Cover image for The Alex studies : cognitive and communicative abilities of grey parrots
The Alex studies : cognitive and communicative abilities of grey parrots
Pepperberg, Irene M. (Irene Maxine)
Publication Information:
Cambridge, Mass. : Harvard University Press, 1999.
Physical Description:
x, 434 pages : illustrations ; 25 cm
Format :


Call Number
Material Type
Home Location
Item Holds
QL696.P7 P46 1999 Adult Non-Fiction Non-Fiction Area

On Order



Can a parrot understand complex concepts and mean what is says? Since the early s, most studies on animal-human communication have focused on great apes and a few cetacean species. Birds were rarely used in similar studies on the grounds that they were merely talented mimics -that they were, after all, birdbrains. Experiments performed primarily on pigeons in Skinner boxes demonstrated capacities inferior to those of mammals; these results were thought to reflect the capacities of all birds, despite evidence suggesting that species such a s jays, crows, and parrots might be capable of more impressive cognitive feats.

Author Notes

Scientist Irene Maxine Pepperberg was born in Brooklyn, New York on April 1, 1949. She received her B.S. from the Massachusetts Institute of Technology and her M.A. and Ph.D. from Harvard University. Pepperberg is an adjunct psychology professor at Brandeis University and active in wildlife conservation. She is also the president and founder of The Alex Foundation, a non-profit organization.

Pepperberg's studies focus on animal cognition, animal behavior, and comparative psychology, and she is well-known for her successful work in teaching Alex, an African Grey Parrot, a large vocabulary and the ability to identify objects by color, shape, number, and material.

Pepperberg has published many scholarly books and articles, which appear in journals including Animal Cognition and the Journal of Comparative Psychology. She also wrote the New York Times-bestseller Alex and Me: How a Scientist and a Parrot Uncovered a Hidden World of Animal Intelligence--and Formed a Deep Bond in the Process.

(Bowker Author Biography)



Chapter One Can We Really Communicate with a Bird? Let a man decide upon his favorite animal and make a study of it, learning its innocent ways. Let him learn to understand its sounds and motions. The animals want to communicate with man, but Wakan-Tanka does not intend they shall do so directly--man must do the greater part in securing an understanding. Brave Buffalo, a Teton Sioux (Densmore 1918:172) In this chapter I describe the training and testing techniques I developed to study cognition and communication in Grey parrots. I first provide a brief history of the projects and failures that preceded my research and what I deduced were the possible reasons for these failures. I describe the one project to succeed in training birds to acquire human speech, albeit without comprehension, and the likely reasons for its success. I explain how my techniques were derived from a theoretical framework that examined the effect of social interaction on learning; I describe this theoretical framework in detail. I conclude with a somewhat technical description of the testing procedures and rigorous controls I used to determine what my parrot had learned and to ensure that his responses were based on his understanding of the questions and concepts and not on extraneous cues. Previous Failures to Establish Two-Way Communication with "Talking" Birds I was not the only scientist to reason that the vocal ability of mimetic birds, coupled with their considerable intelligence, should enable them to learn to communicate with humans using speechlike sounds. Mowrer (1950, 1952, 1958) was one of the first to investigate this possibility. He studied mynahs (Gracula religiosa) , magpies (Pica pica) , crows (species not reported), and several psittacids (budgerigars, Melopsittacus undulatus; a Yellow-headed parrot, Amazona ochrocephala ; and a Grey parrot), and used standard psychological techniques--basically the operant methodology described in the previous chapter--prevalent at the time. Interestingly, although his birds were trained according to principles of association and reward, they were neither socially isolated nor placed in operant chambers for their sessions.     Mowrer (1952, 1969) used procedures such as the following to teach a bird to produce "Hello" upon the appearance of its trainers. The first step was to make the bird totally dependent upon its trainers for food, water, and social attention--that is, a bird received food and water only in the presence of its trainers. Next, trainers produced "certain characteristic noises" (p. 264), for example, "Hello," followed by a positive action such as a trainer appearing from behind a screen or uncovering the cage. The noise would then be considered a signal (in the parlance of the behaviorists, a "conditioned stimulus") for subsequent appearance of the positive action (a reward, or "primary reinforcer"). According to the behaviorists, moreover, the sound itself becomes reinforcing, if only secondarily (the "secondary reinforcer"), because of what it predicts (Mowrer 1954). If a bird began to produce the sound itself, it received this secondary reinforcement. Production of the sound would, theoretically, increase more quickly if an additional reward occurred for vocalizing (another "primary reinforcer," such as a favorite food). Thus another association would be formed, between vocalizing and a lessening of hunger, and a bird would be expected to produce the sound with increasing frequency. Mowrer introduced several different words and phrases, but the reward for all vocalizations was food. The idea was that, after a bird emitted vocalizations with some frequency, it could be trained to produce the utterance only in the original, appropriate context (on the appearance of the trainer) by providing the food only when the vocalization was emitted in such a situation.     Mowrer's birds acquired few vocalizations. His use of food rewards that directly related neither to the task being taught nor the skill being targeted (such as saying "Hello" when the trainer appeared) probably delayed or possibly prevented learning: Most likely, birds confounded the label of the object or action to be taught with that of the unrelated food reward (Pepperberg 1978, 1981; see Bruner 1978; Greenfield 1978; Miles 1983). That is, his birds apparently connected reproduction of human sounds with the inevitable appearance of food (a salient object to a hungry bird) rather than with their actual referents, for example, "Hello" and the appearance of the trainer. Birds clearly did not realize that a trainer's appearance was the relevant stimulus for producing "Hello." Attempts to obtain food by producing "Hello" when a trainer was already in place would eventually fail: The trainer would consider that the vocalization had been used inappropriately and provide no reward; a bird's production of the strange sound ("Hello") would consequently lessen (in behaviorist terms, "be extinguished"). Moreover, because a bird received food for whatever it produced, it may have stopped learning after acquiring one or two utterances that were sufficient for decreasing its hunger.     Some researchers, possibly believing that Mowrer's social setting was responsible for his failure, attempted to train mimetic birds under more rigorous operant conditions. Ginsburg (1960) managed to place some natural vocalizations of budgerigars under stimulus control--had birds respond with a particular sound after experiencing a particular stimulus--and Gramza (1970) trained budgerigars to mimic tones and musical phrases, but neither tried to replicate Mowrer's goal of achieving interactive communication. Other scientists placed mynahs in soundproof boxes and played recordings of human vocalizations followed by presentation of food pellets; their birds also failed to acquire much in the way of trained vocalizations, although Ginsburg's birds learned to produce two simple utterances in two distinct contexts (Ginsburg 1963; Gossette 1969; Grosslight, Zaynor, and Lively 1964; Grosslight and Zaynor 1967). The puzzling aspect of these failures was why these birds, which were so vocal in the wild and which learned allospecific vocalizations so readily in the informal setting of a home (Amsler 1947; Boosey 1947; Hensley 1980), were incapable of significant vocal learning under well-controlled laboratory conditions.     The answer was not immediately forthcoming and required three breakthroughs, one in data collection and two in research methodology (Pepperberg 1988c). First, researchers had to document the experiences and behavior of wild mimetic birds. Second, laboratory scientists had to be shown how their training paradigms differed from mimetic birds' natural learning situations and how adapting laboratory conditions to reproduce a natural environment might affect their results. Third, a theoretical framework was needed to integrate data from the wild into a valid training procedure. Learning by Mimetic Birds in Natural Environments and in Certain Laboratory Settings Until the late 1960s, little evidence existed for imitation or for any vocal learning for wild mimetic birds. Researchers thus wondered if vocal learning in captivity was an artifact of unnatural housing conditions (Thorpe 1964). At that time, however, Nottebohm found that Orange-winged Amazon parrots (Amazona amazonica) nesting within audible range of one another in Trinidad often shared calls, whereas colonies of the same species living in separate habitats had completely different dialects. Dialects were so different as to suggest initially (and erroneously) that birds inhabiting the various areas were different species (Nottebohm and Nottebohm 1969). These data made sense only if parrots learned their vocalizations through interactions with parents, flock members, and other organisms (Nottebohm 1970)--that is, if parrots mimicked, but generally other birds of the same species. At roughly the same time, Bertram (1970) showed that wild mynahs predominantly shared calls with and appeared to learn vocalizations from neighboring mynahs and, on occasion, other animals in their habitat (Tenaza 1976). Wickler (1976, 1980) subsequently suggested that the extensive duetting often observed in mated parrot pairs came about through a complex socialization process involving learning: To create a duet unique to the pair, both individuals synergistically adapt and adjust their repertoire (Mebes 1978).     Before too long, some laboratories incorporated social interactions into their experimental designs. If Nottebohm (1970) was correct, some interaction between birds--or at least observation--should be involved in the acquisition process. To determine optimal conditions for allospecific learning, Todt (1975a) used Grey parrots to investigate what might happen if training involved social interaction. He developed the model/rival (M/R) technique in which humans assume roles played by psittacine peers in the wild. Humans thus demonstrate to the parrots the types of interactive vocalizations to be learned. In Todt's procedure, one human is exclusively the principal trainer of each parrot, asking questions and providing increased visual and vocal attention for appropriate responses. Another human is exclusively the model for the parrot's behavior and simultaneously the parrot's rival for the attention of the principal trainer. So, for example, the trainer says, "What's your name?" and the human model/rival responds, "My name is Lora." Such human interchanges are similar to duets observed between parrots in large aviaries (Mebes 1978). Todt's parrots learned the model/rival's response often in less than a day, in striking contrast to the slow and sparse acquisition in operant paradigms (compare Grosslight et al. 1964; Grosslight and Zaynor 1967). The rapidity with which Todt's birds acquired human speech was impressive, but the phrases he used did not allow him to show if a bird understood their meaning. That is, words and phrases did not refer to specific objects or actions, such as "tickle," to which an experimenter could respond by scratching the bird's head. Thus Todt's birds may have learned a human-imposed form of antiphonal duetting (an elaborate form of contact calling for interacting with social peers; Thorpe and North 1964; Thorpe 1974) or a simple conditioned response (e.g., Lenneberg 1971, 1973). Also, Todt's parrots vocally interacted solely with their particular trainer and learned only the phrase or sentence spoken by the model/rival, never that of the principal trainer. Todt's intent, however, had been not to train birds to communicate meaningfully with humans, but only to determine optimal learning conditions. Subsequent studies showed that another mimid, the starling (Sturnus vulgaris) , acquires human utterances only in conjunction with social interaction (West, Straud, and King 1983), and that white-crowned sparrows (Zonotrichia leucophyrs) learn allospecific song (e.g., of a strawberry finch, Amandava amadava ) from live but not taped tutors (Baptista and Petrinovich 1984, 1986).     Social interaction, therefore, seemed important for learning, but none of these studies had established actual communication between humans and birds. In addition, scientists wondered about the extent of social interaction necessary for learning communication skills. At present, some researchers argue that only minimal input is needed even for children to learn to communicate appropriately (Goldin-Meadow 1997, but see Menyuk, Liebergott, and Schultz 1995; Locke and Snow 1997); and when I began my research, isolated white-crowned sparrows had been shown to learn conspecific song after hearing only a limited number of taped renditions (12 songs/day for 21 days; Petrinovich 1985). Why then should we be concerned with social interaction? The reason is that social interaction is a part of social context and we need to determine its specific role in learning, particularly allospecific learning. Context includes, for example, the quality and quantity of input during interaction, the relative status of individuals who are interacting, the environment in which interaction occurs, and the feedback a learner receives for attempting the behavior it is learning (Pepperberg 1993b). Interaction thus may not provide all components necessary for learning to communicate, but may provide components that facilitate such learning. Two examples illustrate this point (Pepperberg 1997).     First, birds tutored in a laboratory may not understand how to use their learned vocalizations. We now know that birds learn not only what to sing but also when to sing and the appropriate context for specific songs (King and West 1983, 1989; Kroodsma 1988; Spector, McKim, and Kroodsma 1989); such learning occurs when birds are taught in a social context (review in Brown and Farabaugh 1997). We also know that wild birds respond less vigorously to songs of laboratory-reared conspecific isolates than to songs of wild conspecifics (see, e.g., Thielcke 1973; Shiovitz 1975; Searcy, Marler, and Peters 1985). Something appears to be missing in the overall singing behavior when it is learned in isolation in a laboratory.     Second, social interaction may not be the reason learning occurs but rather may facilitate learning or modify the speed and amount of learning; moreover, the type and amount of interaction may be important (see, e.g., Pepperberg 1985). For example, positive correlations often exist between high model status and the rate and amount learned in both humans (Mischel and Liebert 1967; Bandura 1977) and birds (Payne 1978, 1981, 1982, 1983; Mundinger 1979; Baptista and Morton 1982; Snow and Snow 1983; McGregor and Krebs 1984). Boys imitated actions of camp mates regarded as having high status by their peers far more often than actions of other group members (Lippitt, Polansky, and Rosen 1952; cf. Dollinger and Thelen 1978); birds may react similarly to different song tutors (Payne and Groschupf 1984). And if no single model seems "superior" and modeled patterns are diverse, both humans (Bandura 1977) and birds (see, e.g., Laws 1994) may combine elements of modeled patterns to develop a repertoire. Furthermore, a live tutor may simply be more effective (Pepperberg 1997): Certain birds learn more from live tutors than from tapes; some birds choose to learn songs of live tutors with whom they can interact rather than songs presented by tape or noninteractive tutors (Waser and Marler 1977; Kroodsma 1978; Todt, Hultsch, and Heike 1979; Payne 1981; Kroodsma and Pickert 1984b; note Brown and Farabaugh 1997). Data showing, for example, that juvenile Bewick's wrens (Thryomanes bewickii) , ostensibly before learning their final adult song pattern, countersing with adults using "nearly perfect renditions of the adult songs" (Kroodsma 1974:360), suggest that the effect of social interaction on song acquisition is not a laboratory artifact (see Baptista 1983 for comparable data on white-crowned sparrows).     Clearly, social interaction is important for learning, but is not the only factor involved, and we must return to issues of social context and type of input. Todt (1975a), for example, varied neither types of input nor contexts in which his birds were taught. Might these factors also be important for learning a communication code? Researchers as diverse as Piaget (1952), Vygotsky (1962), and Bandura (1971a) had described how intellectual, environmental, and social contexts affect human learning; might context affect a bird's ability to learn to communicate with humans? Contextual Effects on Learning For Piaget (1952), all intellectual development arises from a subject's continuous interaction with its specific environment. Development proceeds as novel experiences are assimilated into and modify preexisting patterns, which themselves were formed during the subject's reactions to prior experiences. A crucial factor is that the subject's environment provides experiences that encourage such "assimilation" and "accommodation" (see, e.g., Doré and Dumas 1987). Though not directly stated by Piaget, an idea inferred from his writings is that a concept or behavior is more likely to be assimilated if it has functional value for the student, particularly if this functionality is demonstrated explicitly (Wadsworth 1978; see Bandura 1971a; Merrill et al. 1996). For example, teaching the concept of color might proceed more quickly if a subject is shown that different colors represent different flavors of candy: The subject might be motivated to distinguish among differently colored candies and to learn to label red to obtain the cherry flavor, and yellow to refuse the lemon.     A somewhat different but complementary idea has been inferred from Vygotsky's writings (1962, 1978): that "assimilation" and "accommodation" involve use of intellectual skills developed in one context to solve problems in a new one (e.g., see Rozin 1976; Rogoff 1984; Wertsch 1985). A crucial factor is then the existence of appropriate contextual support, or "scaffolding" (Bruner 1977): a bridge showing how to transfer skills between familiar and novel situations by explicit demonstration or emphasis of features shared by the two contexts (Rogoff and Gardner 1984). Scaffolding may involve comments or questions such as "What did we say when we saw Grandma yesterday?" when a child sees a neighbor and a parent is trying to instill appropriate greeting behavior. Similarly, Bandura (1971a) suggests that learning is most effective when subjects observe demonstration of--and themselves practice--novel, targeted behavior in a familiar context. Thus a child might learn a new shape most efficiently if the teacher uses a familiar object of that shape and presents the task in the context of previously learned shapes and familiar objects.     This idea makes a strong case for involving environmental context in the learning process, particularly with respect to communication: Communication is, after all, a social process, and thus acquisition, as noted above, should occur most readily in a social situation. Consequently, researchers conducting studies to determine how organisms learn to communicate must delineate the exact role (or roles) played by social input and its context, and must search for specific mechanisms through which such input can influence learning. In particular, the material discussed above suggests that learning to communicate--acquiring the ability to convey intent and respond to the intent of other individuals (see Smith 1977)--whether the communication is between humans, between birds, or between humans and animals, requires more than simple interaction with a tutor. I have previously argued that for true communication to develop, tutoring must expose a student to three components of the desired communication code (Pepperberg 1991): (1) the code's semantics (the meaning of its individual elements), (2) its syntax (the appropriate rules for combining its elements), and (3) its pragmatics (how the code is used and how its use affects other individuals). How do these components affect learning and what input provides all three components? Determining how different input affects learning can be simplified if a conceptual framework already exists that identifies the critical factors to be studied and provides a prescription for teaching. I believe that human social psychology, in the form of social modeling theory (see, e.g., Bandura 1971a, 1977), provides just such a framework for studies of communication (Pepperberg 1986a,b, 1988c). Social Modeling Theory and Its Application to Birds Social modeling theory was the outcome of social psychologists' successful attempts to determine the underlying mechanisms of what they considered real-world learning, in contrast to the behavioristic associationist paradigm. Social psychologists, for example, reasoned that an animal whose survival depended upon differential reinforcement of trial-and-error learning (the basis for the operant paradigm) would rarely live long enough to learn the task in question (Bandura 1971a); they posited similar bases for cultural learning in humans. They proposed that "provision of models not only serves to accelerate the learning process, but also, in cases where errors are dangerous or costly, becomes an essential means of transmitting behavior patterns" (Bandura 1963:54). Other aspects of modeling theory devolved from these same researchers' efforts to devise procedures to enable humans to overcome strong inhibitions or phobias; the idea was that by determining how the toughest learning problems were resolved, scientists could make inferences about general learning mechanisms (Bandura 1971b). Eventually these ideas were used to teach aspects of communication to humans (see, e.g., Brown 1976; Snow and Hoefnagle-Höhle 1978; Pepperberg 1981, 1985; Fey 1986).     Social modeling theory systematically identifies specific contributions to the learning process that may not otherwise be easily distinguished. The theory encompasses several levels of interaction (see, e.g., Pepperberg 1993b). One level separates learning situations into three functionally distinct categories: two types of active input--social modeling and social interaction--and the passive phenomenon of observational learning. A second level involves determining the optimal form of input for a given type of learning. For the purposes of this chapter, the modeling situation provides the best examples for explaining the theory.     Within every aspect of the modeling paradigm, the learner is an active force in the learning process, and trainers motivate as well as model. Unlike a subject in an operant paradigm, who learns to respond to a few simple stimuli (e.g., a ball, a block) with actions acquired by trial and error (e.g., point to one item rather than to another) and is rewarded with an unrelated item (e.g., food), a learner in a modeling paradigm uses the modeled act and the modelers to direct the course of action. Optimally, modelers signal which environmental aspects should be noted, emphasize common attributes--and thus possible underlying rules--of diverse acts, provide contextual reasons for the acts, and demonstrate consequences of the acts (note the parallels with the work of Piaget and Vygotsky). An example is a modeled interaction for a child who lacks language, such as an autistic individual. The child is shown a tray of objects and observes the following interactions between two trainers: Trainer 1: (points to a block) "What's this?" Trainer 2: "Block." Trainer 1: "Yes, that's a block . Here's the block ." The block is given to Trainer 2, who plays with it for a few moments. The roles of questioner and receiver are then exchanged; later, a ball becomes the targeted object in the session. The interactions clearly depict speaker-listener-respondent relationships, in which both an action and the consequences of the action are clearly demonstrated by an interactive tutor and model.     Social modeling theory thus emphasizes how attention, comprehension, and motivation affect learning. The theory consists of a set of principles--the second level referred to above--that describe the optimal form of social input for any type of learning (Bandura 1971a, 1977). Because some principles presume the existence of cognitive abilities (such as symbolic coding, cognitive organization, and mental rehearsal) that cannot be unquestionably attributed to all nonhumans (cf. Griffin 1985), not all principles may be relevant for animal studies. I believe, however, that a subset of four principles is clearly applicable to avian vocal learning (see, e.g., Pepperberg 1986a,b).     One principle states that the student's level of competence must be taken into account (note Jouanjean-L'Antoëne 1997). Human children, for example, most easily and most often imitate whatever is just slightly beyond their current abilities (Piaget 1954a; Ryan 1973; Scollon 1976; Nelson 1978, cited in Krashen 1983; Krashen 1980, 1982; Kuczaj 1982a,b, 1983; Masur 1988). Interactions that model a new behavior that differs only slightly from an existing behavior or that encode only slightly novel information are most easily learned. Thus children generally acquire additional color labels more quickly than their first color labels. Some songbirds seem to respond similarly, in that they more easily acquire allospecific vocalizations that resemble their natural songs in, for example, tonal pattern or syntax (see Pepperberg 1997 for detailed examples).     A corollary is that for learning to continue, tutor/models must constantly adjust their demonstration to take into account--and continue to challenge--a student's increasing knowledge. Finding the appropriate level of input can be difficult: Input that is too simple may be ignored because the learner loses interest; input that is too advanced and not understood may similarly be ignored. Thus trainers and student must work in concert. Trainers must interact with the student, determine the student's current level, and make upward or downward adjustments; the process is recursive and continuous. Humans do learn most effectively from tutors who actively adjust their input (see, e.g., Bruner 1983; Dunham, Dunham, and Curwin 1993; Tomasello 1992; Menyuk, Liebergott, and Schultz 1995; Peters 1996; Landry et al. 1997). For songbirds, evidence is circumstantial, because we do not know if adults intentionally teach song to juveniles. A female cowbird (Molothrus ater ater) does, however, actively and continually direct male singing by a wingstroke display (West and King 1988), but the extent to which she adjusts her input is not easily analyzed (see, e.g., Caro and Hauser 1992). Researchers have also studied whether different types of adult-juvenile interactions and different models presented by adult male-female pairs during the juveniles' song acquisition period affect learning in birds like zebra finches ( Taeniopygia guttata; see, e.g., Williams 1990; Williams, Kilander, and Sotanskl 1993; Mann and Slater 1994, 1995; see also Zann 1997), but much remains to be learned.     A second principle of social modeling theory states that modeling must help the student understand how new material relates to current problems and what advantage is conferred by learning new material (again, note the parallels with Piaget and Vygotsky). Thus training is most effective when two conditions are met: (1) the student sees and then practices the targeted behavior under conditions similar to those in its regular environment, and (2) the appropriate use and consequence of the behavior are explicitly demonstrated (Bandura 1971, 1977; Brown 1976; Harris et al. 1986; Pierce and Schreibman 1995; note Schwartz and Terrell 1983; Baptista and Petrinovich 1984, 1986; Payne, Payne, and Doehlert 1984; Lock 1991). Thus, for example, a child who is shown exactly how to request a toy, or a bird that sees exactly how a particular vocalization is used in an aggressive encounter, are both likely to learn more readily than if they were in situations without such demonstrations.     A third principle states that the more intense the interaction between a student and its models, the more effective is the training. Intensity--the extent to which tutors arouse a response in a student (Bandura 1977; Locke and Snow 1997)--is determined from direct observation of the interactants (e.g., by recording emotional reactions) or from indirect measures (e.g., blood pressure or hormone levels). One implication, supported by data reviewed in Pepperberg and Neapolitan (1988), is that, for both humans and birds, intense interaction requires one or more tutors per student. Of course, increasing the intensity of interaction may not always increase learning: Overly nurturant models may inhibit learning by preventing a student from experimenting on his or her own (see Rogoff 1990; note the tie-in to the corollary concerning challenging a student intellectually); overly aggressive models may arouse fear or counter-aggression strong enough to block processing of any input (see Casey and Baker 1993 for a possible example for white-crowned sparrows). For some birds, such as zebra finches, intense interaction may affect which tutor is chosen, not necessarily how much is learned (Jones and Slater 1996).     The fourth principle states that if inhibition or resistance exists toward learning, the first three principles are even more important. The fourth principle is particularly relevant to what I call exceptional communication (Pepperberg 1985): communication characterized by vocal learning that is unlikely to occur in the normal course of development. For birds, exceptional learning can take two forms. One is use of allospecific vocalizations by subjects generally expected to acquire functional use of only conspecific vocalizations (e.g., contextual use of the song of unrelated species; see Baptista 1988 and Baptista et al. 1981 for data on exceptional song learning in, respectively, a wild song sparrow, Melospiza melodia , and a Lincoln sparrow, Melospiza lincolnii ). The second is age-independent acquisition of vocalizations in species generally recognized as having a limited "sensitive phase" for vocal learning (for laboratory studies on white-crowned sparrows, see Baptista and Petrinovich 1984, 1986; Petrinovich 1988; Jones, ten Cate, and Slater 1996; cf. Marler 1970; for possible parallels in the wild, see Baptista 1985; Baptista and Morton 1988; review in Baptista and Gaunt 1994; cf. Nelson, Marler, and Morton 1996; Nelson 1998). The term "exceptional" implies some resistance toward acquiring the targeted behavior. Thus, for exceptional learning to occur, social modeling theory predicts that tutor/models be even more attuned to the student's level, that interactions be even more intense, and that demonstrations be even more explicit as to real-world uses and consequences of a targeted behavior than during normal learning. This principle has been important for analyzing exceptional learning in some oscine birds (see, e.g., Neapolitan and Pepperberg 1988; Pepperberg and Schinke-Llano 1991), and we'll see how it is similarly critical for understanding how I adapted social modeling theory to teach a Grey parrot to communicate with humans and to use this communication to examine his cognitive abilities (e.g., Pepperberg 1990d). (Continues...) Copyright © 1999 Irene Maxine Pepperberg. All rights reserved.

Table of Contents

1 Introduction: In Search of King Solomon's Ring
2 Can We Really Communicate with a Bird?
3 Can a Parrot Learn Referential Use of English Speech?
4 Does a Parrot Have Categorical Concepts?
5 Can a Parrot Learn the Concept of Same/Different?
6 Can a Parrot Respond to the Absence of Information?
7 To What Extent Can a Parrot Understand and Use Numerical Concepts?
8 How Can We Be Sure That Alex Understands the Labels in His Repertoire?
9 Can a Parrot Understand Relative Concepts?
10 What Is the Extent of a Parrot's Concept of Object Permanence?
11 Can Any Part of a Parrot's Vocal Behavior Be Classified as "Intentional"?
12 Can a Parrot's Sound Play Assist Its Learning?
13 Can a Parrot's Sound Play Be Transformed into Meaningful Vocalizations?
14 What Input Is Needed to Teach a Parrot a Human-based Communication Code?
15 How Similar to Human Speech Is That Produced by a Parrot?
16 How Does a Grey Parrot Produce Human Speech Sounds?
17 Conclusion: What Are the Implications of Alex's Data?