Work in this area is undeniably fascinating; brain damage of various types can be deeply disruptive to the maintenance of personal identity in, frankly, really wacky ways. Damage to regions of the right hemisphere (RH) has been implicated in delusional misidentification disorders such as Capgras syndrome and Cotard’s delusion, as well as in anosognosic deficits.
To take these one by one, in Capgras the patient recognizes a loved one but insists they are an imposter (sometimes to tragic ends; see Blount’s (1986) case where the patient decapitated his father looking for batteries in the “robot’s” head). In Cotard’s delusion the patient believes they are dead. In anosognosia, the patient denies a deficit; their left hand is paralyzed but the patient may insist their arm is just tired or that they don’t wish to move the arm.
Note in these deficits, and in previous work on self-face recognition, researchers have suggested a unique role for the RH.
Research in self recognition has proceeded with cognitive psychological studies of neurologically intact participants and of patients undergoing the Wada test to isolate language processing, presumably prior to surgical treatment of epilepsy. Typical stimuli include face photographs, and some sort of explicit judgement of whether the depicted face is famous (or known) or of the self is required. In some tasks, face stimuli are digital morphs between two faces. Judgements are made with either hand; with central presentation of the stimuli, psychologists assume that response speed is reflective of faster processing by the hemisphere that controls that hand. Faster left-hand responses indicate efficient processing of the stimuli by the RH and show the immediate ability of RH motor commands to influence left-hand responding without transmission delays across the corpus callosum.
Turk, Heatherton, Kelley, Funnell, Gazzaniga and Macrae (2002, Nature Neuroscience) studied a famous split-brain patient (JW). JW saw lateralized presentations of face stimuli; because of his callosal disconnection surgery, presentations in the left visual field are initially processed by the RH, as in all of us, but computations are isolated and cannot cross the severed corpus callosum. The converse is true for right visual field presentations. Here, JW was presented with morphs between his face and that of Michael Gazzaniga (stimuli ranged between 0% JW and 100% JW in 10% increments). Gazzaniga is well known to JW by virtue of many years together at Dartmouth. The title of the paper is Mike or Me?, and this is precisely the judgement that JW was required to make, in separate blocks of trials.
The data gathered here was the proportion of “me” responses in the “Is this Me?” condition as a function of how much “me” was in the morph and the proportion of “Mike” responses in the “Is this Mike?” condition as a function of how much “Mike” was in the morph. Chi-squared analyses revealed a significant “Mike” bias in the RH and a significant “me” bias in the left hemisphere (LH).
Just to clarify then, in this work we are seeing (in the disconnected hemispheres of a spilt-brain patient) a LH bias for self-face recognition.
I have several problems with this study, not the least of which is that we are looking at a split-brain patient. If we are interested in the liklihood of a “default” hemisphere for recognition of self-faces, perhaps we ought not to study seriously epileptic patients, in whom there is bound to be abnormal cerebral organization.
But here’s my problem beyond that fact, and I think it affects all morphed facial stimuli studies. The cartoon below may help with my point:
—– Intermediate morphs here —–
Note there is much more changing than the relationships between the facial features that constitute an “identity code”. There are serious featural differences between the stimuli — JW is a darker stimulus with darker hair; clearly, morphs with darker top portions can be reliably associated with JW. We know from extensive work that the LH and RH differ in their processing of faces (see Bruno Rossion and colleagues’ work in the Journal of Cognitive Neuroscience, 2000). Identification of parts in a face, or “working with” facial features, is a lateralized process preferentially subserved by the LH. Thus we haven’t identified a site of self-recognition per se, only an area of brain or a hemisphere particularly responsive to features (in this case, darkness) within a face. If JW treated the task featurally it may not be surprising to see an LH bias for self. Similar concerns arise in cross-race facial identification studies, where race may be coded as a feature first and be responsible for the robust cross-race recognition difficulties observed (especially by the more feature-coding LH).
The description of the data leaves a great deal to be desired, unfortunately. A careful look shows substantive false alarming, and all by the RH. In the “Is this Me?” condition, the RH false alarmed 8% of the time when there was 0% JW in the image, and in the “Is this Mike?” condition the RH false alarmed 20% of the time when there was no MG in the image!
There are other potentially interesting data patterns, too. The LH self bias and the RH other bias emerge differently with the changing of the morphs. JW’s RH has very few “me” responses until the proportion of JW is 70%. At that point, there is still a LH bias until the proportion of JW in the image for “me” trials reaches 100%, where the LH and the RH are both answering “me” consistently. On the other hand, JW’s LH makes a fair number of “Mike” responses through the morphs with up to 60% Mike. At that point, there is no RH bias through 100% Mike.
These later points deserve scrutiny, and bring many questions. How can we subtract out the influence of false alarms? How can we analyze the data with a statistical procedure that allows testing for hemispheric differences at each morph level, as chi-squared can cleary not handle? Finally, and most importantly I think, we have to think outside the box and perhaps get away from using morphed images in self-face recognition research, if indeed the use of such images encourages featural or part-based responding.