We conducted two studies to examine parameters of social attention in contingency awareness training using switch activation with individuals who had multiple profound disabilities. In Study 1 we compared leisure devices and social attention as reinforcing stimuli with 5 individuals. Results indicated the reinforcing qualities of social attention over leisure devices with 2 individuals and documented the importance of session length in training. In Study 2 we investigated idiosyncratic behaviors as indicators of responsiveness with 3 of the 5 original participants as they activated switches. Behavior changes during switch activation versus nonactivation times in the leisure device and social attention conditions suggested volitional movement supporting contingency awareness and preference. Implications for clinical practice are discussed.
In this paper we describe the latest in a series of studies focused on improving methods for revealing the learning potential of individuals with multiple profound impairments. To assess learning potential with this population, researchers must use instruments designed to address whether the individuals respond to reinforcement contingencies, and, if so, whether they show a preference among reinforcers. That is, do some stimuli in the contingencies (i.e., social attention) produce behavior change more reliably than others (e.g., music)? People with multiple profound disabilities have severe limitations in their motor development, sensory systems, cognitive skills, and communication abilities (Kobe, Mulick, Rash, & Martin, 1994). These individuals are nonverbal and have no other reliable, symbolic means of communication (Snell, 2002). This combination of challenges significantly hinders their ability to obtain assistance with basic needs, indicate preferences, make requests, and other important communicative actions. Indeed, in numerous prior studies, many of these individuals showed behavior that represented purposeful responding or learning; most of them were completely dependent on caregivers for survival (e.g., Green, Reid, Canipe, & Gardner, 1991). This array of conditions results in a basic challenge to act on their behalf in an informed manner. Thus, particularly in adulthood, it is not uncommon for many individuals to spend much of their day on the periphery of formal activities, with little or no opportunity for independent behavior.
Despite the absence of foundational information, some dedicated activity coordinators and therapists do provide opportunities for independent behavior, usually in the form of an adaptive switch. An adaptive switch is a switch that is modified for unconventional modes of closure, such as a head turn, head raise, slight finger or palm pressure, and so forth. Therapists place these switches in locations where the individual might be able to make contact and close the switch with a volitional movement. In the event that such movement might occur, the therapist arranges for switch closure to produce pleasant and potentially reinforcing feedback, such as music from an adapted audio tape player or massage from a vibrating pillow. Many individuals in these programs close the switch from time to time and experience the pleasant feedback. In nearly as many cases, however, it is not clear whether the switch closures are voluntary or are incidental to uncoordinated movements, startle responses, or seizure activity.
Even formal testing for the volitional nature of the switch closures leads to inconclusive results. In formal studies, tests usually have two conditions: (a) extinction conditions, in which the switch is not connected to any source of stimulation and closures “presumably” produce no stimulus change (e.g., Realon, Favell, & Lowerre, 1990); and (b) conditions in which the switch is connected to a source of stimulation (e.g., tape player) and switch closures produce activation of the source. Other researchers have replaced the extinction condition with a condition in which stimulation is delivered noncontingently (e.g., Lancioni et al., 2003; Leatherby, Gast, Wolery, & Collins, 1992). Learning is inferred when switch closures consistently occur more often or for longer periods in the condition with contingent stimulation. A problem with interpretation of such data is that variables other than the arranged consequences may be controlling responding. For example, many adaptive switches produce sensory feedback directly, such as sudden changes in switch-plate pressure and audible “clicks” at the moment of switch closure (M. Saunders, Smagner, & Saunders, 2003). Because these brief events may be reinforcing, they may maintain responding in both test conditions, leading to the conclusion that there is no discrimination of the change in conditions and, hence, no learning or contingency awareness (e.g., Watson, 1966; R. Saunders & Saunders, 2005).
Another approach to testing for contingency awareness is to offer different forms of feedback across test sessions and observe for differential switch use, which would suggest preference (e.g., Wacker, Berg, Wiggins, Muldoon, & Cavanaugh, 1985; for research reviews, see Cannella, O'Reilly, & Lancioni, 2005; Lohrmann-O'Rourke & Browder, 1998). Similar problems in data interpretation arise with this method as well; when response rates are similar across several sources of stimulation, is it because they are enjoyed equally or is it because none of the responding is intentional? Thus, the dedicated therapist continues searching for a source of stimulation that will produce differential responding.
Social attention is often considered as one possibility for examining preference and, therefore, contingency awareness (Ivancik & Bailey, 1996; Leatherby et al., 1992; Sobsey & Reichle, 1989; Wacker, Wiggins, Fowler, & Berg, 1988). The typical protocol for social outcomes is to have the switch activate a signal or taperecorded verbal message. Wacker et al., for example, provided students with a switch that when closed activated an audiotape with a recording of the teacher's name. Social attention was provided for at least 5 s or for as long as the tape was activated. Results showed that 5 of the 6 students closed the switch as many as 800 times in a 15-minute session during contingent attention conditions. Sobsey and Reichle (1989) “bridged” switch closure and social attention with a buzzer rather than a taped message. They found not only that the buzzer plus attention increased responding, but further that social attention alone and the buzzer alone increased responding, relative to conditions with no scheduled contingent events.
Another approach to assessment of preference and contingency awareness is to examine indices of happiness (e.g., Favell, Realon, & Sutton, 1996; Green, Gardner, & Reid, 1997). Such indices include smiling, laughing, and sudden motor movements indicative of excitement. These indices have been applied to comparisons within sensory stimulation programs (e.g., Green & Reid, 1996), switch-use programs (e.g., Ivancik, Barrett, Simonow, & Kimberly, 1997), or both (Lancioni, O'Reilly, Campodonico, & Mantini, 2002). The prediction is that these indices will show different frequencies of occurrence across exposures to preferred and nonpreferred stimuli. Unfortunately, in the one study in which the researchers evaluated such indices in a group of individuals with minimal motor movement, none showed differences on the indices. A concern about this approach is that the indices of happiness have focused on conventional behaviors, assuming that these behaviors will capture the response of all participants. Given that individuals with multiple profound disabilities present with a complex array of idiosyncratic behaviors, perhaps an approach that examined changes in these individual behaviors would be more sensitive for investigating preferred stimuli.
Despite these challenges, assessment of learning and preference through switch closure and demonstration of contingency awareness is important in planning the education, treatment, and day-to-day support of persons with multiple profound disabilities. Discovering that social attention can be reinforcing is important because often traditional items such as food, liquids, or electronic leisure devices do not function as reinforcers. Switch closure established with social attention as the reinforcer also has the potential to be shaped into a response with communicative functions, such as signaling a request or a need (Schweigert, 1989; Schweigert & Roland, 1992). Furthermore, individuals with multiple profound disabilities have few opportunities to engage in dyadic interaction and often do not develop conventional social signals, such as smiling, when others interact with them (Scheigert & Roland, 1992). Thus, a first step in building a social interaction and communication repertoire could be establishing switch closure as a reliable means of gaining attention and interaction.
The foregoing discussion sets the context for our primary purposes in designing the following two studies: to (a) assess for the possible reinforcing effects of social attention for adults with multiple profound disabilities and (b) explore responses to social attention as a means of examining contingency awareness. The individuals who participated had not shown contingency awareness on prior tests with leisure-device activation as the consequence of switch closure. That is, these individuals were observed to close their switches as often when no leisure device was connected to the switch (extinction) as when switch closures activated a device. They also had not shown differential responding across many training sessions with adaptive switches and several different devices. Following Wacker et al.'s (1985) suggestion to examine social attention experimentally, in Study 1 we employed an alternating-treatments design, examining contingency awareness in training sessions across two switch-closure outcomes: social attention and leisure device. Second, Cannella et al. (2005) suggested that more research needed to be conducted to determine why preference assessments fail with some individuals. One hypothesis was that assessment session lengths have been too short for learning to occur or for evidence of preference to emerge. Thus, we conducted the present alternating treatments across three training-session lengths. In Study 2, we further explored the reinforcing effects of social attention during preference testing. This was accomplished by broadening the concept of indices of happiness to indices of anticipation, arousal, and happiness in a systematic comparison of changes in individual behavioral topographies across social and leisure device feedback conditions for periods proximate to and not proximate to switch closure. The results from Study 2 confirm and extend the results from Study 1.
Experimental Design Overview
We employed a single-subject alternating-treatment design with 5 participants to compare switch use when closures were followed by either contingent social attention or activation of a leisure device, such as a vibrator. During the social-attention condition, switch closures turned on a voice output device that played a short audiotaped request that elicited 10 s of attention from the experimenter. During the leisure-device condition, switch closure initiated 10 s of electronic leisure-device activation. Switch closures that occurred during the playing of the recorded message or during the consequence (attention, device activation) had no effect on the length of the consequences; therefore, these closures were nonfunctional. Three participants began in the social interaction condition and consequences were alternated thereafter; 2 participants began in device activation, with alternation following.
Switch use was monitored across three exposure lengths. Exposure Length 1 consisted of one session per day for 4 days. Each session was 15 min in length and was parsed into three 5-min phases across which the social and device outcomes were alternated. Thus, participants could produce device activation or social interaction for a total of up to 30 min each in Exposure Length 1. Exposure Length 2 consisted of one 15-min session per day across 6 days, but in each session either social attention or device activation was scheduled, but not both. Thus, participants could produce up to nearly 45 min each of social attention and device activation in Exposure Length 2. Exposure Length 3 consisted of 12 sessions, 1 per day of 15 min each. In the first 3 sessions, social attention was scheduled; in the second 3 sessions, the leisure device was scheduled. The sequence was then repeated, yielding a total of 90 min of training in each condition.
Participants and Setting
Five adults with profound multiple impairments who resided at a skilled nursing facility participated. They were nonverbal and unable to communicate with symbols, which was determined by current speech–language pathology reports from their medical files. Functioning level, known sensory status, and motor dysfunction are described in Table 1. All participants had several years of prior switch experiences in recreation therapy programs. Each participant's program was designed by the rehabilitation team after detailed preference testing and observation to identify the optimal type of stimulation as well as extensive switch trials to rule out inappropriate switch types and placements. Nonetheless, these individuals participated in these programs for years without demonstrating contingency awareness with leisure devices and, thus, without showing preferences among the specific devices. The results of a test for contingency awareness with one of our participants, and which are typical for this group, are shown in Figure 1.
In addition to the above criteria, we selected participants based on residential staff members concurrence that the individuals appeared to enjoy social attention. Concurrence was derived from the results of a social interaction profile completed on each resident by at least two staff members who worked with them on a daily basis. The profile provided us with a description of the resident's typical emotional behaviors (e.g., smiling or laughing) as well as potentially interactive behaviors (e.g., eye contact, motor movement in response to staff behavior). A copy of the profile is available from the corresponding author.
Equipment and Data Collection
Participants previously had been formally assessed to determine type and position of the switch for their switch-use programs. The leisure device for each of them was selected from archival reinforcer preference testing data. The devices chosen had not been shown to be preferred, but some switch closures occurred in activity sessions in which these devices were employed to deliver sensory consequences. Subsequent to selection, several informal sessions were conducted with the chosen device with each participant and observed for rates of use that were different than what the archival data had shown. All participants' pre-experimental rates were consistent with their archival data.
We derived the type of social interaction to which the participant positively responded from their social interaction profile and by observing the reaction of the participant to talking, singing, whispering, physical contact, verbal praise, and experimenter movement. Table 2 summarizes this information and includes a description of each participant's switch type and leisure device employed during the leisure-device conditions of the study. A battery-operated interface (see M. Saunders et al., 2003, for a description) was used to record all switch closures and session length. A second interface recorded only functional switch closures (i.e., closures that started the reinforcement timer). The interfaces were arranged electrically in line between the participant's switch and the voice output device or leisure device. The voice output produced by switch closure was a prerecorded message that said, “Come talk to me.” All sessions were videotaped.
We did not conduct a formal baseline with the leisure device. As stated above, pre-experimental rates of switch closure to operate the device were consistent with rates recorded in numerous activity periods in previous months and years. The earlier tests for contingency awareness, such as the one shown in Figure 1, were inconclusive. Although switch closures to activate the device may have been volitional, the tests failed to show that switch closures were a function of their consequences. Moreover, device activation did not appear to produce arousal or increased motor activity that might lead to involuntary switch closure because switch closure rates were no higher during activation conditions than during extinction conditions. Thus, in Study 1 the conditions in which switch closures produced activation of the leisure device were considered to be the null or baseline condition: We had no expectation that switch closure rates in the device condition would increase over the course of the study. We did, however, predict that higher or elevated switch closure rates in conditions with social outcomes would demonstrate a discrimination between the contingencies, with contingency awareness shown for the social contingency.
During the social interaction condition, the experimenter stood behind and out of sight of the participant. When the participant closed the switch, the voice output message was activated; the experimenter stepped into view directly in front of the participant, and provided 10 s, measured with a stopwatch, of social interaction. To avoid producing accidental switch closures, the experimenter approached from the side of the resident opposite the side with the switch. Thus, the participant would not turn into the switch at the sight of the experimenter. The experimenter also avoided interactions that might have prohibited subsequent switch use, such as tickling someone in a way that would result in body movement leading to being out of position to close the switch. During the leisure device condition, switch closure resulted in 10 s of activation of the participant's leisure device. A 5-in-1 Switch Modifier® (Enabling Devices) controlled the period of activation.
A prompt procedure was used with all participants. A prompt consisted of one assisted switch closure, a verbal statement notifying the participant of the consequence for switch closures, and the consequent 10-s reinforcement period. The prompt was provided at the start of each session and at the beginning of the second and third 5-min segment of a session if no switch closures had occurred since the previous prompt. This procedure ensured that the participant came into contact with changes in scheduled outcomes in Exposure Length 1 and served to demonstrate in Exposure Lengths 2 and 3 that the scheduled outcomes were not changing within each session.
Procedural fidelity was evaluated for 25% of the sessions for each participant and evaluated by comparing the experimenter's performance with the planned performance. An impartial observer reviewed the videotaped sessions, making judgments of compliance according to a plan sheet and recording checks for experimenter behavior that matched the plan and a minus for those omitted. The plan sheet detailed the equipment setup, use of prompts, target behavior (switch closure), and events following the target behavior (Billingsley, White, & Munson, 1980). Percentage of compliance was calculated by dividing the emitted plan components by the total number of intended procedural plan components and multiplying by 100. The overall percentage of compliance for procedural fidelity was 99%.
Rate of Switch Use and Test Length
None of the participants demonstrated substantial differences in frequencies of switch closure in Exposure Length 1 (not shown in a figure). As shown in Figure 2, and as we anticipated, none of the participants showed any systematic change in the rate of responding across device outcome sessions. However, as also shown in Figure 2, participant (P) 2 showed a consistent pattern of higher activation in social attention conditions than in leisure device conditions in Exposure Length 2. In Exposure Length 3, P2 again showed a higher frequency of switch closures in social interaction than for the leisure device conditions. P1, P3, and P4 continued to show no consistent differences across social and device outcome conditions in Exposure Lengths 2 and 3. Participant 5 showed an emerging preference for social interaction, however, in Exposure Length 3.
Nonfunctional Switch Closures
All participants produced some nonfunctional switch closures; that is, switch closures during the 10 s of social interaction or device activation. On average, for every functional switch closure produced by the participants, two nonfunctional switch closures were produced. The range is performance of P4, who produced 1 nonfunctional closure for every functional closure, to P1 who produced 2.7 nonfunctional closures per functional closure. As we report below, participants exhibited more overall motor movement during the 10-s period of activation than at times distant from functional switch closures. Thus, some of the nonfunctional closures might be explained by closures incidental to increased physical activity during activation. Second, the switches employed with these participants were very sensitive to contact. Because these individuals did not have well-coordinated movements, functional and nonfunctional closures could occur as a function of multiple closures in a burst.
The most important finding in Study 1 is that 2 of the participants (P2 and P5) showed through switch-closure data that each preferred social interaction as an outcome over activation of their leisure device. These demonstrations of preference were the first ever indications that these individuals could discriminate among switch-closure outcomes (i.e., demonstrate an awareness of the different contingencies). Stated in more behavior analytical terms, these individuals demonstrated that they could learn. Prior to this study, nothing in their clinical records supported such an inference.
The results also suggest that evidence of preferences is more likely to emerge in switch-use data when each test condition is presented for multiple consecutive sessions. The present results are not conclusive in that regard, however, as another interpretation would be that cumulative exposure to the two conditions was responsible (i.e., if Exposure Length 1 had been continued long enough, the effects seen with P2 and P5 would have eventually emerged without a change in exposure length).
The lack of differential responding for P1, P3, and P4 suggests that their responding may have been involuntary and that switch closures were accidental. Further, a related interpretation is that the activation of both the leisure device and the audiotaped message increased their overall arousal level, leading to excitation with increased motor movements, leading to additional accidental closures. That is, despite our inferences to the contrary from the previous tests for contingency awareness, device activation may have entrained motor movement.
Research to date suggests that some individuals with multiple profound disabilities can control events in their environment, as has been documented through adaptive switch studies and contingency awareness testing. Further, the switch-use methodology can reveal whether individuals have a preference for environmental control across several outcomes (Dittilo, 1986; Fisher et al., 1992; M. Saunders et al., 2005; Schweigert, 1989). This research has shed light on the cognitive and potentially communicative abilities of this population. Nevertheless, major complications to the research have been mixed results across participants, as in Study 1, or equivocal results within individuals. Thus, adaptive-switch assessments and interventions have promise, but further research on more reliable protocols is needed. More precisely, adaptive-switch interventions might be improved if they better captured the idiosyncratic nature and array of behaviors that individuals within this population produce during switch use. Study 2 was conducted to explore this hypothesis.
In Study 2 we focused on whether Study 1 had captured more than just switch-closure data; that is, was further information relative to the learning characteristics and preferences of the participants available? Specifically, because the sessions in Study 1 had been videotaped, we wondered whether close analysis of the video records would support, contradict, or expand the conclusions from the switch-closure data. To address these questions, we employed a series of observation, data-collection, and data-reduction protocols using the video available from Study 1. Limited by resources, however, in Study 2 we were restricted to follow-up on only 3 participants from Study 1. Participants 1, 2, and 3 were chosen (a) for homogeneity of experience, as each had begun Study 1 in the social condition; and (b) because the primary experimenter in Study 1 reported anecdotally that these 3 participants exhibited more physical activity that could be objectively measured than did P4 and P5 during the experimental sessions.
Data were collected from 60 minutes of videotaped sessions for each participant. The 60 minutes consisted of the final 15-minute session from each condition (social, device) from Exposure Length 3. There was one exception because Participant 3 did not complete the 12th session (i.e., session length < 15 min). In his case, data were collected from the 11th session. The Study 2 methodology consisted of the following steps: (a) description and enumeration of participant behaviors, with assessment of interobserver agreement; (b) assignment of these behaviors to predominant categories, with assessment of interobserver agreement; and (c) calculation of the rates of each category within various time segments across conditions (social and device).
Description and Enumeration of Behaviors
An observer, referred to hereafter as the primary observer, viewed the tapes and recorded descriptions of the behaviors exhibited by each participant. We chose this approach to allow the primary observer to document the array and number of idiosyncratic behaviors that the participants exhibited without the restriction of coding for a predetermined set of behaviors. Descriptions primarily included information about the body part that was moved and the direction, speed, and extent of the movement (e.g., moves head slightly to the right, raises arm slowly, and moves tongue in and out of mouth for approximately 5 s). Descriptions also included the sounds that the participant made, volume, and estimated duration. Switch activations were also recorded.
Transcription of observations
Observations were dictated while the videos were being viewed. Dictation was fed into a speech-recognition software program, Dragon Naturally Speaking®. This tool allowed for observations to be recorded into a text format in real time. Following completion of all structured observations, we inserted time markers into the text documents. One marker identified the moment of each functional switch closure. Other markers identified the start and stop of a “window of time” surrounding each functional switch closure. Times were taken from the digitized time displays on the videotapes. Each window of time, referred to hereafter as an activation window, was 40 s in length, consisting of a 20-s period immediately before a functional switch closure and a 20-s period immediately following the switch closure. We chose 40 s for a window based on the average amount of time that it appeared to take participants to initiate and complete movements, especially related to functional switch closure (about 20 s), and then adding an equal amount of time after functional switch closure. In instances where functional switch closures occurred less than 40 s apart, the total time between the two switch closures was divided in half. For example, if two switch activations occurred within 30 s of one another, the first 15 s became the postclosure period for the preceding switch closure and the second 15 s became the preclosure interval for the subsequent switch closure.
Agreement on description and enumeration
A protocol was followed to verify the content of the original observations recorded during the viewing of the videotapes. The verification was completed by examining portions of the primary observer's original transcriptions and comparing them to transcriptions created by a second observer. This observer independently collected data on two sessions per participant, or half that of the primary observer. Two 1-min segments per participant were selected for comparison between the two observers. The segments were selected based on the primary observer's ability to align the two transcripts for comparison. Overall, the two observers had 82% agreement (range = 79% to 88%). Disagreements generally occurred when one person included more instances of a behavior than did the other rather than a disagreement on which behaviors were produced.
Following completion of the description and enumeration of individual behaviors, these behaviors were grouped into categories based on similar attributes, which were classified primarily by the body part that was moved and the type of the movement or sound that was observed. For instance, all individual types of head movements were now grouped into one category, head movement. For each participant, the behavior categories that accounted for 5% or more of the total number of recorded behaviors were identified as predominant behavioral categories. Table 3 provides the definitions of the predominant behavior categories and examples of the behaviors that comprised them.
Agreement on identification of categories
A protocol was followed to verify the creation of predominant behavioral categories by the primary observer. This verification was completed by having a third observer independently view a selection of session videotapes for each participant and identify the main categories of movements and behaviors based on these observations. These behavior categories were then compared to the categories that were identified by the primary observer. For P1, the third observer identified six categories, including four of the five categories that the primary observer had identified. For P2, the third observer identified six categories, including five of the six categories that the primary observer had identified. For P3, the third observer identified eight categories, including five of the six categories that the primary observer had identified.
Agreement on categorization of behaviors
Next, a protocol was used to assess agreement for the quantitative data. This was done to verify the primary observer's behavioral coding (i.e., placement of individual behaviors into categories). For this procedure, a fourth observer independently reviewed the primary observer's transcript for one session for each participant. Her task was to place the participant's recorded behaviors into the primary observer's operationally defined predominant behavioral categories. We analyzed interjudge agreement for placing behaviors into categories using Cohen's kappa, a statistic that considers chance within the calculation of proportion of agreement between two independent observers (Hollenbeck, 1978). For this statistic, kappa values of .75 and above are considered excellent (Cicchetti & Sparrow, 1981). For P1, P2, and P3, Cohen's kappa coefficients of .85, .92, and .90, respectively, were achieved.
Calculating Behavior Rate
Finally, the rates of behaviors in each predominant behavioral category were calculated across (a) time segments (activation windows and times of nonactivation), (b) pre- and postswitch closure periods within the activation windows, and (c) within activation windows by condition (social and device outcomes).
Five predominant behavioral categories were identified for P1. These five categories accounted for 541 (89.87%) of his 602 total behaviors. Six categories were identified for P2, accounting for 530 (91.07%) behaviors of her 582 total behaviors. Six categories were identified for P3, accounting for 286 (80.08%) of his 354 total behaviors. The 3 participants shared four predominant behavioral categories; head movement, eye movement, mouth movement, and smile. P1 and P3 shared the category arm movement, and Ps 2 and 3 shared the category vocalization. Participant 2 was the only individual to have breathing (i.e., altered respiration) as a predominant behavioral category.
Figure 3 presents the rate data from the activation windows contrasted with all other time in the sessions for each participant. This figure shows that for P1, the behaviors for all five predominant behavioral categories occurred at a higher rate during the activation windows than during times of nonactivation. For P1, the rate of head movements during activation windows was higher by nearly 5 responses per minute (i.e., 6.8 vs. 2.2) than during times of nonactivation. On the other hand, the rate of smiling during activation windows was higher by a difference of only 0.9 responses per minute. Yet, that difference makes the response rate during activation about four times the rate during nonactivation times—a larger relative change than with head movement. Thus, the differences across time periods can be viewed from two perspectives: relative and absolute. For P1, these data came from 16.8 min of activation windows and 42.5 min of nonactivation time. Figure 3 shows that for P2, behaviors from five of the six predominant behavioral categories occurred at a higher rate during activation windows than during times of nonactivation. For this participant, these data come from 30.5 min of activation windows and 29.5 min of nonactivation time. Figure 3 also shows that for P3, behaviors from all five predominant behavioral categories occurred at a higher rate during activation windows than during times of nonactivation; these data come from 13 min of activation windows and 46.7 min of nonactivation time.
We used a Poisson regression to determine whether the difference between the two segments was statistically significant. A Poisson regression is a statistical model that is used to describe data that are generated from rate data across conditions from segments of time of unequal length. Data from all participants were combined, yielding a statistical analysis of the group as a whole. The coefficient estimate was .87, p < .001, showing that the rate of behaviors was higher during the activation windows than during times of nonactivation and that this difference was statistically significant.
Figure 4 displays the rate data for the two time periods (pre- and post-switch closure) within the activation windows. As can be seen, behaviors for P1 in four of the five behavioral categories occurred at a higher rate during the preclosure time period than during the postclosure period. Similarly for P2, behaviors in four of the six behavioral categories occurred at a higher rate during the pre- than during the post-switch time period. The two categories with the highest rates, head movements and eye movements, showed a slight reverse pattern. Overall, however, P2's pre- and post-switch closure data are quite similar. Results for P3 in the figure show that behaviors from five of the six predominant behavioral categories occurred at a higher rate during the post- than during the pre-switch time period. Behaviors from the category arm movements, which was the category with the greatest amount of activity overall, showed the reverse pattern, in which behaviors occurred at a higher rate during the pre- than the post-switch time period.
Figure 5 displays the data for the two feedback conditions (social vs. device) during the activation windows for each participant. Only data from the activation windows were analyzed because all participants generally showed higher rates of predominant behavioral categories during the activation windows than the times of nonactivation. As can be seen, for P1, four of the five predominant behavioral categories occurred with a higher rate in the social condition than the nonsocial or device condition. For P2, four of the six predominant behavioral categories occurred at a higher rate during the social condition. For P3, all five predominant behavioral categories occurred at a higher rate during the social condition. A Poisson regression was used to determine whether the differences between the two conditions were statistically significant. Data from all participants were combined for this analysis. The coefficient estimate was .84, p < .001, showing that the rate of behaviors was significantly higher during the social feedback condition than during the device feedback condition.
The behavioral observations and descriptions resulted in identification of a wide variety of movements and sounds produced by each participant in Study 2. They were generally active throughout the sessions, consistently producing observable changes in behavior. These behaviors consisted primarily of movements of body structures (i.e., head, arm), but also included vocalizations. All participants produced behaviors with multiple body parts, demonstrating an array of motor movements. In addition, they all demonstrated variation of these movements and sounds in direction, degree, and duration.
There were similarities and differences in the types of behavior categories that were identified across participants. The primary similarities included the categories head movement, eye movement, mouth movement, and smile. The similarities may indicate that certain body structures are more likely to be active for all individuals or that only limited sets of behaviors can be identified with a view of just the upper body. The differences in categories across participants and individual behaviors sorted into each category per participant, however, illustrate the individuality of each participant.
Examination of the frequencies of behaviors for all categories illustrates several points. First, by looking at each individual, we see that a small set of behavior categories comprise the majority of the total recorded behaviors. That is, several categories account for most of the individual behaviors; arguably, these are the most significant categories. Second, all participants demonstrated an overall pattern of higher rates of behavior during activation windows than during times of nonactivation. Third, all participants showed some differences in the rates of behaviors between times directly before switch closure versus times directly following switch closure, but the patterns were not consistent. Participant 1 showed a higher rate of behaviors in the preclosure time period, P3 in the postclosure period, and P2 showed small and inconsistent differences. Similarly, P1 and P3 showed a higher rate of behaviors in the social condition than in the nonsocial condition, and P2 showed a difference only in one category, head movement.
It is fortuitous that P2 was included in Study 2, because her data may set the benchmarks for interpreting the data of P1 and P3. In Study 1, P2 showed, through switch-use data, that (a) she preferred social stimulation and through that demonstration, that (b) her functional switch closures were volitional. Thus, her Study 2 data make sense. First, she had increased activity levels around functional switch closures, suggesting anticipation of the effects of switch closure and arousal as a function of the subsequent stimulation (of either type). Her data in Figure 2 suggest that anticipation and arousal played equal parts in her activity levels. Her data in Figure 3 suggest that she enjoyed both outcomes, but given a choice (Study 1, Figure 1), she would respond more frequently when the outcome was social. From a clinical perspective, the composite data set indicates a positive prognosis for further treatment and skill development.
Participant 1 did not show a preference in Study 1. The combination of increased activity levels during activation windows (Figure 2) with more of that activity prior to switch closure than after (Figure 2), however, suggest some degree of intentionality in switch closure. His data in Figure 3 suggest that the majority of his increase in post-switch-closure activity (Figure 2) derived from postclosure social-interaction periods. Thus, from a clinical perspective, the composite data set suggest that contingency awareness may be nascent, that social interaction has potential for becoming a reinforcer, and, thus, one should continue with data-based intervention with this individual.
Participant 3 also did not show a preference in Study 1. Like P1, P3's increased activity levels proximate to functional switch closure suggest a possible volitional aspect to his switch closures (Figure 3). His data as shown in Figures 4 and 5, taken together, suggest that there is considerably more arousal from social stimulation than from device activation. There is one notable exception in the data from Figures 4 and 5. In Figure 4, it can be seen that P3's arm movements are greater in the preclosure than in the postclosure period. Re-review of the videotapes revealed that switch closures in the social condition were frequently preceded by a pronounced lifting of the left arm and a turning of the torso and head into the switch on the right. Switch closures in the device condition were less often preceded by the exaggerated arm movement. A clinically sound interpretation is that switch closures in the social condition were “recruited” from volitional arm and torso movement; switch closures in the device condition were more likely incidental to simple head-turning. This interpretation is supported by the data displayed in Figures 4 and 5 on mouth movements. Re-review of the videotape also showed exaggerated opening of the mouth occurring concurrently with the exaggerated arm movement, thus, evidence of a multi-element recruitment pattern. Clinically, these data yield another excellent prognosis.
The foregoing interpretations are just that— interpretations. Are they correct? Only additional, systematic, data-based interventions will answer those questions. Yet, they do provide valuable information with considerable face validity that would be extremely useful for clinicians. Insight into the capabilities and preferences of this population is a challenge to obtain. The methodology we employed in these studies suggests a way to begin revealing information about individuals with multiple profound disabilities that heretofore has been undiscovered. Even so, the methodology should be viewed as an evolving one, particularly one with several practical issues to be addressed. For one, much work remains for standardizing a protocol for assessing contingency awareness, reinforcer preferences, and any special effects that contingent social attention brings to the assessment equation. Second, the technology for data collection and data analysis needs to have a better user interface. Our data-collection interfaces were built solely for our research use by a medical instruments developer. Similar devices are not yet available to clinicians via adaptive equipment vendors. Further, as we build an assessment protocol, we must also build a user-friendly data-interpretation protocol to guide clinicians as well as family members and others. Through these steps, we have the potential to reveal and build upon a previously undiscovered learning potential in many more individuals with multiple profound disabilities and to expand their ability to interact with the environment and others.
The authors are indebted to the staff of Program Area Team N at Fircrest Regional Habilitation Center (Seattle, WA) for their support of the research project. The study also was supported in part by funds from National Institute of Child Health and Human Development Grants HD018955 and HD02528 to the Schiefelbusch Institute for Life Span Studies, University of Kansas. Requests for reprints should be sent to Richard R. Saunders, Life Span Institute, 1052 Dole, University of Kansas, 1000 Sunnyside Ave., Lawrence, KS 66045. firstname.lastname@example.org