Multilevel Development: The Ontogeny of Individual and Group Behavior
Abstract and Keywords
The aim of this chapter is to bring back into focus the primacy of behavioral analysis both for its own sake and for integrating and understanding the discoveries made at genetic and neural levels of organization. This chapter argues that this integrative aspect of behavior requires tools and techniques that match the sophistication of those used by research focused more directly on genetic and neural levels. The chapter surveys the behavioral ontogeny of laboratory rats using a dialectic process that interweaves the study of individual and group behavior. The chapter ends with a preview of the directions technology and integration may take in the study of the behavioral development of individual and of groups.
Behavioral neuroscience is a young field that has rapidly generated a large body of knowledge. The growth rate on one side of this body, however, its behavioral side, has been slower than on its now larger, “neuro” side. We think that the behavioral side can—and will—grow dramatically and change in important ways that will benefit the entire field.
In this chapter, we are particularly concerned with multiple levels of analysis and organization. Whether a behavioral neuroscientist emphasizes an organism’s behavior, or a neural system, or some region of the genome that is affected by brain-mediated experience, the subject matter is that of a complex, dynamic system and there arise common conceptual barriers that have been difficult to penetrate (Alberts, 2002).
Reductionism is one traditional approach to such complex phenomena. Typically, one begins at the level of the individual organism and proceeds to look down to lower levels of organization (e.g., neural or endocrine systems, genomic systems, or even proteomic events). Such reductionistic studies involve stripping away the organism and most of its parts or from the parts with which it interacts (i.e., other organisms). This is a powerful strategy for revealing lower-level mechanisms but too often organismal behavior is lost and cannot be retrieved from the reductionistic levels.
Researchers have also been looking up. While remaining anchored at the level of the individual organism, these investigators cross to “higher” levels, usually involving one or more other organisms. In studies of mammalian development, the mother–infant relations can constitute a functional, integrated unit. In species that give birth to multiple off spring, the aggregate of off spring may itself constitute a group.
The nonreductionist strategy has paid good dividends. As we will see, the rewards include (p. 476) discoveries of how mechanisms on a lower level of organization (the individual) create, sometimes additively and at other times multiplicatively, a separable and higher level of organization (the group). We are learning how the two levels, the level of the individual and the emergent level of the group, interact and affect one another. This amounts to an advance in multilevel integration, important to all of behavioral neuroscience. After all, our ultimate goal is not simply to reduce behavior to the expression of genes or the firing of neurons, but rather to understand how lower-level entities such as genes and neurons interact to play a generative role in the emergence of individual and group behavior.
We think that behavioral studies will contribute importantly to expanding and refining our ability to conceptualize how multiple levels of organization operate and, especially, how they develop. Understanding dynamic, living systems must include understanding how they are formed and maintained at multiple levels of organization.
The ascendance of behavioral studies will occur largely in response to the challenge of linking genes, brain, and behavior. Behavior is paramount: It is bodies and interacting bodies that behave, not brains or genes. To discover the links, it is essential to cross levels of organization and to forge interdisciplinary connections. Indeed, multilevel integration is central to behavioral neurobiology itself. The present chapter was written in part to explore some of the ways in which behavioral studies within neurobiology can lead the way toward multilevel understanding and facilitate new and powerful integrative perspectives. It was also written to explore ways in which behavioral methods and approaches can be used as powerful tools for integrating lower-levels of organization with higher, behavioral levels.
Current directions in neurobiology and genetics have been driven by the development of new technologies that have enabled the exploration of previously unexplorable problems. We will argue that analogous technological developments are occurring for behavior that are allowing behavioral studies to take a new and leading role in behavioral neurobiology, largely because new methods have opened up new ways of seeing behavior and, importantly, of handling the massive sets of data that are characteristic of complex systems, such as living organisms and groups of organisms. Thus, we will begin with an overview of some of the technological innovations for the study of behavior.
By themselves, they can do little to solve the integrative problem we pose. With guiding precepts, however, we believe that these behavioral tools (and others yet to be developed) can provide a conceptual framework for the integration of individual and group behavior across multiple levels and through development.
Studies of the Norway rat (Rattus norvegicus) have probably contributed the most to our knowledge pertinent to understanding multilevel development and behavior. The present chapter will reflect the generous contributions of this species, which has long been appreciated for its adaptability to various laboratory conditions, its rapid and dramatic postnatal development, and its rich behavioral repertoire. Mus musculus, of course, is now growing in popularity due to rich knowledge of its genome, but R. norvegicus best helps us tell the story that comprises the present chapter. Indeed, both species typify the gap in levels of analysis this chapter aims to bridge conceptually and methodologically.
The dialectic of this review begins with an overview of new tools that enable new views of individuals and groups. We then turn to precepts to guide our investigative journey from individual to group and group to individual. By reviewing episodes in the study of ontogenies, we will discover how powerful the dialectic process of moving from part to whole, individual to group, and group to individual can be. We then return to some of the new tools and illustrate how they yield new insights and facilitate our integrative aims.
New Tools Enable New Views
New technologies have changed the study of behavior. So profound are these technology-enabled changes that they have engendered new conceptualizations of behavior and of causality, especially in relation to behavioral development. The advent of new views can be found on many levels of analysis, but is especially evident in the study of groups. Behavior in the context of groups is complicated, but new technological tools have made such complex observations increasingly possible and precise. Here we note a few of the more influential technological advances that have altered the ways we view behavior.
Video: Capturing Behavior and Making Time Elastic
The advent of video imaging and video recording, both analog and digital, has steadily trans-formed the study of behavior. Video enables images (p. 477) to be captured (i.e., recorded and stored), replayed, reanalyzed, duplicated, and shared. Infrared capabilities enable imaging in the dark. Video equipment has become increasingly miniaturized and remarkably inexpensive, replacing the older, celluloid film-based technologies that were bulky, slow, and expensive.
Video technologies also enable us to capture behavior in “elastic time.” Standard, commercial formats provide about 30 frame-like shots per second. Time-lapse video, in contrast, samples at a specified, reduced rate, perhaps only 5 times per second. When playback is at the standard rate, the images from each second move at an accelerated pace. Observers can record without direct presence, perhaps continuously for 24 h or more, and then compress time during playback and see the behavior without interruption in a viewing session of 1 or 2 h. With these methods, we see more, we collect more, and come to understand a lot more.
Another way of appreciating the power of a time-lapse view is to contrast it with another, time- sampled view of behavior. “Time-sampling” is accomplished by observing an animal or group of animals at intervals, usually predetermined, fixed intervals. Time-sampled observations are usually recorded “live” with a keyboard or, perhaps voice recording, so they are limited to behaviors that can be discerned reliably in real time. In the same 15 min devoted to time-sampling one target, the timelapse user can sample portions of each second and accumulate six hours of such data! These time-lapse records can be reexamined (frame by frame, if necessary), duplicated, and shared.
Video can also expand time by slowing down the stream of behavior. Such approaches are helpful when the behavior of interest occurs very rapidly or consists of small rapidly expressed components. Even after more than 20 years of observing cowbirds courting, it was not until West and King (1988) used high-speed video to expand time and examine the microstructure of the birds’ behavior, could they discern the female’s brief “wingstroke” that serves as a reinforcing signal, indicating an effective song from the male.
Other high-speed and high-throughput methods, beyond the scope of the present chapter, have been employed productively. These include kinematic methods enabled by video and other digitized tracking technologies, which provide two- and three-dimensional tracking of limb and joint positions. These data, which rapidly become voluminous during most applications, yield quantitative descriptions of movement-in-time that are used to analyze behavioral form and plasticity.
Appearing on the market now are a variety of video-/computer-based systems featuring software designed to recognize and categorize an animal’s behavior. These systems not only can track movement and thus measure activity, but with pattern recognition capabilities, they can reliably categorize activities such as rearing and grooming. Such systems have great promise in replicating the tasks undertaken by some human observers, but more work is needed.
Computerization: Do More, Store More, See More
There are two correct views on the contribution of computerization to the study of behavior. The first, commonly held view, is that computers help us collect more data, store it, and analyze it. Such computerization greatly facilitates what we already do. We can do more per unit time, and so we do. Unbelievably, computer speed has evolved at the awesome pace prescribed by Moore’s law, which states that computational speed will double every 18 months and may continue to do so for another 250 years (Lloyd, 2000). Computers have become really fast and will likely continue doing so.
Linked to blazing computational speed has been a corresponding evolution of affordable storage space. Memory got cheap and it continues to get cheaper. So, we can do more and store more. Because analytic speed has increased, we can collect and store digital video, handle vast data sets for three-dimensional, kinematic analyses, or on-line pattern recognition.
The second and more recent correct view is that computerization has evolved to a level that it now enables to see things in ways we could not have imagined before. Speed and storage boundaries now allow us to collect so much more data that there have been raised new questions about the organization and analysis of such data sets. For example, we have collected precise measures of position, orientation, and activity on each of eight rat pups, as they interact in a huddle. From the formidable data set generated by the group, we could for the first time, describe the dynamic behavior of the group as a whole, by “seeing” for the first time, the flow of aggregon formation—the 22 different combinations of bodies in contact that can be (and are) displayed by eight rat pups in a huddle. Later in this chapter, we examine and discuss the behavior of groups.
(p. 478) Perhaps it is time to start developing a new field called psychobioinformatics. We construe psychobioinformatics to include any technology, technique, or method that allows us to extract, organize, integrate, and analyze behavioral and biological data at multiple levels of organization and timescales. These new tools include not only the imaging technologies just mentioned but also a variety of new algorithms that allow us to use computers in new ways to study behavior. For example, progress is being made in computer-automated tracking and identification of behavior for specific applications (Delcourt et al., 2006; Noldus, Spink, & Tegelenbosch, 2001; Tsibidis & Tavernarakis, 2007), but much more remains to be done (Schank & Koehnle, 2007).
Robotics and Computer Simulation: Hypothesis Testing by Construction
We believe that thorough analysis of behavior and the integration of behavioral processes at different levels require more than traditional hypothesis testing. Empirical methods are necessary but probably insufficient to manage the complexities of behavioral problems. Additional approaches are needed. Whereas traditional empirical approaches accomplish hypothesis testing by falsification, in robotics hypothesis testing is by construction. This approach embodies the methodological precept that if you understand how something works, you should be able to build something like it to test your understanding.
Robotic and simulation-based approaches to hypothesis testing are just beginning to emerge, so it is premature to review and evaluate them. Instead, we can best describe an example and consider its contributions to knowledge of behavior and development. In short, this approach is based on demonstration of the possible, with reliance on elements that resemble biological realities.
Some of robotics’ most powerful manifestations grow out of computational modeling. Central processor unit (CPU) technology is at the heart of both approaches. With the increasing power of CPUs and their miniaturization, we not only can construct multilevel simulation of robots and agents, but it is also possible to construct physical models (i.e., robots). Such robotic applications require the construction of predictive and explanatory models that connect multiple levels of organization. By building both virtual and physical models, we can test by construction (and not merely falsification) the theories and mechanisms we propose as explanations of multilevel development. That is, we can build model systems that behave like we think natural systems behave and then compare the behavior of our artificial systems to natural systems.
In this section, we present some general and basic formulations—precepts—that will help clarify themes that prevail throughout the research described in the remainder of the chapter. These precepts arise from findings propelled by research tools such as those discussed in our early remarks and analyses guided by innovative developmental perspectives. Although our discussions will focus on developmental and behavioral aspects of behavioral neurobiology, we think that the precepts are broadly applicable to the field, especially in the quest for improvements in multilevel integration.
Precept 1: Ontogeny Is a Collective Noun
Ontogeny is not a single process; rather, ontogeny comprises a number of interacting processes at different levels of organization. Thus, the term ontogeny is actually a collective noun referring to interacting ontogenies within ontogenies. That development proceeds on multiple levels is obvious when we begin to enumerate things that develop: neurons, neural systems, functional capabilities, individual behavior, and group behavior.
Such a multitude of levels presents patterns of activities. Obviously, there are interactions among units or “agents” on each level (e.g., neurons, systems, individuals). Less obvious but no less important, however, is that each of these levels can affect others. Behaviors and phenomena are emergent from lower-level interactions and therefore cannot be adequately understood without understanding the dynamic interactions of lower-level components. Interactions across levels are also apparent.
The perspective we will take in this chapter is that development is multilevel and emergent. Our focus will be on two special levels and their interactions: individual and group behavior. We will show that studying individual and group behavior is important as subject and as paradigm not only for understanding development but also for other integrative efforts. The behaving individual is analyzed at a fundamental level and such analysis is strategically positioned for reduction to other familiar, lower levels. Conveniently, the level of the individual is also below another important level, that of the group.
(p. 479) Nevertheless, focusing on just these two levels does not make our task simple. An ontogeny at any level contains ontogenies within it. Our task will not only be to describe these ontogenies and their mechanisms, but also how they interact across and within levels. Because our perspective is one of emergence, it seems untenable to consider ontogenies as programmed in genomes. Rather, they emerge from interacting parts and from ontogenies at numerous levels of organization including the genome. How ontogenies emerge from these interactions is heavily constrained by context.
Precept 2: Development Occurs in Context
Contextual features constrain and generate patterns of behavior. These contextual features include all kinds of environmental elements, such as temperature, gravity, light, sound, and chemical stimuli. Some less obvious, but influential features include other organisms, the geometry of the physical context, body size and shape, as well as the placement and orientation of receptors. Thus, context, as we define it, is always present and behavior is never context-free.
Development is sustained and constrained by context. The mammalian uterus is a context, as is the amniotic sac. Alberts and Cramer (1988) traced the rats’ early postnatal development as a series of adaptations to series of contextual changes that comprised a series of “ontogenetic niches.” These were, in sequence, the uterus, the exterior of the mother’s body, the clump of siblings in the nest, and finally, the social coterie outside of the nest, first experienced by weanlings.
There have been a variety of essays that have explored the interrelations of development, adaptation, and context. Interestingly and importantly, in mammalian development, the study of behavior development is often best investigated in the context of social groups. This implies that mechanisms such as gene expression and regulation are also embedded in social contexts. Gene expression and regulation in a developing animal influences and is influenced by context. In this view, phenotypic expression during development is not merely gene expression in developmental context but it better characterized as a web of interactions within and between levels and contexts. What genes are expressed and how they are expressed are downwardly influenced and regulated by social context (Schank, 2001).
Precept 3: Groups Are Functional Entities
Conspecifics located together at the same time constitute a group. But there are different kinds of groups. For purposes of the present discussion, we will recognize two kinds: groups that are (a) the sum of their parts (i.e., aggregations independent of their interactions) and (b) groups that are more than the sum of their parts (i.e., aggregations emergent from their interactions, some of which can be adapted groups).
In the present framework, the most basic, minimally organized group structure emerges when individuals act or interact in a manner that merely results in physical association with one another. In such instances, we find each individual capable of autonomous behavior. Individuals respond to others, or perhaps they respond to a common cue in the environment, and thus enter into association with others through simple movements and/or sensory interactions (Fraenkel & Gunn, 1940). When such actions and interactions result in a group and the organization and characteristics of the aggregate are the basic, linear sum of the individuals, we have a simple group.
In contrast, emergent and adapted groups are attained when the aggregate displays functional characteristics or capabilities that are different or beyond those of the individuals comprising it. We say that the “group is more than the sum of its parts” and that it is an adapted group. Such groups define themselves in terms of the integration of their activities or the specialized, emergent, functional characteristics of the group. Specialized “functional characteristics” are demonstrable during interactions between levels, especially when one level alters or regulates the other. To use a distinction introduced by evolutionary biologist, Williams (1966), who discussed such properties of aggregations of insects, we will primarily focus on the distinction between a group of adapted pups and an adapted group of pups.
It should be kept in mind, however, that patterns of organization can emerge for which there is no obvious adaptive function. This can occur anytime context is changed, since changes in context affect emergent organization. Nevertheless, our proposed dichotomy of kinds of groups (i.e., group of adapted pups versus an adapted group of pups) is useful, for it helps us to discriminate between different forms of interaction and see their functional consequences. When we recognize such differences, we can better ask how they arise. Much of this chapter consists of different approaches to recognizing and analyzing groups and their organization.
(p. 480) Precept 4: Development Is Emergent
We identify emergence as a cause of development. To be sure, we are again invoking emergence, but previously, in Precept 3, we identified emergence as instances in which a new level of organization arises. The developmental emergence that we recognize here arises on the same level of organization as the interactions that induce the new form or function. This is a framework in which behavior can cause or invent new forms of behavior. This is a framework in which development is emergent.
Development emerges when tissues interact in ways that “induce” new elements (Saunders, 1982). Development emerges when tissue accumulates to produce changes in size and shape. It is often the case that the activity of such growing forms constitutes interaction among cells or systems and these interactions comprise information that alters form and function (Goodwin, 1994). The development of sensory systems provides a wealth of such examples (Gottlieb, 1976). Similarly, exercising nerves, muscles, and bones constitutes interactions and information across the body from which emerge stronger, more efficient, and adaptive movements (Thelen, 1988; Thelen, Kelso, & Fogel, 1987). Development also emerges from interactions across levels, just as was recognized earlier in Precept 3.
Both uses of emergent—as a cause of invention on another level of organization (Precept 3) and as a cause of development (Precept 4)—are valid. In fact, these two forms of emergence typically occur simultaneously. With the development of new capabilities on levels below that of the organism, new ontogenies emerge on higher levels of organization, i.e., on the level of the individual. As developing individuals interact, individual ontogenies create new interactions and thus developmental change is made more likely on the level of the group. Similarly, interactions on the level of the group can create new conditions within which there arises new behavior by the group. And, as we shall see, new developments on the level of the group can also exert downward influences to the level of the individual.
Together, the interactions on multiple levels constitute development. As proffered by Precept 1, there are ontogenies within ontogeny and the interactions that comprise these levels are developmental information. These elements of information comprise the nonprogrammed emergent instructions that determine development.
Behavioral Development on the Levels of the Individual and the Group: Episodes of Discovery
In the next two sections, we review research from our laboratories, from our collaborations, and from the laboratories of others in a manner that we think can help illustrate and illuminate dialectic processes of ontogenesis on the level of the individual and the group. In addition, we hope to demonstrate that there is much that can help us conceptualize and think about multilevel development by “looking up” from the level of the individual to the level of the group. When this upward view is combined with the more traditional downward, reductionistic perspective, we see new integrative possibilities.
We first examine some of the more traditional, descriptive accounts of development on the level of the individual. Even simple description of individual development is not that simple. For example, whether individuals are viewed in their normal context (in situ) or in some approximation of the normal context, or contrastingly, out of [normal] context—in isolation (ex situ)—assumes importance.
Behavior and Development on the Level of the Individual
Descriptive Accounts of Rat Development
Williard Small’s (1899) diary-like report was probably the first systematic description of the development of the domesticated rat. His account is based both on direct observations of pups in a nest box and of individuals removed from the nest for inspection, sometimes with quasiexperimental probes to test whether, for instance, a pup can detect a particular odorant. Presented as a description of “psychic development of the young white rat,” Small included discussion of dimensions of behavior that are purely constructs, such as curiosity, fearfulness, greed, and the like. Small did not attempt to operationalize such constructs, so this aspect of his work is almost fanciful by today’s standards. Small made a lasting contribution, however, and many of his observations of early sensory function have lasted as reliable accounts. His narrative remains a valuable read. It is filled with thoughtful percepts and integrative overviews of development.
Bolles and Woods (1964) conducted an informal type of time-sample study with 13 caged dams and litters. They observed and took notes on entire litters of pups by watching all of the individuals, (p. 481) documenting the ages at which different behaviors were displayed. Featured in this now classic report were “reflex figures” such as body flexions and sniffing, “functional activities” including eating, drinking, grooming, and climbing, as well as a few “social behaviors.” The study provides an overview of “onsets” in pups, such as the onset of walking, running, and grooming behaviors. They also observed the behavior of single, “focal” pups in an additional six litters. Again, the data were mostly qualitative and focused on behavioral categories.
Alberts’ (1984, 2005) reviews of sensory development and general development in rat exemplify another type of descriptive account. The approach used in such accounts is based on assembling and synthesizing findings from many different investigations, from different laboratories, conducted at different times and with a variety of different techniques. From such diversity, the goal is to create an idealized, unified overview of an “average” pup moving through a series of developmental transitions.
Altman and Sudarshan (1975) provided a kind of overview of motor development by testing pups of different ages with a battery of tests and measures. Some were largely observational and others were motoric challenges, such as balancing on narrow supports or hanging onto a wire. Their developmental description is designed to define a set of developmental milestones to characterize ontogenetic change and achievement.
Mechanistic Reactions that Create Complex Behavior, EX SITU
Early in the twentieth century, research groups dedicated to a behavior-based, mechanistic perspective were motivated by a grand view of an evolutionary phylogeny of plant and animal behavior. In this framework, streams of complex behavior are composed of simple responses to stimuli in the immediate environment. A variety of basic reactions to stimuli were identified, including tropisms, taxes, and kineses (Fraenkel & Gunn, 1940).
Nevertheless, there is power in such mechanistic approaches. Recently, we tested individual, 10-dayold rat pups’ behavior on a uniform, temperature- controlled surface that was tilted at a very modest angle (8° or less). Pups were observed to move downhill (Alberts, Motz, & Schank, 2004). In other words, we saw positive geotaxis on substrates tilted at 4° and 8°. Figure 23.1 illustrates the phenomenon of positive geotaxis on modest inclines. With increasing slopes, pups increasingly orient and move downhill. This finding contrasts sharply with the “classic” result reported extensively by Crozier and his colleagues (e.g., Crozier & Pincus, 1926, 1936). It now appears that in their zeal, Crozier and his associates rushed to conclusions that were confounded methodologically and colored by preconceptions (cf., Krieder & Blumberg, 1999; Motz & Alberts, 2005).
Using an automatic frame-grabbing technique and software that helped measure, store, and analyze the x-y coordinates of the pup’s snout, shoulders, and rump, we were able to examine position, angular changes in position, orientation, contact with walls, and velocity. We found that pups initially displayed nondirected activities, the frequency and velocity of which were independent of inclination. The size and configuration of the testing arena made it likely that the pups’ initial movements would lead to contact with a wall, usually within 30 s. Once in contact with a wall, pups on an incline (p. 482) were likely to orient downhill, as illustrated in Figure 23.1. The probability and extent of positive geotaxis on an inclined, walled surface increased as a function of angle of inclination. Pups maintained contact and thus maintained downhill orientation. Average movement velocities during the 8 s following wall contact were significantly greater than during a comparable time sample for pups that had not yet encountered a wall. The contact-induced increase in speed, coupled with their orientation, accounts for the characteristic “positive geotaxis.”
A coordinated, complex response (orientation and movement downhill) can result from an assembly of simpler, separable reactions to proximal cues. It was possible to diagnose and study this form of behavioral organization only by the careful control of numerous other factors such as temperature, texture, lighting, and the vibrations of nearby equipment that influenced the pups’ responses. Clearly, the behavior of an individual involves multiple inputs, simultaneously at play. There is power and parsimony in analyses that focus on the proximal cues in an individual’s immediate context that evoke or shape its behavior.
Development of Behavioral Arousal and Activity
Campbell and associates miniaturized some of the classic stabilimeters for monitoring general locomotor activity of single animals and collected systematic, quantitative data under controlled conditions. As an animal traverses the floor of a stabilimeter cage, it tips the floor and activates a microswitch and counter. It was not especially surprising that the pups’ activity increased after postnatal day (P) 5, when locomotor competence improved dramatically, but it was remarkable that activity surged 10-fold on P15, as can be seen in Figure 23.2. Particularly unexpected was the dramatic decline in activity after P15 (see upper line in Figure 23.2). The P15 peak and subsequent diminution in activity became the focus of subsequent studies of the ontogeny of forebrain inhibition (Campbell, Lytle, & Fibiger, 1969; Campbell & Mabry, 1973; Moorcroft, Lytle, & Campbell, 1971).
Campbell and associates used pharmacological manipulations as well as lesions of forebrain structures to test the hypothesis that the decrease in activity after P15 reflected the development of descending inhibition on brainstem structures that produced the arousal reflected in locomotor activity. Sure enough, when cholinergic forebrain activity was disrupted, the age-related decrease in activity was disrupted, and activity was potentiated. The overall concept was that a variety of stimuli and agents can nonselectively increase arousal in the young pup, and therefore its general activity level. Then, beginning around P15, coincident with the maturation of neural structures such as descending inhibitory projections to the brain stem, behavioral modulation begins (e.g., Fibiger, Lytle, & Campbell, 1970).
(p. 483) This picture of an early phase of unrestrained activity followed by reduced levels of activity appeared to be highly robust and reliable until Randall and Campbell (1976) examined the same developmental phenomenon in social settings, viz., in the presence of the mother or littermates or both. Fifteen-day-old pups presented an entirely different picture. Behavioral activity of rat pups tested in either of the two social setting did not show the distinctive P15 peak seen reliably in studies of isolated pups. Instead, when the development of behavioral arousal was measured in a social context, pup activity increased gradually and steadily from day 5 to day 30, as shown by the filled points (broken line) in Figure 23.2. This shocking finding, from the same laboratory that had faithfully promoted a purely central nervous system (CNS)based explanation for the development of behavioral arousal, was followed by a study examining “the role of environmental stimuli” in the ontogeny of behavioral arousal (Campbell & Raskin, 1978). Once again, it was found that a variety of contextual stimuli, including the presence of bedding material or nest odors, contributed to the modulation of activity in the isolated rat pup.
It was concluded from this long and rigorous line of research that the ontogeny of behavioral arousal is neither a purely brain maturation phenomenon nor a simple product of arousal-inducing external cues. Instead, the frequency, intensity, and duration of activities displayed by rat pups depend on a confluence of identifiable factors in the pup’s brain, body, history, as well as its momentary state. This panoply of variables is best combined into a view of situated behavior, in which behavioral expression emerges from continuous interactions of organismin-environment.
Development of Ingestive Behavior
In this section, rather than review, we select a set of findings that characterize development of ingestion when it is studied in the individual rat pup in and out of its social context. We will parse suckling into components, showing that the expression of these complex, adaptive behaviors can be understood as a system containing relatively simple and localized responses. Nonsuckling ingestion will also be examined. Once again, we see behaviors that are based on context-dependent, locally organized reactions.
The development of ingestion actually comprises two, separable ontogenies. One is the development and dissolution of suckling. The second is the onset and development of independent ingestion (i.e., feeding and drinking). The two processes are remarkably independent, in terms of the sensory controls, physiological cues, neural systems, and responses to pharmacological manipulations (see Hall & Williams, 1983; Chapter 22).
Before an infant can suckle, it must locate and then attach to a nipple. Olfactory cues are key. Infant rats rendered anosmic do not suckle (cf., Alberts, 1976). Intact rat pups do not attach to the nipples of a dam if her ventrum has been thoroughly washed (Teicher & Blass, 1976). Applying to the clean maternal ventrum, the distillate of the original wash reinstates suckling, indicating that a key olfactory cue has been removed and replaced. Without the correct olfactory input, even a deprived infant will simply rest near a lactating dam and not display any form of searching or dam- directed behavior.
Amniotic fluid is or contains the necessary olfactory cues for the newborn rat’s suckling sequence (Teicher & Blass, 1976; Blass & Teicher, 1980). It is instructive to note that during parturition, amniotic fluid is abundant and spread on the dam’s body and around the natal environment. Odors that have become familiar in utero may be recognized in the new and dramatically different ex utero world and thus serve a crucial role in bridging the two environments.
The formative mechanism that prepares a newborn rat to respond to suckling involves learning. Amniotic olfactants—whether natural or inserted by an experimenter’s injection—are learned by the fetus. If a lemon scent (citral) is added to each amniotic sac on embryonic day (E)17 and the pups are delivered at term, on E22, citral proves to be the necessary cue for nipple attachment. For the citralconditioned pups, even natural amniotic fluid stimulus is not sufficient to activate the suckling sequence (Pedersen & Blass, 1983).
It is easy to conclude (albeit prematurely and incorrectly) that because olfactory input is necessary for nipple attachment and suckling, and because suckling from a washed dam can be reinstated by painting her nipples with an odor known to the pup, that nipple odor elicits apprehension of a teat. However, the critical and oft unappreciated finding is that a conditioned perinatal odor (i.e., one that was experienced in utero and during parturition) merely in the atmosphere is sufficient to facilitate nipple attachment to a washed dam (Pedersen & Blass, 1981; Pedersen, Williams, & (p. 484) Blass, 1982). Under such conditions, there was no gradient to follow and no odor source emanating from the dam’s body. Activation was the key.
Behavioral activation, initially general and undirected, facilitates nipple attachment because odor cues potentiate movements that bring pups to the dam’s body. The pups’ head movements and probing brings the infants’ perioral area in contact with a nipple, at which point perioral reflexes produce oral grasping of the teat. Thus, general activation of behavior, embodied in a pup with age-related sensory and motor characteristics, situated in a niche composed of a mother’s ventrum, leads directly to topographically distinct behavior and a functional, adaptive outcome.
The Automatic Infant
In situ ingestion is reflexive and nonregulated. Friedman (1975) found that the limiting factor of suckling intake for P10 rats is the availability of mother’s milk. In the “normal” context of suckling from the dam, intake stops when the mother’s milk supply is depleted. When P10 rats were given access to a series of milk-laden dams (10 days postpartum), they continued to suckle and to gain weight.
Hall and Rosenblatt (1977) installed a cannula through which they could deliver milk into a pup’s mouth, under its tongue. These cannulated pups moved freely attached to the nipples of an anesthetized dam. The experimenters then delivered milk through the cannula in long pulses (0.10 mL/15 s). Milk was delivered until the pups ceased ingesting it. The most stunning result was that the youngest pups showed no satiety. Even after consuming such large quantities of milk that breathing was difficult, the engorged 5-day-olds struggled back to a nipple.
P10 rats with full stomachs showed an average increase in latency to attach; 15- and 20-day-olds would abandon the nipple after reaching satiety. Hall and Rosenblatt (1977) describe the gradual diminution of the stretch response. Even “nipple shifting” (i.e., releasing a suckled nipple and switching to another one) only appears after P10, which seems to be another sign that internal cues are beginning to exert behavioral control along with the external, olfactory and perioral stimuli (Hall & Rosenblatt, 1977).
Precocious Feeding EX SITU
For this discussion, feeding consists of nutrient ingestion from a source other than the mother’s mammary glands. There are several dazzling demonstrations that infant rats are capable of independent, regulated ingestion. This discovery emphasizes how the infant rat can express independent, regulated ingestion, even though such behavior is never expressed under normal conditions during early postnatal life.
Hall (1979b) provided some of the founding demonstrations of precocious feeding in rats. Here he used the same cannula system described earlier, except the pup was in a warm incubator with no mother present and had been deprived for 0.5, 7, or 22 h. Testing involved a series of slow infusions. Pups consumed the diet, and intake increased with deprivation, up to about 78% of the infusate or 2% of their body weight.
The drawings in Figure 23.3 depict some of the movements evoked by the oral infusions. Hall noted that such activities included mouthing, (p. 485) tongue protrusions, and probing “the surface in front of them.” These postures are rarely seen when pups are with the dam and littermates. That setting, though highly variable, nevertheless contains a basic set of olfactory, tactile, contour, and thermal cues that engage the pups’ movements and shape an entirely different topography of response. With a dam present, for example, the pups in Figure 23.3 would be observed to stretch and tread in response to the oral infusion and, if detached from the nipple, its aroused behavior would be directed at the dam. The pup would paddle and squirm into contact, and then push and probe into the dam’s ventrum until a nipple was encountered. It would be grasped and sucking would resume.
A common feature in every successful demonstration of precocious ingestion is that infant is tested alone, away from the mother. In other words, regulatory competence is masked in the presence of the mother and litter. This is a stunning reminder of the potency of context. Pups can regulate their ingestion, if an appropriate niche (albeit an unusual one) is created.
Behavior and Development on the Level of the Group
By recognizing suckling and feeding as separable processes, we also see two examples of how the presence of social stimuli—the mother in some cases and littermates in others—creates a new order in each individual. This “new order” is evidence for the emergence of a new, functional level of organization, as evidenced by downward regulation, from higher- to-lower levels. In this case, we see litter → individual, and mother → off spring, forms of downward regulation. We will survey a few examples of such processes on the level of the group.
Downward Regulation of Suckling Intake Exerted by Litter Cohort
The amount of milk obtained by pups begins to decrease between P15 and P20, although the amount of milk produced by the mother does not. Indeed, when pups begin to ingest solid food, there is milk available from the dam’s mammary ducts (Thiels, Cramer, & Alberts, 1988). The dissolution of suckling comes from a combination of the pups’ interactions with one another and the dynamics of the mother–litter relations (Thiels & Alberts, 1991).
Thiels et al. (1988) revealed the power of social setting when two 20-day-old rats were integrated into a litter of six 15-day-olds. Under this arrangement, the 20-day-olds consumed as much milk or more than their new littermates and significantly more than pups in homogenous, 20-day-old litters. Thiels et al. (1988) considered several aspects of the dam’s behavior that could affect milk availability, but these did not account for the observed differences.
Also significant is that 20-day-old rats embedded in a litter of 15-day-olds exhibited nonsuckling behavior that also resembled that of the younger pups. For instance, 20-day-olds placed among 15-day-olds were more docile and oriented toward the dam more than did 20-day-olds with same-age littermates.
Observations of the behavior of 15- and 20-day- old litters placed nearby an anesthetized dam provided additional insight into the way the groups acted and interacted as a function of age. The proportion of pups attaching to the dams’ nipples was higher and the latency to the first attachment was shorter in 15-day-olds compared to 20-day- olds. It appears that the presence of 15-day-old littermates acts as a form of social facilitation for suckling in the 20-day-olds. In contrast, milk consumption by 15-day-olds was relatively unaffected by the presence of 20-day-old of companion pups (Thiels et al., 1988).
Perhaps the most dramatic, if not most bizarre, demonstration of potent social influences was the demonstration that social stimulation from younger pups can override the diminishing tendency of older pups to attach to the nipple. Individual postweaning-age pups, housed with a series of preweaning-age mothers and their 16- to 21-dayold litters, continue to suckle until they are as old as 70 days of age, well beyond the time of normal weaning (Pfister, Cramer, & Blass, 1986). It is noteworthy that older pups never attach to a nipple when none or fewer than half of the younger pups are attached.
Weaning at the Level of the Group
All mammalian off spring wean, i.e., they cease suckling and commence independent feeding. We have seen that the physical and physiological bases of suckling and feeding develop within each individual. Synchronized and coordinated with these individual ontogenies are transitions in maternal physiology and behavior (Alberts & Gubernick, 1983; Rosenblatt & Lehrman, 1963 ). In addition, there are also ontogenies on the level of the group. Thus, the diminution of suckling and the transition to feeding occurs within complex and dynamic social niches.
(p. 486) A rat pup’s initial egressions from the nest are pivotal events in the weaning process because food is encountered remotely—in other areas of a burrow or in the outside world. The age at which pups leave the nest and begin to sample solid food varies, depending on many contextual factors, but weaning is often underway around P20. What factors account for the pups leaving the nest?
Video observations of rats in a variety of laboratory-based habitats have provided much information on the weaning process (Alberts & Leimbach, 1980; Cramer, Thiels, & Alberts, 1990; Galef & Clark, 1971; Gerrish & Alberts, 1996, 1997; Thiels, Alberts, & Cramer, 1990). In the course of observing weanling rats’ initial egressions from a nest box into an adjacent open field, we noticed that their forays from the nest often occurred during periods of heightened general activity. More careful, empirical examination revealed a pattern of postsuckling activation.
Shortly after a mother rat completes a nursing bout and leaves the nest, there is often a gradual increase in activity among the litter that can build into an explosion of activity, termed postsuckling behavioral arousal (Gerrish & Alberts, 1997). The aroused litter often displays tumultuous activity, dramatically obvious in a time-lapse videograph. The litter seethes and roils. With increasing turbulence within the nest, individual pups may be seen climbing over and revolving around the group. During this accelerated activity, individuals may zoom out of their orbital path, leaving the group and heading away. We studied this phenomenon in 20-day-old litters in a nest that was attached, via a couple of short tunnels, to an outside field where food and water were located. This postsuckling behavioral arousal was typically the prelude to leaving the nest and entering the field (Gerrish & Alberts, 1997).
The phenomenon can be produced by presenting P20 rats with an anesthetized dam, and then separating each one from a nipple by a tug to its tail. (A nondislodging tail tug was used as a control.) Loss of orally grasped nipple predicts the onset of activation. Figure 23.4 summarizes some experiments that specified the stimulus conditions for the phenomenon of postsuckling behavioral arousal. The behavioral effect was profound and led directly to food sampling (Gerrish & Alberts, 1997; see also Alberts & Leimbach, 1980).
Social Contexts Determine Diet Selection
Ingesting safe, nutritious food and avoiding toxic substances (especially the poisons intended for unwanted commensals) is a primary challenge in the life of wild R. norvegicus (Barnett, 2001). Galef and Clark (1971) showed that rat pups completely avoid eating a food that is avoided by the adult members of their colony and prefer foods preferred by adults in the colony. In one of their elegant experiments, rat pups emerged from nest boxes and approached adults that were feeding from a bowl containing a distinct, nutritious Diet A and avoiding another bowl containing a distinct, nutritious, Diet B, which the adults had previously experienced as toxic. (When pups were present, Diet B lacked toxic content.) Pups fed on Diet A hundreds of times and never sampled Diet B, even when a full bowl of this safe, palatable food was located as little as 3 in. from Diet A!
The adults never prevented the pups from sampling Diet B nor did they “mark” the food or its bowl with a chemical deterrent. In fact, Galef and Clark (1971) showed that pups did not learn an aversion to Diet B; they simply learned eating what the adults ate (Diet A).
Such powerful and reliable specificity proved to be formed by remarkably simple and nonspecific mechanisms. Basically, when weanlings approach other rats, this simply brings them into the vicinity of food. Pups choose to eat the same food that others are eating. Preexposure to cues that they can recognize from mother’s milk may bias their choice (Galef & Henderson, 1972) as can an association of food odors with carbon disulfide, a constituent of rat breath (Galef, Mason, Petri, & Bean, 1988). Together, these findings suggest that social factors are the predominant guiding force for diet selection and much of the rats’ feeding behavior. These factors are influential only because the pups are so exquisitely sensitive and responsive to them.
Ontogeny of Huddling
For R. norvegicus, huddling, or contact behavior, is a life-long behavior. Huddling among littermates begins immediately after birth; huddling with the dam also begins at about this time, as the mother broods and nurses her litter of propagules. Pups remain huddled together in the nest during maternal excursions until around P15 when the coherence of the group lessens, but does not fade. Under normal, group-living, colonial conditions, adult R. (p. 487) norvegicus continue to engage in a variety of behaviors that involve cutaneous contact. These habits have earned R. norvegicus the label of “contact species” (Barnett, 1963).
A naturally occurring clump of pups is not a just a heap of bodies. There is usually activity somewhere within the group. Time-lapse views are especially revealing, for they enable a human observer to see the rhythms and flow of individual behavior within the group. In one observational study, groups of littermates were placed in a bowl-shaped nest for observation. Focal pups within each litter were marked for individual identification and the groups were observed with time-lapse videography at a record:playback ratio of 12:1. Under these conditions, the huddle is a nearly continuously seething mass. Figure 23.5 depicts the amount of time a focal pup was observed on the surface of the group. Data in panels A and C, and B and D depict individual pups in the same clump so, by mentally sliding the paired panels together, it is possible to envision the alternations and in-phase movements of these individuals.
One answer, derived from studies of 10-dayold rats, is that pup flow can be understood as a group manifestation of individual regulations of exposed body surface shown in Figure 23.5. Litters were placed in nests that were either cool (24°C) (p. 488) or warm (36°C). There were two focal pups, one experimental and the other the control. The experimental animal was injected with an anesthetic before the test and its littermate control was given a saline injection. Both were placed on the surface of their huddle for time-lapse video recording. The time that each focal pup was exposed on the huddle surface was quantified.
It is unusual to run a behavioral study in which the subject is anesthetized, but the point here was the fate of the anesthetized, nonparticipating pup. This pup provided a telling diagnostic. In the room- temperature nest, the unanesthetized pups actively burrowed down into the depths of the litter, thereby pushing the anesthetized pup to the surface of the huddle where it remained almost continuously. Thus, under these conditions, the anesthetized pup was a “floater.” In contrast, its active littermate control periodically appeared and disappeared at the surface, as is characteristic of normal pup flow.
That this is a thermally determined, regulatory activity was further demonstrated by the fate of the anesthetized pup in the warm nest (right-hand panel, Figure 23.6). Now, as unanesthetized pups emerged at the surface of the nest to lose heat, the anesthetized pup becomes a “sinker,” disappearing into the depths of the group. Thus, we see here a reversal of the direction of pup flow caused by a change in thermal conditions.
This group regulatory behavior can be explained in terms of individual adjustments. Developmental analyses of this phenomenon could prove informative (e.g., Sokoloff & Blumberg, 2001), because there are dramatic developments of individual behavioral thermoregulation in rats (Farrell & Alberts, 2007; Hoffman, Flory, & Alberts, 1999a, 1999b; Pfister, 1990).
We have been discussing the behavior of individual rat pups, sometimes seeing them as an individual within a complex family setting, often in contexts that were specially designed to isolate the rat pup from many of the myriad stimuli that can simultaneously affect pup behavior. Now we turn to the group, especially the group of littermates that comprise the huddle, the pup’s first social setting.
By identifying the aggregation of littermates as a “huddle,” we immediately invoke a term associated with defense against cold. The connotation is appropriate. Infant rats, often considered ectothermic, thus lacking self-produced body heat, were (p. 489) shown to be warmer and to have lower metabolic rates as a function of group size. Energy conservation by huddling was impressive in scale (Alberts, 1978).
How is this group effect on metabolic heat production achieved? We now know that even newborn rats can activate endogenous deposits of brown adipose tissue, a thermogenic organ, and thereby generate body heat (cf., Blumberg, 2001). But, due in part to the pups’ lack of insulative fur and meager subcutaneous fat, they are susceptible to rapid loss of body heat. The pup’s large body surface in relation to its small body volume or mass is the prime factor for such heat loss (Alberts, 1978). By huddling, a pup can greatly reduce its exposed surface area, thus reducing the surface:mass ratio, which governs heat transfer.
It is important to ask whether the metabolic benefits of huddling have functional significance. Small huddles of four littermates were placed in a temperature-regulated compartment. By changing the pups’ ambient temperature from 20°C to 40°C and back to 20°C over the course of 2 h and measuring the exposed surface area of the group every 10 min, we obtain a picture of the group surface dynamics. Figure 23.7 summarizes the results. At each age, the group surface area was regulated behaviorally to increase and decrease with ambient temperature. These results show that rat pups behave in a manner that produces a form of group regulatory behavior. In this instance, the phenomenon is evidenced as the systematic control of the group’s physical properties that control heat loss and heat gain.
Group Regulatory Behavior
We have characterized this form of group regulatory behavior as an example of niche construction (Alberts, 2007; Alberts & Schank, 2006) whereby huddles of pups create a microenvironment in which the average surface:mass ratio of each individual is reduced. This effectively reduces the magnitude of the environmental challenge, and makes (p. 490) the pup’s limited thermogenic abilities more effective (see also Blumberg, 2001; Sokoloff, Blumberg, & Adams, 2000). Therefore, in this niche of bodies, each pup benefits and can express its capabilities with greater efficacy. In this way, the group becomes each individual’s niche, and in this niche, their physiological size and capabilities combine to make them more competent than they are as single individuals.
Reconstructing Individual and Group Development: New Tools and Techniques
We can begin to synthesize the diverse body of knowledge about rat pups by constructing models of group and individual behavior. Constructing even relatively simple models of interacting infant rats leads to detailed and unexpected issues. Here we provide a glimpse of what may be possible with these new tools and techniques.
In a typical study, which simplifies many contextual variables, litters of eight rat pups are placed in the stalls of a starting corral (Figure 23.8) centered in an arena (8 × 12 in.) that is a well-controlled, warm environment (e.g., 34°C floor; 34°C air). When the corral is removed, the pups can move freely and aggregate. Video images of pups’ movements and aggregations provide a surfeit of information, from which we sample to make comparisons with virtual or physical models. We have found that image sampling every 5 s is sufficient to collect meaningful data concerning individual and group behavior (May et al., 2006; Schank, 2008; Schank & Alberts, 1997, 2000; Schank, May, Tran, & Joshi, 2004).
Data on locomotion are obtained by recording pup positions as x–y coordinates in images; these data allow for analysis of activity, pup contact, wall contact, and aggregation (Schank & Koehnle, 2007). For example, how pups aggregate into patterns of contact with other pups can be measured as aggregon patterns (i.e., the distribution of individuals in contact groups formed on the surface of an arena) or as subgroups (i.e., the number of contact groups that form, which can range from 1 to the number of individuals in the group; Schank, 2008; Schank & Alberts, 1997, 2000). There are 22 possible aggregons combinations of 8 pups, as shown in the aggregon index of Figure 23.9 and eight possible subgroups, each consisting of subsets of aggregon combinations (Schank, 2008; Schank & Alberts, 2000). The dynamics of aggregon and subgroup frequencies provide detailed and diagnostic descriptions of group behavior (Schank, 2008).
To understand how blind and deaf pups could form the aggregon patterns, we built an agent- based model (Schank, 2008; Schank & Alberts, 1997, 2000). (p. 491) Pup movement was simplified: With each time-step, a pup could move to one of eight adjacent locations in the virtual arena. The rules of behavior were simple: (1) if one or more pups are located in front or on the side of a pup’s head, then there is an increased probability that the pup moves towards that location. Similarly, we assumed that (2) pups moved toward the walls of the arena directly in front of them. The two other rules governed activity: (3) A pup remains active as a function of the number of active and inactive pups it contacts and (4) becomes inactive again as a function of the number of active and inactive pups contacted. These rules are conceptually simple. Where a pup moves and whether it is active is situation- dependent.
To find a model that behaves like real pups involves determining specific values for the probabilities of moving toward pups or toward wall and for the activity rules described in the previous paragraph. This can be accomplished by systematically simulating the parameter values, or in more complex cases, using optimization techniques such as simulated annealing or genetic algorithms (Schank, 2008; Schank & Alberts, 2000). We thereby fit movement and activity parameters to the aggregative patterns of groups of pups. As illustrated in Figure 23.10, the best fit model generates aggregon frequency distributions that mimic those of the real 7-day-old rat huddles using only the movement and activity rules previously described. We explored the landscape of results derived from changing parameters of the model and have shown that this is a robust phenomenon (Schank & Alberts, 1997).
It is important to emphasize that the model contains no rules of group behavior, but only rules for individual movement and activity. Nonetheless, in the absence of group instructions, pups aggregate and age-specific patterns of aggregation emerge (Schank, 2008; Schank & Alberts, 2000).
Perhaps the most interesting result of this model was the identification of social context-dependent (p. 492) activity during early development. Seven-day-olds exhibited the same probabilities of staying active or becoming active independent of the number or the activity state of pups contacted. Ten-day-olds, however, showed a clear and systematic activity relationship to the number and the activity state of pups contacted. We believe that the coupling of activity states characteristic of the 10-day-olds reflects the onset of sociality and reveals a fundamental development from the complete autonomy exhibited by the 7-day-olds in this context.
Converging on our interpretation of a developmental transition into sociality are the results of a separate study in which we infused either oxytocin (OT) or an OT antagonist (Vasotocin) intracister- nally in 7- and 10-day-old pups that were tested under group conditions (Alberts, 2007; Odya, Sokoloff, & Alberts, 2002). We used OT because it is widely associated with modulating social behavior in a variety of species and behavioral contexts (e.g., Ferguson, Young, & Insel, 2002; Goodson, 2005; Insel & Fernald, 2004; Lim & Young, 2006), though there have been few developmental studies (cf., Carter, 2003). In this preliminary study, we measured group “size” (i.e., total area subtended by the aggregons) and found that in 10-day-olds OT treatment promoted proximity and group coherence whereas the antagonist decreased group coherence below levels seen in the control groups. Figure 23.11 illustrates these results. Thus, we believe that this research with rats (a) defines and validates emergent group behavior, (b) reveals an ontogeny of group behavior, and (c) provides quantitative and qualitative methods for isolating and synthesizing the rules and mechanisms for the patterns of aggregation we have observed.
Hypothesis Testing by Construction
Hypothesis testing in behavioral neuroscience has traditionally been accomplished by falsification. There is art to this kind of science: Beautiful experiments are devised to falsify, and thereby reject, well-articulated null hypotheses. Most often, we seek precise and specific hypotheses, which can be thoroughly tested and rejected. As noted early in this chapter, however, integrative problems as complex as understanding development of behavior on (p. 493) multiple levels needs more than a series of tests of specific hypotheses because the problem of integrating multiple levels is too complex. At this point, we need to synthesize the available information. We need to construct a system analog or model that we can test and analyze. We think that modeling is a valuable and important approach that can be used rigorously, and can contribute in special ways to a fuller understanding of complex processes. As mentioned above, one goal of constructing models is to integrate levels by creating models that behave like the natural systems. When we successfully construct models that behave like a natural living system, using biologically valid elements, we demonstrate a sufficiency or completeness of knowledge regarding a complex reality in relation to its simpler components. Such models become a source of testable hypotheses.
Models come in a variety of forms and compositions, ranging from pure mathematical entities to computation or even physical models (Koehnle & Schank, 2003). Robotic models pose a special challenge in this regard but offer special rewards. One reason for building physical robotic models of animals—in addition to building only virtual models—is to emulate physical interactions of bodies with their environment. Ideally, such interactions can be modeled virtually (Bish, Joshi, Schank, &Wexler, 2007) but in practice, the virtual models remain only approximations for physical properties, which may therefore miss important physical interactions. Moreover, physical robots also allow for the gradual evolution of increasingly complex robotic models of animals by adding components and then testing the new phenotype.
We see special value in using robots to learn about multilevel development using individual and group behavior as the prime phenomenological levels. Rat development is beautifully suited to this strategy. The infant’s limited sensory and motor repertoires direct the modeler to begin with simple robots. Then, to account for the increased sensorimotor complexity that emerges in subsequent developmental states (e.g., the development of visual and auditory function), the modeler can reengineer the robots to include such sensory systems. Each stage requires testing and reengineering based on results and new data.
To build robots that behave like infant rats required us to model pup morphology and physical contact. These preliminary requirements proved instructive and provided lessons not readily learned from the virtual models alone.
We learned that the shape of a robot’s body (Figure 23.12) is important because it determines how the robot interacts with corners, walls, and other robots in an arena (see Figure 23.12, top), and thus whether a robot can behave like a pup. A 10-day-old rat pup measures about 7.6 × 2.5 cm, which is too small to model with available mechanical and electrical hardware. Thus, we built robots scaled to a rat pup’s length-to-width ratio (3:1). The overall dimensions for the robots were 32.4 x 10.8 cm. We built a total of eight robots (May et al., 2006). Rat pups, up to 10 days old, primarily use their back legs for forward locomotion (Altman & Sudarshan, 1975). Robots with rear-driven wheels and differential drive on the chassis create forces on the body that resemble those of young rat pups (May et al., 2006; Schank et al., 2004).
We believe that the flexibility of pups’ bodies and the rigidity of robots’ bodies resulted in different patterns of motion. For example, rat pups often (p. 494) leave corners in an arena by flexing their body in an arc rather than backing up. Because flexing to turn is not possible for our robots, we simulated this maneuver by allowing a robot to back up slightly and then turn. We achieved different circular arc motions by shifting the center of motion along the lateral axis of the robot, which affected the overall turning trajectory. This center could be adjusted to accommodate data for pups at different ages. Because our strategy is to start with the “simple” and work to the complex on a developmental timescale, our robots modeled only an infant rat’s dominant sensory modality during the first 12 days after birth, namely its tactile sense. To model tactile sensitivity, we mounted microswitch bump sensors around the skirt to mimic tactile sensory fields of rats (Figure 23.12, bottom). Because the majority of infant rat interactions occur rostrally, we mounted a number of adjustable sensors near the front of the robot. A total of 14 sensors were installed, with half located on the head and half along the body. To increase the range of each sensor, we mounted brass metal strips on them, which allowed sensing at all points along the robot’s body.
The central processor of a robot implements sensorimotor rules that simulate the brain of an animal. There are many levels of abstraction for modeling a brain, ranging from a simple reactive architecture that deterministically relates peripheral sensory input and motor output to neural networks models with increasingly realistic representation of neural anatomy and physiology. Thus, a simple and natural starting point for modeling contact and aggregation in infant rats with robotic modes is to implement a reactive-control architecture that is responsive to contact alone. This approach—which is a simplified implementation of a brain—resembles the contextual simplification and control that we used when we began the testing of (real) rat pups in a warm, flat, evenly illuminated arena (Schank & Alberts, 1997; 2000).
A typical robot run, using this thigmotaxic-reactive architecture, produced patterns of movement that were highly stereotyped. Robots repeatedly followed walls, circling the arena over and over again, resembling the “stereotypies” often observed in animals in zoos. It was clear that a purely deter-ministic, thigmotaxic-reactive architecture could not explain the behavior of pups in an arena. There were, however, runs in which the robots’ behavior did not appear so stereotypical. A robot’s behavior can be influenced indeterminately when, for example, dust on the substrate causes wheels to slip slightly, or if there are small differences in the servo motors that run each wheel. This suggested to us that randomness may be an important factor in generating behavior.
We then decided to test robots that moved completely randomly with no influence on their behavior from their tactile sensors. To produce a random control robot, a robot at each time step t (the interval between time steps was set to 2 s) randomly chooses one of 10 locomotor behaviors produced by modulating the two servo motors running the wheels: (1) forward; (2) 45° left; (3) 90° left; (4) 135° left; (5) 45° right; (6) 90° right; (7) 135° right; (8) backup left; (9) backup right; and (10) do nothing. To our great surprise, results from both individual and multiple robot experiments revealed that a simple random architecture with no contingency relationships between sensory inputs and motor outputs coupled with body- environment constraints accounted for a large proportion of the behavioral patterns observed in both individuals and groups. Individual random robots followed walls, got stuck in corners, escaped them, and visited a variable number of corners in each run. Individuals in groups produced the same movements and even aggregated in similar ways as illustrated in Figure 23.13 (see Table 23.1 for comparisons among all metrics).
It is striking that a randomly moving robot with a body shape similar to a pup placed in an arena is so constrained by arena geometry and body shape that its behavior alone and in a group is very similar to rat pups. But real infant rats have sensory systems that respond to tactile input. Surely their behavior cannot be purely random? Probably the best interpretation of these results is that body morphology and its interactions with the environment can generate emergent patterns of aggregation even though the robots are using no rules at all to move about! We are currently simulating various ways in which these robots can have flexible bodies because having rigid bodies greatly constrains movement and is unrealistic as a model of flexible bodied rat pups. Our initial results are that flexibility does make a difference and that flexible bodied robots, which move randomly, do not match pup behavior nearly as well though body morphology still contributes substantially to the patterns of aggregation we observe (May, 2007). We are now at the stage where we can begin to model the contribution of simple sensorimotor rules with an understanding of how morphology and the physical structure of the environment constrain behavior. (p. 495)
Final Overview: Individuals and Groups during Natural Selection
Recognition of groups as coherent, functional units has a long history in biology. By 1872 and with each edition of the Origin of the Species, Darwin increasingly recognized the importance of natural selection operating at the level of the group. Yet, in the 1960s, adaptation at the level of the group was strongly challenged—so strongly in fact, that the idea was essentially rejected outright and considered an anathema if raised for discussion (Darlington, 1980). The demise of group selection can be attributed, at least in part, to overemphasis on the idea that the interests of the individual are sacrificed or subordinated to those of the group. After group selection was narrowed and isolated this way, there was room to popularize mutually exclusive viewpoints. Proponents of the view that natural selection is limited to the level of the individual scored big with inclusive fitness theory (Hamilton, 1964, 1975) and selfish gene theory (Dawkins, 1976; Williams, 1966). (p. 496)
Table 23.1 Comparison of Robots to 7-Day and 10-Day Rats on All Metrics
Pup Age vs. Robots Pup Age
Individual wall contact
7 days = Robots = 10 days
Individual corner contact
7 days < Robots = 10 days
Individual center contact
7 days > Robots > 10 days
Individual distance moved
7 days < Robots = 10 days
7 days > Robots = 10 days
Group wall contact
7 days < Robots = 10 days
Group corner contact
7 days < Robots > 10 days
Group center contact
7 days > Robots < 10 days
7 days < Robots < 10 days
Robots either statistically match 10-day-old pups or are intermediate between 7− and 10-day-old pups, except for Group corner contact and Group center contact.
There is now, however, a growing revival of interest in group selection as an important force in evolution (e.g., Wade, 1976; Wilson, 1975, 1983), even among some previous detractors (Wilson & Wilson, 2007). We see the precepts and much of the research discussed in this chapter as amenable with the more modern ideas about selection and adaptation at the level of the group.
In contemporary treatments of the topic, efforts are made to recognize that selection occurs simul-taneously on the levels of individual and group: “… anything that is good for the group must be good for one or more of the individuals in it” (Darlington, 1980, p. 140). As we have found, the key to understanding group function lies in detailed attention to individuals in it. The basic argument is that when group advantages are achieved, the advantages are experienced by individuals in the group, not just by the group itself. And, if more than one such group exists, then the relative advantages (improved fitness of individuals) of one over the other can contribute to the selection of those individuals and hence, an adaptive advantage to the groups comprising those individuals. When group selection operates positively at both individual and group levels, evolution can be rapid (e.g., Wilson & Wilson, 2007).
Group adaptation and group selection is multileveled and context-dependent, much like the individual and group processes outlined in the present chapter. We see great potential in a coherent, multilevel and multi-timescale view of developmental and evolutionary thinking. Behavioral analyses can forge new paths into this terrain.
Alberts, J. R. (1976). Olfactory contributions to behavioral development in rodents. In R. L. Doty (Ed.), Mammalian olfaction, reproductive processes, and behavior (pp. 67–93). New York: Academic Press.Find this resource:
Alberts, J. R. (1978). Huddling by rat pups: Group behavioral mechanisms of temperature regulation and energy conservation. Journal of Comparative and Physiological Psychology, 92, 231–245.Find this resource:
Alberts, J. R. (1984). Sensory-perceptual development in the Norway rat: A view toward comparative studies. In: R. Kail & N. Spear (Eds.), Comparative perspectives on memory development (pp. 65–101). New York: PlenumFind this resource:
Alberts, J. R. (1985). Ontogeny of social recognition: An essay on metaphor and mechanism in behavioral development. In: E. S. Gollin (Ed.), Comparative development of adaptive skills: Evolutionary implications (pp. 65–101). Hillsdale, NJ: Erlbaum,.Find this resource:
Alberts, J. R. (2002). Simply complex: Essentialism trumps reductionism. Current Neurology and Neurosciences Reports, 2, 379–381.Find this resource:
Alberts, J. R. (2005). Infancy. In I. Q. Whishaw & B. Kolb (Eds.), The behavior of the laboratory rat: A handbook with tests (pp. 266–277). New York: Oxford University Press.Find this resource:
Alberts, J. R. (2007). Huddling by rat pups: Ontogeny of individual and group behavior. Developmental Psychobiology, 49, 22–32.Find this resource:
Alberts, J. R., & Cramer, C. P. (1988). Ecology and experience; Sources of means and meaning of developmental change. In: E. M. Blass (Ed.), Handbook of behavioral neurobiology, Vol. 9: Behavioral ecology and developmental psychobiology (pp. 1–39). New York: Plenum Press.Find this resource:
Alberts, J. R., & Gubernick, D. J. (1983). Reciprocity and resource exchange: A symbiotic model of parent-off spring relations. In: L. A. Rosenblum & H. Moltz (Eds.), Symbiosis in parent-off spring relations (pp. 7–44). New York: Plenum Press.Find this resource:
Alberts, J. R., & Leimbach, M. P. (1980). The first foray: Maternal influences on nest egression in the weanling rat. Developmental Psychobiology, 13, 417–430.Find this resource:
Alberts, J. R., & May, B. (1984). Non-nutritive, thermotactile induction of olfactory preferences for huddling in rat pups. Developmental Psychobiology, 17, 161–181.Find this resource:
Alberts, J. R., Motz, B. A., & Schank, J. C. (2004). Positive geotaxis in rats: A natural behavior and a historical correction. Journal of Comparative Psychology, 118, 123–132.Find this resource:
Alberts, J. R., & Schank, J. C. (2006). Constructing ontogenetic niches. AlifeX Conference. Bloomington, IN.Find this resource:
Altman, J., & Sudarshan, K. (1975). Postnatal development of locomotion in the laboratory rat. Animal Behaviour, 23, 896–920.Find this resource:
Barnett, S. A. (1963). The rat: A study in behaviour. London: The Camelot Press Ltd.Find this resource:
Barnett, S. A. (2001). The story of rats: Their impact on us, our impact on them, Crows Nest, Australia: Allen and Unwen.Find this resource:
Bish, R., Joshi, S., Schank, J., & Wexler, J. (2007). Mathematical modeling and computer simulation of a robotic rat pup. Mathematical and Computer Modeling, 54, 981–1000.Find this resource:
Blass, E. M., & Teicher, M. H. (1980). Suckling. Science, 210, 15–22.Find this resource:
Blumberg, M. S. (2001). The developmental context of thermal homeostasis. In E. M. Blass (Ed.), The Handbook of behavioral neurobiology. Volume 13: Developmental (p. 497) psychobiology, developmental neurobiology and behavioral ecology: Mechanisms and early principles (pp. 199–228). New York: Plenum Press.Find this resource:
Bolles, R. C., & Woods, P. J. (1964). The ontogeny of behavior in the albino rat. Animal Behaviour, 12, 427–441.Find this resource:
Campbell, B. A., & Mabry, P. D. (1973). The role of catecholamines in behavioral arousal during ontogenesis. Psychopharmacolgia, 31, 253–264.Find this resource:
Campbell, B. A., Lytle, L. D., & Fibiger, H. C. (1969). Ontogeny of adrenergic arousal and cholinergic inhibitory mechanisms in the rat. Science, 166, 637–638.Find this resource:
Campbell, B. A., & Randall, P. K. (1976). Ontogeny of behavioral arousal in rats: Effect of maternal and sibling presence. Journal of Comparative and Physiological Psychology, 90, 453–459.Find this resource:
Campbell, B. A., & Raskin, L. A. (1978). Ontogeny of behavioral arousal: The role of environmental stimuli. Journal of Comparative and Physiological Psychology, 92, 176–184,Find this resource:
Carter, C. S. (2003). Developmental consequences of oxytocin. Physiology & Behavior, 79, 383–397.Find this resource:
Cramer, C. P., Thiels, E., & Alberts, J. R. (1990). Weaning in Norway Rats: I. Maternal Behavior patterns. Developmental Psychobiology, 23, 479–494.Find this resource:
Crozier, W. J., & Pincus, G. (1926). The geotropic conduct of young rats. Journal of General Physiology, 10, 257–269.Find this resource:
Crozier, W. J., & Pincus, G. (1936). Analysis of the geotropic orientation of young rats. X. Journal of General Physiology, 20, 111–124.Find this resource:
Darlington, P. J., Jr. (1980). Evolution for naturalists: The simple principles and complex realities. New York: John Wiley & Sons.Find this resource:
Darwin, C. R. (1872). The origin of species by means of natural selection, or the preservation of favoured races in the struggle for life (6th ed.). London: John Murray.Find this resource:
Dawkins, R. (1976). The selfish gene. Oxford: Oxford University Press.Find this resource:
Delcourt, J., Becco, C., Ylieff, M. Y., Caps, H., Vandewalle N., & Poncin P. (2006). Comparing the EthoVision 2.3 system and a new computerized multitracking prototype system to measure the swimming behavior in fry fish. Behavior Research Methods, 38, 704–710.Find this resource:
Farrell, W. J., & Alberts J. R. (2007). Rat behavioral thermoregulation integrates with nonshivering thermogenesis during postnatal development. Behavioral Neuroscience, 121, 1333–1341.Find this resource:
Ferguson, J. N., Young, L. J., & Insel, T. R. (2002). The neuroendocrine basis of social recognition. Frontiers in neuroendocrinology, 23, 200–224.Find this resource:
Fibiger, H. C., Lytle, L. D., & Campbell, B. A. (1970). Cholinergic modulation of adrenergic arousal in the developing rat, Journal of Comparative and Physiological Psychology, 72, 384–389.Find this resource:
Fraenkel, G. S., & Gunn, D. L. (1940). The orientation of animals. London: Oxford University Press.Find this resource:
Friedman, M. I. (1975). Some determinants of milk ingestion in suckling rats. Journal of Comparative and Physiological Psychology, 89, 636–647.Find this resource:
Galef, B. G., Jr., & Clark, M. M. (1971). Social factors in the poison avoidance and feeding behavior of wild and domesticated rat pups. Journal of Comparative and Physiological Psychology, 75, 341–357.Find this resource:
Galef, B. G., Jr., & Henderson, P. W. (1972). Mother’s milk: A determinant of the feeding preferences of weaning rat pups. Journal of Comparative and Physiological Psychology, 78, 213–219.Find this resource:
Galef, B. G., Jr., Mason, J. R., Preti, G., & Bean, N. J. (1988). Carbon disulfide: A semiochemical mediating socially-induced diet choice in rats. Physiology and Behavior, 42, 119–124.Find this resource:
Gerrish, C., & Alberts, J. R. (1996). Environmental temperature modulates onset of independent feeding: Warmer is sooner. Developmental Psychobiology, 29, 483–495.Find this resource:
Gerrish, C., & Alberts, J. R. (1997). Post-suckling behavioral arousal contributes to weaning in Norway rats (Rattus norvegicus). Journal of Comparative Psychology, 111, 37–49.Find this resource:
Goodson, J. L. (2005). The vertebrate social behavior network: Evolutionary themes and variations. Hormones and Behavior, 48, 11–22.Find this resource:
Goodwin, B. (1994). How the leopard changed its spots. New York: Scribner.Find this resource:
Gottlieb, G. (1976). The roles of experience in the development of behavior and the nervous system. In G. Gottlieb (Ed.), Studies in the development of behavior and the nervous system (Vol. 3, pp. 25–54). New York: Academic Press.Find this resource:
Hall, W. G. (1979a). Feeding and behavioral activation in infant rats. Science, 205, 206–209.Find this resource:
Hall, W. G. (1979b). The ontogeny of feeding in rats: I. Ingestive and and behavioral responses to oral infusions. Journal of Comparative and Physiological Psychology, 93, 977–1000.Find this resource:
Hall, W. G., & Rosenblatt, J. S. (1977). Suckling behaior and intake control in the developing rat. Journal of Comparative and Physiological Psychology, 91, 1232–1237.Find this resource:
Hall, W. G., & Williams, C. L. (1983). Suckling isn’t feeding, or is it? Advances in the Study of Behavior, 13, 219–254.Find this resource:
Hamilton, W. D. (1964). The genetical evolution of social behavior: I and II. Journal of Theoretical Biology, 7, 1–52.Find this resource:
Hamilton, W. D. (1975). Innate social aptitudes in man: An approach from evolutionary genetics. In R. Fox (Ed.), Biosocial anthropology (pp. 133–153), London: Malaby Press.Find this resource:
Hoffman, C. M., Flory, S. G., & Alberts, J. R. (1999a). Neonatal thermotaxis improves reversal of a thermally-reinforced instrumental response. Developmental Psychobiology, 34, 87–99.Find this resource:
Hoffman, C. M, Flory, S. G., & Alberts, J. R. (1999b). Ontogenetic adaptation and learning: A developmental constraint in learning for a thermal reinforcer. Developmental Psychobiology, 34, 73–86.Find this resource:
Insel, T. R., & Fernald, R. D. (2004). How the brain processes social information: Searching for the social brain. Annual Review of Neuroscience, 27, 697–722.Find this resource:
Koehnle, T. J., & Schank, J. C. (2003). Power tools needed for the dynamical toolbox. Adaptive Behavior, 11, 291–295.Find this resource:
Krieder, J. C., & Blumberg, M. S. (1999). Geotaxis in 2-weekold Norway rats (Rattus norvegicus). Developmental Psychobiology, 35, 35–42.Find this resource:
Lim, M. M., & Young, L. J. (2006). Neuropeptidergic regulation of affiliative behavior and social bonding in animals. Hormones and Behavior, 50, 506–517.Find this resource:
Lloyd, S. (2000). Ultimate physical limits to computation. Nature, 406, 1047–1054.Find this resource:
May, C. J. (2007). Modeling the behavior of infant Norway rats (Rattus norvegicus). Unpublished doctoral dissertation, University of California, Davis.Find this resource:
May, C. J., Schank, J. C., Joshi, S., Tran, J., Taylor, R. J., & Scott, I. (2006). Rat pups and random robots generate (p. 498) similar self-organized and intentional behavior. Complexity, 12 (1), 53–66.Find this resource:
Moorcroft, W. H., Lytle, L. D., & Campbell, B. A. (1971). Ontogeny of starvation-induced behavioral arousal in the rat. Journal of Comparative and Physiological Psychology, 75, 59–67.Find this resource:
Motz, B. A., & Alberts, J. R. (2005). The validity and utility of geotaxis in young rodents. Neurotoxicology and Teratology, 27, 543–544.Find this resource:
Noldus, L. P., Spink, A. J., & Tegelenbosch, R. A. (2001). EthoVision: A versatile video tracking system for automation of behavioral experiments. Behavioral Research Methods, Instrumentation and Computation, 33, 398–414.Find this resource:
Odya, E. C., Sokoloff, G., & Alberts, J. R. (2002). The effects of oxytocin on the aggregation of infant rats. Orlando, FL: Society for Neuroscience.Find this resource:
Pedersen, P., & Blass, E. M. (1981). Olfactory control over suckling in albino rats. In R. N. Aslin, J. R. Alberts, & M. R. Petersen (Eds.), The development of perception: Psychobiological perspectives (pp. 359–381). New York: Academic Press.Find this resource:
Pedersen, P., & Blass, E. M. (1983). Prenatal and postnatal determinants of the 1st suckling episode in albino rats. Developmental Psychobiology, 15, 349–355.Find this resource:
Pedersen, P. E., Williams, C. L., & Blass, E. M. (1982). Activation and odor conditioning of suckling behavior in 3-day-ole albino rats. Journal of Experimental Psychology: Animal Behavior Processes, 8, 329–341.Find this resource:
Pfister, J. F. (1990). The development of responses to hot and cold challenges in the Norway rat: Preferences, regulation, and learning Unpublished doctoral dissertation, Indiana University, Bloomington, IN.Find this resource:
Pfister, J. P., Cramer, C. P., & Blass, E. M. (1986). Suckling in rats extended by continuous living with dams and their preweanling litters. Animal Behaviour, 34, 415–420.Find this resource:
Randall, P. K., & Campbell, B. A. (1976). Ontogeny of behavioral arousal in rats: Effect of maternal and sibling presence. Journal of Comparative and Physiological Psychology, 90, 453–459.Find this resource:
Rosenblatt, J. S., & Lehrman, D. S. (1963). Maternal behavior in the laboratory rat. In H. L. Rheingold (Ed.), Maternal behavior in mammals (pp. 8–57). New York: Wiley.Find this resource:
Saunders, J. W. (1982). Developmental biology. Patterns, problems, principles. New York: Macmillan.Find this resource:
Schank, J. C. (2001). Beyond reductionism: Refocusing on the individual with individual-based modeling. Complexity. 6, 33–40.Find this resource:
Schank, J. C. (2008). The development of locomotor kinematics in neonatal rats: An agent-based modeling analysis in group and individual contexts. Journal of Theoretical Biology, 254, 826–842.Find this resource:
Schank, J. C., & Alberts, J. R. (1997). Self-organized huddles of rat pups modeled by simple rules of individual behavior. Journal of Theoretical Biology, 189, 11–25.Find this resource:
Schank, J. C., & Alberts, J. R. (2000). The developmental emergence of coupled activity as cooperative aggregation in rat pups. Proceedings of the Royal Society of London, 267, 2307–2345.Find this resource:
Schank J. C., & Koehnle T. J. (2007). Modeling complex biobehavioral systems. In: Laubichler, M. D., & Muller G. B, (Eds.), Modeling biology: Structures, behaviors, evolution (pp. 219–240). Cambridge, MA: MIT Press.Find this resource:
Schank, J. C., May, C. J., Tran, J. T., & Joshi, S. S. (2004). A biorobotic investigation of Norway rat pups (Rattus norvegicus) in an arena. Adaptive Behavior, 12, 161–173.Find this resource:
Small, W. S. (1899). Notes on the psychic development of the young white rat. American Journal of Psychology. 11, 80–100.Find this resource:
Sokoloff, G., & Blumberg, M. S. (2001). Competition and cooperation among huddling infant rats. Developmental Psychobiology, 39, 65–75.Find this resource:
Sokoloff, G., Blumberg, M. S., & Adams, M. M. (2000). A comparative analysis of huddling in infant Norway rats and Syrian hamsters: Does endothermy modulate behavior? Behavioral Neuroscience, 114, 585–583.Find this resource:
Teicher, M. H., & Blass, E. M. (1976). Suckling in newborn rats: Eliminated by nipple lavage, reinstated by pup saliva. Science, 193, 422–425.Find this resource:
Thelen, E. (1988). Dynamical approaches to the development of behavior. In: J. A. S. Kelso, A. Mandell, & M. F. Schlesinger (Eds.), Dynamical patterns in complex systems (pp. 348–369). Singapore: World Scientific.Find this resource:
Thelen, E., Kelso, J. A. S., & Fogel, A. (1987). Self-organizing systems and infant motor development. Developmental Review, 7, 39–65.Find this resource:
Thiels, E., & Alberts, J. R. (1991). Weaning in the Norway rat: Relation between suckling and milk, and suckling and independent ingestions. Developmental Psychobiology, 24, 19–38.Find this resource:
Thiels, E., Alberts, J. R., & Cramer, C. P. (1990). Weaning in Norway rats: II. Pup behavior patterns. Developmental Psychobiology, 23, 495–510.Find this resource:
Thiels, E., Cramer, C. P., & Alberts, J. R. (1988). Behavioral interactions rather than milk availability determine decline in milk intake of weanling rats. Physiology and Behavior, 42, 507–515.Find this resource:
Tsibidis G. D. & Tavernarakis, N. (2007). Nemo: A computational tool for analyzing nematode locomotion. BMC Neuroscience, 8(1), 86–92.Find this resource:
Wade, M. J. (1976). Group selection among laboratory populations of tribolium. Proceedings of the National Academy of Sciences, 73, 4604–4607.Find this resource:
West, M. J., & King, A. P. (1988). Female visual displays alter the development of male song in the cowbird, Nature, 334, 244–246.Find this resource:
Williams, G. C. (1966). Adaptation and natural selection. Princeton, NJ: Princeton University Press.Find this resource:
Wilson, D. S. (1975). A theory of group selection. Proceedings of the National Academy of Sciences. 72, 143–146.Find this resource:
Wilson, D. S. (1983). The group selection controversy: History and current status. Annual Review of Ecology and Systematics, 14, 159–187.Find this resource:
Wilson, D. S., & Wilson, E. O. (2007). Rethinking the theoretical foundations of sociobiology, Quarterly Review of Biology, 82, 327–348.Find this resource: