Music Discriminations by Pigeons - PsycNET - American ...

Journal of Experimental Psychology: Animal Behavior Processes 1984, Vol. 10, No. 2, 138-148

Copyright 1984 by the American Psychological Association, Inc.

Music Discriminations by Pigeons Debra Porter and Allen Neuringer Reed College Pigeons learned to discriminate between complex musical sequences. Two birds responded at more than 80% correct in a single-operandum discrimination learning task when the S+ was a 1-min excerpt of Bach flute music and the S- was a 1min excerpt of Hindemith viola music (Experiment 1). Four pigeons responded at more than 70% correct when they were required to peck the left of two disks during presentations of any portion of a 20-min Bach organ piece and to peck the right disk during any portion of Stravinsky's Rite of Spring for orchestra (Experiment 2). These discriminations were learned slowly. However, the birds generalized consistently and independently of the instruments involved when presented with novel musical excerpts (Experiment 3). They preferred the left "Bach disk" when ,novel excerpts from Buxtehude and Scarlatti were introduced and the right "Stravinsky disk" when novel excerpts from Eliot Carter, Walter Piston, and another Stravinsky work were introduced. People responded similarly. Therefore the pigeon's response to complex auditory events may be more like the human's than is often assumed.

Studies of stimulus control of animal behavior have been extended from "simple" physical dimensions, such as intensity, wavelength, frequency, or angle (Blough & Blough, 1977; Mackintosh, 1974), to more complex stimuli that, although varying in sometimes poorly understood ways along multiple di^ mensions, can be related to human concept formation (e.g., Mervis & Rosph, 1981). Herrnstein and his colleagues haye shown that pigeons discriminate complex and abstract visual concepts. When required to peck a response disk when a picture containing a person is projected on a small screen, the birds learned the concept of "person" and responded correctly to pictures they had never before seen (Herrnstein & Loveland, 1964). Additional work showed that birds learned the complex concepts of angularity, trees, leaves, fish, and water (Cerella, 1979; Herrnstein, 1979; Herrnstein & de Villiers, 1980; Herrnstein, Loveland, & Cable, 1976;Lubow, 1974;Malott & Siddall, 1972). Such demonstrations are important, for they show the extent to which animals and people perceive and behave simThis article is based on a thesis presented by the first author to Reed College in partial fulfillment of the requirements for a BA degree. Requests for reprints should be sent to Allen Neuringer, Department of Psychology, Reed College, Portland, Oregon 97202.

ilarly. Humanlike animal discriminations have also been shown when auditory stimuli are used. Hulse, Cynx, & Humpal (1984) showed that starlings could discriminate between sound rhythms and pitches and that they generalized like human subjects. D'Amato and Salmon (1982) showed that both monkeys and rats we,re able to discriminate between simple auditory "tunes." The S+ stimulus was a sequence of six frequencies, the first four rising, the fifth falling, and the last again rising, with the complete stimulus taking 1.8 s. The S— stimulus was a 2-s period of 17 monotonically descending and 16 monotonically ascending frequencies. The present study extends the D'Amato and Salmon research to a more complex auditory domain and to possibly a more unlikely subject. There were two main questions: First, can pigeons discriminate between two specific episodes of classical music? Second, if yes, can they also respond differentially to broad categories of musical stimuli, as do people? The pigeon provides a relatively strong test of whether animals can learn to adequately discriminate complex auditory patterns. For although pigeons are known to discriminate between pigeon calls (Beer, 1970) and aspects of their audiograms are similar to human audiograms (Harrison & Furumoto, 1970; Heise, 1953), a number of researchers have concluded that pigeons are relatively poor at utilizing au138

MUSIC DISCRIMINATIONS BY PIGEONS

139

ditory cues (e.g., Mackintosh, 1974; Shettle- reinforcers were scheduled to average 6/min, approximately the same frequency as under the fixed-ratio schedule; then worth, 1972).

reinforcement frequency was decreased to 2/min for 35

Experiment 1 The first experiment asked whether pigeons could discriminate between excerpts from two pieces of classical music, one by Bach and the other by Hindemith.

Method Subjects Two White Carneaux pigeons (No. 25 and No. 70) were maintained at 80% of their normal body weights. They had continuous access to grit and water in their home cages but received grain only once per day after the experimental sessions, Both birds had previous experience in a variety of operant conditioning experiments but had not been exposed to auditory conditioning tasks.

Apparatus A Lehigh Valley pigeon chamber, 30 X 35 X 40 em, contained a front panel with a yellow transilluminated response disk located 9 cm above and 5 cm to the left of a reinforcement dispenser that occasionally provided 3 s of access to mixed grain. A small speaker was mounted behind the front panel, and musical stimuli were presented with a Craig Model 2603 or Sony Model TC-105 tape recorder. The stimuli consisted of a 1-min excerpt from J. S. Bach's Prelwle in C Minor for flute, which was continually repeated to comprise the S+, and a 1-min excerpt from P. Hindemith's Sonata, Op. 25, No. 1 for viola, which was repeated to comprise S—. The two stimuli were selected because they are: appreciably different to adult human listeners. Relay equipment in a separate room controlled all aspects of the experiment. Although intensities were not measured, the musical stimuli were characterized as "loud" by human listeners.

Procedure Phase 1. The Bach and Hindemith pieces alternated on the average of once per min, with a 5-s to 3-min range of intervals. When Bach played, S+ contingencies were in effect, and every 15th peck to the disk produced reinforcement (FR 15). When Hindemith played, the S- contingencies were in effect and food was unavailable. To assure that responses in S- were not accidentally reinforced, the stimuli did not switch from S- to S+ until at least 10 s had elapsed without a response. Each of 33 sessions terminated after 50 reinforcements had been presented. Phase 2. To assure that the behaviors were not being controlled exclusively by fixed-ratio schedule cues—that is, if the 15th peck were reinforced, that would be a cue for S+ contingencies, and if the 15th peck were not reinforced, S- was in effect—the contingency in S+ was changed to a variable-interval (VI) schedule under which pecks were reinforced variably in time. For a few sessions.

Phase 3. Two control sessions were inserted during the VI phase to test whether the birds were discriminating between attributes of the musical stimuli or whether artifacts were responsible for the behaviors. During the first session, the tape recorders were disconnected from the speaker while all other contingencies were the same as in Phase 2. During the second session, blank tapes replaced the music—the recorders remained on but transmitted no music—and all other contingencies were the same as in Phase 2.

Results and Discussion Both birds learned the Bach-Hindemith discrimination, as shown in Figure 1, and over the last five sessions of the first phase (FR reinforcement), they responded correctly 77% and 89% of the time. (This percentage equals the response rates in S+ divided by the sum of the response rates in S+ and S-.) When the reinforcement schedule was changed to VI, performances were temporarily disrupted, but the birds again came to respond more frequently to Bach than to Hindemith, and over the last five sessions of this phase they responded correctly 86% and 82% of the time (Figure 1). Either disconnecting the tape recorders or substituting blank tapes caused the percentage of correct responses to fall immediately (see points marked "a" and "b" in Figure 1), although in both cases, and for both birds, the percentages were above 50%. The schedule contingencies or some artifact in the soundpresenting apparatus may have provided partial cues. However, the large difference in the percentage of correct responses between these control sessions and the-music sessions indicates that the discriminations were largely based on the music. Experiment 2 The second experiment attempted to confirm and extend the above findings. First, much longer musical compositions served as discriminative stimuli to help rule out the possibility that the pigeons were responding to a single, simple differentiating attribute of the two discriminative cues. For example, if only one of the stimuli in Experiment 1 happened to have a note of a given frequency or intensity,

140

DEBRA PORTER AND ALLEN NEURINGER

that note and that alone might have guided the discriminations. In the present experiment, the length of the excerpts comprising S+ and S- stimuli was increased from 1-min to more than 20-min. Since a trial consisted of an average of 1 min, the exact nature of the stimulus varied greatly from trial to trial. Second, to extend the generality of the findings, pigeons of three different breeds were used. Third, again to extend the generality (and to provide for the generalization testing to follow), a choice-assessment method was used rather than the single-operandum method in Experiment 1. Method Subjects Two White Carneaux (No. 96 and No. 18), a Schietti Modena (No. 67), and two Silver King (No. 16 and No.

62) pigeons were maintained at 85% of their free-feeding weights. These pigeons had not been previously used in auditory discrimination tasks, but all had previous experimental training.

Apparatus A chamber similar to that used in Experiment 1 contained two Gerbrands response disks located 15 cm apart and 20 cm above the floor and a reinforcement dispenser, centered beneath the two disks, which provided 4 s of access to mixed grain. The keys, lighted by 7.5-W bulbs, and the houselight, an overhead 7.5-W bulb, were illuminated at all times except during reinforcement. Music was transmitted from two General Electric Model 8415 cassette tape recorders and was played through an overhead, 15-cm diameter speaker. Two musical stimuli—Bach's Toccatas and Fugues in D Minor (BWV565) andF(BWV 540) for organ, and Stravinsky's Rite of Spring for orchestra-^were recorded on a Sony model TC-130 stereo cassette recorder. As in Experiment 1, the two stimuli were judged by the experimenters to differ appreciably along many dimensions. A Dynamic Compression attachment on the tape recorder limited the decibel range. When the

90 r

Sessions Figure 1. Percentage of responses during the S+ stimulus (Bach) divided by the sum of responses in S+ and S- (Hindemith) as a function of sessions for each of two pigeons (No. 25 and No. 70). (The schedule of positive reinforcement in S+ was a fixed ratio schedule at first and then a variable interval. Music was omitted at points "a" and "b.")

MUSIC DISCRIMINATIONS BY PIGEONS music was played into the experimental chamber, the intensity was in the 90 dB range (80 to 96 dB) when measured by a Realistic music/sound-level meter a few inches from the speaker at about the level of a bird's head. Relay equipment in a separate room controlled the experiment.

Procedure Phase 1. The two musical stimuli were programmed to alternate randomly on the average of once per min with intervals between 2.5 s and 3.7 min, When Bach played, pecks on the left disk were reinforced variably in time on the average of once every 30 s (VI 30-s schedule of reinforcement). When Stravinsky played, pecks on the right disk were reinforced according to the same VI 30-s schedule. Pecks to the inappropriate disk—the right during Bach and the left during Stravinsky—were not reinforced and, in addition, were followed by a minimum interval during which neither reinforcements nor music alternations could occur. The duration of this penalty for errors was slowly increased from 0 to 10 s over the course of about the first 25 sessions and then kept at 10 s for the remainder of the experiment. Between 60 and 80 sessions were given, each lasting 45 min or until 50 reinforcements had been obtained, whichever, came first. Phase 2. After an average of 70 sessions, a no-music control session was inserted. Blank tapes replaced the normal Bach and Stravinsky tapes so that no music played into the experimental chamber, but the contingencies of reinforcement were the same as in Phase 1, that is, if the bird pecked the left disk when the "Bach tape recorder" was operated (although now with a blank tape), reinforcement was given on a VI 30-s schedule, and if it pecked on the right disk when the "Stravinsky tape recorder" was operated (also with a blank tape), pecks were again occasionally reinforced. As in Experiment 1, this controlled for artifacts. The single no-music session was followed by three to six sessions in which the musical stimuli were again presented, as in Phase 1. Phase 3. There followed a second control—this time for intensity—in which two or three sessions were given where the loudness of the music was decreased about 10 dB to the 80 dB range (70-86 dB) measured a few inches from the speaker. A final one to three sessions of normal intensity completed this experiment.

Results and Discussion

141

accuracy attained with simple color discriminations (Honig, 1969), it should be noted that during the terminal sessions each bird (except for No. 18) responded for periods of up to 15 min at more than 90% correct. These periods of almost perfect responding were intermixed with periods when the birds either changed disks several times during one music interval or, more likely, pecked one disk alone; they remained there for 7 or 8 min despite the absence of reinforcement and absence of stimulus alternations. When blank tapes were substituted for the musical stimuli in Phase 2, performances fell to 38%, 40%, 45%, and 38% correct for the four birds, respectively. During the next session, when the Bach and Stravinsky pieces were returned, performances returned to their normal levels. Thus the musical stimuli were essential for the discriminative behaviors. The less than 50% "correct" performance in the absence of musical stimuli was probably due to the specific contingencies involved. Recall that errors not only prohibited reinforcement but also kept the stimuli from switching, When the intensity of the music was decreased in Phase 3, performances were slightly disrupted (see Figure 3). For all birds, the percentage of correct responses was lowered during the first session of decreased intensity, but the percentage of correct responses then increased during the second session and fell within the normal intensity range. Thus it is unlikely that absolute intensities were major controllers of the discriminative performances. Experiment 2 confirmed and extended Experiment L Pigeons apparently used complex music as discriminative stimuli. Although the present discrimination was learned slowly and only reached a 70%-75% level of correct responding, other more common stimuli, when presented in the context of a complex learning task, led to comparable speeds of acquisition and levels of proficiencies in pigeons. For example, similar learning curves were obtained when pigeons learned to discriminate arbitrary relationships between line orientations and associated hues (Carter & Eckerman, 1975).

Figure 2 shows that the percentage of correct responses—denned as the sum of responses left when Bach was playing plus responses right when Stravinsky was playing divided by total correct plus incorrect responses—improved slowly for four of the five birds and, over the last five sessions, attained an average of 74%, 73%, 70%, and 76% correct responses, respectively. The fifth bird, No. 18, continued to respond at less than 60% correct and thereExperiment 3 fore was not continued in later phases of the experiment. Although 70%-75% accuracy is Due to the length and complexity of the relatively low compared, for example, to the musical selections in Experiment 2, it is un-

142


likely that the discriminations were based on pitch, and so forth. To test whether the pigeons specific notes or sequences. The musical stim- were responding to categories of music, the uli differed along numerous dimensions, for next experiment presented the four pigeons example, in one case the instrument was an ,that completed Experiment 2 with brief organ, whereas in the other there was a com- "probe" interludes of a variety of musical plete symphony orchestra, and there were dif- pieces never before played to the subjects. The ferences in tempo, timbre, richness of sound, excerpts were chosen from works by Buxte-

80r

40

10

20

30

40

50

60

70

80

Figure 2. Percentage of correct responses as a function of sessions for each of five pigeons. (A correct response was to the left disk when Bach was played and to the right disk when Stravinsky was being played. Correct responses were divided by total responses to generate the percentage of correct responses.)

143


Method hude, Vivaldi, and Scarlatti—these being, according to human listeners, more similar to Subjects Bach than Stravinsky—and by Eliot Carter, Walter Piston, and a different Stravinsky Four of the five pigeons used in Experiment 2 served piece—these being more similar to the Ex- as subjects in Experiment 3. periment 2 Stravinsky. The novel excerpts included -a variety of instruments—the pieces Apparatus by Buxtehude and Piston being for organ, the The apparatus was the same as in Experiment 2 except Scarlatti for harpsichord, the Vivaldi for violin that a third tape recorder, Craig Cassette Model 2603, was and orchestra, the Carter for chamber ensem- added. All novel probe stimuli were played on this recorder transmitted through the same speaker as the standard ble, and the second Stravinsky piece for or- and Bach and Stravinsky pieces. chestra. The basic question is whether pigeons who learned to respond differentially to musical Procedure pieces by Bach and Stravinsky would respond Phase 1. Sessions began with the same Bach and Strain a meaningful, or systematic manner when vinsky selections used in Experiment 2, which were alpresumably novel musical excerpts were in- ternated variably in time. When Bach played, the' birds occasionally reinforced for pecking the left "Bach troduced. In the presence of the novel piece, were whereas when the Stravinsky piece played, the birds a high proportion of pecks on the left disk, or disk," were occasionally reinforced for pecking the right "Stra"Bach disk," would imply a generalization be- vinsky disk." (Throughout this section, we will refer to tween the new piece and Bach. Likewise, a these standard discriminative stimuli from Experiment 2 high proportion of pecks to the right "Stra- simply as Bach and Stravinsky.) Sessions began with 5 of discrimination contingencies identical to those in vinsky disk" would indicate a similarity be- min Experiment 2. There followed a 3-min probe period that tween the novel excerpt and Stravinsky's The was subdivided into six 30-s intervals; the example that Rite of Spring. If, of course, the birds had follows illustrates the method. learned to respond to sequences or other spe- The probe, in this illustration, began with 30 s of Strafollowed by 30 s of a novel piece (in this example cific attributes of the original Bach and Stra- avinsky, Buxtehude organ work), and another 30 s of Stravinsky. vinsky stimuli, it might be expected that they There followed another 30 s of Stravinsky, 30 s of Bach, would respond randomly to the novel probes and a final 30-s period of Stravinsky. Thus responding during the novel excerpt was compared (a) with responding or not respond at all.

100 #16

#62

#96

#67

90 80

70

V** 60 50

Sessions Figure 3, Percentage of correct responses—responses to the left when Bach was played plus responses to the right when Stravinsky was played divided by total responses—for four pigeons under two intensities of music. (The Xs represent the normal 90 dB intensity range and the Os represent a 10-dB decrease.)

144


during Stravinsky excerpts that preceded and followed the novel piece and (b) with responding during a Bach excerpt that was analogously preceded and followed by Stravinsky. Throughout the six 30-s intervals comprising the probe, reinforcement was withheld so as to ensure that these generalization tests were not influenced by reinforcement. The main question was whether responding during the novel Buxtehude excerpt was similar to that during Stravinsky, during Bach, or neither. Following the 3-min probe interval there was an approximately 3-min interval of the standard Bach and Stravinsky alternation during which reinforcers were again occasionally presented for correct responses. Another 3min probe interval was then presented, and this sequence of 3-min probes (reinforcers withheld) and approximately 3 min of standard discrimination periods (reinforcers occasionally presented for correct responses) was continued throughout the session. During each experimental session, five to eight probe intervals were given. Excerpts from only one novel piece were presented during any given session. After every few sessions containing probes, one or two sessions were given containing only the standard Bach and Stravinsky pieces under the Experiment 2 contingencies to maintain high response strengths. The novel musical stimuli presented during the probes were a novel Stravinsky work, his Firebird Suite for orchestra, a Buxtehude organ work, a Vivaldi concerto for violin and orchestra, Eliot Carter's Sonata for Flute, Oboe, Cello and Harpsichord, Walter Piston's Chromatic Study on the Name of Bach for organ, and a Scarlatti harpsichord sonata. The Buxtehude, Scarlatti, and Vivaldi works were composed before 1750 and share attributes of the baroque style with the standard Bach composition; the Piston and Carter pieces, and the Stravinsky Firebird Suite were all composed after 1900 and share attributes of a contemporary musical style. In the probe example above, the novel piece was surrounded by Stravinsky, that is, Stravinsky-novel-Stravinsky, and this was compared with a standard Stravinsky-BachStravinsky sequence. This order of presentation was counterbalanced, that is, the Stravinsky-novel-Stravinsky sequence was used for the Buxtehude, Vivaldi, Carter, and Scarlatti probes, and the Bach-novel-Bach sequence was used for the Firebird, Piston, Vivaldi, and Scarlatti probes. In the latter cases, therefore, Bach-novel-Bach was compared with Bach-Stravinsky-Bach. Note that the Scarlatti and Vivaldi novel probes1 were presented in both the Stravinsky-novel-Stravinsky and the Bach-novel-Bach sequences for comparison purposes. Furthermore, whereas the standard-novel-standard component of the probes preceded the standard-standard-standard components in two cases—Buxtehude and Firebird Suite—the order was reversed in the remaining four instances—the Vivaldi, Piston, Carter, and Scarlatti excerpts. The number of probes given per novel excerpt ranged from 5 to 17, depending on the consistency of a bird's responses to that novel piece, that is, until a bird demonstrated consistent preference for at least five consecutive probes, Phase 2. In the generalization tests in Phase 1, the sudden changes from standard-to-novel-to-standard might have influenced responding. To help control for effects of sudden change, Phase 1 compared responding in the standard-novel-standard sequence to responding in a standardstandard-standard sequence in which the sudden changes

in music were the same. But a second possible confound arises in that reinforcers were withheld for the entire 3min probe interval, and this, too, might have influenced responding. To help control for absence of reinforcement effects, Phase 1 counterbalanced the ordering of sequences: The standard-standard-standard sequence sometimes preceded, and other times followed, the standard-novel-standard sequence. In Phase 2, we further examined the possible influences of change in stimuli and absence of reinforcement by presenting the standard Bach piece as though it were a novel excerpt. That is, we compared a Bach-BachBach sequence with a Bach-Stravinsky-Bach sequence. Here, all excerpts were from the originally learned standard Bach and Stravinsky pieces so as to explore the effects of the probe procedure itself. Phase 3. Finally, to compare pigeons with people, we asked seven college students with varying musical backgrounds to rate the similarity between the same probe pieces and the original Bach and Stravinsky standard pieces. The students listened to 1.5 min of Bach and Stravinsky each, named "A" and "B," respectively. They then listened to 30 15-s excerpts from the same novel pieces presented to the pigeons (six 15-s excerpts from each of five novel probe pieces) and, after hearing each excerpt, the subjects were asked to mark "A" or "B" on their answer sheets depending on whether the excerpt sounded "more similar to A or B."

Results and Discussion The top panel of Figure 4 ishows performances under the first probe condition, which contained a novel piece by Stravinsky, his Firebird Suite for orchestra. Date points above 50% indicate that the bird is responding more on the "Bach disk" (left disk) than on the "Stravinsky disk" (right disk) and date below 50% indicate more responses to the "Stravinsky disk." The median of all probes containing this excerpt is plotted for each bird. One of the two curves drawn shows the percentage of responses under the Bach-Firebird Suite-Bach sequence (the dashed lines), whereas the other shows the percentage of responses under the standard Bach-Stravinsky-Bach sequence (the solid lines). The important results are that, for all subjects, Bach-Firebird Suite-Bach curves are similar to the standard Bach-Stravinsky-Bach curves. Furthermore, the average curve well represents the individuals. All birds responded to the Firebird Suite as they did to the standard Rite of Spring, although the birds presumably had never before heard the Firebird Suite. To conserve space, only the averages of the four subjects' performances are shown in the bottom of Figure 4 for the remaining probes. Again, the averages shown well represent the

145


individual performances. The results indicate that the birds were probably not responding to attributes unique to the standard pieces but rather to some general and probably complex musical attribute. At first, it appeared that the birds were responding on the "Bach disk" to organ music (the Bach and Buxtehude pieces) and on the-"Stravinsky disk" to orchestral or

chamber music (the Stravinsky standard, Stravinsky's Firebird Suite, Vivaldi, and Carter works). However, responses to the Piston organ work, a contemporary piece, were primarily on the "Stravinsky disk," a judgment similar to that made by human observers (see below); and responses to the Scarlatti harpsichord sonata were primarily on the "Bach disk."

Average x o a

2 3

1 2 3

PROBE

1 2 3

Novel Probe Sequence Standard Sequence Standard Bach Stimulus Standard Stravinsky Stimulus Prabe Stimulus

2 3

PERIODS

Figure 4. The top shows the percentage of responses on the left disk by each of four pigeons and the average of the four birds when Stravinsky's Firebird Suite was the novel probe piece. (Values above 50% indicate more responses on the left "Bach disk" than on the right "Stravinsky disk." Values below 50% indicate relatively more responses on the "Stravinsky disk." When the squares [the percentage of responses to the novel probe stimulus] are relatively close to the circles [the percentage of responses to the Stravinsky stimulus], as in the upper graphs, the birds responded to the novel probe as they did to the standard Stravinsky. When the squares are relatively dose to the Xs [the percentage of responses to the standard Bach stimulus], the birds responded to the novel piece as they did to the standard Bach. The bottom of the figure shows averages of the four subjects' percentages of responses under the remaining seven novel probe conditions.)

146


Figure 5 shows performances for each subject under the control Bach-Bach-Bach versus Bach-Stravinsky-Bach probes for each subject. The center Bach excerpt in the Bach-BachBach sequence is analogous to the novel stimulus in the previous tests. In this case, however, the center stimulus is from the standard Bach piece. Two results should be noted. First, responding to the central Bach excerpt was similar to responding to the other Bach excerpts, whereas responding to the central Stravinsky excerpt was quite dissimilar to the central Bach excerpt. This result would be expected if the birds were responding to qualitative musical cues rather than simply responding to any abrupt changes in the compositions. Second, there was a tendency in three of the subjects for the percentage of responses on the "Bach disk" to decrease with the three successive intervals of Bach, something also seen in the novel probe intervals, for example, in the Carter probe condition. Because reinforcement was withheld during the three consecutive 30-s Bach-Bach-Bach probe intervals, responding had a tendency to approach 50%. But the basic probe method was validated in that the birds responded similarly to the center Bach excerpt as to the surrounding Bach excerpts; and the birds responded differently to the Stravinsky excerpt than to the surrounding Bach excerpts.

#67 1 2 3

#96

#62

#16

I 2 3 I 2 3 PROBE PERIODS

I

2

3

Figure 5. Percentage of responses on the left, or "Bach," disk for four pigeons when excerpts from the standard Bach piece replaced novel excerpts. (The Xs show percentage of responses to the left disk when Bach was played, and the circles show percentage of responses to the left disk when Stravinsky was played.)

The results of asking people to make similarity judgments on these pieces of music were as follows: 95% of their responses were B (Stravinsky-like) when Stravinsky's Firebird Suite played, 100% were A (Bach-like) for Buxtehude, 81 % were A for Vivaldi, 100% were B for Carter, and 76% were B for Piston (the Scarlatti piece was omitted from this test, but we assume that people would respond to Scarlatti as similar to Bach). Thus except for the Vivaldi violin concerto, people placed musical compositions in the same category as did pigeons. We cannot currently explain the difference between people and pigeon judgments with respect to Vivaldi, nor do we know which attributes were most important in determining the pigeon's choices. But neither is it known exactly how people make musical discriminations (e.g., Shuter, 1968). Discussion This is the first demonstration that pigeons can learn to discriminate between complex musical stimuli. Two birds in Experiment 1 learned to respond when a 1-min Bach selection was played and to withhold responses in the presence of a 1-min Hindemith excerpt. Experiment 2 extended this research to different and longer pieces, used a variety of breeds of pigeons, and measured choices. Four of five birds responded correctly between 70% and 75% of the time, that is, pecked the left disk when Bach was played and the right disk when a Stravinsky piece was played. That the discriminations were not due solely to artifacts of the music-presenting equipment was shown by substituting blank tapes for the music. That the discriminations were not based solely on the absolute intensities of the two pieces was shown by decreasing the intensities without appreciable loss in performance. That the discriminations were relatively difficult was shown by responding that remained below 80% when 20-min Bach and Stravinsky pieces served as discriminative stimuli and by comparison with previous studies in which pigeons learned simple auditory discriminations, for example, discriminations between two simple tones (Jenkins & Harrison, 1960). Exact comparisons are difficult, however, because of differences in procedures. After experience with the Bach and Stravinsky pieces in Experiment 2, the birds cat-


egorized novel musical excerpts in ways generally analogous to human categorizations. That is, pigeons generally responded on the Bach disk to short novel excerpts from Buxtehude and Scarlatti and on the Stravinsky disk to short novel excerpts from the contemporary composers, Piston and Carter, and to a novel Stravinsky piece, his Firebird Suite. Throughout these generalization tests, reinforcements were withheld so as to prohibit learning. In one case the birds differed from people, and it is noteworthy that all four birds differed in the same way: They responded to a novel Vivaldi excerpt as if it were more like Stravinsky than like Bach. This result occurred twice, once when the novel Vivaldi was surrounded by standard Bach music, and once when it was surrounded by standard Stravinsky music, and the results were the same. Thus there may be a consistent difference in how people and pigeons categorize music. Whether this difference was due to the specific experiences of the present birds*—would the Vivaldi have been categorized differently if the pigeons had experienced different standard stimuli?— or to some attribute of the physical stimulus must be determined by future research. However, the apparent similarity in the categorization of people and pigeons with respect to five of the six novel musical episodes is consistent with the important research by Herrnstein and his colleagues on visual categorizations in pigeons. After experience with a relatively small set of visual stimuli, pigeons appear capable of generalizing to an extremely large set of additional stimuli, for example, people, trees, angles, and so forth. As in the present case, the stimuli are extremely complex, thereby making identification of the contrplling stimulus variables exceedingly difficult. Future research must attempt to define the controlling stimuli. Were the pigeons responding mainly to attributes along a simple physical dimension, for example, to intensity or frequency, to relational attributes of the stimuli (see Hulse et al., 1984), or to broader categories or prototypes (Herrnstein & de Villiers, 1980; Resch, 1978)? Recent research by D'Amato and Salmon (1982) and by Hulse et al. (in press, 1984) suggest that these questions might profitably be confronted through simplifying the musical stimuli. For example, a simple tune could be generated and loudness, pitch, tempo and relational aspects then varied sys-

147

tematically. It should be noted, however, that responses to such minimal stimuli may be controlled differently than are responses to complex and rich stimuli that vary continually along multiple dimensions. An alternative mode of analysis, one appropriate for complex stimuli, is suggested by Schneider's (1972) work on pigeon color categorization and by Blough's (1982) demonstration that pigeons' responses to letters of the alphabet could meaningfully be placed in a multidimensional similarity space. If, indeed, pigeons and people similarly categorize a wide variety of musical stimuli, the ability to control the prior experience of the pigeon subjects would facilitate studies of the genesis of musical categorization. A related issue concerns whether the discriminations and categorizations were in some sense "natural" (see Herrnstein & de Villiers, 1980; Mervis & Rosch, 1981). The slowness of acquisition in Experiments 1 and 2 would seem to counterindicate a natural category interpretation. But the systematic categorizations of the novel probe stimuli ice Experiment 3, in the absence of any reinforcement for those categorizations, might indicate that the birds entered the experiments with a differential potential to establish broad categories similar to those of people. In conclusion, a methodological observation. The general assumption that animals are incapable of such complex discriminations and categorizations as those studied here can partly be traced to a reaction against anecdotal and anthropomorphic description;; of complex animal behaviors common in the latter part of the 19th century. The reaction was seminally represented by Lloyd-Morgaci's (1894) canon of parsimony, which adjures psychologists to invoke only the simplest capacities necessary to explain an animal's behavior and to avoid, whenever possible, attributing human characteristics to animals. An alternative hypothesis, suggested by ethologists from Aristotle (see the Historia animalium, Bk. 8, Pt. 2, p. 588a, lines 18-36; cited in Mcfceon, 1941) through Romanes (1883) to Lorenz (1974), is that animals are as similar to people as existing evidence permits. Both the simplicity hypothesis of experimental psychologists and the similarity hypothesis of etnolojjists should lead, through research, to the same identification of animal capacities. If a researcher begins by hypothesizing simple capacities, research

148


might proceed through a series of positive demonstrations of increasingly sophisticated behaviors until negative results are obtained. On the other hand, if similar-to-human capacities are hypothesized, progress may be in the opposite direction—a series of disconfirming results—until positive demonstrations are forthcoming. Both of these strategies, if wrongly used, lead to error. Many observers have prematurely proclaimed similarities between animals and people when the demonstration of similarity was an artifact of method (see Rosenthal, 1965). But the opposite error can be equally pernicious: Absence of reports concerning complex behavior does not imply that an animal is incapable of such behavior. These considerations suggest a general principle. The parsimony of Lloyd-Morgan's canon or Occam's razor may produce powerful, general, and aesthetic explanations when researchers are attempting to explain previously demonstrated phenomena. However, the probability of discovering new behaviors, capacities, or phenomena may be optimized by hypothesizing that the currently unknown behaviors or capacities of another organism are as similar to our own as is permitted by existing evidence. The hypothesis must then, of course, be tested. By taking this "similarity strategy," we have shown a generally unexpected agreement between the discriminative responses of pigeons and people to a sophisticated human art form.

Heise, G. A. (1953). Auditory thresholds in the pigeon. American Journal oj'Psychology, 1, 1-19. Herrnstein, R. J. (1979). Acquisition, generalization, and discrimination reversal of a natural concept. Journal of Experimental Psychology: Animal Behavior Processes, 5, 116-129. Herrnstein, R. J., & de Villiers, P. A. (1980). Fish as a natural category for people and pigeons. In G. H. Bower (Ed.), The psychology of learning and motivation (Vol. 14). New York: Academic Press. Herrnstein, R. J., & Loveland, D. H. (1964). Complex visual concept in the pigeon. Science, 146, 549-551. Herrnstein, R. J., Loveland, D. H., & Cable, C. (1976). Natural concepts in pigeons. Journal of Experimental Psychology: Animal Behavior Processes, 2, 285-311. Honig, W. K. (1969). Attentional factors governing the slope of the generalization gradient. In R. M. Gilbert & N. S. Sutherland (Eds.), Animal discrimination learning. New York: Academic Press. Hulse, S. H., Cynx, J., & Humpal, J. (1984). Absolute and relative pitch discrimination in serial pitch perception by birds. Journal of Experimental Psychology: General, 113, 38-54. Hulse, S. H., Cynx, J., & Humpal, J. (in press). Cognitive processing of pitch and rhythm structures by birds. In H. Roitblat, T. Bever, & H. Terrace (Eds.), Animal cognition. Hillsdale, NJ: Erlbaum. Jenkins, H. M., & Harrison, R. H. (1960). Effects of discrimination training on auditory generalization. Journal of Experimental Psychology, 59, 246-253. Lloyd-Morgan, C. (1894). An Introduction to comparative psychology. London: W. Scott. Lorenz, K. Z. (1974). Analogy as a source of knowledge. Science, 185, 229-234. Lubow, R. E. (1974). Higher-order concept formation in the pigeon. Journal of the Experimental Analysis of Behavior, 21, 475-483. Mackintosh, N. J. (1974). The psychology of animal learning. New \brk: Academic Press. Malott, R. W., & Siddall, J. W. (1972). Acquisition of the people concept in pigeons. Psychological Reports, 31, 3-13. References McKeon, R. (Ed.). (1941). The basic works of Aristotle. Beer, C. G. (1970). Individual recognition of voice in the New York: Random House. social behavior of birds. In D. S. Lehrman, R. A. Hinde, Mervis, C. B., &Rosch, E. (1981). Categorization of natural & E. Shaw (Eds.), Advances in the study of behavior objects. Annual Review of Psychology, 32, 89-115. (Vol. 3). New York: Academic Press. Rosen, E. (1978). Principles of categorization. In E. Rosch Blough, D. S. (1982). Pigeon perception of letters of the & B, B. Lloyd (Eds.), Cognition and categorization. alphabet. Science, 218, 397-398. Hillsdale, NJ: Erlbaum. Blough, D., & Blough, P. (1977). Animal psychophysics. Romanes, G. (1883). Animal intelligence. New York: ApIn W. K. Honig & J. E. R. Staddon (Eds.), Handbook pleton. ofoperant behavior. Englewood Cliffs, NJ: Prentice-Hall. Rosenthal, R. (1965). Clever Hans: A case study of scientific Carter, D. E., & Eckerman, D. A. (1975). Symbolic matchmethod. In O. Pfungst (Ed.), Clever Hans. New York: ing by pigeons: Rate of learning complex discriminations Holt, Rinehart & Winston. predicted from simple discrimination. Science, 187, Schneider, B. (1972). Multidimension scaling of color dif662-664. ference in the pigeon. Perception and Psychophysics, 12, Cerella, J. (1975). Visual classes and natural processing 373-378. in the pigeon. Journal of Experimental Psychology: Hu- Shettleworth, S. J. (1972). Constraints on learning. In man Perception and Performance, 5, 68-77. D. S. Lehrman, R. A. Hinde, & E. Shaw (Eds.), Advances D'Amato, M. R., & Salmon, D. P. (1982). lime discrimin the study of behavior. New \ork: Academic Press. ination in monkeys (Cebus apetta) and in rats. Animal Shuter, R. (1968). The psychology of musical ability. LonLearning & Behavior, 10, 126-134. don: Methuen. Harrison, J. B., & Furumoto, L. (1970). Pigeon audiograms: Comparison of evoked potential and behavioral Received February 11, 1983 thresholds in individual birds. Journal of Auditory Research, 1, 33-42. Revision received May 2, 1983