Sources and References

Primary texts

• Gregory Berns, How Dogs Love Us: A Neuroscientist and His Adopted Dog Decode the Canine Brain (2013)
• Alexandra Horowitz, Inside of a Dog: What Dogs See, Smell, and Know (2009); Our Dogs, Ourselves: The Story of a Singular Bond (2019)
• Marc Bekoff, The Emotional Lives of Animals (2007); Minding Animals: Awareness, Emotions, and Heart (2002)
• Heather Douglas, Science, Policy, and the Value-Free Ideal (2009)
• Helen Longino, Science as Social Knowledge: Values and Objectivity in Scientific Inquiry (1990)
• Patricia McConnell, The Other End of the Leash: Why We Do What We Do Around Dogs (2002)
• Jean Donaldson, The Culture Clash (1996, rev. 2005)
• Temple Grandin and Catherine Johnson, Animals in Translation: Using the Mysteries of Autism to Decode Animal Behavior (2005)
• Konrad Lorenz, King Solomon’s Ring (1949); Man Meets Dog (1950)
• Peter Singer, Animal Liberation (1975)
• Tom Regan, The Case for Animal Rights (1983)
• The Belmont Report: Ethical Principles and Guidelines for the Protection of Human Subjects of Research (1979), National Commission for the Protection of Human Subjects
• Russell and Burch, The Principles of Humane Experimental Technique (1959) — source of the Three Rs framework
• Nagasawa et al., “Oxytocin-gaze positive loops and the coevolution of human-dog bonds,” Science 348:6232 (2015)
• Thomas Nagel, “What Is It Like to Be a Bat?” Philosophical Review 83:4 (1974)

Online resources

• Stanford Encyclopedia of Philosophy (SEP): “Animal Consciousness,” “Scientific Objectivity,” “Values in Science,” “Research Ethics”
• Canadian Council on Animal Care (CCAC) guidelines
• American Veterinary Medical Association (AVMA) guidelines on animal welfare

Chapter 1: Science and Ethics — What Can We Learn from Dogs?

1.1 Introduction: Why Dogs?

Dogs occupy a singular position in human civilization. They are, in the words of Alexandra Horowitz, “wolves who noticed us” — animals whose evolutionary trajectory became intertwined with our own roughly 15,000 to 40,000 years ago. Yet for most of the history of Western science, the inner lives of dogs were treated as philosophically and scientifically inaccessible: useful objects of behavioral study, perhaps, but not subjects of genuine psychological investigation.

This course takes a different starting point. By examining the science of dogs — what researchers have discovered about canine cognition, emotion, sensory experience, and social relationships — we simultaneously illuminate a set of pressing philosophical questions about the nature of science itself. What counts as evidence? Whose values shape what gets studied, and how? When is it permissible to use animals in research? How should scientific knowledge interact with public policy and everyday practice?

Dogs are not merely a convenient case study. They are, in a real sense, the ideal case study for a course at the intersection of science, ethics, and society. Our emotional investment in dogs makes the ethical stakes visceral; their cognitive sophistication makes questions of consciousness urgent; and the long history of human-dog relations makes questions of power, exploitation, and obligation unavoidable.

1.2 Why Dogs Make an Ideal Science-Society Lens

1.2.1 Uniqueness at the Intersection

Very few topics sit simultaneously at the heart of cutting-edge empirical research, public emotional investment, ethical controversy, and policy debate. Canine science does. Unlike, say, high-energy physics or molecular genetics — fields where public engagement is thin and emotional stakes are low — dog science is something millions of people care about deeply, often before they have ever heard a philosopher’s name.

This creates a distinctive pedagogical opportunity. When we ask whether Gregory Berns’s fMRI data establish that dogs feel love, we are not asking an abstract philosophical puzzle: we are asking something that matters to dog owners, to animal welfare advocates, to research ethics boards, and to policy-makers considering the legal status of animals. The stakes are genuine, and that genuine-ness forces philosophical rigor in a way that purely hypothetical examples often fail to.

There is also a methodological (方法论) advantage. Because dog science spans behavioral ecology, comparative neuroscience, applied animal behavior, and cognitive ethology, it offers a cross-section of scientific practice broad enough to illuminate the full range of philosophical themes in this course: from the structure of experimental design to the sociology of scientific communities to the politics of science communication.

1.2.2 The Emotional Investment Problem and Opportunity

The emotional salience of dogs is not merely a pedagogical convenience — it is itself philosophically interesting. The fact that most people bring strong prior beliefs about dog cognition and dog welfare to any encounter with canine science raises questions about confirmation bias (确认偏见), motivated reasoning, and the conditions under which emotional investment helps or hinders good epistemic practice.

Feminist philosophers of science have argued that the posture of emotional detachment traditionally valorized in science is not actually available to researchers — and that pretending otherwise is epistemically costly. The alternative is not uncritical enthusiasm but what Donna Haraway calls partial perspective (部分视角): acknowledging one’s position and the values that shape it, and subjecting them to critical scrutiny. Dog science, precisely because its practitioners are often emotionally invested, is an excellent laboratory for developing this capacity.

1.3 A Brief Natural History of Human-Dog Relationships

1.3.1 The Domestication Timeline

The domestic dog (Canis lupus familiaris) is the earliest domesticated animal. Although the precise date and location of domestication remain contested, the current best evidence points to a divergence from wolf ancestors somewhere between 15,000 and 40,000 years before present, almost certainly in Eurasia. Some researchers argue for a single domestication event; others propose multiple independent domestication episodes in different regions.

The fossil record is difficult to interpret because the morphological differences between early dogs and wolves were subtle, and behavioral differences leave no direct fossil trace. Molecular phylogenetics has become the dominant tool for reconstructing domestication history, but ancient DNA studies have repeatedly revised the picture. As of the mid-2020s, a broad consensus locates the initial domestication event in eastern Eurasia, possibly in what is now China or Siberia, sometime during the Upper Paleolithic — a period of sustained contact between modern humans and wolves around human encampments.

What is particularly striking about the timing is that domestication appears to have preceded agriculture. Dogs were not first domesticated as herding or guarding animals; they entered human society as hunting companions and camp followers before humans had settled communities to guard. This temporal fact suggests that the initial driver of domestication was social rather than purely instrumental — an attraction between two social species, each of which may have benefited from the association in ways that were initially incidental.

1.3.2 Key Studies in Canine Cognition Arising from Domestication

One of the most important research programs in canine science directly concerns the cognitive consequences of domestication. A series of studies beginning in the late 1990s, primarily led by Ádám Miklósi and József Topál at the Family Dog Project in Budapest, established that dogs possess social cognitive abilities that wolves, chimpanzees, and other closely related species do not: specifically, the capacity to use human communicative gestures — particularly pointing — to locate hidden food.

This finding is philosophically significant. The capacity to follow a human point is not simply a matter of perceptual acuity; it requires understanding the communicative intent behind the gesture. Young children develop this capacity early in the first year of life. Dogs, uniquely among non-human animals, appear to develop it without specific training. The Budapest studies demonstrated that this capacity emerges in the first weeks of life, before puppies have had extensive experience with human pointing, suggesting that it is either an innate feature of the domesticated phenotype or that dogs acquire it through a very short developmental window. Wolves raised identically to dogs by human handlers do not reliably develop the same capacity.

Brian Hare and colleagues at Duke University extended this research program, arguing that dogs have been specifically selected for cooperative communication (合作沟通) with humans — not just tolerating human proximity but actively seeking human social input. Hare’s research demonstrated that dogs outperform even chimpanzees, our closest relatives, on tasks requiring the use of human social cues. This is remarkable because chimpanzees exceed dogs on virtually every other measure of raw cognitive ability. The implication is that domestication has produced a specialized social cognitive niche in dogs rather than general intelligence enhancement.

These findings have direct implications for the ethics of dog research. If dogs are genuinely adapted for human social relationships — if attending to humans is, in a meaningful sense, what dogs are for — then research designs that isolate dogs from human social contact, or that treat dogs as merely convenient laboratory animals, may be both ethically problematic and scientifically suboptimal.

1.3.3 Dogs as Social and Cultural Artifacts

Dogs have served as hunters, herders, guards, war animals, status symbols, therapeutic companions, and scientific instruments. This diversity of roles reflects the fact that dogs are not simply biological entities but social artifacts (社会人工制品) — beings whose meaning, treatment, and welfare are constituted in part by the cultural frameworks within which they exist.

Konrad Lorenz, the Nobel-winning ethologist, wrote extensively about human-dog relationships in King Solomon’s Ring and Man Meets Dog. While his science has been superseded in important ways, his central insight — that dogs are genuinely social beings whose behavior can only be understood in the context of their relationships — anticipates much of contemporary canine science.

1.4 Science as a Form of Inquiry

1.4.1 The Basic Structure of Scientific Reasoning

Science is a systematic method for generating reliable knowledge about the natural world. At its core, it involves the formulation of hypotheses (假设), the collection of evidence through observation and experiment, and the revision of beliefs in light of that evidence. But this bare description conceals enormous complexity. What counts as a good hypothesis? How much evidence is enough? Who gets to decide?

Philosophy of science (科学哲学) is the branch of philosophy concerned with the foundations, methods, and implications of scientific inquiry. In this course, we approach philosophy of science not as a purely theoretical exercise but as a practical tool for understanding real scientific practice — including the science of dogs.

1.4.2 Scientific Communities and Social Epistemology

One of the most important developments in twentieth-century philosophy of science was the recognition that science is not an activity of isolated individuals but a social enterprise (社会事业). Helen Longino’s work on social epistemology is central here. In Science as Social Knowledge, Longino argues that scientific objectivity is not a property of individual scientists or their methods alone but emerges from the structure of scientific communities — specifically, from communities that maintain norms of open criticism, uptake of criticism, shared standards, and equality of intellectual authority.

This social dimension of science has profound implications for ethics. If scientific knowledge is produced by communities, then questions about who belongs to those communities, whose perspectives are represented, and what values are embedded in research programs become epistemically relevant — not merely politically convenient.

1.4.3 Using Dogs to Illuminate Science

Each chapter in this course uses a specific domain of canine science to illuminate a philosophical theme. The neurobiology of canine love raises questions about research ethics and the permissibility of animal subjects. Canine olfaction research raises questions about scientific objectivity and the role of values. Dog training science raises questions about the relationship between scientific knowledge and social practice. Together, these case studies build a picture of science as a fundamentally human — and therefore ethically freighted — activity.

Chapter 2: Research Ethics — Procedural Frameworks and Animal Subjects

2.1 The Development of Research Ethics

2.1.1 Historical Background

Modern research ethics did not emerge in a vacuum. It developed in response to documented abuses: the Nazi medical experiments at Nuremberg, the Tuskegee syphilis study in the United States, and numerous other cases in which the interests of research subjects were subordinated to the interests of researchers and institutions. These abuses generated both legal responses — the Nuremberg Code (1947), the Declaration of Helsinki (1964) — and philosophical efforts to articulate principles that should govern research involving human subjects.

2.1.2 The Belmont Report

The Belmont Report (贝尔蒙特报告), published in 1979 by the National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research, articulates three foundational principles for research ethics:

Respect for Persons (尊重人): Research subjects are autonomous agents entitled to make their own decisions. This principle generates the requirement of informed consent (知情同意) — the requirement that subjects be given sufficient information about a study to make a voluntary, competent decision about participation.

Beneficence (善行): Researchers and institutions have an obligation to maximize benefits and minimize harms. This is not merely an aspiration but a structured obligation that shapes study design, risk assessment, and the conduct of research.

Justice (正义): The benefits and burdens of research should be fairly distributed. Historically, vulnerable populations — prisoners, the poor, racial minorities — have borne disproportionate burdens as research subjects while enjoying few of the benefits.

2.2 Animal Research Ethics

2.2.1 The Moral Status of Animals

When we move from human to animal subjects, the Belmont framework does not straightforwardly apply. Animals cannot give informed consent in the relevant sense. But this does not mean that animal interests are morally irrelevant. The philosophical debate over moral status (道德地位) concerns which entities have interests that generate moral obligations on the part of others.

Several positions are relevant here:

Sentientism holds that sentience — the capacity for subjective experience, particularly pleasure and pain — is the morally relevant criterion. Peter Singer’s utilitarian framework, developed in Animal Liberation (1975), holds that we are required to give equal consideration to the equal interests of all sentient beings, regardless of species. Singer does not argue that animals and humans must be treated identically — only that their comparable interests must be weighted equally. The capacity for suffering, not species membership, is what matters morally. Singer calls the alternative — granting lesser moral weight to animal suffering simply because it occurs in a non-human animal — speciesism (物种歧视), a prejudice analogous to racism or sexism. Applied to dog research, Singer’s framework requires that we ask: does the expected benefit of this research, weighted impartially across all affected parties (including the research dogs), outweigh the expected harm? Laboratory housing, restraint, and experimental procedures that cause fear or pain must be entered into the utilitarian calculus on the same basis as the benefits to humans.

Rights-based approaches, associated with Tom Regan’s The Case for Animal Rights (1983), hold that animals who are “subjects-of-a-life” — who have beliefs, desires, memory, perception, an emotional life, preference and welfare interests, and an individual welfare that matters to them independently of its utility to others — possess inherent value (内在价值) and rights that cannot be overridden merely for the aggregate benefit of others. Regan’s position is more demanding than Singer’s. Where Singer might permit animal research that produces a sufficiently large benefit, Regan argues that the rights of individuals cannot be traded off against aggregate welfare gains. On Regan’s view, using dogs as research subjects — even with refinement and reduction — treats them as mere receptacles for value rather than as ends in themselves. The morally correct position, on Regan’s account, is the total abolition of harmful animal research, not its careful management.

The contrast between Singer and Regan maps onto a fundamental divide in ethical theory. Singer’s utilitarianism is a consequentialist (结果主义) framework: the morality of an action is determined entirely by its consequences. Regan’s rights theory is deontological (义务论的): certain actions are wrong regardless of consequences, because they violate the inherent rights of persons (human or otherwise). In practice, both frameworks generate strong presumptions against most current animal research, though they differ in whether those presumptions are in principle overridable.

Contractarian approaches ground moral status in membership in a moral community of beings capable of reciprocal obligation. On strict contractarian views, animals lack moral status as rights-bearers, though they may be protected instrumentally — because their suffering distresses humans who care about them, or because including them in our moral circle reflects virtues we wish to cultivate.

2.2.2 Relational Ethics and Animal Research

A third framework, increasingly influential in bioethics and animal studies, is relational ethics (关系伦理学). Rather than grounding moral status in intrinsic properties of the individual (sentience, being a subject-of-a-life), relational approaches argue that moral obligations arise from relationships — and that the nature and history of a relationship shape the specific obligations it generates.

Clare Palmer’s work in Animal Ethics in Context (2010) develops this framework in application to animals. Palmer argues that we have stronger obligations to animals with whom we have relationships of dependence, co-habitation, or historical entanglement than to wild animals we have never encountered. Dogs are the paradigm case: millennia of domestication have created a relationship of deep mutual entanglement. We bred dogs into dependence on human social structures; we shaped their cognitive and emotional architecture to make them maximally attentive to us. This historical relationship generates obligations that go beyond what a purely individualist moral framework would recognize.

In the context of research ethics, relational approaches suggest that dogs are not simply any sentient beings who happen to be enrolled in studies: they are animals whose entire evolved psychology is organized around human social relationships. Using that psychology as a research tool — enrolling pet dogs in laboratory studies, exploiting their human-oriented attention for experimental purposes — engages a specific set of relational obligations. At minimum, the relational framework demands that research designs acknowledge the social and emotional needs of dogs as beings adapted for human relationship, not merely biological entities with a capacity for suffering.

2.2.3 The Three Rs Framework

The most influential procedural framework for animal research ethics is the Three Rs (三个R原则), developed by William Russell and Rex Burch in The Principles of Humane Experimental Technique (1959):

The Three Rs framework has become embedded in law and institutional policy across many jurisdictions. In Canada, the Canadian Council on Animal Care (CCAC) requires that all federally funded animal research adhere to Three Rs principles, assessed through institutional animal care committees (IACCs).

Each R applies in specific and instructive ways to dog research. For Replacement: the shift from laboratory-housed dogs to studies of owned pet dogs living in their normal home environments (as in Berns’s fMRI project) does not eliminate dogs as subjects but replaces the invasive laboratory housing model — a significant welfare gain. Computer modeling of olfactory detection can replace some behavioral trials. For Reduction: meta-analyses of published dog cognition studies have revealed that many early studies were severely underpowered, using small samples that produced unreliable estimates. Larger, better-designed multi-lab studies can generate more reliable knowledge with fewer total animal-subject exposures. For Refinement: the design of enriched laboratory environments for research dogs — socialization programs, access to human companionship, outdoor time, play — reduces chronic stress and produces more ecologically valid behavioral data. The welfare gain and the scientific gain align: a chronically stressed dog does not provide a good window into normal canine cognition.

2.2.4 Critique and Expansion of the Three Rs

The Three Rs framework is procedural rather than prohibitive: it does not rule out any particular category of research but requires that researchers minimize harm within the research they undertake. Critics from animal rights traditions argue that this is insufficient — that some research is simply impermissible regardless of how carefully it is conducted.

A more moderate critique focuses on the implementation of the Three Rs. Refinement in particular is highly context-dependent: what counts as minimizing distress for a dog in a laboratory setting involves complex empirical questions about canine cognition, social behavior, and stress physiology — questions that are themselves the subject of active scientific investigation.

Chapter 3: The Neurobiology of Canine Love

3.1 Gregory Berns and the Dog fMRI Project

3.1.1 Scientific Background

Gregory Berns is a neuroscientist at Emory University whose research program on canine cognition began with a striking methodological innovation: training awake, unrestrained dogs to lie still in a functional magnetic resonance imaging (fMRI) scanner. Published initially in 2012 and expanded in his popular science book How Dogs Love Us (2013), Berns’s work represents a watershed moment in the science of animal minds.

Functional MRI (功能性磁共振成像) measures brain activity by detecting changes in blood oxygenation — the BOLD (blood-oxygen-level-dependent) signal. Regions of the brain that are more active consume more oxygen, producing a detectable signal. The technique has been used extensively in human cognitive neuroscience but had rarely been applied to awake animals prior to Berns’s work, because the method requires the subject to remain motionless.

3.1.2 Methodological Challenges: Training Dogs to Hold Still

The central methodological challenge of the Berns project was not data analysis but subject preparation. An fMRI scanner is a loud, enclosed, and unfamiliar environment; head movement of even a few millimeters renders BOLD data unusable due to motion artifacts. In human studies, subjects can be instructed verbally; with dogs, a different approach was required.

Berns and his team, working with professional dog trainers, developed a multi-month behavioral preparation protocol. Dogs were first desensitized to the sounds of the scanner using audio recordings played at progressively increasing volume. They were then trained to rest their chins in a custom-molded head support designed to minimize movement. Ear protection was fitted and conditioned as a neutral stimulus. Finally, dogs were habituated to the full scanner environment — including the enclosed tube — through gradual exposure paired with high-value food rewards.

The training protocol itself is scientifically significant for several reasons. First, it demonstrates that the dogs were voluntary participants: they could and did decline to enter the scanner on days when they were unwilling. The voluntary nature of participation distinguishes this research design sharply from studies that use restraint. Second, the extensive behavioral preparation — typically twelve to eighteen months of training before data collection began — means that the researchers developed an unusually detailed understanding of each individual dog’s temperament, preferences, and behavioral repertoire. Third, the protocol was based entirely on positive reinforcement, consistent with the scientific consensus on effective and welfare-positive training.

The motion correction challenge remained significant even with well-trained dogs. Berns’s team developed a custom motion-correction algorithm adapted for canine neuroanatomy, and applied strict inclusion criteria: trials in which the dog’s head moved more than 3 mm in any direction were excluded from analysis. The resulting dataset, while relatively small by human neuroscience standards, was methodologically rigorous in ways that earlier attempts at animal neuroimaging had not been.

3.1.3 The Question of Canine Love

The phrase “canine love” might seem to beg important philosophical questions. But Berns is careful to distinguish the scientific claim — that dogs possess neural structures and processes that are homologous to those associated with love and affection in humans — from the stronger philosophical claim that dogs experience love in a phenomenologically rich sense.

The scientific claim rests on comparative neuroscience (比较神经科学): the principle that neural structures conserved across evolutionary lineages likely serve similar functional roles. The caudate nucleus, the limbic system more broadly, and the oxytocin-based bonding system are ancient mammalian features. Dogs and humans share these systems, and Berns’s data suggest they activate in dogs under circumstances analogous to those that activate them in humans.

What does it mean to say that a dog “loves” its owner? The word “love” in English covers a range of psychological phenomena: attachment, preference, positive affect, motivated social approach, distress at separation. Berns’s neural data are most directly relevant to the affective and motivational components — the hedonic positive valence of the owner’s presence, and the motivational salience of that presence. They do not directly address the more cognitively complex components of love in humans, such as narrative self-representation, idealization, or long-term planning around a relationship. The cautious interpretation — and the one best supported by the available evidence — is that dogs experience something functionally analogous to the hedonic and motivational aspects of attachment, implemented in neural systems homologous to those supporting comparable states in humans. Whether this constitutes “love” in the full human sense depends on philosophical commitments about what love requires that the neuroscience alone cannot resolve.

3.1.4 Methodological and Ethical Implications

Berns’s research ethics framework is worth examining carefully. His dogs were trained, not coerced; they could and sometimes did leave the scanner voluntarily. This design reflects a sophisticated application of Three Rs thinking: by working with owned pets in a non-invasive paradigm, Berns eliminated the need for laboratory-housed animal subjects entirely. The study thus serves as an example of replacement in practice — not replacing dogs as subjects but replacing the invasive, laboratory-housing model with one that integrates naturally into the animals’ existing lives.

3.2 Oxytocin, Bonding, and the Science of Attachment

3.2.1 The Oxytocin System

Oxytocin (催产素) is a neuropeptide produced in the hypothalamus that plays a central role in mammalian social bonding. Originally studied in the context of mother-infant bonding and pair bonding in monogamous species, oxytocin research has expanded substantially in recent decades. The “oxytocin loop” — the reciprocal positive feedback between gazing, touching, and oxytocin release — has been documented in human mother-infant pairs.

3.2.2 The Nagasawa et al. 2015 Study

A landmark 2015 study by Miho Nagasawa and colleagues at Azabu University in Japan, published in Science (Nagasawa et al., “Oxytocin-gaze positive loops and the coevolution of human-dog bonds,” Science 348:6232), demonstrated that the oxytocin bonding loop operates between dogs and their human owners in a way that closely parallels the human mother-infant bond.

The study design was elegant. Dog-owner pairs were brought into a room and allowed to interact freely for thirty minutes. Urine samples collected before and after the interaction were analyzed for oxytocin metabolites. The key variable was the amount of time each dog spent gazing at the owner’s face during the interaction session. The finding: dogs who gazed at their owners for longer periods showed greater oxytocin increases, and — crucially — their owners showed greater oxytocin increases in parallel. The gaze-oxytocin relationship was bidirectional and reciprocal: more gaze predicted more oxytocin in both parties.

To test the causal direction, Nagasawa’s team used an intranasal oxytocin administration paradigm. Dogs who received intranasal oxytocin (rather than saline control) subsequently gazed at their owners for significantly longer periods, and their owners’ oxytocin levels rose correspondingly. Oxytocin caused more gazing, which caused more oxytocin — a positive feedback loop strikingly parallel to the documented loop in human mother-infant pairs.

The wolf comparison was the study’s most striking result. Wolves raised by humans from birth — hand-reared in conditions similar to domestic dogs — did not show the oxytocin-gaze positive loop. Wolves gazed at their human handlers less than dogs, and neither wolves nor their handlers showed significant oxytocin increases following the interaction. This result strongly suggests that the oxytocin-mediated gaze bond is not simply a consequence of individual socialization with humans but is a specifically domesticated adaptation — a feature of the dog genome shaped by selection during domestication.

The Nagasawa study is philosophically significant in several respects. First, it provides a plausible proximate mechanism for the observed strength of dog-human attachment — not merely behavioral co-occurrence but a shared neurochemical positive feedback system. Second, the wolf comparison provides evidence that this mechanism is an evolutionary adaptation to life with humans rather than a learned behavior. Third, the study is methodologically notable for measuring oxytocin in both species simultaneously, treating the dog-human dyad as the unit of analysis rather than either party alone — a design that reflects a relational rather than individualist framing of the research question.

3.2.3 Philosophical Implications: What Do These Findings Mean?

The oxytocin findings raise a cluster of philosophical questions. First, there is a question of reduction: does showing that love (in dogs and humans) is mediated by a particular neurochemical system explain love, or merely describe one of its mechanisms? Philosophers of mind debate whether neural correlates of mental states constitute explanations or merely redescriptions.

Second, there is a question of inference: the oxytocin data establish a correlational and partially causal story about a physiological process. But what licenses the inference that this process is accompanied by subjective experience — that there is something it is like to be a dog gazing at its owner?

This is, in essence, the hard problem of consciousness (意识的困难问题), articulated by David Chalmers: even a complete physical account of the mechanisms of experience would leave unexplained why any physical process is accompanied by subjective experience at all. The hard problem applies with particular force to animal minds, where we cannot rely on verbal report.

The cautious interpretation of the oxytocin data is accordingly modest: the data establish that dogs and humans are neurochemically coupled in their social interactions in a way that is homologous to attachment bonding in other mammalian dyads. Whether this coupling is accompanied by subjective experience of affection — whether there is something it is like to be the dog in this loop — is a further question that the neurochemical data alone cannot answer.

Chapter 4: Values in Science — Objectivity and Its Limits

4.1 The Value-Free Ideal

4.1.1 What Is the Value-Free Ideal?

The value-free ideal (价值无涉理想) holds that science, properly conducted, is — or should aspire to be — free of value judgments. On this view, the epistemic virtues of science (accuracy, consistency, explanatory power, predictive success) are entirely distinct from ethical, social, and political values, and the latter have no legitimate role in scientific inference. This ideal has deep roots in logical empiricism and in the public legitimacy of science as a source of authoritative, neutral knowledge.

4.1.2 Why the Value-Free Ideal Fails

Heather Douglas’s Science, Policy, and the Value-Free Ideal is the most sustained contemporary critique of this position. Douglas argues that the value-free ideal, while motivated by genuine concerns about the distortion of science by non-epistemic interests, is both descriptively false (science has never been value-free) and normatively problematic (a value-free science would be inadequate to guide policy).

The descriptive failure is illustrated throughout the history of science. Funding priorities shape which questions get investigated; theoretical frameworks shape which data are collected; classification systems embed contestable assumptions about what differences matter. In the history of animal behavior science, the dominance of behaviorism from the 1920s through the 1970s was not a neutral methodological choice but a value-laden commitment: it privileged parsimony in one direction (no mental state posits) while ignoring parsimony in another (the implausibility of radically divergent mental lives in closely related species). The choice to dismiss animal cognition as unscientific was itself shaped by values — specifically, a social-scientific ideal of rigor that derived its prestige from association with physics and chemistry.

The normative failure is equally important. If science is to guide policy — as Douglas argues it must — then scientific conclusions always have inductive risk: the risk that they will be wrong. When scientists characterize their findings as merely descriptive, value-free reports, they obscure the normative choices embedded in their methods and the asymmetric costs of different kinds of error. Douglas argues that scientists have a responsibility to acknowledge these choices and their value-ladenness, rather than sheltering behind a false claim of neutrality.

4.1.3 Types of Values in Science

Douglas draws a crucial distinction between two ways that values can operate in science:

Direct roles: values function as reasons for accepting or rejecting a scientific conclusion. This is the form of value-influence that the value-free ideal prohibits, and rightly so — if a scientist accepts or rejects a hypothesis because it supports a preferred policy outcome, this is a corruption of the epistemic process.

Indirect roles: values function in choices about research design, data classification, and the characterization of evidence — not as reasons for believing conclusions but as guides to the management of uncertainty and the weighing of risks.

4.2 Helen Longino and Social Objectivity

4.2.1 The Social Constitution of Scientific Knowledge

Helen Longino’s social epistemology (社会认识论) offers a different but complementary critique of the value-free ideal. In Science as Social Knowledge, Longino argues that individual scientists inevitably bring background assumptions and values to their work — assumptions that are not themselves warranted by the data but that shape what counts as evidence for what. Objectivity is not achieved by eliminating these assumptions (which is impossible) but by subjecting them to structured social criticism.

4.2.2 Background Assumptions in Dog Science

A background assumption is a claim that is taken for granted within a research community — not explicitly tested or argued for, but operative in the design of studies, the interpretation of data, and the framing of conclusions. Longino’s central contribution is to show that background assumptions are not merely incidental to science but partly constitutive of it: they determine what counts as evidence for what.

A specific example from dog cognition research illustrates the point. Throughout much of the twentieth century, animal behavior researchers operated with the background assumption that mentalistic explanations (心理主义解释) — explanations that invoke beliefs, desires, intentions, or emotions as causal factors — were scientifically inadmissible. This assumption was often presented as a neutral methodological commitment to parsimony. But it was not neutral: it reflected a philosophical position (reductionist behaviorism) and a set of social values (the prestige of physics-style science within biology).

When researchers began studying dog-human social cognition in the 1990s, they brought a different set of background assumptions: that mental state attribution is scientifically respectable under appropriate constraints, that evolutionary continuity licenses inference from human to animal mental states, and that convergent evidence from multiple methodologies is more reliable than any single method. These different assumptions did not merely change the conclusions of research — they changed what counted as a research question in the first place. The question “can dogs understand human pointing as a communicative act?” could not be asked, let alone answered, within the strict behaviorist framework.

4.2.3 Application: Values in Canine Consciousness Research

The case of canine consciousness research illustrates both Douglas’s and Longino’s points with particular clarity. For much of the twentieth century, the dominant framework in animal behavior research was strict behaviorism (行为主义): the view that scientific psychology should restrict itself to observable behavior and refrain from positing internal mental states as explanatory constructs. This methodological commitment was presented as a neutral, scientific constraint — the value-free choice.

But the commitment to behaviorism was not value-neutral. It embodied assumptions about what counts as a scientifically respectable explanation, who has the standing to make claims about animal minds, and what the costs of error are. The behaviorist framework systematically underestimated the cognitive and emotional complexity of animals — including dogs — not because the data demanded it but because the background assumptions did.

Marc Bekoff’s work on cognitive ethology (认知行为学) represents a deliberate effort to replace these background assumptions with a framework that takes animal minds seriously as scientific objects. His advocacy for a science that is both rigorous and open to evidence of animal consciousness reflects a position that values and epistemology are inseparable.

4.3 Feminist Philosophy of Science

4.3.1 Situated Knowledge and Standpoint Epistemology

Longino’s social epistemology connects to a broader tradition in feminist philosophy of science (女性主义科学哲学). Donna Haraway’s concept of situated knowledge (situated knowledge, 情境知识) holds that all knowledge is produced from a particular social location — a position that is shaped by gender, race, class, institutional affiliation, and other factors. The pretense of a “view from nowhere” — a purely objective, disembodied perspective — is itself an ideological construction that serves particular interests.

Sandra Harding’s standpoint epistemology (立场认识论) extends this point: groups that occupy marginalized social positions may, under certain conditions, have epistemic advantages — access to aspects of reality that are invisible from dominant positions. This has implications for who should be included in scientific communities and whose perspectives should inform research questions.

4.3.2 Feminist Philosophy of Science Applied to Animal Cognition Research

Feminist philosophers of science have applied these frameworks specifically to the study of animal behavior. Lynda Birke and others have argued that the tradition of viewing animals as passive objects of study — rather than as active subjects of their own lives — reflects a broader pattern of objectification that has operated in both the study of non-human animals and the study of marginalized human groups. The parallel is not merely rhetorical: the same posture of detachment that produced bad science about women’s psychology also produced bad science about animal minds.

A concrete example: early studies of dog-human attachment often treated the dog’s behavior purely as a dependent variable — a measure of human influence — rather than as active social behavior in its own right. Studies asked “how do humans shape dog behavior?” rather than “how do dogs and humans mutually influence each other?” The asymmetric framing was not forced by the data; it reflected prior assumptions about the relative agency and relevance of human and dog perspectives. Feminist-inflected relational framings — treating the dyad rather than either party as the unit of analysis — have generated different, and arguably richer, scientific questions.

Haraway’s own work on dog-human relationships in The Companion Species Manifesto (2003) argues that the relationship between dogs and humans is a site of genuine co-constitution (共同建构): humans and dogs have shaped each other over millennia, and understanding either requires understanding the relationship. This is not sentimentality but a substantive claim about the unit of biological and social analysis.

4.3.3 Application: Who Studies Dogs, and Why?

The sociology of canine science is instructive. Women have been disproportionately prominent in the field — from Patricia McConnell to Alexandra Horowitz to the many women who have contributed to applied animal behavior. Whether this reflects a genuine epistemic advantage arising from different socialization around animal relationships, or merely reflects the undervaluing of research on “soft” topics like animal emotions, is itself an interesting question. What is clear is that the questions asked in canine science — and the way they are framed — are not politically neutral.

Chapter 5: Canine Olfaction, Cognition, and the Question of Animal Consciousness

5.1 The Dog’s Nose: A Window onto Umwelt

5.1.1 Olfactory Anatomy and Capacity

The dog’s nose contains approximately 300 million olfactory receptor neurons, compared to roughly 6 million in humans. The olfactory epithelium (嗅觉上皮), the specialized tissue lining the nasal cavity where these receptors are located, covers an area of roughly 130–170 cm² in dogs versus 5 cm² in humans. The olfactory cortex accounts for a proportionally much larger share of the canine brain than in humans — approximately 12.5% of total brain volume compared to less than 1% in humans.

Dogs can detect odors at concentrations estimated to be 10,000 to 100,000 times lower than the human threshold, though estimates vary substantially by odorant compound. More precisely, dogs appear to have lower detection thresholds (检测阈值) for many biologically significant compounds — particularly fatty acids and other volatile compounds associated with biological organisms — than for arbitrary chemical stimuli. This suggests that canine olfaction has been shaped by evolution to be maximally sensitive to ecologically relevant information.

The anatomy of canine odor sampling is sophisticated in ways that go beyond raw receptor density. Dogs can control airflow through their nostrils independently; each nostril samples a slightly different region of the olfactory world, enabling olfactory stereopsy (嗅觉立体感) — the three-dimensional localization of odor sources, analogous to binocular vision in spatial localization. The Jacobson’s organ (雅各布森器官), or vomeronasal organ, provides a secondary olfactory channel specialized for detecting non-volatile compounds including pheromones; its function in dogs is not yet fully understood but is thought to mediate some aspects of conspecific communication and social recognition.

Dogs also have a distinctive pattern of nasal airflow during sniffing: they can sniff at rates of up to five times per second, and the expired airflow creates turbulence that stirs the surrounding odor environment, effectively sampling a larger volume of air per unit time than passive inhalation would permit. This active sampling (主动采样) behavior is controlled by specialized nasal muscles and respiratory patterns; experienced scent-detection dogs show highly efficient sniffing protocols adapted to the specific search task.

5.1.2 The Umwelt Concept

Alexandra Horowitz devotes sustained attention to olfaction in Inside of a Dog, using it to introduce the concept of umwelt (环境世界) — a term developed by the Baltic German biologist Jakob von Uexküll to describe the subjective, species-specific perceptual world of an organism. For von Uexküll, every animal lives in its own umwelt, a bubble of perceptible and meaningful stimuli defined by the species’ sensory and motor capacities.

5.1.3 Alexandra Horowitz’s Umwelt Research

Alexandra Horowitz’s research program at the Dog Cognition Lab at Barnard College, Columbia University, is distinctive in its sustained focus on the dog’s perspective — on what the world looks, smells, and feels like from inside canine experience. Her book Inside of a Dog is organized around the umwelt concept as a structuring idea: each chapter attempts to reconstruct some aspect of the dog’s perceptual and cognitive world from available scientific evidence.

Horowitz’s empirical work on dog cognition has addressed several questions relevant to the umwelt framework. Her widely publicized study on the “guilty look” in dogs (2009) found that the facial expression owners interpreted as guilt — flattened ears, lowered tail, averted gaze — was not correlated with whether the dog had actually transgressed, but with whether the owner believed and acted as if the dog had. Dogs showed “guilty” behaviors in response to owners scolding them, regardless of whether they had done anything wrong. The study is often cited as evidence that dogs do not experience guilt as an emotion; more cautiously, it demonstrates that the owner’s interpretation of canine facial expression is unreliable as evidence of the dog’s internal state.

Horowitz’s later work on dog olfactory cognition, including studies of nose-work and olfactory enrichment, has argued that allowing dogs to engage in extended olfactory investigation — sniffing at their own pace, following odor trails — is both intrinsically rewarding for dogs and cognitively enriching. This research has direct welfare implications: walks that allow dogs to stop and smell their environment, rather than proceeding at human pace, may be significantly more satisfying to dogs than walks that prioritize human exercise goals. The umwelt framework makes this point scientifically tractable: if the dog’s world is organized primarily around olfactory information, then practices that deny dogs access to olfactory experience are denying them access to their primary mode of engagement with the world.

5.1.4 Applied Olfaction: Detection, Diagnosis, and the Limits of Explanation

Dogs’ olfactory capacities have been exploited for a remarkable range of applied purposes: tracking, explosive and drug detection, search-and-rescue, and — increasingly — medical detection. Dogs have been trained to detect cancers (lung, colorectal, prostate), hypoglycemia in diabetic patients, and COVID-19 infection, often at accuracy rates comparable to or exceeding standard diagnostic tests.

These applications raise interesting questions for the philosophy of science. The detection capability exists; it can be documented in controlled trials. But the mechanism — precisely what volatile compounds the dogs are detecting — is often poorly understood. Science can establish that a phenomenon exists before it can explain why it exists. The relationship between these two kinds of scientific achievement is itself philosophically interesting.

5.2 Animal Consciousness: The Hard Problem in the Wild

5.2.1 Philosophical Frameworks

The question of animal consciousness has both scientific and philosophical dimensions. Scientifically, the question is whether and to what degree non-human animals have subjective experiences — whether there is something it is like to be a dog, a bat, or a bee. Philosophically, the question is whether and how we could ever know the answer.

Thomas Nagel’s famous paper “What Is It Like to Be a Bat?” (1974) argued that the subjective character of experience — what he calls phenomenal consciousness (现象意识) — is irreducible to any physical or functional description. Even a complete neuroscientific account of bat echolocation would not tell us what it is like, from the inside, to experience the world through echolocation. Nagel’s argument is directed primarily against physicalist theories that claim to reduce mental states to physical states without remainder. His point is not that bats are unconscious, or that we cannot know anything about their experience — it is that there is an aspect of experience (its subjective, qualitative character) that resists third-person, objective characterization.

Applied to dogs, Nagel’s argument generates a specific epistemic problem: we can characterize the neural correlates of canine olfactory processing in considerable detail, we can measure behavioral responses to odors, we can even measure the neurochemical consequences of olfactory stimulation — but none of this tells us what it is like to smell a fire hydrant from a dog’s perspective. The phenomenal quality of the experience — its smell-ness, its intensity, its richness as a source of information — is accessible, if at all, only from the inside.

This does not mean we should be agnostic about dog consciousness. Nagel explicitly allows that bats are conscious, and that echolocation involves phenomenal experience of some kind. His argument is about the limits of third-person scientific characterization, not about the existence of animal consciousness. The appropriate response to Nagel’s argument is epistemic humility — recognition of the limits of our knowledge — combined with a principled default toward assuming that animals with complex nervous systems and behavioral repertoires have phenomenal experience of some kind.

Cambridge Declaration on Consciousness (2012): A group of prominent neuroscientists, including Stephen Hawking as a witness, signed a declaration stating that non-human animals possess the neurological substrates that generate consciousness, including all mammals and birds. While not a scientific finding in itself, the declaration represents a significant shift in scientific consensus.

5.2.2 Marc Bekoff and Cognitive Ethology

Marc Bekoff is the foremost advocate of cognitive ethology (认知行为学) — the scientific study of animal minds — as a serious scientific enterprise. Bekoff argues that the evidence for rich emotional and cognitive lives in non-human animals is extensive, convergent, and methodologically robust: behavioral evidence, neuroanatomical homology, evolutionary continuity, and neurochemical similarity all point in the same direction.

Bekoff’s methodological stance is guided by what he calls biocentric anthropomorphism (以生命为中心的拟人论) — the careful, evidence-constrained attribution of mental states to animals, grounded in knowledge of their biology, ecology, and evolutionary history. This is not the naive anthropomorphism of projecting human emotional categories onto animals without regard for the differences between human and animal psychology; it is the principled application of what we know about the evolution of nervous systems and behavior to make informed inferences about animal minds.

Bekoff’s research on animal emotions draws on the concept of primary emotions (初级情绪) — evolutionarily ancient affective states (fear, rage, lust, care, seeking, panic, play) proposed by the neuroscientist Jaak Panksepp. Panksepp’s research demonstrated that these emotional systems are implemented in subcortical brain structures that are conserved across mammals, and that they can be identified by characteristic behavioral signatures, neurochemical profiles, and brain lesion patterns. Bekoff argues that the Pankseppian framework provides a scientific basis for attributing at least these primary emotional states to dogs and other mammals, independent of the more contentious question of whether animals have sophisticated self-aware consciousness.

5.2.3 Temple Grandin and Sensory-Based Cognition

Temple Grandin’s contributions to animal cognition are distinctive in that they are grounded in her own experience of autism — a form of cognitive organization that she argues is more similar to animal cognition than neurotypical human cognition. In Animals in Translation, Grandin argues that many animals (including dogs) are primarily sensory thinkers (感官思维者) — that their cognition is organized around specific, concrete sensory impressions rather than the abstract categories that dominate neurotypical human thought.

This hypothesis, if correct, has implications for both the scientific study of animal consciousness and the ethics of animal welfare. Animals may be more sensitive to specific sensory triggers — particular sounds, smells, visual patterns — than we typically assume, and their distress may arise from sensory overwhelm rather than from the kind of abstract anticipatory anxiety that characterizes human suffering.

Chapter 6: Dog Training — Science, Practice, and Policy

6.1 The Science of Learning and Its Application

6.1.1 Behaviorism and Classical Conditioning

The science of dog training is grounded historically in learning theory (学习理论) — specifically, in the behaviorist tradition associated with Ivan Pavlov, John Watson, and B.F. Skinner. Classical conditioning (经典条件反射) is the process by which a neutral stimulus becomes associated with a biologically significant stimulus, eventually eliciting a conditioned response. Pavlov’s famous experiments showed that dogs could be conditioned to salivate in response to a bell that had been repeatedly paired with food.

Operant conditioning (操作条件作用) extends this framework: behavior is shaped by its consequences. Behaviors followed by positive consequences (reinforcement) become more frequent; behaviors followed by negative consequences (punishment) become less frequent. Skinner’s systematic development of operant conditioning principles provided the theoretical basis for modern behavior modification.

The four-quadrant framework is a formal, descriptive taxonomy of how consequences affect behavior. It does not by itself prescribe which quadrant should be used in any given training situation. The prescriptive question — which methods are most effective and most welfare-positive — requires additional empirical and ethical considerations.

6.1.2 The Effectiveness Evidence

A growing body of empirical research compares training outcomes across methods that rely predominantly on positive reinforcement versus those that rely substantially on aversive methods (positive punishment or negative reinforcement). The evidence is consistent and substantial: positive reinforcement-based methods produce equivalent or superior behavioral outcomes on most measures, with significantly lower rates of fear responses, aggression, and stress-related behaviors.

A 2020 study by Ana Catarina Vieira de Castro and colleagues (published in PLOS ONE) compared dogs trained by aversive methods, mixed methods, and reward-based methods on measures of training effectiveness, stress behaviors, and urinary cortisol levels. Aversive-method dogs showed significantly higher rates of stress behaviors (yawning, lip-licking, lowered tail) during and after training sessions, and higher cortisol levels, without any advantage in training outcome. The study provided direct physiological evidence that aversive training methods cause measurable stress even outside of acute punishment episodes — consistent with the hypothesis that aversive training produces conditioned fear responses to the training context itself.

6.1.3 Patricia McConnell and the Ethological Turn

Patricia McConnell’s The Other End of the Leash represents a significant development in the popular science of dog training: the integration of ethology (行为学) — the study of animal behavior in its natural context — with learning theory. McConnell draws on her training as a zoologist and her experience as a certified applied animal behaviorist to argue that effective dog training requires understanding not just the mechanics of conditioning but the evolved social and communicative repertoire of dogs as a species.

McConnell emphasizes that dogs and humans are both highly social mammals, but we are social in different ways. Humans are primarily visual primates (视觉灵长类动物); dogs are primarily olfactory and auditory communicators. The mismatches between human and canine communicative styles — humans talk continuously, dogs communicate in bursts; humans approach face-to-face, dogs approach obliquely — generate many of the problems that bring dogs to behavioral consultants.

6.1.4 Jean Donaldson and the Science-Based Training Movement

Jean Donaldson’s The Culture Clash is a polemical and influential intervention in dog training culture. Donaldson argues that traditional dog training — rooted in dominance hierarchies, physical compulsion, and punishment — is not merely ethically problematic but scientifically bankrupt: it misrepresents dog cognition, produces inferior behavioral outcomes, and causes unnecessary suffering.

Donaldson’s central concept is the characterization of dogs as “amoral, hedonistic, and brilliant” (非道德的、享乐主义的、聪明的) — not moral agents who need to be dominated, but associative learners who respond to consequences. From this scientific characterization, she derives an ethical and practical conclusion: positive reinforcement-based training is not only kinder but more effective, because it works with the animal’s actual cognitive capacities rather than against them.

6.2 The Dominance Debate

6.2.1 The Alpha Wolf Model

From approximately the 1950s through the 1990s, an influential strand of dog training theory held that dogs are pack animals who instinctively seek to establish dominance hierarchies, and that effective training requires the human to establish “alpha” status. This dominance model (支配模型) drew on early ethological studies of wolf pack behavior, including work influenced by the ethological tradition of Lorenz, which described rigid linear dominance hierarchies in wolf packs.

The dominance model generated training practices — alpha rolls, scruff shakes, physical compulsion, the “dominance down” — that are now understood to be both ethically problematic and scientifically ungrounded. The alpha roll, in which the dog is forced onto its side or back, was claimed to mimic dominance displays in wolf packs. In fact, wolf body postures in which a subordinate animal rolls over are typically voluntary appeasement gestures, not forced submissions — and the claim that performing them on dogs communicates social dominance to the dog was never tested empirically.

6.2.2 The Scientific Debunking of Dominance Theory

Subsequent research revised the picture substantially. David Mech, whose early work on wolf behavior had contributed to the dominance model, subsequently argued — prominently and publicly — that the original research was based on studies of captive wolves: groups of unrelated individuals forced into artificial association in zoo enclosures. These captive groups did develop coercive hierarchies, because unrelated wolves competing for resources in a confined space have no basis for the cooperative relationships that characterize natural wolf packs. Free-ranging wolf packs, Mech documented, are typically family groups — breeding pairs and their offspring — with cooperative rather than coercive social structures. The “alpha” pair are simply the parents; their priority access to resources reflects parental status rather than won dominance.

The scientific debunking of the dominance model has been explicit and consistent. The American Veterinary Society of Animal Behavior issued a position statement in 2008 opposing the use of dominance-based training techniques, citing both the lack of scientific support and the evidence of harm. The statement noted that the theoretical foundation of dominance-based training — the claim that dogs are attempting to achieve social dominance over their owners — has no credible empirical basis in the behavior of either domestic dogs or wolves.

More fundamentally, dogs are not wolves. The 15,000+ years of domestication have produced an animal with a distinctive social orientation toward humans — one that cannot be adequately described by wolf pack dynamics. The selective breeding (选择育种) that produced dogs specifically favored individuals who were attentive to and compliant with human social cues, not animals who challenged human authority. Dogs are pre-adapted to defer to humans; the problem of “dominance” is largely a misinterpretation of normal canine social behaviors through a species-inappropriate theoretical lens.

6.2.3 The Science-Policy Gap in Dog Training

Despite the scientific consensus in favor of positive reinforcement-based training and against aversive methods, there is a significant and persistent gap between this consensus and actual practice — both among professional trainers and among dog owners. This gap is instructive for the philosophy of science and science communication.

Part of the explanation is institutional: dog training is essentially unregulated in most jurisdictions. Unlike veterinary medicine, which requires licensure and continuing education, anyone can call themselves a dog trainer regardless of educational background, training methods employed, or outcomes achieved. This absence of regulatory infrastructure means that the scientific consensus has no formal mechanism of implementation.

Part of the explanation is cultural: the dominance model is deeply embedded in popular culture, perpetuated by television programs featuring charismatic trainers who demonstrate dramatic behavioral interventions using physical corrections. These programs are not peer-reviewed; they are entertainment products, and they are selected for dramatic visual impact rather than scientific accuracy.

The science-policy gap in dog training is a specific instance of a general problem in knowledge translation (知识转化) — the challenge of moving scientific consensus into practice. The parallel with other science-policy gaps (climate change, vaccine safety, nutritional science) is instructive: in each case, entrenched practices, cultural narratives, and economic interests resist the uptake of scientific findings, and the deficit model of science communication — if we just tell people the facts, they will change their behavior — fails.

6.3 Science, Policy, and Dog Training Regulation

6.3.1 The Policy Gap

Despite the scientific consensus in favor of positive reinforcement-based training and against aversive methods, there is significant variation in the regulatory environment governing dog training. Unlike veterinary medicine, dog training is essentially unregulated in most jurisdictions — anyone can call themselves a dog trainer, regardless of educational background or training method.

This gap between scientific knowledge and policy is itself a philosophically interesting case study. It raises questions about how scientific knowledge is — and should be — translated into regulatory action, and about the role of professional organizations, consumer protection law, and public education in bridging that gap.

6.3.2 Animal Welfare Law and Its Limits

Animal welfare legislation typically prohibits cruelty — the deliberate infliction of unnecessary suffering — but does not mandate any particular standard of care beyond this minimum. The challenge for dog training policy is that the harms of aversive training methods are often diffuse, cumulative, and non-obvious: they manifest as anxiety, fear, and reduced welfare over time rather than as acute physical injury.

Chapter 7: Science and Society — Knowledge, Power, and Responsibility

7.1 The Social Contract of Science

7.1.1 Science as a Public Enterprise

Modern science is largely funded by public money and justified by appeal to public benefit. This creates what philosophers and sociologists have called the social contract of science (科学的社会契约): an implicit agreement between the scientific community and society, in which scientists receive resources, autonomy, and social prestige in exchange for producing knowledge that benefits the public and for maintaining the integrity of the scientific enterprise.

The terms of this contract are contested. Scientists have historically claimed that autonomy — freedom from direct political or public oversight — is necessary for good science, because public pressure can distort the production of knowledge. Critics argue that unchecked autonomy allows science to serve particular interests rather than the public, and that greater public participation in research priority-setting would produce a science more responsive to genuine social needs.

7.1.2 The Deficit Model and Its Failures

The dominant model of science communication through much of the twentieth century was the deficit model (缺陷模型): the public lacks scientific knowledge; the solution is to provide more information. On this model, public skepticism toward science is a product of ignorance, and the remedy is education. The model implies a one-way flow of information: scientists produce knowledge, communicators transmit it, and a passive public receives and accepts it.

The deficit model has been extensively critiqued in science studies. Empirical research consistently fails to find a strong positive correlation between scientific knowledge and acceptance of scientific consensus on contested issues. People who score highly on tests of scientific literacy are not significantly more likely to accept consensus positions on climate change, GMO safety, or nuclear power — and in some studies, higher science literacy predicts more polarized, not more convergent, attitudes. The explanation, researchers suggest, is that people use scientific literacy not primarily to update their beliefs impartially but to construct more sophisticated defenses of their existing positions.

The deeper problem with the deficit model is its unidirectional, asymmetric framing. It assumes that scientists have knowledge and the public has ignorance; that the appropriate relationship is communication from the former to the latter; and that the only legitimate outcome is public acceptance of scientific consensus. This framing obscures the extent to which science itself is shaped by public values, interests, and priorities — and the extent to which public skepticism may sometimes reflect legitimate concerns about the conduct, funding, or application of science rather than mere ignorance.

7.1.3 Public Engagement with Science

The successor to the deficit model is a cluster of frameworks grouped under the label of public engagement with science (科学公众参与), or more recently responsible research and innovation (负责任的研究与创新). These frameworks share the view that the appropriate relationship between science and society is dialogic rather than one-way: the public should be able to influence research priorities, raise concerns about research designs and applications, and participate in the governance of science rather than merely consuming its outputs.

In the context of animal research, public engagement takes several forms. Animal care and use committees increasingly include non-scientific lay members — a recognition that judgments about acceptable risk to animal subjects involve value questions that are not the exclusive province of scientists. Public consultations on animal research policy bring in stakeholder perspectives including animal welfare advocates, farmers, pet owners, and philosophers. Citizen science projects — including projects that enroll owned dogs as research subjects — create new forms of participation in which members of the public contribute data and, increasingly, help frame research questions.

The dog cognition research community has been unusually successful at public engagement, partly because dog owners are highly motivated participants and partly because the research questions — what do dogs feel, understand, and experience? — are questions that many people care about. The challenge is ensuring that this public interest translates into genuine participation in research governance rather than mere enthusiasm for findings that confirm existing beliefs.

7.2 Ethics of Science Communication

7.2.1 Responsibilities of Scientists as Communicators

Scientists who publish popular books — Berns, Horowitz, Bekoff, McConnell, Grandin — occupy a distinctive position. They are simultaneously scientific researchers, subject to the norms of scientific inquiry, and public communicators, subject to different (though related) standards of accuracy, accessibility, and responsibility. The tension between these roles is not always easy to navigate.

Overstating findings: Popular science accounts of dog cognition research frequently overstate the certainty or generality of experimental findings. A study showing that dogs prefer familiar human odors under specific laboratory conditions becomes, in popular press, “dogs love their owners’ smell.” The inflation of findings in translation from technical to popular discourse is a pervasive problem in science communication.

Understating uncertainty: Conversely, popular science accounts sometimes obscure the genuine uncertainty that characterizes the frontier of canine cognition research — uncertainty about the interpretation of neural data, the generalizability of laboratory findings, and the nature of animal consciousness — in favor of a more confident, narrative-friendly picture.

7.2.2 Democratic Dimensions of Science

Heather Douglas argues that the relationship between science and policy is not simply one of knowledge transfer. Scientists make value-laden choices — about what to study, how to study it, and how to characterize their results — that have political and ethical implications. Recognizing this does not undermine the authority of science; it clarifies what that authority is based on and what its limits are.

The precautionary principle (预防原则) is relevant here. In contexts of scientific uncertainty with potentially serious consequences, the precautionary principle suggests erring on the side of caution. In the context of animal research, this might suggest: given genuine uncertainty about the degree of animal suffering caused by experimental procedures, assume a higher rather than lower level of suffering, and design studies accordingly.

7.3 Science, Ethics, and the Future of Human-Dog Relations

7.3.1 What the Science Demands of Us

The convergent evidence from behavioral ecology, comparative neuroscience, and cognitive ethology supports the conclusion that dogs have rich inner lives: they experience positive and negative emotional states, they form genuine social bonds, they have preferences and aversions, and they suffer when treated badly. This is not merely a sentimental claim but a scientific one, supported by multiple converging lines of evidence.

If this scientific picture is approximately correct, it has normative implications. At minimum, it supports taking dog welfare seriously as a genuine moral concern rather than merely an expression of sentimental attachment. More substantively, it supports the reform of practices — in research, training, commercial breeding, and everyday care — that cause unnecessary suffering.

7.3.2 Persistent Questions

This course has raised as many questions as it has answered, which is as it should be. Among the questions that remain open:

What is the appropriate threshold for animal consciousness, and what kind of evidence can resolve disagreements about it? How should scientific uncertainty be weighted in policy decisions about animal welfare? What obligations, if any, do scientific communities have to ensure that their research programs are responsive to animal welfare concerns? How should competing values — the value of scientific knowledge, the value of animal welfare, the value of public trust in science — be weighed against one another?

7.3.3 Synthesis: Connecting the Course’s Central Themes

The philosophical analysis developed in this course converges on an integrated perspective that connects each of its central themes. The story of canine science is, at its core, a story about the inseparability of empirical inquiry and ethical commitment.

Consider how the themes interlock. The domestication of dogs — explored in Chapter 1 — created an animal uniquely adapted for human social life, whose cognitive and emotional architecture is organized around the human gaze, voice, and attention. This fact about dogs’ nature makes the ethics of using them as research subjects distinctively complex (Chapter 2): we are asking beings adapted for trust and cooperation to participate in procedures they did not choose, using a social relationship that domestication itself created. The Berns fMRI project (Chapter 3) is exemplary precisely because it took seriously the co-evolutionary relationship between dogs and humans, designing studies around dogs’ voluntary participation rather than their biological availability as subjects.

The value-ladenness of science (Chapter 4) — documented through the history of behaviorism’s suppression of animal mind research and its replacement by cognitive ethology — connects directly to the recovery of dogs as genuine subjects of experience rather than merely behavioral mechanisms. It was not purely new data that changed the field; it was the uptake of new values — the value of taking animal experience seriously, the epistemic legitimacy of attributing mental states to evolutionarily related species — that created the conditions under which the data could be interpreted differently.

The umwelt concept (Chapter 5) reminds us that scientific knowledge of dogs is always knowledge from the outside — a third-person reconstruction of a first-person reality that remains, in Nagel’s sense, partially inaccessible. This epistemic humility is not a counsel of despair; it is a prompt for methodological creativity and ethical caution. If we cannot fully know what a dog’s experience is like, we have additional reason to err on the side of welfare rather than convenience.

The science-policy gap in dog training (Chapter 6) illustrates what happens when scientific knowledge fails to engage the social, cultural, and institutional conditions of its uptake. The lesson is not that science is powerless to change practice, but that changing practice requires attending to more than the quality of evidence — it requires understanding the social epistemology of the communities that must receive and act on that evidence.

Finally, the science communication challenges explored in Chapter 7 circle back to the starting point: dogs are sites of intense human emotional investment, and that investment is both an obstacle and a resource for good science. It is an obstacle when it generates motivated reasoning, confirmation bias, and resistance to complicating evidence. It is a resource when it motivates rigorous inquiry, sustained public engagement, and genuine concern for the welfare of the animals being studied.

The science of dogs is, in this sense, a microcosm of science in society more generally. It illustrates, in vivid and emotionally accessible form, the central lessons of philosophy of science: that knowledge is socially produced; that values pervade scientific practice; that the ethics of research is inseparable from its epistemology; and that the relationship between scientific communities and the broader public is one of reciprocal obligation rather than one-way transmission.

Summary Table: Key Concepts by Chapter

Glossary of Key Terms