Calculating Frequency with ‘Dead Air Time’

This may be one of the advantages of timings over recordings, well, some types of recordings, and some types of timings. On the other hand, sometimes recordings may have the advantage over timings. Anyway, I want to deal with a problem in how we measure frequency.

Let’s start with an example. Suppose the learner starts a timing and makes a response every second for the first 10 seconds.  OK, it we stopped the timing after 10 seconds the frequency would be one per second, or, on the chart, 60 per minute, a fairly high frequency.

Now, suppose we hadn’t ended the timing right then and there, but left the timer running. And that after the 10th second, the learner made no more responses.  Dead air, so to speak. As the seconds tic by, the frequency will go down. Right?

Right.  After 20 seconds the learner’s frequency would be 10 responses in 20 seconds. Multiply by three, and we get a frequency of 30 per minute, half of what we had.

If we continued timing, or recording, after 30 seconds the frequency would be down to 20 per minute. (Remember, in this example the learner made 10 responses in the first 10 seconds and then stopped responding, even though the clock is now continuing to run.)

If this scenario played out for a full minute, the learner’s frequency would drop to 10 per minute. And if the same scenario continued for the next minute, the learner would be down to 5 per minute.

So, question:

What’s the kid’s real frequency?  60 per minute, 30 per minute, 20 per minute, 10 per minute, 5 per minute, 1 per minute, what?

By manipulating the clock, or following some rules about using the clock, the representation of the kid’s frequency can be changed without the kid ever doing anything about it, one way or another.  In other words, you can get any frequency you want simply by ensuring that some “dead air time” (time when the person is not responding) is included in, or removed from, the determination of the frequency.

Where timings gain a bit of an advantage is when they span very brief durations of time.  If someone conducts a 10-second timing, there isn’t a whole lot of time left over for any “dead air time” to creep in and thus start changing the apparent frequency.  On the other hand, if you activated a response recorder and several minutes passed by before the learner started responding, you have to decide whether to include that “dead air time” in the calculation of the frequency.

So, what’s the real frequency?

This problem has been brought up from time to time in Behavior Analysis, and it is sometimes offered as a crippling blow to frequency as a measure of behavior.

The wishy-washy answer is, “it depends.”

Dead air time happens during timings and during recordings of behavior. It can occur as a pause for a few seconds within a timing, after which the behavior resumes.

My guess is that if we were to apply Lindsley’s “the child knows best” principle, that we then let the learner’s behavior dictate the frequency, and not some arbitrary rules, unless there is some compelling reason to have the rules trump the principle.

So, abiding by the principle, in my view, in the example above the kid’s frequency was 1 per second for 10 seconds, and that’s it. It’s not 60 per minute because for 50 seconds out of that minute the kid remained mute. Nor was it 10 per minute, because that’s applying an arbitrary rule.  The figure of 5 per minute (in the example, recording or timing for two minutes instead of one) is just as applicable and just as valid and just as totally arbitrary.  And thus just as worthless.

So, how do we chart 1 per second for 10 seconds?  Put a dot at the 60 per minute line?  That seems to be what’s done now if, say, a 10-second timing was conducted and the result of 1 per second obtained.  But, that might be stretching matters a bit, or more than a little bit. In the case of the example above, extrapolating 1 per second for 10 seconds to 60 per minute would be clearly incorrect and unreasonable, given that the learner didn’t maintain that pace for a full minute.

So, how do we chart it? 

This is what’s not clear to me.  Perhaps this is where there’s a weakness in the charting conventions. Currently, we have the Record Floor, aka these days as a Counting Time Floor, and the Behavior Floor (another PT relic from the past, rarely used anymore).  Neither floor applies to the situation above where we are confronted with maybe having to include “dead air time” in the calculation of a frequency. Perhaps we need a third floor, an “Actual Floor,” or something. The notion of an “Actual Floor” would be predicated on the learner’s behavior, not on tactical considerations from the perspective of the teacher, researcher, or other person conducting observation and measurement. Accordingly, an “Actual Floor” would require its own symbol on the chart, what I am not sure.

Or, as an alternative, when a timing, or a recording, falls to less than a minute, perhaps we should avoid charting the frequency in terms of count per minute. (I can almost feel the tidal wave of disagreement there!). In that case, perhaps a different chart, one with Count per Second up the left, would work better?

Otherwise, we’ll end up with a situation where the figure given for a learner’s natural frequency can be arbitrarily changed at will, on paper or on a computer, without the learner’s actual behavior ever changing.

I am open to suggestions on this, and I think this would be a fruitful and compelling topic for discussion for Precision Teaching. — JE





2 Responses to “Calculating Frequency with ‘Dead Air Time’”

  1. Dick Briggs Says:


    The issue is right on target, in particular if you are building duration. However, the concept should also be used to better identifying minute behaviors and possible relationships/intervention plans.

    I would suggest keeping the SSC but instead of the vertical lines being daily, use them for time slices. This would allow one to observer behavior across time slices but during the same intervention.

    Quick analogy. Micowave pop corn.
    Use the vertical lines in several charts, such as 5 seconds, 10 seconds, 30 seconds, 1 minute, 5 minutes (burned results).

    Time the popping of one bag. Chart the number of kernals popped (and maybe unpopped) for each time slice and compare the results. Can you notice a difference between timing slices? If so, can an intervention be designed to increase or decrease the pattern.

    Example: Petit mall – during stimuls/response activity, would comparing various time slice charts better pinpoint the “dead air”? How often are they occurring, what is triggering, what intervention would be best?

    Example: Knee surgery and rehab. Sitting down, raise leg to be parallel to the floor. Do you do 3 sets of 10 repetitions over x minutes? How much rest between sets is most helpful in meeting the goal? How fast should each leg lift be performed. Do I do a burst of leg lifts with 3 second intervals? Do I do them with 5 second intervals? Initially and over time, which therapy program would result in quicker healing.

    I have to tun to help my neighbor move. I’ll think of more examples.

  2. John Eshleman Says:

    Dick, sorry for the delay in getting back to you.

    You raise some good points. My point had more to do with how we measure frequency. When do we start the clock? When do we stop it? What is the actual frequenc of the behavior?

    The “dead air time” refers to time when the behavior of interest has either stopped, but the clock continues to run, or to time before the behavior of interest has started, but the clock’s already been started.

    One way around this, of course, would be to use a cumulative record, because you can then excise off the flat portion of the record before or after the run of the behavior. That solution would work, in concept. Except that hardly anyone uses cumulative records anymore, and there aren’t any convenient ways of generating them outside the operant laboratory. (Perhaps cell phones, PDA’s, or iPods, even, could be applied to this purpose some day.)

    The upshot of all this is that it’s possible to derive frequencies that do not depend entirely on the behavior of the organism; frequencies that are a joint measurement product of the behavior and of the recording method used. — JE

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: