Results, Part 2

In the previous lab note, I presented in detail the results from the first experiment.

To recap, birds trained to peck at an icon as soon as it appeared on the screen. Each presentation consisted of one of 4 objects presented at one of 4 screen locations. Birds were allocated to one of four groups: Location, for which the sequence order of locations at which the objects appeared was consistently repeated; Object, for which the sequence of objects repeated but in random locations; Both, for which both object and location repeated consistently; and Neither, for which the sequence of both object and location was randomized.

The primary results of Experiment 1 were a) the finding that birds learned a consistent location but not a consistent object order; b) the finding that if both location and object order were presented in a consistent sequence, then birds DID encode object order; and c) the cool result of object-location binding in birds in Group Both.

This left us with the question: Why did birds learn the consistent location order when object order was random, but not object order when location order was random?

One possibility is that object order, on its own, is not as salient as a consistent location order. To test this hypothesis, in Experiment 2 we decided to increase the amount of times the pigeon must peck at the object when it appears in order to get food reinforcement. The same birds that were used in Experiment 1 were used in Experiment 2. In Experiment 1, the bird only had to peck the object once to get its food. In this experiment, we required each bird to peck the object from 1-5 times, with an average of 3 pecks at the object required to earn reinforcement.

What did we find? Let’s take a look at the results.

 

Remember, we’re looking at the time cost as measured by how much slower the bird is at pecking the object on a test trial where the order of object, location, or both was varied compared to baseline training performance. The higher the bar, the greater the time cost in reaction time (RT).

A few interesting things can be gleaned from this graph. First, we finally have an effect of changing object order to differ from the order used in training for birds in Group Object! That is, the birds in Group Object, for which the object order but not the location order was consistent during training, now show a slower RT to probe test trials when the object is presented out of its training order. These birds didn’t show such an effect in Experiment 1. The only difference between experiments was the amount of times the bird had to peck at, and therefore process, the object. This appears to have worked! This means that the lack of an effect of changing the object order for Group Object in Experiment 1 was due to object order being less salient than location order. Increasing the amount of time the bird processes the object increased its salience, and therefore enabled the birds to encode object order.

Another interesting result was the loss of the binding effect in Group Both. Remember, in Experiment 1, these birds showed less of an RT cost when the object and location order were switched together during testing in a way that preserved the object-location association (Both), than when either was switched alone or if both were switched together but in a way that broke the object-location association (No Binding). It’s not immediately clear why the increase in peck requirement would make the binding effect go away. One possible explanation is that the longer time spent pecking at each object necessitated lengthening the time between sequential presentations of location. The increased time between visiting different quadrants of the screen could have reduced the saliency of location information, thereby reducing binding of object-place associations. An argument against this account was the fact that the increased peck requirement did NOT prevent birds in Group Location from encoding location information.

Regardless of the reason that binding went away, we did confirm our starting hypothesis that the failure of initially encoding object order in Group Object was due to the lower salience of this information when location order was random.

In the next update, we’ll fill you in on the results of Experiment 3, which tested whether a consistent sequence order during training was necessary for binding of object-location associations.

Until next time!

Aaron Blaisdell