Principles and application of the receptor noise model of color discrimination: a comment on Olsson et al.
Principles and application of the receptor noise model of color discrimination: a comment on Olsson et al.
复制标题
颜色辨别受体噪声模型的原理和应用：对 Olsson 等人的评论。
DOI：
10.1093/beheco/arx153
复制
发表时间：
2018
影响因子：
2.4
通讯作者：
Osorio D
中科院分区：
文献类型：
--
作者：
Osorio D
颜色为科学提供了两种面孔：我们可以精确地根据动物的光感受器兴奋来确定动物可以获得的光谱信息，但另一个人，更不用说另一个物种，实际上是如何体验颜色的问题，仍然是一个最受欢迎的哲学难题。我们如何才能弥合这一差距？Olsson和他的同事(2017)解释了如何将颜色视觉模型参数化，这是我们20年前提出的(Osorio和Vorobyev 1996；Vorobyev 1997；Vorobyev和Osorio 1998)，自那以后已经受到对颜色感兴趣的行为生态学家的欢迎(Maia等人。2013年)。在这里，我们简要地看看该模型背后的原则，以及它们对其应用的影响。上世纪90年代的S，我们对色觉的进化很感兴趣。例如，为什么人类和蜜蜂的光感受器具有不同的光谱敏感性。为了回答这个问题，我们考察了颜色视觉的理论基础。优雅的早期模型，赫尔曼·冯·赫姆霍尔茨的线元素和Erwin Schrödinger对其进行的修改，都假设感知受光感受器噪声的限制。即使我们对颜色处理一无所知，这些模型也可以使用。光感受器产生的噪声应限制(或匹配)性能的假设源于基本工程原理(Sterling和Laughlin，2015)。然而，对Helmholtz和Schrödinger线元素的预测与实验明显不一致，因此颜色科学家得出结论，颜色辨别受到视觉系统后期产生的噪声的限制。因此，对颜色阈值的分析可以用来揭示光感受器阶段以外的颜色视觉机制。在实验证据的指导下，我们修改了亥姆霍兹线元，增加了仅基于色度来区分颜色的假设(Osorio和Vorobyev 1996；Vorobyev 1997；Vorobyev和Osorio 1998)。所得到的方程描述了理想的色度检测器的性能，该色度检测器不在颜色处理的感受器后阶段添加噪声，并且忽略了颜色信号的非色差分量。我们最初的意图是，对实际颜色阈值与我们模型预测的偏差的分析，将使我们能够对蜜蜂的颜色处理机制做出推断(Vorobyev，1997)。令人惊讶的是，偏差太小，无法分析--该模型完全符合蜜蜂的光谱敏感性(Vorobyev等人)。2001年)，它也与使用边缘模糊的大刺激测试的人类的光谱敏感性一致(Vorobyev和Osorio 1998)。该模型在预测阈值方面的成功表明，在许多动物中，在颜色处理的后期添加的噪声很少，即颜色处理几乎是理想的，并突出了在视觉行为中使用彩色和非彩色信号之间的区别。感受器噪声模型是有用的，因为它基于简单的生理原理，并允许我们解释光感受器光谱敏感性的功能(Osorio和Vorobyev，1996)。这为比较和进化研究提供了很好的基础。模型预测是一种“零假设”，用来测试中枢大脑机制的证据，例如对手通道、颜色分类或天生的颜色偏好(Scholtyssek等人)。2016)。该模型已经过测试，特别是在鸟类身上(见附文)。尽管如此，对动物辨别非常相似颜色的能力的测试可以合理地假设，受体的反应相对于平均光照水平的波动是线性的。…的分歧在哪里
Color presents 2 faces to science: we can precisely specify the spectral information available to an animal in terms of its photoreceptor excitations, but the question of how another human, let alone another species, actually experiences color remains a favorite philosophical conundrum. How can we bridge this gap? Olsson and his coworkers (2017) explain how to parametrize a model of color vision, which we proposed 20 years ago (Osorio and Vorobyev 1996; Vorobyev 1997; Vorobyev and Osorio 1998), and has since become popular with behavioral ecologists interested in color (Maia et al. 2013). Here, we look briefly at the principles underlying the model, and their implications for its application. In the 1990’s, we were interested in the evolution of color vision. For example, why human and bee photoreceptors have different spectral sensitivities. To answer this question, we looked at the theoretical foundations of color vision. Elegant early models, Hermann von Helmholtz’s line element and its modification by Erwin Schrödinger, assume that perception is limited by photoreceptor noise. These models can be used even if we know nothing about color processing. The assumption that performance should be limited by (or matched to) noise originating in the photoreceptors follows from fundamental engineering principles (Sterling and Laughlin 2015). However, predictions of the Helmholtz and Schrödinger line elements markedly disagreed with experiments, and color scientists therefore concluded that color discrimination is limited by the noise originating later in the visual system. It follows that analysis of color thresholds can be used to reveal the mechanisms of color vision beyond the photoreceptor stage. Guided by experimental evidence, we modified the Helmholtz line element by adding the assumption that colors are discriminated on the basis of chromaticity alone (Osorio and Vorobyev 1996; Vorobyev 1997; Vorobyev and Osorio 1998). The resulting equations describe performance of an ideal chromatic detector that does not add noise at the postreceptor stages of color processing and ignores the achromatic component of color signal. We first intended that the analysis of the deviations of actual color thresholds from predictions of our model would allow us to make inferences about mechanisms of color processing in the honeybee (Vorobyev 1997). Surprisingly, the deviations were too small to analyze—the model fitted perfectly the spectral sensitivity of the honeybee (Vorobyev et al. 2001), it also agreed with the spectral sensitivity of humans tested using large stimuli with blurred edges (Vorobyev and Osorio 1998). The success of the model in predicting thresholds indicates that in many animals little noise is added at the later stages of color processing, that is, color processing is nearly ideal, and highlights the distinction between the uses of chromatic and achromatic signals in visual behavior. The receptor noise model is useful because it is based on straightforward physiological principles, and allows us to interpret the function of photoreceptor spectral sensitives (Osorio and Vorobyev 1996). This provides an excellent basis for comparative and evolutionary studies. Model predictions are a “null hypothesis” against which to test evidence of central brain mechanisms, such as opponent channels, color categorization, or innate color preferences (Scholtyssek et al. 2016). The model has been tested, especially in birds (see accompanying paper). Nonetheless tests of an animal’s ability to discriminate very similar colors can reasonably assume that receptor responses are linear with respect to fluctuations about the mean light level. Where differences …