Behavioral Interventions
Behav. Intervent. 16: 111±125 (2001)
DOI: 10.1002/bin.79
John T. Rapp*, Claudia L. Dozier and James E. Carr
University of Nevada, Reno, Nevada, USA
The pica of a 6-year-old girl diagnosed with autism was initially shown to persist in the absence of
socially mediated consequences. In an attempt to provide a competing source of oral stimulation, we
used a stimulus preference assessment to identify food items that were subsequently presented
noncontingently. However, the noncontingent schedule could not be thinned to a practical variation
while still maintaining reductions in pica. A subsequent multielement evaluation of response blocking
and verbal reprimands demonstrated that neither intervention both produced and maintained low levels
of pica. Verbal reprimands were then used in conjunction with noncontingent food presentation, but this
intervention did not produce signi®cant reductions in pica. Suppression of pica was ultimately obtained
in both a clinic setting and in the child's natural environment using contingent, varied auditory
stimulation. The results are discussed in the context of the `least restrictive alternative' model of
treatment selection. Copyright # 2001 John Wiley & Sons, Ltd.
Aberrant behavior maintained by automatic reinforcement has been deemed inappropriate for a variety of reasons (e.g., interference with educational training); however,
such behavior is sometimes permitted in controlled contexts (see, e.g., Charlop, Kurtz,
& Greenberg-Casey, 1990; Rapp, Dozier, Carr, Patel, & Enloe, 1999). However, pica, a
behavior often found to occur in the absence of social consequences, is not acceptable at
any level in any stimulus context due to health risks. These risks include intestinal
blockage, poisoning, and parasitic infection (Fisher, Piazza, Bowman, Kurtz, Sherer, &
Lachman, 1994; Foxx & Martin, 1975; Moncrieff, Koumides, Clayton, Patrick,
Renwick, & Roberts, 1964). Therefore, successful assessment and treatment of
automatically reinforced pica is of particular importance. Historically, however, this
behavior problem has been dif®cult to treat. As it is typically classi®ed, the behavior of
pica involves the ingestion of a wide range of inedible items. The variability in ingested
stimuli might contribute to the dif®culty in treating this behavior.
*Correspondence to: John T. Rapp, Psychology Department, University of Florida, Gainesville, FL 32611, USA.
E-mail: [email protected]¯.edu
Copyright # 2001 John Wiley & Sons, Ltd.
J. T. Rapp et al.
Piazza et al. (1998) recently conducted a multiphase study with three individuals
who exhibited severe pica. After a functional analysis suggested that each
participant's pica was maintained (at least in part) by a source of automatic
reinforcement, the authors conducted stimulus preference assessments with food
items that were similar (i.e., `matched') or dissimilar (i.e., `mismatched') to objects
the individuals typically ingested. During subsequent treatment evaluations,
continuous (i.e., noncontingent) access to matched stimuli reduced pica to nearzero levels for each participant. Similarly, Piazza, Hanley, and Fisher (1996) provided
noncontingent access to a preferred food item in an attempt to reduce the cigarette
pica of one participant. However, reductions in cigarette pica were observed only
after the addition of a mild punishment procedure (i.e., response interruption), which
was paired with the presentation of a purple card. In the presence of this card, pica
reductions maintained across four generalization settings.
In the present investigation, we ®rst conducted a functional analysis (Iwata,
Dorsey, Slifer, Bauman, & Richman, 1982) to determine the variable(s) maintaining
the pica of a young girl diagnosed with autism. Subsequently, we conducted a
stimulus preference assessment to identify food items that might compete with pica.
Finally, we attempted to treat the participant's pica using a noncontingent food (NCF)
procedure similar to that reported by Piazza et al. (1996). After this intervention
failed to eliminate pica, we evaluated the use of a variety of punishment procedures
according to the `least restrictive alternative' model of treatment selection (Johnston
& Sherman, 1993). Consequently, treatment progressed through the following
interventions based upon level of intrusiveness: (a) NCF (®xed-time (FT) schedule
delivery of a preferred food item) without punishment, (b) response blocking, and (c)
verbal reprimands (evaluated in a multielement design), (d) NCF combined with
contingent verbal reprimands (a `mild' punishment procedure), and (e) various levels
of contingent auditory stimulation (a more `intense' punishment procedure). The
result of this progression was the reduction of pica to near-zero levels across multiple
stimulus conditions.
Participant and Dependent Variable
Marisa was a 6-year-old girl diagnosed with autism who received 30 h of individual
training per week in a university-sponsored program. Her receptive language skills
consisted of approximately 100 words, while her expressive language was limited to
Copyright # 2001 John Wiley & Sons, Ltd.
Behav. Intervent. 16: 111±125 (2001)
Functional assessment and treatment of pica
approximately eight words and several pictorial mands. Marisa had an extensive
history of engaging in pica, which involved the ingestion of inedible objects (e.g.,
rocks, dirt, grass, balloons). Such objects were detected in Marisa's feces as often as
three times per week.
For the purposes of this investigation, pica was de®ned as the insertion of
simulated pica objects (which were placed on the ¯oor) into the participant's mouth.
Because of the potential dangers of consuming inedible objects, and the absence of
medical staff supervision, we chose to bait Marisa's therapy room with edible objects
(i.e., broken pieces of animal crackers) that resembled inedible objects found in
Marisa's therapy room (e.g., pieces of ceiling tiles).
Data Collection and Interobserver Agreement
All sessions in this phase were conducted in Marisa's therapy room (9 feet12
feet) and were 5 min in duration. Data were collected with a video camera and scored
using a real-time recording (RTR) procedure (Rapp, Carr, Miltenberger, Dozier, &
Kellum, 2001). RTR involves the second-by-second recording of target behavior(s)
and relevant events by scoring the exact time (from the VCR counter) of onset and
offset of the events. With this method, videotaped sessions were scored on a secondby-second basis for occurrence and nonoccurrence of the target behavior using a
scoring sheet marked for 300 s. Two independent observers scored 33% of the
videotaped sessions in this phase. The mean point-by-point interobserver agreement
(IOA) scores for occurrences and nonoccurrences of pica were 97% (range, 93.3% to
100%) and 99.5% (range, 99.3% to 100%), respectively.
Experimental Design and Procedures
A functional analysis (Iwata et al., 1982) was conducted to assess the variable(s)
that possibly maintained Marisa's pica. A multielement design was used to assess
levels of pica under four different experimental conditions. These conditions were no
interaction, attention, demand, and a control (play) condition. For each session, the
room was baited with approximately 50 items that could be ingested. All sessions
were conducted in the same room.
During the no-interaction condition, Marisa independently played with toys while
a therapist read Marisa's therapy book. The therapist did not interact with Marisa
(i.e., no consequences were provided for pica). This condition assessed levels of pica
in the absence of social consequences. In the attention condition, Marisa was also
given access to several toys while a therapist read Marisa's therapy book. Contingent
on instances of pica, the therapist delivered a verbal reprimand (e.g., `Marisa, not in
mouth'). This condition evaluated social positive reinforcement in the form of
Copyright # 2001 John Wiley & Sons, Ltd.
Behav. Intervent. 16: 111±125 (2001)
J. T. Rapp et al.
attention as a potential variable maintaining pica. During the demand condition, a
therapist presented Marisa with a nonpreferred academic training task (i.e., one that
was frequently correlated with problematic behavior). Contingent on each instance of
pica, the therapist withdrew the training task, turned away from Marisa, and withheld
academic instructions for 20 s. Following this break period, the therapist re-initiated
the task. Correct responses were followed by brief verbal praise, while other behavior
was ignored. The purpose of this condition was to evaluate whether social negative
reinforcement, in the form of the removal of a nonpreferred academic task, was a
variable maintaining pica. The control (play) condition involved providing Marisa
continuous access to toys in her room, as well as social interaction on an FT 15 s
Results and Discussion
Figure 1 shows high and undifferentiated levels of pica across the no-interaction
(Mˆ7.3 per min), control (Mˆ6.9 per min), and demand conditions (Mˆ4.3 per
min). In addition, slightly lower levels of pica were observed in the attention
condition (Mˆ2.6 per min) in which verbal reprimands were provided contingent on
instances of pica. Within-session analysis suggests that contingent social disapproval
Figure 1. Occurrences of pica across functional analysis conditions.
Copyright # 2001 John Wiley & Sons, Ltd.
Behav. Intervent. 16: 111±125 (2001)
Functional assessment and treatment of pica
may have served as a mild punisher, which increased the mean interresponse time
(IRT) of Marisa's pica during this condition. Similarly, Marisa was initially
compliant with prompts early in the demand condition sessions (e.g., for the ®rst
minute), but eventually engaged in pica. This initial compliance may account for the
slightly lower level of pica observed in this condition. Based on the functional
analysis, we concluded that Marisa's pica was maintained by non-socially mediated
variables (i.e., automatic reinforcement).
Dependent Variables
The de®nition of pica was the same as in phase 1. Food ingestion (scored separately
for each food item) was de®ned as the placement, with subsequent chewing and
swallowing, of a piece of food into Marisa's mouth.
Data Collection and Interobserver Agreement
Data collection, session length, and scoring were the same as in phase 1. In
addition, IOA was calculated in the same manner as in phase 1. A second observer
scored 30% of the sessions of this phase. The mean IOA score for nonoccurrences of
pica was 100% (no pica occurred during this analysis). Similarly, the mean IOA
scores for occurrences and nonoccurrences of food ingestion were 96% (range, 93%
to 99%) and 99.2% (range, 98.9% to 100%), respectively.
A paired-stimulus preference assessment (Fisher, Piazza, Bowman, Hagopian,
Owens, & Slevin, 1992) was conducted to identify food items that might compete
with the reinforcement produced by pica. Eight food items (graham crackers, carrot
sticks, bread sticks, strawberry gelatin, un¯avored gelatin, banana, rice cakes, and
tofu) were selected for this assessment based upon whether these items were similar
(e.g., hard, crunchy) or dissimilar (e.g., soft, chewy) to the various non-food items
Marisa typically ingested (similar to Piazza et al., 1998). These food items were cut
or broken into comparably sized pieces and then presented three times (on a plate), in
quasi-random pairings, across three separate days.
For each session, the two food items were presented to Marisa (after she was seated
at her dining table) on separate plates (side presentation was altered randomly). In
Copyright # 2001 John Wiley & Sons, Ltd.
Behav. Intervent. 16: 111±125 (2001)
J. T. Rapp et al.
addition, pica objects were placed on the ¯oor on both sides of Marisa to determine
whether the food items would compete with reinforcement produced by pica. During
these sessions the number of food and pica items ingested was recorded.
Results and Discussion
The results from the stimulus preference assessment show that Marisa consistently
selected and ingested rice cakes (Mˆ3.0 pieces per min), followed by strawberry
gelatin (Mˆ2.7 per min), graham crackers (Mˆ2.3 per min), bread sticks (Mˆ1.0
per min), and tofu (Mˆ0.3 per min). Pieces of carrot sticks, banana, and un¯avored
gelatin were never selected. The data indicate that Marisa consistently selected rice
cakes, thus suggesting that she had a relative preference for this food item.
Interestingly, although pica objects were available, and Marisa did not interact with
some pairs of items (neither item was selected), she did not engage in pica during any
of these sessions. Based on these results, rice cakes were used during the subsequent
NCF phase.
It is unclear why Marisa did not engage in pica when she was presented two food
items with which she did not interact. Because of the brevity of these sessions, it is
unknown how long Marisa would have remained seated in her chair and abstained
from pica (in the presence of what appeared to be two nonpreferred food items).
Based on clinical observations and the results obtained in the demand condition (see
phase 1), Marisa typically did not remain seated for periods greater than
approximately 5 min. During the demand condition, once Marisa left her chair, she
always engaged in pica. As such, we speculate that the absence of pica during the
paired-food sessions might have been the result of a prior history of receiving
preferred foods (when less preferred items were refused), close supervision and
prevention of pica during meals (when seated at a table), or both. In short, the absence
of pica during these sessions appeared to be the result of stimulus control of sitting
and proximity of the therapist, rather than the presence of two nonpreferred food
After conducting stimulus preference assessments for three individuals who
exhibited automatically reinforced pica, Piazza et al. (1998) provided noncontingent
access to the item identi®ed as most preferred for each. We attempted to
systematically replicate this outcome by providing Marisa access to rice cakes (the
most preferred food item from the stimulus preference assessment) on a dense
schedule to compete with pica. After obtaining reductions with a dense food
Copyright # 2001 John Wiley & Sons, Ltd.
Behav. Intervent. 16: 111±125 (2001)
Functional assessment and treatment of pica
presentation schedule, we then attempted to thin the schedule. When thinning was
ineffective for producing zero levels of pica, more intrusive interventions were
The use of contingent auditory stimulation (e.g., 110±130 dB) has been shown to
function as both a positive punisher and a negative reinforcer (Azrin, 1958). More
recently, less intense contingent auditory stimulation has been utilized via the
awareness enhancement device (AED) to eliminate automatically reinforced handto-head behaviors, such as hair pulling (Rapp, Miltenberger, & Long, 1998), and digit
sucking (Ellingson et al., 2000). When worn by the respective individual, the AED
emits an auditory stimulus (78 dB) contingent on the placement of the individual's
hand within 6 inches of his or her head. Because the AED delivers the prescribed
consequence independent of a social agent, it appears to be useful for reducing
behavior maintained by automatic reinforcement. However, the AED delivers a
contingent tone only after the behavior occurs for 1±2 s and, therefore, it is optimally
applied to hand-to-head behaviors (e.g., hair pulling, digit sucking) that occur for
extended periods of time. By contrast, pica is most often (as in the present case) a
discrete hand-to-head behavior. Due to this fact, the AED currently does not lend
itself well to detecting and delivering an auditory consequence for occurrences of
pica. Therefore, in this phase of the investigation, an auditory stimulus that was
controlled by a therapist was applied contingent on instances of pica during sessions
when Marisa was alone.
Dependent Variables
The de®nition of pica was the same as in previous phases. Pica attempts were
de®ned as the movement of either hand that contained a potential pica object above
Marisa's waistline. Pica approaches were de®ned as a combination of the following
behaviors: (a) Marisa moving either of her hands toward a pica object, (b) Marisa
looking down toward a pica object, (c) Marisa leaning forward toward a pica object,
and (d) Marisa bending at the knees when pica objects were at her feet.
Data Collection and Interobserver Agreement
Data collection and scoring were the same as in previous phases. In addition, IOA
was calculated in the same manner as in previous phases. A second observer scored
28% of the sessions. Mean IOA scores for the occurrences and nonoccurrences of
pica, pica attempts, and pica approaches (combined) were 93.6% (range, 92.5% to
100%), and 99.6% (range, 97.9% to 100%), respectively.
Copyright # 2001 John Wiley & Sons, Ltd.
Behav. Intervent. 16: 111±125 (2001)
J. T. Rapp et al.
Experimental Design and Procedures
Treatment sessions were conducted in Marisa's training room and were 5 min in
duration. Marisa's therapy room was baited with the same simulated pica objects that
were used in previous phases. Although toys were present in this room (placed in bins
and shelves), these items were not readily accessible to Marisa until the second
auditory stimulation phase was implemented. This alteration was made following our
observation that Marisa attempted to acquire these items after experiencing the
auditory stimulus (contingent on an instance of pica). Near the end of our treatment
evaluation, 15±20 min probes were conducted in Marisa's training room and in the
backyard of her home (a problematic setting for pica). A multielement design (to
evaluate the milder punishers) with a series of reversals was used to evaluate the
effects of four different treatments on the occurrence of pica.
Baseline. The no-interaction sessions conducted during the functional analysis
constituted the baseline phase. During these sessions the room was baited with
simulated pica objects and a therapist was seated at a desk engaged in work activities.
No consequences were provided for pica.
Noncontingent food (NCF). During the NCF condition, a therapist presented a
piece of a rice cake (determined in phase 2) on an FT schedule to Marisa during nointeraction sessions (a therapist interacted with Marisa only to provide the food item).
Based on the mean IRT observed during baseline, we initially presented the food item
on an FT 10 s schedule. We attempted to thin this schedule by 10 s when the level of
pica was at or below 0.2 per min for two consecutive sessions. The schedule was
made denser (i.e., returned to a previous schedule value) following two consecutive
sessions above the same criterion. Attempts to thin the schedule were terminated
following the failure of the initially thinned schedule (FT 20 s) to achieve the
criterion, a return to the original NCF schedule (FT 10 s), and a subsequent failure of
the thinned schedule.
Response blocking. In this condition, the therapist physically guided Marisa's
hand away from her mouth contingent on attempts to engage in pica. Otherwise,
the therapist did not interact with Marisa or provide consequences for other
Verbal reprimand. In this condition, the therapist delivered a statement of
disapproval (as in the attention condition during phase 1) to Marisa contingent on
pica attempts. Otherwise, the therapist did not interact with Marisa or provide
consequences for other behavior.
Copyright # 2001 John Wiley & Sons, Ltd.
Behav. Intervent. 16: 111±125 (2001)
Functional assessment and treatment of pica
NCF‡Verbal reprimand. This phase was conducted to assess the effects of a
combination treatment that consisted of verbal reprimands and NCF. Although both
NCF and verbal reprimands produced reductions in pica when implemented alone,
neither resulted in the elimination of pica. Given that response blocking and verbal
reprimands produced similar reductions in Marisa's behavior, the latter was selected
for use with NCF due to its relative ease of implementation. Based on the results
obtained during NCF, the criterion for thinning during this phase was altered
depending on the schedule in place. We initially started with same response criterion
used in NCF, but then reduced the rate requirement to zero for ®ve consecutive
sessions at FT 10 s. Thereafter, we attempted to thin this schedule by 5 s when the
level of pica was at or below 0.2 per min for three consecutive sessions.
Alone. This condition was conducted to assess levels of pica when Marisa was
alone. During these sessions, Marisa was alone in her therapy room and was viewed
by two graduate students via a small glass window and TV monitor. No social
consequences were provided for pica.
Contingent auditory stimulation After NCF, response blocking, and verbal
reprimands were evaluated, our treatment consisted of three separate phases of
contingent auditory stimulation. During these phases, a 4 inch6 inch speaker was
placed on the wall in Marisa's training room. Because a therapist was not required to
dispense edibles on an FT schedule, these phases of the study took place when Marisa
was alone. Marisa was observed on a monitor by a graduate student who controlled a
CD player that contained a disc of pre-recorded tones. During each session, Marisa
wore a yellow wristwatch that was intended to obtain inhibitory stimulus control over
pica (similar to Piazza et al., 1996) based on its pairing with the auditory stimulus.
However, it should be noted that the effectiveness of the wristwatch in obtaining
stimulus control over pica was not experimentally evaluated in this study.
In the ®rst auditory stimulation phase (AS1), an 83 dB tone was presented for 2±3 s
(on an FR-1 schedule) contingent on instances of pica and pica attempts.
During the second auditory stimulation phase (AS2), the tone and the target
behaviors remained the same; however, the volume of the tone was increased to
90 dB. In addition, during the third session of this phase, although levels of pica
began to subside, it became apparent that once an object was in Marisa's hand, she
would ingest it (despite the contingent auditory stimulus). As placement of the item
in Marisa's mouth appeared to be the terminal (reinforced) response in a chain of
behaviors, we decided to deliver the auditory consequence for pica approaches
(behavior that occurred earlier in the response chain) in order to prevent the terminal
pica response. Finally, it appeared that the wristwatch had not developed stimulus
control properties (i.e., Marisa engaged in pica while wearing the watch); therefore,
Copyright # 2001 John Wiley & Sons, Ltd.
Behav. Intervent. 16: 111±125 (2001)
J. T. Rapp et al.
we attempted to increase the salience of this stimulus object by placing it in a freezer
(to decrease its surface temperature) for several hours prior to each session. Although
Marisa would look at the watch after it was placed on her wrist, its temperature did
not appear to cause her any discomfort.
In the third auditory stimulation phase (AS3), the target behaviors remained the
same, but the auditory stimulus was changed such that four distinct tones were
presented in a quasi-random order during the 2±3 s delivery. This was similar to
the stimulus variation procedure reported by Charlop, Burgio, Iwata, and Ivancic
(1988). Due to the use of varied tones, the volume ranged from 90 to 94 dB during
this phase.
Assessment/Treatment probes. After producing zero rates of pica in Marisa's
therapy room, we assessed the effectiveness of the AS3 intervention during ®ve
extended probe sessions. Based upon caregiver report, the most problematic setting in
which Marisa engaged in pica was in her backyard during unsupervised playtime.
This playtime was typically provided for 15±20 min periods each day. Naturalistic
probes were initially 5 min in duration and were extended to 20 min. Next, we
simulated Marisa's backyard (in her therapy room) by providing stimuli that she
commonly ingested, as well as appropriate play objects, from that environment and
assessed the effectiveness of the AS3 intervention under these conditions. Marisa's
training room was furnished with one square foot of dirt, 40 square feet of sod, a
small trampoline (to simulate a larger trampoline that was in her backyard), and
several toys that were often available in her backyard. Finally, a naturalistic probe
was conducted in the backyard of Marisa's home. If pica was observed during any of
these sessions, the auditory stimulus was delivered as described in the AS3 condition.
This was done to help promote generality of the response suppression.
Results and Discussion
Figure 2 shows that following high levels of pica during the initial baseline
(Mˆ7.7 per min), NCF (Mˆ0.9 per min) on an FT 10 s schedule produced an
immediate reduction to near-zero levels of pica. However, this reduction failed to
maintain after the NCF schedule was twice thinned to FT 20 s. Moreover, to be of
clinical signi®cance, a reduction to zero levels of pica was desirable. After a return to
baseline resulted in highly variable levels of pica (Mˆ2.0 per min), neither response
blocking (Mˆ0.8 per min) nor verbal reprimands (Mˆ0.7 per min) produced
clinically signi®cant reductions in pica (the ®nal sessions of each condition indicated
increasing trends). After the third baseline phase (Mˆ4.3 per min), NCF combined
with verbal reprimands produced reductions in pica to near-zero levels during dense
schedules of NCF. However, similar to the original NCF phase (without verbal
Copyright # 2001 John Wiley & Sons, Ltd.
Behav. Intervent. 16: 111±125 (2001)
Functional assessment and treatment of pica
Figure 2. Occurrences of pica across baseline and treatment conditions.
reprimands), gradual thinning of the NCF schedule, even with FR-1 verbal
reprimands, did not maintain near-zero levels of pica.
Following high levels of pica during the ®rst baseline phase in which she was alone
(Mˆ7.3 per min), the implementation of AS1 failed to reduce Marisa's pica
(Mˆ11.1 per min). In fact, it is possible that the tone used in this condition positively
reinforced pica, evidenced by the higher-than-baseline rates that resulted. An
increase in the intensity of the auditory stimulus during the AS2 phase resulted in a
gradual reduction to zero rates of pica (Mˆ0.2 per min). A baseline probe conducted
after AS2 resulted in an immediate increase in pica to a mean of 4.4 per min. The reimplementation of AS2 was marked by an initial decrease in pica; however, a return
to zero rates of pica was not achieved. Problems with the auditory equipment during
this phase resulted in a variable schedule of punishment that closely resembled a
variable-ratio-3 schedule. It is speculated that the schedule produced by equipment
failure, the possibility that Marisa habituated to the tone, or both, accounted for the
persistence of Marisa's responding during this phase. It should also be noted that both
the use of the cold watch and the presentation of the auditory stimulus for
`approaches' began during the third session of this phase. This fact combined with
equipment problems renders interpretation of the data from this phase dif®cult.
Copyright # 2001 John Wiley & Sons, Ltd.
Behav. Intervent. 16: 111±125 (2001)
J. T. Rapp et al.
The introduction of AS3 was marked by the immediate and sustained reduction of
pica to zero or near-zero rates (Mˆ0.01 per min). In addition, extended probes of
15 min and 20 min also resulted in zero rates of pica. Similarly, 15 min probes that
were conducted in the training room with dirt and sod, and a naturalistic probe that
was conducted in Marisa's backyard, resulted in zero and near-zero rates of pica. We
hypothesized that zero rates of pica, as well as pica attempts (the auditory stimulus
was not presented), during the naturalistic probe may have been due to inhibitory
stimulus control exerted by the wristwatch. However, this effect was not experimentally assessed during this investigation.
Although toys were accessible (they were in bins on shelves) prior to the AS2
phase, Marisa never manipulated objects that were present in her room when pica
objects were simultaneously available. However, once reductions in pica were
produced with contingent auditory stimulation, the percentage duration of object
manipulation increased during the AS2 (Mˆ56.6%) and AS3 (Mˆ77.2%) phases,
and maintained during the extended sessions (Mˆ51.4%) and naturalistic context
probes (Mˆ94%). Moreover, following the contingent presentation of the respective
auditory stimulus, Marisa began to manipulate an object within 7 s (the time required
for her to walk across the room) for 43.8% of pica approaches, attempts, and
occurrences. This may suggest that pica was a highly preferred behavior in Marisa's
repertoire and that pairing the auditory stimulus with this behavior made it less
preferred (through punishment); thus, an increase in another highly preferred
behavior was observed. Put differently, when pica was unavailable to Marisa, toy
manipulation became a preferred stimulus event. Alternatively, object manipulation
may have been associated with the absence of the tone, though this position is less
tenable as Marisa did attempt to engage in pica while simultaneously manipulating
This study illustrates the progression from less to more intrusive interventions for
the persistent pica of a young child diagnosed with autism. When NCF, response
blocking, and verbal reprimands produced inadequate reductions in pica, a
contingent and varied auditory stimulus was utilized, which resulted in the
suppression of pica. However, there were at least four limitations of the treatment
evaluation that are worthy of discussion. First, the effects of the ®nal auditory
stimulus intervention (AS3) were not replicated. A withdrawal and a subsequent reimplementation of the AS3 intervention would have provided stronger evidence for
attributing the clinically signi®cant reductions of pica to this intervention. In
addition, a conclusion about the effectiveness of the AS3 could have been more
Copyright # 2001 John Wiley & Sons, Ltd.
Behav. Intervent. 16: 111±125 (2001)
Functional assessment and treatment of pica
convincing if the intervention had been carried out for a longer period of time.
Likewise, as with the AS2 intervention, it is possible that the suppressive effects
obtained with this intervention may have been lost following a return to baseline.
However, this intervention was conducted in a clinic setting, which involved time
constraints as well as caregiver opposition to an additional reversal.
Second, the delivery of the auditory stimulus contingent on `pica approaches',
beginning in the AS2 phase, was inconsistent with the response topography that
resulted in auditory stimulus delivery in previous phases. It is possible that blocking
and verbal reprimands may have been more effective had they been provided earlier
in the chain of behavior (i.e., contingent on pica approaches rather than pica
attempts). However, the treatment evaluation was conducted to reduce the occurrence
of a potentially dangerous behavior while still permitting other behavior; this
alteration was made to increase the potential effectiveness of the procedure.
Third, the use of simulated pica objects (i.e., edible objects) may have made it
dif®cult for Marisa to distinguish between objects that were consumable and items
deemed unacceptable (because they were on the ¯oor) for ingestion. However,
Marisa was only allowed to eat food items that were either delivered to her by a
therapist or were placed on a plate while she was seated at a table. In addition, Marisa
engaged in speci®c behaviors when she picked up objects from the ¯oor (actual pica
objects or baited food) that she did not when she retrieved objects from her plate or
from a therapist. These behaviors included jumping around the pica objects before
reaching to the ground, then quickly grabbing the object and running to a corner of
the room (presumably to avoid detection or a reprimand) to ingest it. These particular
behaviors suggest that Marisa discriminated between these `edible' objects and
inedible ones.
Finally, the acceptability of the procedures utilized in this study was not evaluated.
As a number of punishment procedures were used during the treatment evaluation,
assessment of treatment acceptability would have provided further support of the use
of more intrusive procedures for life-threatening behavior. In light of this limitation,
it should be emphasized that Marisa's guardian was always aware of and approved of
the programmed contingencies for Marisa's pica.
The results of this investigation contribute to the literature in several ways. First,
the use of NCF was shown to be ineffective for eliminating pica with schedules
thinner than FT 10 s. Data from this study suggest that even highly preferred stimuli,
provided on response-independent schedules, may not effectively compete with
automatically reinforced pica. Second, this study is one of very few to use varied
consequences (tones) to eliminate a target response. More speci®cally, this study
adds to growing literature on the use of contingent auditory stimulation to eliminate
automatically reinforced behavior. Third, the use of `natural stimuli' (i.e., dirt and
sod) represents a novel attempt to integrate relevant stimuli into analogue
Copyright # 2001 John Wiley & Sons, Ltd.
Behav. Intervent. 16: 111±125 (2001)
J. T. Rapp et al.
assessments in order to evaluate the generality of behavior suppression. Finally, as a
whole, this study provides a logical progression from less to more intrusive interventions for life-threatening behavior. Although reinforcement-based procedures are
obviously preferred approaches, the use of punishment procedures may become
necessary when less intrusive measures fail to eliminate dangerous behavior.
Unfortunately, unlike many behaviors maintained by socially mediated consequences, a rich and effective reinforcement-based technology for behavior that is
incompatible with pica has yet to be developed. Future research could evaluate the
bene®ts of and guidelines for using mild aversive interventions for life-threatening
behavior that is resistant to less intrusive procedures. Similarly, research could focus
on evaluating the presence of response hierarchies (e.g., Green & Striefel, 1988) so
that suppression of the aberrant behavior, as seen in this investigation (with more
intrusive interventions), might actually increase more socially desirable behavior
(e.g., toy manipulation).
John Rapp and Claudia Dozier are now at The University of Florida. James Carr is
now at Western Michigan University. The authors thank Kimberley Enloe and Ryan
Pabico for their assistance with data collection.
Azrin NH. 1958. Some effects of noise on human behavior. Journal of the Experimental Analysis of
Behavior 1: 183±200.
Charlop MH, Burgio LD, Iwata BA, Ivancic MT. 1988. Stimulus variation as a means of enhancing
punishment effects. Journal of Applied Behavior Analysis 21: 89±95.
Charlop MH, Kurtz PF, Greenberg-Casey F. 1990. Using aberrant behavior as reinforcers for autistic
children. Journal of Applied Behavior Analysis 23: 163±181.
Ellingson SA, Miltenberger RG, Stricker J, Garlinghouse M, Roberts J, Galensky TL, Rapp JT. 2000.
Analysis and treatment of ®nger sucking. Journal of Applied Behavior Analysis 33: 41±52.
Fisher WW, Piazza CC, Bowman LG, Hagopian LP, Owens JC, Slevin I. 1992. A comparison of two
approaches for identifying reinforcers for persons with severe and profound disabilities. Journal of
Applied Behavior Analysis 25: 491±498.
Fisher WW, Piazza CC, Bowman LG, Kurtz PF, Sherer MR, Lachman SR. 1994. A preliminary
evaluation of empirically derived consequences for the treatment of pica. Journal of Applied Behavior
Analysis 27: 447±457.
Foxx RM, Martin ED. 1975. Treatment of scavenging behavior (coprophagy and pica) by
overcorrection. Behaviour Research and Therapy 13: 153±162.
Green G, Striefel S. 1988. Response restriction and substitution with autistic children. Journal of the
Experimental Analysis of Behavior 50: 21±32.
Copyright # 2001 John Wiley & Sons, Ltd.
Behav. Intervent. 16: 111±125 (2001)
Functional assessment and treatment of pica
Iwata BA, Dorsey MF, Slifer KJ, Bauman KE, Richman GS. 1982. Toward a functional analysis of selfinjury. Analysis and Intervention in Developmental Disabilities 2: 3±20. (Reprinted in 1994, Journal
of Applied Behavior Analysis 27: 197±209.)
Johnston JM, Sherman RA. 1993. Applying the least restrictive alternative principle to treatment
decisions: a legal and behavioral analysis. The Behavior Analyst 16: 103±115.
Moncrieff AA, Koumides OP, Clayton BE, Patrick AD, Renwick AC, Roberts GE. 1964. Lead poisoning
in children. Archives of Diseases in Childhood 39: 1±13.
Piazza CC, Fisher WW, Hanley GP, LeBlanc LA, Worsdell AS, Lindauer SE, Keeney KM. 1998.
Treatment of pica through multiple analyses of its reinforcing functions. Journal of Applied Behavior
Analysis 31: 165±189.
Piazza CC, Hanley GP, Fisher WW. 1996. Functional analysis and treatment of cigarette pica. Journal of
Applied Behavior Analysis 29: 437±449.
Rapp JT, Carr JE, Miltenberger RG, Dozier CL, Kellum KK. 2001. The use of real-time recording to
enhance the within-session analysis of functional analysis data. Behavior Modi®cation 25: 79±93.
Rapp JT, Dozier C, Carr JE, Patel M, Enloe K. 1999. Analysis and treatment of automatically reinforced
behaviors displayed by a boy with autism. In Reinforcement-Based Interventions for Severe Behavior
Disorders, symposium presented at the meeting of the Florida Association for Behavior Analysis,
Iwata B (chair).
Rapp JT, Miltenberger RG, Long ES. 1998. Augmenting simpli®ed habit reversal with an awareness
enhancement device: a preliminary investigation. Journal of Applied Behavior Analysis 31: 665±668.
Copyright # 2001 John Wiley & Sons, Ltd.
Behav. Intervent. 16: 111±125 (2001)