In his Essay Concerning Human Understanding, John Locke takes up a question that had been posed to him by the Dublin politician and intellectual William Molyneux. Molyneux had asked Locke whether a man born blind who had learned to distinguish cubes from spheres by touch would, upon having his sight restored and being presented with the same objects, be able to match seen shape with felt. In the Essay, Locke agrees with Molyneux that the answer is “no”: the sensory ideas of sight and touch are intrinsically too different, they argued, for the connections between them to be uncovered except by building up associations between the different modalities. And with few exceptions – among them G.W. Leibniz (in the Nouveaux Essais) and a handful of more recent theorists (see Thomson 1974, Evans 1985, Campbell 1996, Noe 2004, Gallagher 2005) – the majority of those who have reflected on this question have taken something like this view.
Molyneux’s question is an empirical one, and so is resolvable in principle by experimental means. But this has proven to be more difficult than it might seem, as the handful of relevant studies have resulted in conflicting reports, and few if any have been carried out in a way that satisfies all the relevant methodological scruples. (For some discussion, see the afterword to Morgan 1977.) A paper by Richard Held and colleagues (available here, with supplementary information here) in the May 2011 issue of Nature Neuroscience, which purports to resolve the issue once and for all, is unfortunately no exception to this trend. Held et al. tested three congenitally blind individuals who had had their sight restored through the efforts of Project Prakash, a humanitarian organization that works with disabled children in India. The subjects were tested within 48 hours post-surgery, immediately after their bandages were removed, using stimuli consisting of several pairs of Lego blocks. Unsurprisingly, they were able to distinguish the objects by touch; somewhat more surprisingly, they were also able to distinguish them by sight as soon as their bandages were removed; but then in the crucial condition they were at or barely above chance when asked to match or distinguish seen objects and felt ones. They could, however, do this much successfully after a few days with the bandages off. The authors conclude that “the answer to Molyneux’s question is likely negative”:
The newly sighted subjects did not exhibit an immediate transfer of their tactile shape knowledge to the visual domain. This finding has important implications for bimodal perception. Whatever linkage between vision and touch may pre-exist concomitant exposure of both senses, it is insufficient for reconciling the identity of the separate sensory representations. However, this ability can apparently be acquired after short real-world experiences. (Held et al. 2011, p. 552)
So it is just as Locke and the rest of the classic empiricists predicted: the newly-sighted subjects could see the stimuli well enough to distinguish matching pairs from non-matching ones, but were unable to match visual percepts with haptic ones prior to the opportunity to build up a pattern of associations.
The trouble, however, is that success in this experimental paradigm demanded more than a mere haptic-visual match. For sight and touch differ not just in their phenomenal character, but also in the way that spatial information is made available in each (the “point of view”, as it were): perceiving the shape of an object through touch involves moving one’s hand or hands all around its surface, but in this study visible object presentation involved displaying only a single side of the stimulus object, from a distance of 12 inches away. (The authors report that subjects were allowed “to adjust their distance or viewpoint” so long as they remain seated, but they don’t say how often subjects did this – and the anecdotal data on newly-sighted subjects (e.g. Sacks 1995) suggests that they are unlikely to have done this very much.) To succeed in matching seen objects with felt ones (as opposed to simply discriminating between visible stimuli on the basis of their facing sides), the subjects had therefore to construct a perspective-invariant three-dimensional representation of the seen objects on the basis of this extremely limited visual information. While visually normal subjects have no trouble constructing perceptual representations of this sort, it is no surprise to find that those who have been blind from birth cannot.
Indeed, in another study carried out under the auspices of Project Prakash, Ostrovsky et al. (2009) found that even several weeks or months after the restoration of vision, subjects presented with minimally complex visual stimuli (such as pictures of a cube with three facing sides or images of common objects) were entirely unable to say even how many objects they were viewing. As the authors write, “subjects’ responses were driven by low-level image attributes; [when asked to point to objects] they pointed to regions of different hues and luminances as distinct objects. This approach greatly oversegmented the images and partitioned them into meaningless regions, which would be unstable across different views and uninformative regarding object identity. A robust object representation is difficult to construct on the basis of such fragments” (Ostrovsky et al. 2009, p. 1487). (Unfortunately for our purposes, subjects were not asked simply to distinguish between object pairs.) How unsurprising, then, that individuals at an even earlier stage post-surgery might have been unable to construct such robust representations for complex three-dimensional shapes!
Put slightly differently, the key point is that Molyneux’s question is not the question whether a person born blind will be able to see as soon as his or her eyes are opened (as it were). Rather, the question is whether, given that a person has had his or her sight restored well enough to make out what is in front of him or her, those intact visual percepts will have spatial content in common with perceptions of touch. This is why Held et al. took the time to see whether their subjects could differentiate between pairs of stimuli presented just to vision; but as Gareth Evans (1985, p. 380) once noted, this alone is not enough to establish visual figure perception – especially of a complex three-dimensional shape.
How might an experiment be conducted that could answer Molyneux’s question less ambiguously? Here are two suggestions. First, in the study cited above Ostrovsky et al. report that newly-sighted individuals’ judgments of object identity were more accurate when stimuli were moving, as the additional visual information provided by object motion helped them “link together parts of an object and segregate them from the background” (p. 1488). So it would be interesting to test haptic-visual matching in newly-sighted subjects when visual stimuli are made to rotate and move around. Secondly and more straightforwardly, we might just opt for two-dimensional stimuli in place of three-dimensional ones: Ostrovsky et al. report that subjects were able to count objects successfully when the stimuli consisted of non-overlapping line drawings, and several authors have reported that newly-sighted subjects could recognize two-dimensional stimuli while being unable to identify three-dimensional objects by sight.* A better question to ask, then, might be: Supposing a man born blind, and taught by his touch to distinguish between a circle, and a square impressed in the same material, and nighly of the same bigness, so as to tell, when he felt one and t’other, which is the square, and which the circle; and supposing now the square and the circle placed on a table, and the blind man to be made to see; quaere, whether by his sight, before he touched them, he could now distinguish, and tell, which is the circle, which the
* For example: Sacks’s patient Virgil could read some letters from the Snellen eye chart, though he could not identify the shapes of wooden blocks or even simple squares and circles (Sacks 1995, pp. 115, 126). Sacks (pp. 122-123) also notes that Valvo’s patient H.S. could identify single letters but not whole words, and that Gregory’s patient S.B. – who when blind had told time by feeling the hands of a watch with no glass – could read the time via a clock on the wall. It is remarkable to me that Sacks of all people does not see how these kinds of differences should complicate our answers to Molyneux’s question.
Campbell, J. 1996. Molyneux’s question. Philosophical Issues 7, 301-318.
Evans, G. 1985. Molyneux’s question. In G. Evans, Collected Papers (New York: Oxford University Press), 364-399.
Gallagher, S. 2005.
Held, R., et al. 2011. The newly sighted fail to match seen with felt. Nature Neuroscience 14, 551-553.
Morgan, M.J. 1977. Molyneux’s Question: Vision, Touch, and the Philosophy of Perception. New York: Cambridge University Press.
Noe, A. 2004. Action in Perception. Cambridge: The MIT Press.
Ostrovsky, Y., et al. 2009. Visual parsing after recovery from blindness. Psychological Science 20, 1484-1491.
Sacks, O. 1995. To see or not to see. In Sacks, An Anthropologist on Mars (New York: A.A. Knopf), 108-152.
Thomson, J.J. 1974. Molyneux’s question. Journal of Philosophy 71, 637-650.