Psychophysical Investigations Into The Neural Basis Of Synaesthesia (2001) PDF

doi 10.1098/rspb.2000.1576 Psychophysical investigations into the neural basis of synaesthesia...

doi 10.1098/rspb.2000.1576 Psychophysical investigations into the neural basis of synaesthesia V. S. Ramachandran* and E. M. Hubbard Center for Brain and Cognition, University of California, San Diego, 9500 Gilman Drive, LaJolla, CA 92093- 0109, USA We studied two otherwise normal, synaesthetic subjects who `saw' a speci¢c colour every time they saw a speci¢c number or letter. We conducted four experiments in order to show that this was a genuine perceptual experience rather than merely a memory association. (i) The synaesthetically induced colours could lead to perceptual grouping, even though the inducing numerals or letters did not. (ii) Synaesthetically induced colours were not experienced if the graphemes were presented peripherally. (iii) Roman numerals were ine¡ective: the actual number grapheme was required. (iv) If two graphemes were alternated the induced colours were also seen in alternation. However, colours were no longer Downloaded from https://royalsocietypublishing.org/ on 07 November 2024 experienced if the graphemes were alternated at more than 4 Hz. We propose that grapheme colour synaesthesia arises from `cross-wiring' between the `colour centre' (area V4 or V8) and the `number area', both of which lie in the fusiform gyrus. We also suggest a similar explanation for the representation of metaphors in the brain: hence, the higher incidence of synaesthesia among artists and poets. Keywords: synaesthesia; multisensory; extrastriate visual areas; psychophysics that the visual number grapheme rather than the numer- 1. INTRODUCTION ical concept is required for evoking colours. However, they The bizarre phenomenon of synaesthesia, e.g. `seeing both noted that if they `imaged' the corresponding Arabic sounds' or seeing speci¢c colours upon seeing speci¢c numeral then the corresponding colour was evoked, but numerals (Galton 1880, 1883; Cytowic 1989; Paulesu et al. more faintly than when looking at the actual numerals. 1995; Gray et al. 1997; Harrison & Baron-Cohen 1997), was ¢rst clearly documented by Galton (1880). Galton 2. METHODS AND RESULTS (1880) also noted that it tends to run in families. We now report on several new psychophysical experiments for In order to show conclusively that the evoked colours were exploring the e¡ect and determining its neural locus. We indeed perceptual rather than memories, we asked whether studied two `grapheme-colour' synaesthetes who were synaesthetically induced colours could in£uence perceptual otherwise completely normal. They saw speci¢c colours grouping. Perceptual grouping is a convenient diagnostic test for when looking at speci¢c letters or numerals (subject J.C. determining whether a given feature is genuinely perceptual or experienced colours with both letters and numerals, while not (Beck 1966; Treisman & Gelade 1980; Julesz 1981) (e.g. tilted subject E.R. experienced colours for numerals only). Our lines will segregate and group separately from vertical lines but ¢rst questions were as follows. Is synaesthesia a genuine Ts will not segregate from Ls). In order to explore whether sensory phenomenon? Or is the phenomenon just a synaesthetically induced colours are able to a¡ect grouping, we memory association from early childhood (e.g. from created a 7 5 matrix of 35 graphemes (see ¢gure 1a). The prob- books with coloured numbers)? When some synaesthetes ability of seeing vertical columns or horizontal rows in such a speak of chicken tasting `pointy' are they merely being matrix is ca. 50%. However, by choosing the graphemes care- metaphorical (as in when we say `cheddar cheese is sharp' fully we can bias normal subjects to group on the basis of simi- or that `he wears loud shirts')? larity of shape. Taking advantage of the fact that the Anecdotally, we were convinced that the e¡ect in our synaesthetic subjects often saw the same (or very similar) colour two subjects was a sensory one. First, the subjects said they in two di¡erent graphemes, we chose graphemes for each `saw' the colours spatially in the same location as the synaesthetic subject such that grouping by colour, for example, grapheme and said `it is not just memory'. Second, when would lead to vertical organization whereas grouping by simi- shown half-tone black and white photographs of fruits and larity of shape would lead to a horizontally biased organization vegetables (e.g. a cucumber, tomato, orange and banana) as in normals (compare ¢gure 1b). they readily identi¢ed the fruit, but said that no colours We tested two synaesthetes (J.C. and E.R.) and a total of 20 were evoked (e.g. when looking at the banana, `it reminds control subjects (ten for each synaesthete). Displays were me of yellow, but I do not actually see yellow the way I see constructed for each synaesthete so that each would evoke yellow when you show me an F'). This argues against the similar colours in our synaesthetes. We tested each synaesthete memory hypothesis or at least the simple version of it with displays that would induce red and green or yellow and (although, as we shall see, the memory, metaphor and blue percepts. If synaesthetically induced colours are truly cross-wiring interpretations are not quite as incompatible perceptual and, therefore, are able to a¡ect grouping, we would with each other as they might at ¢rst seem). Furthermore, expect that synaesthetic subjects would be in£uenced by their neither subject experienced colours with Roman numerals evoked colours while controls would group solely on the basis of or when graphemes were traced on their palms, implying similarity of shape. The displays were presented in pseudo-random order such * Author for correspondence ([email protected]). that all possible combinations of horizontal and vertical Proc. R. Soc. Lond. B (2001) 268, 979^983 979 & 2001 The Royal Society Received 21 November 2000 Accepted 3 January 2001  980 V. S. Ramachandran and E. M. Hubbard Neural basis of synaesthesia (a) (b) 3 8 3 8 3 8 3 3 8 3 8 3 8 3 7 0 7 0 7 0 7 7 0 7 0 7 0 7 3 8 3 8 3 8 3 3 8 3 8 3 8 3 7 0 7 0 7 0 7 7 0 7 0 7 0 7 3 8 3 8 3 8 3 3 8 3 8 3 8 3 Downloaded from https://royalsocietypublishing.org/ on 07 November 2024 Figure 1. (a) Schematic of displays used to test whether synaesthetically induced colours are able to a¡ect grouping. Black graphemes (1.58  2.28) on a white background were presented until the subject responded. The spacing between graphemes was 4.08 horizontally and 3.48 vertically. The total display was 34.78  25.58. Control subjects consistently grouped this display horizontally (grouping the 3s with the 8s). (b) However, subject E.R. perceived 3s and 7s as red and 8s and 0s as green. She therefore grouped the display vertically, consistent with her synaesthetically induced colours. (a) grouping were presented once to each subject; this produced a 100 J.C. total of 144 trials. The subjects were not informed of the purpose control of the experiments. They were simply asked to indicate whether 80 the display appeared to group horizontally or vertically. In order to be certain that the subjects understood the instructions, 60 prior to testing they were initially presented with a simple dot 40 display in which there was an unambiguous grouping into rows or columns on the basis of proximity. grouping (percentage of trials) 20 Both synaesthetes grouped with their induced colours while control subjects generally grouped on the basis of shape. Subject 0 J.C. consistently grouped the displays with his induced colours red–green red–green blue–yellow blue–yellow in 90.97% of trials, while control subjects showed a slight bias to colour shape colour shape grouping the displays on the basis of form for each of the two display types (red ^ green and blue ^ yellow) (see ¢gure 2a). (b) 100 Because of di¡erences in the graphemes selected, the di¡erence E.R. between J.C.'s control population and E.R.'s was signi¢cant control 80 (F3,39 ˆ 4.71 and p 5 0.01), so we did not collapse over control populations. There were no signi¢cant di¡erences between the 60 conditions corresponding to the red ^ green and blue ^ yellow strategies among J.C.'s control population (t19 ˆ 1.8 and 40 p 4 0.05). We therefore present the results for each subject 20 pooled over the red ^ green and blue ^ yellow display types (see ¢gure 2 for individual comparisons). The di¡erence in grouping 0 direction between J.C. and control subjects was signi¢cant red–green red–green blue–yellow blue–yellow (t21 ˆ 4.46 and p 5 0.001). colour shape colour shape Subject E.R. consistently grouped the displays with her induced colours in 86.75% of trials, while control subjects were Figure 2. Individual comparisons for synaesthetes versus signi¢cantly biased to group in the opposite direction, grouping controls on the grouping task. The colours are those that on the basis of similarity of shape in 62.4% (t9 ˆ 3.29 and were experienced by our two synaesthetes ( J.C. and E.R.) p 5 0.01) and 63.9% (t9 ˆ 4.00 and p 5 0.01) of trials (see when presented with our grouping displays. The scores are ¢gure 2b). Again, there were no signi¢cant di¡erences between presented as the percentages of trials in which subjects the red ^ green and blue ^ yellow displays for E.R.'s control popu- grouped either on the basis of induced colours or on the lation (t19 ˆ 0.3 and p ˆ 0.77), so we collapsed over display type. basis of similarity of shape. White bars indicate control The di¡erence between E.R. and the controls was signi¢cant subject performance while shaded bars indicate synaesthete (t21 ˆ6.09 and p 5 0.001). Our results therefore suggest that the performance. The error bars indicate the standard error of the synaesthetically induced colours are genuinely perceptual rather mean. No standard error is reported for the synaesthetes since they were unique observations. These comparisons were than being merely associative. individually signi¢cant for J.C. (red^green, t9 ˆ 4.25 and In our second experiment, we used stimuli analogous to the p 5 0.01 and blue^yellow, t9 ˆ 2.52 and p 5 0.05) and for Ishihara (1997) pseudo-isochromatic tests for colour vision. As a E.R. (red^green, t9 ˆ 5.14 and p 5 0.001 and blue^yellow, preliminary demonstration, we had a matrix of small, black 5s t9 ˆ 3.37 and p 5 0.01). See the text for pooled results. scattered randomly on a white screen. Embedded among these Proc. R. Soc. Lond. B (2001)  Neural basis of synaesthesia V. S. Ramachandran and E. M. Hubbard 981 Figure 4. Schematic showing that cross-wiring between area V4 and the number-grapheme area in the fusiform gyrus Downloaded from https://royalsocietypublishing.org/ on 07 November 2024 Figure 3. A matrix of Hs forming a triangle shape embedded might be the neural basis of grapheme-colour synaesthesia. in a background of distractors. Normals ¢nd it di¤cult to Area V4 is shown in red and the number-grapheme area is detect the triangle. Synaesthete J.C. did so easily (for him the shown in green (after Rickard et al. 2000). Area V4 was Hs were seen as green, the background Fs were seen as yellow de¢ned in one of the authors (E.M.H.) using standard and the Ps were seen as red). Figures were presented in blocks functional magnetic resonance imaging methods for de¢ning of eight trials with the same grapheme used to create the retinotopic visual areas (Engel et al. 1994, 1997); the location embedded ¢gure and two other graphemes used to create the of the number area was estimated from the results of Rickard remainder of the matrix throughout each block. The display, et al. (2000) (average coordinates for left hemisphere, 29.5, which consisted of 44^48 graphemes (each 0.680.88), did not 54.8 and 12.0). make use of the full screen, but instead used the central 13.28 10.08 area, which is indicated by a black outline. The four shapes they had seen via a keyboard button press (see the subjects were told which grapheme would compose the legend to ¢gure 3 for details). The two synaesthetes were signi¢- embedded form at the beginning of each block. Embedded cantly better (mean per cent correct 81.25%) (t40 ˆ 2.07 and forms, which were composed of six to eight graphemes, were p 5 0.05) than the controls. Performance for the control subjects presented in randomly selected locations near the centre of the screen and were ca. 6^78 wide and 4^58 high. Rectangles were was better than chance (59.4%), indicating that they understood clearly di¡erent from squares, with the width of each the task, but found it di¤cult to ¢nd the embedded shape. Since rectangle being twice its height. Each subject was given two synaesthetes were better and signi¢cantly so than normals, we practice trials prior to the ¢rst trial, one in which the display conclude that the phenomenon is genuinely perceptual and not remained on screen until a keyboard press indicated that the confabulatory in origin, nor based on memory associations. subject was ready to continue and one timed presentation in Next, we had our two synaesthetes (J.C. and E.R.) look at a order to familiarize the subjects with the procedure in each ¢xation spot in the centre of the screen while we moved the trial. The block presentation order was counterbalanced target grapheme (e.g. a 5 or an E) gradually outwards. To our across subjects. Twenty normal controls were tested for each surprise, the subjects no longer saw the numbers as coloured at a synaesthete in order to determine baseline performance and critical distance! We repeated the experiment for both subjects assess potential di¡erences in task di¤culty. Because the after having enlarged the numbers in order to scale for eccentri- graphemes used to create the embedded ¢gure were tailored separately for each synaesthete, two-tailed t-tests were used city (Anstis 1998). Graphemes that elicited red, green, blue or for determining whether group di¡erences existed in the two yellow were presented for 1s while subjects ¢xated a central ¢xa- tasks. The groups were not signi¢cantly di¡erent from each tion cross. This presentation time was chosen in order to mini- other (t39 ˆ 0.88 and p ˆ 0.39). We therefore pooled the results mize the possibility of eye movements while allowing the of the 40 control subjects. synaesthetic subjects enough time to experience their colours (see below). The subjects indicated which number was presented was a set of black 2s (mirror images of 5s) forming a global and then indicated what, if any, colour they experienced. One of shape such as a square or a triangle. Normal subjects cannot four graphemes was presented at one of three eccentricities: 5.5, segregate the 2s from the 5s since they are too similar (Beck 10.9 and 12.18. The graphemes were scaled for eccentricity and 1966; Treisman & Gelade 1980; Julesz 1981) but we found that were 3.18 2.18 (height  width) when presented near ¢xation, the synaesthetic subjects could do so easily and discern the while at the periphery they were 4.78 3.08 (height  width). global shape. Both subjects identi¢ed all graphemes perfectly, even at 12.18 We conducted a more formal experiment along these lines on eccentricity. Subject E.R. experienced no colours when numerals our two synaesthetes (J.C. and E.R.) and 40 normal controls (20 were presented beyond 118 eccentricity in either visual ¢eld for each synaesthete). One of four di¡erent shapes (a square, while subject J.C. experienced no colours when graphemes were rectangle, triangle or diamond) composed of small graphemes presented beyond 118, but only for graphemes presented in his was embedded in a matrix of other small graphemes (¢gure 3). left visual ¢eld. This observation also argues strongly against The graphemes were again chosen so that each synaesthete would metaphor or memory. Why would the grapheme fail to evoke a experience either red ^ green or blue ^ yellow pairs (e.g. the ¢gure memory of a colour when it is still clearly visible? Memories would appear red to each synaesthete while the background ordinarily show positional invariance. would appear green or vice versa). Each stimulus matrix was We propose instead that synaesthesia arises from cross-wiring presented for 1s after which subjects would indicate which of the between adjacent brain maps in a manner analogous to the Proc. R. Soc. Lond. B (2001)  982 V. S. Ramachandran and E. M. Hubbard Neural basis of synaesthesia `remapping' of face to hand that occurs in area S1 following arm This observation again supports the cross-wiring hypoth- amputation (Ramachandran et al. 1992; Ramachandran & esis rather than the memory association hypothesis (you Rogers-Ramachandran 1995; Ramachandran & Hirstein 1998). cannot `remember' a colour you have never seen!). Further Such cross-wiring could occur between colour areas V4 (Lueck study of this subject may throw light on the old philoso- et al. 1989; Zeki & Marini 1998) or V8 (Hadjikhani et al. 1998), phical conundrum raised by Moyleneux (see Locke 1690) which are located in the fusiform gyrus and the `visual number as to whether someone can experience completely novel grapheme' area, which is directly adjacent to it in the same `qualia' never before experienced. gyrus (Pesenti et al. 2000; Rickard et al. 2000) (see ¢gure 4). We Synaesthesia has too often in the past been dismissed as are currently exploring this possibility using functional childhood memories or `mere metaphor', which is an magnetic resonance imaging. Given that area V4 emphasizes example of the classic fallacy of trying to explain an central vision (Gattass et al. 1988), the cross-wiring might also enigma (synaesthesia) with a mystery (metaphor). Since be expected to a¡ect central vision predominantly, even though our understanding of the neural representation of meta- the representation of area V4 extends over 35^408 of the phor is still in its infancy, explaining synaesthesia as mere contralateral visual ¢eld (Gattass et al. 1988). Consistent with metaphor is unlikely to be a fruitful strategy. Indeed, our this hypothesis, we found that, when presented with real colours results suggest that we can turn the problem on its head in the periphery, both synaesthetes were able to identify the and argue that understanding synaesthesia (a concrete Downloaded from https://royalsocietypublishing.org/ on 07 November 2024 colours correctly in 100% of trials. perceptual e¡ect the anatomical locus of which can be We then explored the temporal dynamics of the evoked potentially pinned down) can provide an experimental colours. We presented J.C. and E.R. with a black grapheme lever for understanding the neural basis of metaphors. (visual angle 2.581.88) at ¢xation and alternated it with a Metaphor involves linking one conceptual map (e.g. taste second grapheme at speeds varying from 1 to 20 Hz. At low of cheese) with another seemingly unrelated one (e.g. speeds (less than 4 Hz), the subjects experienced their colours tactile; `sharp' or `£at'), perhaps as a way of economizing alternating with the graphemes. The number form could still be on computational burden (Lako¡ 1987). In addition, it seen alternating at speeds higher than 5 Hz (up to ca. 10 Hz), but has been shown (Ullman 1945; Williams 1976) that the colours were no longer experienced. sensory adjectives undergo systematic shifts from one sensory domain to another (e.g. loud colours or bitter cold). 3. DISCUSSION Perhaps a genetically based excess of connections in Taken collectively, these results strongly imply that some individuals leads to both synaesthesia and a propen- synaesthesia is a genuine perceptual e¡ect, which is sity for metaphor and, hence, the higher incidence of possibly caused by cross-wiring in speci¢c brain areas. synaesthesia among artists, poets and novelists (Root- Since the phenomenon runs in families, it is tempting to Bernstein & Root-Bernstein 1999). It has not escaped our speculate that a mutation (perhaps X-linked; see Bailey notice that this hypothesis may also explain why we ¢nd & Johnson 1997) causes excessive proliferation (or defec- certain tastes and smells `disgusting' and scrunch up our tive pruning) of neural connections between adjacent noses (Darwin 1872) but also speak of someone being brain maps, e.g. between area V4 and the number area in morally `disgusting' and make the same face (e.g. if a the fusiform gyrus (our two synaesthetic subjects both drunk makes an unwanted sexual pass at a woman). Is it had parents who were synaesthetic). If so, one can go entirely a coincidence that smell and taste `maps' are in from a single gene to a speci¢c brain area to detailed the orbitofrontal frontal lobes, the same place where psychophysics all in a single `preparation' ö a strategy `maps' for moral disgust might lie (Lane et al. 1997; that has been tried with modest success in Drosophila (e.g. Northo¡ et al. 2000)? Cutforth & Gaul 1997) but not in humans. It is common knowledge that there is a great amount of heterogeneity We thank Francis Crick, Geo¡rey Boynton, Patricia among synaesthetes (which is perhaps caused by cross- Churchland, Eric Altschuler, Diane Rogers-Ramachandran wiring at di¡erent stages) and it remains to be seen how and Julia Fuller Kindy for discussions and John Wixted for general the results reported here are. Regions concerned statistical consultation. This research was supported by grant with more abstract numerical concepts and more sophisti- NIH MH 60474 from the National Institutes of Health. cated colour processing both lie near the angular gyrus. If cross-wiring occurred at this later stage, then the numerical concept itself might be expected to evoke REFERENCES speci¢c colours. Anstis, S. 1998 Picturing peripheral acuity. Perception 27, 817^825. We have recently begun testing a third synaesthete (S.S.) Bailey, M. E. S. & Johnson, K. S. 1997 Synaesthesia: is a genetic who experiences colours evoked by faces (both photo- analysis feasible? In Synaesthesia: classic and contemporary readings graphs and cartoon faces), the saturation and hue of which (ed. S. Baron-Cohen & J. E. Harrison), pp. 182^207. Oxford, vary monotonically with facial expression (e.g. happy UK: Blackwell Publishers, Inc. evokes red while sad evokes green). This would be consis- Beck, J. 1966 E¡ect of orientation and of shape similarity on tent with our speculations on cross-wiring in the fusiform perceptual grouping. Percept. Psychophys. 1, 300^302. Cutforth, T. & Gaul, U. 1997 The genetics of visual system gyrus, which is also known to contain cells that respond to development in Drosophila: speci¢cation, connectivity and faces (Rolls & Tovee 1995; Kanwisher et al. 1997). Intrigu- asymmetry. Curr. Opin. Neurobiol. 7, 48^54. ingly, this subject is colour anomalous (s-cone weakness), Cytowic, R. E. 1989 Synaesthesia: a union of the senses. New York: but he sees `weird' number-evoked colours that are `extra- Springer. spectral' for him (i.e. he cannot see them in the real world, Darwin, C. 1872 The expression of the emotions in man and animals. but experiences them in response to graphemes or faces). London: John Murray. Proc. R. Soc. Lond. B (2001)  Neural basis of synaesthesia V. S. Ramachandran and E. M. Hubbard 983 Engel, S. A., Rumelhart, D. E., Wandell, B. A. & Lee, A. T. Frackowiak, R. S. 1989 The colour centre in the cerebral 1994 FMRI of human visual cortex. Nature 369, 525. cortex of man. Nature 340, 386^389. Engel, S. A., Zhang, X. & Wandell, B. A. 1997 Colour tuning in Northo¡, G. (and 16 others) 2000 Functional dissociation human visual cortex measured with functional magnetic reso- between medial and lateral prefrontal cortical spatiotemporal nance imaging. Nature 388, 68^71. activation in negative and positive emotions: a combined Galton, F. 1880 Visualised numerals. Nature 21, 494^495. fMRI/MEG study. Cerebr. Cortex 10, 93^107. Galton, F. 1883 Inquiries into human faculty and its development. Paulesu, E., Harrison, J., Baron-Cohen, S., Watson, J. D. G., London: Dent & Sons. Goldsetein, L., Heather, J., Frankowiak, R. S. J. & Frith, Gattass, R., Sousa, A. P. B. & Gross, C. G. 1988 Visuotopic C. D. 1995 The physiology of coloured hearing: a PET activa- organization and extent of V3 and V4 of the macaque. J. tion study of colour-word synaesthesia. Brain 118, 661^676. Neurosci. 8, 1831^1845. Pesenti, M., Thioux, M., Seron, X. & De Volder, A. 2000 Gray, J. A., Williams, S. C. R., Nunn, J. & Baron-Cohen, S. Neuroanatomical substrates of Arabic number processing, 1997 Possible implications of synaesthesia for the hard ques- numerical comparison, and simple addition: a PET study. J. tion of consciousness. In Synaesthesia: classic and contemporary Cogn. Neurosci. 12, 461^479. readings (ed. S. Baron-Cohen & J. E. Harrison), pp. 173^181. Ramachandran, V. S. & Hirstein, W. 1998 The perception of Oxford, UK: Blackwell Publishers, Inc. phantom limbs: the D. O. Hebb lecture. Brain 121, 1603^1630. Hadjikhani, N., Liu, A. K., Dale, A. M., Cavanagh, P. & Ramachandran, V. S. & Rogers-Ramachandran, D. 1995 Tootell, R. B. H. 1998 Retinotopy and colour sensitivity in Touching the phantom. Nature 377, 489^490. Downloaded from https://royalsocietypublishing.org/ on 07 November 2024 human visual cortical area V8. Nature Neurosci. 1, 235^241. Ramachandran, V. S., Rogers-Ramachandran, D. & Stewart, Harrison, J. E. & Baron-Cohen, S. 1997 Synaesthesia: a review M. 1992 Perceptual correlates of massive cortical reorganiza- of psychological theories. In Synaesthesia: classic and contem- tion. Science 258,1159^1160. porary readings (ed. S. Baron-Cohen & J. E. Harrison), Rickard, T. C., Romero, S. G., Basso, G., Wharton, C., Flitman, pp. 109^122. Oxford, UK: Blackwell Publishers, Inc. S. & Grafman, J. 2000 The calculating brain: an fMRI study. Ishihara, S. 1997 Ishihara's tests for colour de¢ciency. Tokyo: Neuropsychologia 38, 325^335. Kanehara & Co., Ltd. Rolls, E. T. & Tovee, M. J. 1995 Sparseness of the neuronal Julesz, B. 1981 Textons, the elements of texture perception, and representation of stimuli in the primate temporal visual their interactions. Nature 290, 91^97. cortex. J. Neurophysiol. 73, 713^726. Kanwisher, N., McDermott, J. & Chun, M. M. 1997 The fusi- Root-Bernstein, R. & Root-Bernstein, M. 1999 Sparks of genius: form face area: a module in human extrastriate cortex the thirteen thinking tools of the world's most creative people. Boston: specialized for face perception. J. Neurosci. 17, 4302^4311. Houghton Mi¥in Co. Lako¡, G. 1987 Women, ¢re and dangerous things. University of Treisman, A. M. & Gelade, G. A. 1980 A feature-integration Chicago Press. theory of attention. Cogn. Psychol. 12, 97^136. Lane, R. D., Reiman, E. M., Ahern, G. L. & Schwartz, G. E. Ullman, S. 1945 Romanticism and synaesthesia: a comparative 1997 Neuroanatomical correlates of happiness, sadness, and study of sense transfer in Byron and Keats. Publ. Mod. Lang. disgust. Am. J. Psychiatr 154, 926^933. Assoc. Am. 60, 811^827. Locke, J. 1690 Essay concerning human understanding, ch. 8. Williams, J. M. 1976 Synesthetic adjectives: a possible law of London: E. Mory. semantic change. Language 52, 461^478. Lueck, C. J., Zeki, S., Friston, K. J., Deiber, M. P., Cope, P., Zeki, S. & Marini, L. 1998 Three cortical stages of colour Cunningham, V. J., Lammertsma, A. A., Kennard, C. & processing in the human brain. Brain 121, 1669^1685. Proc. R. Soc. Lond. B (2001)