Visual masking and the dynamics of human perception, cognition, and consciousness A century of progress, a contemporary synthesis, and future directions

The 1990s, the “decade of the brain,” witnessed major advances in the study of visual perception, cognition, and consciousness. Impressive techniques in neurophysiology, neuroanatomy, neuropsychology, electrophysiology, psychophysics and brain-imaging were developed to address how the nervous system transforms and represents visual inputs. Many of these advances have dealt with the steady-state properties of processing. To complement this “steady-state approach,” more recent research emphasized the importance of dynamic aspects of visual processing. Visual masking has been a paradigm of choice for more than a century when it comes to the study of dynamic vision. A recent workshop (http://lpsy.epfl.ch/VMworkshop/), held in Delmenhorst, Germany, brought together an international group of researchers to present state-of-the-art research on dynamic visual processing with a focus on visual masking. This special issue presents peer-reviewed contributions by the workshop participants and provides a contemporary synthesis of how visual masking can inform the dynamics of human perception, cognition, and consciousness.

Correspondingly, single cell recordings and brain imaging techniques combined with visual masking have provided new insights about which brain areas are involved in conscious and unconscious vision (Dehaene, Naccache, Le Clec' H, Koechlin, Mueller, Dehaene-Lambertz et al., 1998;Macknik, this volume;Pinel et al., 2001;Rolls & Tovée, 1994). This and related work shows that even during the processing of unconscious inputs, large networks can be recruited. Thus, both psychological experiments and brain research on visual masking suggest that unconscious vision plays an important role in human cognition and can be studied in a rigorous way.
Visual processing. Masking has been used to study detailed properties of the visual system itself. The application of masking to visual processing encompasses a broad range of areas including the perception of contour (Francis, this volume), motion (Öğmen, this volume; Otto, this volume), colour, pattern (Herzog, this volume), stimulus brightness (Rudd, this volume), and spatial location (Breitmeyer, this volume).
Given the strong interest in masking and the frequency of its use as a tool for investigating perceptual, cognitive, and neurophysiological systems, it is perhaps surprising to note that there is currently no generally agreed-upon theory of the mechanisms that are involved in producing masking effects. Researchers who use masking as a tool to explore other issues generally have the implicit theory that the mask interrupts processing or interferes with detection of the target properties. However, these ideas are generally not rigorously investigated (usually because the researcher is actually interested in something other than masking per se).
Likewise, even though masking effects have been studied for nearly a century, there remains much debate about which properties of masking are fundamental and which properties reflect parametric variations of common mechanisms. For example, many older studies asked subjects to report the perceived brightness of a target stimulus, while modern studies tend to ask subjects to make some kind of discrimination of a target. While such changes in criterion content are known to produce quantitative changes in the data, whether these quantitative changes reflect fundamental differences in the underlying processes remains largely unknown.
http://www.ac-psych.org Correspondingly, the study and use of masking is often hampered by variability of methods, such as divergent stimuli and experimental tasks that are used by different laboratories. Without time-consuming replications and comparisons among these differences, nobody can tell with certainty whether the procedural differences matter.

THe WorksHop GoAls
The goal of the proposed workshop was to bring together researchers who are interested in understanding the mechanisms that produce the many different types and effects of masking. Through intense interaction between groups and researchers, we hoped to gain new insights into current research and identify ideas and methods that would lead to an improved understanding of the role and mechanisms of masking effects at both the behavioural and neurophysiological levels. Because masking plays an integral part in the study of many aspects of cognition, the outcome of the meeting promised to provide new insights into many different areas of human cognition, especially studies of consciousness. To achieve these goals, we invited speakers from a variety of backgrounds.
In addition to sharing state-of-the-art research, we asked the participants to make connections across different domains, identify a framework for discussing visual masking and related topics, raise general questions about the topic, and to promote theoretical speculations. We wanted to look for possible connections and big unanswered questions, and to think about what types of research would really change how researchers consider or use masking techniques, dynamic vision, and consciousness. Hence, the current proceeding was aimed to be a comprehensive and controversial overview of the field rather than a collection of loosely related empirical results. Despite these controversies, it was generally acknowledged that metacontrast masking provides an advantageous experimental procedure for the study of the neural correlates of consciousness with fMRI because the (u-shaped) masking function is able to rule out a variety of candidate brain areas. In essence, one can scan the brain for u-shaped activity functions, because the activity marker of conscious perception (reflected in the masking function) is correlated in a predictable but non-linear (and, hence, nontrivial) manner with a specific independent variable (SOA,

identify future ways to study consciousness
Similar to the development of neurophysiological studies, the study of consciousness has dramatically improved in recent years with new techniques and theories. The properties of masking have contributed substantially to these developments, but perhaps have not been used to their full potential. The meeting attendees were asked to consider how masking effects can best be used to study issues of consciousness in the future. The most exhaustive procedure of residual conscious prime perception uses the same stimuli, responses, and response mappings in the prime discrimination task as in the target response task, but asks participants to discriminate the masked prime in the prime discrimination task instead of responding to the visible target. Chance performance (d-prime equal to zero) in such an exhaustive prime discrimination task has become somewhat of a standard prerequisite for establishing unconscious vision (e.g., Kiefer, this volume). There were also lively discussions about phenomenal experiences (i.e., qualia, the way things appear in consciousness). Breitmeyer noted that phenomonologically a target may be invisible, but still produce a high d-prime measure. An example of this is in feature inheritance effects where the target is invisible, but some of its features are visible in the mask (Herzog, this volume; Otto, this volume; for modelling, see Hamker, this volume). Ulrich Ansorge pointed out that these types of phenomena link some masking paradigms to mask priming effects. A particularly important phenomenological experience is that of time itself.
In some situations properties of a target are modified in perceived time and order as well as in spatial appearance (cf. Scharlau, this volume).
Finally, it was noted that there is a good chance that conclusions from different experimental masking paradigms and across studies of healthy participants and neuropsychologically impaired subjects could be directly compared with controlled procedures (Breitmeyer, this volume).

Model development
To Öğmen, this volume), feedforward models (VanRullen, this volume) and processing models that hypothesized interactions but did not include quantitative calculations (Bachmann, this volume; Enns & Oriett, this volume).
Many of the quantitative models can be compared to each other, but it is less clear how to compare the quantitative and non-quantitative models.
Greg Francis (this volume) discussed experimental data that are evidence against a wide variety of quantitative models. He argued that a fundamental flaw of all current models is that they one-sidedly focus on the temporal dynamics and do not appropriately take into account processing of specific spatial aspects of test and masking stimuli, such as grouping phenomena. Herzog (this volume) arrives at the very same conclusion from empirical grounds. Pattern as well as metacontrast masking seems to involve complex spatial interactions that may be best explained with spatial perceptual grouping. Still, Bruce Bridgeman (this volume) and Hermens and Ernst (this volume) showed that simple neural network models can simulate many masking conclUsions In conclusion, the workshop generated a lively discussion and exchange of ideas. The articles in this special issue provide an extended exposition of these contributions. A century of research provides a wealth of information about visual masking, yet we acknowledge that our understanding of masking remains limited.
The situation is not very different from other areas of vision, such as form perception. One can draw a "bad news / good news" lesson from this comparison. The bad news is, of course, that our current knowledge does not lead to a simple set of laws or rules that can provide a general understanding of visual masking. The good news is that this failure appears to stem from the fact that masking is not a relatively isolated peculiarity of vision but instead is a complex phenomenon with important implications for many areas of vision science. It involves an extremely broad coverage of visual phenomena including surface, depth, and contour processing, perceptual grouping, attention, contextual effects, awareness, and priming. It has been used to understand many properties of both normal and abnormal visual function. Thus, we expect that our understanding of masking will progress hand in hand with other aspects of visual science with reciprocal and synergetic contributions.
After the workshop, there were a series of further discussions that were carried out through e-mail communications. This included an interesting challenge to the community to predict masking effects. We have archived these discussions (along with some photos of the workshop) at http://lpsy.epfl.ch/VMworkshop/ under the Follow-ups section.