The flash-lag effect (FLE) is an empirical fact. Understanding the effect, however, requires us to build conceptual models that can range from the biophysical description of neuronal interactions to global cognitive schemes. Whichever model may be conceived, it should explain the FLE and related findings, rely on sound physiological grounds, adopt the smallest set of assumptions, and predict novel phenomena.

Motion extrapolation, as originally proposed by Romi Nijhawan (Reference Nijhawan1994), nicely explains the flash-lag phenomenon in its purest form. However, in order to accommodate that explanation with discordant empirical findings, Nijhawan, in addition to splitting perceptual extrapolation in two successive steps, now resorts to at least three different putative mechanisms: (1) the role of lateral communications between neighboring retinotopic locations, (2) the impact of visual transients generated by offsets, and (3) a dynamical interplay between motor and visual processes. Therefore, Nijhawan's effort to fit motion extrapolation into challenging empirical data has compelled him to adopt a piecewise mosaic of physiological functions. Also, even though Nijhawan does a good job in weaving the explanatory content of his account, he leaves aside a deeper exploration of its predictive potential.

Both empirical and theoretical evidences point to the ability of basic sensory operations to generate the FLE (Baldo & Caticha Reference Baldo and Caticha2005; Berry et al. Reference Berry, Brivanlou, Jordan and Meister1999; Erlhagen Reference Erlhagen2003). In fact, a simple neural network model built upon a small set of physiologically grounded assumptions was able to replicate the standard FLE (under several conditions), as well as its dependence on stimuli luminance, priming, trajectory, and predictability (Baldo & Caticha Reference Baldo and Caticha2005). This model also helps us evaluate the relative underlying involvement of spatial sources (lateral interactions between neighboring receptive fields) and temporal sources (time delays required to activate the neurons in the network or to broadcast signals among them).

However, despite their physiological appeal, simplistic models are still unable to encompass the phenomenological richness portrayed by this illusion, such as has been observed in the chromatic (Nijhawan Reference Nijhawan1997; Sheth et al. Reference Sheth, Nijhawan and Shimojo2000), spatial frequency (Sheth et al. Reference Sheth, Nijhawan and Shimojo2000), and auditory (Alais & Burr Reference Alais and Burr2003) domains. This wide-ranging manifestation of the FLE would call for the contribution of higher, more integrative, underlying mechanisms. Over the last few years, we have been refining a conceptual framework in which attention plays a role in contributing to the FLE. As pointed out by Nijhawan himself, “distances that separate the representations are large, so the information exchange between the representations is time consuming” (target article, sect. 9.1, para. 1). Actually, the first conceptual alternative challenging the motion extrapolation account was based on the time consumed by a unidirectional shift of (spatial) attention from the flashed location to the moving location (Baldo & Klein Reference Baldo and Klein1995). Meanwhile, whereas the role of attention in causing the FLE remained elusive, the modulation of the FLE by attentional factors was gradually being established (Baldo & Namba Reference Baldo and Namba2002; Baldo et al. Reference Baldo, Kihara, Namba and Klein2002; Chappell et al. Reference Chappell, Hine, Acworth and Hardwick2006; Namba & Baldo Reference Namba and Baldo2004; Sarich et al. Reference Sarich, Chappell and Burgess2007).

More recently, we have advanced our attentional framework by proposing that the FLE could originate from the time needed for attention to bind the flash and moving percept into a unitary whole (Baldo & Klein, in press). Starting with the detection of an abrupt event (a stationary flash or a beep, for instance), a shift of attention from a stationary object to a changing (“moving”) object has to occur in feature rather than physical space in order to bind them together into a unitary percept. This object-based attentional shift would require some time to be carried out, regardless of any spatial separation between both visual stimuli, and would naturally lead to a percept consisting of a changing stimulus further ahead along its “trajectory” (equivalently, we can also consider a non-directional attentional spread over the object, in feature space, which starts when an abrupt-onset stimulus is presented – though not necessarily from where it is presented).

It is easy to see that this feature-based “attentional” explanation is not in conflict with the findings concerning the flash-initiated and flash-terminated conditions, as claimed before in relation to the purely spatial attention shift (Alais & Burr Reference Alais and Burr2003; Khurana & Nijhawan Reference Khurana and Nijhawan1995): Whereas the time required for the attentional binding will lead to the FLE in the former condition, no FLE will be observed in the latter (in opposition to motion extrapolation's predictions), since the moving stimulus never reaches a position beyond that where it disappeared or stopped.

The relationship between delays required to bind a spatially extended object and time-consuming shifts (or spread) of attention has been extensively reported (Chappell et al. Reference Chappell, Hine, Acworth and Hardwick2006; Enns & Oriet Reference Enns and Oriet2004; Houtkamp et al. Reference Houtkamp, Spekreijse and Roelfsema2003; Kanai et al. Reference Kanai, Sheth and Shimojo2004; Roelfsema et al. Reference Roelfsema, Lamme and Spekreijse2000; Sarich et al. Reference Sarich, Chappell and Burgess2007). Because the scheme we propose involves an attentional binding in feature space, our account unifies empirical findings observed in a broad class of visual features (such as position, color, luminance, and spatial frequency), as well as in crossmodal phenomena (Alais & Burr Reference Alais and Burr2003). It is worth noting that only cognitive-oriented models are presently able to capture the entire perceptual spectrum revealed by the FLE. Nijhawan's motion extrapolation and the attentional binding account we presently offer are the main conceptual frameworks embodying this class of integrative models. Our proposal, however, is favored by a more parsimonious set of assumptions, a deeper physiological root, and a greater predictive power.

Eventually, cognitive models are to be brought down to the wiring structure and functional operations of neuronal lattices. It is conceivable that lateral interactions giving rise to spatio-temporal facilitations, as evidenced by simple networks (Baldo & Caticha Reference Baldo and Caticha2005) and also assumed by Nijhawan, could be further elaborated and extended to perceptual processing other than visual. A desirable connection between the fundamental rules of neural functioning and the large-scale expression of cognitive integration could thus be envisaged.

In conclusion, it is encouraging to recognize that the decade-long dispute over the underpinnings of the FLE has already contributed toward refining, discarding, or amalgamating fragmented and contradictory conceptions. Whatever the final consensus turns out to be, Nijhawan's legacy to the current debate on visual perception will endure.