Teaching Children with Autism to Identify Known and Unknown Information across Self and Others

Abstract

This study evaluated procedures for teaching three children diagnosed with autism spectrum disorder the perspective-taking skill of identifying known and unknown information by others based on what they were sensing across all five senses: see, taste, feel, hear, and smell. Using a multiple baseline across participants design, this study evaluated a training package consisting of rules, multiple exemplar training, error correction, and reinforcement. The treatment package successfully taught participants to identify known/unknown information based on what individuals sensed. Generalization across untrained stimuli and people was observed from baseline to posttraining for all participants.

Perspective taking involves observing the behavior of others and subsequently responding to the private events that they would typically have in given situations and/or predicting their future actions (LeBlanc et al., 2003). Autism spectrum disorder (ASD) is characterized by pervasive deficits in social communication and interaction repertoires, which categorically include deficiencies in social-emotional reciprocity; nonverbal communicative behaviors; and developing, maintaining, and understanding social relationships (American Psychiatric Association, 2013). This often leads to individuals with ASD displaying related deficits in the varying aspects of social skills to be mediated between oneself and others, including taking on another’s perspective (Baron-Cohen et al., 1985).

The research literature proposes varying definitions of perspective taking with some differing theoretical underpinnings. Although there have been theoretical accounts of the development of perspective-taking repertoires from traditional developmental and cognitive psychological research in theory of mind (ToM), there are also behavioral accounts.

Theory of Mind Account

The vast majority of existing literature on assessing and teaching perspective-taking skills to individuals with ASD comes from ToM research. ToM refers to an individual’s understanding and inferences of the mental states of another individual (e.g., knowledge or thoughts, emotions, intentions, beliefs, desires) and the subsequent use of such information in order to: interpret what another person may say, make sense of their behavior, and make predictions about their future behavior (Howlin et al., 1999; Ozonoff & Miller, 1995).

ToM researchers have described five different levels of understanding informational states that are suggested to function as prerequisites for individuals learning to take the perspective of another (Howlin et al., 1999), including: (1) identifying what others sense; (2) identifying how stimuli appear to others depending on their position in space; (3) identifying who knows information based on what they are sensing or experiencing/“seeing leads to knowing”; (4) predicting the future behavior of another person based on information gathered in steps 1–3 (“Where will he look for the bicycle?”); and (5) identifying others beliefs about the environment (Hadwin et al., 1996). With respect to step 5, individuals are described as having a true belief when their description of events matches reality (e.g., He thinks his bicycle is in the garage, and indeed, it is) and a false belief when their description of events does not match reality (He thinks his bicycle is in the garage but his sister borrowed the bicycle without his knowledge; Schneider et al., 2015).

In ToM studies conducted by Baron-Cohen et al. (1985, 1986, 1989), false-belief tasks (also referred to as the “Sally-Anne task”) were used to evaluate perspective taking across children of varying populations (e.g., ASD, intellectual disabilities, and typically developing controls of a lower mental age). They found that although the majority of individuals with intellectual disabilities and typically developing controls passed such tests, approximately 80% (16/20) of those with ASD did not accurately answer questions pertaining to the beliefs of others. An interesting pattern also emerged wherein the children with ASD frequently answered perspective-taking questions inaccurately by responding to their own knowledge instead of responding with respect to the perspective of another individual. These findings have remained consistent across research in this area, culminating ToM researchers to suggest that ToM impairments are core, central features of deficiency in children with ASD (Leslie & Frith, 1988; Perner et al., 1989).

Behavioral Accounts and Research

There are also accounts for perspective taking within behavioral research representative of Skinner’s (1953, 1957) work, as well as a post-Skinnerian account of human language and cognition posed by relational frame theory (RFT; Hayes et al., 2001). Behavior analytic research regarding perspective taking is distinct from research conducted in more traditional psychological perspectives in the particular way in which it views the acquisition of the skill. Although traditional psychology suggests that children do not demonstrate ToM until they have reached a developmental milestone, behavior analysis attributes perspective taking to specific learning histories or a shared interaction with physical environmental events (Spradlin & Brady, 2008). For example, in the same way that individuals learn to tact their own private events by referring to accompanying public events, perspective taking involves tacting the private events of others by referring to accompanying public events (Schlinger, 2009).

Much of recent behavioral research on perspective taking has adopted an RFT account of human language and cognition, which expands conceptual analyses to view the development of perspective taking as relational responding. In particular, an RFT account suggests that the behavior of relating, in this case being able to distinguish between one’s perspective and another’s, is learned operant behavior that can be strengthened. In other words, relating is behavior under the antecedent control of contextual cues (e.g., pronouns such as “I” and “you”) that were established through a history of multiple exemplar training (MET; Barnes-Holmes et al., 2004).

Behavior analytic research has successfully taught various perspective-taking skills using behavioral intervention procedures such as MET and rules. MET involves inclusion of stimulus variations during teaching in an effort to promote generalization across situations (Cooper et al., 2020). MET has been used to teach children to identify others’ false beliefs (e.g., Charlop-Christy & Daneshvar, 2003), respond appropriately to metaphorical speech (Persicke et al., 2012), respond to false-belief tasks in the natural environment (Dhadwal et al., 2021), and tact what others are sensing (Gould et al., 2011; Hahs, 2015; Welsh et al., 2019).

The use of rules has been included in previous behavioral research for teaching perspective-taking skills. Rules are used to indicate a contingency that exists so that individuals do not have to experience the contingencies directly. Instead, the expected consequences are described in the rule (Cooper et al., 2020). Rules have been used to teach children to use socially appropriate deception (Bergstrom et al., 2016), and treatment packages including the use of both rules and MET have been used to teach individuals with ASD to recognize sarcasm (Persicke et al., 2013), respond to deceptive statements (Ranick et al., 2013), resolve social conflicts (Suarez et al., 2021), respond appropriately to disguised mands (Najdowski et al., 2017), and infer and respond to the play preferences of others (Najdowski et al., 2018).

Sensing and Knowing Research

In one of the tacting what others are sensing studies mentioned earlier, Gould et al. (2011) evaluated the effects of MET on teaching children with ASD the identification of what others can see. The table-top procedure involved the presentation of a picture of a person, surrounded by four stimuli oriented above, below, left, or right of the person. The participant was then to follow a fading arrow prompt showing the depicted person’s eye gaze and was asked to identify the stimulus they were seeing. The results revealed that all participants learned to identify the correct stimulus being seen, and generalization of this skill was observed with untrained pictures, but tacting with real people in vivo was limited.

Hahs (2015) replicated and extended the study conducted by Gould et al. (2011) in a variety of ways. First, the necessity of a full verbal repertoire was assessed; second, the response requirement of participants was assessed when extended to include above and below alongside left and right, rather than left or right alone; third, the addition of a gestural prompt was assessed; and fourth, the environment was differentially assessed as the study was conducted in an educational setting with peers and supervisors. The results demonstrated that two of three participants demonstrated generalization to untrained stimuli, whereas the third did not. Furthermore, generalizability to the natural environment, in vivo and with live people, either did not occur or was limited. These studies both sought to use an MET approach to teach participants to tact what others could see on the basis of public behavior (e.g., eye gaze), but neither moved beyond the sense of seeing.

Welsh et al. (2019) even further extended this line of research to include teaching children with ASD to tact stimuli others are experiencing across all senses (i.e., seeing, tasting, feeling, hearing, and smelling) in the natural environment, in vivo, and with live people. Through MET, participants were taught to attend to public stimuli correlated with the private events of sensing (Welsh et al., 2019). The results revealed that this teaching procedure was efficacious for teaching all participants to accurately tact what others could sense, and generalization from baseline to posttraining was observed across untrained people and stimuli for all participants.

Identification of what others are sensing has been suggested to be a potential prerequisite for identifying who knows information based on what they are sensing or experiencing/“seeing leads to knowing” (Howlin et al., 1999; Welsh et al., 2019). “Seeing leads to knowing” indicates an individual demonstrates the skill of identifying that other people only know things they have experienced themselves (directly or indirectly; Howlin et al., 1999). Therefore, mastery of this skill would involve an individual’s identification that they can or cannot tact seeing something (e.g., placed within a container) unless they have seen that it was placed there, as well as the fact that others can or cannot tact seeing something (e.g., placed within a container) unless they have also seen it placed there. In particular, “people only know about things they have seen. If they can’t see something, then they don’t know about it” (Howlin et al., 1999, p. 244).

An approach to assessing this skill is through scenario-based self-judgment and other-judgment trials. In self-judgment, the individual is engaged in playing a hiding game with varying empty boxes and stimuli that differ in feature (e.g., color, size). The individual is then told that one of the objects is going to be hidden in the box, and they are instructed to either watch it be placed or keep their eyes closed. Then, knowledge (e.g., “Do you know what is in the box?”) and justification (e.g., “How do you know?”/“Why don’t you know what is in the box?”) questions are asked. In other-judgment, the individual is encouraged to play the same hiding game, but with another individual, who is shown the stimuli to be placed in the empty box and either remains to watch its placement or is instructed to close their eyes. Then, knowledge (e.g., “Does [other person] know what is in the box?”) and justification (e.g., “How does [he/she] know?” / “Why doesn’t [he/she] know what is in the box?”) questions are asked. Responding with accuracy to self- and others-judgment questions, the individual will explain that they or another individual either knew what stimulus was placed in the box or did not, based on whether or not they or another individual had observed the placement of the stimulus.

Accurately identifying who knows information based on what they are sensing or experiencing seems relevant and important for predicting the actions of others based on their knowledge obtained through experience. For example, “people think things are where they saw them. If they didn’t see something then they won’t know where they are” (Howlin et al., 1999, p. 248). Therefore, if Mom saw her sunglasses on her desk, one can predict that the desk is where she will look for her sunglasses. If she did not see her daughter borrow her sunglasses, she will not know they are no longer there.

Identifying what others know based on their experiences is helpful for successful social interactions. In the sunglasses example, Mom may become annoyed when she cannot find her sunglasses. Her daughter’s recognition that her mother may look for the sunglasses on the desk, may prompt the daughter to inform Mom that the sunglasses are being borrowed. That is, when others do not know information, they may need to be provided with information (e.g., Grandpa is not looking down while on a walk and does not know there is dog feces ahead; Dad is wearing headphones and does not know the doorbell is ringing).

Purpose

No behavioral research has evaluated methods for teaching individuals with ASD this skill. The purpose of the current study was to evaluate the effects of a training package that included rules, MET, error correction, and reinforcement on participants’ identification of what they and others (1) know and how they know the information and (2) do not know and why they do no not know the information. In line with Schlinger’s (2009) suggestion to allow participants to come into close contact with the relevant accompanying public stimuli, leading question and experiential prompts were used during training. Furthermore, the current study used an RFT-informed approach by incorporating MET as a feature of the teaching procedure.

Method

Participants and Settings

Three children diagnosed with ASD participated in this study. Informed consent was obtained by all participants using a consent form approved by Pepperdine University’s Institutional review board (IRB). Kyle was an 11-year-old, Latin American, English-speaking male living in an upper-middle-class neighborhood. He was receiving behavioral intervention from a community-based agency for 20 hr per week in his home and was performing below age equivalence across domains on the Vineland Adaptive Behavior Scales 3 (Sparrow et al., 2016).

Joy was a 7-year-old, white, English-speaking female living in a middle-class neighborhood. She was receiving behavioral intervention from a community-based agency for 20 hr per week in her home and was performing at a level three on the Verbal Behavior Milestones Assessment and Placement Program (VB-MAPP; Sundberg, 2008) at the time of the study.

Odin was a Chinese American, English-speaking, 6-year-old male living in an upper-middle-class neighborhood. He was receiving 35.5 hr per week of behavioral intervention in a school setting. His intelligence quotient (IQ) identified from the Wechsler Intelligence Scale for Children, Fifth Edition (WISC®-V; Wechsler, 2014) was 93, and his test scores on the Expressive One-Word Picture Vocabulary Test (EOWPVT; Martin & Brownell, 2011) and Peabody Picture Vocabulary Test (PPVT-4; Dunn & Dunn, 2007) were 93 and 85, respectively.

All participants communicated using full sentences and had the following prerequisite skills related to the study’s targeted skills: (a) tacting all stimuli targeted; (b) responding correctly to pronouns; (c) responding to names of familiar people; (d) tacting items sensed (When asked, “What do/does you/he/she/I see/taste/feel/hear/smell?”); (e) responding to and using negation (don’t); (f) answering cause-and-effect questions (Why?); and (g) demonstrating rule-governed behavior. They also had the prerequisite skills necessary to answer the questions that were presented during sessions. For example, if the question asked whether something was wet or dry, the participant already had that vocabulary (wet, dry) in their repertoire.

One session per day was conducted 1–3 days per week during their regularly scheduled ABA-based intervention sessions. For Kyle and Joy, trials were conducted in various rooms and areas of the home, including the participants’ bedrooms, kitchen, living room, bathroom, front yard, and back yard. For Odin, trials were conducted in various rooms and areas of the school, including the classroom and an outdoor area facing the playground just outside of the classroom. The rooms contained everyday living items that were normally in the rooms; therefore, multiple distracting stimuli were present (e.g., radio on, brother playing) during sessions. Sessions were conducted in various rooms, thus the participant and others in the room sat or kneeled on the floor, sat on the couch or a chair, stood, and sometimes sat at a table in the classroom when trials were presented.

Response Measurement and Interobserver Agreement

Identifying whether self or others knew given information and the reason they knew or did not know the information was the dependent variable. Responses were recorded as correct if the participant accurately responded within 3 s by identifying whether information was known or unknown by a target person based on the stimulus that the target person was or was not sensing and if the participant identified the correct sense the target person was using to gain this knowledge. For example, a participant was asked, “Does (name) know that music is playing?,” and the participant responded, “Yes.” The experimenter then asked, “How does (name) know this?,” and the participant responded, “Because (name) can hear it.” For a trial to be scored as correct, responding to both questions was required to be accurate. In particular, a response was recorded as incorrect if either (1) the participant did not accurately identify whether or not the target person knew information (e.g., a participant was asked if [name] knew music was playing [while wearing headphones] and the participant responded with “no” instead of “yes”); (2) did not identify the correct sense being used to gain the information (e.g., a participant was asked how [name] knew music was playing [while wearing headphones] and the participant responded, “Because he can see/taste/feel/smell it.”); or (3) if the participant did not initiate a response within 3 s of delivery of the instruction.

Interobserver agreement (IOA) was collected by two independent observers using trial-by-trial agreement. Trials were recorded as an agreement when the observers agreed on whether the trial was correct or incorrect. IOA was calculated by dividing the number of agreements by agreements plus disagreements and multiplying by 100%. IOA was collected for 52.94% of sessions for Kyle, 60% of sessions for Joy, and 33% of sessions for Odin. Mean IOA agreement equaled 100%, 96.67% (range: 90%–100%), and 99.28% (range: 95%–100%) for Kyle, Joy, and Odin, respectively.

Experimental Design and Procedure

A nonconcurrent multiple baseline across participants design was implemented to examine the effects of the intervention.

General Procedure

Twenty trials were conducted each session. See Table 1 for the specific distribution of trials across people. The target person being observed each trial by the participant was positioned approximately 6–8 ft away from the participant. Experimenter one discreetly instructed the target person in vivo how to engage with the target stimuli prior to the presentation of each question. All instructions were provided in a manner in which the participants were unable to observe. Additional people were in the room to serve as distractors, and they were not instructed on how to behave. During each trial, experimenter one asked the participant whether or not a target person knew specific information about a stimulus (i.e., “Do/Does you/I/[person] know [information]?”) and how they knew (i.e., “How do/does you/I/[person] know?”) or why they did not know (i.e., “Why don’t/doesn’t you/I/[person] know?”) the information.

Table 1.

Distribution of Known (K) and Unknown (U) Stimuli across People during Study Phases

	Baseline	Intervention	Novel Person Probe	VR-3	Posttraining
Kyle	Mom (5U, 4K) Dad (3U, 3K) Experimenter 1 (5U, 3K) Experimenter2 (5K, 1U) Kyle (13U, 15K)	Experimenter1 (23U, 17K) Kyle (16U, 25K)	Experimenter 1 (2U, 2K) Experimenter 3 (5U, 5K) Kyle (3U, 3K)	Experimenter1 (9U, 10K) Kyle (11U, 10K)	Mom (5U, 4K) Dad (1U, 1K) Experimenter 1 (2U, 3K) Experimenter 2 (4U, 7K) Kyle (14U, 14K)
Joy	Mom (7U, 8K) Sibling (6U, 3K) Experimenter 1 (2U, 2K) Experimenter 2 (6U, 9K) Experimenter 3 (1U, 1K) Experimenter 4 (2U, 1K) Joy (5U, 5K)	Experimenter 1 (43U, 36K) Joy (42U, 39K)	Dad (4U, 4K) Experimenter 1 (3U, 3K) Joy (3U, 3K)	Experimenter 1 (10U,10K) Joy (5U, 5K)	Mom (5U, 8K) Sibling (2U, 1K) Experimenter 1 (4U, 4K) Experimenter 2 (5U, 6K) Experimenter 5 (3U, 5K) Joy (19U, 18K)
Odin	Experimenter 1 (7U, 7K) Experimenter 2 (7U, 6K) Classmate 1 (7U, 3K) Classmate 2 (14U, 8K) Experimenter 3 (5U, 7K) Experimenter 4 (1U, 2K) Classmate 3 (1U, 1K) Odin (39U, 41K)	Experimenter 1 (48U, 40K) Odin (35U, 37K)	Experimenter 1 (3U, 2K) Experimenter 5 (4U, 5K) Odin (3U, 3K)	Experimenter 1 (10U, 10K) Odin (10U, 10K)	Experimenter 1 (2U, 4K) Classmate 1 (1U, 2K) Classmate 2 (5U, 4K) Experimenter 3 (5U, 4K) Classmate 3 (1U, 1K) Odin (5U, 4K)

Different stimuli were presented each trial so that the same stimulus was never repeated (see Table 2 for sample stimuli presented for each sense). To determine the order in which questions about senses would be presented and which people would engage in the sensing behavior a die was rolled. For example, if the experimenter rolled a three for “sense,” it was associated with "feel." Once the order of senses were determined, the person to be observed was determined by the roll of a die. For example, if the experimenter rolled a four for “person,” it was associated with “experimenter 2.” For each sense, a coin flip was used to determine the order in which known versus unknown information pertaining to a person was presented. For example, if the experimenter flipped heads, the trial contained a question about known information, and if the experimenter flipped tails, the trial involved a question about unknown information. The order was determined semirandomly, using the following rules: (1) the same person could not be presented more than twice consecutively; (2) known and unknown information could not be presented more than twice consecutively; (3) the same sense could not be presented more than twice consecutively; and (4) each of the five senses were required to be presented four times throughout the 20 trials.

Table 2.

Sample Stimuli across Senses

	Setup	Question	Response	Question	Response
Unknown
See	Hide a crayon or marker in hand.	Do you know what color the crayon in my hand is?	No.	Why don’t you know?	Because I can’t see it.
	Show child a picture of preferred character on your phone. Do not let (person) see.	Does (person) know who this is?	No.	Why doesn’t (person) know?	Because (person) can’t see it.
	Experimenter facing in the opposite direction from the child. Have child pick out a toy without showing you.	Do I know which toy you are holding?	No.	Why don’t I know?	Because you can’t see it.
Taste	Take a bite out of a breakfast cracker (do not let the learner see the packaging of the crackers or the flavor).	Do you know if my crackers have a blueberry or a grape flavor?	No.	Why don’t you know?	Because I can’t taste it.
	Give the child something to drink with (person) in the room.	Does (person) know what flavor your drink is?	No.	Why doesn’t (person) know?	Because (person) can’t taste it.
	Dip a pretzel stick in a sauce (maybe guacamole, ranch, or cheese sauce). Then, pop the pretzel in the learner’s mouth.	Do I know if the sauce is spicy?	No.	Why don’t I know?	Because you can’t taste it.
Feel	Bring child into the bathroom and turn on the faucet (don't let child see what knob you turned on).	Do you know if the water is hot or cold?	No.	Why don't you know?	Because I can’t feel it.
	Give the learner a damp towel.	Does (person) know if the towel is wet or dry?	No.	Why doesn’t (person) know?	Because (person) can’t feel it.
	Give the learner a bag and have him put whatever toys they want to place in it. Do not watch them place the toys in the bag. Then have the learner zip up the bag.	Do I know if the bag of toys will be heavy or light?	No.	Why don’t I know?	Because you can’t feel it.
Hear	Whisper to another person in front of the child and make sure the child cannot hear.	Do you know what I am saying to (person)?	No.	Why don’t you know?	Because I can’t hear it.
	Play music on a low volume while a third person is in another room.	Does (person) know we are playing music?	No.	Why doesn’t (person) know?	Because (person) can’t hear it.
	Have the learner turn on a radio station while wearing headphones or earbuds.	Do I know if the radio station is on a commercial break?	No.	Why don’t you know?	Because you can’t hear it.
Smell	Bring novel scented lotion and have parent smell it, do not let the child smell it.	Do you know if my lotion is a minty or sweet scent?	No.	Why don’t you know?	Because I can’t smell it.
	Stand far away from (person) but still in view.	Does (person) know if I am wearing perfume right now?	No.	Why doesn't (person) know?	Because (person) can’t smell it.
	Instruct the learner to smell a vanilla candle (ensure that the lid is off).	Do I know what type of candle you have?	No.	Why don’t I know?	Because you can’t smell it.
Known
See	Hold an object up in front of the learner.	Do you know how many fingers I’m holding up?	Yes.	How do you know?	Because I can see it.
	(Person) looking at something on their phone.	Does (person) know what picture is on their phone?	Yes.	How does (person) know?	Because (person) can see it.
	Turn the light on in the oven without the learner seeing it.	Do I know if the light is on in the oven?	Yes.	How do I know?	Because you can see it.
Taste	Tell child to close their eyes and have them taste a familiar food.	Do you know what food this is?	Yes.	How do you know?	Because I can taste it.
	(Person) is sucking on a lollipop.	Does (person) know what flavor the lollipop is?	Yes.	How does (person) know?	Because (person) can taste it.
	Have a third person pop a chip into your mouth.	Do I know if this chip has a cheesy or spicy flavor?	Yes.	How do I know?	Because you can taste it.
Feel	Give the child a paper towel to wipe their hands with.	Do you know if the paper towel is wet or dry?	Yes.	How do you know?	Because I can feel it.
	(Person) picks up a closed box.	Does (person) know if the box is heavy or light?	Yes.	How does (person) know?	Because (a person) can feel it.
	Pick a leaf off a tree or bush outside and then crush it in your hand. Make sure the learner observes this.	Do I know if this leaf is sticky?	Yes.	How do I know?	Because you can feel it.
Hear	Say, “That’s refreshing” after you take a sip of water.	Do you know if I am saying “That’s refreshing” after I drink my water?	Yes.	How do you know?	Because I can hear it.
	(Person) is standing right outside the door but is not looking inside the room. Child claps.	Does (person) know what you are doing?	Yes.	How does (person) know?	Because (person) can hear it.
	Hand the learner your phone and play a classical song at a low volume in which there is no singing (instrumental). Then, stand at a distance from the phone and learner.	Do I know if someone is singing in the song?	Yes.	How do I know?	Because you can hear it.
Smell	Place a piece of mint candy near the learner’s nose (but do not let the learner see it).	Do you know if this is a mint or sour candy?	Yes.	How do you know?	Because I can smell it.
	Have (person) close their eyes and smell oranges.	Does (person) know what food this is?	Yes.	How does (person) know?	Because (person) can smell it.
	Wear a hat for a couple of minutes then cover your face with the hat.	“Do I know if this hat has a scent like my hair?”	Yes.	How do I know?	Because you can smell it.

Given that the target response was contingent upon identifying known or unknown information based on a sensing behavior, the experimenter ensured that the target person was engaging in a sensing behavior that was different than the participant when relevant. This was done to control for the participant simply tacting known or unknown information based on what they were sensing. For example, if the experimenter asked what the target person knew based on something the target person could feel, the experimenter made sure that the participant did not touch or hold the item prior and during the presentation of the trial. In another instance, if the experimenter asked what a target person knew based on something the target person could hear, the experimenter made sure that the participant was unable to hear the stimulus being sensed. In particular, the experimenter would have the target person sensing the stimulus wear headphones with the volume at a level that could not be heard by the participant and in the absence of extraneous sounds (i.e., television on if the target stimulus was a song on a phone) that could be heard by the participant before the trial was presented. The hearing sense, however, was not fully controlled for, as there were mistakenly some trials, during which, the participant was also able to hear what the target person could hear.

We also tried to ensure that participants would not be able to answer questions based on a sense unrelated to the target sense. For example, when presenting a damp towel for the sense of “feel” we ensured that the towel did not “look” damp. Or, when presenting a drink or food, the package was not present so that it would not be possible to discern if an item was spicy, sweet, or a particular flavor (cranberry vs. fruit punch). We also attempted to control for learning history by avoiding the presentation of stimuli that a person would know the information of based on their previous history with the item. For example, if a candle or lotion was presented, it was not one that the target person had experienced before, so they would have to rely on their current experience (as opposed to previous experience) to “know” about it.

Baseline

The general procedures outlined earlier were implemented and no feedback or reinforcement was provided for responses. Five to eight people (e.g., the participant, experimenters, a parent, sibling, and/or classmates) served as target people (see Table 1).

Training

During training, only the perspectives of the participant and experimenter one were presented. Perspectives of others were not included so that the remaining people presented in baseline could be saved to test for generalization across people during novel person probes and posttraining.

Training sessions entailed a treatment package consisting of rules, MET (new stimuli being sensed were presented each trial), error correction, and reinforcement. Before each session, an informal preference assessment was conducted during which two stimuli (different items were included each time) considered to be preferred in the past were presented to the participant, and the participant was asked if they wanted to earn one of the stimuli. The item selected combined with praise was then provided contingent upon the participant responding correctly to both questions (“Do/Does you/I[person] know [information]?” and “How/Why . . . ?”) on a continuous reinforcement schedule (CRF). In the event that the participant responded correctly to only one of the questions, praise was provided to the participant for the answer to that question but the preferred item was not provided.

At the beginning of each session a rule was provided: “When you or another person knows something, it is because you or the other person can see, taste, feel, hear, or smell it. When you or the other person do not know something, it is because you or the other person cannot see, taste, feel, hear, or smell it.” Then, the 20 trials were initiated. Once the participant’s responding reached 80% correct for one session, this rule was no longer presented at the start of each session and was instead only used during step one of the error correction procedure.

A four-step error correction procedure followed incorrect responses and consisted of: (1) a rule reminder; (2) a leading question; (3) experiential prompt; and (4) full vocal model. For example, in step 1 of the error correction procedure, when the participant provided an incorrect response, the experimenter presented the participant with a rule reminder (see rule above) and then re-presented the question. If the participant continued to respond incorrectly, step 2 of the error correction procedure was implemented wherein the experimenter asked the participant a leading question such as, “Can I see (stimulus)?” If the participant responded correctly to this leading question (i.e., said, “Yes”), the original question (e.g., “Do I know the color of the box on the table?”) was re-presented. If the participant did not respond correctly to the leading question or continued to respond incorrectly when the original question was repeated, step 3 of the error correction procedure was implemented wherein the participant was positioned in the same viewpoint as the experimenter (experiential prompt), and the original question (e.g., “Do I know the color of the box on the table?”) was re-presented. If the participant still responded incorrectly, the original question was repeated and a full vocal model (step 4) was immediately provided (“Say, ‘Yes’” and “‘Because you can see it’”).

During the first session of training, only two senses (see and taste) were presented in a semirandom order. When the participant responded with at least 80% accuracy for one session with the first two senses, a third sense was added (feel) into the next session. Once the participant responded with at least 80% accuracy for one session with the three senses, a fourth sense (hear) was added. Finally, once the participant responded with at least 80% accuracy for one session with four senses, a fifth sense (smell) was added. When the participant responded with at least 80% accuracy to the random rotation of all five senses for one session, a novel person probe was conducted.

The purpose of the novel person probe was to ensure generalization to a novel person had occurred prior to moving to posttraining. Reinforcement was provided on a CRF schedule during the novel person probe but error correction was not implemented. Experimenter one still asked all the questions during trials, but the second person’s perspective was a novel person. It was planned that if the participant responded below 80% accuracy with the novel person, then the novel person would be introduced into the training phase until 80% accuracy was achieved; however, this was not required for any of the participants.

If the participant responded with at least 80% accuracy to the novel person probe, the experimenter thinned the reinforcement schedule to a variable ratio 3 (VR-3). Posttraining was introduced when the participant responded with at least 80% accuracy for two consecutive sessions in the VR-3 phase.

Posttraining

Posttraining procedures were identical to baseline. Three to four of the baseline sessions were repeated in posttraining, wherein the same people and stimuli used in baseline were re-presented, and none of the stimuli trained in intervention were presented. Posttraining was conducted in order to test for generalization across untrained people and stimuli.

Results

Figure 1 displays the percentage of correct responses for Kyle (top panel), Joy (middle panel), and Odin (bottom panel) during baseline, training, training with a VR-3 reinforcement schedule, and posttraining.

Fig. 1.

Percentage Correct Responding to Knowing Questions across Baseline, Training, Variable-Ratio 3 Reinforcement Schedule, and Posttraining. Note. Open squares represent novel person probes. Arrows with numbers represent the number of target senses included in the labeled and subsequent unlabeled sessions

Kyle

During baseline, Kyle’s percentage of correct responding ranged from 10% to 15%. During training, there was an immediate increase in correct responding. During the sixth training session, Kyle met the mastery criterion of 80% to 100% correct for one session with all five senses. During the novel person probe, Kyle’s responding was 100% correct. Following this, the reinforcement schedule was thinned to a VR–3, and Kyle’s performance continued to be 100% correct. During the posttraining sessions, Kyle’s responding ranged from 90% to 100% correct.

Joy

During baseline, Joy’s percentage of correct responses ranged between 15% and 45%. During training, there was an immediate increase in correct responding. During the eighth session of training, Joy met the criterion of 80% to 100% correct for one session with all five senses, so a novel person probe was conducted, during which, Joy responded with 90% accuracy. Then, the reinforcement schedule was thinned to a VR-3, and Joy’s responding was between 85% and 90% correct. During posttraining, Joy’s performance was between 75% and 95% correct.

Odin

During baseline, Odin’s percentage of correct responding ranged from 10% to 30%. During training, there was an immediate increase in correct responding. During the ninth training session, Odin met the criterion of 80% to 100% correct responding for one session with all five senses required to move on to the novel person probe, during which, his performance was 80% correct. Following this, the reinforcement schedule was thinned to a VR–3, and Odin’s responding ranged from 80% to 90% correct. During the posttraining sessions, Odin’s performance ranged from 75% to 100% correct.

Discussion

The three participants learned to identify information that was known and unknown and how/why the information was known or unknown to themselves and others using a training package consisting of rules, MET, error correction, and reinforcement. Generalization across stimuli and people was observed in that participants responded with high levels of accuracy when the same stimuli and some of the people presented during baseline, but not presented during training, were re-presented during posttraining.

These results are consistent with previous research demonstrating MET to be efficacious for teaching perspective-taking skills (e.g., Bergstrom et al., 2016; Dhadwal et al., 2021; Lovett & Rehfeldt, 2014; Najdowski et al., 2017; Persicke et al., 2012; Persicke et al., 2013; Ranick et al., 2013; Suarez et al., 2021). Also consistent with a study conducted by Welsh et al. (2019), only one experimenter’s perspective was required to be trained in order to observe generalization to untrained people. That is, all participants responded between 80% and 100% correct to a novel person probe after being trained to criterion with one person. Then, participants also went on to respond 75%–100% correct during posttraining when multiple untrained people’s perspectives were presented. Findings were also consistent with previous research demonstrating that skills related to perspective taking can be taught in the natural environment with live people rather than at a table top or using videos (e.g., Dhadwal et al., 2021; Najdowski et al., 2017; Najdowski et al., 2018; Suarez et al., 2021; Welsh et al., 2019).

A potential limitation of the current study is that we did not conduct the same skills assessment with each of the participants as a part of the study. Instead, we reported their skill levels based on the most recent skills assessment conducted as a part of their regular behavioral intervention program. Future research may consider administering the same assessment (based on verbal behavior or language/cognition) for all participants.

An additional limitation is that procedural integrity data were not collected. Because sessions were conducted in the natural environment with distractions occurring in the room, the experimenter was able to discreetly instruct the other experimenters and family members as to what behaviors to engage in prior to each trial without the participant noticing. In addition, during the training sessions, only one experimenter implemented procedures. Maintenance data were also not collected, therefore it is unknown if participants continued to exhibit their skills over the coming weeks and months postintervention. Future research should collect data on procedural integrity and maintenance of learned skills.

Another limitation is that the current study did not measure or teach participants to respond socially with others after identifying whether and how/why they had or did not have knowledge of information. Perspective taking refers to observing the behavior of others followed by responding to the private events they are presumed to have experienced and/or predicting their future behavior (LeBlanc et al., 2003). Future research could probe whether participants appropriately identify the likely future behavior of others based on their knowledge as well as whether participants’ exhibit a socially appropriate response after identifying what others know/do not know and what others are likely to do given/not given access to that information. If participants fail probes, this skill could be taught directly. In addition, future research may investigate exactly which steps need to be taught to obtain socially appropriate behavior. The targets in this study may not require teaching in order to result in meaningful social behavior. Even though this is a limitation of the current study, the skill of identifying what others know, in and of itself, appears to be relevant in social language and may be a skill worth targeting in its own right. For example, statements of others’ knowledge occur in conversations, for example, when saying “She doesn’t know the meeting location has been changed.”

Given the differing theoretical accounts for the acquisition of perspective-taking skills, future research could attempt to identify the necessary skills to teach versus those that may not need direct intervention in order to result in meaningful changes in social interactions related to perspective taking. For example, it is unknown which of steps 1–3 of the five steps ToM researchers believe to be involved in teaching “seeing leads to knowing” are required to be intact for individuals to exhibit step 4 (i.e., predicting future behavior of another person [“Where will he look for his bicycle?”] based on the information gathered in steps 1–3). Future research could also identify the necessary procedures to include during intervention. For example, the role that MET and/or rules play in the acquisition of perspective-taking skills could be examined. Although MET has been a common approach to teaching perspective-taking skills in the behavior analytic literature, future research could attempt to identify the relative value of MET. Likewise, whether rules are necessary or improve the effects of an intervention could be evaluated. By definition, rule-governed behavior suggests that when one has contact with the rules that describe contingencies, they can respond to rules effectively without needing to have direct contact with the contingencies described by the rule (Skinner, 1969).

In summary, identification of what individuals know is an important skill for successful social interaction. The use of a training package consisting of rules, multiple exemplar training, error correction, and reinforcement was efficacious for teaching the participants in the current study to identify what they and others know. Future research should explore how this skill relates to both social behavior and other skills within a larger perspective-taking repertoire.

Data Availability

Data will be made available on reasonable request.

Declarations

Conflict of Interest

The authors declare that they have no conflict of interest.

Informed Consent

Informed consent was obtained by all human participants using a consent form approved by Pepperdine University’s IRB.