Vision science – a field with obvious importance for the study of art – has engaged in debate between reductionists and holists over recent decades, wherein the former camp advocates the study of reduced and isolated visual stimuli such as bars and gratings, while the latter group advocates the study of naturalistic stimuli, such as natural scenes, that encompass many stimulus dimensions and replicate characteristic aspects of the natural world (Felsen & Dan Reference Felsen and Dan2005; Pinto et al. Reference Pinto, Cox and DiCarlo2008; Simoncelli & Olshausen Reference Simoncelli and Olshausen2001). This debate parallels the universalist-contextualist debate that animates Bullot and Reber's (B&R's) article, for indeed their contextualism is a variant of holism, albeit an especially radical one.

A number of features of the debate in vision science are illustrative. First, few if any scientists dismiss the viewpoint of the opposing side, as B&R do in relation to universalism. Reductionists have shown limitations in some holistic thinking, but have generally done so without rejecting it outright. Reductionists' chief complaint is that in using fully natural stimuli, we lose the ability to parametrically manipulate them – which is a problem also faced by the zealous contextualism of B&R. However, even ardent reductionists accept that the ultimate test of their theories is to see how they fare in natural settings (Rust & Movshon Reference Rust and Movshon2005).

But although holists have proven that reduced stimuli can lead to incomplete models of the visual system (Olshausen & Field Reference Olshausen and Field2004; Reference Olshausen and Field2005), they nevertheless accept fundamental reductionist claims. Holists would agree that to suggest that no “correct” knowledge can be gleaned without perfectly elaborated context is to deny that we can, in the vision science analogy, understand or predict any dimension of response to natural scenes using reductionist models. This is demonstrably not the case (David & Gallant Reference David and Gallant2005; David et al. Reference David, Vinje and Gallant2004).

Moreover, vision science has managed to synthesize reductionism and holism. This trend and parallel ones in other relevant areas of neuroscience (e.g., Lewicki Reference Lewicki2002) should serve as models for the psychological study of art. Rather than divide the field into “ahistorical psychologism” and its converse, historical philosophism, we should seek integrated approaches.

For example, measurement of reduced properties of naturalistic stimuli can grant novel and unexpected insights – with respect to vision and to art. The basic statistical properties of natural scenes such as spatial frequency spectrum characteristics have been shown to be regular, and this regularity influences mammalian vision via evolutionary demands for efficient neural coding (Field Reference Field1987; Reference Field1994). Regularity exists despite the common impression that natural scenes are limitlessly diverse – indeed, this naïve view went mostly unchallenged until the 1980s. However, we now know that natural scene regularities shape systems including retinal and cortical coding, object segmentation, attention, and so forth (see Geisler Reference Geisler2008).

Examining reduced aspects of art while retaining a degree of naturalism is likewise essential to scientific understanding of this unique and defining human trait. By measuring low-level statistical properties in samples of world artwork from many cultures and time periods, we find that art also has regularities. In particular, nearly all paintings, like natural scenes, show scale invariant (1/f) spatial statistics (Graham & Field Reference Graham and Field2007; Reference Graham and Field2008; Redies et al. Reference Redies, Hasenstein and Denzler2007) – again, despite apparent heterogeneity. This means artist output is constrained by evolved aspects of the visual system: images lacking such regularities (e.g., very blurry images, or random, white noise images) are difficult for the system to process, because of its evolved coding strategies. Such images are in a way imperceptible. No artist or movement would last long making only, for example, white noise images, because they would be indistinguishable – even though there are far more possible white noise images than there are particles in the universe (Graham & Field Reference Graham, Field and Squire2009). Thus, certain types of art are a priori unlikely to be made or appreciated. Such fundamental knowledge is revealed without reference to historical context, but does derive from the study of basic, shared properties in natural exemplars and – crucially – from consideration of their relation to the brain.

Moreover, if we defer to historical context – to the exclusion of reductionist empiricism – we can come to mistaken conclusions. Consider Jackson Pollock: we know from historical documentation that Pollock's paintings were created using drip techniques that employed significant randomness. Indeed, what made his art so avant-garde – even compared to earlier automatist art – was precisely this randomness (Chave Reference Chave and Karmel1999). Though Pollock retained a degree of deliberate design, the randomness of his art is today seen as essential to the appreciation of his work, as B&R note. Thus, taking the stance of historical philosophism, we might conclude that such paintings prove our visual system can appreciate random patterns so long as we comprehend the appropriate context.

However, when we examine Pollock independently of “causal data” and historical context, and instead test his work with respect to basic properties relevant to human vision, we see that in fact Pollock's paintings are not truly – or even approximately – random. They show robust scale invariant spatial statistics, which are mostly indistinguishable from those of natural scenes, representational art, and nonrepresentational art (Graham & Field Reference Graham and Field2008). Pollock thus shares fundamental properties with other art styles, which are in turn shaped by visual coding. We can even suppose that if they were truly random, his paintings would not have been appreciated – neither in his time nor ours. This gives us a rather different perspective on the appreciation of Pollock's work.

B&R's arguments can be challenged on their own philosophical terms as well: for example, which experts are we to trust with regard to “correct” context, and when do we declare such stories unassailable? Rigid contextualism invariably leads to revisionism: because the “relevant facts” change with greater perspective – consider that Pollock was dismissed as an unserious showboat in his time by serious critics and artists – we often cannot appreciate context until we have created mythology, which is surely anathema to B&R's demand for historical accuracy.

B&R's strain of utopian philosophy is of little relevance in the empirical sciences. Yet accounting for naturalism is surely warranted – in the scientific study of art, as in vision science. The solution in both fields is to integrate holistic and reductionist approaches.