Audience Control over Children’s Honest Reports

Abstract

We evaluated audience control over children’s honest reports using a reversal (ABA or ABAB) design. Four typically developing children performed a computer game in which they had to shoot a target and then report on their performance during and at the end of each session. Baseline assessed the accuracy of their reports in the absence of an experimenter. During the audience condition, an adult was present in the room and observed the child during the task. Participants accurately reported their errors when an adult was present, whereas they lied about their performance by systematically reporting errors as correct responses when an adult was absent. Honest reports about their total score at the end of the session also increased in the presence of the audience member. These results suggest that the presence of an adult exerted control over children’s honest/accurate reports. We discussed the reasons why the presence of an adult may have served as a discriminative stimulus for honest reports.

The positive relation between past and present behavior, whether verbal or not, has been labeled correspondence (da Silva & Lattal, 2010; ). We refer to say-do correspondence when individuals do what they said they would (Israel & O'Leary, 1973; Risley & Hart, 1968) and to do-say correspondence (da Silva & Lattal, 2010; Lloyd, 2002; Ribeiro, 1989) when they say or accurately report what they have done. These two distinct, yet related areas of research have been of interest to behavior analysts for quite some time, as they have implications for understanding socially relevant phenomena such as honesty, reliability, compliance, and truthfulness (e.g., Baer et al., 1984; Doepke et al., 2003; Lloyd, 2002; Paniagua, 1989, 1990; Sauter et al., 2020; Sparling et al., 2011; Shillingsburg et al., 2017, 2019; Stocco et al., 2021).

When it comes to do-say correspondence, caregivers expect their children to report their past behavior accurately or honestly when asked about their performance at school, how they got hurt, whether some strange person tried to approach them in a public area, and so on (Shillingsburg et al., 2019). Honest verbal reports are also important in police investigations and court proceedings which may involve children’s eyewitness testimony (Doepke et al., 2003; Sparling et al., 2011). Caregivers rank honesty as the most desirable quality for their children (Alwin, 1989) and consider lying a problem behavior (Paniagua, 1989; Stouthamer-Loeber, 1986). As such, behavioral investigations on the variables that lead to honest reports are warranted.

Research on do-say correspondence has focused on the specific conditions under which self-reports about what a participant has done are either accurate (i.e., honest) or inaccurate (i.e., dishonest; Antunes & Medeiros, 2016; Balog et al., 2019; Brino & de Rose, 2006; Cortez et al., 2013, 2014, 2017, 2019; Critchfield & Perone, 1990, 1993; Domeniconi et al., 2014; Mazzoca & Cortez, 2020; Oliveira et al., 2016; Ribeiro, 1989; Sauter et al., 2020). In a seminal study, Ribeiro (1989) evaluated self-reports about recent past behavior of eight preschool children. Participants were first invited to play with some toys. After playtime, a second experimenter held up a picture of a toy and asked, “Did you play with the [toy name]?” During baseline, when no specific verbal report was reinforced, participants showed nearly perfect correspondence by saying “yes” or “no” to the toy with which they had played or not, respectively. However, when reinforcement (praise and tokens) was made contingent upon reports about having played with certain items, participants started to say “yes” when presented with pictures of toys with which they had not played during the session. Subsequently, when reinforcers were made contingent upon the correspondence between their nonverbal (play) and verbal (report) behaviors (i.e., correspondence training), accurate reports increased again. Thus, it seems that during baseline, when self-reports were “honest,” they were primarily functioning as tacts under control of the response products of participants’ previous behavior (i.e., nonverbal stimuli). On the other hand, when reinforcement was made contingent upon specific verbal reports, participants verbal behavior came under the influence of motivating operations, functioning primarily as mands (Ribeiro, 1989; Skinner, 1957).

Another source of control for participants’ self-reports is the presence of listeners who mediate reinforcement for verbal responses. Given their correlation with differential availability of reinforcement for verbal behavior, listeners acquire discriminative control over behavior. Skinner referred to this type of antecedent control over verbal behavior as audience control (Skinner, 1957). When listeners punish verbal responses, their presence may acquire negative audience control, in that their presence may abate verbal behavior (e.g., the speaker may emit fewer, or no responses). Negative audiences may also evoke avoidance or escape behaviors, as is the case of saying something different from what has been previously punished, or lying (Fonai & Serio, 2007; Skinner, 1957). When discussing how verbal behavior is often a product of a multitude of variables, Skinner (1957) refers to algebraic summation in which positive and negative audiences may both affect verbal behavior at the same time. This form of stimulus (i.e., convergent) control may lead to response forms that are more likely to be reinforced and less likely to be punished by the audience. For instance, when referring to an individual who died, the tact “passed away” may be more likely to be evoked than the tact “died” when in the presence of a relative of the deceased. This is because the presence of the relative may function as a negative variable that abates any verbal response that may be offensive or unpleasant.

Despite the importance of audience variables in increasing the frequency, as well as selecting the specific topography of verbal responses, very few studies have been conducted on this topic (Brino & de Rose, 2006; Cortez et al., 2019; Dixon et al., 2019; Silverman et al., 1986; Singer-Dudek et al., 2020; Stocco et al., 2014). Stocco et al. (2014), for example, compared the effects of constant versus varied reinforcement schedules across audience members to evaluate audience control over college students’ reports about private events. During one of their conditions (i.e., constant), both experimenters delivered points contingent on participants saying a pseudoword when a specific visual stimulus was present. In another condition (i.e., varied), one experimenter delivered points for reporting a pseudoword in the presence of a visual stimulus, and a second experimenter delivered points for saying a different pseudoword in the presence of the same visual stimulus. Results showed that all participants’ reports came under audience control in the varied condition, but reports in the constant condition were similar across experimenters.

Considering the literature on the effects of an audience when reporting past events, Brino and de Rose (2006) asked four children between the ages of 7 and 11 with a history of school failure to report their recent performance on a computer-based academic task. Participants read words aloud and, after hearing the correct response, reported whether they had read correctly or incorrectly by clicking on a green or red square, respectively. The selection-based responses of clicking the green and red squares served the same function as topography-based responses such as, “I got it right” or “I got it wrong,” respectively (Michael, 1985). During Condition A, participants remained alone in the experimental room while performing the read-report task. At the end of each session, they had access to a computer game for a period equal to the duration of the session. In that condition, participants always reported their performance accurately (i.e., honestly) when they read correctly. However, participants almost always reported their performance incorrectly (i.e., lied) when were told by the computer that they had read incorrectly (i.e., they consistently clicked on the green square instead of the red square). Condition B assessed the effects of having the experimenter present during the task, whereas Condition C evaluated the effects of differentially reinforcing reports contingent upon their correspondence with the previous actions in the absence of the experimenter. During this condition, the amount of time the child had access to a computer game varied according to the number of accurate reports of errors made during the session. The experimenter added 1 min of game access for each honest report of errors made by the child during the session. Results showed that both conditions B and C were effective in increasing honest reports of errors. According to the authors, it is possible that during the condition in which the experimenter was absent, and reinforcement (game access) was noncontingent, participants rarely reported having made an error because of their history of punishment for making errors. In this case, lying served a negative reinforcement function (i.e., avoided punishment). When the experimenter was present, dishonest reports decreased and honest reports increased, which was also likely due to a preexperimental history of reinforcement and punishment for these types of reports.

Although Brino and de Rose (2006) showed an audience effect over children’s honest reports, their results should be interpreted with caution. First, task difficulty was never manipulated, which led to a disproportionate number of errors, and in turn, differences in opportunities to report errors across conditions. For example, one participant (Jose) made 24 errors in condition A and 89 errors in condition B. The higher number of errors in one of the conditions could have led to an increase in the number of dishonest reports (e.g., Critchfield & Perone, 1990; Critchfield & Perone, 1993; Domeniconi et al., 2014). Moreover, the order and the number of conditions varied across participants, while some conditions were never replicated within participants (e.g., one participant was exposed to conditions A, B, and C, whereas another was only exposed to condition A). Finally, the effects of the audience over self-reports were confounded by other variables such as differential reinforcement for corresponding reports.

In a follow-up study, Cortez et al. (2019) used the same task to evaluate the effects of three different audiences (computer, experimenter, and child) on the accuracy of self-reports by six participants between 7 and 11 years old. During baseline, participants remained alone in the experimental room while performing the read-report task. At the end of each session, they had access to a gift contingent upon their participation. Differently from Brino and de Rose (2006), all participants accurately reported their past behavior (reading performance) across trials; that is, they all emitted honest reports even when they read incorrectly. During the audience conditions, besides reporting their performance within each trial by clicking either the green or red squares after completing each task (i.e., trial), participants also had to report their total score to the audience at the end of each session. The experimenter used the number of honest reports of correct readings emitted during the session to establish the participant’s total score. At the end of each session, the computer displayed the total score earned during each session for approximately 10 s, after which one of the audiences (computer, experimenter or peer) asked, “How many points did you score?” However, the number of points required to have access to the preferred item was manipulated by the experimenter in a way that participants could only access them by reporting a higher score than the one obtained. If participants reported an equal (i.e., corresponding) or a lower score than the one actually obtained, they had access to an item of medium preference. During the computer condition, participants had to report (say) the number of points they earned during the session to the computer. During the experimenter condition, participants had to report their final score to an adult who entered the room. Finally, during the peer condition, participants had to report their final score to a peer. No consequences were delivered contingent upon accurate or inaccurate reports of errors during any of these conditions which were presented in a fixed order (i.e., computer, experimenter, and peer). Results showed more honest (accurate) reports to the computer, followed by the experimenter, and finally to another child, suggesting that participants’ frequency of honest reports varied as a function of the type of audience. However, these results could have been confounded by possible sequence effects, and there were no replications across conditions.

Despite evidence suggesting the effects of specific audiences over honest reports, the studies mentioned above (Brino & de Rose, 2006; Cortez et al., 2019) lacked adequate experimental control. Moreover, the fact that these studies were published in Portuguese may limit access to those who do not speak that language. For this reason, the purpose of the current study was to extend this line of research by evaluating the effects of a specific audience (adults) on children’s honest reports of correct and incorrect responses, using a reversal design, while also attempting to control for the number of errors made during the experimental task and the order of conditions.

Method

Participants, Setting, and Materials

Three Brazilian boys, Fausto, Julio, and Diego, and one girl, Rita, between 7 and 9 years old participated in the study. They had no known disabilities and attended the same school. They were recruited from a research program located at the Universidade Federal de São Carlos aimed at increasing reading and writing skills. Although participants did not have a formal diagnosis, they had all been referred to this program because of some deficits in reading and writing skills. All recruitment and experimental procedures were approved by the university’s Institutional Review Board. Across all participants and conditions, a male graduate student served as the audience.

Experimental sessions were conducted in a 3 x 3 m room containing a desktop computer, a microphone, speakers, and a desk with three colored cardboard boxes that contained gifts (toys and school supplies). The experimenter was located in an adjacent room where he could observe participants through a one-way mirror, as well as through a computer monitor that mirrored the one being used by the participant in the experimental room. The experimenter remained in the observation room with the lights off throughout all experimental conditions. Sessions were conducted three days a week and lasted 3 to 5 min each, with one to two sessions per day, depending on participants’ availability.

The current study used a shooting game developed by Cortez et al. (2014) to evaluate the correspondence between behavior (doing) and reporting (saying), available in two versions: shooting ducks or spacecrafts. All participants were exposed to both tasks during the study; however, the number of exposures to each version varied according to participants’ preference. In other words, to maintain motivation, participants were allowed to choose either game during a session. The computer game automatically recorded participants’ shooting performances (i.e., a hit or a miss) and their self-reports, which consisted of clicking either a green or a red square to report whether they had or had not shot the target, respectively, during each trial (selection-based response). The software also allowed the experimenter to manipulate task difficulty in an attempt to control for the number of errors made by the participants. Such manipulation was made by increasing or decreasing the number of obstacles, and the distance between the target and the shooter, as well as the number, size, and speed of targets. The experimenter manipulated task difficulty between sessions based on participants’ shooting performance in the previous session as an attempt to equate the number of errors. At the end of the session, a counter located on the center of the computer screen was used to display the total score obtained by the child for 10s, after which it disappeared. The points were contingent upon performance in the game and not on self-report accuracy. A smartphone was also used to record participants’ vocal reports of the total number of points obtained in each session. Participants received small toys and school supplies at the end of each session, regardless of their performance.

Response Definition and Experimental Design

Dependent variables included honest reports of misses and hits during the computer game. Honest reports were defined as clicking the red square after having missed the target or clicking the green square after hitting the target. Dishonest reports were defined as clicking the green square after having missed the target or clicking the red square after hitting the target. A second measure of honesty included participants’ vocal reports (final report) about the total number of points earned (total score) at the end of each session. A final report was defined as dishonest when participants reported a total score that differed from the one displayed on the computer (i.e., the total score actually achieved), and the final report was defined as honest when participants reported the same number of points displayed by the computer.

The experiment employed a reversal design (Kazdin, 2011) with two conditions. During baseline/control, participants completed the task (i.e., computer game) alone; during the audience condition, participants completed the task in the presence of an adult.

General Procedure

During all experimental conditions, each trial consisted of a shoot-report sequence. Participants shot a target by horizontally moving the mouse to direct the shooter and then by pressing the mouse’s left button to fire. Hitting the target produced auditory and visual feedback in the form of a duck falling or a space monster exploding. Missing the target produced a different sound. In each trial, participants had to shoot only one target. After every single shot and its corresponding auditory and visual feedback, the participant had to report their performance. The following recorded instruction was provided: “Did you hit it? If you did, click the green square; if not, click the red square,” while the text “Did you hit it?” appeared in the upper center of the screen, with the green (yes) and red (no) squares right below it. The position of the squares (right or left) randomly varied across trials. Participants reported their performances by clicking either the green (yes) or red (no) squares. Figure 1 depicts the trial (shoot-report) sequence for each version of the game. At the end of each session, the computer displayed the total score earned during the session for approximately 10 s, after which the experimenter entered the room and asked, “How many points did you score?” As the total score was displayed on the screen for only 10 s, there was no visual cue that participants could rely on when reporting their score to the experimenter. Regardless of the accuracy of participants’ vocal reports, the experimenter responded with a neutral statement (“ok”) and allowed participants to choose a small toy or a school supply which they could keep. Preference was assessed at the beginning of the study by directly asking participants which toys or school supplies they would want to receive as a gift for their participation.

Fig. 1.

Example of computer game versions and self-report screen in each trial. Note. Different game versions are represented, with shooting ducks in the top left and shooting spacecrafts in the top right. The self-report screen presented after each trial is represented by the screen in center of the second row

Baseline and audience sessions consisted of 10 easy and 10 difficult trials, to increase the likelihood that the participants would hit the target at least 10 times, earning a total of 10 points at the end of the session (e.g., Cortez et al., 2014). The experimenter manipulated the level of difficulty between sessions by increasing or decreasing the following parameters: number and speed of the targets, number of obstacles, distance between the shooter and the target, and target’s acceleration (constant or non-constant). For example, easy trials were usually programmed with up to four targets simultaneously presented in low speed, using a short distance between the shooter and the targets, constant acceleration, and a low number of obstacles (one or two). On the order hand, difficult trials presented only one or two targets simultaneously, in high speed and non-constant acceleration. The distance between the shooter and the target and the number of obstacles were also increased. Although task difficulty was manipulated this way to attempt to equate the number of errors across sessions, the actual number of errors varied as participants could have hit the target during a difficult trial and missed the target during an easy trial. Despite this variability, the number of errors did not vary significantly across sessions (see Fig. 2), suggesting that this manipulation was successful in controlling for task difficulty.

Fig. 2.

Percentage of target shooting errors, honest reports of correct responses and errors during baseline and audience conditions for all participants. Note. The numbers under each chart indicate the total score (TS) actually earned at the end of each session and participants’ final report (FR) about the total score. Asterisk indicates the session in which Rita saw the experimenter through the one-way mirror

Experimental Conditions

Pretraining

The purpose of this condition was to teach participants to play the computer game and the shoot-report sequence. Pretraining was kept brief (i.e., the experimenter conducted only one 10-trial session) to prevent a long history of reinforcement for reporting their performance, as has been done in previous studies in this area (e.g., Cortez et al., 2013, 2014, 2017, 2019; Domeniconi et al., 2014; Mazzoca & Cortez, 2020; Oliveira et al., 2016). Before the first trial, the experimenter said:

Now you will learn how to play the game. You will have to move the mouse to direct the shooter and then you will have to press the left mouse button to fire. Pay attention to the instructions given by the computer. I will stay here in case you need any assistance.

Trials were like those presented during the baseline and audience conditions. During the two or three initial trials, depending on the participants’ responses, the experimenter delivered social consequences (e.g., “that is it,” “you got it”) every time a report was honest (i.e., corresponded with their performance) to ensure the task was understood (do-say sequence). If participants looked at the experimenter and asked a question, the experimenter instructed the participant to listen carefully to the computer’s message and then delivered social consequences following the first two to three trials in which the participant gave an honest report (a maximum of two social consequences were delivered contingent on reports for each type of trial – hitting or missing the target). If participants emitted dishonest reports during pretraining, the experimenter implemented a correction procedure. For example, if participants missed a target and reported their performance as a hit (clicking in the green/yes square), the experimenter blocked access to the mouse and asked participants, “Why did you click the green square? Did you hit the target?” After the participants’ answers (e.g., “no, I did not”), the experimenter then reminded participants of the instruction presented by the computer (i.e., “Did you hit it? If you did, click on the green square; if not, click on the red square,”) and asked which square (red or green), they should have clicked on. Following this correction procedure, the experimenter allowed participants to continue playing the game. All participants reported their correct and incorrect responses honestly during pretraining trials, so the correction procedure was not implemented in the present study.

Baseline

During this condition, participants remained alone in the experimental room while completing the shoot-report sequence (i.e., computer game task). At the end of each session, the experimenter entered the room and asked, “Now that you have finished the game, tell me: how many points did you score?” No consequences followed participants’ reports, although they could choose an item as discussed above. Stability criterion was set at a maximum of 15% variation in honest report levels (primary dependent variable) during the computer game task across three consecutive sessions. Additionally, report accuracy had to remain below 40% in each of these three consecutive sessions.

Audience Condition

A novel adult, who had no preexperimental history with the participants, was present in the experimental room while participants were completing the task (i.e., computer game). As during baseline, the experimenter remained in the adjacent room. The adult (audience) held a clipboard and sat behind participants, taking data and watching them throughout the session. If participants asked him a question or tried to interact, the adult said, “I can’t talk to you right now. Keep playing the computer game, please.” At the end of the session, the experimenter returned to the room and, in the presence of the audience member, asked participants how many points they obtained in that session. Consequences, level of difficulty, and stability criterion were the same as in baseline (i.e., 15% variation in correspondence levels across three consecutive sessions).

Interobserver Agreement (IOA)

A second observer collected data on participants’ final session reports, while all other data collection was automated. An agreement was scored when both the observer and experimenter recorded the same verbal report. IOA was assessed for a minimum of 86.5% of sessions for each participant and calculated by taking the number of agreements divided by the number of agreements plus disagreements and multiplying by 100. IOA was 100% for the four participants. Given that tasks were automated, with the exception of the final total score report, we did not collect procedural fidelity data.

Results

Figure 2 shows the percentage of honest reports of correct (hits) and incorrect (misses) responses during the game for all participants across conditions. Gray bars indicate the percentage of errors (misses) during the task. White squares represent the percentage of honest reports of hits, and black triangles represent the percentage of honest reports of misses. The numbers under each chart indicate the total score (TS) actually obtained by the participants and their final report (FR) on the total score.

Participants honestly reported correct hits at 100% during all experimental conditions. In other words, when they hit the target, they accurately reported having done so, regardless of whether an adult was present. During the initial baseline, Julio, Diego, and Rita honestly reported between 72% and 100% of errors during the first session, as well as the total final score obtained at the end of the session. However, for these three participants, honestly reporting errors decreased throughout the condition. Only Diego started reporting higher scores than the ones actually earned at the end of four out of five sessions. Fausto honestly reported his errors only on 20% of the trials during the first baseline session, with his reports decreasing to 0% accuracy during the subsequent two sessions. At the end of each session, Fausto also reported higher final scores than what he had obtained at the end of two out of three baseline sessions. Julio and Rita honestly reported their final total score regardless of the experimental condition.

The presence of an adult serving the role of the audience during task completion increased the percentage of honest reports of errors to 100% for all participants, with Fausto and Julio making a few dishonest reports during the first and second sessions, respectively. Final total session reports to the experimenter were also honest across participants. During the return to baseline, honest reports of errors decreased to 0% for the first session for all participants, remaining at zero for Fausto and Julio, near zero (M = 8.47%) for Diego, and increasing for Rita, who honestly reported her errors at 100% during the last baseline session. Diego and Fausto emitted, respectively, one and two dishonest reports about the final total score (final report). Both participants lied on their final total score only when they had obtained less than eight points (i.e., six or seven points).

Rita’s pattern of responding during the return to baseline (i.e., 100% accurate) could have been due to seeing the primary experimenter through the glass in the adjacent room (the asterisk in Fig. 2 indicates the session in which Rita saw the experimenter). During the last session of the second baseline condition, Rita left the chair, put her head against the glass and waved to the experimenter. The experimenter pretended not to see her and did not mention anything about such occurrence. In the following sessions, Rita remained seated and did not put her head against the glass anymore. When exposed to the audience condition again, Rita’s levels of honest reports of errors were 100% across all sessions and remained at this level at the return to baseline.

Due to participants’ availability, the audience condition was reintroduced to Diego and Rita only. Following reintroduction of this condition, the percentage of honest reports of errors went back to 100%, with final session reports also being honest, replicating the initial implementation.

Discussion

We investigated whether the presence of an adult (i.e., audience) would exert control over children’s honest reports. Results showed that when an adult was present during the session, participants honestly reported their errors and final total session scores. On the other hand, when no adult was present, participants lied; that is, they reported their errors as correct responses, and two of four participants inflated their final session total scores.

Previous studies on audience control of reported events suggested that children tend to emit dishonest reports in the absence of an audience and tell the truth when in the presence of an audience, especially if the audience consists of an adult (e.g., Brino & de Rose, 2006; Cortez et al., 2019). However, given that these studies lacked a certain level of experimental control, their results should be interpreted with caution. The current study, which employed greater experimental control, confirmed these findings. Results from this study also replicated other findings from the do-say correspondence literature (Balog et al., 2019; Brino & de Rose, 2006; Cortez et al., 2013, 2014, 2017, 2019; Domeniconi et al., 2014; Mazzoca & Cortez, 2020; Oliveira et al., 2016), in that participants’ dishonest reports were suppressed by the presence of an unknown adult, likely due to a preexperimental history of punishment for making (and subsequently) reporting errors. One interpretation of these findings is that children receive aversive consequences (e.g., verbal reprimands, physical punishment) when they accurately report errors (e.g., low grades), which could contribute to the decreased likelihood of reporting errors in the future. This may establish caregivers (i.e., the listeners) as discriminative stimuli for punishment (i.e., negative audiences), which may in turn increase the frequency of lying in their presence (Paniagua, 1989).

Cortez et al. (2014) showed that accurate/honest reports may also vary as a function of the type of task, in that children were more likely to report inflated scores when reporting their performance on academic tasks (e.g., reading and math) and less likely to do so when reporting their performance on leisure tasks (e.g., music). This could be because children may receive more punishment for making errors during academic tasks rather than leisure activities. Moreover, children may have a history of receiving positive social reinforcement for reporting successes. These contingencies may serve to distort children’s tacts to the point that their verbal behavior may no longer be under control of a discriminative stimulus (i.e., the actual event), but rather the specific motivating operation related to either positive or negative reinforcement, making it a lie (Skinner, 1957).

For all participants in the current study, the presence of an adult led to an immediate increase in accurate reports; that is, participants immediately ceased lying about their performance during the game or during their final report (total score) to the experimenter. Given that the specific adult present during task completion never delivered consequences contingent upon task performance or report accuracy, its audience effects over verbal reports may be due, as mentioned above, to extra-experimental histories. Given that caregivers consider lying as inappropriate and problematic behavior (Stouthamer-Loeber, 1986), it is possible that these children may have experienced some form of reinforcement or punishment (i.e., timeout, removing privileges, verbal and/or physical punishments) delivered by an adult for honest and dishonest reports, respectively (Paniagua, 1989).

Thus, presence of an adult during the session may have served as a discriminative stimulus correlated with reinforcement for reporting successes (i.e., positive audience) and a discriminative stimulus correlated with punishment for reporting errors (i.e., negative audience). If this were the case, the adult may have acquired audience control over participants’ reports due to their similarity with previous adults who may have either reinforced honest reports or punished dishonest ones. Additionally, because participants were exposed to a pretraining condition during which the experimenter provided reinforcement for corresponding reports (albeit very briefly), it is possible that the experimenter became a positive audience for honest reports. Thus, the novel adult may have served the same (audience) function as previous adults due to similarities with previous audiences which have imposed the same types of contingencies for truth-telling (Spradlin, 1985). Future research should assess the effects of both positive and negative audiences over children’s honest reports while directly manipulating the contingencies that would lead to this type of control.

If participants had previously experienced reinforcement (from adults) for reporting honestly and punishment for reporting dishonestly on multiple occasions (i.e., multiple-exemplar training), then the correspondence between doing and saying may be automatically reinforcing (e.g., Ribeiro, 1989), and the lack of correspondence may be automatically punishing or akin to extinction. If this were the case, participants would have been honest across all conditions, regardless of the presence or absence of an audience, which did not happen. However, it may be possible that any automatic conditioned reinforcer arising from emitting correspondent reports (i.e., being honest) may be conditional upon the presence of an audience. This would explain why children emitted more honest reports in the presence of the adult. This is a plausible analysis that may aide in explaining “feeling good about telling the truth” or “feeling bad about lying” for certain individuals (e.g., Palmer, 1996). This type of analysis has yet to be considered in this area of research and should be the focus of future investigations.

The results of the current study contribute to the literature by showing the differential effects of an audience variable in the context of assessing and teaching children’s accurate reports, a skill that is crucial in their development (Greer & Ross, 2008). Even though the behaviors of accurately saying what you did and doing what you say are differentially reinforced by the verbal community very early in a child’s development (Catania, 2013), these repertoires have been shown to be easily disrupted. For instance, when consequences are made contingent upon specific verbal behavior such as reporting correct responses only, stimulus control may be eventually lost (Balog et al., 2019; Brino & de Rose, 2006; Cortez et al., 2013, 2014, 2017, 2019; Domeniconi et al., 2014; Mazzoca & Cortez, 2020; Oliveira et al., 2016; Ribeiro, 1989; Stocco et al., 2021), in that participants would lie (Skinner, 1957). Very few studies in behavior analysis have explored the process by which stimulus control of verbal behavior is either distorted or lost due to the delivery of specific consequences. However, if the most important reinforcement for the tact consists of how the listener reacts to it, then the specific actions taken by a listener in response to a tact may greatly contribute to its distortion (Skinner, 1957). Thus, if a child receives praise contingent upon reporting successes, these types of reports may increase despite the child having actually been unsuccessful.

Considering that lying and being dishonest have been major topics of concern for caregivers (Crawshaw, 2015; Gervais et al., 2000; Paniagua, 1989), identifying their controlling variables may allow for the development of strategies to promote honest statements. Previous research has shown that when reinforcement is contingent upon statements that correspond with previous behavior (correspondence training), children are more likely to accurately report what they have done, even if it involves reporting a mistake (Baer, 1990; Bevill-Davis et al., 2004; Cortez et al., 2019; Lloyd, 2002; Paniagua, 1990; Ribeiro, 1989). Therefore, parents and caregivers should reinforce self-reports of errors, mistakes, and inappropriate behaviors, so children are more likely to tell the truth. It is possible that when parents deliver aversive consequences contingent upon reporting having made a mistake, the future likelihood of reports rather than mistakes will be reduced. Thus, prevention or intervention programs for lying should focus on the establishment of correspondence between verbal and nonverbal relations, in which reinforcement is made contingent upon corresponding reports and reinforcement is withheld for noncorresponding reports (da Silva & Lattal, 2010; Lattal & Doepke, 2001). Because both verbal behavior and its related events are accessible to caregivers, “the listener can monitor the presence versus the absence of the particular relation under consideration” (Paniagua, 1989, p. 975).

Although the effectiveness of correspondence training has been shown in several basic and translational studies (e.g., Balog et al., 2019; Brino & de Rose, 2006; Cortez et al., 2014; Cortez et al., 2013; Cortez et al., 2017; Domeniconi et al., 2014; Ribeiro, 1989), its application outside of laboratory settings has been limited. Future translational and applied studies should consider the use of hidden cameras (Sauter et al., 2020) or one-way mirrors (e.g., Cortez et al., 2014; Cortez et al., 2019) to observe participants complete a variety of tasks that would later have to be reported. Experimenters may also rely on observable response products that may serve as strong predictors of the reported event (Stocco et al., 2014). Considering that it could be unrealistic to some parents or professionals to continuously observe children’s behaviors to verify whether their reports correspond to what they actually did, future research should investigate whether intermittent observations would be sufficient to increase the accuracy of reports. Previous studies on do-say correspondence have shown that thinning the schedule of reinforcement for corresponding reports from continuous to intermittent was successful in maintaining a high frequency of accurate reports in sessions conducted 30 and 60 days after the original intervention (Cortez et al., 2013; Cortez et al., 2017).

Previous research has shown that honestly reporting or do-say correspondence may be a “skill” or higher-order operant (e.g., Cortez et al., 2014; Greer & Ross, 2008; Luciano et al., 2001; Matthews et al., 1987). Cortez et al. (2014), for example, conducted correspondence training for one of four tasks (reading, math, computer game, and music) and saw increases in accurate reports across all tasks. Future research should investigate whether correspondence training in the presence of one audience or multiple audiences (multiple exemplar training) would generalize to novel audiences.

A possible limitation of the study was that Rita’s behavior did not return to initial baseline levels during the final withdrawal. This is likely a function of Rita seeing the primary experimenter through the one-way mirror. Despite this, we observed a reversal during the first return to baseline and a subsequent replication of the audience effects. Although it is possible that participants may have reacted to nonvocal responses emitted by the adult present in the room (e.g., head shakes, nods, facial expressions), he had been trained to remain neutral and not react to any of the participants’ initiations. However, future research should collect procedural fidelity data on the audience’s nonvocal responses during the session.

Future research should also replicate our procedures with additional participants to increase the generality of our findings. Furthermore, studies should establish the discriminative function of different audiences as a way to reduce the influence of extra-experimental variables. The effects of the number of audience members, as well as their specific characteristics (e.g., gender, age, clothing) could also be manipulated. We hope this work motivates applied researchers to investigate and evaluate interventions for lying from a behavioral perspective. Although some initial efforts have been made to evaluate some procedures to teach children with autism to report past events (Naoi et al., 2007; Shillingsburg et al., 2017; Shillingsburg et al., 2019), no study with such a population investigated the role of audience control.

Finally, it is important to note that the terms honest and dishonest were used to describe (i.e., tact) specific behaviors, namely accurate and inaccurate reports, respectively, and were never meant to imply participants’ characteristics. Honesty is a label used to describe observed behaviors (e.g., truth-telling) and should never serve as the explanation for these behaviors (e.g., children lied because they are dishonest). Instead, specific environmental variables, such as the ones investigated in this study, are responsible for the types of behaviors that are often subsumed under the label of honesty. It is our hope that future researchers continue to investigate important psychological and social constructs, such as truth-telling and lying, from a behavior-analytic (i.e., radical behaviorist) perspective.

Declarations

Conflict of Interest

The Authors declare that they have no conflict of interest.

Ethical approval

All procedures conducted in the present study were conducted in accordance with the ethical standards of the institutional research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Informed consent

Informed consent was obtained from all participants’ parents before the beginning of the experiment. We also obtained participant’s assent.