Revising the empirical picture--论文代写范文精选

调用不同的效应细胞，不同的行为,但没有明显的理由排除这是有作用的，收敛和发散型的基础，进一步感觉运动的复杂性,从跟踪和分离表示，可以看得更清楚更精细,但仍相对基本情况。下面的essay代写范文讲述了这一问题。

Abstract
The abstract structure of Sterelny’s theory has been considered at some length, and it has been suggested that the late separation of indication and control uncertainty is implausible. If his empirical survey is right however, then the models of figure 2 & figure 16 are supported by the evidence. This section re-examines the empirical picture.

The early evolution of decoupled representation The early evolution of nervous systems is a useful point of comparison for Sterelny’s account of representational complexification. As a very simple multicellular animal without a nervous system, sponges provide the baseline state. In sponges cells called myocites perform both sensory and motor functions (figure 17a), contracting to regulate water flow through the body of the sponge. Because they perform both sensory and motor functions myocites are called independent effectors (Swanson 2003, p. 16). 

The next grade of complexity is illustrated in Cnidaria, where sensory and effector functions have separated (figure 17b). Sensory neurons, derived from ectoderm, have specialized for sensory discrimination. This allows much greater specialization for sensory functioning, and it allows the cells to be positioned in locations best suited for sensing, as opposed to motor actions. But in addition the separation of function allows a given sensory neuron to innervate multiple effector cells (called divergence), and it allows a given effector cell to receive input from multiple sensory cells (called convergence). This is shown in figure 17c. These properties allowed dramatic advances in sensorimotor complexity, which included specialized sensory systems dedicated to particular classes of stimuli, such as chemicals, temperature and light (Swanson 2003, p. 21).

Comparison of figure 17c with 2b and 2c suggests that proto or elementary versions of both of Sterelny’s grades of representational complexity are present in the Cnidarian nervous system. In 17c B1 takes input from S1 and S2, so has at least part of the structure of robust tracking, and S1 innervates B1 and B2, thus having the form of decoupled representation. Calling different effector cells different behaviors might seem a stretch, but there is no obvious reason to rule this out and it is a short step from divergent innervation of effector cells to influencing multiple distinct overt behaviors (illustrated in examples discussed below).

Convergence and divergence form the basis for further sensorimotor complexification, and both robust tracking and decoupled representation can be seen more clearly in more elaborate, yet still relatively basic, cases. Figure 18a shows an intuitive example of a core perceptual mechanism, hierarchical feature analysis, from Hubel and Wiesel’s (1962) account of hierarchical feature construction in the cat visual cortex. Neurons in the lateral geniculate nucleus (LGN) are sensitive to dot shaped stimuli. Neurons in the striate cortex (or V1) called simple cells take input from multiple LGN neurons and are sensitive to oriented lines. It is not depicted in 18a, but Hubel and Wiesel showed an additional stage: some simple cells integrate from LGN cells sensitive to dark dots, and hence are sensitive to dark oriented lines, whilst others integrate from cells sensitive to light dots, and are sensitive to light oriented lines. 

Complex cells integrate across light- and dark-sensitive simple cells, and are sensitive to lines of a particular orientation, whether light or dark. Downstream areas of the visual system extract more complex features: for instance in primates neurons in MT are sensitive to motion information of various kinds, whilst neurons in the temporal cortex are sensitive to object features and identity. Visual processing in primates is organized into major partially parallel streams, beginning with M, P and koniocellular pathways from the retina through LGN and extending to the dorsal and ventral streams (Reid 2003).

Hierarchical feature analysis is an efficient and powerful way to construct more elaborate forms of perception from the starting point seen in Cnidaria. Individual receptor cells are the basic sensing unit on which complex perception is based, which means that the discrimination of complex stimuli must be achieved through comparison across multiple receptor cells. A given receptor cell is often highly ambiguous with respect to possible external signal sources, but this ambiguity can be reduced by downstream integration, as shown in 18a (and 12b). By iterating integration across multiple stages, increasingly complex high order information can be extracted, corresponding to increasingly complex features of distal sources. 

The relevant point here is that such feature representations will generally count as cases of both robust tracking and decoupled representation. The robust tracking aspect is apparent from 18a: the features are differentiated by means of multiple pathways back to the source. The decoupled representation aspect isn’t diagrammed, but there are strong reasons to think that many feature representations can contribute to multiple behaviors. In the case of vision, e.g., many states and entities will be represented in terms of combinations of location, movement, form, and color information. Likewise, in the case of gustation many items experienced orally will be represented in terms of combinations of sweet, salty, sour and bitter features. Each feature will participate in the representation of many different specific things, and thereby contribute to different behaviors. In other words, a given feature can point to different behaviors as part of different feature complexes, just as depicted in figure 13.

Hierarchical feature integration probably emerged early in the evolution of specialized perceptual systems, in part because of the constraints of perceiving complex stimuli with individual receptor cells as the base sensory unit. Comparison across receptors then provides the only way to discriminate features more complex than individual receptors can detect. In the case of eyes, the organization of the eye, retina, and downstream pathways enables highly structured comparisons. This integrative organization begins very simply; as Land and Nilsson point out, “Even the simple pit eye of a planarian flat worm…has some ability to compare intensities in different directions” (2002, p. 4). 

The evolution of eyes was already well advanced by the early Cambrian (Schoenemann 2006), and Land and Nilsson suggest that visually guided predation may have been the trigger for the Cambrian explosion (2002, p. 2). This will have required relatively complex integrative information processing mechanisms because a mobile predator must identify its prey and contend with effect of its own movement on perception. ‘Nystagmus’, the ability to hold a fixed point of focus during movement and to rapidly shift to a new point, allows stable perception during movement. It is found in contemporary crustaceans, and Schoenemann argues that the stalk eyes of the crustacean Leanchoilia, found in the early Cambrian, may have supported nystagmus and, combined with strong swimming ability, allowed a predatory lifestyle (2006, p. 311).

Neural convergence is at least a necessary condition for robust tracking, and the evidence concerning nystagmus in Leanchoilia indicates that this potential was being exploited in relatively sophisticated ways in some of the earliest multicellular animals. Divergence supports decoupled representation, and by extending the argument at the end of the last section it can be predicted that there should have been an early elaboration of divergence to form increasingly complex, clearly articulated decoupled representations. 

Specialized neurons provide a long range, point-to-point communication mechanism that can serve either convergence or divergence just as easily. The appearance of specialized perceptual systems was, in effect, an adaptive investment in specialized mechanisms for indication uncertainty reduction. Distal perception requires both convergence and divergence, since convergence is needed for complex feature discrimination, and the effective discrimination of distal things requires flexible integration of feature information. Animals also have a pressing need for control uncertainty reduction, and convergence applied to motor control provides multiconditionalization, a powerful form of behavior targeting. Since perceptual information will often be relevant to more than one behavior, mechanisms for distributing perceptual information to multiple behaviors should be favored.

Pavlovian and operant learning are mechanisms of this kind. The perceptual discriminations that provide the perceptual raw materials for associative learning count as decoupled representations because, prior to learning, they aren’t coupled to particular behaviors, and with ongoing learning they can be flexibly uncoupled (extinction) and coupled to new behaviors. The phylogenetic distribution of associative learning is very wide – for instance, the sea snail Aplysia californica is a model organism for research on the molecular mechanisms of associative learning (e.g., Kandel and Hawkins 1992) – which points to early evolutionary origins. 

Taken together, these lines of evidence cast doubt on the stage structure of the 2a → 2b → 3b model – robust tracking and decoupled representation appear together, rather than in sequence – and they indicate that the empirical interpretation of the theory described in section 2.2.1.2 is mistaken. As predicted by the argument at the end of section 3.2.2.3, multiconditionalization and decoupled representation appear near the beginning of the evolution of multicellular animals, rather than near the end. The abstract structure of Sterelny’s account of the conditions that select for decoupled representation is correct: information arrives in a piecemeal fashion, and may be relevant to more than one behavior (p. 78). However these circumstances aren’t unique to the social worlds of great apes. In the early evolution of neurally-based sensorimotor systems information arrives in a piecemeal fashion, in part because the base sensory unit is the individual receptor cell, and because the discrimination of distal stimuli typically depends on integrating many different signals. The features discriminated can be relevant to multiple behaviors, so are functionally decoupled.（essay代写)

51Due网站原创范文除特殊说明外一切图文著作权归51Due所有；未经51Due官方授权谢绝任何用途转载或刊发于媒体。如发生侵犯著作权现象，51Due保留一切法律追诉权。
更多essay代写范文欢迎访问我们主页 www.51due.com 当然有essay代写需求可以和我们24小时在线客服 QQ:800020041 联系交流。-X（essay代写)