Freeman T C
School of Psychology, Cardiff University, PO Box 901, CF10 3YG, Cardiff, UK.
Vision Res. 2001 Sep;41(21):2741-55. doi: 10.1016/s0042-6989(01)00159-6.
By adding retinal and pursuit eye-movement velocity one can determine the motion of an object with respect to the head. It would seem likely that the visual system carries out a similar computation by summing extra-retinal, eye-velocity signals with retinal motion signals. Perceived head-centred motion may therefore be determined by differences in the way these signals encode speed. For example, if extra-retinal signals provide the lower estimate of speed then moving objects will appear slower when pursued (Aubert-Fleischl phenomenon) and stationary objects will move opposite to an eye movement (Filehne illusion). Most previous work proposes that these illusions exist because retinal signals encode retinal motion accurately while extra-retinal signals under-estimate eye speed. A more general model is presented in which both signals could be in error. Two types of input/output speed relationship are examined. The first uses linear speed transducers and the second non-linear speed transducers, the latter based on power laws. It is shown that studies of the Aubert-Fleischl phenomenon and Filehne illusion reveal the gain ratio or power ratio alone. We also consider general velocity-matching and show that in theory matching functions are limited by gain ratio in the linear case. However, in the non-linear case individual transducer shapes are revealed albeit up to an unknown scaling factor. The experiments show that the Aubert-Fleischl phenomenon and Filehne illusion are adequately described by linear speed transducers with a gain ratio less than one. For some observers, this is also the case in general velocity-matching experiments. For other observers, however, behaviour is non-linear and, according to the transducer model, indicates the existence of expansive non-linearities in speed encoding. This surprising result is discussed in relation to other theories of head-centred motion perception and the possible strategies some observers might adopt when judging stimulus motion during an eye movement.
通过将视网膜和追踪眼动速度相加，就可以确定物体相对于头部的运动。视觉系统似乎很可能通过将视网膜外的眼动速度信号与视网膜运动信号相加来进行类似的计算。因此，以头部为中心的感知运动可能由这些信号编码速度的方式差异所决定。例如，如果视网膜外信号提供的速度估计较低，那么在追踪时移动物体看起来会更慢（奥伯特 - 弗莱施尔现象），而静止物体则会朝着与眼动相反的方向移动（菲勒内错觉）。以前的大多数研究认为，这些错觉的存在是因为视网膜信号准确地编码了视网膜运动，而视网膜外信号低估了眼动速度。本文提出了一个更通用的模型，其中两种信号都可能存在误差。研究了两种类型的输入/输出速度关系。第一种使用线性速度传感器，第二种使用非线性速度传感器，后者基于幂律。结果表明，对奥伯特 - 弗莱施尔现象和菲勒内错觉的研究仅揭示了增益比或幂比。我们还考虑了一般的速度匹配，并表明在理论上，线性情况下的匹配函数受增益比限制。然而，在非线性情况下，尽管存在一个未知的缩放因子，但单个传感器的形状得以揭示。实验表明，奥伯特 - 弗莱施尔现象和菲勒内错觉可以用增益比小于1的线性速度传感器得到充分描述。对于一些观察者来说，在一般的速度匹配实验中也是如此。然而，对于其他观察者来说，行为是非线性的，根据传感器模型，这表明在速度编码中存在扩展的非线性。结合以头部为中心的运动感知的其他理论以及一些观察者在眼动过程中判断刺激运动时可能采用的策略，对这一惊人结果进行了讨论。
