SPED 8013 | Chapter 4: Measuring Behavior

Definition of Measurement

  • The process of applying quantitative labels to observed properties of events using a standard set of rules

Researchers Need Measurement

  • How scientists operationalize empiricism
    • Without measurement, science is guesswork and opinion
  • Applied behavior analysts measure behavior to answer questions
    • Basis for talking about behavior

Practitioners Need Measurement

  • To evaluate effects of intervention
    • Before and after treatment
    • During treatment
  • To guide decision making
  • To prevent mistakes
    • Continue ineffective treatment
    • Discontinue effective treatment

Benefits of Measurement

  • Optimize effectiveness
  • Verify legitimacy of treatments
  • Identify and end use of pseudoscience
  • Accountability
  • Meet ethical standards

Measurable Dimensions of Behavior

  • Dimensions are distinct features that can be measured
  • Three fundamental properties
    • Repeatability or countability: Behavior can be counted
    • Temporal extent: Duration
    • Temporal locus: When behavior occurs

Measures Based on Repeatability

  • Count
    • Number of responses emitted during an observation period
    • Reported as a frequency count
    • Measures of count alone do not provide sufficient information for analysis

In other words, a measure of count only would provide the chart on the left where the calling out behavior appears to be worsening. By including a second dimension of measurement, which is duration of observation we are able to calculate the rate of calling out as a percentage of minutes. This gives a more accurate indication of the rate of behavior as observed in the above right chart.

  • Rate/Frequency
    • Ratio of count per observation period
      • More meaningful than count alone
      • Include counting time for reference
      • Rate of correct and incorrect helpful in skill development
      • Reported as number per standard unit of time
        • Guidelines for Using Rate
          • Take complexity of response into account
          • Useful measure for free operants
          • Not appropriate for responses within discrete trials, i.e.. when you are restricting someone’s ability to respond it is not fair to use rate as a measurement
          • Not appropriate for continuous behavior over extended period. Example provided, Out of seat, to count number of times the student is out of seat when the behavior should indeed be measuring duration of time spent out of seat versus how many times the student was out of seat. In this case rate would be an inappropriate measure.
  • Celeration
    • Measure of the change in rate of responding per unit of time
    • Reported using Standard Celebration Chart
    • Captures behavior acceleration and deceleration

Measures Based on Temporal Extent

  • Duration
    • The amount of time a behavior occurs
    • Total duration per session
    • Duration of each occurrence
    • Reported in standard time units
    • Count and duration measures provide different pictures of the same behavior (see below graph, where left graph indicates frequency and second graph indicates duration)
    • A more appropriate choice for continuous behavior over extended period. Recall the example provided, Out of seat, to count number of times the student is out of seat when the behavior should indeed also be measuring duration of time spent out of seat versus how many times the student was out of seat. (i.e., student gets out of seat twice by operational definition of “out of seat” but returns to seat quickly one day, and then the next day only gets out of seat once but only returns to the seat near the end of class)

Measure Based on Temporal Locus

  • Response Latency
    • Measure of elapsed time between onset of stimulus and initiation of response
  • Interresponse time
    • Amount of time that elapses between two consecutive instances of a response class
    • Direct measure of temporal locus and related to rate

Derivative Measures

  • Percentage
    • A ratio formed by combining the same dimensional qualities
    • Expresses proportional quantity
    • Ex. Proportion of correct to incorrect
    • Ex. Proportion of observation intervals when behavior occurred
  • Trials-to-Criterion
    • Measure of the number of response opportunities needed to achieve a predetermined level of performance
    • Other measures can be used to determine trials-to-criterion (e.g. rate)
    • Typically calculated post facto
    • Used to compare effectiveness

Considerations for Using Percentage

  • Often misunderstood, used incorrectly
  • Most accurate with divisor of 100 or more
  • Percentage may be misleading
  • Limited use because has no dimensional quantity
  • Sets artificial limits on behavior change

Definitional Measures

  • Used to determine whether instances have occurred, then recorded using another method
    • Topography
      • They physical form or shape of a behavior
      • Not a fundamental dimensional quality of behavior
    • Magnitude
      • The force or intensity with which a response is emitted
      • Important parameter for some response classes
        • Eg voice, volume
      • Not a fundamental dimensional quality of behavior

Procedures for Measuring Behavior

  • Typically involve one or a combination of these three:
    1. Event Recording
    2. Timing
    3. Time Sampling Methods

Event Recording

  • Procedures for detecting and recording the number of times a behavior is observed
  • Devices Include:
    • Data sheets w/Pens & Pencils, wrist counters, digital counters, masking tape, paper clips, etc.

Considerations for Event Recording

  • Event recording is easy to do
  • Behavior must have discrete beginning and ending points
  • Rate must not be too high
  • Inappropriate for behaviors with long duration

Timing

  • Procedures to measure duration, response latency, and inter response time
    • computer systems, stopwatch, wall clocks, tape recorder
  • Duration
    • Precise recording of the amount of time a given behavior occurs
  • Response latency and inter response time
    • Precise recording of duration between events of interest

Time Sampling

  • Variety of methods for observing and recording behavior during intervals or at specific moments in time
  • Observation is divided into intervals, presence or absence of behavior recorded for each interval

Time Sampling: Whole-Interval Recording

  • Used to measure continuous behavior
  • Brief intervals (5-15 seconds)
  • At end of interval, record if behavior occurred throughout
  • Risk of underestimation
  • Reported as percentage of intervals when behavior occurs
    • Typically used for behaviors that you are seeking to increase, and this measure gives a conservative estimate of the intervention

Time Sampling: Partial-Interval Recording

  • At end of interval record if behavior occurred at any time during the interval
  • Multiple occurrences scored as one
  • Allows recording of multiple behaviors
  • Risk of overestimation
  • Reported as percentage of intervals when behavior occurred
    • Typically used for behaviors that you are seeking to decrease, and this measure gives a conservative estimate of the intervention

Time Sampling: Momentary Time Sampling

  • Record whether the behavior is occurring at the end of the interval
  • Does not require undivided attention
  • Misses much behavior
    • Best for continuous behavior, not low frequency, or short duration
  • Reported as a percentage of intervals when behavior occurred

Time Sampling: Planned Activity Check

  • PLACHECK
  • Variation of momentary time sampling
    • Measures behavior of individuals within a group
  • At end of interval, measure number of students engaged in target activity

Guidelines for Time Sampling

  • Use a timing device to signal beginning and end of observation
    • Increase accuracy
    • Not distracted by watching a stopwatch
  • Record a response for every interval (e.g. yes or no)
    • Prevents losing your place with blank intervals

Time Sampling Artifactual Variability

  • Artifact is something that appears to exist because of the way it is examined or measured
  • Time sampling provides estimate of actual occurrences
  • Different procedures produces different results
  • Differences produce variability in data (indicated in below data chart)

Measuring Behavior by Permanent Product

  • Measuring behavior after it has occurred by measuring its effects on the environment
  • All previous procedures can be applied to permanent product measurement
  • Products can be natural or contrived

Advantages of Permanent Product Recording

  • Practitioner free to do other tasks
  • Possible measurement of otherwise inaccessible behavior
  • More accurate, complete, continuous
  • Easier data collection (IOA, treatment integrity)
  • Measurement of complex behavior–can view repeatedly)

Determining Appropriate Use

  • Is real-time measurement needed?
    • Moment to moment decisions required
  • Can behavior be measured by permanent product?
    • Each occurrence must produce the same product
    • Product can only be produced by target behavior
    • SIB (Self injurious behavior)
  • Will a contrived product affect the behavior?
    • Reactivity effects
  • Cost to obtain and measure the permanent product?
    • Availability, cost and effort of generating the product

Computer-Assisted Measurement

  • Data collection and analysis software combined
    • Multiple systems available
    • Sophisticated and easy to use
    • Laptops, hand-held computers, PDAs
  • Simultaneous recording of multiple behaviors across multiple dimensions