Thirty-two zebra finches were tested (16 males and 16 females); 8 were tested in the pitch group, 12 in the duration group, and 12 in the intensity group. All groups had an equal number of males and females within the group. All zebra finches were at least 160 days old at the beginning of the experiment. The animals were bred and reared at the Leiden University animal breeding facility, where they were housed in single sex groups on a 13.5 L:10.5 D schedule at 20–22 °C. Food, water, grit and cuttlebone were available ad libitum. During the experiment, food was used as reinforcement and therefore only available after a correct trial. Food intake was monitored daily, and additional food was provided whenever necessary.
The experiments took place in individual operant conditioning cages, which were placed in separate sound-attenuated rooms. Each room was illuminated by a fluorescent tube that emitted a daylight spectrum on the same 13.5 L:10.5 D schedule as was used in the breeding facility. A speaker (Vifa 10BGS119/8) was located 1 m above the centre of the cage. The operant conditioning cages were constructed of mesh wire sides with a back wall and floor of foamed PVC. The back wall supported three horizontally aligned pecking keys and a food hatch above them, all easily accessible from various perches. The pecking keys were fitted with red LED lights. Birds needed to peck on the middle key to initiate a trial and stimulus playback. Depending on the nature of the playback, the bird had to either peck on the key on the left or the key on the right within 30 s. A correct response was followed by 8 s of food access (seeds identical to their regular diet), and an incorrect response was followed by 15 s of darkness.
The birds were trained to discriminate between stimuli that each consisted of two duplets of tones (i.e. four tones organized in an ABAB structure). Half of the stimuli consisted of two duplets with iambic stress, and the other half of two duplets with trochaic stress. For one group of birds, the stress was created by changes in pitch, for a second group by changes in duration, and for a third group by changes in intensity (see Fig. 1 for an example). Each bird received a set of four stimuli with iambic stress and four with trochaic stress. In the pitch condition, the iambic stimuli had a low–high–low–high pattern and the trochaic stimuli a high–low–high–low pattern. The high tones were always 25% higher than the low tones within the same quadruplet. All tones for this condition were 60 ms long and had an intensity of 70 dB. In the duration condition, the stimuli were organized in a similar fashion, the iambic stimuli had a short–long–short–long pattern, and the trochaic stimuli had a long–short–long–short pattern. Each of the four training stimuli within one category (iambs or trochees) started with a different tone duration, and the long tones were 50% longer than the short tones within the same quadruplet. As the initial group of eight birds trained with these stimuli showed very poor learning, we decided to test four additional zebra finches with a stimulus set in which the long tones were 100% longer than the short tones. All tones in this condition had a pitch of 3 kHz and an intensity of 70 dB. In the intensity condition, the iambic stimuli had a soft–loud–soft–loud pattern and the trochaic stimuli a loud–soft–loud–soft pattern. The loud tones were 5 dB louder than the soft tones. Here also the learning during training was poor, and hence, we again added an additional four zebra finches that were trained with stimuli in which the loud tones were 8 dB louder than the soft tones. All tones in this condition were 60 ms long and had an intensity of 3 kHz. In all conditions, the pure tones were separated by a 60-ms silent interval. For each condition, four different training sets were created to avoid pseudoreplication (see Table 1 for an example of the training stimuli). The pitches, durations and intensities of the tones were all chosen to be within the hearing range of the zebra finches. Furthermore, the differences between the tones within each condition have been shown to be audible for the birds (Spierings and ten Cate 2014).
There were four different test conditions, three testing a potential bias of the birds (Tests 1, 2 and 3) and one control condition (Test 4, Table 2). In Tests 1, 2 and 3, the zebra finches heard long sequences of alternating tones. If they perceived these to be organized in an iambic way, they were expected to categorize them as they did with the iambic training stimuli. If they grouped the tones as trochees, they should respond similarly as to the trochaic training stimuli. Test 4 also had long strings, but consisting of one single tone. These strings could not be grouped based on the alternations, which means that non-random responses of the birds would indicate a response preference for one of the keys or a perceptual grouping bias extended to non-alternating sounds.
More specifically, the stimuli for Test 1 consisted of the same tones as those used for the second and third training quadruplets (see Table 1; “Appendix”), only now in a string of 26 tones. As in the training stimuli, the tones alternated with high–low, long–short or loud–soft configurations. However, unlike the training stimuli, these long strings started and ended with a 1.3-s fade, obscuring how the string started. This prevented the birds from simply comparing the start or end tones of a string to the training stimuli. Moreover, some test strings started with a stressed tone and others with an unstressed tone. All tones were again separated by 60-ms silent intervals. The stimuli for Tests 2 and 3 were constructed similarly to those of Test 1, but now consisted of new tones. These tones were of either a different pitch (for the pitch condition), duration (for the duration condition) or intensity (for the intensity condition) than the tones from the training stimuli. Test 2 had tones within the range of the training tones, and Test 3 consisted of one tone that was on the edge of the range of the training tones and tones that were higher and lower, longer and shorter, or louder and softer than the training tones (see “Appendix”). As in Test 1, these strings were 26 tones long and had a 1.3-s fade in and fade out. Test 4 consisted of three different test strings, all containing one of the three tones that also occurred in the second and third training strings (see “Appendix”). These test strings did not have alternating tones, but had a repetition of a single tone. Like the strings of Tests 1, 2 and 3, they were 26 tones long, separated by 60-ms silent intervals and with a fade in and fade out of 1.3 s. As there were four different training sets for each condition, there were also four test sets per condition to match the stimuli from the training.
Each zebra finch was first trained on the go-left/go-right design with two unfamiliar zebra finch songs. They received a reward for pecking on the left key after hearing one song, and on the right key after hearing the other song. When they reached a criterion of over 75% correct responses to both songs for three consecutive days, they proceeded to the training.
During training, the zebra finches had to discriminate between four stimuli with iambic stress and four stimuli with trochaic stress by pecking on either the left or the right key after the stimulus was played. For half of the birds, the key for iambs was on the right side of the cage and the key for trochees on the left; for the other half of the birds, this was switched. If an individual only used one of the response keys instead of both, the programme was set to repeat a stimulus that received an incorrect response until the bird gave the correct response. This setting would be on for <24 h, motivating the animal to use both response keys. Training continued until the birds reached a criterion of >75% correct for three consecutive days or when they reached 20,000 trials without having 3 consecutive days with more than 55% correct responses. Those birds that reached the learning criterion then proceeded to the test phase. One bird from the duration condition and one bird from the intensity condition did not reach the criterion, but their performance was above 60% correct for 10 consecutive days. These two birds also proceeded to the test phase.
In the test phase, 20% of the trials were non-reinforced test stimuli, presented in a random order within a test block. The other 80% of the trials remained reinforced training trials. The test items were organized in two sequentially presented test blocks, one with the stimuli of Tests 1, 2 and 3 and the second one with the stimuli of Test 4. A bird moved to the next test block after each test stimulus in the block had been presented 40 times.
The responses of the birds to the training and test stimuli were calculated as proportions of responses to the key for iambs and the key for trochees per stimulus (number of responses/number of trials). It was also possible for the birds not to respond within 30 s of initiating a trial, in which case a “no response” was recorded. The proportion of no responses was calculated as the number of non-responses divided by the number of trials. For the training, we calculated the average responses to all iambic training stimuli and the average response to the trochaic stimuli per bird. For the test, we calculated the average response towards the different stimuli within the test trials of that condition. The three response proportions (iambic, trochaic and no response) always added up to be a hundred percentage per stimulus. These data were analysed with a generalized linear model (glm) with test item (all tests and the training iambic and trochaic stimuli) as fixed effect and the individual as the random measure. Pairwise comparisons were made between the proportions of responses to the iambic and the trochaic key for each test and the two training sets by using a Tukey’s post hoc test, corrected for multiple testing. When there was only one individual tested in a condition, their responses were analysed with a pairwise t test between their responses to the iambic and the trochaic stimuli.