Avian Visual Cognition

Stimulus Control in Complex Visual Arrays

  Michael E. Young & Edward A. Wasserman
Departments of Psychology, Southern Illinois University & University of Iowa

 avcrule.gif (935 bytes) 

Next Section:
Simultaneous Same-Different Discrimination









It should come as no surprise that birds are able to categorize visual stimuli as a function of their perceptual similarity (e.g., Blough, 2001). If things look alike, then it is clear that they should be treated alike. Abstract categorization, however, differs from perceptual categorization in terms of the stimulus features that are necessary to perform the categorization. Perceptual categorization relies on the members of a category being similar to one another; for example, birds look more like other birds than they look like dogs, chairs, or computers (see the contributions by Blough, Huber, and Urcuioli, this volume). Abstract categorization is required when responding must rely on the relations among a group of visual stimuli; for example, are the items all the same or all different? Are they displayed in a symmetric or asymmetric arrangement? Are they aligned in order of size? Abstract categories thus transcend the individual items that make up the visual display and require a focus on the relations among those items.

In this chapter, we will present our recent work on the pigeon's ability to categorize pictorial displays as involving items that are the same as or different from one another. You will be given the opportunity to work through some sample experiments and be shown the results of the pigeon studies. By the time you are done, you will know why pigeons have so much difficulty learning to categorize 2 items as the same as or different from one another and why they find it much easier when there are 16 items that are all the same or all different. We have also begun work on some human analog experiments and you will discover that pigeons and people often behave quite similarly on complex cognitive tasks.

Chapter Outline & Navigation

I.   Introduction

II.  Simultaneous Same-Different Discrimination
       Mixture manipulations

III. Information Theoretical Account of Same-Different Discrimination

IV.  Successive Same-Different Discrimination
Mixture manipulations
          List length manipulations
          One vs. three mechanisms

V.   Conclusions
Recent related work with other nonhuman animals
          Related work with humans
          Current and future directions      

VI.  References

I. Introduction 

Human beings and other animals are constantly confronted with an extraordinarily complex array of external stimuli. Yet, sense is somehow made of these varied stimuli. One way of reducing the demands on an organism's sensory and information-processing systems is for it to treat similar stimuli as members of a single class; by so doing, substantial cognitive economy can be achieved, thus freeing the animal's adaptive machinery to deal with other competing exigencies of survival. In addition, categorical processing permits an organism to identify novel stimuli as members of a particular class and to generalize knowledge about that class to these new members. Thus, an organism need not be bound to respond to only those stimuli with which it has had prior experience, thereby enhancing its ability to cope with a continually changing world.

Although theorists often extol the adaptive virtues of categorization and conceptualization, we remain far from understanding exactly how organisms process stimuli so as to partition the world into classes of related objects and events. Indeed, given the early writings on the subject, we should wonder whether nonhuman animals are even capable of conceptual behavior. Over a century ago, Charles Darwin noted, "If one may judge from various articles which have been published lately, the greatest stress seems to be laid on the supposed entire absence in animals of the power of abstraction, or of forming general concepts" (Darwin, 1871/1920, p. 84). Twenty-five years later, Lloyd Morgan denied animals the ability to behave conceptually. To do so, he said, requires that we "neglect all that is variable and focus the attention on the uniform relation. [Then] we have reached a conception, and this conception is not concrete, particular, and individual, but abstract, general, and of universal application" (1896, p. 263). Morgan believed that only adult humans (not even children) are capable of conceptualization. The vestiges of this belief still persist in the field of experimental psychology.

Several recent lines of research are radically changing that initial opinion (Wasserman, 1995). One of those new lines of research concerns the acquisition of the same-different concept by animals. The classification of two or more items as the same as or different from one another requires a level of abstract conceptualization previously thought to be unique to human beings.

Evidence of this ability in nonhuman primates is now substantial (e.g., Premack, 1976, 1983; Wright, Santiago, Urcuioli, & Sands, 1983); however, initial evidence of same-different conceptualization in the pigeon was promising, but weak. Indeed, early attempts to train pigeons to discriminate same from different stimuli led to rather pessimistic conclusions: "Because [a Same-Different concept] transcends the items themselves there is a possibility that it may be a difficult concept to learn. There is the possibility that monkeys can abstract the task and learn the Same-Different concept whereas pigeons cannot" (Wright et al., 1983, p. 316).

In one of the earliest studies of pigeon same-different categorization, Wright et al. (1983) trained pigeons to peck one key when two slides of real objects were the same as one another and to peck a second key when the two slides were different from one another. After nearly 20,000 trials of training using 210 different pictures, the pigeons averaged over 80% correct to the familiar pictures, but accuracy dropped to only 62% correct to an unfamilar set of pictures.

Edwards, Jagielo, and Zentall (1983) trained pigeons to peck one key when two line-drawn circles or pluses were presented and to peck a second key when a plus and circle were presented. After 2,208 trials, the pigeons averaged 87% correct; but, upon transfer to differentially reinforced novel displays, accuracy dropped to only 59% during the first 48-trial transfer session in Experiment 1 (where the novel stimuli were red and green colors) and to only 57% during the first 48-trial transfer session in Experiment 2 (where the novel stimuli were flashing or steady white lights). These test results suggest that the pigeon's original learning was primarily item specific.

Most of the prior research on abstract conceptualization in animals has involved the classification of just two visual items as the same as or different from one another (e.g., Edwards et al., 1983; Santiago & Wright, 1984; Wright, Santiago, Sands, Kendrick, & Cook, 1985; Wright et al., 1983). Two items represent the smallest number necessary for a same-different classification; but, it is also possible, and perhaps even easier, to make same-different judgments with displays involving a greater number of items. Wasserman, Click here to view Figure 1Hugart, and Kirkpatrick-Steger (1995) recently reported an experiment in which pigeons were taught to classify displays involving 16 rather than 2 items that were either all the same as one another or all different from one another (Figure 1). The displays were 4 x 4 arrays of 16 computer icons that were chosen from a set of 16 possible icons; the Same displays involved a single randomly selected icon repeated 16 times, whereas the Different displays involved all 16 icons in one of 16 different spatial configurations. Pigeons received food pellet reinforcement for pecking one of two buttons ("same") in the presence of Same displays and for pecking the second button ("different") in the presence of Different displays. The same-different discrimination was rapidly acquired and it supported strong transfer to novel 16-item displays; accuracy on trained displays averaged 83% correct after approximately 13,000 trials and 71% correct on new displays created from a set of 16 untrained icons.

Recent work in our laboratory has made significant strides toward understanding how nonhuman animals solve these simultaneous same-different discriminations (Young & Wasserman, 1997; Young, Wasserman, & Garner, 1997). We have also extended our research to investigate the pigeon's discrimination of lists of successively presented identical items from lists of successively presented different items (Young, Wasserman, & Dalrymple, 1997; Young, Wasserman, Hilfers, & Dalrymple, 1999). In this chapter, we will present a summary of our published work on the same-different concept, introduce some of our latest findings, and consider comparative issues by sharing the results of our recent studies on human discrimination of simultaneous same-different displays.

Next Section Simultaneous Same-Different Discrimination