Home » setting criteria

Shortening the Intermission Between Cue and Behavior: Proofing for Latency

At this point in the proofing series, we've addressed "The 3 D's" as well as precision.

Now it's time to work on latency.

What is latency?
Latency is the time interval between when you give the cue and when the dog starts to respond to the cue.  Latency is different from speed, which will be the next entry in the proofing series.  Speed is the time between the initiation of the behavior and the completion of the behavior.  

Living with Monte has taught me all about high latency (and low speed, coincidentally).  I would ask for a sit, and before his butt started to lower to the ground had time to make a home made lasagna (including a home made sauce), pay the bills, watch the seasons pass, teach a few classes, and catch the first half of a Knicks game.  

The common advice of trainers is to "click only the fastest sits."  The problem many trainers run into is developing a framework for comparison:  in order to be able to judge which sits are the "fastest," you need to establish a baseline latency.  

I had students in class work on this the other night.  I passed around sheets of memo paper; each with two columns, one for the trial session number and one for the dog's response time.

I told each student to select a behavior that had a high latency (a significant response time between the cue and the initiation of the behavior).  I told them to cue the behavior five to ten times.
I instructed them to cue the behavior, and immediately start mentally singing ABC's in their head, stopping when the dog began to respond to the cue. 

Remember, when working on latency we are trying to decrease the time between the cue and the initiation of the behavior, so we are not clicking when the dog's bum hits the floor but when they first start to lower their hind quarters (using the ABC's as a reference tends to help people "count" at a reliable pace each trial).  After each trial, the student was to write the letter in the appropriate column. 

Incidentally, you may use any song that is to your liking, provided it has a consistent rhythm.

When the students had recorded 5-10 trials, we used a chart which lists the number corresponding to each letter of the alphabet.  

So if we did five trials, the chart may look like this:

Trial 1:  D (4)
Trial 2:  G (7)
Trial 3:  B (2)
Trial 4:  I (9)
Trial 5:  C (3)

We divide the total by the number of trials, establishing a baseline latency of five.  Our lowest latency response was two, and this is our goal latency (we know that the behavior can be offered this quickly).

All responses occurring at or below baseline should be reinforced on a 1:1 continuous ratio.  

Cue the behavior, mentally sing your ABC's, and reinforce those responses which come at E or earlier.  If your dog fails to respond within the designated time span, you have two options:

a)  use a hand target to get your dog up so that you can set them up for another trial


b)  do not click, but toss a treat away, to reset your dog for another opportunity to earn reinforcement.  

Repeat this until you are clicking and reinforcing 4 out of 5 responses at or below the current baseline.

When you've reached this point, it's time to reassess our baseline.

For the next five trials, our record-keeping form may look like this:

Trial 1:  B (2)
Trial 2:  F (6)
Trial 3:  C (3)
Trial 4:  E (5)
Trial 5:  E (5)

Now our baseline time is 4.2.  For the sake of convenience, we'll round this down to four.   A count of two is still our lowest response time, and thus remains our goal latency.

We repeat the process, reinforcing those responses which come at or before D.  When 80% of the responses are successful at this level, establish a new baseline, and repeat the protocol.  Remember, we can continue lowering our baseline until we reach our goal latency.

If in this round of repetitions we receive a response that came out to 1 or 1.5, we can lower our goal latency according but the procedure used to get there remains the same.

In my work with my own dogs and those of my clients, I have come to understand that each dog is going to have a different goal latency for a particular behavior.  I can pretty much bet my house that Mokie will always respond faster to cues than Monte.  In general, his movements are much slower and hers are much more rapid.  

However, I have had great results using this procedure to reduce latency for both of my dogs, and my clients seem to really appreciate having a good record keeping and time measurement system to help them judge which sits or downs actually are the "fastest" and thus worthy of reinforcement.  

It should only take a few sessions to start seeing a reduction in latency.

A bonus component of any of these proofing techniques is that once you proof a couple of different behaviors for relevant aspects of fluency, you'll start noticing that your dog's responses to all cues will generalize much more quickly for those "fluency factors."  Your dog is learning to learn!

Stay tuned for the next installment in the proofing series...Issuing Speeding "Click-its!"