acknowledging essentials, inferences, and all that you are ... · acknowledging essentials,...

Gregory P. Hanley. Ph.D., BCBA-D

Acknowledging essentials, inferences, and all that you are willing to not know when

functionally analyzing behavior

BABATOctober, 2016

For more information and free materials, go to:

www.practicalfunctionalassessment.com

Be wary of

incontrovertible truths.

“All models are wrong;

some are useful.”

Box & Draper, 1987, p. 424

Functional assessment

process to determine the variables influencing problem

behavior

Some Assumptions

Problem behavior is an operant

Certain situations potentiate certain consequences

Goal of a Functional Assessment

Identify the consequences that maintain problem behavior

Identify the situations that evoke the behavior

In order to treat problem behavior

Functional Assessment Process

Functional AnalysisObserve while manipulating

Indirect Assessment Interview

Descriptive AssessmentObserve

Defining features of the

Standard Functional AnalysisMultiple test conditions

Uniform test conditions

Isolated test contingencies

Reinforce dangerous behavior only

Toy-play control condition

Pro

ble

m B

eh

avio

rP

er

Min

ute

Sessions

Example of a

standard functional analysis

How do we know this is the standard functional analysis?

(Jessel, Hanley, Ghaemmaghami, 2016)

1965-2000 (Hanley et al., 2003)

64% SFAs1 out 3 with modifications

2001-2012 (Beavers et al., 2014)

85% SFAs1 out of 7 with modifications

Multiple test conditions





Is the Standard Functional Analysis Effective?

Does it lead to a differentiated analysis?

Literature reviews:

Hanley et al. (2003): approx. 94%

Beavers et al. (2014): approx. 92%

Case series:

Hagopian et al. (2014): 47%

Slaton et al. (2016): 44%


Does it leads to larger treatment effect sizes?

Campbell (2003)

Higher PZD when Rx was based on “EFA”

But, these larger effects were almost exclusively obtained when researchers implemented the treatments in controlledsettings under rich schedules of reinforcement


Not one study showing practical outcome in relevant context (e.g., school, home, community) with only a function-based treatment

when a standard functional analysis was used


Not one study showing the social acceptability

of the process



Not one study showing a socially-validated effect on problem behavior


Apparent solution to ineffectiveness:

Excessive elaboration

Elaboration of the standard functional analysis (SFA)

Prior to a SFA

Formal Descriptive assessments

Preference assessments

Demand Assessments

Manuals outlining extensive team-based processes Following a failed SFA

Slight and systematic deviations from the SFA core procedures

Following a failed SFA-based treatment

Stimulus avoidance assessments

More preference assessment and reinforcer assessments

Despite the Excessive Elaboration of the Standard Functional Analysis…

Not one study showing…

…a practical outcome in a relevant context

…the social acceptability of the process

…a socially-validated effect on problem behavior

“Since all models are wrong, the scientist cannot obtain a ‘correct’ one by excessive elaboration.

Just as the ability to devise simple but evocative models is the signature of the great scientist so overelaboration is often the mark of mediocrity.”

George Box, 1976, p. 792

Research to practice gap Possible reason:

Because the outcomes are mediocre even when the process is elaborate

Functional analysis has been around for approx. 50 years (e.g., Lovaas et al., 1965; Sailor et al., 1968)

Standard functional analyses have been around 34 years (Iwata et al., 1982)

Over 300 studies containing over 500 standard functional analyses have been published (Jessel et al., 2016)

Yet, 55 to 65% of practitioners recently surveyed reported never conducting a functional analysis (Oliver et al., 2016; Roscoe et al., 2015)

Interview-informed

Synthesized Contingency

Analysis

IISCA

Standard

Functional

Analysis




Reinforce dangerous behavior


Interview-Informed

Synthesized Contingency

Analysis

Single-test condition

Individualized test conditions

Synthesized contingencies

Reinforce precursors to and

dangerous behavior

Test-matched control

Sessions

1 2 3 4 5 6

0

1

2

3

4

Escape/Tangible/Attention

Zeke

Pro

ble

m B

ehav

ior

per

min

An IISCA

1. An open-ended interview is always part of the process

Primary goals are to identify:

a) co-occurring topographies of problem behavior

b) events/interactions that appear to routinely evoke problem behavior

c) interactions that follow problem behavior and are reported to stop it

Interviews allow for discoveries

which can then be verified (or not) in the IISCA

Some Important Aspects of the IISCA

2. We synthesize multiple contingencies into one test condition

which contingencies and the specific materials and interactions are informed by the interview

Some Important Aspects of our Approach

Why might problem behavior occur?Single contingencies:1. Attention or toys (social-positive reinforcement)

2. Escape/avoidance (social-negative reinforcement)

3. Sensory/non-social (automatic reinforcement)

Combinatorial contingencies:

1. Attention and Toys

2. Escape to toys

3. Escape to toys and attention

4. Escape to automatic reinforcement

5. Compliance with mands

6. Escape to access to rituals, preferred conversations

7. Etc…..

Case Example (Bob, 8 yo, dx: Autism)Therapist: Sandy JinSetting: Clinic

Inextricable synthesis

Hypothesis:

Bob engages in meltdowns and aggression in order to obtain:

“His way” in the form of escape from adult instructions and access to preferred ways of interacting with electronics or academic materials

Pro

ble

m B

ehav

ior

per

Min

Sessions

0

1

2

3

4

5

6

1 2 3 4 5

0

2

4

6

8

10

12

14

Escape /

Tangible

Escape /

Tangible

Bob(Ipad context)

Bob(Math context)

Elective synthesis

(with initial verification)

Hypotheses:

Gail engages in meltdowns and aggression in order to obtain:

preferred (tangible) items and maternal attention

0

1

2

3

4

Tangible /

Attention

Analyst

Mother

Analyst

Mother

Analyst

Gail

Pro

ble

m B

ehav

ior

per

Min

0

1

2

3

4

Tangible

Sessions

2 4 6 8 10 12 14

0

1

2

3

4

Test

Control

Meltdowns

Col 46 Attention

Case Example (Gail, 3 yo, dx: PDD-NOS)Therapist: Nicholas VanselowSetting: Clinic

Why synthesize?

1. Seems to emulate the ecology better

Why synthesize?


2. Isolated contingencies sometimes do not control behavior whereas synthesized contingencies do.

• Call et al., 2005

• Dolezal & Kurtz, 2010

• Hanley et al., 2014

• Ghaemmaghami et al., 2016

• Mueller et al., 2005

• Slaton et al., 2016

• Slaton et al., 2016


0

2

4

6

8Test

Control

Food,attention

Addison

Attention

Tangible (food)

Control

0

2

4

6

8

10

Escape totangibles,attention

Jay

Control

Attention

Tangible

Escape

0

2

4Escape tocompliance withrequests

Franklin

Escape

Requests

Control

2 4 6 8

0

1

2

3

4Escape tocompliancewith requests

Ned

2 4 6 8 10 12

Control

EscapeRequests

IISCA Decoupled IISCASynthesized Isolated SynthesizedType of

contingency

Pro

ble

m b

ehav

ior

per

min

Sessions

Slaton et al., 2016

Why synthesize?


2. Isolated contingencies sometimes do not control behavior whereas synthesized contingencies do.

3. Doing so leads to effective action—meaningful treatment effects

– Hanley et al., 2014, Santiago et al., 2016; Ghaemmaghami et al., 2016

Some reasonable questions:

Have IISCAs been replicated?

(I.e., Do they have generality?)

Yes.

From Jessel, Hanley, and

Ghaemmaghami (JABA, 2016)

0

4

8

12Will

TestControl

Wayne Allen Kat (Cxt 1)Sam

0

2

4

6Jack (Cxt 1) Keo

Kristy Jim

Roxy

0

2

4

6Alex (Cxt 2) Chris

Jeff Zeke Kat (Cxt 2)

0

1

2

3

4 Mike Mitch

Gary Jian Earl

0.5

1.0

1.5

2.0

0

Paul Dan

Alex (Cxt 1) Beck

Sid

2 6 10

0.5

1.0

1.5

2.0

0

Lee

2 4 6

Steve

1 3 5

Jesse

1 3 5

Carson

1 3 5

Jack (Cxt 2)

Sessions

Pro

ble

m b

ehav

ior

per

min

Sessions

Pro

ble

m b

ehav

ior

per

min

ute

From Rajaraman et al. (2016)

Has the process been socially validated?

Yes.

Social Acceptability Questionnaire Results

Ratings

Questions Gail Dale Bob Mean

1. Acceptability of assessment procedures

7 7 7 7

2. Acceptability of treatment packages

7 7 7 7

3. Satisfaction with improvement in problem behavior 7 7 6 6.7

4. Helpfulness of consultation 7 7 7 7

Note. 7=highly acceptable, highly satisfied, or very helpful

1=not acceptable, not satisfied, or not helpful

from Hanley et al., JABA, 2014

Has the process been socially validated?

Yes.

from Santiago et al., JADD, 2016

Have socially validated treatments been developed from the IISCA?

Have socially validated effects been achieved from the IISCA?

Social Acceptability Questionnaire Results

Ratings

Questions Gail Dale Bob Mean

1. Acceptability of assessment procedures

7 7 7 7

2. Acceptability of treatment packages

7 7 7 7

3. Satisfaction with improvement in problem behavior 7 7 6 6.7

4. Helpfulness of consultation 7 7 7 7

Note. 7=highly acceptable, highly satisfied, or very helpful

1=not acceptable, not satisfied, or not helpful

from Hanley et al., JABA, 2014

Socially validated treatments and outcomes.

Yes.

Socially validated treatments and outcomes.

Yes.


0

1

2

3

4

0

1

2

3

4

Com

ple

x F

CR

per

min

0

1

2

3

4

Rein

forcem

ent (%

)

0

20

40

60

80

100

Pro

ble

m B

ehav

ior

per

min

0

1

2

3

4

Sessions

5 10 15 20 25 30 35 40 45 50 55 60 65

Res

ponse

to

Inst

ruct

ions

(%)

0

20

40

60

80

100

Treatment Extension

1 2

Compliance

Noncomp.

Levels3

BL FCT + EXT Denial and Delay Tolerance Training

Simple FCR Complex FCR

Zeke

Sim

ple

FC

R

per

min

Tole

rance

Res

ponse

per

min

Response Chaining


Treatment:

Unpredictable and intermittent reinforcement of

communication, toleration, and compliance

Implemented by relevant caregivers in relevant contexts who impose relevant and historically challenging routines

Let’s compare.Standard

Functional Analysis IISCA

Number of differentiated analyses 500+ x 34

Social validation of assessment process x

Social validation of derived Rx x

Social validation of outcome x

Towards a more efficient analysis

0 200 400 6000 200 400 600

IISCA

Latency-based

Brief

Trial-based

Other

Standard

Single-testIISCA

N = 115

N = 456

N = 21

N = 64

N = 6

N = 34

N = 8

Analysis duration (min)

Func

tiona

l an

alysi

s fo

rmat

Min Max

Jessel et al., 2016


Functional AnalysisIISCA





Efficiency x (average: 245 min) (average: 25 min)

Levels of Analytic Control

Strong Test condition: Consistently elevated rates Control condition: Zero or near-zero rates

Moderate Test condition: Some zeros or near-zero ratesControl condition: Zero or near-zero rates

Weak Test Condition: Variable but higher ratesControl condition: Lower but non-zero rates

Pro

ble

m B

ehav

ior

per

Min

Sessions

1 2 3 4 5 6

0

1

2

3

4

Test

Control

Pro

ble

m B

ehav

ior

per

Min

Sessions

1 2 3 4 5 6 7 8 9 10

0

1

2

3

4

Test

Control

Pro

ble

m B

ehav

ior

per

Min

Sessions

1 2 3 4 5 6 7 8 9 10

0

1

2

3

4

Test

Control

Interpretation and Implications of Varying Levels of Analytic Control

Strong experimental control

Interpretation: Have access to all reinforcers and EOs

Implication: Can turn off problem behavior with reinforcement; should achieve meaningful outcome with only function-based Rx

Pro

ble

m B

ehav

ior

per

Min

Sessions

1 2 3 4 5 6

0

1

2

3

4

Test

Control

Interpretation and Implications of Varying Levels of Analytic Control

Weak experimental control

Interpretation: All reinforcers for problem behavior are not identified

Implication: Variable responding will persist throughout skill development, probably necessitating punishment or arbitrary reward system

Pro

ble

m B

ehav

ior

per

Min

Sessions

1 2 3 4 5 6 7 8 9 10

0

1

2

3

4

Test

Control

The necessity of punishment when function-based treatments are made more practical is commonly reported:Fisher et al., 2003, Hagopian et al.,1998, Hanley et al., 2005, Wacker et al., 1990

Towards a more controlled analysis

0 25 50 75 100

IISCA

Latency-based

Brief

Trial-based

Other

Standard

Single-testIISCA

N = 21

N = 64

N = 6

N = 34

N = 8

no weak moderate strong

N = 115

N = 456

Percentage of applications with

control

Functio

nal

anal

ysi

s fo

rmat








% of analyses with strong control x (26%) (92%)

But these conclusions are from literature reviews.

Within-subject comparison is best next step.

From Fisher et al., 2016Differentiation SFA: 4 of 5IISCA: 4 of 5

Escape to Tangibles & Attention

Escape / Tangible

Disagreement

Many believe that a standard functional analysis is a goldstandard.

I think: Oh my.

Oh my science.

Main assertion

When there are different outcomes between two analyses, the one that has been around longer does not necessarily trump the findings of the other.

The truth can be found in

effective action.

- Treatment effects

- Speed of process

- Acceptability/adoptability of practice

Slaton, J., Hanley, G., & Raftery, K. (2016)

IISCA vs. Standard Analysis

0

1

2TestControl

Escape to tangiblesand attention

Tangible

Ignore/Alone

PlayEscape

Attention

0

1

2

3Escape to tangiblesand attention

1 2 3 4 5

0

1

2

3Escape totangibles

5 10 15

0

2

4 Escape totangiblesand attention

Diego

Mason

Riley

Pro

ble

m b

ehav

ior

per

min

Sessions

IISCA Standard IISCA

IISCA vs. Standard Analysis

0

1

2

Escape topredictableschedule

Control

Test

EscapePlay

AttentionTangible

Alone

1 2 3 4 5

0

1

2Escape to tangibles,stereotypy, andattention

5 10 15 20 25

0

2

4

6

8

1 2 3 4 5

0

1

2

3Escape to tangiblesand attention

5 10 15

Pro

ble

m b

ehav

ior

per

min

Kyle

Jonah

IISCA Standard

Sessions

IISCA vs. Standard AnalysisIgnore/Alone

EscapeTangible

Play

Attention

0

2


0

2


0

2

4Escape to rituals

2 4 6

0

1Escape to tangibles

5 10 15 20

Pro

ble

m b

ehav

ior

per

min

Sessions

IISCA Standard

Dylan

Emily

Chloe

Jeff

0

2

4

ControlTest

Escape totangibles

Treatment Comparison Results

0

1

2

3

4

5

FCR

BL FCT + EXT

Escape totangibles,attention

Problembehavior

BL FCT + EXT

Escape

5 10

0

1

Escape totangibles

BL FCT + EXT

5 10

Attention

BL FCT + EXT

IISCA- based treatment Standard-based treatment

Pro

ble

m b

ehav

ior

per

min

Sessions

Emily

Jeff

Treatment Comparison Results

5 10 15 20 25

0

2

4

6

Escape torituals

BL FCT + EXT

5 10 15 20 25

Escape

BL FCT + EXT

0

1

2

3Escape totangibles

Escape

Tangibles

IISCA- based treatment Standard-based treatment

Pro

ble

m b

ehav

ior

per

min

Sessions

Chloe

Dylan








% of analyses with strong control x (26%) (92%)

Within-S comparison: % differentiated x (57%) (93%)

Within-S comparison: % initial Rx success x (50%) (100%)

The supposed problem(s) with the IISCAImprecision

Do not know the specific operant class to which any particular topography of problem behavior belongs.

Do not know whether some part(s) of the synthesized contingency are irrelevant

Do not know whether behavior is maintained by pos or neg sr

Sometimes cannot neatly describe or compartmentalize the controlling variables

Consider the effective action without knowing these things

Imprecision is not Unique to the IISCAInterpretive ambiguity from an IISCA

Partly synthesized contingencies populate SFAs

Interpretive ambiguity from an SFA

The supposed precision of a SFA is but an illusion

Antecedent PB Consequence Interpretation

No Attention / No tangible No Attention / Tangible Might be evoked by low attn. or lack of tangible, or both and maintained by attn. or tang or both

Attention / No tangible Attention / Tangible Might be maintained by tangible or attention or both

Antecedent PB Consequence Interpretation

No Attention / No tangible Attention / Tangible Behavior may be controlled by one, the other, or both

“All models are wrong; the practical question is how

wrong do they have to be to not be useful.”

Box & Draper, 1987, p. 424

Recognize the Historical Significance

of the Standard Functional Analysis• Moved us from behavior modification to behavior analysis

– Taught us our professional humility

• Inspired us to transcend description and prediction to control

– Allowed us to be scientific practitioners

• Showed us how to create stable and controlled baselines

• Allowed us to discover and enhance treatments

To achieve the humane promise of a function-based treatment and a socially valid outcome

From a functional analysis:

What can I safely infer?

What must I know?

What do I not need to know?

That which I can safely infer via my functional analysis:

Response class membership

Problem Behaviors reported to co-occur

(in order of concern)

1. SIB

2. Aggression

3. Disruptive Behavior

4. Disruptive vocalizations

5. Whining/complaining

This analysis shows all forms of problem behavior are evoked and maintained by same synthesized contingency.

This happens every time we conduct this sort of analysis.(Warner et al., 2016)

This happens every time anybody else conduct this sort of analysis (Smith and Churchill, 2002, Borrero & Borrero, 2008, Herscovitch et al., 2009)

That which I can safely infer via my functional analysis:

Response class membership

If parents or teachers report that other topographies of problem behavior precede or co-occur with the primary topography of problem behavior, they are likely to be maintained by the same reinforcers.

I will infer response class membership and use their response to intervention (RTI) as verification

That which I must know via my functional analysis:

That I can reliably turn problem behavior off with the presentation of the reinforcer(s)

That I can reliably turn problem behavior on with the presentation of the evocative event

(And that the reinforcers and evocative events were identified by other people relevant to the behaver)

That which I do not need to know via my functional analysis:

The single operant function of each problem behavior

Whether problem behavior is maintained by positive or negative reinforcement

Whether some element of a synthesized contingency is a “true” contingency or merely a “false positive”

That which I do not need to know via my functional analysis:

The single operant function of each problem behavior

Whether problem behavior is maintained by positive or negative reinforcement

Whether some element of a synthesized contingency is a “true” contingency or merely a “false positive”

Whether I can neatly compartmentalize the operation in the analysis into a tidy generic class of reinforcement

(e.g., social positive, social negative, attn, tang, esc, etc.)

My gold standardGeneral and socially validated behavior change

by relevant people in relevant contextsBaer, Wolf, & Risley, 1968

I achieve it by being able to turn on and off problem behavior in an analysis informed by caregivers.

For more information go to:


Many thanks to my Practical Functional Assessment Research and Practice group:

Nick Vanselow, Sandy Jin, Laura Hanratty, Joana Santiago,

Mahshid Ghaemmaghami, Joshua Jessel, Jessica Slaton,

Robin Landa, Christy Warner, Shannon Ward, Tanya Mouzakes

Adithyan Rajaraman, Ellen Gage, Holly Gover, & Kelsey Ruppel

Enjoy your weekend.

Work to be done

1. Teaching people how to discover relevant reinforcement contingencies via open-ended interviews

2. Teaching people how to design analyses from results of interviews

3. Determining a more systematic process for progressing from inconclusive to strongly differentiated analyses

4. Determining whether a valid treatment outcome can be obtained despite a sometimes unreliable process

5. Determining the probability of a socially valid outcome for heterogeneous groups of children who engage in severe problem behavior (i.e., RCT of the IISCA & skill-based treatment process)

acknowledging essentials, inferences, and all that you are ... · acknowledging essentials,...

Documents