PSYC 3210 :: Laboratory Exercise - Category Discrimination Training

Objective

To train a starling to peck a key whenever it hears songs from a specific individual, and not to peck to songs from other individuals.

Introduction

Last week, your group trained a starling to eat from a food hopper whenever it heard the sound of the hopper raising or saw a cue light that reliably preceded the hopper. Some of your birds may have begun to peck at the cue light, an example of a conditioned response that emerges because the starling associates the cue light with food arriving in the hopper. Over the course of the last week, your group’s starling has been further trained with an automated computer program called shape to associate that pecking behavior with food reinforcement.

You will now introduce discriminative stimuli to the operant behavior, which inform the bird about whether its responses will be rewarded. One set of stimuli will be associated with reward (S+) and another set of stimuli will not (S-). The stimuli you’ll be using are 10-second segments of starling song produced by different individuals. Each starling’s song is individually distinctive, and starlings can easily learn to classify songs by which individual sang them.

Your scientific goal over the next two sessions is to determine what elements of the song starlings use in this classification. This kind of question is frequently addressed experimentally in two phases. The first phase is called category training: you will present several exemplars of one category with reinforcement (S+), and several exemplars of another category without reinforcement (S-). Eventually your bird will learn to respond to the S+ stimuli and withhold responses to S- stimuli. Once this happens, you’ll start the probe or transfer experiment, in which you present several novel or modified exemplars of the S+ and S- categories. You can choose these exemplars to test specific questions, like “what happens if I play the stimulus backwards?”

Procedure

Spectrogram Inspection

Your first task is to familiarize yourself with the structure of the song stimuli. Each person should complete this exercise on his or her computer, but you can discuss questions within your group. You don’t need to record your answers to the questions below, but you should make sure to answer them all, as you will be using this information to design stimuli for next week.

Preparation: Go to Collab and download the files in the “Starling Songs” folder. You will also need software for viewing and editing sound stimuli. We’ll use a free program called Audacity, which you can download from this link.

Once you’ve installed the program, open the st500_4.wav file. You will initially see the recording as an oscillogram, which is a 1-dimensional plot of the sound pressure level as a function of time. It’s generally more useful to view the recording as a spectrogram, which is a 2-dimensional plot with time on the x axis, frequency on the y axis, and stimulus power indicated by color. Select “Spectrogram” from the dropdown menu to the left of the plot.

audacity screenshot 1

In either the spectrogram or oscillogram view, you can select sections of the file by dragging with the mouse. Click the icon with the green arrow to play your selection. Zoom in on the first 3-5 seconds of the song using the cursor and the “View/Zoom to Selection” menu item. Select “Spectrogram Parameters” from the dropdown menu to the left of the plot. Adjust the following parameters and describe how it changes the appearance of the spectrogram:

Pay particular attention to how Window Size affects the plot. Compare very short and very long windows. How are frequency and time resolution related to the window size? Pick a set of parameters that allows you to see good detail in the plot. You should be able to see clear vertical, horizontal, and curving lines. Try to get something that looks like the image below:

audacity screenshot 2

Scroll through the spectrogram and look for repeated elements called motifs. It’s helpful to listen to small segments of the song while looking at the spectrogram. There will usually be several renditions of a motif in a row, and they will look and sound similar. About how many distinct motifs are there in the song? How many times, on average, does the bird repeat each motif?

Focus on the last two motif renditions of the recording (see below). By selecting small segments of the motifs and by comparing the two renditions, describe what each of the elements (“notes”) labeled in the figure below sound like. It may help to find elements that look similar in other motifs so you can listen to them in different contexts. How many notes are found in both motifs? Before moving on, check your answers with the instructor or a TA.

motifs and notes

Compare this recording with st500_2, another recording from the same bird. How many motifs are found in both recordings? How many motifs are unique? How similar is the sequence? Listen to both recordings back to back - how similar do they sound to you? Now open a recording from a different bird (any file that doesn’t start with st500) and compare it to the two recordings you already have open. Are there are any motifs shared between birds? What about the order?

For the next part of the assignment, your group has been assigned one of the singers as an S+ stimulus. Open all the files from this bird in Audacity and look at the spectrograms. Choose 5 different recordings to use as exemplars, saving the remainder for next week’s probe experiment.

Category discrimination training

Preparation

  1. Assign tasks to each member of the group: timing, data recording, and apparatus operation.
  2. The data recorder should download the data sheet for this week. Enter the group members’ names and the subject ID.
  3. The apparatus operator should connect to the controller using the USB cable as before and navigate to http://192.168.7.2:8000
  4. Verify that the sound output works by clicking on organfinale.wav in the stimulus control section.
  5. In addition to the 5 S+ stimuli you picked, choose 10 additional stimuli to serve as your S- category. These should be recordings from several birds that are not the same as your S+ bird. Make a list of your S+ and S- stimuli and submit it to the instructor. We will use this list to continue training the birds.

Training

  1. On each trial, select a stimulus to play using the following procedure: Flip a coin to determine whether you will play an S+ or an S- stimulus. If the coin is heads, randomly choose one of the S+ stimuli; if the coin is tails, randomly choose one of the S- stimuli. Enter the stimulus name in the data sheet.
  2. Signal the bird to start a trial by turning on the central green cue light.
  3. When the bird pecks the center key, start the stimulus playing by clicking the stimulus name. Ignore any responses the bird makes during playback.
  4. After the stimulus ends, turn off the center light and watch to see if the bird pecks the right key. Wait for up to 20 seconds. If the bird pecked, and the stimulus was S+, raise the hopper. Otherwise, do nothing. Record the bird’s response and the latency (from the end of the stimulus). Code the bird’s response as a 0 for no peck and a 1 for peck.
  5. Repeat steps 1–4. For each set of 10 trials, calculate the percent correct to track the bird’s progress.

Notes:

Finishing up

  1. Disconnect your computer from the controller
  2. Clean up around the cage
  3. The recorder distributes copies of the data to each member of the group

What’s next?

As a group, complete the Probe Design assignment. Over the week ahead, we will continue training the birds on the categories and prepare them for the probe experiment.