin errors). At this point, the next conditionbegan.

Response DefinitionsA correct selection-based response was

defined as a point to the target symbol in acomparison array of 2 to 12 symbols (depend-ing on the individual participant) within 5 s of

CARL T. SUNDBERG et al.604

the presentation of the sample stimulus. Aselection-based tact involved pointing to a sym-bol when an object was presented (e.g., whenshown the plastic piece and asked “What’sthis?”, the participant pointed to the corre-sponding symbol). A selection-based intraver-bal involved pointing to a symbol when aspoken word was presented in the verbalframe “Which one is____?” as the sample stim-ulus (e.g., when asked “Which one is mojam?”,the participant pointed to the correspondingsymbol).A correct topography-based response was

defined as the participant emitting the corre-sponding hand sign within 5 s of the presenta-tion of the target stimulus. For a response tobe recorded as correct, it must have been aclose enough approximation to the desiredresponse that it was easily distinguishable fromthe other responses in the participant’s reper-toire. A topography-based tact involved a spe-cific motor response when an object waspresented as the antecedent stimulus(e.g., when shown a LEGO block configura-tion and asked “What’s this?”, the participantpulled her left ear). A topography-based intra-verbal involved a specific motor responsewhen a spoken word was presented in the ver-bal frame “What is the sign for____?”(e.g., when asked “What is the sign for jihba?”,the participant pulled her left ear).

ProcedureIn general, all participants were exposed to

three different selection-based arbitrary MTStasks involving nonsense words, symbols, andobjects, and one topography-based task involv-ing hand signs, nonsense words, symbols, andobjects. Each of the participants experiencedthe four conditions in a different order. Thedependent variable was the number of errorsmade per round in each of the four condi-tions. Comparisons were made after errorshad stabilized using a repeated acquisitiondesign (see Tables 2 and 3 for examples ofrepeated acquisition training).Independent variable: Low-verbal group.

Two participants in the low-verbal group weretaught to select the corresponding symbolwhen presented with an object and asked“What is this?” They were taught theseselection-based tact relations across three dif-ferent configurations of the comparison array.

In a separate topography-based condition, par-ticipants were taught to make a specific handsign when presented with a correspondingobject. The other two participants in the low-verbal group were taught to select the corre-sponding symbol when presented with a non-sense word. They were taught these selection-based intraverbal relations across three differ-ent configurations of the comparison array. Ina separate topography-based condition, theywere taught to make a hand sign when pre-sented with a nonsense word.

All participants in the low-verbal group pro-gressed through three selection-based condi-tions: fixed location, random location, and allsymbols the same (tact disruption). Individual-ized symbol boards (see Appendix A) wereused because the array size differed for eachparticipant (from two to six symbols). In thefixed-location condition, all symbols were dis-tinct and they remained in the same positionon the board until mastered. One board wasused for each participant for all trials. In therandom-location condition, all symbols weredistinct, but their position on the board wasshuffled after each correct response. Sixposter boards were used, each containing aunique arrangement of the same set of stimuli.The six boards were stacked in front of theparticipants. After every correct response, theexperimenter removed the poster board andput it on the bottom of the stack. Thus, thesymbol positions repeated every sixth correctresponse. In the same-symbol condition, allthe symbols on the board were identical, andone board was used for all trials for each

Table 2

An example of three rounds of word–symbol relations forDebra from the low-verbal group

Symbols

Round Window Horizontal SquareTripleSquare

1 mook rac mig doe2 doe mig jib dro3 slo dro rook jib

Note. For Round 1, when presented with the verbal stimu-lus “mook” the correct selection would be the symbol thatlooked like a window out of an array of four symbols. Theverbal stimulus “rac” goes with the horizontal symbol, etc.For subsequent rounds the same symbols are used but cor-respond to different nonsense words. Two new nonsensewords replaced two existing words each round.

605COVERT VERBAL MEDIATION

participant. The same-symbol condition wasdesigned to disrupt self-generated tacts bymaking all the comparison stimuli exactly thesame, and establishing their location as thetarget source of control.Participants proceeded through the phases

in various orders. Two participants in the low-verbal group started with the fixed-locationcondition, then progressed through therandom- location, same-symbol, and hand-signconditions. The two other participants startedwith the hand-sign condition, then progressedthrough the fixed-location, random-location,and same- symbol conditions, and thenrepeated the hand-sign condition. Both partic-ipants in the low-verbal group who started withthe hand-sign condition (Molly and Jim)repeated this condition at the end because itwas determined by the experimenters that thecondition change may have occurred beforethe errors stabilized.A table containing the word–symbol and

word–object relations was used to determinethe relations for each round (see Table 2 for asample of word–symbol relations). BecauseMolly only had two stimulus sets, the relationssimply alternated for each round. For exam-ple, if “Rook” was related with symbol 1 and“Poe” was related with symbol 2 in round one,then in round two, “Rook” was related withsymbol 2 and “Poe” was related to symbol1. Jim, Debra, and Eric had three, four, andsix stimulus sets, respectively. For each round,the symbol or the sign (depending on the con-dition) was constant, but two new one-syllablenonsense words were rotated into the sets.Independent variable: High-verbal group.

All four participants in the high-verbal group

were taught to select the corresponding symbolwhen presented with a two-syllable nonsenseword across three different configurations ofthe comparison array. Each board containedan array of 12 symbols. They were also taughtto make a hand sign when presented with atwo-syllable nonsense word in the topography-based condition. The progression through theconditions was the same as that of the low-verbal group.

For participants in the high-verbal group,packs of 10 data sheets were used (designed toaccommodate 30 rounds for each participant).For each round, the symbol or the sign(depending on the condition) was constant onthe data sheet for all 30 rounds, but eight non-sense words from the previous rounds (fourfrom the last and four from the one before)and four new words were rotated in to create12 all-new target relations.

Two participants in the high-verbal groupwent through the conditions in the followingorder: fixed location, random location, samesymbol (tact disruption), and sign. The othertwo started with the sign condition, then pro-gressed through the fixed-location, random-location, and same-symbol condition, and thenrepeated the sign condition for the same rea-son as stated above. A table containing theword–symbol and word–object relations wasused to determine the relations for eachround for every participant (see Table 3 for asample of word–symbol relations).

Selection-based pretraining. Pretraining forthe selection-based condition started with asymbol board being placed in front of a partic-ipant. For the two participants in the low-verbal group who were trained with objects

Table 3

An example of the word–symbol relations for five rounds for the high-verbal participants

Symbols

Round Vertical Cross Stairs Maze Window Square3 Quarter H Square Check I Horizontal

1 neeba peeba qeeba reeba paba quaida raca saba baba cafa daida haca2 paba quaida raca saba baba cafa daida haca seeba teeba veeba weeba3 baba cafa daida haca seeba teeba veeba weeba taba waca vaida vaca4 seeba teeba veeba weeba taba waca yaida vaca vihba daca naiad pafa5 taba waca yaida vaca vihba daca naiad pafa jafa hayba fuba meeba

Note. The nonsense words listed in each round were the verbal stimuli presented that correlated with the symbols listedat the top of each column. For example, in round one, when presented with the verbal stimulus “neeba” the correctselection was the vertical symbol. When presented with the verbal stimulus “peeba” the correct selection was the crosssymbol. The same symbols were used in every round but their correlation to a word changed. For example, in round twoit was the verbal stimulus “paba” that correlated to the symbol called vertical.

CARL T. SUNDBERG et al.606

(the tact relation), the experimenter placedeach of the objects on the symbols one at atime while saying “This goes with this.” Objectswere left on the board for 5 s. For the six par-ticipants who were trained using words (theintraverbal relation), the experimenterpointed to each symbol (at a rate of one every2 s) while saying “This is _____.” (e.g., “Biba”).The participant was then told that the experi-menter would hold up an object or present aword and the participant would have 5 s topoint to the pattern that went with that object.Data collection began immediately after thisdemonstration.Topography-based pretraining. Pretraining

for the topography-based condition startedwith a demonstration of each hand sign and arequest that the participant imitate the handsign. This process was repeated five times.Next, if the participants were trained usingobjects (the tact), the experimenter wouldpoint to each object and make the corre-sponding hand sign while saying “This is this.”If the participants were trained using words(the intraverbal relation), the experimenterwould make the hand sign for each word whilesaying “This is ____.” (e.g., “Biba”). The partic-ipant was then told that the experimenterwould hold up an object (if a tact) or presenta word (if an intraverbal) and the participantwould have 5 s to make the hand sign thatwent with that object or word.Fixed-location condition: Selection-based

training. The two participants in the low-verbalgroup who were trained on the tact relation(Molly and Jim) were shown the objects orLEGO block configurations one at a time andtheir corresponding symbol (as described inpretraining). Next, the experimenter ran-domly selected one of the items and presentedit to the participant along with their symbolboard and said “Which one is this?” (object/LEGO). For the other six participants whowere trained on the intraverbal relation, theprocess was the same except instead of anobject or a LEGO block configuration, a non-sense word was presented. For the participantsin the high-verbal group, if the correct symbolwas selected, the experimenter said “yes” andpresented the next trial. For the participantsin the low-verbal group, if the correct symbolwas selected, the experimenter praised theparticipants (e.g., “Yeah, good job!”) and pre-sented the participant with a nickel. If the

response was incorrect, the experimenter said,“No, this goes here” or “No, this is ____” whilepointing to the correct symbol.

For the high-verbal group, all 12 words wereeach presented once and then repeated in adifferent order. This occurred until the masterycriterion was met. When the criterion was met,the round ended and either the session endedor a new round began with 12 new word–symbol relations. Four rounds were conductedfor each participant during each session. Thisprocess was repeated until the participant’s per-formance stopped improving (i.e., the numberof errors had stabilized). For the participants inthe low-verbal group, the above procedureswere the same; all objects or words were pre-sented (2, 3, 4, or 6, depending on the partici-pant) once and repeated until the criterion wasmet. However, only one round per day was con-ducted with these participants.

Random-location condition: Selection-basedtraining. The random location condition wasconducted in the same way as the fixed-location condition except that a correctresponse produced a new trial involving a newsymbol board. This new board had the samesymbols, but the symbols were in differentlocations.

Same-symbol condition: Selection-basedtraining. The same-symbol (tact disruption)condition was conducted in a manner similarto the fixed- and random-location conditionexcept all the symbols on the board lookedexactly alike.

Hand-sign condition: Topography-basedtraining. The hand-sign condition started withthe pretraining on the signs (as describedabove). After pretraining, the participants werepresented with one object or one word andasked “What is the sign for this?” or “What isthe sign for ____?” (e.g., “Pog”). As with theprevious conditions, all of the objects or wordswere presented one at a time and thenrepeated. This process continued until the cri-terion was met; then, either the session endedor a new round began with new word–sign orobject–sign relations. This process was repeateduntil the participant’s performance stabilized.One round per session was conducted with theparticipants in the low-verbal group and fourrounds per session were conducted with theparticipants in the high-verbal group.

Reward contingency. The participants inthe low-verbal group were paid $1.00 per six

607COVERT VERBAL MEDIATION

min of participation and were given a nickelfor each correct response. Participants in thehigh-verbal group were paid $5.00 per ses-sion at the end of each session. All partici-pants or their guardians signed an informedconsent form.

MeasurementThe first hand sign or the first symbol indi-

cated following the presentation of the targetstimulus was recorded. Correct and incorrectresponses were recorded on the correspond-ing data sheets until the mastery criterion fora round was met. At that point, a new roundbegan if time permitted (high-verbal group),or the session ended (low-verbal group).

Exit Interview and Talk-aloud ProcedureAt the conclusion of the study, exit inter-

views and talk-aloud procedures were con-ducted (Ericsson & Simon, 1993; Potter et al.,1997). Three of the participants in the high-verbal group (Gina, Julie, and Erin) wereavailable for the interviews and talk-aloud pro-cedures (the participants in the low-verbalgroup did not have sufficient language skillsfor these activities). Each participant was

asked a series of questions designed to iden-tify strategies used and the extent of theirmediating verbal and nonverbal behavior.Also of interest was the extent to which medi-ated behavior was correlated with differentialperformance across conditions, or if any par-ticular type of covert behavior was correlatedwith differential performance across condi-tions. A mock session was also conducted. Theparticipants were asked to talk aloud as theywere presented a block of stimulus presenta-tions from the experiment. They wereinstructed to try to overtly vocalize anyresponses that they may have covertly emittedduring the actual experiment (e.g., given thesymbol that looked like stairs: “Miba. Thatsounds like maybe. Maybe I’ll go upstairs.”).They were also instructed to report any rela-tions for which they did not engage in anyconsistent and specific verbal behavior.

Results

Low-verbal group. The frequency of errorsduring the last four rounds of each conditionis presented in Figure 1 for Jim, a representa-tive participant from the low-verbal group.Data for each participant in the low-verbal

Sign Condition

0

10

20

30

40

50

60

70

Fre

qu

ency

of

Err

ors

Last Four Rounds in Each Condition

Jim

Selection-based

Fixed Condition

Selection-based

Random Condition

Selection-based

Same Condition

Fig. 1. The frequency of errors and mean lines for each of the last four rounds across three selection-based MTS con-ditions and one topography-based hand-sign condition for participant Jim from the low-verbal group.

CARL T. SUNDBERG et al.608

group are presented in Appendix C. Duringthe selection-based fixed-location condition,Jim met the criterion for all relations. How-ever, in the random location condition inwhich the symbols were in different locationson each trial, his mean frequency of errorsmore than doubled and he failed to meet thecriterion for any of the relations. Thus, Jimacquired conditional discriminations when thecomparison symbols were in the same locationon each trial, but not when in different loca-tions on each trial.Jim’s best performance of the three

selection-based conditions was in the same-symbol condition (tact disruption). These datademonstrate that symbol location rather thansymbol topography was the relevant source ofcontrol for Jim’s selection behavior. Jim’s bestperformance of all conditions was during thetopography-based sign condition. One possibleexplanation for Jim’s better performance withhand signs is that the topography-based tasksinvolved only simple discriminations, whereasall the selection-based tasks involved condi-tional discriminations.Figure 2 presents the mean frequency of

errors for the low-verbal group’s last fourrounds of each condition. The group errorpattern was similar to Jim’s data. For theselection-based tasks, the random-location con-dition was the most difficult for the partici-pants in the low-verbal group. The fixed-location and same-symbol conditions were

easier than the random- location condition.Three of the four participants failed to masterany of the relations in the random-locationcondition, but mastered all of the relations inthe fixed-location and same- symbol conditions(see Appendix C).

One participant, Molly, who demonstratedthe lowest verbal skills and worked with anarray of only two symbols, acquired a peculiarresponse pattern and failed to demonstratereliable conditional discriminations in any ofthe selection-based conditions. However, inthe topography-based hand-sign condition, sheperformed in a manner commensurate withthe other participants in the low-verbal group.The mean frequency of errors for thetopography-based hand-sign condition for allthe participants in the low-verbal group showsthat their best overall performance was in thehand-sign condition, whereas their worst per-formance was in the random-location condi-tion (Fig. 2).

High-verbal group. The frequency of errorsduring the last 10 rounds of each of the fourconditions is presented in Figure 3 for Julie, arepresentative participant from the high-verbalgroup. Data for all participants from the high-verbal group are presented in Appendix D. Acomparison of the selection-based fixed loca-tion and the selection-based random locationshows little difference between these two con-ditions. Thus, unlike Jim from the low-verbalgroup, Julie was not adversely affected by

0

10

20

30

40

50

60

70

Conditions

Selection-based

Fixed Condition

Selection-based

Random Condition

Selection-based

Same Condition

Sign

Condition

Low-Verbal

Mea

n F

req

uen

cy o

f E

rro

rs

Fig. 2. Mean frequency of errors for the low-verbal group for the last four rounds across the three selection-basedMTS conditions and one topography-based hand-sign condition.

609COVERT VERBAL MEDIATION

randomization of symbol location; in fact, sheperformed better in the random conditionthan the other MTS conditions.However, in the same-symbol condition in

which the possibility of tacting the symbols wasdisrupted by making all the symbols identical,Julie’s error rate more than tripled. These datademonstrate that topographically distinct

comparison stimuli can affect MTS perfor-mance (see also Slattery, Stewart, & O’Hora,2011). Julie’s lowest error rate was for thetopography-based hand-sign condition.

Figure 4 presents the high-verbal group’smean frequency of errors for the last fourrounds of each condition. The group’s errorpattern was similar to Julie’s data. For the

0

2

4

6

8

10

12

14

16

18

20

22

24

Fre

quen

cy o

f E

rrors

Last Ten Rounds in Each Condition

Selection-based

Fixed Condition

Selection-based

Random Condition

Selection-based

Same Condition

Sign

Condition

Julie

Fig. 3. The frequency of errors and mean lines for each of the last 10 rounds across three selection-based MTS condi-tions and one topography-based hand-sign condition for participant Julie from the high-verbal group.

0

2

4

6

8

10

12

14

Conditions

Selection-based

Same Condition

High-Verbal

Selection-based

Random Condition

Sign

Condition

Selection-based

Fixed Condition

Mea

n F

req

uen

cy o

f E

rro

rs

Fig. 4. Mean frequency of errors for the high-verbal group for the last 10 rounds across the three selection-basedMTS conditions and one topography-based hand sign condition.

CARL T. SUNDBERG et al.610

selection-based tasks, performance was similarduring the fixed- and random-location condi-tions. The highest number of errors for threeout of the four high-verbal participantsoccurred during the same-symbol condition(see Appendix D). One participant, Gina,demonstrated approximately the same perfor-mance for all the selection-based conditions(although in the exit interview she reportedthat the same-symbol condition was the hard-est). The data for three participants in thehigh-verbal group (Julie, Erin, and Amanda)suggest that the distinct patterns on each sym-bol were an important source of stimulus con-trol, regardless of whether the symbol’slocation was constant. The scores for the fourparticipants in the topography-based conditionwere approximately equal to their scores inthe selection-based fixed- and random-locationconditions, but better than their scores in theselection-based same-symbol condition.Comparisons between groups. The most

noteworthy difference between the two groupswas their opposite performance on the same-symbol and the random-location conditions.The high-verbal group was adversely affectedby the tact disruption in the same-symbol con-dition, but the low-verbal group was not. How-ever, the low-verbal group was adverselyaffected by the randomization of the locationof the symbols, whereas the high-verbal groupwas not. These data suggest that different con-tingencies were operating for high-verbal andlow-verbal groups in the selection-based condi-tions. Interestingly, there was little differencebetween the two groups in the topography-based hand-sign condition.

Exit interview and talk-aloud procedure.Each of the three available high-verbal partici-pants were asked a series of questionsdesigned to identify strategies and the extentof mediating verbal and nonverbal behaviorduring the experimental procedures (seeTable 4 for the questions). The participantswere then asked to talk aloud as they were pre-sented a sample block of trials from the study.

All three participants reported using thesame basic strategy during the fixed and ran-dom conditions. The participants created arecognizable name or action for each arbitrarysample word or symbol, and for each compari-son symbol. Then they altered the word to aphrase or something that would connect theinitial stimulus to a sign or comparison sym-bol. All three participants indicated that muchof the variance in errors between rounds was aresult of the extent to which they could comeup with names and good connections betweenthe stimuli. For example, when the sampleword was “biba” and the symbol was an emptysquare, the participant would convert the word“biba” to “bebop” and think of the square as adance floor. The resulting self-generated con-nection was “Bebop on a dance floor.” Allthree participants identified the same-symbolcondition as the hardest because they couldnot use their strategies to create names for thesymbols and come up with a story to connectthe sample stimulus with the comparison stim-ulus. The results of a mock session, in whichthe participants were asked to talk aloud asthey were presented with a block of stimuli,confirmed the strategies reported by the par-ticipants. The talk-aloud activity also con-firmed the lack of verbal strategies in thesame-symbol condition (e.g., “Bocam. Huh, Iwonder where that goes”).

Discussion

The present study found that disruptingmediating verbal behavior in an arbitrary MTSpreparation adversely affected performancefor high-verbal participants, but not for low-verbal participants. The participants from bothgroups acquired similar conditional discrimi-nations, but it is likely that different contin-gencies were operating for each group. Thedata suggest that high-verbal participantsdepended on mediating verbal behavior assupplementary sources of stimulus control for

Table 4

Protocol analysis questions

1. Which condition did you find the easiest?

2. Why do you think that condition was the easiest?

3. Did you use any general or specific strategies to findthe right symbol?

4. Why did you think there was such variation in scoreswithin a condition?

5. Which condition did you find the hardest?

6. What do you think made that condition difficult?

7. Did you have any trouble using your general orspecific strategies to find the right symbol?

8. Did you have any problem distinguishing visually oneposition from another on the board during the samecondition?

611COVERT VERBAL MEDIATION

more efficient MTS selection behavior. Low-verbal participants, however, presumably werenot affected by the verbal disruption proce-dure because their limited verbal abilities pre-cluded them from benefitting fromsupplementary verbal behavior during MTSacquisition. For these participants, it is morelikely that they acquired the conditional dis-criminations by unmediated contact with theexperimenter-manipulated contingencies, orby some form of nonverbal mediation.The results from the high-verbal group are

consistent with previous research demonstrat-ing that the response products of mediatingbehavior can provide supplemental sources ofstimulus control in MTS performance for rats(e.g., Laties et al., 1965), pigeons(e.g., Sidman, 1960), monkeys (e.g., Hodoset al., 1962), children (Parsons et al., 1981),and college students (e.g., Santos, Ma, &Miguel, 2015). In addition, the current studyreplicates and extends previous research onjoint control that showed disrupting mediatingverbal behavior impairs performance(e.g., Clough et al., 2016; DeGraaf & Schlin-ger, 2012; Gutierrez, 2006; Lowenkron, 2006;Sidener & Michael, 2006).The high-verbal participants brought an

extensive history of verbal and nonverbal reper-toires to the experimental sessions. As a resultof this history and the immediate contingen-cies, these participants emitted verbal behaviorduring the trials that produced supplementarySDs that occurred along with the programmedcontingencies, and putatively participated inevoking MTS selection behavior as a multiplycontrolled response (Blough, 1959; Michael,Palmer, & Sundberg, 2011). Figure 5 shows aninterpretation of how a high-verbal partici-pant’s verbal behavior and other sources ofstimulus control may interact during the earlystages of MTS acquisition. A mediated path toa selection response is presented in the upperpanel of Figure 5, and an unmediated path toselection is presented in the lower panel. Themediated path is based on Skinner’s (1957)analysis of verbal behavior and Lowenkron’s(1998) analysis of joint control. Following themediated path, a trial begins on the left side ofthe diagram with the presentation of the arbi-trary auditory sample stimulus (e.g., “Beeba”)that ultimately functions as a verbal SD (VSD1)with three separate behavioral effects (diver-gent multiple control).

As shown in Figure 5, two effects are evoca-tive, and one effect is function-altering. Theevocative effects are observed by an anteced-ent’s demonstration of an immediate differen-tial increase or decrease of a behavior. Oneevocative effect of the sample stimulus is dem-onstrated by an immediate overt or covertechoic verbal response (VR1). The participantsconfirmed this effect in their exit interviews.The second evocative effect of the samplestimulus is that it evokes an overt nonverbalscanning response (NVR1; i.e., looking for thecorrect comparison stimulus). The conse-quences for these two behaviors are likelyautomatic reinforcement related to hearingone’s own echoic response, finding the tar-geted item, or making progress toward solvingthe MTS problem (Vaughan & Michael,1982). The third effect of the initial verbal SD

is a function-altering effect that produceschanges in the function of other stimuli(Michael, 1995, 2004; Schlinger & Blakley,1987, 1994). This effect occurs following a fewrepeated contacts with the MTS reinforcementcontingencies in which the presentation of thesample stimulus (VSD1) alters one of the stim-uli in the comparison array to an SD functionand establishes the other stimuli as S-deltas.

All of the high-verbal participants reportedin the exit interviews that they initiallyrepeated the nonsense words (an echoic) andconverted them to something more recogniz-able and easier to distinguish from the others.Figure 5 shows this activity in which theresponse product of the participant’s echoicresponse may also function as an SD (VSD2)that evokes a self-echoic response (VR2). Theconversion behavior occurs when the echoicresponse product generates VSD3, whichevokes a self-intraverbal response (VR3). Forexample, upon hearing the sample stimulus“Beeba,” one participant reported that over afew trials she converted it to “Baby.” The resultof this conversion was an emergent self-generated intraverbal relation.

All of the participants also reported givingnames to the stimuli in the comparison arraysto make them easier to remember (thus gen-erating additional emergent verbal behaviorsand verbal SDs). For example, the symbol withframed lines (NVSD1) was given the label “win-dow” (VR4) by the three participants who wereavailable for the exit interview. These exitreports support Horne and Lowe’s (1996)

CARL T. SUNDBERG et al.612

observation that participants performing onMTS tasks often engage in overt or covert ver-bal behavior about the task and how to per-form on it. The current interpretationprovides some specifics about what the partici-pants might be talking about and how this talkmight affect selection behavior. Figure 5 showsthat the response product from this emergenttact generates a verbal SD (VSD6), and it thenparticipates in joint control with the intraver-bal relation described below.After a small number of exposures to the

direct contingencies, the participantsreported coming up with a phrase to connectthe two stimuli (termed “intraverbal naming”by Horne & Lowe, 1996). Figure 5 shows thecovert self-intraverbal response product of

VR3 (saying “baby”) generating VSD4 (hearingas a self-listener “baby”) and evoking a sec-ond intraverbal response, VR5 (saying to one-self “in the window”), with the resultingresponse product producing VSD5. The par-ticipants reported that they covertly repeatedthe phrase “baby in the window,” now as aself-echoic (VR6) producing VSD6 as theyscanned the comparison array. Upon encoun-tering the comparison stimulus with framedlines (NVSD1), the previously acquired andself-generated tact “window” is evoked (VR4),and the resulting response product generatesVSD6. When the covert tact and intraverbalboth contain the same primary response form(“window”), the instantaneous confluence ofthese two separate antecedent events

Fig. 5. A mediated path to a selection response is presented in the upper panel, and an unmediated path to selectionis presented in the lower panel. The mediated path begins on the upper left side of the diagram with the presentation ofthe arbitrary auditory sample stimulus (e.g., “Beeba”), and concludes with a selection response and reinforcement. Themediated path contains a sequence of verbal events that set up joint control that evokes a selection response. The unme-diated path begins with the same sample stimulus, but shows how the selection response could also be acquired by directcontact with the contingencies, which is likely the path taken by the low-verbal participants.

613COVERT VERBAL MEDIATION

constitutes the discriminable event (NVSD2)identified as joint control (Lowenkron,1998). This emerged nonverbal event is thesimilarity of the two response products andtacted (VR7) as such by the speaker andtagged on to primary tact (“window”) as anautoclitic tact of sameness in the form ofselection behavior (Lowenkron, 1998). Thereinforcement for these mediating behaviorsinvolves a history of both experimenter-manipulated direct reinforcement and puta-tive automatic reinforcement.Proponents of the RFT and stimulus equiva-

lence accounts often refute a verbal media-tional account to explain complex behavioron the basis of the short latency between thesample stimulus and the response (e.g., Saun-ders & Green, 1996). However, it is importantto distinguish between the current emission ofa target response and the variables involved inthe relevant learning history regarding thatresponse. Eventually, the supplemental stimuliare no longer necessary, and stimulus controlis transferred to the target verbal stimulus(i.e., the response is fluent, or memorized).Indeed, the participants reported in the exitinterviews that once they learned the connec-tion between two stimuli, they no longerneeded their strategies.When the procedure was changed to the

same-symbol condition, disrupting the tact ele-ment of joint control impaired performancefor high-verbal participants, but not for low-verbal participants. For high-verbal partici-pants when all 12 symbols in the comparisonarray looked exactly alike, tacting was difficult.Figure 6 shows that without a consistent tactfor a comparison stimulus (e.g., “Window”),the second type of covert intraverbal cannotoccur (e.g., “Baby in the window”). Althoughthe self-echoic (e.g., “Beeba”) and first type ofintraverbal (e.g., “Baby”) could still occur,without the primary tact, joint control, intra-verbal naming, and the secondary autoclitictact cannot occur, and no help is provided tothe high-verbal participant in finding the cor-rect comparison stimulus. At this point, theparticipants had to switch to the unmediatedpath that required more trials. This differenceis perhaps why the participants reported inthe exit interviews that the same-symbol condi-tion was the most difficult.Autoclitic verbal behavior is critical to joint

control, but an alternative interpretation of its

specific role is possible. Lowenkron (1998)suggested that the selection response be classi-fied as a descriptive autoclitic. While it seemsclear that autoclitic behavior is involved in thesequence of events, it is possible that theautoclitic response is not in the form of selec-tion behavior, but rather it is a critical step inthat direction. Consider that in the sequenceof events in a joint- control account, the com-parison stimulus must control two completelydifferent behaviors, occurring in rapid succes-sion. The exact same static stimulus (e.g., thewindow) first controls a tact, but then immedi-ately controls selection behavior, in that orderonly. The nonverbal stimulus cannot evokethe selection response first because withoutthe primary tact, the necessary SD generatedby its response product is not available to com-bine with the self-echoic (“Baby in the win-dow”) and set up joint control. Without jointcontrol, the descriptive autoclitic tact cannotoccur (e.g., “That’s it!”), both of which arenecessary to alter the functional effect of aspecific stimulus in the comparison array, suchthat it sets up the conditional discriminationthat ultimately evokes correct selectionbehavior.

Figure 7 shows this alternative interpreta-tion. When the self-echoic (VR6) and tact(VR4) conditions evoke the same responseform, thus producing the same responseproduct (VSD6), the confluence of these vari-ables generates the onset of joint control(NVSD2) and evokes secondary verbal behav-ior in the form of an autoclitic tact (VR7)(e.g., “That’s it,” similar to Sidman, Cresson, &Wilson-Morris’, 1974, reported “aha”response). The response product of theautoclitic tact produces VSD7. VS

D7 has both

evocative and function-altering effects;namely, it evokes (or maintains) scanningbehavior (NVR2) and alters the function ofthe window from a nonverbal SD evoking atact to a verbal SD that participates in evokingselection behavior. Thus, it is not until theverbal response product of the autoclitic tactoccurs (e.g., self-hearing of “That’s it”) alongwith the visual comparison stimulus, that aconditional discrimination is set up and selec-tion behavior is evoked by VSD9 as a selection-based intraverbal. This alternative accountaccommodates the rapid change in the evoca-tive effect of the nonverbal stimulus in thecomparison array.

CARL T. SUNDBERG et al.614

For the high-verbal participants, at least19 different putative emergent behavioral rela-tions can be identified as occurring at variousstages of acquisition between the presentationof the sample stimulus and the selection of acomparison stimulus (Fig. 5). None of theseantecedents, behaviors, or consequences werespecifically targeted by the experimenters’ pro-grammed contingencies. Seven types of verbalbehavior putatively emerged for the high-verbal participants during various stagesof acquisition: self-echoic1, self-intraverbal1,self-tact, self-intraverbal2, self-echoic2, self-autoclitic behavior, and an overt selection-based intraverbal response. In addition,11 types of stimulus control were identified asputatively emergent: eight new verbal SDs ema-nating from the response products of the puta-tive covert verbal behavior, two new nonverbal

SDs evoking the primary and autoclitic tacts,and most importantly, the multiple causationthat produced joint control and its resultingSD evoking an autoclitic tact of sameness.Finally, a new type of consequence putativelyemerged in the form of automatic reinforce-ment related to sameness (e.g., “That’s it”). Itis also possible that other types of automaticconsequences emerged during the experimen-tal activity, including automatic negative rein-forcement related to escape (e.g., vanHaaren, 2015).

Many of the emergent forms of verbalbehavior involved topography-based relationsevoked on the way to emitting a selection-based response. This effect provides supportfor the view that selection-based verbal behav-ior in a MTS preparation may be dependenton topography-based verbal behavior

Fig. 6. The same mediated and unmediated paths shown in Figure 5 are presented, but show that without a tact for aspecific comparison stimulus, joint control, intraverbal naming, and the secondary autoclitic tact cannot occur. Themediated path comes to a dead end and no help is provided to the participant. At this point, the participant has to relyon the unmediated path.

615COVERT VERBAL MEDIATION

(Lowenkron, 1991; Polson & Parsons, 2000;Sundberg & Sundberg, 1990; Wraikatet al. 1991). Without a specific response topog-raphy, nothing is available to mediate transferbetween variables in a selection-based task andparticipants must rely on the unmediated con-tingencies or some form of nonverbalmediation.The high-verbal participants and the low-

verbal participants performed similarly in thetopography-based hand-sign condition. Highverbal participants reported in the exit inter-views that this was the easiest condition. Inter-estingly, all of the low-verbal participantsperformed better in the topography-basedhand-sign condition than in the selection-based random condition, replicating theresults of Sundberg and Sundberg (1990) andWraikat et al. (1991). These findings have clin-ical implications for the use of sign languageversus selection-based icon systems for teach-ing language to nonvocal intellectually dis-abled persons.The results from the current study can also

provide some insight regarding the discrepancybetween those who easily demonstrate stimulusequivalence, relational framing, and derivedrelations, versus those who struggle or fail (seeJEAB’s January, 2014 special issue). In general,the lower the verbal and intellectual abilitiesdemonstrated by participants, the more diffi-cult it is to obtain emergent relations, althoughnot impossible, even with nonvocal humansand nonhumans (e.g., Lowenkron, 1984; McIn-tire, Cleary, & Thompson, 1987; Saunders &

Spradlin, 1989; Schusterman & Kastak, 1993;Sundberg & Sundberg, 1990; Swisher &Urcuioli, 2018; Zentall, Wasserman, & Urcuioli,2014). The current data suggest that it is a par-ticipant’s mediating verbal behavior that pro-vides valuable supplementary sources ofstimulus control that participate in evokingselection behavior and producing emergentrelations. These types of self-generated verbalstimuli are less available to low-verbal partici-pants and nonhumans, although even a smalldegree of verbal or nonverbal mediation maybe effective in generating emergent behavior.

A potential limitation of the current study isthat the primary behavior of interest is covertand cannot be directly measured. Rather, therole of covert verbal behavior can only beinferred by observing orderly effects in otherindirect measures (Palmer, 2011; Skinner,1953). Another limitation is that it would havebeen more informative to run the talk-aloudprocedure prior to the interview to guardagainst potential order effects. However, boththe reports in the exit interview and theresponses during the talk-aloud procedureswere consistent with the overt verbal behaviorexhibited by the participants during theexperiment.

Previously it was noted that there are fourdifferent conceptual explanations of emergentstimulus–stimulus relations. Lowenkron’s(1998) joint control account and Horne andLowe’s (1996) naming theory both maintainthat a participant’s mediating verbal behaviorscan function as causal variables and should be

Fig. 7. An alternative interpretation of the autoclitic shows joint control evoking a discriptive autoclitic tact of same-ness (“That’s it”). The response product of the autoclitic tact alters the function of the stimuli in the comparison array,establishing one stimulus as VSD8 that along with VSD7, participates in a conditional discrimination and selection behav-ior is evoked by VSD9 as a selection-based intraverbal.

CARL T. SUNDBERG et al.616

accounted for in the explanation of emergentrelations. Proponents of equivalence theoryand RFT tend to dismiss the relevance of aparticipant’s mediating verbal behavior as asource of stimulus control when examiningemergent relational behavior (e.g., Hayeset al., 2001; O’Hora, Barnes-Holmes, &Stewart, 2014; Quinones & Hayes, 2014;Saunders & Green, 1996; Sidman & Tailby,1982; Slattery, Stewart, & O’Hora, 2011). How-ever, the current data suggests that verbal par-ticipants emit verbal behavior, much of itcovert, during the trials. The response prod-ucts of these self-produced verbal responsescan function as verbal SDs that evoke otherverbal and nonverbal behaviors, including theterminal response. Without an account of thecausal role of a participant’s verbal mediation,we argue equivalence theory and RFT areincomplete explanations of MTS performanceand emergent relational behavior with high-verbal participants.What emerges in an arbitrary MTS prepara-

tion with high-verbal participants are newself- speaker and self-listener behaviors, alongwith joint control. These emergent behaviorsand their sources of control interact with theexperimenter-manipulated contingences andevoke selection behavior as a multiply con-trolled verbal response. Skinner’s (1957)analysis of verbal behavior along with Low-enkron’s (1998) analysis of joint control pro-vide a plausible and parsimonious account ofemergent stimulus–stimulus relations. Inaddition, a verbal behavior account allowsfor a more precise identification of thecausal variables that underlie emergentbehaviors, thereby engendering more effec-tive language assessment and interventionprograms for individuals experiencing lan-guage disorders.

References

Barlow, D. H., & Hayes, S. C. (1979). Alternating treat-ments design: One strategy for comparing the effectsof two treatments in a single subject. Journal of AppliedBehavior Analysis, 12, 199-210. https://doi.org/10.1901/jaba.1979.12-199

Blough, D. S. (1959). Delayed matching in the pigeon.Journal of the Experimental Analysis of Behavior, 2, 151-160. https://doi.org/10.1901/jeab.1959.2-151

Boren, J. J., & Devine, D. D. (1968). The repeated acquisi-tion of behavioral chains. Journal of the ExperimentalAnalysis of Behavior, 11, 651-660. https://doi.org/10.1901/jeab.1968.11-651

Clough, C. W., Meyer, C. S., & Miguel, C. F. (2016). Theeffects of blocking and joint control training onsequencing visual stimuli. The Analysis of VerbalBehavior, 32, 242-264. https://doi.org/10.1007/s40616-016-0067-1

DeGraaf, A., & Schlinger Jr., H. D. (2012). The effect ofjoint control training on the acquisition and durabil-ity of a sequencing task. The Analysis of Verbal Behavior,28, 59-71. https://doi.org/10.1007/BF03393107

Dougher, M., Twohig, M. P., & Madden, G. J. (2014). Edi-torial: Basic and translational research on stimulus–stimulus relations. Journal of the Experimental Analysis ofBehavior, 101, 1-9. https://doi.org/10.1002/jeab.69

Eckerman, D. A. (1970). Generalization and responsemediation of a conditional discrimination. Journal ofthe Experimental Analysis of Behavior, 13, 301-316.https://doi.org/10.1901/jeab.1970.13-301

Ericsson, K. A., & Simon, H. A. (1993). Protocol analysis: Ver-bal reports as data. Cambridge, MA: MIT Press.

Ferster, C. B., & Skinner, B. F. (1957). Schedules of reinforce-ment. New York, NY: Appleton-Century-Crofts.

Greer, R. D., & Longano, J. (2010). A rose by naming:How we may learn to do it. The Analysis of Verbal Behav-ior, 26, 73-106. https://doi.org/10.1007/BF03393085

Gutierrez, R. D. (2006). The role of rehearsal in joint con-trol. The Analysis of Verbal Behavior, 22, 183. https://doi.org/10.1007/BF03393038

Hayes, S. C., Barnes-Holmes, D., & Roche, B. (2001). Rela-tional frame theory: A post-Skinnerian account of humanlanguage and cognition. New York, NY: Plenum.

Hodos, W., Ross, G. S., & Brady, J. V. (1962). Complexresponse patterns during temporally spaced respond-ing. Journal of the Experimental Analysis of Behavior, 5,473-479. https://doi.org/10.1901/jeab.1962.5-473

Horne, P. J., & Lowe, C. F. (1996). On the origins of nam-ing and other symbolic behavior. Journal of the Experi-mental Analysis of Behavior, 65, 185-241. https://doi.org/10.1901/jeab.1996.65-185

Horne, P. J., Lowe, C. F., & Randle, V. R. (2004). Namingand categorization in young children: II. Listenerbehavior training. Journal of the Experimental Analysis ofBehavior, 81, 267-288. https://doi.org/10.1901/jeab.2004.81-267

Laties, V. G., Weiss, B., Clark, R. L., & Reynolds, M. D.(1965). Overt “mediating” behavior during temporallyspaced responding. Journal of the Experimental Analysisof Behavior, 8, 107-116. https://doi.org/10.1901/jeab.1965.8-107

Laties, V. G., Weiss, B., & Weiss, A. B. (1969). Furtherobservations on overt “mediating” behavior and thediscrimination of time. Journal of the Experimental Anal-ysis of Behavior, 12, 43-57. https://doi.org/10.1901/jeab.1969.12-43

Lowenkron, B. (1984). Coding responses and the generali-zation of matching to sample in children. Journal ofthe Experimental Analysis of Behavior, 42, 1-18. https://doi.org/10.1901/jeab.1984.42-1

Lowenkron, B. (1988). Generalization of delayed identitymatching in retarded children. Journal of the Experi-mental Analysis of Behavior, 50, 163-172. https://doi.org/10.1901/jeab.1988.50-163

Lowenkron, B. (1989). Instructional control of generalizedrelational matching to sample in children. Journal ofthe Experimental Analysis of Behavior, 52, 293-309.https://doi.org/10.1901/jeab.1989.52-293

617COVERT VERBAL MEDIATION

Lowenkron, B. (1991). Joint control and the generaliza-tion of selection-based verbal behavior. The Analysis ofVerbal Behavior, 9, 121-126. https://doi.org/10.1007/BF03392866

Lowenkron, B. (1998). Some logical functions of joint con-trol. Journal of the Experimental Analysis of Behavior, 69,327-354. https://doi.org/10.1901/jeab.1998.69-327

Lowenkron, B. (2006). Joint control and the selection ofstimuli from their description. The Analysis of VerbalBehavior, 22, 129-151. https://doi.org/10.1007/BF03093035

McIntire, K. D., Cleary, J., & Thompson, T. (1987). Condi-tional relations by monkeys: Reflexivity, symmetry,and transitivity. Journal of the Experimental Analysis ofBehavior, 47, 279-285. https://doi.org/10.1901/jeab.1987.47-279

Michael, J. (1985). Two kinds of verbal behavior plus apossible third. The Analysis of Verbal Behavior, 3, 1-4.https://doi.org/10.1007/BF03392802

Michael, J. (1995). What every student in behavior analysisought know: A system for classifying the multipleeffects of behavioral variables. The Behavior Analyst, 18,272-284. https://doi.org/10.1007/BF03392714

Michael, J. (2004). Concepts and principles of behavior analysis(2nd ed.). Kalamazoo, MI: Association for BehaviorAnalysis International.

Michael, J., Palmer, D. C., & Sundberg, M. L. (2011). Themultiple control of verbal behavior. The Analysis of Ver-bal Behavior, 27, 3-22. https://doi.org/10.1007/BF03393089

Miguel, C. F., Petursdottir, A. I., Carr, J. E., & Michael, J.(2008). The role of naming in stimulus categorizationby preschool children. Journal of the Experimental Analy-sis of Behavior, 89, 383-405. https://doi.org/10.1901/jeab.2008-89-383

O’Hora, D., Barnes-Holmes, D., & Stewart, I. (2014). Ante-cedent and consequential control of derivedinstruction-following. Journal of the Experimental Analy-sis of Behavior, 102, 66-85. https://doi.org/10.1002/jeab.95

Palmer, D. C. (2011). Consideration of private events isrequired in a comprehensive science of behavior. TheBehavior Analyst, 34, 201-207. https://doi.org/10.1007/BF03392250

Parsons, J. A., & Ferraro, D. P. (1977). Complex interac-tions: A functional approach. In: B. Etzel, S.W. Bijou,J.M. LeBlanc, & D.M. Baer (Eds.), New developments inbehavioral research: Theory, method, and application(pp. 237-245). Hillsdale, NJ: Lawrence ErlbaumAssociates.

Parsons, J. A., Taylor, D. C., & Joyce, T. M. (1981). Precur-rent self-prompting operants in children: “Remem-bering.” Journal of the Experimental Analysis of Behavior,36, 253-266. https://doi.org/10.1901/jeab.1981.36-253

Polson, D. A., & Parsons, J. A. (1994). Precurrent contin-gencies: Behavior reinforced by altering reinforce-ment probability for other behavior. Journal of theExperimental Analysis of Behavior, 61, 427-439. https://doi.org/10.1901/jeab.1994.61-427

Polson, D. A., & Parsons, J. A. (2000). Selection-based ver-sus topography-based responding: An important dis-tinction for stimulus equivalence. The Analysis of VerbalBehavior, 17, 105-128. https://doi.org/10.1007/BF03392959

Potter, B., Huber, S., & Michael, J. (1997). The role ofmediating verbal behavior in selection-based respond-ing. The Analysis of Verbal Behavior, 14, 41. https://doi.org/10.1007/BF03392915

Quinones, J. L., & Hayes, S. C. (2014). Relational coher-ence in ambiguous and unambiguous relational net-works. Journal of the Experimental Analysis of Behavior,101, 76-93. https://doi.org/10.1002/jeab.67

Randell, T., & Remington, B. (1999). Equivalence rela-tions between visual stimuli: The functional role ofnaming. Journal of the Experimental Analysis of Behavior,71, 395-415. https://doi.org/10.1901/jeab.1999.71-395

Santos, P. M., Ma, M. L., & Miguel, C. F. (2015). Trainingintraverbal naming to establish matching-to-sampleperformances. The Analysis of Verbal Behavior, 31, 162-182. https://doi.org/10.1007/s40616-015-0040-4

Saunders, K. J., & Spradlin, J. E. (1989). Conditional dis-crimination in mentally retarded adults: The effect oftraining the component simple discriminations. Jour-nal of the Experimental Analysis of Behavior, 52, 1-12.https://doi.org/10.1901/jeab.1989.52-1

Saunders, R. R., & Green, G. (1996). Naming is not (nec-essary for) stimulus equivalence. Journal of the Experi-mental Analysis of Behavior, 65, 312-314. https://doi.org/10.1901/jeab.1996.65-312

Schlinger, H. D., & Blakely, E. (1987). Function-alteringeffects of contingency-specifying stimuli. The BehaviorAnalyst, 10, 41-45. https://doi.org/10.1007/BF03392405

Schlinger, H. D., & Blakely, E. (1994). A descriptive taxon-omy of environmental operations and its implicationsfor behavior analysis. The Behavior Analyst, 17, 43-57.https://doi.org/10.1007/BF03392652

Schoenfeld, W. N., & Cumming, W. W. (1963). Perceptionand behavior. Psychology: A Study of a Science, 5,213-252.

Schusterman, R. J., & Kastak, D. (1993). A California sealion (Zalophus californianus) is capable of formingequivalence relations. The Psychological Record, 43, 823-839. https://doi.org/10.1007/BF03395915

Shimp, C. P., & Moffitt, M. (1977). Short-term memory inthe pigeon: Delayed-pair-comparison procedures andsome results. Journal of the Experimental Analysis ofBehavior, 28, 13-25. https://doi.org/10.1901/jeab.1977.28-13

Sidener, D. W., & Michael, J. (2006). Generalization ofrelational matching to sample in children: A directreplication. The Analysis of Verbal Behavior, 22, 171-181.https://doi.org/10.1007/BF03393037

Sidman, M. (1960). Tactics of scientific research. New York,NY: Basic Books, Inc.

Sidman, M. (1971). Reading and auditory-visual equiva-lences. Journal of Speech and Hearing Research, 14, 5-13.https://doi.org/10.1044/jshr.1401.05

Sidman, M. (1994). Equivalence relations and behavior: Aresearch story. Boston, MA: Authors Cooperative.

Sidman, M., Cresson, O., Jr., & Wilson-Morris, M. (1974).Acquisition of matching to sample via mediated trans-fer. Journal of the Experimental Analysis of Behavior, 22,261-273. https://doi.org/10.1901/jeab.1974.22-261

Sidman, M., & Tailby, W. (1982). Conditional discrimina-tion vs. matching to sample: An expansion of the test-ing paradigm. Journal of the Experimental Analysis of

CARL T. SUNDBERG et al.618

Behavior, 37, 5-22. https://doi.org/10.1901/jeab.1982.37-5

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

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

Skinner, B. F. (1974). About behaviorism. New York: Knopf.Slattery, B., Stewart, I., & O’Hora, D. (2011). Testing for

transitive class containment as a feature of hierarchi-cal classification. Journal of the Experimental Analysis ofBehavior, 96, 243-260. https://doi.org/10.1901/jeab.2011.96-243

Stewart, I., McElwee, J., & Ming, S. (2013). Language gen-erativity, response generalization, and derived rela-tional responding. The Analysis of Verbal Behavior, 29,137-155. https://doi.org/10.1007/BF03393131

Stromer, R. (1996). On the experimental analysis of nam-ing and the formation of stimulus classes. Journal ofthe Experimental Analysis of Behavior, 65, 250-252.https://doi.org/10.1901/jeab.1996.65-250

Sundberg, C. T., & Sundberg, M. L. (1990). Comparingtopography-based verbal behavior with stimulusselection-based verbal behavior. The Analysis of VerbalBehavior, 8, 31-41. https://doi.org/10.1007/BF03392845

Sundberg, M. L. (2014). Verbal behavior milestones assessmentand placement program: The VB-MAPP. Concord, CA:AVB Press.

Swisher, M., & Urcuioli, P. J. (2018). Reflexivity withoutidentity matching training: A first demonstration.

Journal of the Experimental Analysis of Behavior, 109, 125-147. https://doi.org/10.1002/jeab.302

van Haaren, F. (2015). Automatic negative reinforcement:Its possible role in problem behavior and treatmentimplications. Behavior Analysis: Research and Practice,15, 161-170. https://doi.org/10.1037/bar0000020

Vaughan, M. E., & Michael, J. L. (1982). Automatic rein-forcement: An important but ignored concept. Behav-iorism, 10, 217-227.

Wilson, M. P., & Keller, F. S. (1953). On the selective rein-forcement of spaced responses. Journal of Comparativeand Physiological Psychology, 46, 190-201. https://doi.org/10.1037/h0057705

Wraikat, R., Sundberg, C. T., & Michael, J. (1991). Topog-raphy-based and selection-based verbal behavior: Afurther comparison. The Analysis of Verbal Behavior, 9,1-17. https://doi.org/10.1007/BF03392856

Wulfert, E., Dougher, M. J., & Greenway, D. E. (1991).Protocol analysis of the correspondence of verbalbehavior and equivalence class formation. Journal ofthe Experimental Analysis of Behavior, 56, 489-504.https://doi.org/10.1901/jeab.1991.56-489

Zentall, T. R., Wasserman, E. A., & Urcuioli, P. J. (2014).Associative concept learning in animals. Journal of theExperimental Analysis of Behavior, 101, 130-151. https://doi.org/10.1002/jeab.55

Received: January 27, 2017Final Acceptance: April 4, 2018

619COVERT VERBAL MEDIATION

Appendix A: Examples of symbol boards used with low-verbal and high-verbal participants.

Appendix B: Words and signs used with low-verbal and high-verbal participants

Word lists for the two low-verbal group participants in the word–sign and word–symbolconditions.

Debra Eric

po po mac pig sorook slo jack jig doeflo blo pac rigclo crow zook migsac tro rac flobo mook dro sook

CARL T. SUNDBERG et al.620

Appendix C: Errors for each of the last four rounds per phase for low-verbal group

Word lists for the two low-verbal group participants in the word–sign and word–symbolconditions.

mo jim rook jookzig sac zoo sike

Signs for participants in the low-verbal group.

Molly Jim Debra Ericpat head withright hand

pat head with right hand pat head with right hand pat head with right hand

touch nose withright finger

touch nose with right finger touch nose with right finger touch nose with rightfinger

open one palm forward slap table with right hand open one palm forwardlift right elbow (90� from body) slap table with right hand

lift right elbow (90� frombody)

put right fist into left palm

Final word list for participants in the high-verbal group.

paba kaba taba saba baba daba habalaba maba naba buba cuba duba fubabeeba keeba deeba heeba bihba kihba dihbafihba bayba cayba dayba fayba jeeba leebameeba neeba hayba jayba layba mayba bacadaca haca jacaa naca faca laca macapayba nayba rayba sayba hafa peeba queebareeba paca raca quaca soca baida caidadaida haida seeba teeba veeba weeba tacawaca yaca vaca huba juba muba nubajaida laida maida naiad luba quba pubaruba hihba jihba lihba mihba paida quaidaraida saida bafa cafa dafa nihbapihda rihda taida quihba yaida zaida

Signs for participants in the high-verbal group.

slide right hand down left arm from elbow to handraise left arm and pointtouch nose with right fingeropen one palm forwardpet imaginary animalput right fist into left palmslap table with right handlife right elbow (90� from body)hook index and middle fingers of both hands to each othertap underneath chin with right handcircle index fingers of both hands around each othercover mouth with right hand

Participant

Phase Molly Jim Debra Eric Grand Mean

Fixed 16 16 9 1169 68 14 274 10 35 5569 11 25 59x = 39.5 x = 26.1 x = 20.8 x=38 x = 31.1

Random 12 64 95 6468 39 63 69

621COVERT VERBAL MEDIATION

Appendix D: Errors for each of the last 10 rounds per phase for high-verbal group

Participant

Phase Molly Jim Debra Eric Grand Mean

15 60 72 7562 57 67 76x = 39.3 x = 55 x = 74.3 x=71 x=59.9

Same 64 12 20 1064 17 36 4259 16 32 374 23 14 34x = 65.3 x = 17 x = 25.6 x=22.3 x=32.5

Sign 18 291 5425 335 2x = 12.3 x = 29.5 x=20.9

Sign 18 5 30 8858 13 51 4513 7 8 2261 11 33 25x = 37.5 x = 9 x = 30.5 x = 45 x=30.5

Participants

Phase Amanda* Gina Julie Erin Grand Mean

Fixed Last 10 Prior 10 Last 10 Last 10 Last 10 Amanda’s Last 1013 7 13 8 222 19 6 3 1017 10 9 1 621 16 8 3 5 x=9.522 7 1 3 5 Amanda’s Prior 1039 4 8 4 04 14 1 2 433 9 37 1 5 x = 7.511 12 8 5 27 13 13 4 12x = 18.9 x = 11.1 x = 10.4 x = 3.4 x=5.1

Random 12 14 2 1319 8 2 56 21 0 616 5 6 08 8 1 73 3 1 07 16** 1 22 2 17 6 57 1 4x = 8.7 x = 10.7 x = 2.2 x=4.3 x=6.1

Same 10 13 14 3414 13 5 142 4 1 1722 15 23 1817 19 7 236 6 2 1715 6 4 117 7 3 1317 11 12 1417 1 4 19x = 12.7 x = 9.5 x = 7.5 x=18 x=11.9

CARL T. SUNDBERG et al.622

Participants

Phase Amanda* Gina Julie Erin Grand Mean

Sign 30 8 0 5 x=710 6 3 713 5 2 321 3 2 139 12 0 512 9 0 13 8 1 106 11 5 196 2 1 37 9 0 11x = 11.7 x = 7.3 x = 1.4 x = 7.7

*The first cell includes the errors from participant Amanda’s last 10 rounds and from the previous10 rounds of the fixed phase. Both group means are included for this phase.**Only seven rounds were con-ducted for participant Gina during the random phase.

623COVERT VERBAL MEDIATION