The Auckland Pigeon Lab Michael Davison, Ph.D. The Experimental Analysis of Behaviour Research Unit, Faculty of Science, The University of Auckland Auckland, New Zealand
(Left to Right) Rear: Douglas Elliffe, Michael Davison. (Left to Right) Front: Ph.D. students Michelle Banicevich and Nathalie Boutros.

My part of this lab has been running since 1969, when I came to Auckland as a Lecturer.  At that time, Ivan Beale already had some pigeon experiments running.  I managed to get 4 sets of Grason-Stadler relay equipment, sufficient to run a couple of experiments, and a few initial graduate students (Bill Temple, Mary Foster, Brenda Lobb, Austin Trevett).  My graduate students not only ran experiments, they also made programming equipment – 28 v solid-state modules using HTL—and soon we had many more experiments going.  Over the years, we have been through 5v logic panels, 6809 computers running FLEX and SLAVIC (4-K slave basic with control functions, written by one of our techs),  and PDPs 8E, 8A, and 11/73 all running Time-shared Super SKED.  It was when the main board of the 11/73 went down and we suddenly had to find $10,000 for a new one that I realized that we had to change – to PC compatibles, and to MED-PC© which, by that time, was mature and, as we quickly found, easy and effective to use.  We now use MED-PC© Version 4 for all of our experiments, and PowerBASIC and Corel QuattroPro for data analysis.

We have a dedicated pigeon holding room (Figure 1) that also is the experimental room—as you will see—running 16 experiments with 6 pigeons in each.  We are doing this in a way that is unusual.  Thanks to an influx of money a few years ago, we were able to convert all of our large pigeon home cages into experimental environments with 3-6 experimental keys and a single feeder in each.  The stimuli vary between just red, green and white/yellow LEDs to small VGA units on which we can show more complex stimuli, and a few MED 12-color units (ENV-124AM).  But almost all of our pigeon-side equipment is home made—because we used to have mechanical and electronics technicians available at no cost (save the cost of material).  This is no longer the case, unfortunately.  The computer interface side is purely MED-Associates, and we have been in the happy position of being able to replace the older interface units with new ones over the last few years.  Other changes, like the change from MED-PC© Version 2 to MED-PC© Version 4, have been driven by our university insisting on replacing our PCs every 3 years, and not providing EISA slots in the new machines.

We have 6 PCs running the experiments, and each one runs up to 6 experiments.  All experiments start and run automatically (using BATCH files run by the Windows XP Task Scheduler) overnight, starting at about 00.30 hrs after the room lights come on at midnight.  Mostly, the six birds in an experiment run simultaneously, and experiments are run successively ending about 6.30 am.  This is done by reconfiguring the MED interfaces for each successive experiment within the BATCH files.  For every session for every bird in every experiment we collect the time of every event to 0.01-s resolution and an associated code for the type of event.  Additionally, we have a set of 4 MED pigeon boxes (ENV-007) and interfacing for 2nd and 3rd year teaching labs, which are run conventionally (i.e., the pigeons are carried to the boxes for the session).

The lab is a joint operation between myself and Douglas Elliffe – we share all equipment, and graduate students too.  One of us (yes, that does include Doug and myself) is responsible for checking the data and weighing all the pigeons on each day of the week, so the lab is a very cooperative venture.  I think it particularly important that the academic staff do their parts in this—this leads to much better research and research supervision.

Our research has been overwhelmingly concerned with all aspects of choice and behaviour allocation.  Thus, we use concurrent schedules, conditional-discrimination procedures, multiple schedules, foraging procedures, and so on.  This focus explains our usual, but not exclusive, use of simple stimuli – though, as I mentioned above, we do have some more complex stimulus options, and have, in the past, used a programmable monochromator and full VGA screens for some research using complex stimuli in conditional discrimination and aspects of stimulus control.  Our major contributions have been producing data and theories of choice – we worked initially on the generalized matching relation, then developed the contingency-discriminability model, and as spin-off from these, two theories of conditional discrimination (Davison & Tustin, 1978; Davison and Nevin, 1999).  More recently, we have been looking at the dynamical effects of changing concurrent-schedule reinforcer ratios within sessions, and have discovered both a series of very local effects of reinforcer deliveries (so called “preference pulses”) and remarkable order in choice as affected by sequences of reinforcers (Davison and Baum, 2000; 2003; Krageloh, Davison, & Elliffe, 2005).  We have also discovered that learning—adaptation to new environmental contingencies—can be very fast indeed (Figure 2).  Just 3-5 reinforcers can put choice onto a new stable level.  More recently, we have started investigating choice between 4 alternatives, and started looking again at the local dynamics of choice by arranging feedback functions  between choice and the subsequent probability of reinforcement.

Our research over the last 7 years has led us to reconceptualize the Law of Effect, the way in which we think reinforcement works, and to question conventional wisdom in this area.  Two experiments (Krageloh, Davison, & Elliffe, 2005; Davison & Baum, 2006) have suggested that reinforcers do not increase the probability of responses that they follow; rather they serve to predict future reinforcement conditions.  A reinforcer effect (preference pulse) may only be obtained if a reinforcer predicts that there will be more of the same for that response at that location; if it predicts a lowered conditional probability of future reinforcement, that reinforcer may decrease subsequent response rate for that response, and lead to an increased probability of a different response. 

References

Davison, M., & Baum, W.M. (2000). Choice in a variable environment:  Every reinforcer counts.  Journal of the Experimental Analysis of Behavior, 74, 1-24.

Davison, M., & Baum, W.M. (2003). Every reinforcer counts:  Reinforcer magnitude and local preference. Journal of the Experimental Analysis of Behavior, 80, 95-129.

Davison, M. & Baum, W.M. (2006). Do conditional reinforcers count? Journal of the Experimental Analysis of Behavior, 86, 269-283.

Davison, M., & Nevin, J.A. (1999). Stimuli, reinforcers, and behavior:  An integration.  Journal of the Experimental Analysis of Behavior, 71, 439-482.

Davison, M.C., & Tustin, R.D. (1978).  The relation between the generalised matching law and signal-detection theory.  Journal of the Experimental Analysis of Behavior, 29, 331-336.

Krägeloh, C.U., Davison, M., & Elliffe, D. (2005).  Local Preference in concurrent schedules: The effects of reinforcer sequences.  Journal of the Experimental Analysis of Behavior, 84, 37-64.