The Riddle of Anosognosia: Does Unawareness of Hemiplegia Involve a Failure to Update Beliefs?

cortex 49 (2013) 1771e1781

Available online at www.sciencedirect.com

Journal homepage: www.elsevier.com/locate/cortex

Research report The riddle of anosognosia: Does unawareness of hemiplegia involve a failure to update beliefs?

Roland Vocat a, Arnaud Saj b and Patrik Vuilleumier a,b,* a Laboratory for Behavioral Neurology & Imaging of Cognition, Department of Neuroscience, Medical School, University of Geneva, Switzerland b Clinic of Neurology, University Hospital of Geneva, Geneva, Switzerland article info abstract

Article history: Anosognosia for hemiplegia (AHP) is defined as a lack of awareness for motor incapacity Received 1 February 2012 after a brain lesion. The causes of AHP still remain poorly understood. Many associations Reviewed 10 May 2012 and dissociations with other deficits have been highlighted but no specific cognitive or Revised 30 August 2012 neurological impairment has been identified as a unique causative factor. We hypothe- Accepted 20 October 2012 sized that a failure to update beliefs about current state might be a crucial component of Action editor Giuseppe Vallar AHP. Here, we report results from a new test that are compatible with this view. We Published online 27 November 2012 examined anosognosic and nosognosic brain-damaged patients, as well as healthy controls, on a task where they had to guess a target word based on successive clues, with Keywords: increasing informative content. After each clue, participants had to propose a word solu- Anosognosia for hemiplegia tion and rated their confidence. Compared to other participants, anosognosic patients were AHP abnormally overconfident in their responses, even when information from the clues was Awareness insufficient. Furthermore, when presented with new clues incongruent with their previous Denial response, they often stuck to their former “false” beliefs instead of modifying them. This Hemiplegia impairment was unrelated to global deficits in reasoning or memory, and all patients Stroke eventually identified the correct solution of riddles after the last, fully informative, clue. These results suggest that a deficit in the generation and adjustment of beliefs may be a key factor contributing to the occurrence and persistence of anosognosia, when associated with concomitant losses in motor, proprioceptive, and/or attentional functions. Patients may remain unaware of their deficit partly because they cannot “update” their beliefs about current state. ª 2012 Elsevier Ltd. All rights reserved.

1. Introduction dysfunction. As initially described by Babinski (1914), the most common presentation is anosognosia for hemiplegia (AHP), Anosognosia is a striking neurological symptom that mani- where patients fail to recognize a severe motor deficit despite fests as unawareness and lack of concern of for their deficit direct confrontation during neurological examination. (hemiplegia, hemianopia, neglect, etc.) in brain-damaged pa- This lack of awareness can affect the motor deficit itself tients, and that cannot be explained by global cognitive (e.g., “I’m able to move my left arm”) but also its consequences

* Corresponding author. Neurology & Imaging of Cognition, Department of Neuroscience & Clinic of Neurology, University Hospital (HUG), 24 rue Micheli-du-Crest, CH-1211 Geneva, Switzerland. E-mail address: [email protected] (P. Vuilleumier). 0010-9452/$ e see front matter ª 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.cortex.2012.10.009 1772 cortex 49 (2013) 1771e1781

for a particular action (e.g., “I can open a bottle”) and/or its 1995; Gilmore et al., 1992; Lu et al., 2000, 1997), neuroana- general implications in everyday life (e.g., “I can go back to tomical studies have pointed to an involvement of many work”). The cognitive and neural mechanisms of this phe- different brain regions including the parietal lobe (Bisiach nomenon are poorly understood, and remain difficult to study et al., 1986; Heilman, 1991), insula (Karnath et al., 2005), pre- in a systematic way due the predominance of symptoms in motor, motor and sensory areas (Berti et al., 2005), thalamus acute stages and their heterogeneity (Vocat et al., 2010). Several (Starkstein et al., 1992), or more complex cortico-subcortical theoretical accounts of AHP have been proposed (Berti et al., circuits of motor control (Pia et al., 2004). Our own work 2005; Davies et al., 2005; Heilman, 1991; Levine et al., 1991; (Vocat et al., 2010) also highlighted a role for distributed le- Ramachandran, 1995; Weinstein and Kahn, 1955) but none of sions in areas associated with motor function, spatial atten- them appears sufficient to explain all the manifestations of tion, as well as interoceptive and affective functions, anosognosia, and the frequent dissociations with other disor- consistent with a multi-componential model of AHP. ders or between different domains of anosognosia itself (Vocat However, although the ABC model (Vuilleumier, 2004)or2- and Vuilleumier, 2010; Vuilleumier, 2000). Babinski (1914) factor delusional theory (Davies et al., 2005) made the hy- initially emphasized that AHP was often associated with pothesis that AHP might be critically linked to impairments in hemianesthesia and proprioceptive loss, possibly preventing the generation and adjustment of one’s beliefs in face of un- direct sensory experience of weakness, whereas Bisiach et al. certainty, no study has directly tested the ability of patients to (1986) insisted on a particular relation between AHP and check their belief accuracy and/or change their current be- spatial neglect which might disrupt conscious awareness of liefs. If anosognosics still rely on their past beliefs about their left hemibody. The discovery theory of Levine et al. (1991) health and are not able to modify them according to the new proposed that AHP could emerge due to a conjunction of condition (e.g., due to losses in sensory feedback, intentions, global cognitive impairment and impaired proprioception. A or bodily representations), they might indeed fail to correctly role for motor neglect was also highlighted as a possible acknowledge their current handicap. Here, we therefore causative factor by Heilman (1991), but related to a suppression designed a new neuropsychological test allowing us to of motor intention. Likewise, a deficit in comparator mecha- address this issue. Further, we aimed at using a task simple nisms that match the predicted movements (based on inten- and brief enough to be administered to patients at the bedside, tion) and the actual movement (based on sensory feedback) even in the acute stage. Our test required participants to make was also recently implicated (Bottini et al., 2010; Fotopoulou guesses in conditions where they should feel uncertain about et al., 2008; Frith et al., 2000). Finally, for Weinstein and Kahn their responses and could not be confident about their beliefs. (1955) and Turnbull et al. (2002), motivational or psychologi- Our main aim was to observe how patients with AHP could cal defense mechanisms could not be excluded. change their beliefs when confronted with incongruent in- More recent hypotheses (Davies et al., 2005; Vuilleumier, formation contradicting these beliefs. 2004) have suggested multiple causative factors. According to this view, AHP might emerge through various combinations of 2 (or more) deficits. Davies et al. (2005) proposed that AHP 2. Method could resemble delusions, requiring the presence of a first impairment that prompts delusional beliefs and a second 2.1. Population impairment that interferes with processes of belief evaluation which would otherwise allow rejecting the delusional belief. We tested 11 control participants (mean age 62.8 9.6, Similarly, Vuilleumier (2004) put forward a general “ABC 8 women, all right-handed), with no history of neurological or model” of anosognosia, where a combination of processes psychiatric diseases, and a group of 9 consecutive patients involving appreciation, belief, and check operations might with a first right hemispheric stroke and a full left hemiplegia. subserve awareness for motor losses, or cause anosognosia or The severity of AHP in these patients was measured using the denial when damaged in different ways in different patients. classical scale of Bisiach et al. (1986). Among them, 5 patients According to this view, the ability to modify one’s beliefs (mean age 56.4 16.8, 3 women, all right-handed) were able to together with the aptitude to monitor performance and out- give a good spontaneous description of their plegia (Bisiach comes in case of uncertainty might play a critical role in AHP, scale rating ¼ 0) and constituted the nosognosic group; in combination with other primary deficits in sensory, motor, whereas 4 other patients (mean age 60 10.2, 2 women, all and/or attentional functions (Venneri and Shanks, 2004). A right-handed) had clear and stable anosognosia for their recent elegant study (Jenkinson et al., 2009) also showed that hemiplegia (Bisiach score ¼ 3 in all of them) and were there- patients with AHP are impaired in reality monitoring pro- fore included in the anosognosic group. We did not test pa- cesses not strictly related to movement, but also extending to tients with intermediate Bisiach scores (ratings ¼ 1or2) visual perception. In keeping with multifactorial causes, a because it has been shown that measures of AHP may be recent prospective study (Vocat et al., 2010) showed that unreliable in these cases and represent only partial or fluc- different combinations of deficits and lesions have a different tuating deficits in awareness (Baier and Karnath, 2005). impact on the presence of AHP, with proprioceptive loss being most strongly correlated with the severity of AHP in the acute 2.2. Riddle test stage after the stroke, but other cognitive impairments being more crucial for its persistence in chronic stages. The task required participants to guess 10 target words. For Moreover, besides a general predominance of right hemi- each of these words, a succession of 5 verbal clues was pro- spheric dysfunction in AHP (Adair et al., 1995; Breier et al., vided in turn. These clues were brief sentences presented cortex 49 (2013) 1771e1781 1773

visually on a laptop in front of the participants and were calculated a global score of correct reasoning which was ob- concurrently read aloud by the experimenter to ensure correct tained by summing the number of riddles that were finally comprehension. The clues gave information about semantic resolved (with a correct answer given before or after the fifth or physical attributes of the targeted word that did not contain clue). Finally, for each participant, we computed a measure of any explicit motor imagery or gestural component. Each clue their level of confidence for the 1st, 2nd, 3rd, 4th and 5th clue, appeared on the screen one after the other. They remained by averaging their ratings after each clue in each of the 10 displayed until the end of the riddle for a given word target. riddles. Following the presentation of each clue (see example in Fig. 1, These different scores were then averaged and compared and Appendix for a complete list of the riddles), participants between the 3 groups of participants: healthy controls, had to propose verbally a word that could possibly correspond nosognosic hemiplegic patient, and anosognosic hemiplegic to the current clue(s). If they could not find any word, they had patients. Statistical comparisons between the 3 groups on the possibility to advance to the next clue but were encour- the different scores were performed by using the non- aged to search for a possible word (all participants were parametrical ManneWhitney U-test, with a p-value generally able to do so and only rarely refused to guess a threshold for significance fixed at .05. In addition, individual word). After a guess was uttered, the experimenter typed the comparisons of response confidence level were made between proposed word on the laptop and validated it. Then, partici- the score of each anosognosic patient and both the control pants had to rate their level of confidence in their guess. This and the nosognosic groups using the Crawford statistical test. rating ranged from “0” (completely unsure) to “8” (completely sure) on a vertical scale presented visually. Patients gave their 2.3. Additional clinical testing level of confidence by reporting the number and a verbal label associated with it (e.g., “0, completely unsure”). If no word was Neurological deficits, as well as executive and memory func- proposed (10.7% of trials on average), the confidence rating tions were also assessed in patients using several standard- was skipped. This was then followed by the next clue (see ized clinical tests that could be administered at the bedside example in Fig. 1). After the fifth and last clue, the rating of (see also Vocat et al., 2010). These included in particular tactile confidence for the last given guess was followed by the display sensation [score based on 12 stimulations; 3 repetitions of 2 of the “correct” word target corresponding to the definition of types of stimulation (touch or sting) on 2 sites (hand or all preceding clues. No other feedback on accuracy was given elbow)], proprioception [score based on 9 trials; 3 joints until this final clue. (middle finger, wrist and elbow), 3 times each], visual field and Importantly, the 5 successive clues were chosen to have a visual extinction [scores based on 4 stimulations, with small progressive informative content. The first was always finger movements across each hemifield], and spatio- ambiguous enough to have multiple potential answers, while temporal orientation. Visuo-spatial neglect was assessed the next successive clues gave more and more precise infor- using a composite score from 3 tests: star cancellation from mation on the target, in such a way that the possible answers the Behavioral Inattention Test (Wilson et al., 1987), multiple were less and less numerous. The fifth and last clues were line bisection (Schenkenberg et al., 1980), and copy of the chosen to leave no doubt about the correct answer to the GainottieOgden figure (Ogden, 1985). Several tests were used riddle. Fig. 1 shows an example of the whole succession of to assess general frontal functions such as verbal fluency clues for 1 of the 10 riddles. The riddles were presented in a [using names of animals and words beginning with the letter random order counterbalanced across participants. The test “M”, 1 min each (Thuillard and Assal, 1991)], digit span was administrated approximately one week after the stroke. (Wechsler, 1981), and mental flexibility [category sorting task Each response of each participant after each clue was (Weigl, 1927)]. In addition, we also assessed global cognitive classified as “correct response” (if it corresponded to the cor- functioning with MMSE (Folstein et al., 1975), providing sepa- rect word solution of the riddle), “false response” (if it did not rate measures for verbal memory (word-list subtest of the correspond), “repetition” (if the same false response was given MMSE), and working memory (subtraction subtest of the as for the preceding step), and “no response” (if the patient MMSE). The means from the 2 patient subgroups (anosogno- could not report any word). For each participant, we also sics vs nosognosics) on each of these scores were compared by non-parametrical ManneWhitney U-test with a threshold of significance at p < .05.

2.4. Analysis of brain lesions

We used the same procedure as previously used in larger studies on brain lesion-behavior relationships in right- hemisphere patients (Verdon et al., 2010; Vocat et al., 2010). The location and extent of brain damage was delineated in each patient, based on MRI scans obtained after the first week Fig. 1 e Example of one of the 10 riddles. Five clues were post-stroke (5.8 1.2 days on average). MRI scans were ob- provided in successive order, until the correct response tained with a 3 T apparatus (Siemens Trio). The usual thick- was given. All clues remained visible on the screen until ness was 6 mm (interslice gap: 1 mm). We used the T1 and T2 resolution of the riddle, so as to minimize working images, including T2-weighted fluid attenuated inversion re- memory demands. covery (Flair) sequences, for optimal identification of the 1774 cortex 49 (2013) 1771e1781

lesion areas. We selected brain scans showing the greatest and controls, respectively) and for the third clue ( p ¼ .032 and extent of lesions in all patients. The lesion area was manually p ¼ .001, compared to nosognosics and controls, respectively). reconstructed using individual images and transferred onto a Following the fourth clue, the certainty levels in anosognosics standardized brain template with the MRIcro software were still significantly different from controls ( p ¼ .006) but (Rorden and Brett, 2000, www.mricro.com). This was verified did not reliably differ from nosognosics ( p ¼ .190) even though by a trained neurologist who was blind to the patients’ per- the mean rating values were still numerically higher in ano- formance. We then examined the overlap of the normalized sognosics (see Fig. 3). By contrast, on the fifth and final clue lesion areas for the different subgroups of patients (see Fig. 2). where the riddle could generally be resolved without ambi- To estimate the critical anatomical regions differentially guity, no difference was found between anosognosics and damaged in the 2 patient subgroups, we compared the lesion other participants, nosognosics ( p ¼ .413) or controls overlap for the nosognosic and the anosognosic patients with ( p ¼ .138). Importantly, the confidence rating values reported a subtraction procedure and chi-squared analysis. Due to our by nosognosics and controls did not differ for any of the 5 small population, these data are mainly reported for clues (.510 < p < .913). completeness and comparison with other studies (e.g., This overconfidence was seen in each and every anosog- Karnath et al., 2004; Berti et al., 2005). nosic patient as compared to other participants (see Fig. 3b), as verified by additional single-case analyses. For the 3 first clues, all 4 patients showed significant differences (Crawford test) 3. Results with the control group (.001 < p < .05), with only one single comparison showing only a tendency for one patient after the All participants including patients completed the whole pro- first clue ( p ¼ .084). For the differences with the nosognosic cedure with 10 riddles in a single session. Subjective reports group, p-value ranged from .001 to .01 for the first clue, from during the test indicated that they enjoyed the task and were .001 to .09 for the second clue (3 anosognosics patients out of 4 motivated in finding the correct word targets. This motiva- showing a significant effect) and from .02 to .27 for the third tional factor was important to ensure the validity of our clue (2 anosognosics out of 4 with a significant effect). These measure. The 3 different groups were not significantly analyses confirm that each and every anosognosic patient different in term of mean age [anosognosics (58) vs nosogno- showed a significant difference (or a strong tendency) in sics (62), p ¼ .999; anosognosics vs controls (61), p ¼ .503; confidence that predominated for the 3 first clues relative to nosognosics vs controls, p ¼ .606], and the 2 patient groups both the control and the nosognosic patients. In line with the (anosognosic vs nosognosic) were tested after a similar time group data, controls and nosognosic patients did not differ. delay since stroke onset (11 and 11.25 days respectively, We next analyzed the quality of responses given to each p ¼ .937). clue in each riddle (50 in total). Overall, anosognosics showed A word solution was proposed and a level of confidence significantly fewer correct responses than nosognosics assigned by the participant on most trials (90% on average). ( p ¼ .032). A similar tendency was found when comparing We first analyzed the ratings of confidence (ranging from 0 to anosognosics with controls ( p ¼ .078). Furthermore, when we 8) for responses given to the 5 successive clues (averaged examined the nature of these errors (see Fig. 4a), we observed across all riddles). As illustrated in Fig. 3a, patients with that anosognosics repeated an incorrect response from a anosognosia showed a significantly higher degree of “cer- preceding trial much more often than both nosognosics tainty” after the first clue than both nosognosics ( p ¼ .016) and ( p ¼ .016) and controls ( p ¼ .006), whereas the 2 latter groups healthy controls ( p ¼ .006). The same result persisted for the made the same amount of repetitive responses ( p ¼ .51). Thus, second clue ( p ¼ .016 and p ¼ .003, compared to nosognosics anosognosics often failed to modify their responses when

Fig. 2 e Brain lesions in each patient group. The overlap of damaged areas is shown for (a) 5 patients with left hemiplegia but no AHP, and (b) 4 patients with left hemiplegia plus AHP. Color codes represent the number of patients with lesions affecting the same area. cortex 49 (2013) 1771e1781 1775

Fig. 3 e Confidence ratings reported for each of the 5 successive clues (mean for the 10 riddles). The level of confidence was obtained after each guess proposed by the participant, and could range from 0 (completely unsure) to 8 (completely sure). The mean confidence ratings are plotted for each group (a; controls, nosognosics, and anosognosics) or detailed for the four anosognosic patients (b).

confronted with a new clue that was incongruent with their (correct in all patients), as well as visuo-spatial neglect tasks previous guess (e.g., proposing “clock” after clue No. 1 and ( p ¼ .997), verbal memory ( p ¼ .178), and working memory repeating it again after clue No. 3 in the example given in ( p ¼ 1.000). Most importantly, no difference was found for Fig. 1, Table 1). tests probing frontal lobe functions, including verbal fluency Remarkably, however, when we computed the number of (categorical: p ¼ .386; phonological: p ¼ 965) and digit span riddles that were eventually resolved on the last trial (i.e., the ( p ¼ .720). Remarkably, the patients did not differ either on accuracy of the final word proposed), we found no significant measures of reasoning and mental flexibility in the Weigl difference in anosognosics as compared with nosognosics category sorting test ( p ¼ .539). MMSE scores were also com- ( p ¼ .111) or controls ( p ¼ .412), while the 2 latter did not differ parable in anosognosic and nosognosic patients (27.5 and 28.5, either ( p ¼ .583). This final resolution outcome demonstrates respectively, p ¼ .132). The “normal range” of these scores that general reasoning abilities and semantic knowledge were indicates that both groups of patients showed no severe not affected in anosognosic patients. cognitive alteration. In addition, none of them showed a se- We also examined performance on several additional vere amnesia, profound neglect, and/or massive executive neurological and neuropsychological tests. Compared to non- dysfunction. anosognosic patients, the anosognosics showed no significant Finally, we performed an exploratory anatomical analysis difference on any of these tests, including tactile sensation comparing the lesions overlap of anosognosic and nosognosic ( p ¼ .555), proprioception ( p ¼ .601), visual field loss ( p ¼ .116), patients. Using a standard subtraction analysis (see Fig. 5a), visual extinction ( p ¼ .345), spatio-temporal orientation the maximal lesion overlap in affected patients fell in the 1776 cortex 49 (2013) 1771e1781

Fig. 4 e Different types of response given to the riddles. Repetitions correspond to false words that were proposed more than once to successive clues in the same riddle. Resolutions correspond to correct responses given after the ﬁfth (ﬁnal) clues. The number of repetitions and resolutions (mean for the 10 riddles) are plotted for each group (a; controls, nosognosics, and anosognosics) or detailed for the four anosognosic patients (b).

anterior parietal-temporal cortex junction and subcortical supramarginal gyrus (MNI coordinates: 49, 43, 24). Thus, white matter (MNI coordinates: 40, 42, 24). An additional both analyses pointed to the parieto-temporal junction and statistical analysis with a voxel-by-voxel chi-square proce- accord with other studies indicating that this region is dure (see Fig. 5b) showed a peak of lesion differences in the frequently lesioned in AHP and related disorders (Bisiach et al., 1986; Feinberg et al., 2010; Vocat et al., 2010). Table 1 e Number of each type of response for each participant. A response was classified as “Correct response” (if it corresponded to the correct word solution of the riddle), “False response” (if it did not correspond), 4. Discussion “Repetition” (if the same false response was given as for the preceding step), and “No response” (if the patient could By using a novel and simple “riddle” test, our study reveals not report any word). that patients with AHP present with a specific impairment in evaluating the certainty of their guesses and adjusting to new Correct False Repetition No response response response conflicting information, unlike other patients who also suffer from right-brain damage and left hemiplegia but show no Controls 19.64 20.00 5.00 5.36 anosognosia. Although our sample was relatively small, this Nosognosics 21.60 17.00 3.40 8.00 pattern was observed in group comparisons as well as in Anosognosics 13.75 23.75 10.50 2.00 single-case analyses, with at least a statistical tendency for cortex 49 (2013) 1771e1781 1777

Fig. 5 e Anatomical analysis comparing the lesions of anosognosic and nosognosic patients using a standard subtraction analysis (a) and a voxel-by-voxel chi-square procedure (b).

each and every patient (see Fig. 3b). These finding suggests trial and new information, rather than reject their current that deficits in cognitive monitoring mechanisms beyond in- beliefs and make a new guess. In other words, while nosog- ternal sensorimotor control may contribute to AHP. nosics or controls would usually change their mind and find In our task, the beginning of each riddle was fully uncertain another word that could suit better the succession of clues, and could not be expected to be successfully resolved. The anosognosics tended to persist with their preceding response first clue was always evasive enough to create, in healthy and find in the new clues any thin link that could justify the participants, a feeling of doubt that should make them pro- given word. Remarkably, during testing, we observed that pose a word with caution and have low confidence in their these patients did not simply persevere on their verbal re- initial response. These conditions of poor informative content sponses; but they often tended to give reasons to do so and in a guessing context were created with the purpose to verbally formulate ad hoc explanations. For example, for the “simulate” the degraded information about current state (due riddle illustrated in Fig. 1, an anosognosic patient answered to neglect, hemianopia, hemianesthesia, etc.) that is typically “bread” after the first clue, but then said, after the second clue: associated with AHP. Thus, we could observe, in controlled “it is just as I thought, it is small bread.”; and after argued: and comparable conditions, how healthy people, nosognosic “yes, like when we cut the bread, with the knife, it makes a stroke patients, and anosognosic patients would behave when regular noise.” These connections were often non-obvious asked to produce and evaluate guesses based on partial but their justification offered with some logic. It is therefore knowledge. Critically, however, the task and the riddles were unlikely that the impaired ability to change belief reflected a totally unrelated to motor functioning. Strikingly, we found purely cognitive deficit in reasoning. These issues echo some that during the first 3 clues, where informative content was of the debates in the current research on error monitoring in still poor, anosognosics exhibited an abnormally high over- healthy people, where the nature of error and feedback confidence in their initial response, instead of being cautious monitoring functions are still unclear and associated with and uncertain like controls and nosognosics were. Remark- both sensorimotor and motivational factors (Holroyd and ably, the confidence ratings of anosognosic patients after their Coles, 2002; Ullsperger and von Cramon, 2004), as we have first guess (mean ¼ 5.7 on a 0e8 scale) was already close to discussed elsewhere (Vocat and Vuilleumier, 2010). Moreover, maximum levels, and actually similar to those made by other this behavior is reminiscent of the fanciful justifications often participants on the penultimate (4th) clue (mean ¼ 5.6), just provided by anosognosic patients (e.g., “my hand is asleep.”) before resolution of the riddle. The nosognosic patients were when they are confronted with their weakness during much less certain about their first guess (mean ¼ 1.9). It thus neurological examination (Assal, 1983; Nathanson et al., 1952). appeared as if the anosognosics could not experience “doubt” Importantly, such tendency to overconfidence and persis- anymore. tence in current beliefs was not accompanied by a more The abnormal confidence in anosognosics was not only general impairment in executive functions, vigilance, spatial “immediate” when a new belief was formed about the target attention, or memory (as measured with our clinical bedside word, but it also failed to be appropriately revised when sub- testing). In particular, no evidence for greater deficits in sequent clues were provided and conflicted with the previous frontal lobe functions was found on clinical tests of reasoning response. Even in the presence of incongruent new informa- and mental flexibility (Weigl sorting test) or word fluency tion, anosognosics seemed to stick to their initial “false” be- (both phonological and semantic), which are typically highly liefs, rather than modify them. Significantly more often than sensitive to prefrontal lesions of various origins (Lezak et al., nosognosics or controls, anosognosic patients tended to 2012). Thus, although we did not use standard neuropsycho- repeat a same incorrect response when they had to deal with a logical tests of high-level executive function, there is no new clue that did not exactly confirm their first guess. They reason to think that our riddle task was more difficult, ab- typically preferred to find “non-obvious” but still “plausible” stract, or demanding on memory, visual-spatial, or linguistic connections between the false word provided on a preceding functions than some of the other bedside tasks; if anything, 1778 cortex 49 (2013) 1771e1781

the Weigl sorting test used here (requiring the flexible for- that AHP is typically transient after acute brain injury: as the mation of categories based on shape properties) would actu- patients progressively recover from their neurological deficits ally be anticipated to be more abstract and cognitively and receive more informative feedbacks on their current sate, challenging. Moreover, the present findings accord with pre- their awareness of paralysis increases and their anosognosia vious studies that reported no significant correlation between tends to disappear. This view also supports the potential anosognosia and several frontal lobe deficits (Anderson and benefits of daily confrontation with the deficit to enhance Tranel, 1989; Starkstein et al., 1992; Vocat et al., 2010). How- awareness and facilitate rehabilitation (Fotopoulou et al., ever, because our testing was limited by clinical constraints 2009; Ownsworth et al., 2006). due the acute neurological settings, it would interesting in On a more general level, our findings accord with the pro- future studies to explore whether the deficits observed here posal that some aspects of AHP after right-brain damage does or does not correlate with standard frontal tests as well might reflect the role of the left hemisphere as a “story-teller” as more specific decision tasks and self-monitoring tests that (Ramachandran, 1995) or “interpreter” (Gazzaniga, 2000) that are not routinely tested in the clinic. uses any available information in a given situation and builds Therefore, the specific impairment in evaluating and cor- the most plausible “story” with it (see also Assal, 1983). In this recting current belief states, as we observed here in anosog- process, relevant congruent information tends to be high- nosics, might point to some dysfunction in higher-level lighted while incongruent elements are ignored or sup- executive functions that depend on brain areas distinct from pressed, as if they were pulled out of consciousness. The role those concerned with standard frontal tasks. We suspect that of the right hemisphere (e.g., mediated by fronto-striatal or this more specific deficit might reflect damage to prefrontal insula circuits) might constitute a “devil’s advocate” that help cortical areas (Coltheart, 2010; Goel et al., 2009) or fronto- evaluate and reject inappropriate beliefs generated in the left striatal circuits (Samejima and Doya, 2007) that are known hemisphere, by signaling self-relevant errors and promoting to be recruited by self-monitoring, error detection, or set the resolution of uncertain inferences (Coltheart, 2010; Goel shifting conditions, and also frequently found to be damaged et al., 2009). In patients with anosognosia, the right hemi- in patients with AHP (Pia et al., 2004; Vocat et al., 2010). spheric lesion might disrupt this “devil’s advocate” function Damage to the right insula is also common in AHP (Vocat et al., and reduce the ability to check and change current beliefs 2010; Baier and Karnath, 2008) and might also contribute to about one’s own state despite the presence of incongruent abolish predictive affect signals related to uncertainty and information. Due to a lack of error signals and subsequent error risk (Preuschoff et al., 2008; Singer et al., 2009). Further check operations, the patients may be abnormally confident research is needed to better understand the exact cognitive in their beliefs and fail to integrate new information in order mechanisms underlying belief formation and verification, as to modify them when confronted with incongruency. Hence, well as their neural substrates (Harris et al., 2008). Interest- anosognosics might be unable to change their past beliefs ingly, functional neuroimaging work in healthy volunteers (“my left arm is fully functional”) simply because they have no suggests that judgments of disbelief (“false” vs “true” re- grounds to do so at both the sensory-motor and affective- sponses) made in response to statements from various do- motivational levels. This account of AHP partly converges mains (e.g., semantic “devious means friendly” or personal with the “discovery hypothesis” put forward by Levine et al. “you were born in New York”) correlate with activation of (1991), but we surmise that the discovery failure may reflect insula and anterior cingulate regions associated with error a more specific impairment in belief monitoring abilities monitoring (Harris et al., 2008; Kapogiannis et al., 2009), as rather than from a general cognitive defect or dementia (as well as areas in the right temporo-parietal junction recruited proposed by Levin et al.); and that an abnormal experience of by perspective taking and shifting (Sommer et al., 2007), all motor weakness may not only be caused by losses in propri- commonly damage in patients with AHP than without AHP oception but also influenced by other deficits including spatial (Vocat et al., 2010), particularly when AHP persists. Interest- neglect, impaired motor intention, amnesia, etc. (Vuilleumier, ingly, our own overlap analysis (see Fig. 5) showed that this 2004; Davies et al., 2005), perhaps in different combinations in region, the right temporo-parietal junction, was more different patients (see Vocat and Vuilleumier, 2010 for the frequently damaged in anosognosic patients (although these description of a multifactorial model of self-monitoring and results are difficult to interpret since they are based on a small anosognosia). This variety of factors may account for differ- sample of patients). ences in terms of the specificity and selectivity of AHP Importantly, we found that after receiving all 5 clues with commonly observed in the clinic (Cocchini et al., 2010; progressively increasing informative consent, anosognosics Fotopoulou et al., 2008; Marcel et al., 2004). Importantly, like all patients were able to find the correct word solution on the last previous studies investigating neuropsychological correlates step as often as the controls or nosognosics did. This suggests of AHP (e.g., Vocat et al., 2010; Marcel et al., 2004), our finding that anosognosics had no general problem in reasoning but of a strong association between anosognosia and a deficit in required a repeated signal of errors, or a larger incongruence beliefs updating does not demonstrate that the latter is causal. between a new clue and the previous guess, in order to prompt As proposed for other deficits, such impairment is likely to be a re-appraisal of their preceding responses and to trigger a neither sufficient not necessary to produce anosognosia but new solution. The threshold to evoke doubt about the current we suspect it could be crucial in some cases, particularly in belief might be pathologically increased in these patients, combination with other deficits disrupting a normal experi- presumably due to their lesion site or perhaps other factors ence of motor function (e.g., due to concomitant losses in including personality or past experiences (Weinstein and feedback or feedforward control mechanisms). Importantly, a Kahn, 1953). This observation is also consistent with the fact deficit in belief shifting alone would not be sufficient to cause cortex 49 (2013) 1771e1781 1779

anosognosia (for hemiplegia and/or other concomitant defi- cognitive substrates implicated in belief monitoring, and cits) but likely to provoke denial only (or predominantly) for a better understand their interaction with other deficits function for which direct information (e.g., motor intention or affecting motor monitoring in AHP. proprioceptive signals) is degraded or unavailable for the patient (see also Vuilleumier, 2004; Levine et al., 1991). Other recent theories about AHP have highlighted the importance of motor planning and motor intention disrup- Acknowledgments tions (Berti et al., 2005; Heilman, 1991; Vallar et al., 2003), or anomalies in body ownership and motor agency (Baier and This work is supported by grants from the Swiss National Karnath, 2008). These theories all agree with the fact that Science Foundation to PV (SNF 32003B-108367, GLAASS: elementary neurological deficits (in motor or sensory func- Geneva-Lausanne Anosognosia in Acute Stroke Study) and a tion) are not sufficient to fully explain AHP, but hypothesize fellowship from the E. Boninchi Foundation to RV. that some impairment in higher-level monitoring processes must be involved. In our study, however, we further suggest Appendix A. The 10 riddles that the latter high-level deficits may not necessarily imply specific functions related to motor actions or attention, but Riddle 1: Airplane can be observed in other domains when patients are con- I have wings fronted with partial and uncertain information e as demon- I can fly strated here in our new “riddle” test. Although this disruption I can move very fast alone is neither necessary nor sufficient to lead to anosog- I exist in various sizes nosia, we believe that it may provide a “fertile field”, for the I have wheel emergence of full-blown anosognosia when co-existing with other deficits (including in motor intention and comparator Riddle 2: Tooth processes) and that it could contribute to the severity and/or I’m white persistence of anosognosia. Likewise, using an ingenious re- I have deep roots ality monitoring task, Jenkinson et al. (2009) reported that both I need daily maintenance patients with AHP and those without AHP failed to discrimi- I’m covered with enamel nate between motor memories internally generated by in- Some children keep me for the little mouse tentions versus externally generated by actual execution, while only the patients with AHP additionally failed to judge Riddle 3: Carrot the reality of past visual imagination versus experience, sug- I am a food gesting a more general deficit in monitoring processes beyond I am very cultivated motor function only. I am a vegetable Our study is not without limitations. In particular, I am usually orange although we found consistent anomalies in confidence re- The rabbits adore me sponses in every patient with AHP (with at least a strong statistical tendency), our sample was relatively small (leading Riddle 4: Key to relatively low statistical power for some of our analysis). I can be of metal Testing acute stroke patients at the bedside with specific and I can be carry in a hand quantitative experimental tasks is challenging, and AHP tends It’s better not to lose me to disappear after a few days or even hours in many cases I protect against thieves (Vocat et al., 2010). Nonetheless, future studies need to recruit I allow people to return home larger groups with different degrees of anosognosia and different types of neurological deficits, as well as to explore Riddle 5: Cow the role of more specific decision making processes. Having I am an animal of the female gender established this phenomenon of overconfidence in a guessing I am sometimes black and white task for the first time, we hope that future research will I can be sacred investigate and confirm similar behaviors in other neuropsy- Sometimes I wear a bell chological syndromes associated with anosognosia (e.g., I chew when I have nothing else to do amnesia, blindness). Preliminary data with our “riddle task” administered to an anosognosic patient with cortical blind- Riddle 6: Heart ness suggest the exact same pattern of findings: over- My weights is approximatively 300 g confidence and difficulties to update wrong beliefs. Thus, with I produce a regular sound a different kind of anosognosia, with no relation to motor Sports makes me feel excited functions and a different distribution of lesions, the same I am usually on the left rather on the right side higher-level cognitive processes associated with belief for- Lovers often draw me mation and verification seemed also to be crucially involved. The role of such processes has certainly been underestimated Riddle 7: Shadow and understudied until recently. Therefore, a major goal for Without the sun, I could not survive future studies on anosognosia should be to clarify the neuro- You often takes advantage of me when it’s hot 1780 cortex 49 (2013) 1771e1781

You can’t lose me or catch me Cocchini G, Beschin N, Fotopoulou A, and Della Sala S. Explicit I can tell you the time and implicit anosognosia or upper limb motor impairment. e I look like you traits to traits Neuropsychologia, 48(5): 1489 1494, 2010. Coltheart M. The neuropsychology of delusions. Annals of New York Academy Science, 1191: 16e26, 2010. Riddle 8: Garden Davies M, Davies AA, and Coltheart M. Anosognosia and the two- I need sun, water and fresh air factor theory of delusions. Mind and Language, 20(2): 209e236, I love flowers but also vegetables 2005. I change with the seasons Feinberg TE, Venneri A, Simone AM, Fan Y, and Northoff G. The I can be hanging or secret neuroanatomy of asomatognosia and somatoparaphrenia. e Sometimes, professionals take care of me Journal of Neurology, Neurosurgery and Psychiatry, 81(3): 276 281, 2010. Folstein MF, Folstein SE, and McHugh PR. “Mini-mental state”. A Riddle 9: Matches practical method for grading the cognitive state of patients for I can be in a pocket or in a box the clinician. Journal of Psychiatric Research, 12(3): 189e198, I change color after use 1975. I fear water Fotopoulou A, Rudd A, Holmes P, and Kopelman MD. Self- My red head can cause great damage observation reinstates motor awareness in anosognosia of hemiplegia. Neuropsychologia, 47(5): 1256e1260, 2009. I can be replaced by the shock of 2 stones Fotopoulou A, Tsakiris M, Haggard P, Vagopoulou A, Rudd A, and Kopelman M. The role of motor intention in motor awareness: Riddle 10: Broom An experimental study on anosognosia for hemiplegia. Brain, I am a current object 131(12): 3432e3442, 2008. I have a rigid part, the other soft Frith CD, Blakemore SJ, and Wolpert DM. Abnormalities in the The cleaners use me frequently awareness and control of action. Philosophical Transactions of e I am essential to witches Royal Society London Biology Science, 355(1404): 1771 1788, 2000. Gazzaniga MS. Cerebral specialization and interhemispheric My modern competitor is the vacuum cleaner communication: Does the corpus callosum enable the human condition? Brain, 123(7): 1293e1326, 2000. Gilmore RL, Heilman KM, Schmidt RP, Fennell EM, and Quisling R. e references Anosognosia during Wada testing. Neurology, 42(4): 925 927, 1992. Goel V, Stollstorff M, Nakic M, Knutson K, and Grafman J. A role for right ventrolateral prefrontal cortex in reasoning about Adair JC, Gilmore RL, Fennell EB, Gold M, and Heilman KM. indeterminate relations. Neuropsychologia, 47(13): 2790e2797, Anosognosia during intracarotid barbiturate anesthesia: 2009. Unawareness or amnesia for weakness. Neurology, 45(2): Harris S, Sheth SA, and Cohen MS. Functional neuroimaging of 241e243, 1995. belief, disbelief, and uncertainty. Annals of Neurology, 63(2): Anderson SW and Tranel D. Awareness of disease states following 141e147, 2008. cerebral infarction, dementia, and head trauma: Standardized Heilman KM. Anosognosia: Possible neuropsychological assessment. Clinical Neuropsychologist, 3(4): 327e339,1989. mechanisms. In Prigatano G and Schacter DL (Eds), Awareness Assal G. Non, je ne suis pas paralyseé, c’est la main de mon mari. of Deficit After Brain Injury. Clinical and Theoretical Issues. New Schweizer Archiv fu¨r Neurologie, Neurochirurgie, und Psychiatrie, York: Oxford Universitary Press, 1991. 133(1): 151e157, 1983. Holroyd CB and Coles MG. The neural basis of human error Babinski J. Contribution a` l’e´tude des troubles mentaux dans processing: Reinforcement learning, dopamine, and the error l’he´miple´gie organique (anosognosie). Revue Neurologique, 27: related negativity. Psychological Review, 109(4): 679e709, 2002. 845e848, 1914. Jenkinson PM, Edelstyn NM, Drakeford JL, and Ellis SJ. Reality Baier B and Karnath HO. Incidence and diagnosis of anosognosia monitoring in anosognosia for hemiplegia. Conscious Cognition, for hemiparesis revisited. Journal of Neurology Neurosurgery and 18(2): 458e470, 2009. Psychiatry, 76(3): 358e361, 2005. Kapogiannis D, Barbey AK, Su M, Zamboni G, Krueger F, and Baier B and Karnath HO. Tight link between our sense of limb Grafman J. Cognitive and neural foundations of religious ownership and self-awareness of actions. Stroke, 39(2): belief. Proceeding National Academy of Sciences of the United States 486e488, 2008. of America, 106(12): 4876e4881, 2009. Berti A, Bottini G, Gandola M, Pia L, Smania N, Stracciari A, et al. Karnath HO, Baier B, and Nagele T. Awareness of the functioning Shared cortical anatomy for motor awareness and motor of one’s own limbs mediated by the insular cortex? Journal of control. Science, 309(5733): 488e491, 2005. Neuroscience, 25(31): 7134e7138, 2005. Bisiach E, Vallar G, Perani D, Papagno C, and Berti A. Unawareness Karnath HO, Fruhmann Berger M, Ku¨ ker W, and Rorden C. The of disease following lesions of the right hemisphere: anatomy of spatial neglect based on voxelwise statistical Anosognosia for hemiplegia and anosognosia for hemianopia. analysis: A study of 140 patients. Cerebral Cortex, 14(10): Neuropsychologia, 24(4): 471e482, 1986. 1164e1172, 2004. Bottini G, Paulesu E, Gandola M, Pia L, Invernizzi P, and Berti A. Levine DN, Calvanio R, and Rinn WE. The pathogenesis of Anosognosia for hemiplegia and models of motor control: anosognosia for hemiplegia. Neurology, 41(11): 1770e1781,1991. Insights from lesional data. In Prigatano G (Ed), The Study of Lezak MD, Howieson DB, Bigler ED, and Tranel D. Anosognosia. New York: Oxford University Press, 2010. Neuropsychological Assessment. 5th ed. USA: OUP, 2012. Breier JI, Adair JC, Gold M, Fennell EB, Gilmore RL, and Lu LH, Barrett AM, Cibula JE, Gilmore RL, Fennell EB, and Heilman KM. Dissociation of anosognosia for hemiplegia and Heilman KM. Dissociation of anosognosia and phantom aphasia during left-hemisphere anesthesia. Neurology, 45(1): movement during the Wada test. Journal of Neurology 65e67, 1995. Neurosurgery and Psychiatry, 69(6): 820e823, 2000. cortex 49 (2013) 1771e1781 1781

Lu LH, Barrett AM, Schwartz RL, Cibula JE, Gilmore RL, De´mences et syndromes de´mentiels. Approche neuropsychologique. Uthman BM, et al. Anosognosia and confabulation during the Paris: Masson, 1991. Wada test. Neurology, 49(5): 1316e1322, 1997. Turnbull OH, Jones K, and Reed-Screen J. Implicit awareness of Marcel AJ, Tegner R, and Nimmo-Smith I. Anosognosia for plegia: deficit in anosognosia? An emotion-based account of denial of Specificity, extension, partiality and disunity of bodily deficit. Neuropsychoanalysis,4:69e86, 2002. unawareness. Cortex, 40(1): 19e40, 2004. Ullsperger M and von Cramon DY. Neuroimaging of performance Nathanson M, Bergman PS, and Gordon CG. Denial of illness: Its monitoring: Error detection and beyond. Cortex, 40(4e5): occurrence in one hundred consecutive cases of hemiplegia. 593e604, 2004. Archives of Neurology and Psychiatry, 68(3): 380e387, 1952. Vallar G, Bottini G, and Sterzi R. Anosognosia for left-sided motor Ogden JA. Anterior-posterior interhemispheric differences in the and sensory deficits, motor neglect, and sensory loci of lesions producing visual hemineglect. Brain and hemiinattention: Is there a relationship? Progress in Brain Cognition, 4(1): 59e75, 1985. Research, 142: 289e301, 2003. Ownsworth T, Fleming J, Desbois J, Strong J, and Kuipers P. A Venneri A and Shanks MF. Belief and awareness: Reflections on a metacognitive contextual intervention to enhance error case of persistent anosognosia. Neuropsychologia, 42(2): awareness and functional outcome following traumatic brain 230e238, 2004. injury: A single-case experimental design. Journal of Verdon V, Schwartz S, Lovblad KO, Hauert CA, and Vuilleumier P. International Neuropsychology Society, 12(1): 54e63, 2006. Neuroanatomy of hemispatial neglect and its functional Pia L, Neppi-Modona M, Ricci R, and Berti A. The anatomy of components: A study using voxel-based lesion-symptom anosognosia for hemiplegia: A meta-analysis. Cortex, 40(2): mapping. Brain, 133(3): 880e894, 2010. 367e377, 2004. Vocat R, Staub F, Stroppini T, and Vuilleumier P. Anosognosia for Preuschoff K, Quartz SR, and Bossaerts P. Human insula hemiplegia: A clinical-anatomical prospective study. Brain, activation reflects risk prediction errors as well as risk. Journal 133(12): 3578e3597, 2010. of Neuroscience, 28(11): 2745e2752, 2008. Vocat R and Vuilleumier P. Neuroanatomy of impaired body Ramachandran VS. Anosognosia in parietal lobe syndrome. awareness in anosognosia and hysteria: A multi-component Conscious Cognition, 4(1): 22e51, 1995. account. In Prigatano G (Ed), The Study of Anosognosia. New Rorden C and Brett M. Stereotaxic display of brain lesions. York: Oxford University Press, 2010. Behavioural Neurology, 12(4): 191e200, 2000. Vuilleumier P. Anosognosia. In Bogousslavsky J and Cummings JL Samejima K and Doya K. Multiple representations of belief states (Eds), Behavior and Mood Disorders in Focal Brain Lesions. and action values in corticobasal ganglia loops. Annals of New Cambridge University Press, 2000. York Academy Science, 1104: 213e228, 2007. Vuilleumier P. Anosognosia: The neurology of beliefs and Schenkenberg T, Bradford DC, and Ajax ET. Line bisection and uncertainties. Cortex, 40(1): 9e17, 2004. unilateral visual neglect in patients with neurologic Wechsler D. Wechsler Adult Intelligence Scale-Revised (WAIS-R): impairment. Neurology, 30(5): 509e517, 1980. Manual. San Antonio: TX: The Psychological Corporation, Singer T, Critchley HD, and Preuschoff K. A common role of insula 1981. in feelings, empathy and uncertainty. Trends in Cognitive Weigl E. Zur Psychologie sogenannter Abstraktionsprozesse. Sciences, 13(8): 334e340, 2009. Zeitschrift fu¨r Psychologie, 103: 2e45, 1927. Sommer M, Dohnel K, Sodian B, Meinhardt J, Thoermer C, and Weinstein EA and Kahn RL. Personality factors in denial of illness. Hajak G. Neural correlates of true and false belief reasoning. Archives of Psychiatry and Neurology, 69(3): 355e367, 1953. NeuroImage, 35(3): 1378e1384, 2007. Weinstein EA and Kahn RL. Denial of Illness. Symbolic and Starkstein SE, Fedoroff JP, Price TR, Leiguarda R, and Robinson RG. Physiological Aspects. Springfield, Illinois, USA: Charles C. Anosognosia in patients with cerebrovascular lesions. A study Thomas publisher, 1955. of causative factors. Stroke, 23(10): 1446e1453 , 1992. Wilson B, Cockburn J, and Halligan P. Development of a Thuillard F and Assal G. Donneés neuropsychologiques chez le behavioral test of visuospatial neglect. Archives of Physical sujet aˆge´ normal. In Habib M, Joanette Y, and Puel M (Eds), Medicine and Rehabiliation, 68(2): 98e102, 1987.