What is variability and why do you as a behavior analyst care?

Variability from Chartlytics on Vimeo.

Statistical variability is taught in school: range, standard deviation, and variance are familiar terms. But what about behavioral variability? What is it? And more importantly, what is it telling us about our intervention?

Behavioral Variability

Johnston and Pennypacker (2009) define behavioral variability as “Variations in features of responding within a response class, as well as variation in summary measures of that class (p. 355). This means we can look at variability in two ways:

1. Variations within a response class
2. Variations of response class measures across observations

Consider this soccer pinpoint example: kicks ball in net. The frequency, duration, and latency of this behavior will vary within a response class. In your own backyard, variability may be caused by topography. Or if you watched USA women at the 2016 Olympics, you could see variations in force applied to the ball during the shootout.

If you’re rooting for a team (or player) at the 2018 World Cup in Russia, you may see variability across observations. The frequency will likely change – the team may have a higher frequency of “kicks ball in net” in game one than in game two.

Similarly, challenging behavior often varies across observations in a week. On Monday, a student might display 6 instances of “pinches peer” during a 30-minute instructional period. On Thursday, the frequency might be 1 instance during the same 30-minute instructional period.

Bounce: Visualization, and why you should care.

Suppose you want to visualize variability. That’s bounce: “variability visually displayed as a dispersal of charted performances bouncing around the celeration line” (Kubina & Yurich, 2012).

On a Standard Celeration Chart, we can quantify bounce. (The Chartlytics digital chart automatically calculates it for you.) If we chart Max’s performances of kicks ball in next, we’ll find a dispersal or “bounce envelope” of x3.1… which enables you to project the range of future performances.

Bounce provides critical information on stimulus control.

If a learner knows how to do something well, the bounce envelope will be narrow. Narrow bounce means strong environmental control/influence. Wider bounce means less, inconsistent or mixed environmental control/influence.

For example, suppose child see-writes spelling word in one classroom, where she is given candy from the teacher. Her bounce is tight; she typically completes all spelling words. But in her new classroom, she gets a high-five for see-writes spelling word. Then her bounce envelope is wider; the reinforcement changed, so the behavioral variability and bounce envelope did.

For acceleration targets, Precision Teachers have found the following data on bounce envelopes.

• Very strong control: x1 to x3  – easily find what influences the behavior
• Strong control: x3 to x6
• Moderate control: x6 to x10
• Weak, inconsistent control: x10 and above – difficult to discover what’s influencing the behavior

Note that deceleration bounce data usually has a wider envelope.

Research example: Systematic preference assessment for identifying R+ tangibiles

Preference assessments are used to identify preferred stimuli that may function as reinforcement. Two types have the best predictive validity: paired-stimulus and multiple stimulus without replacement. Reachers accessed how these simulus preferences work for different topographies of attention, (Lang et. al., 2014).

First, researchers used pair-wise preference assessment to identify high and low preferred forms of attention (e.g., fist bump, verbal praise, thumbs up, verbal praise + fist bump, and control/nothing), This helped figure out the hierarchy of preference.

Then, researchers implemented  multiple-stimulus without replacement before discrete trial training. They used ABAB design. In one condition, clients would get a low-preferred form attention contingent on correct response. In another condition, clients would get the high-preferred of attention contingent upon correct response.

Rick recharted data from the study to see what further information could be gleaned. In the figure below, the green lines represent bounce; and the black Xs represent instances of challenging behavior.

In the “low-preferred” condition, the celeration was x2.02 (for decel data) and bounce was x3.96. In other words, the challenging behavior was going up, but the control was pretty strong and predictable.

In the “high preferred” condition, the celeration was /2.06 (for decel data) and bounce was x14.4. In other words, the behavior was decreasing, but there’s a lot of variability. The x14 is a BIG DEAL and visually jumps out.

In “low prefered” again, you can see celeration of x1.19, and bounce of x7.09. And in “high preferred” again, celeration again is x1.93 with a tight bounce of x1.18.

Implications for clinical practice

So why should behavior analysts care about variability, or bounce?

1. Bounce shows you the degree of influence from your intervention,
2. When implementing interventions, look at a narrowing of bounce (bounce change) to show a dimension of impact. If you’re gaining control, you’re doing something right — even if challenging  behavior isn’t decreasing.

References

Johnston, J. M. & Pennypacker, H. S. (2009). Strategies and tactics of behavioral research. New York: Routledge.

Kubina, R. M. & Yurich, K. K. L. (2012). The precision teaching book. Lemont, PA: Greatness Achieved.

Lang, R., van der Werff, M., Verbeek, K., Didden, R., Davenport, K.,  Moore, M., Lee, A., Rispoli, M., Machalicek, W., O’Reilly, M., Sigafoos, J., & Lancioni, G. (2014). Comparison of high and low preferred topographies of contingent attention during discrete trial training. Research in Autism Spectrum Disorders, 8. 1279–1286. DOI: 10.1016/j.rasd.2014.06.012

Want to learn more about variability ? Curious about how behavior fluency works in your field or on your projects? Let us know in the comments!

Already watched the webinar video? Take BACB Type 2 CEU quiz here. Worth 1 credit.

bSci21 readers can also click here get a special 14 webinar deal – 14 CEUs for only $99 through Chartlytics!

Chartlytics simplifies data collection, standardizes decision-making, and projects outcomes — so behavior professionals can help their clients reach goals faster. We make it easy to use the Standard Celeration Chart and Precision Teaching techniques with our digital platform and services. www.chartlytics.com

*Paid content by Chartlytics.