Search

Of the basic qualities that define perception perhaps the most important are darkness and lightness. An animal may see forms poorly, have little or no color vision, minimal depth perception, or even be effectively blind (e.g., some species of bats, moles, and mole rats). But most animals, many plants with and even bacteria discriminate light and dark.

The renewed enthusiasm for neural networks in the 1970s was largely based on a computational means of altering the weights of network connections by a mathematical feedback technique called “back propagation.” This advance introduced a way of continuously modifying the weights of network connections based on feedback from the network outputs. Enthusiasm for this approach continues unabated today because, at least in part, this is what brains seem to be doing.

Unless one goes looking for trouble, it is reasonable to assume that objective reality exists and that physical instruments measure it accurately, or at least as accurately as currently possible. This reality is defined by laws of physics and, with respect to the understanding the behavior of biological agents on Earth, we need not worry much about deeper issues that physicists are still grappling with (e.g., how quantum events are related to macroscopic events; the role of observation in reification; whether information subsumes matter and energy). For present purposes objective reality can be restricted to classical physics. In contrast, perceptions—whether conscious or as unconscious equivalents—are subjective.

An obvious way nervous systems differ from electronic computers is brains give rise to perceptions, defined as what we subjectively experience, while computers do not. This fact, however, does not mean that neuroscientists know how perceptions arise, what they are or whether non-biological machines could ever give rise to them. For many investigators nervous system is taken to construct representations of the physical world. Based on what has already been said, however, this option is implausible. Moreover, there are profound discrepancies between the physical sources of stimuli and what we perceive.

The concept of stimuli as energy from the external or internal environment impinging on receptors is not as simple as it may seem. Sensory systems include pre-neural filters and amplifiers that alter the energy before it reaches receptor neurons. Once the receptors have transduced the resulting energy, the peripheral neural signals interact with other neurons in an increasingly complex hierarchy. What is passed on to these poorly defined higher order systems—the welter of neural connections that lie between inputs and outputs—are “stimuli” in their own right. To complicate matters even further, these higher order effects elicit feedback to the lower order stations they came from in a never-ending loop.

Significance The foundations of human music have long puzzled philosophers, mathematicians, psychologists, and neuroscientists. Although virtually all cultures uses combinations of tones as a basis for musical expression, why humans favor some tone combinations over others has been debated for millennia. Here we show that our attraction to specific tone combinations played simultaneously (chords) is predicted by their spectral similarity to voiced speech sounds. This connection between auditory aesthetics and a primary characteristic of vocalization adds to other evidence that tonal preferences arise from the biological advantages of social communication mediated by speech and language., Musical chords are combinations of two or more tones played together. While many different chords are used in music, some are heard as more attractive (consonant) than others. We have previously suggested that, for reasons of biological advantage, human tonal preferences can be understood in terms of the spectral similarity of tone combinations to harmonic human vocalizations. Using the chromatic scale, we tested this theory further by assessing the perceived consonance of all possible dyads, triads, and tetrads within a single octave. Our results show that the consonance of chords is predicted by their relative similarity to voiced speech sounds. These observations support the hypothesis that the relative attraction of musical tone combinations is due, at least in part, to the biological advantages that accrue from recognizing and responding to conspecific vocal stimuli.

Brains as Engines of Association seeks an operating principle of the human brain and is divided into four parts. The first part (“What Nervous Systems Do for Animals”) is intended to set the stage for understanding the emergence of neural systems as promoting what all organisms must accomplish: survival and reproduction. The second part (“Neural Systems as Engines of Association”) lays out the general argument that biological sensing systems face a daunting problem: they cannot measure the parameters of the world in the way physical instruments can. As a result, nervous systems must make and update associations (synaptic connections) on the basis of empirical success or failure over both evolutionary and individual time. The third part (“Evidence that Neural Systems Operate Empirically”) reviews evidence accumulated over the past 20 years that supports this interpretation in vision and audition, the sensory systems that have been most studied from this or any other perspective. Finally, the fourth part (“Alternative Concepts of Neural Function”) considers the pros and cons of other interpretations of how brains operate. The overarching theme is that the nervous systems of humans and every other animal operate on the basis associations between stimuli and behavior made by trial and error over species and lifetime experience.

Dealing successfully with objects and conditions in the world entails another basic visual category: the geometry of objects. This domain is also rich in perceptual phenomenology that needs to be explained. The problem in understanding perceived geometry is much the same as the problem of understanding lightness values and colors: biological vision lacks the tools—in this case rulers, protractors, radar guns, laser range scanners—needed to measure geometrical reality. As a result, the geometries we see always differ from the measured parameters of the physical world. The aim of this chapter is to consider geometrical examples that further support the conclusion that vision operates by ranking perceptual values according to the frequency of occurrence of biologically useful stimulus patterns.

If the operating principle of nervous systems is making trial-and-error associations among neurons by accumulated learning over evolutionary and individual time, a next step is to define “associations” more exactly and to review the ways they are created. Neurobiologists have generally focused on the cellular mechanisms that alter synaptic structure, physiology, and biochemistry as a result of individual experience (“neural plasticity”), while psychologists have approached neural associations from the perspective of behavior. Neuroscientists of all stripes agree, however, that associations are based on the connections among neurons and neuronal populations. This chapter examines more specifically how neural connectivity evolves, how it arises during development, and how it is modified over animal lifetimes.

Definitions of the term “animals” in dictionaries and textbooks are surprisingly vague. The characteristics usually mentioned are eukaryotic, multicellular, heterotrophic, sexually reproducing, and capable of rapid and independent movement. But some or all of these properties are characteristic of many organisms in the other kingdoms of life on Earth. In fact, the major distinguishing feature of animals in most cases is the presence of a nervous system. But if nervous systems are indeed one of the main attributes that distinguish organisms in the animal kingdom, what exactly are nervous systems and what advantages do they bring? Without at least some provisional answers, seeking the operating principle of neural systems would be futile.

Basic to the question of whether or not the brain and the rest of the human nervous system have a simple operating principle are some central facts about biology and its relation to neuroscience. What nervous systems do is best appreciated in the context of what all organisms must accomplish in order to survive and prosper, with or without neural assistance. Although the author’s understanding of these issues is no more than that of any other student who pays a modicum of attention to the broader sweep of scientific progress, this chapter considers some points of consensus. The aim is to situate the quest for a principle of neural function in the context of biology writ large.

The reason for using vision as an example in the previous three chapters is that more is known about the human visual system and visual psychophysics than about other neural systems. But this choice begs the question of whether other systems corroborate the evidence drawn from vision. Is the same empirical strategy used in other sensory systems to contend with the same problem (i.e., the inability of animals to measure the actual properties of the world)? Based on accumulated anatomical, physiological, and psychophysical information, audition is the best bet in addressing this question in another modality. This chapter examines whether the perception of sound can also be explained empirically as a way to deal with a world in which the physical parameters of sound sources can’t be apprehended.

Perceived motion is defined as the apparent speed and direction of objects that are translating and/or rotating in three-dimensional space. It has long been clear, however, that the perceived speeds and directions of moving objects disagree with physical measurements of motion. With respect to speed, the flash-lag effect has been a major focus; with respect to direction, the emphasis has been on the effects of apertures. These phenomena—and many others—raise the question of how we routinely succeed in the world despite blatant discrepancies between the motions we see and the physical speeds and directions of the objects we must deal with. Resolving this puzzle presents another body of evidence pertinent to whether empirical ranking is the way nervous systems link objective and subjective domains.

Nervous systems employ some or all of the senses to convert energy at the level of receptor cells into neural information. This conversion generates electrochemical signals carried centrally that, via a series of additional neurons, determine behavior. The most obvious behaviors are those mediated by skeletal muscles, cardiac muscles, smooth muscles, and glands. But motor responses are members of a longer list that includes attention, perception, emotion, memory, thought, motivation, and others. How to describe these additional “systems,” and how to decipher what the relevant circuits do and how they do it, are unanswered questions. This chapter describes the main features of neural processing, asking whether the principle of neural function is simply to make associations that lead to useful behaviors.

A major challenge in neuroscience today is to decipher the operating principle of the brain and the rest of the nervous system in the same straightforward way that biologists have come to understand the functions of other organs and organ systems (e.g., the cardiovascular system, the digestive system, and so on). The argument here has been that the function of nervous systems is to make, maintain, and modify neural associations that ultimately promote survival and reproduction in a world that sensory systems can’t apprehend. In this way, we and other animals can link the subjective domain of perception to successful behavior without ever recovering the properties of the world. Neural function on a wholly empirical basis may be the key to understanding how brains operate.

Although understanding neural functions has progressed at a remarkable pace in recent decades, a fundamental question remains: How does the nervous system relate the objective world to the subjective domain of perception? Everyday experience implies that the neural connections on which we and other animals depend link physical parameters in the environment with useful responses. But that interpretation won't work: biological sensory systems cannot measure the physical world. Whereas something is linking sensory inputs to useful responses, it is not the physical world that instruments measure. How, then, have we animals met this challenge, and what is it that we end up perceiving? The purpose of this chapter is to suggest how nervous systems have evolved to deal with the inability to convey the objective properties of the real world.

What, then, is the evidence that sensory systems link stimulus inputs to useful responses empirically as a means of generating successful behavior in a physical world that the senses cannot measure? This chapter focuses on evidence derived from studies of lightness and color in vision, the brain system that has been most extensively studied in this regard. The argument here, and in the following chapters that consider other perceptual qualities and systems, is that evolved circuitry based on accumulated experience with frequency of occurrence of biologically useful stimuli accomplishes this feat. This strategy, called empirical ranking theory, explains why the qualities we perceive are always at odds with physical measurements.

Given the argument and evidence that the operating principle of nervous systems is to generate neural associations on a wholly empirical basis, it makes sense to compare this framework with other ideas that neuroscientists have proposed for understanding brain function. These concepts fall into several broad categories: (1) neural operation based on detecting, computing, and representing stimulus features as such; (2) neural computation and representation based on statistical inferences about physical reality; and (3) neural operation based on efficient and/or predictive coding. Another concept pertinent to all these ideas, including neural operation as empirical association, is whether nervous systems are carrying out computations. This section reviews these conceptual options and some implications that follow.

Goffinet’s complaints (1) about our vocal similarity hypothesis (2) are unwarranted on both practical and theoretical grounds. First, the issue of tuning is a red herring. We used standard just intonation intervals to evaluate the consonance of chords because their role in music across historical and cultural boundaries is foundational (3). We limited the chords tested to combinations of the 12 intervals of the chromatic scale for practical reasons: additional ratios and/or tuning systems would have increased the chords that had to be evaluated to an unreasonable number. More generally, our results are unlikely to depend on a particular tuning system. The popular chords in our study … [↵][1]1To whom correspondence should be addressed Email: dan.bowling{at}univie.ac.at. [1]: #xref-corresp-1-1

The basis of musical consonance has been debated for centuries without resolution. Three interpretations have been considered: ( i ) that consonance derives from the mathematical simplicity of small integer ratios; ( ii ) that consonance derives from the physical absence of interference between harmonic spectra; and ( iii ) that consonance derives from the advantages of recognizing biological vocalization and human vocalization in particular. Whereas the mathematical and physical explanations are at odds with the evidence that has now accumulated, biology provides a plausible explanation for this central issue in music and audition.