The functional anatomy of the McCollough contingent colour aftereffect

Transcription

1 Brain imaging NeuroReport 10, 195±199 (1999) WE report two functional magnetic resonance imaging (fmri) experiments which reveal a cortical network activated when perceiving coloured grids, and experiencing the McCollough effect (ME). Our results show that perception of red±black and green±black grids activate the right fusiform gyrus (area V4) plus the left and right lingual gyri, right striate cortex (V1) and left insula. The ME activated the left anterior fusiform gyrus as well as the ventrolateral prefrontal cortex, and in common with colour perception, the left insula. These data con rm the critical role of the fusiform gyrus in actual and illusory colour perception as well as revealing localized frontal cortical activation associated with the ME, which would suggest that a `top-down' mechanism is implicated in this illusion. NeuroReport 10:195±199 # 1999 Lippincott Williams & Wilkins. Key words: Colour perception; Functional MRI; McCollough effect; Ventrolateral pre-frontal cortex The functional anatomy of the McCollough contingent colour aftereffect J. Barnes, CA R. J. Howard, 1 C. Senior, M. Brammer 2 E. T. Bullmore, 2 A. Simmons 3 and A. S. David Departments of Psychological Medicine, 1 Old Age Psychiatry, 2 Biostatistics and Computing and 3 Neuroimaging, Institute of Psychiatry and King's College School of Medicine and Dentistry, De Crespigny Park, Denmark Hill, London SE5 8AZ, UK CA Corresponding Author Introduction In 1965, McCollough described a colour after-effect that is contingent on the orientation of the stimuli presented [1]. After an induction period consisting of viewing two alternating patterns of chromatic gratings at 908 to each other, complementary colour after-effects contingent on orientation are seen when subjects view achromatic gratings. Explanations for this include local fatigue and/or error correction within speci c sets of neurons in the primary visual cortex (V1) [1], associative learning [2] and higher order perceptual learning effects [3]. The ME has also been reported to depend on wavelength of light coming from the grating rather than the perceived colour implicating the cells in V1 rather than the colour constancy cells of V4 [4]. The human visual system has a modular organization in which colour is represented separately from other visual properties. Functional neuroimaging and studies of patients with acquired de cits in colour perception has enabled the human colour area (V4) to be localized to the lingual and fusiform gyri in ventromedial occipitotemporal cortex or more speci cally, to the fusiform gyrus [5]. Higher level de cits and colour imagery may involve adjacent areas of the occipito-temporal cortex [6,7]. Unlike simple after-images which arise from the # Lippincott Williams & Wilkins retina the ME shows some inter-ocular transfer [8]. Proposed sites for the ME have included the lateral geniculate nucleus, primary visual cortex or the temporal lobes, although patients with known memory de cits due to Alzheimer's disease or bilateral medial temporal lobectomy show MEs of strength and duration comparable to normal controls [9]. It has been suggested that the site of the colour component of this effect could be the colour-sensitive cells found in the `blobs' of area V1 [10] which may interact with other cells in V1 sensitive to orientation and spatial frequency, and it is a change in connection strengths between these cells that give rise to MEs [11]. Patients with visual form agnosia following carbon-monoxide poisoning and traumatic cortical blindness [12] show the normal ME despite problems in recognising the simplest of geometric forms and discriminating the orientation of objects. Residual primary visual cortical function was presumed to account for these results. The present study reports two experiments which examined the neural substrate of the ME utilising functional magnetic resonance imaging (fmri). The rst involved contrasting coloured vs achromatic grids at 458, and was expected to produce net fusiform/lingual activation. The second contrasted achromatic grids (intended to elicit the illusion of colour) vs achromatic grids at 458 (see Fig. 1), in Vol 10 No 1 18 January

2 J. Barnes et al. order to show net activation speci c to the ME. We predicted that this would include fusiform and lingual areas, as well as additional cortical areas speci c to any cognitive components involved in the phenomenon. Such additional activation would have strong implications for a plausible neural mechanism for the ME. Materials and Methods Cerebral activation associated with the McCollough effect was investigated in seven neurologically normal volunteers (mean age 32 years; range 23±39). All were pre-tested to ensure adequate uncorrected vision and susceptibility to the ME. Prior to scanning each subject underwent induction of the ME which consisted of viewing test patterns displayed on a computer. The subjects viewed in free vision, two test patterns (as in Experiment 1, target phase A) alternately for 9 s each with a 1 s gap between patterns for a total induction of 30 min (see Fig. 1). Although the intensity of the ME was not recorded for each individual subject, after the induction period all subjects reported experiencing the illusory colour when viewing vertical and horizontal grids. All subjects were scanned within 15 min of induction. During the experiment the stimuli were projected by a computer controlled projector system onto a screen placed 2 m in front of the subjects' eyes, and viewed through a prismatic mirror. Each experiment comprised an ABAB design with ve repeats of 30 s of target and 30 s presentation of control stimuli. The target phase was viewed rst by the subjects in each of the experiments (see Fig. 1). Image acquisition: Gradient-echo echoplanar MR images were acquired using a 1.5 T GE Signa System (General Electric, Milwaukee, WI, USA) tted with advanced NMR hardware and software (ANMR, Woburn, MA, USA). Daily quality assurance was carried out using an automated quality control procedure. Images were acquired encompassing the entire temporal, occipital, frontal and parietal lobes with a search volume of voxels. In each of 14 non-contiguous 7 mm slices (0.7 mm gap), parallel to the inter-commissural (AC-PC) plane, 100 T2 ± FIG. 1. In the target (A) phase of Experiment 1, subjects were showed separate test patterns of red and black horizontal grating with a spatial frequency of 1.5 cycles/deg for 4 s, a 1 s blank white screen, followed by a green and black vertical grating for 4 s again followed by a white screen. In the reference (B) phase subjects were presented with a grating inclined at 458 to the right or left. The gratings alternated, one every 4 s separated by a 1 s blank white screen. In Experiment 2, the target (A) phase consisted of achromatic horizontal and vertical gratings which alternated every 4 s with a 1 s gap, while the reference (B) phase was identical to Experiment 1. All subjects reported seeing a pinkish hue when shown the horizontal gratings and a light green hue when viewing the vertical grating. 196 Vol 10 No 1 18 January 1999

3 The functional anatomy of the McCollough contingent colour after-effect FIG. 2. A generic brain activation map (GBAM) of the McCollough effect showing activations during Experiment 2. Activated clusters (in red) were in the fusiform gyri (right: 35, 47, 13; left: 40, 53, 7), ventrolateral prefrontal cortex (right: 46,22, 2; left: 40,36, 2) and left insula ( 49,11,4). weighted MR images depicting BOLD contrast [13] were acquired with TE ˆ 40 ms, TR ˆ 3000 ms. At the same session, a 43 slice, high-resolution inversion recovery echoplanar image of the whole brain was acquired in the AC-PC plane with TE ˆ 73 ms, TI ˆ 180 ms, TR ˆ ms. Image analysis: Slight subject motion during fmri can cause changes in T2 -weighted signal intensity unrelated to changes in BOLD contrast. An image realignment procedure was therefore applied and allowance made for spin history effects [14,15]. The power of periodic signal change at the (fundamental) A/B frequency of stimulation was estimated by iterated least squares tting a sinusoidal regression model to the motion-corrected time series at each voxel of all images [16]. The false-positive activation ratio was set such that 30 random type one errors would be expected in the area of the brain examined. The threshold was set by randomization [15]. This corresponded to voxels with a probability of false positive activation, All parametric maps of FPQ were then registered in the standard space [17] and generic activation for the group was then robustly determined [15]. Results The coloured patterns in Experiment 1 produced a small but signi cant increase in signal intensity in the fusiform gyrus (BA19), the insula and primary visual cortex only. The pattern seen here, including the left insula activation, is consistent with previous colour perception neuroimaging studies which utilized `Mondrian' patterns [6] (Table 1). The coordinates suggest the location to be area V4, as determined previously [5]. Studies using multicoloured Mondrian stimuli show that the location of V4 in the left hemisphere can vary between individuals by as much as 34 mm in the anteroposterior (y) axis and 24 mm in the dorsoventral (z) axis, and even after averaging across subjects, variations in location of 10.7 mm in the y axis and 3.95 mm in the z axis are evident [5]. In experiment 2, all subjects reported seeing colours when viewing achromatic vertical and horizontal lines (the ME). The illusory colour activated Table 1. Regional activations in colour perception and McCollough effect. Anatomical region No. Voxels Co-ordinates (x,y,z) Colour perception Fusiform gyrus 4 29, 58, , 52, 7 Primary visual cortex 5 6, 92,4 Insula 2 32, 19, 2 McCollough effect Fusiform gyrus 6 35, 47, , 53, 7 Ventrolateral prefrontal cortex 33 40,36, ,22, 2 Insula 6 49,11,4 Vol 10 No 1 18 January

4 the left anterior fusiform gyrus, but signi cantly, no activation was seen in the primary visual cortex (V1). Of particular note is the robust activation outside the temporo-occipital cortex including the right and left ventrolateral prefrontal cortex (Fig. 2). Discussion We have mapped the cortical activation of subjects when viewing illusory colour as in the McCollough contingent after-effect. This was made possible by scanning each individual performing the McCollough task and an equivalent colour perception task in the same session. Perception of the visual attribute of colour in the perception condition (Experiment 1) was associated with activity in an area described in previous studies as being V4 [5]. An illusion of the same attribute produced a different pattern of activation most strikingly, there was lack of activity in the area of V4. The viewing of the McCollough stimuli instead activated the left anterior fusiform gyrus, an area previously identi ed as having a role in the colour naming [18]. It should be noted that the ME is not a simple retinal after-effect. These have been studied using fmri and also appears to activate the fusiform gyrus [19,20]. Activation of the primary visual areas (V1/V2) was evident in Experiment 1 when subject viewed the red±black, green±black grids. Activation in this area has been noted in previous imaging studies of colour [5,21]. Interestingly, this V1/V2 activation was absent in the illusory condition. This could be as a result of V1/V2 being equally activated in both the target and reference phases or that the colour seen in the ME is due to a different mechanism involving higher cognitive aspects. A further explanation as to why illusory colour does not activate V1/V2 might be that the cells in this area seem to be important in wavelength discrimination [22], a role that might be irrelevant in the perception of illusory colour. We did not predict extensive frontal activation speci cally in the ventrolateral prefrontal cortex when viewing illusory colour, which we assume, re ects additional processing underlying the ME. Activity in such areas has been associated with visual working memory [23], and when comparing normally coloured objects with their black and white counterpart [24]. This pre-frontal activation cannot be attributed to `bottom up' effects such as the fatiguing of neurons in the early visual areas. This prompts us to speculate that the process that produces the colour illusion is a process related to a `top-down' mechanism. It is clear that the activation seen in the frontal lobes is due to more complex colour tasks as it was not seen in simple viewing of coloured patterns. The interpretation of the activation seen in the McCollough effect is not simple, as there is no agreement between the colour centre in humans and monkeys. Recently, there have been claims that the colour centre in the human that is homologous to the colour area in primates is located in area TEO rather than in V4 of the macaque cortex and as such has been given a new name V8. This is clearly a topic of ongoing debate [25±27]. Our main aim was to compare veridical and illusory colour perception in humans rather than to comment on the functional equivalence of human and primate cortical activity. Our ndings instead are relevant to the issue of a colour processing stream [5] that may be hierarchical in nature, where the fusiform gyrus has a essential role, and additional areas of activation such as frontal and temporal cortex may represent higher levels of processing of colour. What remains to be investigated is to what extent activation in these regions correlates with the subjective magnitude of ME experienced. Conclusion We have mapped the activation associated with the McCollough illusion and demonstrated that the viewing of this illusion is associated with activity within the anterior fusiform gyrus, close to the area V4, previously reported as functionally specialised for the early processing of colour. Although activation in the ventrolateral prefrontal cortex during this illusion was not predicted it does tend to suggest that a `top-down' processing mechanism is implicated in this colour contingent after-effect. References J. Barnes et al. 1. McCollough C. Science 149, 115±116 (1965). 2. Allen LG and Siegel S. Cognition 64, 207±222 (1997). 3. Bedford FL. Cognition 64, 23±230 (1997). 4. Thompson P and Latchford G. Nature 320, 525±526 (1986) 5. McKeefry DJ and Zeki S. Brain 120, 2229±2242 (1997) 6. Howard RJ, ffytche DH, Barnes J et al. NeuroReport 9, 1019±1023 (1998). 7. De Vreese LP. Neuropsychologia 29, 1±18 (1991). 8. MacKay DM and MacKay V. Nature 242, 477±479 (1973). 9. Savoy RL and Gabrieli JD. Percept Psychophys 49, 448±455 (1991). 10. Livinstone M and Hubel D. J Neurosci 4, 309±356 (1984). 11. Savoy RL. Percept Psychophys 36, 571±576 (1984). 12. Humphrey GK, Goodale MA, Corbetta M et al. Curr Biol 5, 545±551 (1995). 13. Ogawa S, Lee TM, Kay AR and Tank DW. Proc Natl Acad Sci 3, 9868±9872 (1990). 14. Friston KJ, Williams SCR, Howard R et al. Magn Reson Med 35, 346±355 (1996). 15. Brammer MJ, Bullmore ET, Simmons A et al. Magn Reson Imag 15, 763±770 (1997). 16. Bullmore ET, Brammer MJ, Williams SCR et al. Magn Reson Med 35, 261±277 (1996). 17. Talairach, Tournoux. Stuttgart: Thieme Verlag (1988). 18. Martin A, Haxby JV, Lalonde FM et al. Science 270, 102±105 (1995). 19. Sakai K, Watanabe E, Onodera Y et al. Proc R Soc Lond B 261, 89±98 (1995). 20. Hadjikhani N, Liu AK, Dale AM et al. Nature Neurosci 1, 235±241 (1998). 21. Corbetta M, Miezin FM, Bobmeyer S et al. J Neurosci 6, 1811±1828 (1994). 22. Kulikowski JJ, Walsh V, McKeefry DJ et al. Behav Brain Res 60, 73±78 (1994). 23. Courtney SM, Ungerleider LG, Kell K et al. Nature 386, 608±611 (1998). 24. Zeki S and Marini L. Brain 121, 1669±1685 (1998). 198 Vol 10 No 1 18 January 1999

5 The functional anatomy of the McCollough contingent colour after-effect 25. Zeki S, McKeefry DJ, Bartels A and Frackowiak RSJ. Nature Neurosci 1, 335 (1998). 26. Tootell RBH and Hadjikhani N. Nature Neurosci 1, 335±336 (1998). 27. Heywood C and Cowey A. Nature Neurosci 1, 171±172 (1998). ACKNOWLEDGEMENTS: J.B is supported by a UK MRC studentship; C.S is supported by the John MacKintosh Trust, Gibraltar; E.T.B is supported by the Welcome trust. We would also like to thank Chris Andrew, Steve Williams and the staff of the MRI unit. Received 23 September 1998; accepted 13 November 1998 Vol 10 No 1 18 January