Saturday, February 18, 2012

That's Impossible! How the Brain Processes Impossible Objects

Relativity, by M.C. Escher.

The artwork of M.C. Escher is famous for its visual trickery. The human visual system tries to project the two dimensional image onto a three dimensional scene, but the perspective is contradictory: it cannot exist in the real world. These impossible constructions violate the laws of geometry and fascinate consumers of t-shirts, posters, and Apple products.

How does the brain represent these illusory staircases and towers? While a fascinating topic of study in the field of object perception (Levy et al., 2004), Escher prints can make for overly complicated stimuli in neuroimaging experiments. Simpler 2D figures, such as the impossible objects drawn by Swedish artist Oscar Reutersvärd, have been used in fMRI experiments (Soldan et al., 2008).

An extensive collection of 810 impossible objects is available from Impossible World, which is a fantastic resource1 maintained by Vlad Alexeev.

Previous neuroimaging experiments have used the possible/impossible object decision task to study the neural correlates of perceptual priming, an implicit form of memory. Behaviorally, repeated presentation of possible objects results in faster decision times, and this priming effect is smaller (Soldan et al., 2008) or non-existent (Schacter et al., 1995) for impossible objects. Neurally, the phenomenon of repetition suppression, or the reduction in neural activity seen upon repeated stimulus presentation, is thought to reflect facilitated perceptual processing (and perhaps behavioral priming).2 Repetition suppression predicts behavioral priming for possible objects (Habeck et al., 2006):
A set of occipital, parietal, and temporal brain regions decreased their activation across presentations, including bilateral middle occipital gyrus, left precuneus, right supramarginal gyrus, as well as some frontal and thalamic areas, such as right inferior frontal gyrus, left cingulate gyrus, and right thalamus.
However, no such relationship was observed for impossible objects.

The previous studies focused on varieties of repetition priming and whether there is a "structural description system" that facilitates the identification of perceptually coherent objects. A recently published article was specifically interested in the neural basis of impossible figures and how they are represented in the visual cortex (Wu et al., 2012). The stimuli were impossible and possible exemplars of the two-pronged trident (Fig. 1 below), shown at four different angles.

Fig. 1 (Wu et al., 2012). Examples of stimulus figures used in impossible condition and possible condition. (a) Is an impossible figure and (b) is a possible figure [that] resembles the former.

The paper started by reviewing the basic neuroanatomy of the visual system and its division into dorsal ("where") and ventral ("what") visual streams. Objects are primarily represented in the ventral stream, and the lateral occipital complex (LOC) is one area that seems to be specialized for object recognition. The authors predicted that impossible objects would be difficult for the LOC to process; therefore, additional regions would be recruited:
In the present study, we thought that the 3D structures of impossible figures might be difficult to be represented by object-selective regions (such as the LOC), and the impossible perceptions might be derived from detecting the contradiction in interpretation of the 3D structure. Therefore, we postulated that both the brain regions in the dorsal visual pathway, such as the SPC [superior parietal cortex] related to the perceptual ambiguities resolving and perceptual content modifying, and the brain areas related to the object-selective regions in the ventral pathway would be involved in the impossible figures processing.
Nineteen participants performed the possible/impossible object decision task (30 trials of each condition) while their brains were scanned. Four participants showed repetition priming in the task (first 15 trials of each condition slower than the last 15) and were excluded. The remaining subjects did not show priming.3 Personally, I would have used 30 unique possible and impossible figures to avoid priming effects entirely.

What were the results? As predicted, regions in both dorsal and ventral visual streams showed greater activation for impossible than for possible figures: right superior parietal in the former and right fusiform and inferior temporal gyri in the latter.
The right SPG in the dorsal visual pathway might be related to spatial information processing and the right LOC (FG and ITG) in the ventral visual pathway (the object-selective regions) might be related to the representation of the impossible 3D structure. Therefore, our results indicated that the impossible 3D structure might be difficult to be represented by human visual system, and the impossible perception might be derived from the detecting and resolving the contradiction in the subjects’ interpretations according to different perceptions triggered by 3D cues.

Fig. 2 (Wu et al., 2012). Brain regions showing significant difference between impossible condition and possible condition [FEW-corrected threshold of P < 0.05 at the cluster level (P < 0.001, 10 contiguous voxels cutoff at the voxel level)].

There were no brain regions that showed greater activation for possible objects.

The authors suggested that their ventral stream regions are part of LOC, although this is debatable. In the original study of Malach et al. (1995), LOC is posterior to the inferior temporal focus here, but Grill-Spector et al. (2001) state that:
...the entire region, beginning in lateral occipital cortex and extending anteriorly and ventrally into posterior temporal regions, responds more strongly to intact objects with clear shape interpretations than to control stimuli that do not depict clear shapes.
LOC doesn't seem to differentiate between familiar objects and unfamiliar objects with clear 3D interpretations (e.g. Henry Moore sculptures). At any rate, what's interesting here is that LOC was more active for impossible objects, suggesting that "the 3D spatially impossible structure could, [with difficulty], be represented by the visual system." And that, along with greater activity in the right superior parietal cortex, is how the brain processes impossible objects.


1 Perhaps a link to Impossible Worlds (810 figures as line drawings and grayscale images) can be added to the Tarr Lab database.

2 Interpreting neural repetition suppression effects as a reflection of behavioral priming is complicated, however (Horner & Henson, 2012).

3 However, the interaction effect approached significance (p=.083).


Grill-Spector K, Kourtzi Z, Kanwisher N. (2001).The lateral occipital complex and its role in object recognition. Vision Res. 41:1409-22.

Habeck C, Hilton H, Zarahn E, Brown T, Stern Y. (2006). An event-related fMRI study of the neural networks underlying repetition suppression and reaction time priming in implicit visual memory. Brain Research 1075:133-141.

Horner AJ, Henson RN. (2012). Incongruent abstract stimulus-response bindings result in response interference: FMRI and EEG evidence from visual object classification priming. J Cogn Neurosci. 24:760-73.

Levy EK, Levy DE, Goldberg ME. (2004). Art and the brain: the influence of art on Roger Shepard's studies of mental rotation. J Hist Neurosci. 13:79-90.

Malach R, Reppas JB, Benson RR, Kwong KK, Jiang H, Kennedy WA, Ledden PJ, Brady TJ, Rosen BR, Tootell RB. (1995). Object-related activity revealed by functional magnetic resonance imaging in human occipital cortex. Proc Natl Acad Sci 92:8135-9.

Schacter DL, Reiman E, Uecker A, Polster MR, Yun LS, Cooper LA. (1995). Brain regions associated with retrieval of structurally coherent visual information. Nature 376:587-90.

Wu, X., Li, W., Zhang, M., & Qiu, J. (2012). The neural basis of impossible figures: Evidence from an fMRI study of the two-pronged trident Neuroscience Letters, 508 (1), 17-21 DOI: 10.1016/j.neulet.2011.11.064


Subscribe to Post Comments [Atom]


At February 19, 2012 4:25 AM, Anonymous Anonymous said...

Thank you for such an interesting article.

I had a psychotic episode a while back, it went on for some time, and two things I noticed where that I lost my sense of irony (I couldn't see the funny side of The Producers at all) and I had trouble reading phrases where the word could mean two things eg. "lift" meaning raise up and "lift" meaning elevator. I couldn't seem to make the words make sense.

I have read somewhere that this priming you are talking about is related to the brain being able to choose between two instances when the one word means different things. The brain has to choose between all the different meanings of that word each time a word comes up and priming means that you don't always have to spend a lot of time thinking about the meaning of these words.

Is this in the ball park?

Total amateur but interested.

At February 19, 2012 10:54 AM, Blogger The Neurocritic said...

Anonymous - Thanks for commenting about your experiences. There are indeed different types of priming. Semantic priming is closest to what you're describing. In a typical experiment, the meaning of one word primes the retrieval of related words, so you're faster to identify the second word (e.g., apple-orange vs. apple-window). Interestingly, individuals with schizophrenia tend to show alterations in semantic priming:

"(1) schizophrenia patients have overly inclusive semantic networks... and (2) schizophrenia patients are deficient in their use of semantic context, responding to primed words as if they were unprimed..."

Even more similar to what you're describing is the use of context to select the appropriate meaning of a word: "I took the lift to the top floor" vs. "I got a lift from the positive news." Here, those with schizophrenia can show "disproportionate misinterpretation of subordinate meanings (toast at a wedding)."

So you're definitely in the ball park.

At February 19, 2012 11:06 AM, Anonymous Matt Craddock said...

Thanks for the pointer to the Impossible World, really nice resource!

LOC seems to be generally involved in the recovery of 3d shape, and this role seems to be somewhat independent of the sensory source from which the information is recovered. For example, LOC is involved during haptic recognition of objects and when shape is conveyed by an auditory sensory substitution device . In keeping with this role of LOC as something that recovers 3D shape, I would be tempted to interpret greater activation to impossible objects as continuing attempts to resolve the conflicting information about their shape, which is quite different from a circumstance in which there is *no* object or discernible shape present, and quite different again from the activation of some kind of representation of the shape: I tend to think of LOC as a kind of online processor rather than a "storehouse". The final conclusion "the 3D spatially impossible structure could, [with difficulty], be represented by the visual system." seems like a bit of a restatement of the results; we know it must do *something* in the visual system, because otherwise we wouldn't be able to see it.

Anyway, to give it a positive sumamry: I like the demonstration that LOC is involved in processing both possible and impossible 3d shapes, and the implication that it has to work a bit harder for impossible ones.

At February 19, 2012 11:40 AM, Blogger The Neurocritic said...

Matt Craddock - Thanks for the more detailed interpretation of the LOC finding! And also for the links to the multimodal properties of LOC, which had escaped my attention.

At February 20, 2012 9:08 PM, Blogger weiyan said...

You wrote, "Repetition suppression predicts behavioral priming for possible objects (Habeck et al., 2006)... However, no such relationship was observed for impossible objects."

In the actual article, "In the fMRI results, the data of 4 subjects that showed adaption effect were also deleted."

Maybe it is possible that if you combine all the data you won't be able to find an adaptation effect, but I don't think you can conclude that there is definitely no adaptation effect without seeing their complete set of data. There is, just that they removed those subjects. Small number of subjects, but still relevant.

I'm not nitpicky, I was just a little bit misled.

At February 21, 2012 12:04 AM, Blogger The Neurocritic said...

weyan - Are you confusing two separate studies (Habeck et al. 2006 and Wu et al. 2012)?

Wu et al. did make the assumption that "according to the RT results, we thought that the confusing factor of the adaption might be ruled out." But you're correct, they didn't specifically analyze all their fMRI data to check for adaptation effects.

At February 29, 2012 12:14 AM, Blogger weiyan said...

Thanks for the clarification. I really should try and read slower. ;)


Post a Comment

<< Home

eXTReMe Tracker