The usual way of looking at the flash-lag effect (FLE) is to (implicitly) assume that the brain contains fixed position-maps (e.g., retina-centric, head-centric, etc.) and that incoming retinal signals need to be assigned a place on these maps. Moving objects are in the habit of changing their position while your brain is still trying to pinpoint them on a map. Therefore, it would be a smart move if the brain were to place them on the map at the position they will have reached at pinpointing time. Nijhawan calls this visual prediction. It is relatively easy to envision such a visual prediction for the whereabouts of moving targets. Motion sensors report velocity vectors, and the corresponding position shifts in the mapping process can be (re-)calibrated during the visuomotor experience. It would be smart to do this in the front end of the visual system, so that both the action system and the perception system can profit from the shifted mapping. Lankheet and I have developed a simple implementation of this idea (van de Grind Reference van de Grind, Glicksohn and Myslobodski2006, Fig. 1; cf. Lankheet & van de Grind, in press) in which motion stop signals can curtail the mapping shift. Motion prediction as well as start and stop signals are already available at the retinal level (Berry et al. Reference Berry, Brivanlou, Jordan and Meister1999; Schwartz et al. Reference Schwartz, Taylor, Fisher, Harris and Berry2007b). Further refinement is possible in motion sensors in V1 and middle temporal area (MT) and in the subsequent mapping on a head-centric position map. Cantor and Schor (Reference Cantor and Schor2007) presented arguments and evidence that further support such low-level explanations. Echoes of these ideas can be found in the target article, but in the end Nijhawan prefers an exclusively high-level model, including feedback from the visuomotor action-system to the perceptual system. Here I want to sketch a rather more grandiose low- to middle-level alternative for the prediction hypothesis.
A neurophysiologist can see how a position in the environment or on the retina is mapped onto a position in the brain of a studied subject, but such a comparison is not possible for the subject. Without a homunculus, we cannot read physical map positions in our own brain, so external positions have to be more abstractly coded in terms of processable neural activity. It has been shown by Koenderink (Reference Koenderink, van Doorn, van de Grind and Koenderink1984a; Reference Koenderink1984b) that it is possible to neurally construct such an abstract “local sign” system from scratch, by analyzing incoming information with simple operators. Receptive fields and brain maps set up correlation constraints because of overlap and inclusion relations, and local signs are embedded in the correlation-structure of activity patterns. This requires interaction with the environment to (re-)calibrate local sign information in the brain. Similar remarks hold for the time sense (van de Grind Reference van de Grind2002; Reference van de Grind, Glicksohn and Myslobodski2006). Static coordinates are not suitable as references for the construction and maintenance of such a local sign system. In natural vision, everything always flows (or jumps) on the retina. Therefore, motion information is the best anchor for action control, setting the momentary coordinates. Let us hypothesize that motion is used to build up and continuously recalibrate the local sign system. This makes motion an “absolute” quantity, and position and time relative quantities, as in physics. For the FLE that would mean that the position map is always tuned to motion, so that moving objects are normally seen veridically at their instantaneous physical position, regardless of their processing time. Moving objects thus form the position reference for all non-moving objects. The position of the flash in the FLE is therefore correctly seen to lag the position of the moving reference – a shift that for logical reasons appears to equal the speed times latency difference. The flash localization in the FLE is correct, not an illusion, because at the time that we can report the flash position, that position indeed lags the motion reference position, in terms of momentary physical coordinates.
Interestingly, this was the implicit assumption in some of the earliest papers on the flash-lag effect, by Hazelhoff and Wiersma (Reference Hazelhoff and Wiersma1924; Reference Hazelhoff and Wiersma1925), based on Hazelhoff's Ph.D. thesis of 1923. Hazelhoff wanted to measure the perception time of a flash; that is, how long after a brief flash do we consciously perceive it? He invented a new method of measuring this latency. His subjects had to pursue a black vertical bar moving along a horizontal scale with a speed of 23 degrees per second. When the bar passed point 0, a white line was flashed (10–20 msec), centered on the slightly wider black bar, and the subject had to report the bar's scale-position at the moment the flash was seen. Hazelhoff calculated the presumed perception times for a range of flash luminances and for three subjects. The correspondence to modern findings is striking. Hazelhoff reasoned that the pursued object is always seen at the fovea and is “on” all the time. The flash-signal, on the other hand, reaches some position comparator with a head-centric position label that was valid one perception-time back for the moving eye. Metzger (Reference Metzger1932) showed that pursuit of the moving target is not essential in the Hazelhoff effect. If you fixate and the scale moves continuously, a flash will be shifted the same amount. Metzger criticized Hazelhoff's interpretation by pointing out that motion processing also has a latency, and came up with the latency-difference explanation that is still popular today. My aforementioned hypothesis, that motion provides the reference signal for position and time, means that I think that Metzger was wrong. Of course motion has a processing time, but that can simply be incorporated in the corresponding local sign. Now that study of the FLE has grown into a field with many stake-holders, subtle experimental paradigms, grossly incompatible explanations, and dozens of papers, it may be time to turn to the underlying question: What is the reference for positions in visual processing? My guess is: motion.
