Updating impairments and the failure to explore new hypotheses following right brain damage

Updating impairments and the failure to explore new hypotheses following right brain damage We have shown recently that damage to the right hemisphere impairs the ability to update mental models when evidence suggests an old model is no longer appropriate. We argue that this deficit is generic in the sense that it crosses multiple cog- nitive and perceptual domains. Here, we examined the nature of this updating impairment to determine more precisely the underlying mechanisms. We had right (RBD, N = 12) and left brain damaged (LBD, N = 10) patients perform versions of our picture-morphing task in which pictures gradually morph from one object (e.g., shark) to another (e.g., plane). Performance was contrasted against two groups of healthy older controls, one matched on age (HCO-age-matched, N = 9) and another matched on general level of cognitive ability (HCO-cognitively-matched, N = 9). We replicated our earlier findings showing that RBD patients took longer than LBD patients and HCOs to report seeing the second object in a sequence of morphing images. The groups did not differ when exposed to a morphing sequence a second time, or when responding to ambiguous images outside the morphing context. This indicates that RBD patients have little difficulty alternating between known representations or labeling ambiguous images. Instead, the difficulty lies in generating alternate hypotheses for ambiguous information. Lesion overlay analyses, although speculative given the sample size, are consistent with our fMRI work in healthy individuals in implicating the anterior insular cortex as critical for updating mental models. Keywords Right brain damage · Mental representations · Updating failures · Picture morphing · Exploration Introduction Damage to the right hemisphere results in a range of het- erogeneous impairments, including impoverished spatial James Danckert and Britt Anderson contributed equally to the attention (i.e., neglect; Danckert et al. 2012a), statistical manuscript. learning (Shaqiri and Anderson 2013; Shaqiri et al. 2013), humor appreciation (Brownell et al. 1983), working memory * Elisabeth Stöttinger capacity (Ferber and Danckert 2006; Husain et al. 2001), and elisabeth.stoettinger@sbg.ac.at deficient Theory of Mind (Happé et al. 1999; Griffin et al. Carolyn Louise Guay 2006; Weed et al. 2010). We have argued that these impair- clguay@uwaterloo.ca ments can be parsimoniously explained by a general impair- James Danckert ment in updating mental models (Danckert et al. 2012a, b; jdancker@uwaterloo.ca Filipowicz et al. 2016; Geng and Vossel 2013; Shaqiri and Britt Anderson Anderson 2013; Shaqiri et al. 2013; Stöttinger et al. 2014; britt@uwaterloo.ca Vocat et al. 2012). Department of Psychology, Center for Cognitive Everyday we are confronted with an enormous amount Neuroscience, University of Salzburg, Hellbrunnerstrasse of information. Mental models allow us to deal with this 34, 5020 Salzburg, Austria complexity by compactly representing the regularities that Department of Psychology, University of Waterloo, govern our environment. We use these models to guide our Waterloo N2L 3G1, Canada decisions (e.g., Is the situation dangerous? Is this food edi- Centre for Theoretical Neuroscience, University of Waterloo, ble? Johnson-Laird 2004; Griffiths and Tenenbaum 2012; Waterloo N2L 3G1, Canada Vol.:(0123456789) 1 3 1750 Experimental Brain Research (2018) 236:1749–1765 Tenenbaum et al. 2011). Since the world is in flux, these From our prior study, we know that RBD patients notice models are only useful if we update them when things small changes in the sequences of morphing images. RBD change. While a salient shift in the environment allows us patients, however, incorporated changes into their initial per- to simply react (e.g., a bolt of lightning indicates that it is ceptual representation (e.g., “The shark’s fins are getting time to head indoors), gradual changes in the environment bigger.”), rather than entertaining an entirely novel interpre- require the aggregation of mismatching evidence to provoke tation (e.g., fins have now become wings). Similarly, within a proactive decision (e.g., when does cloud cover become the context of the rock, paper, scissors game, although RBD sufficiently menacing to indicate a coming storm? McGuire patients fail to update to a change in strategy, their own play et al. 2014). choices do reflect the fact that they have noticed something We previously showed that right brain damage resulted has changed. That is, RBD patients rapidly abandon their old in selective impairment of gradual updating in response to play strategy but fail to adopt a new strategy optimized for small, subtle environmental changes. We had left (LBD) the change in their opponent’s bias (Stöttinger et al. 2014). and right brain damaged (RBD) patients play the children’s In combination, these data suggest that the impairment in game of ‘rock, paper, scissors’ (RPS) against a computer updating following RBD may be due to either (1) a failure opponent. The computer initially played randomly before to explore alternative interpretations in light of observed switching to ‘paper’ 80% of the time. While LBD patients changes or (2) a general inability to proactively switch. and controls rapidly adopted the optimal strategy (i.e., scis- To evaluate these hypotheses, RBD and LBD patients as sors), RBD patients failed to respond to the transition, with well as healthy seniors (HCO) were presented with three most patients continuing to play randomly (Danckert et al. versions of our picture morphing task. The hypotheses and 2012b). Similarly, RBD patients showed a severe updating predictions for this study are summarized in Table 1. In the impairment in a gradually morphing picture task, in which gradual condition participants saw four picture sets in which one object (e.g., shark) morphed over several iterations into one object (e.g., shark) morphed into a different object a completely different object (e.g., plane; Fig.  2, top panel). (e.g., plane; replication of Stöttinger et  al. 2014; Fig.  2, The logic was to first provide participants with a mental top row). In the repeat condition, some of these sequences model (e.g., “It’s a shark”) and then evaluate how much were repeated in the reverse order (e.g., plane-to-shark). In evidence was required for them to update to a new mental the random condition, images of four additional morphing model (e.g., “It’s a plane now”). RBD patients needed sig- sequences were presented in random order with all pictures nificantly more evidence (i.e., more pictures) before they intermixed. For example, a participant would see the sec- reported seeing the second object compared to LBD patients ond picture of the gun-hair dryer picture set, followed by and healthy controls. Performance in the RPS task correlated the ninth picture of the spider-sun picture set, and so on with performance in the picture morphing task, indicating (Fig. 2, middle row). We compared the likelihood that par- a general updating impairment across multiple cognitive ticipants reported the first or second object for pictures based domains (Stöttinger et al. 2014). on their order in the morphing sequence (i.e., 1st to 15th While our prior studies showed an updating impair- picture position), and on the type of sequence (i.e., gradual, ment after RBD, they did not unequivocally demonstrate repeat or random). the mechanism for this impairment. Updating to gradually Each of our key hypotheses was addressed in the follow- evolving stimuli requires several different processes: (1) we ing ways: need to be able to detect and accumulate evidence of a mis- match. (2) When the mismatch reaches some threshold we 1. To evaluate whether updating impairments seen in RBD need to explore for potential alternative interpretations. (3) patients can be explained by a failure to explore, we Finally, we need to proactively switch to this new interpre- calculated the difference in performance between the tation. The aim of the research reported here was to test for random–gradual conditions. We know that healthy indi- the integrity of each of these mechanisms in RBD patients. viduals tend to switch categories earlier in the morph Table 1 Expected outcome vs. Exploration Proactive switch Perceptual/attentional bias actual outcome Test (condition) ΔRandom–gradual Repeat Random Expected RBD < LBD&HCO RBD > LBD&HCO RBD ≠ LBD & HCO Actual RBD < LBD&HCO ✓ RBD = LBD&HCO ✗ (?) RBD = LBD & HCO ✗ Summary of expected vs. actual results for all three hypotheses (i.e., exploration, proactive switch, percep- tual/attentional bias) potentially explaining the updating impairment seen in RBD patients. The second row refers to the condition that was used to evaluate the hypothesis. Please note that results are only displayed for HCO matched on cognitive status 1 3 Experimental Brain Research (2018) 236:1749–1765 1751 series when presented in a gradual context (e.g., when rests on the accumulation of gradual changes in a noisy the object is still 60% shark and only 40% plane) than environment; participants proactively decide at which when single images from the same sequence are pre- point their current model is no longer supported by the sented in isolation, outside the morphing context (Egré evidence. Indeed, we found that while RBD patients et al. 2013; Raffman 2011; Stöttinger et al. 2016; see showed a selective updating impairment, there was no Egré et al. 2018 for review). This suggests that healthy difference between RBD and LBD patients in a card individuals are exploring alternative interpretations sorting test akin to the WCST (Danckert et al. 2012b; conditional on the evolving history they have viewed Stöttinger et al. 2014; Piper et al. 2012). (e.g., “I know it was a shark but what else could it be? 3. Lastly, to assess the contribution of general post-stroke A bird?”). In the context of the rock–paper–scissors visual or attentional impairments (Hepworth et al. 2015), game, we found that RBD patients explored a lim- we compared the performance of RBD patients with the ited selection of alternative strategies (Danckert et al. performance of control groups in the random condition. 2012b) and employed less efficient exploration strate- Damage to the right hemisphere often results in an atten- gies (Sepahvand et al. 2014). If updating impairments tional bias towards the ipsilesional side (i.e., neglect; in RBD patients are due to a generic failure to explore Danckert et al. 2012a). This could have potentially hin- alternatives evident across different tasks and domains, dered the capacity of RBD patients to correctly identify we expect them to show no benefit from the contextual small changes in our original study, and consequently information provided by the gradual morphing sequence to update to a new model. If the updating impairment (Egré et al. 2013; Raffman 2011; Stöttinger et al. 2016; in RBD patients is due to a general perceptual and/or Table 1, second row). attentional impairment, we expect RBD patients to per- 2. To evaluate whether RBD patients would always strug- form differently than LBD patients when pictures are gle to proactively shift to a new interpretation, (irrespec- presented individually outside of the morphing context tive of whether they are naïve as to the second object), (i.e., random condition; Table 1, right column). we compared performance in the repeat condition across participant groups. Evidence that RBD impairs the abil- One major assumption of our study is that the benefit ity to make proactive decisions comes from research for the gradual over the random presentation sequence is using ambiguous or bistable figures such as the Rubin’s indicative of active exploration. While this benefit has also face/vase picture or the Necker cube in which healthy been reported elsewhere in the literature (see Egré et al. individuals typically alternate between two mutually 2018 for review), to our knowledge our study was the only exclusive interpretations (see Long and Toppino 2004 one that showed this benefit in the context of morphing for review). Given that the sensory input is stable, the objects (Stöttinger et al. 2016). In the original study, we switch to a new interpretation is internally generated used a between subjects design. We therefore first evalu- by the participant. Research shows that performance ated in an online study whether we can replicate the effect in an ambiguous figures task of this kind is associated when the type of presentation (gradual vs. random) is with the right hemisphere, with damage to right frontal manipulated within participants—a non-trivial point given cortex resulting in significant impairment in switching the ongoing discussion about failures to replicate in psy- to the second interpretation (Meenan and Miller 1994). chological science (Bohannon 2015). Furthermore, right frontal and parietal areas are active during proactive alternations between two percepts (Britz et al. 2009, 2011; Sterzer and Kleinschmidt 2007; Weilnhammer et al. 2013; Zaretskaya et al. 2010). So Methods if RBD patients have a general impairment in proac- tively switching to a new interpretation irrespective of Prior to the patient study, we conducted an online study to whether or not they know what the second object will determine the benefit of gradual presentations in a within be, we expect them to identify the second object later subject design (Stöttinger et al. 2016). The same picture compared to other participant groups in the repeat condi- sets used in the online study were then used in the patient tion (Table 1, third row). study. Besides the need to replicate the effect, the online At this point, it is also worth noting that gradual study also allowed us to have a reference point for the updating should not be confused with set-shifting. performance of healthy, younger individuals. Results of While participants have to react to a salient mismatch the online study are included in the graphs depicting the in the Wisconsin Card Sorting Test (WCST; i.e., an results of patients and healthy seniors for comparison action that was considered “correct” suddenly becomes purposes. “incorrect”; Grant and Berg 1948), updating in our tasks 1 3 1752 Experimental Brain Research (2018) 236:1749–1765 one of the healthy elders was found to have had a past neu- Participants rological illness, and was excluded from further analyses. The final sample of 18 participants consisted of nine who Online study were 70 years or younger (HCO-age-matched: 9 females, mean age 66.35 years, ± 3.35) and nine older than 70 years Seventy-seven participants (33 female) recruited through Mechanical Turk, and between 19 and 53  years of age (HCO-cognitively-matched: 6 female, mean age 78.55 years, ± 3.32). The HCO-cognitively-matched control group had a (mean 35.34, SD 9.62) participated in this study [Cauca- sian/white (80.50%), Hispanic (7.8%), African American nominally lower MOCA score than the HCO-age-matched group, but the two control groups did not differ statisti- (5.2%), East Asian (3.9%)]. Participants received $1.50 for their participation. cally {HCO-cognitively-matched mean = 25.78, ± 2.44 vs. HCO-age-matched mean = 27.78, ± 1.86; [t(16) = 1.96, Patient study p = .068]}. Six of the HCO had a MoCA score between 25 and 22, potentially indicating a mild cognitive impairment Four groups were tested in this study—RBD patients, LBD (Smith et al. 2007; Nasreddine et al. 2005; but see; Lee et al. 2008 and; Luis et al. 2009 for the argument of lower cut-off patients, younger (≤ 70 years; i.e., HCO-age-matched) and older (> 70 years; i.e., HCO-cognitively-matched) healthy points). The MoCA for LBD (mean 24.79, ± 2.36) and RBD seniors. Patients were recruited from the Neurological Patient Database of the University of Waterloo (Heart and patients (mean 24.50, ± 2.84) was lower than that of the HCO-age-matched group (all p’s < .01), but comparable to Stroke Foundation funded). Of the 13 LBD patients three were excluded due to Montreal Cognitive Assessment that of the HCO-cognitively-matched group (all p’s > .25). There was no significant difference in MoCA scores of LBD (MoCA; Nasreddine et al. 2005) scores in the demented range (N = 2; MoCA ≤ 9). This score was considerably and RBD patients [t(20) = 0.18, p = .86]. LBD and RBD were of comparable age [t(20) = 0.87, p = .40]. Both groups lower than the optimal cutoff point for vascular demen- tia (i.e., MoCA < 17; Freitas et al. 2012). One patient was were significantly younger than the HCO-cognitively- matched group (all p’s > .01) but comparable in age to the excluded due to a failure to find any discernible lesion on available brain scans (N = 1). The final sample of LBD HCO-age-matched group (all p’s > .30). Also, there was no significant difference for time since stroke or lesion volume patients comprised ten patients (2 female, mean age 61.28 years, ± 14.47). Nine of these patients were stroke patients. between the RBD and LBD patients (all p’s > .05). The University of Waterloo’s Office of Research Ethics In one patient (#835), brain damage was due to a resection of an arterial vascular malformation. Twelve RBD patients approved the protocol for both studies—online and patient study. Participants on both studies gave informed written (2 female, mean age 65.88 years ± 10.13) participated. All of them were stroke survivors. Demographics for all patients consent prior to participation according to the Declaration of Helsinki, by either clicking on the “I agree” button (online are in Table 2 with lesions shown in Fig. 1. Patients were screened for neglect upon admission to the database and study), or by signing the consent letter (patient study). again prior to the experiment using the Behavioral Inat- tention Test (BIT; Wilson et al. 1987). Six RBD patients Stimuli and design showed neglect at initial screening, with two showing chronic neglect at testing (#284, #744). No LBD patient Each participant saw eight picture sets selected from a larger set validated in an earlier study (Stöttinger et al. 2016; https showed neglect. Six of the ten LBD patients were reported to have aphasia when admitted to the database. One LBD ://osf.io/qi2vg/ ). All images were a standard size (316 × 316 pixels) displayed on a white background. Four sets were patient experienced slight word finding difficulties at testing. He was able to describe all the objects and his answers could presented in a gradual order, four in random order. In the gradual condition one common object morphed over 15 iter- be reliably coded as either the first or the second object. Nineteen healthy older controls were recruited from the ations into a different object (Stöttinger et al. 2014, 2016). Two of these sets were presented again, but in reverse order University of Waterloo’s Research in Aging Participant Pool. This pool recruits community dwelling seniors for participa- (i.e., repeat condition). In the random condition, four picture sets were pre- tion in studies on aging. Thus, our HCO were enrolled based on age. We subdivided our healthy older controls into two sented in a random order. Pictures from these sets were randomly assigned to four new series with 15 pictures subgroups. One matched to our clinical participants based on age, and another based on MOCA scores to equate cognitive in each. Each random series contained the same number of pictures from each picture set and each part within a ability. The former group (HCO-age-matched) was under 70 years of age, while the second subgroup (HCO-cogni- series. That is, each random series had five pictures from the first third of the morphing series (pictures #1 to #5); tively-matched) was over 70 years of age. After recruitment 1 3 Experimental Brain Research (2018) 236:1749–1765 1753 Table 2 Demographics for patients (a) and HCO (b) ID Lesion volume Age (years) Time since stroke MoCA Gender Education (years) (years) (a) Demographics (patients)  LBD patients   110 65 64 8.67 23 Male 12   442 8127 68 12.44 23 Male 14   588 274 68 3.25 28 Male 20   788 328 35 1.84 22 Male 15   828 7374 76 1.44 25 Female 17   835 7573 35 1.39 25 Male 14   838 105 69 1.41 26 Male 23   872 665 71 0.89 23 Male 10   898+ 5396 66 0.48 23 Male 8   902+ 161 61 1.76 29 Female 12  RBD patients   27 21,187 49 8.37 27 Male 15   205 3742 62 6.69 29 Male 14   228 6316 86 6.62 26 Female 11   284 14,203 74 6.95 22 Female 12   489 1693 71 4.36 24 Male 15   729 5758 66 2.30 22 Male 14   744 21,479 73 3.60 25 Male 24   792 5878 64 1.67 24 Male 15   856 232 55 0.87 26 Male 16   874 5026 56 1.16 26 Male 14   932 1520 62 0.26 25 Male 16   946 5694 72 0.20 18 Male 20 ID Age (years) MoCA Gender Education (years) (b) Demographics (healthy controls)  Younger healthy controls (≤ 70 years)   1 61 28 Female 20   46 64 28 Female 17   110 69 27 Female 17   143 68 30 Female 16   193 62 30 Female 12   230 69 28 Female 18   249 70 25 Female 24   351 67 25 Female 16   408 68 29 Female 16  Older healthy controls (> 70 years)   3 76 25 Male 17   32 76 22 Female 13   37 76 27 Female 17   148 81 23 Female 22   206 80 26 Male 20   208 81 28 Female 14   321 81 30 Female 18   369 83 25 Female 13   409 73 26 Male 16 a b Neglect at time of screening ( and at time of testing) 1 3 1754 Experimental Brain Research (2018) 236:1749–1765 Fig. 1 Lesion tracings RBD (a) and LBD patients (b) superimposed on the MNI template. Lesions shown in neurological convention five pictures from the middle (pictures #6 to #10), and five The presentation of each set (gradual or random), the pictures from the end of each series (pictures #11 to #15). order of presentation (random then gradual or vice versa), In each of the eight picture sets, one additional object was as well as order of the series (series 1, 2, 3, 4, 1, 3 vs. 3, presented after the third and twelfth pictures as catch tri- 1, 4, 2, 1, 3), was varied between participants resulting als. Catch trials assessed whether participants were simply in eight versions. Versions were counterbalanced across perseverating (Fig. 2; Stöttinger et al. 2016). participants. 1 3 Experimental Brain Research (2018) 236:1749–1765 1755 Fig. 1 (continued) 1 3 1756 Experimental Brain Research (2018) 236:1749–1765 Fig. 2 Objects morphed over 15 iterations from object 1 into object phing context (e.g., second picture of the gun-hair dryer picture set 2 (gradual condition), or vice versa (repeat condition). In the random the ninth picture of the spider-sun picture set, etc.) condition, objects were presented individually, outside of the mor- sets were removed from further analysis. Importantly, per- Procedure formance in the current online study was the same as in Stöttinger et al. (2016) (all p’s > .05). Online study Patient study The online study questionnaires were designed using Qual- trics©. Participants first filled out demographic questions All assessments and tasks were done in the same order for before being assigned to one of eight task versions using the each patient: BIT, MoCA, picture morphing task. Partici- randomize function of Qualtrics©, and yielding roughly the pants were tested individually in a room at the University same number of participants for each task order (n = 39 did of Waterloo. In a few cases, participants were tested at their random, gradual, repeat and 38 did gradual, repeat, random). own home. Instructions were presented on a screen and Prior to the gradual condition, participants were informed repeated verbally to the participants before each condition they would see six different picture series, containing 17 in the picture morphing task. Participants saw one picture images and that each series would begin with a commonly at a time and were asked to tell us for each picture what known object before changing gradually to show a different they saw. Answers were recorded and transcribed after the object by the end of the series. In the random condition, experiment. Each picture was preceded and followed by a participants were told they would see 68 pictures of objects. fixation cross. Patients and healthy participants received $10 In all conditions, participants saw one picture at a time and per hour of study for their participation. typed their answer underneath the picture. The next picture was revealed by clicking a button at the bottom of the screen. Data analysis On average, participants needed 20 min (± 15 min) to com- plete the questionnaire. Time to complete the questionnaire Behavioral data did not correlate with any of the dependent measures in our task (all p’s > .05). Responses were coded as seeing the “first object”, or seeing There were a small number of technical glitches (< 1% an object other than the first object (i.e., “second object”). In of all image presentations). For 29 image presentations addition, we evaluated whether the catch trials were identi- (0.22%), the wrong picture was presented at the first picture fied correctly. For repeated sequences, the coding as first or position of a series (e.g., a saw instead of a shark). From second object was done based on the ordering used in the our initial study, we know that these pictures are never rated morphing sequence. For example, when a picture set mor- as the “second object”. These items were therefore coded phed from a shark to a plane in the gradual condition, and as “first object” reports. In one version, (a picture of a saw from a plane to a shark in the repeat condition, “shark” was was presented instead of the correct picture at picture posi- coded as first object in the gradual condition, with “plane” tion #10). If this happened between a switch of percepts, coded as first object in the repeat condition. While the cod- the set was removed (seven individual cases). Otherwise, ing was not done blindly, it did use a list of validated pic- it was rated as the answer which preceded and followed the ture terms from Stöttinger et al. (2016; https ://osf.io/qi2vg erroneous presentation. Infrequently, participants failed to /), and was done independently by the first two authors answer (i.e., 0.19% of images). Most of these omissions were with an interrater agreement of 98.98% for the online study preceded and followed by the same response, suggesting no and 99.13% for the patient study. The high agreement is change in the participant’s conscious percept. In ten cases, explained by the fact that there was little opportunity for the omission occurred between a switch in report. These 1 3 Experimental Brain Research (2018) 236:1749–1765 1757 ambiguity. The majority of answers in the online study Statistical analysis were less than three words (one word, e.g., “frog”; 90.45%; two words, e.g., “Jumping man”; 8.81%; or three words, Data were analyzed using repeated measures ANOVA. In the e.g., “man jumping up”; 0.47%). In only 0.28%, the answer online study, the mean percentages of r fi st object reports (aver - included more than three words. Most answers in the patient aged over all sets per condition) were submitted to a repeated study were also less than four words (73%). Participants measures ANOVA with image number (15 morphing images in the patient study, however, were more prone to longer from 100% first object to 0% first object) and condition (grad- answers with 9.44% containing ten words or more. ual vs. random) as within subject factors. Separate repeated On a few occasions (2.18%), patient responses indicated measure analyses for each condition (gradual, random, repeat) more than one object (e.g., “jet turning into a shark”). Most were calculated for the patient study with the image number of these were in the random condition (only 0.63% of indi- (15 morphing images from 100% r fi st object to 0% r fi st object) vidual responses in the gradual or repeat condition). Follow- as a within subject factor and participant group (LBD, RBD ing the procedure of our initial study (Stöttinger et al. 2014), and HCO) as a between subject factor. In Sect. Difference: ran- answers in the gradual and repeat condition were coded as dom minus gradual, we calculate a difference score between “second object” as soon as the second object was mentioned. first object reports in the random and gradual conditions (ran- In 1.54% of individual responses both objects were named in dom–gradual) with positive numbers indicating a benefit for the random condition. In cases where the participant stated gradual presentations. The difference score was entered into explicitly which object he/she preferred (e.g., “It could be a a univariate ANOVA with participant group (LBD, RBD cat or a rabbit. But it looks more like a rabbit”) answers were and HCO) as the independent variable. Analyses were calcu- coded based on the indicated preference. Applying this rule, lated separately—(1) with the HCO-age-matched group, (2) 0.51% of individual answers were coded as “first object”, with the HCO-cognitively matched group as controls and (3) and 0.44% of individual answers were coded as “second restricted to RBD and LBD participant groups. Given that both object”. When answers could not unambiguously assigned patient groups suffered from brain injury this was considered to either object (e.g., “airplane or shark possibly”), answers the most meaningful comparison. Statistically significant main were always coded as “second object” (0.32%). We repeated effects were further analyzed by a post-hoc Bonferroni tests as analyses in Sect. Difference: random minus gradual using implemented in SPSS; t tests were used for post-hoc interac- only the coding of when patient participants first reported tion analyses (Bonferroni corrected for multiple comparisons). the second object (and ignoring a stated preference) with Statistical test were two-tailed and an alpha level of p < .05 was no change in the pattern of results or statistically significant used to determine significance. findings. In cases where the standard statistical tests failed to reject the null, we turned to Bayes Factors to assess whether the null was more probable than the alternative. Specifically, the Lesion tracing Bayes factor allowed us to evaluate whether the differences between RBD and LBD represent (a) evidence for H1 [RBD The most recent available clinical CT (17) or MRI (five) performance being worse than LBD (Bayes factor ≥ 3)], (b) scan was obtained for each patient. All scans were aligned evidence for the null hypothesis of no performance difference to the anterior commissure in SPM8. Lesions were traced (Bayes factor ≤ 0.33) or whether the data were not sensitive manually in MRIcron (Rorden et al. 2007) and spatially nor- enough to confidently distinguish between the two alternatives malized using the Clinical Toolbox in SPM (Rorden et al. (Bayes factor > 0.33 and < 3). Our cut-offs used were those 2012). Common involvement of brain-damaged regions recommended by Dienes (2014). One advantage of Bayes fac- across different patient groups was identified by overlapping tors is their robustness to small sample sizes that are under- individual normalized brain lesions on a standard template powered for conventional analyses (Vadillo et al. 2016). Bayes (i.e., AICHA—An atlas of intrinsic connectivity of homo- factors were calculated using the online-calculator provided topic areas; Joliot et al. 2015) in MRIcron. A summary of the by Dienes http://www.lifesci.susse x.ac.uk/home/Zolt an_Diene location and size of participant lesions was obtained using s/infer ence/Bayes .htm and where effect sizes came from our the descriptive tool. Due to our small sample size, these data previous study (Stöttinger et al. 2014). are underpowered for statistical analyses of lesion location and performance scores [e.g., voxel-based lesion-symptom mapping; (VLSM; Bates et al. 2003)]. Thus, while acknowl- edging the exploratory nature of the data, we include them for comparison with prior reports, and for the purpose of generating structure–function hypotheses (Danckert et al. 2012b; Stöttinger et al. 2014, 2015). 1 3 1758 Experimental Brain Research (2018) 236:1749–1765 patient in the random condition was excluded from further Results analysis as the patient did not recognize one object (i.e., the patient failed to identify the spider in the spider–sun pic- Online study ture set even at the first picture position—100% spider/0% sun). See Table 3 for average performance in all three con- Seventy six of the 77 participants reported all catch trials ditions for all groups (note: smaller numbers reflect better correctly. Mean percentages of first object reports are dis- performance). played in Fig. 3. We refrained from comparing performance in the repeat condition with either the gradual or random Gradual condition conditions as morphing direction was not counterbalanced. A repeated measures ANOVA revealed a significant main Compared to the HCO-age-matched group, results dem- effect for picture position [F (14,1064) = 2543.29, p < .001, onstrated a significant main effect for image number η = .97]. As evident in Fig. 3, probability of first answer [F(14,392) = 353.04, p < .001, η = .93] due to a decrease reports decreased as a function of picture position. A sig- of first object reports as a function of picture position, nificant main effect for condition [F (1,76) = 31.35, p < .001, as well as a significant main effect for participant group η = .29] showed that participants reported significantly [F(2,28) = 9.47, p = .001, η = .40] (Fig.  4, left panel). A fewer first objects in the gradual (mean 7.36, SD 0.76) com- post hoc Bonferroni test showed that age-matched HCO and pared to the random condition (mean 8.00, SD 0.72). The LBD patients had a comparable percentage of first object interaction between condition x picture position showed that reports (p = .16). Age-matched HCO had a significantly participants reported the second object earlier in the grad- lower percentage of first object reports than RBD patients ual compared to the random context [F(14,1064) = 13.04, (p = .001). This analysis also demonstrated a significant p < .001, η = .15]: at picture #7, 8, 9 and 11 probability of interaction between image number and participant group first object reports was significantly lower in the gradual [F(14,392) = 3.53, p < .001, η = .20]. Independent samples t compared to the random condition (all p’s < .05). These find - test (Bonferroni corrected) revealed that both patient groups ings replicate Stöttinger et al. (2016). showed a significantly higher percentage of first object reports at picture image #9 and 10 compared to HCO-age Patients matched controls. The same repeated measures analysis restricted to patient Patient participants identified all catch trials correctly, with groups (i.e., RBD vs. LBD) demonstrated a significant main the exception of one patient on one occasion. One set for one effect for picture position [F (14,280) = 206.97, p < .001, η = .91] and a marginally significant main effect for par - ticipant group: RBD patients tended to have a higher pro- portion of first object reports compared to LBD patients [F(1,20) = 4.12, p = .056, η = .17]. To further evaluate the difference in LBD and RBD first object reports, we calcu- lated the Bayes factor (Dienes 2014). We found a Bayes fac- tor of 3.24, indicating support for a greater updating impair- ment after damage to the right side of the brain. The same analysis conducted with the HCO-cognitively matched group revealed similar results. There was a sig- nificant main effect for image number [F (14,392) = 301.09, p < .001, η = .92] and participant group [F(2,28) = 5.33, p = .011, η = .28]. Age-matched HCO had a comparable percentage of first object reports to LBD patients (p = .88), and a significantly lower percentage of first object reports than RBD patients (p = .01). Fig. 3 Average % of first object reports in the online study col- This analysis also showed a significant interac- lapsed across all picture sets and all participants in each condition. tion between picture position and participant group The x-axis represents picture position (100% first object to 0% first object). Participants stopped reporting the first object earlier, when [F(28,392) = 1.55, p = .038, η = .10]. HCO-cognitively it was presented in the gradual condition (blue line) compared to the matched controls showed a comparable percentage of first random condition (red line). At picture #8 (i.e., vertical dotted line), object reports to LBD patients at all picture positions (all the picture was composed equally of both pictures (for interpretation p’s > .05), and significantly lower percentage of first object of the references to color in this figure legend, the reader is referred to the web version of this article) reports than RBD patients at picture #7 (p = .048) (Fig. 4). 1 3 Experimental Brain Research (2018) 236:1749–1765 1759 Table 3 Average performance of patients (a) and HCO (b) ID Gradual Random Repeat Difference (a) Performance of patients  Left brain damaged patients   110 6.00 8.50 7.50 2.50   442 8.50 8.75 6.50 0.25   588 5.75 6.50 7.00 0.75   788 7.50 8.50 9.00 1.00   828 7.75 7.50 7.00 − 0.25   835 6.25 6.25 6.50 0.00   838 7.50 7.25 9.50 − 0.25   872 7.75 7.00 9.00 − 0.75   898 8.50 9.00 10.00 0.50   902 8.00 7.75 6.50 − 0.25  Right brain damaged patients   27 7.25 8.00 7.50 0.75   205 8.50 7.75 8.00 − 0.75   228 9.50 7.00 8.50 − 2.50   284 10.25 6.75 7.00 − 3.50   489 8.50 9.75 8.00 1.25   729 7.50 7.25 6.00 − 0.25   744 9.25 7.25 10.50 − 2.00   792 9.75 6.75 10.00 − 3.00   856 6.75 7.00 5.50 0.25   874 6.50 5.75 7.00 − 0.75   932 7.25 8.50 11.50 1.25   946 9.25 7.25 10.00 − 2.00 ID Gradual Random Random Difference (b) Performance of healthy controls  HCO (≤ 70 years) (age-matched)   1 6.25 7.25 6.00 1.00   46 5.75 7.00 5.50 1.25   110 6.75 7.25 9.50 0.50   143 6.75 7.50 7.00 0.75   193 6.25 7.50 5.50 1.25   230 6.25 8.00 7.50 1.75   249 6.25 8.50 6.00 2.25   351 6.00 7.75 7.00 1.75   408 7.50 7.25 6.00 − 0.25  HCO (>70 years) (cognitively-matched)   3 5.75 8.00 8.50 2.25   32 6.25 7.25 6.00 1.00   37 7.75 6.80 7.50 − 0.95   148 7.25 7.75 8.50 0.50   206 7.50 8.25 8.50 0.75   208 8.00 7.00 5.50 − 1.00   321 5.00 6.75 6.00 1.75   369 6.75 9.50 8.00 2.75   409 7.00 7.25 8.50 0.25 a b Neglect at time of screening ( and at time of testing): please note that there was no significant difference in any of the dependent measures between participants who showed neglect at screening (or at the time of this testing) and other RBD patients (all p’s > .05) Time since stroke was significantly longer than time since stroke for the other patients [t(20) = 15.58, p < .01]. This, however, did not result in better performance in this patient as evident in the table Italic values represent performance 2 SDs outside the range of controls 1 3 1760 Experimental Brain Research (2018) 236:1749–1765 Fig. 4 Average % of first object reports for gradual (left) random 70, gray lines = HCO-Over 70. Results from the online study are (middle) and repeat (right) conditions. The x-axis represents picture included for comparison purposes (blue dotted lines = HCO-online) position (100% first object to 0% first object). At the vertical dot- (for interpretation of the references to color in this figure legend, the ted line, the picture is composed of 50% first object and 50% second reader is referred to the web version of this article) object. Green lines = LBD, red lines = RBD, black lines = HCO-Under Random condition p < .001, η = .83]. No other main effect or interaction reached significance (all p’s > .05). While RBD and LBD The same analyses conducted for averaged percentage of patients demonstrated comparable performance, a Bayes fac- tor of 0.57 suggest that our data is not sensitive enough to first object reports in the random condition revealed a signif- icant main effect for image number for both analyses—with draw a definite conclusion of no patient group differences. The same repeated measures ANOVA conducted with age-matched HCO as controls [F(14,392) = 403.85, p < .001, η = .94] and with HCO matched on cognitive impairment HCO-cognitively-matched controls revealed no significant effect other than a significant main effect for image posi- [F(14,392) = 361.12, p < .001, η = .93]. Percentage of first object reports decreased as a function of picture position tion [F(14,392) = 141.84, p < .001, η = .84]. No significant difference was found for average percentage of first object in both analyses. No other effects or interactions were sig- nificant (all p ’s > .05) indicating that all groups performed reports between HCO-cognitively matched and LBD (p = 1) or RBD patients (p = .68). Results therefore demonstrate that at equivalent levels (Fig. 4). A Bayes factor of 0.13 for the difference between RBD and LBD patients confirmed the when the control group was matched for cognitive impair- ment, no difference was found between groups (Fig.  4). likelihood of equal performance between the two patient groups. Difference: random minus gradual Repeat condition Healthy individuals see the second object earlier when The percentage of first object reports in the repeat condi- the pictures are presented in a gradual morphing context (Fig. 3). We submitted the difference scores (average pro- tion was submitted to the same repeated measures ANOVA described above and revealed a significant main effect portion of first object reports in random–gradual conditions; Table 3) to two separate univariate ANOVAs with partici- for image number with HCO-age-matched as controls [F(14,392) = 140.03, p < .001, η = .83]: the percentage of pant group (RBD, LBD, and either HCO-age matched or HCO-cognitively matched) as the between subject factor. first object reports decreased as a function of picture posi- tion. We also found a trend towards a significant effect for Larger values for this difference score indicate greater ben- efit from the gradual condition. The analysis with HCO-age participant group [F(2,28) = 2.82, p = .077, η = .17]. HCO- age-matched controls showed a slight, but not significantly matched as a control group showed a significant main effect (Fig. 5) [F(2,28) = 7.74, p = .002, η = .36]. RBD patients lower percentage of first object reports compared to RBD patients (p = .075), and a comparable performance to LBD had significantly smaller difference scores compared to LBD patients [F(1,20) = 4.82, p = .04, η = .19] and HCO-age- patients (p = .76). Calculating the analysis for patient groups only (RBD vs. LBD) revealed a significant main effect for matched controls: F(1,19) = 12.25, p = .002, η = .39; LBD patients tended to have smaller difference compared to age- picture position, again showing a decrease of first object reports as a function of picture position [F(14, 280) = 99.94, matched HCO [F(1,17) = 4.12, p = .058, η = .20]. 1 3 Experimental Brain Research (2018) 236:1749–1765 1761 HCO-cognitively matched group in the repeat condition. Three of these patients had overlap in the parietal cortex (postcentral, supramarginal gyrus, angular gyrus, inferior parietal gyrus, intraparietal sulcus), and the posterior insula and superior temporal gyrus (Fig. 6). LBD patients demon- strated less overlap in their lesions than did the RBD group with only four of ten patients showing common involvement of the putamen. Discussion The aim of the present work was to explore why RBD Fig. 5 Average (95% CI) difference scores (random–gradual) for patients show a selective and differential impairment of LBD (green), RBD (red), HCO-age-matched and HCO-cognitively- mental model updating, and to replicate results of our ear- matched (gray). Results of healthy young participants from the online lier study (Stöttinger et al. 2014). As in our previous study, study are included for comparison purposes (blue) (for interpretation of the references to color in this figure legend, the reader is referred to RBD patients needed significantly more pictures than HCO the web version of this article) before they reported seeing the second object in the gradual morphing condition—even after controlling for general The same analysis done for HCO-cognitively-matched cognitive impairment (Fig. 4). Comparisons between LBD group as controls also showed a significant main effect and RBD groups found clear evidence for worse updating [F (2,28) = 4.85, p = .016, η = .26]. HCO-cognitively in RBD patients (Bayes Factor 3.24). Further, five of the 12 matched controls showed a significantly larger benefit for RBD patients, but none of the ten LBD patients, performed gradual vs. random presentation than did RBD patients well outside the average performance of the control group [F(1,19) = 6.88, p = .017, η = .27] and a comparable benefit (i.e., mean of HCO-cognitively-matched ± 2 SD; Table 3). to LBD patients [F(1,17) = 0.81, p = .381, η = .05]. To test whether the updating impairment in RBD patients reflects (1) an inability to explore new hypotheses, (2) a Healthy controls—age effects general impairment of proactive switching or (3) a more general perceptual and/or attentional impairment, we com- The two HCO groups differed on age in the gradual con- pared three conditions: gradual, random and repeat. All dition only. The HCO-age-matched (≤ 70  years) group groups performed at the same level in the random condition, stopped reporting the first object earlier compared to the indicating that basic perceptual, attentional, and language HCO-cognitively-matched (> 70 years) counterparts as evi- capacities were equivalent in all participant groups (Bayes dent in a significant interaction between image number x factor of 0.13). To test whether RBD performance was due participant group (age matched vs. cognitively matched) to impaired ability to explore alternative interpretations, we [F(14,224) = 1.92, p = .026, η = .11] (Fig.  4, left panel). compared first object reports in the random condition with None of the other conditions revealed a significant differ - the gradual condition. In healthy individuals, we consist- ence between the two control groups (all p’s > .05). ently find that participants report the second object earlier when presented in a gradual context (Stöttinger et al. 2016; Lesion overlay analysis Fig. 3), reflective of active exploration strategies (i.e., “It was a shark but it could also be a plane or a bird.”). In the Among the 12 RBD patients, nine had overlapping lesions online study, we replicated this benefit showing the robust- in the insula (anterior and posterior), rolandic operculum, ness of this effect. This is an important point given the ongo- and the putamen. Five RBD patients were considered poor ing discussion about failures to replicate in social science explorers based on average first object reports and difference (Bohannon 2015). Results in the patient study showed a sig- scores more than two SDs above the mean of the HCO-cog- nificant difference between participant groups in how much nitively matched group (highlighted by gray bars in Table 3). they benefited from the gradual presentation. The strong- Four of these five patients had common involvement in the est effect was seen in healthy age-matched controls. This inferior frontal cortex, precentral gyrus, insula (anterior and benefit decreased with age and was about the same in the posterior), rolandic operculum, superior temporal gyrus, older seniors (matched on general cognitive impairment) as and the putamen (Fig. 6). Four RBD patients showed a poor in LBD patients (Fig. 5). In contrast, RBD patients exhibited general switching performance as indicated by average first a perceptual hysteresis—becoming “stuck” on their initial object reports that were two SDs above the mean of the interpretation even as the pictures gradually morphed into 1 3 1762 Experimental Brain Research (2018) 236:1749–1765 Fig. 6 Lesion overlay maps for LBD and RBD patients (top panel). Bottom panel shows 4 of the 5 RBD patients who were considered poor explorers (left panel) and 3 of the 4 RBD patients, considered poor proactive updaters (right panel). Shading indicates the amount of overlap 1 3 Experimental Brain Research (2018) 236:1749–1765 1763 something else. Combined with our previous work (Danck- uncertain environments has been found by several groups ert et al. 2012b; Stöttinger et al. 2014), this indicates that (Blanchard and Gershman 2018; Ohira et al. 2013, 2014; RBD participants are aware that things are changing in the Laureiro-Martínez et al. 2015). When we presented healthy perceptual representation they are forming, but are defi- individuals with some of the gradually morphing picture sets cient in exploring alternative models to encapsulate those used here in an fMRI experiment, we found that the anterior changes. That is, our RBD patients’ verbal reports of the insula was active not only at the actual time point of reported morphing images indicate that they perceive changes. What object change but also about five seconds before—consistent differentiated them from controls and LBD patients was with the involvement of the anterior insula in the explora- that they interpreted those changes within the context of tion of alternative choices (Stöttinger et al. 2015). We have their initial perceptual representation. Indeed, if the task we recently replicated this finding (Stöttinger et al. 2018). employed depended heavily on detecting the small (~ 4%) By directly comparing HCO groups, we found that the changes from one image to the next, one might assume that ability to recognize the second object is sensitive to age, it would be LBD that would lead to the greatest impairment, with older seniors needing significantly longer than younger as damage to this hemisphere is known to affect local as seniors to identify the second object in the gradual morphing opposed to global image processing (Martinez et al. 1997; condition. This is consistent with findings showing slower Robertson and Lamb 1991). Therefore, we suggest that the switch rates in ambiguous figures as people age (Ukai et al. deficit seen here in RBD patients is indicative of a failure to 2003). Since these groups were selected to match either the adapt to changing circumstances as they struggle to explore age or the cognitive status of the patient groups, it cannot be alternate hypotheses to explain the change. conclusively stated that age affects updating behavior. This Data in the repeat condition are harder to interpret given effect could also reflect general cognitive decline. Indeed, the smaller number of observations (i.e., two instead of four six of our 18 HCO had a MoCA score below 26 with four sets). In the repeat condition, there is no need to explore of these participants in the HCO-cognitively matched group alternate hypotheses—participants know that the plane will (> 70 years). According to the cutoff point initially pub - morph into a shark; they have seen it before. All that remains lished, this would be suggestive of a mild cognitive impair- is for participants to determine when to report seeing the ment (MCI; Nasreddine et al. 2005; Smith et al. 2007). How- shark. Performance of RBD patients as a group was not dif- ever, more recent studies suggest cut-off scores of 22/23 to ferent from that of LBD patients or cognitively matched con- be more sensitive for MCI (Lee et al. 2008; Luis et al. 2009). trols in the repeat condition (Fig. 4). However, our rejection All of our HCO were well within this range. of the proactive switch account rests on the strength of a In summary, our study suggests that updating impair- null result, and our Bayes factor of 0.57 suggests that these ments seen after damage to the right side of the brain most negative data are not strong. Further comparisons of gradual likely represent an impairment in the ability to explore the and repeat conditions will be necessary to evaluate whether alternative interpretations within a slowly changing envi- updating impairments in RBD patients are due only to fail- ronment. A better understanding of this type of cognitive ures of hypothesis exploration, proactive switching, or both. impairment is relevant for developing an improved under- Although RBD patients performed more poorly than HCO standing of post-stroke behavioral deficits and prompting the and LBD patients, it is also evident that performance within study of specific rehabilitation procedures. the RBD patients was heterogeneous (Table 3). Some RBD Acknowledgements Open access funding provided by Austrian Sci- patients performed within the norm of cognitively-matched ence Fund (FWF). This work was supported by the FWF Austrian controls, while others were outside this range. There was Science Fund (#V480-B27) Grant (Eliese Richter Program) to E.S., also heterogeneity in the type of impairment observed, with the Natural Sciences and Engineering Research Council of Canada some patients only impaired in one regard (e.g., exploring) Discovery (#261628-07), Canada Research Chair grants, Heart and Stroke Foundation of Ontario #NA 6999 to J.D. and Canadian Institutes while others had broader impairments (Table 3). Given our of Health Research #219972 operating grant to J.D. and B.A.. The sample sizes the lesion overlay analyses should be regarded above-mentioned funding agencies had no role in the study design, as exploratory. Nevertheless, they are consistent with general data collection and analysis, decision to publish, or the preparation of ideas about the systems important for adapting to chang- the manuscript. ing and uncertain circumstances. A failure to explore was Open Access This article is distributed under the terms of the Crea- mostly associated with damage to frontal regions (frontal tive Commons Attribution 4.0 International License (http://creat iveco operculum, insula), while a general impairment in alternat- mmons.or g/licenses/b y/4.0/), which permits unrestricted use, distribu- ing between interpretations (as indexed by the repeat condi- tion, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the tion) was associated more prominently with parietal dam- Creative Commons license, and indicate if changes were made. age (Fig. 6). In healthy individuals, an association between the right insula—commonly damaged in our patients—and exploration of alternative interpretations in stochastic and 1 3 1764 Experimental Brain Research (2018) 236:1749–1765 Joliot M, Jobard G, Naveau M, Delcroix N et al (2015) AICHA: an References atlas of intrinsic connectivity of homotopic areas. J Neurosci Methods 254:46–59 Bates E, Wilson SM, Saygin AP, Dick F, Sereno MI, Knight RT, Laureiro-Martínez D, Brusoni S, Canessa N, Zollo M (2015) Under- Dronkers NF (2003) Voxel-based lesion–symptom mapping. Nat standing the exploration–exploitation dilemma: an fMRI study of Neurosci 6:448–450 attention control and decision-making performance. Strat Manag Blanchard TC, Gershman SJ (2018) Pure correlates of exploration and J 36:319–338 exploitation in the human brain. Cogn Affect Behav Neurosci Lee Y, Dong Woo L, Cho S et al (2008) Brief screening for mild cog- 18(1):117–126 nitive impairment in elderly outpatient clinic: validation of the Bohannon J (2015) Many psychology papers fail replication test. Sci- Korean version of the Montreal cognitive assessment. J Geriatr ence 349:910–911 Psychiatry Neurol 21:104–110 Britz J, Landis T, Michel CM (2009) Right parietal brain activity pre- Long GM, Toppino TC (2004) Enduring interest in perceptual cedes perceptual alternation of bistable stimuli. Cereb Cortex ambiguity: alternating views of reversible figures. Psychol Bull 19:55–65 130:748 Britz J, Pitts MA, Michel CM (2011) Right parietal brain activity pre- Luis CA, Keegan AP, Mullan M (2009) Cross validation of the Mon- cedes perceptual alternation during binocular rivalry. Hum Brain treal cognitive assessment in community dwelling older adults Mapp 32:1432–1442 residing in the Southeastern US. J Geriatr Psychiatry 24:197–201 Brownell HH, Michel D, Powelson J, Gardner H (1983) Surprise but Martinez A, Moses P, Frank L, Buxton R, Wong E, Stiles J (1997) not coherence: sensitivity to verbal humor in right-hemisphere Hemispheric asymmetries in global and local processing: evi- patients. Brain Lang 18:20–27 dence from fMRI. Neuroreport 8:1685–1689 Danckert J, Stöttinger E, Anderson B (2012a) Neglect as a disorder of McGuire JT, Nassar MR, Gold JI, Kable JW (2014) Functionally dis- representational updating. Psychology of neglect. NOVA Science sociable influences on learning rate in a dynamic environment. Publishers, New York, pp 1–28 Neuron 84:870–881 Danckert J, Stöttinger E, Quehl N, Anderson B (2012b) Right hemi- Meenan JP, Miller LA (1994) Perceptual flexibility after frontal or sphere brain damage impairs strategy updating. Cereb Cortex temporal lobectomy. Neuropsychologia 32:1145–1149 22:2745–2760 Nasreddine ZS, Phillips NA, Bédirian V, Charbonneau S, Whitehead V, Dienes Z (2014) Using Bayes to get the most out of non-significant Collin I, Cummings JL, Chertkow H (2005) The Montreal cogni- results. Front Psychol 5:781 tive assessment, MoCA: a brief screening tool for mild cognitive Egré P, De Gardelle V, Ripley D (2013) Vagueness and order effects in impairment. J Am Geriatr Soc 53:695–699 color categorization. J Logic Lang Inform 22(4):391–420 Ohira H, Matsunaga M, Murakami H, Osumi T, Fukuyama S, Shinoda Égré P, Ripley D, Verheyen S (2018) The sorites paradox in psychol- J, Yamada J (2013) Neural mechanisms mediating association of ogy. In Oms S, Zardini E (eds) The sorites paradox. Cambridge sympathetic activity and exploration in decision-making. Neuro- University Press, Cambridge science 246:362–374 Ferber S, Danckert J (2006) Lost in space: the fate of memory represen- Ohira H, Ichikawa N, Kimura K, Fukuyama S, Shinoda J, Yamada J tations for non-neglected stimuli. Neuropsychologia 44:320–325 (2014) Neural and sympathetic activity associated with explo- Filipowicz A, Anderson B, Danckert J (2016) Adapting to change: the ration in decision-making: further evidence for involvement of role of the right hemisphere in mental model building and updat- insula. Front Behav Neurosci 8:381 ing. Can J Exp Psychol 70:201 Piper BJ, Li V, Eiwaz MA et al (2012) Executive function on the psy- Freitas S, Simoes MR, Alves L, Vicente M, Santana I (2012) Mon- chology experiment building language tests. Behav Res Method treal cognitive assessment (MoCA): validation study for vascular 44:110–123 dementia. J Intern Neuropsychol Soc 18:1031–1040 Raffman D (2011) Vagueness and observationality. In: Ronzitti G (ed) Geng JJ, Vossel S (2013) Re-evaluating the role of TPJ in atten- Vagueness: a guide. Springer, Dordrecht, pp 107–121 tional control: contextual updating? Neurosci Biobehav Rev Robertson LC, Lamb MR (1991) Neuropsychological contributions to 37:2608–2620 theories of part/whole organization. Cogn Psychol 23:299–330 Grant DA, Berg E (1948) A behavioral analysis of degree of reinforce- Rorden C, Karnath HO, Bonilha L (2007) Improving lesion–symptom ment and ease of shifting to new responses in a Weigl-type card- mapping. J Cogn Neurosci 19:1081–1088 sorting problem. J Exp Psychol 38:404 Rorden C, Bonilha L, Fridriksson J, Bender B, Karnath HO (2012) Griffin R, Friedman O, Ween J, Winner E, Happé F, Brownell H (2006) Age specific CT and MRI templates for spatial normalization. Theory of mind and the right cerebral hemisphere: refining the Neuroimage 61:957–965 scope of impairment. Laterality 11:195–225 Sepahvand N, Stöttinger E, Danckert J, Anderson B (2014) Sequential Griffiths TL, Tenenbaum JB (2012) Optimal predictions in everyday decisions: a computational comparison of observational and rein- cognition. Psychol Sci 17:767–773 forcement accounts. PLoS One 9:e94308 Happé F, Brownell H, Winner E (1999) Acquired theory of mind Shaqiri A, Anderson B (2013) Priming and statistical learning in right ‘impairments following stroke. Cognition 70:211–240 brain damage patients. Neuropsychologia 51:2526–2533 Hepworth L, Rowe F, Walker M, Rockliffe J, Noonan C, Howard C, Shaqiri A, Anderson B, Danckert J (2013) Statistical learning as a tool Currie J (2015) Post-stroke visual impairment: a systematic litera- for rehabilitation in spatial neglect. Front Hum Neurosci 7:224 ture review of types and recovery of visual conditions. Phthalmol Smith T, Gildeh N, Holmes C (2007) The Montreal cognitive assess- Res An Intern J 5:1–43 ment: validity and utility in a memory clinic setting. Can J Psy- Husain M, Mannan S, Hodgson T, Wojciulik E, Driver J, Kennard C chiatry 52:329–332 (2001) Impaired spatial working memory across saccades contrib- Sterzer P, Kleinschmidt A (2007) A neural basis for inference in per- utes to abnormal search in parietal neglect. Brain 124:941–952 ceptual ambiguity. Proc Natl Acad Sci 104:323–328 Johnson-Laird PN (2004) The history of mental models. In: Mank- Stöttinger E, Filipowicz A, Marandi E, Quehl N, Danckert J, Anderson telow K, Chung MC (eds) Psychology of reasoning: theoretical B (2014) Statistical and perceptual updating: correlated impair- and historical perspectives. Psychology Press, Hove, pp 179–212 ments in right brain injury. Exp Brain Res 232:1971–1987 Stöttinger E, Filipowicz A, Valadao D, Culham JC, Goodale MA, Anderson B, Danckert J (2015) A cortical network that marks the 1 3 Experimental Brain Research (2018) 236:1749–1765 1765 moment when conscious representations are updated. Neuropsy- Vocat R, Saj A, Vuilleumier P (2012) The riddle of anosognosia: does chologia 79:113–122 unawareness of hemiplegia involve a failure to update beliefs? Stöttinger E, Sepahvand NM, Danckert J, Anderson B (2016) Assess- Cortex 49:1771–1781 ing perceptual change with an ambiguous figures task: normative Weed E, McGregor W, Nielsen JF, Roepstor ff A, Frith U (2010) Theory data for 40 standard picture sets. Behav Res Method 48:201–222 of mind in adults with right hemisphere damage: what’s the story? Stöttinger E, Aichhorn M, Anderson B, Danckert J (2018) The neural Brain Lang 113:65–72 systems for perceptual updating. Neuropsychologia. https ://doi. Weilnhammer VA, Ludwig K, Hesselmann G, Sterzer P (2013) Fron- org/10.1016/j.neuro psych ologi a.2018.03.017 toparietal cortex mediates perceptual transitions in bistable per- Tenenbaum JB, Kemp C, Griffiths TL, Goodman ND (2011) How ception. J Neurosci 33:16009–16015 to grow a mind: statistics, structure, and abstraction. Science Wilson B, Cockburn J, Halligan P (1987) Development of a behavioral 331:1279–1285 test of visuospatial neglect. Arch Phys Med Rehabil 68:98–102 Ukai K, Ando H, Kuze J (2003) Binocular rivalry alternation rate Zaretskaya N, Thielscher A, Logothetis NK, Bartels A (2010) Disrupt- declines with age. Percept Mot Skills 97:393–397 ing parietal function prolongs dominance durations in binocular Vadillo MA, Konstantinidis E, Shanks DR (2016) Underpowered sam- rivalry. Curr Biol 20:2106–2111 ples, false negatives, and unconscious learning. Psychol Bull Rev 23:87–102 1 3 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Experimental Brain Research Springer Journals

Updating impairments and the failure to explore new hypotheses following right brain damage

Free
17 pages
Loading next page...
 
/lp/springer_journal/updating-impairments-and-the-failure-to-explore-new-hypotheses-aQqu8Vv7qv
Publisher
Springer Berlin Heidelberg
Copyright
Copyright © 2018 by The Author(s)
Subject
Biomedicine; Neurosciences; Neurology
ISSN
0014-4819
eISSN
1432-1106
D.O.I.
10.1007/s00221-018-5259-6
Publisher site
See Article on Publisher Site

Abstract

We have shown recently that damage to the right hemisphere impairs the ability to update mental models when evidence suggests an old model is no longer appropriate. We argue that this deficit is generic in the sense that it crosses multiple cog- nitive and perceptual domains. Here, we examined the nature of this updating impairment to determine more precisely the underlying mechanisms. We had right (RBD, N = 12) and left brain damaged (LBD, N = 10) patients perform versions of our picture-morphing task in which pictures gradually morph from one object (e.g., shark) to another (e.g., plane). Performance was contrasted against two groups of healthy older controls, one matched on age (HCO-age-matched, N = 9) and another matched on general level of cognitive ability (HCO-cognitively-matched, N = 9). We replicated our earlier findings showing that RBD patients took longer than LBD patients and HCOs to report seeing the second object in a sequence of morphing images. The groups did not differ when exposed to a morphing sequence a second time, or when responding to ambiguous images outside the morphing context. This indicates that RBD patients have little difficulty alternating between known representations or labeling ambiguous images. Instead, the difficulty lies in generating alternate hypotheses for ambiguous information. Lesion overlay analyses, although speculative given the sample size, are consistent with our fMRI work in healthy individuals in implicating the anterior insular cortex as critical for updating mental models. Keywords Right brain damage · Mental representations · Updating failures · Picture morphing · Exploration Introduction Damage to the right hemisphere results in a range of het- erogeneous impairments, including impoverished spatial James Danckert and Britt Anderson contributed equally to the attention (i.e., neglect; Danckert et al. 2012a), statistical manuscript. learning (Shaqiri and Anderson 2013; Shaqiri et al. 2013), humor appreciation (Brownell et al. 1983), working memory * Elisabeth Stöttinger capacity (Ferber and Danckert 2006; Husain et al. 2001), and elisabeth.stoettinger@sbg.ac.at deficient Theory of Mind (Happé et al. 1999; Griffin et al. Carolyn Louise Guay 2006; Weed et al. 2010). We have argued that these impair- clguay@uwaterloo.ca ments can be parsimoniously explained by a general impair- James Danckert ment in updating mental models (Danckert et al. 2012a, b; jdancker@uwaterloo.ca Filipowicz et al. 2016; Geng and Vossel 2013; Shaqiri and Britt Anderson Anderson 2013; Shaqiri et al. 2013; Stöttinger et al. 2014; britt@uwaterloo.ca Vocat et al. 2012). Department of Psychology, Center for Cognitive Everyday we are confronted with an enormous amount Neuroscience, University of Salzburg, Hellbrunnerstrasse of information. Mental models allow us to deal with this 34, 5020 Salzburg, Austria complexity by compactly representing the regularities that Department of Psychology, University of Waterloo, govern our environment. We use these models to guide our Waterloo N2L 3G1, Canada decisions (e.g., Is the situation dangerous? Is this food edi- Centre for Theoretical Neuroscience, University of Waterloo, ble? Johnson-Laird 2004; Griffiths and Tenenbaum 2012; Waterloo N2L 3G1, Canada Vol.:(0123456789) 1 3 1750 Experimental Brain Research (2018) 236:1749–1765 Tenenbaum et al. 2011). Since the world is in flux, these From our prior study, we know that RBD patients notice models are only useful if we update them when things small changes in the sequences of morphing images. RBD change. While a salient shift in the environment allows us patients, however, incorporated changes into their initial per- to simply react (e.g., a bolt of lightning indicates that it is ceptual representation (e.g., “The shark’s fins are getting time to head indoors), gradual changes in the environment bigger.”), rather than entertaining an entirely novel interpre- require the aggregation of mismatching evidence to provoke tation (e.g., fins have now become wings). Similarly, within a proactive decision (e.g., when does cloud cover become the context of the rock, paper, scissors game, although RBD sufficiently menacing to indicate a coming storm? McGuire patients fail to update to a change in strategy, their own play et al. 2014). choices do reflect the fact that they have noticed something We previously showed that right brain damage resulted has changed. That is, RBD patients rapidly abandon their old in selective impairment of gradual updating in response to play strategy but fail to adopt a new strategy optimized for small, subtle environmental changes. We had left (LBD) the change in their opponent’s bias (Stöttinger et al. 2014). and right brain damaged (RBD) patients play the children’s In combination, these data suggest that the impairment in game of ‘rock, paper, scissors’ (RPS) against a computer updating following RBD may be due to either (1) a failure opponent. The computer initially played randomly before to explore alternative interpretations in light of observed switching to ‘paper’ 80% of the time. While LBD patients changes or (2) a general inability to proactively switch. and controls rapidly adopted the optimal strategy (i.e., scis- To evaluate these hypotheses, RBD and LBD patients as sors), RBD patients failed to respond to the transition, with well as healthy seniors (HCO) were presented with three most patients continuing to play randomly (Danckert et al. versions of our picture morphing task. The hypotheses and 2012b). Similarly, RBD patients showed a severe updating predictions for this study are summarized in Table 1. In the impairment in a gradually morphing picture task, in which gradual condition participants saw four picture sets in which one object (e.g., shark) morphed over several iterations into one object (e.g., shark) morphed into a different object a completely different object (e.g., plane; Fig.  2, top panel). (e.g., plane; replication of Stöttinger et  al. 2014; Fig.  2, The logic was to first provide participants with a mental top row). In the repeat condition, some of these sequences model (e.g., “It’s a shark”) and then evaluate how much were repeated in the reverse order (e.g., plane-to-shark). In evidence was required for them to update to a new mental the random condition, images of four additional morphing model (e.g., “It’s a plane now”). RBD patients needed sig- sequences were presented in random order with all pictures nificantly more evidence (i.e., more pictures) before they intermixed. For example, a participant would see the sec- reported seeing the second object compared to LBD patients ond picture of the gun-hair dryer picture set, followed by and healthy controls. Performance in the RPS task correlated the ninth picture of the spider-sun picture set, and so on with performance in the picture morphing task, indicating (Fig. 2, middle row). We compared the likelihood that par- a general updating impairment across multiple cognitive ticipants reported the first or second object for pictures based domains (Stöttinger et al. 2014). on their order in the morphing sequence (i.e., 1st to 15th While our prior studies showed an updating impair- picture position), and on the type of sequence (i.e., gradual, ment after RBD, they did not unequivocally demonstrate repeat or random). the mechanism for this impairment. Updating to gradually Each of our key hypotheses was addressed in the follow- evolving stimuli requires several different processes: (1) we ing ways: need to be able to detect and accumulate evidence of a mis- match. (2) When the mismatch reaches some threshold we 1. To evaluate whether updating impairments seen in RBD need to explore for potential alternative interpretations. (3) patients can be explained by a failure to explore, we Finally, we need to proactively switch to this new interpre- calculated the difference in performance between the tation. The aim of the research reported here was to test for random–gradual conditions. We know that healthy indi- the integrity of each of these mechanisms in RBD patients. viduals tend to switch categories earlier in the morph Table 1 Expected outcome vs. Exploration Proactive switch Perceptual/attentional bias actual outcome Test (condition) ΔRandom–gradual Repeat Random Expected RBD < LBD&HCO RBD > LBD&HCO RBD ≠ LBD & HCO Actual RBD < LBD&HCO ✓ RBD = LBD&HCO ✗ (?) RBD = LBD & HCO ✗ Summary of expected vs. actual results for all three hypotheses (i.e., exploration, proactive switch, percep- tual/attentional bias) potentially explaining the updating impairment seen in RBD patients. The second row refers to the condition that was used to evaluate the hypothesis. Please note that results are only displayed for HCO matched on cognitive status 1 3 Experimental Brain Research (2018) 236:1749–1765 1751 series when presented in a gradual context (e.g., when rests on the accumulation of gradual changes in a noisy the object is still 60% shark and only 40% plane) than environment; participants proactively decide at which when single images from the same sequence are pre- point their current model is no longer supported by the sented in isolation, outside the morphing context (Egré evidence. Indeed, we found that while RBD patients et al. 2013; Raffman 2011; Stöttinger et al. 2016; see showed a selective updating impairment, there was no Egré et al. 2018 for review). This suggests that healthy difference between RBD and LBD patients in a card individuals are exploring alternative interpretations sorting test akin to the WCST (Danckert et al. 2012b; conditional on the evolving history they have viewed Stöttinger et al. 2014; Piper et al. 2012). (e.g., “I know it was a shark but what else could it be? 3. Lastly, to assess the contribution of general post-stroke A bird?”). In the context of the rock–paper–scissors visual or attentional impairments (Hepworth et al. 2015), game, we found that RBD patients explored a lim- we compared the performance of RBD patients with the ited selection of alternative strategies (Danckert et al. performance of control groups in the random condition. 2012b) and employed less efficient exploration strate- Damage to the right hemisphere often results in an atten- gies (Sepahvand et al. 2014). If updating impairments tional bias towards the ipsilesional side (i.e., neglect; in RBD patients are due to a generic failure to explore Danckert et al. 2012a). This could have potentially hin- alternatives evident across different tasks and domains, dered the capacity of RBD patients to correctly identify we expect them to show no benefit from the contextual small changes in our original study, and consequently information provided by the gradual morphing sequence to update to a new model. If the updating impairment (Egré et al. 2013; Raffman 2011; Stöttinger et al. 2016; in RBD patients is due to a general perceptual and/or Table 1, second row). attentional impairment, we expect RBD patients to per- 2. To evaluate whether RBD patients would always strug- form differently than LBD patients when pictures are gle to proactively shift to a new interpretation, (irrespec- presented individually outside of the morphing context tive of whether they are naïve as to the second object), (i.e., random condition; Table 1, right column). we compared performance in the repeat condition across participant groups. Evidence that RBD impairs the abil- One major assumption of our study is that the benefit ity to make proactive decisions comes from research for the gradual over the random presentation sequence is using ambiguous or bistable figures such as the Rubin’s indicative of active exploration. While this benefit has also face/vase picture or the Necker cube in which healthy been reported elsewhere in the literature (see Egré et al. individuals typically alternate between two mutually 2018 for review), to our knowledge our study was the only exclusive interpretations (see Long and Toppino 2004 one that showed this benefit in the context of morphing for review). Given that the sensory input is stable, the objects (Stöttinger et al. 2016). In the original study, we switch to a new interpretation is internally generated used a between subjects design. We therefore first evalu- by the participant. Research shows that performance ated in an online study whether we can replicate the effect in an ambiguous figures task of this kind is associated when the type of presentation (gradual vs. random) is with the right hemisphere, with damage to right frontal manipulated within participants—a non-trivial point given cortex resulting in significant impairment in switching the ongoing discussion about failures to replicate in psy- to the second interpretation (Meenan and Miller 1994). chological science (Bohannon 2015). Furthermore, right frontal and parietal areas are active during proactive alternations between two percepts (Britz et al. 2009, 2011; Sterzer and Kleinschmidt 2007; Weilnhammer et al. 2013; Zaretskaya et al. 2010). So Methods if RBD patients have a general impairment in proac- tively switching to a new interpretation irrespective of Prior to the patient study, we conducted an online study to whether or not they know what the second object will determine the benefit of gradual presentations in a within be, we expect them to identify the second object later subject design (Stöttinger et al. 2016). The same picture compared to other participant groups in the repeat condi- sets used in the online study were then used in the patient tion (Table 1, third row). study. Besides the need to replicate the effect, the online At this point, it is also worth noting that gradual study also allowed us to have a reference point for the updating should not be confused with set-shifting. performance of healthy, younger individuals. Results of While participants have to react to a salient mismatch the online study are included in the graphs depicting the in the Wisconsin Card Sorting Test (WCST; i.e., an results of patients and healthy seniors for comparison action that was considered “correct” suddenly becomes purposes. “incorrect”; Grant and Berg 1948), updating in our tasks 1 3 1752 Experimental Brain Research (2018) 236:1749–1765 one of the healthy elders was found to have had a past neu- Participants rological illness, and was excluded from further analyses. The final sample of 18 participants consisted of nine who Online study were 70 years or younger (HCO-age-matched: 9 females, mean age 66.35 years, ± 3.35) and nine older than 70 years Seventy-seven participants (33 female) recruited through Mechanical Turk, and between 19 and 53  years of age (HCO-cognitively-matched: 6 female, mean age 78.55 years, ± 3.32). The HCO-cognitively-matched control group had a (mean 35.34, SD 9.62) participated in this study [Cauca- sian/white (80.50%), Hispanic (7.8%), African American nominally lower MOCA score than the HCO-age-matched group, but the two control groups did not differ statisti- (5.2%), East Asian (3.9%)]. Participants received $1.50 for their participation. cally {HCO-cognitively-matched mean = 25.78, ± 2.44 vs. HCO-age-matched mean = 27.78, ± 1.86; [t(16) = 1.96, Patient study p = .068]}. Six of the HCO had a MoCA score between 25 and 22, potentially indicating a mild cognitive impairment Four groups were tested in this study—RBD patients, LBD (Smith et al. 2007; Nasreddine et al. 2005; but see; Lee et al. 2008 and; Luis et al. 2009 for the argument of lower cut-off patients, younger (≤ 70 years; i.e., HCO-age-matched) and older (> 70 years; i.e., HCO-cognitively-matched) healthy points). The MoCA for LBD (mean 24.79, ± 2.36) and RBD seniors. Patients were recruited from the Neurological Patient Database of the University of Waterloo (Heart and patients (mean 24.50, ± 2.84) was lower than that of the HCO-age-matched group (all p’s < .01), but comparable to Stroke Foundation funded). Of the 13 LBD patients three were excluded due to Montreal Cognitive Assessment that of the HCO-cognitively-matched group (all p’s > .25). There was no significant difference in MoCA scores of LBD (MoCA; Nasreddine et al. 2005) scores in the demented range (N = 2; MoCA ≤ 9). This score was considerably and RBD patients [t(20) = 0.18, p = .86]. LBD and RBD were of comparable age [t(20) = 0.87, p = .40]. Both groups lower than the optimal cutoff point for vascular demen- tia (i.e., MoCA < 17; Freitas et al. 2012). One patient was were significantly younger than the HCO-cognitively- matched group (all p’s > .01) but comparable in age to the excluded due to a failure to find any discernible lesion on available brain scans (N = 1). The final sample of LBD HCO-age-matched group (all p’s > .30). Also, there was no significant difference for time since stroke or lesion volume patients comprised ten patients (2 female, mean age 61.28 years, ± 14.47). Nine of these patients were stroke patients. between the RBD and LBD patients (all p’s > .05). The University of Waterloo’s Office of Research Ethics In one patient (#835), brain damage was due to a resection of an arterial vascular malformation. Twelve RBD patients approved the protocol for both studies—online and patient study. Participants on both studies gave informed written (2 female, mean age 65.88 years ± 10.13) participated. All of them were stroke survivors. Demographics for all patients consent prior to participation according to the Declaration of Helsinki, by either clicking on the “I agree” button (online are in Table 2 with lesions shown in Fig. 1. Patients were screened for neglect upon admission to the database and study), or by signing the consent letter (patient study). again prior to the experiment using the Behavioral Inat- tention Test (BIT; Wilson et al. 1987). Six RBD patients Stimuli and design showed neglect at initial screening, with two showing chronic neglect at testing (#284, #744). No LBD patient Each participant saw eight picture sets selected from a larger set validated in an earlier study (Stöttinger et al. 2016; https showed neglect. Six of the ten LBD patients were reported to have aphasia when admitted to the database. One LBD ://osf.io/qi2vg/ ). All images were a standard size (316 × 316 pixels) displayed on a white background. Four sets were patient experienced slight word finding difficulties at testing. He was able to describe all the objects and his answers could presented in a gradual order, four in random order. In the gradual condition one common object morphed over 15 iter- be reliably coded as either the first or the second object. Nineteen healthy older controls were recruited from the ations into a different object (Stöttinger et al. 2014, 2016). Two of these sets were presented again, but in reverse order University of Waterloo’s Research in Aging Participant Pool. This pool recruits community dwelling seniors for participa- (i.e., repeat condition). In the random condition, four picture sets were pre- tion in studies on aging. Thus, our HCO were enrolled based on age. We subdivided our healthy older controls into two sented in a random order. Pictures from these sets were randomly assigned to four new series with 15 pictures subgroups. One matched to our clinical participants based on age, and another based on MOCA scores to equate cognitive in each. Each random series contained the same number of pictures from each picture set and each part within a ability. The former group (HCO-age-matched) was under 70 years of age, while the second subgroup (HCO-cogni- series. That is, each random series had five pictures from the first third of the morphing series (pictures #1 to #5); tively-matched) was over 70 years of age. After recruitment 1 3 Experimental Brain Research (2018) 236:1749–1765 1753 Table 2 Demographics for patients (a) and HCO (b) ID Lesion volume Age (years) Time since stroke MoCA Gender Education (years) (years) (a) Demographics (patients)  LBD patients   110 65 64 8.67 23 Male 12   442 8127 68 12.44 23 Male 14   588 274 68 3.25 28 Male 20   788 328 35 1.84 22 Male 15   828 7374 76 1.44 25 Female 17   835 7573 35 1.39 25 Male 14   838 105 69 1.41 26 Male 23   872 665 71 0.89 23 Male 10   898+ 5396 66 0.48 23 Male 8   902+ 161 61 1.76 29 Female 12  RBD patients   27 21,187 49 8.37 27 Male 15   205 3742 62 6.69 29 Male 14   228 6316 86 6.62 26 Female 11   284 14,203 74 6.95 22 Female 12   489 1693 71 4.36 24 Male 15   729 5758 66 2.30 22 Male 14   744 21,479 73 3.60 25 Male 24   792 5878 64 1.67 24 Male 15   856 232 55 0.87 26 Male 16   874 5026 56 1.16 26 Male 14   932 1520 62 0.26 25 Male 16   946 5694 72 0.20 18 Male 20 ID Age (years) MoCA Gender Education (years) (b) Demographics (healthy controls)  Younger healthy controls (≤ 70 years)   1 61 28 Female 20   46 64 28 Female 17   110 69 27 Female 17   143 68 30 Female 16   193 62 30 Female 12   230 69 28 Female 18   249 70 25 Female 24   351 67 25 Female 16   408 68 29 Female 16  Older healthy controls (> 70 years)   3 76 25 Male 17   32 76 22 Female 13   37 76 27 Female 17   148 81 23 Female 22   206 80 26 Male 20   208 81 28 Female 14   321 81 30 Female 18   369 83 25 Female 13   409 73 26 Male 16 a b Neglect at time of screening ( and at time of testing) 1 3 1754 Experimental Brain Research (2018) 236:1749–1765 Fig. 1 Lesion tracings RBD (a) and LBD patients (b) superimposed on the MNI template. Lesions shown in neurological convention five pictures from the middle (pictures #6 to #10), and five The presentation of each set (gradual or random), the pictures from the end of each series (pictures #11 to #15). order of presentation (random then gradual or vice versa), In each of the eight picture sets, one additional object was as well as order of the series (series 1, 2, 3, 4, 1, 3 vs. 3, presented after the third and twelfth pictures as catch tri- 1, 4, 2, 1, 3), was varied between participants resulting als. Catch trials assessed whether participants were simply in eight versions. Versions were counterbalanced across perseverating (Fig. 2; Stöttinger et al. 2016). participants. 1 3 Experimental Brain Research (2018) 236:1749–1765 1755 Fig. 1 (continued) 1 3 1756 Experimental Brain Research (2018) 236:1749–1765 Fig. 2 Objects morphed over 15 iterations from object 1 into object phing context (e.g., second picture of the gun-hair dryer picture set 2 (gradual condition), or vice versa (repeat condition). In the random the ninth picture of the spider-sun picture set, etc.) condition, objects were presented individually, outside of the mor- sets were removed from further analysis. Importantly, per- Procedure formance in the current online study was the same as in Stöttinger et al. (2016) (all p’s > .05). Online study Patient study The online study questionnaires were designed using Qual- trics©. Participants first filled out demographic questions All assessments and tasks were done in the same order for before being assigned to one of eight task versions using the each patient: BIT, MoCA, picture morphing task. Partici- randomize function of Qualtrics©, and yielding roughly the pants were tested individually in a room at the University same number of participants for each task order (n = 39 did of Waterloo. In a few cases, participants were tested at their random, gradual, repeat and 38 did gradual, repeat, random). own home. Instructions were presented on a screen and Prior to the gradual condition, participants were informed repeated verbally to the participants before each condition they would see six different picture series, containing 17 in the picture morphing task. Participants saw one picture images and that each series would begin with a commonly at a time and were asked to tell us for each picture what known object before changing gradually to show a different they saw. Answers were recorded and transcribed after the object by the end of the series. In the random condition, experiment. Each picture was preceded and followed by a participants were told they would see 68 pictures of objects. fixation cross. Patients and healthy participants received $10 In all conditions, participants saw one picture at a time and per hour of study for their participation. typed their answer underneath the picture. The next picture was revealed by clicking a button at the bottom of the screen. Data analysis On average, participants needed 20 min (± 15 min) to com- plete the questionnaire. Time to complete the questionnaire Behavioral data did not correlate with any of the dependent measures in our task (all p’s > .05). Responses were coded as seeing the “first object”, or seeing There were a small number of technical glitches (< 1% an object other than the first object (i.e., “second object”). In of all image presentations). For 29 image presentations addition, we evaluated whether the catch trials were identi- (0.22%), the wrong picture was presented at the first picture fied correctly. For repeated sequences, the coding as first or position of a series (e.g., a saw instead of a shark). From second object was done based on the ordering used in the our initial study, we know that these pictures are never rated morphing sequence. For example, when a picture set mor- as the “second object”. These items were therefore coded phed from a shark to a plane in the gradual condition, and as “first object” reports. In one version, (a picture of a saw from a plane to a shark in the repeat condition, “shark” was was presented instead of the correct picture at picture posi- coded as first object in the gradual condition, with “plane” tion #10). If this happened between a switch of percepts, coded as first object in the repeat condition. While the cod- the set was removed (seven individual cases). Otherwise, ing was not done blindly, it did use a list of validated pic- it was rated as the answer which preceded and followed the ture terms from Stöttinger et al. (2016; https ://osf.io/qi2vg erroneous presentation. Infrequently, participants failed to /), and was done independently by the first two authors answer (i.e., 0.19% of images). Most of these omissions were with an interrater agreement of 98.98% for the online study preceded and followed by the same response, suggesting no and 99.13% for the patient study. The high agreement is change in the participant’s conscious percept. In ten cases, explained by the fact that there was little opportunity for the omission occurred between a switch in report. These 1 3 Experimental Brain Research (2018) 236:1749–1765 1757 ambiguity. The majority of answers in the online study Statistical analysis were less than three words (one word, e.g., “frog”; 90.45%; two words, e.g., “Jumping man”; 8.81%; or three words, Data were analyzed using repeated measures ANOVA. In the e.g., “man jumping up”; 0.47%). In only 0.28%, the answer online study, the mean percentages of r fi st object reports (aver - included more than three words. Most answers in the patient aged over all sets per condition) were submitted to a repeated study were also less than four words (73%). Participants measures ANOVA with image number (15 morphing images in the patient study, however, were more prone to longer from 100% first object to 0% first object) and condition (grad- answers with 9.44% containing ten words or more. ual vs. random) as within subject factors. Separate repeated On a few occasions (2.18%), patient responses indicated measure analyses for each condition (gradual, random, repeat) more than one object (e.g., “jet turning into a shark”). Most were calculated for the patient study with the image number of these were in the random condition (only 0.63% of indi- (15 morphing images from 100% r fi st object to 0% r fi st object) vidual responses in the gradual or repeat condition). Follow- as a within subject factor and participant group (LBD, RBD ing the procedure of our initial study (Stöttinger et al. 2014), and HCO) as a between subject factor. In Sect. Difference: ran- answers in the gradual and repeat condition were coded as dom minus gradual, we calculate a difference score between “second object” as soon as the second object was mentioned. first object reports in the random and gradual conditions (ran- In 1.54% of individual responses both objects were named in dom–gradual) with positive numbers indicating a benefit for the random condition. In cases where the participant stated gradual presentations. The difference score was entered into explicitly which object he/she preferred (e.g., “It could be a a univariate ANOVA with participant group (LBD, RBD cat or a rabbit. But it looks more like a rabbit”) answers were and HCO) as the independent variable. Analyses were calcu- coded based on the indicated preference. Applying this rule, lated separately—(1) with the HCO-age-matched group, (2) 0.51% of individual answers were coded as “first object”, with the HCO-cognitively matched group as controls and (3) and 0.44% of individual answers were coded as “second restricted to RBD and LBD participant groups. Given that both object”. When answers could not unambiguously assigned patient groups suffered from brain injury this was considered to either object (e.g., “airplane or shark possibly”), answers the most meaningful comparison. Statistically significant main were always coded as “second object” (0.32%). We repeated effects were further analyzed by a post-hoc Bonferroni tests as analyses in Sect. Difference: random minus gradual using implemented in SPSS; t tests were used for post-hoc interac- only the coding of when patient participants first reported tion analyses (Bonferroni corrected for multiple comparisons). the second object (and ignoring a stated preference) with Statistical test were two-tailed and an alpha level of p < .05 was no change in the pattern of results or statistically significant used to determine significance. findings. In cases where the standard statistical tests failed to reject the null, we turned to Bayes Factors to assess whether the null was more probable than the alternative. Specifically, the Lesion tracing Bayes factor allowed us to evaluate whether the differences between RBD and LBD represent (a) evidence for H1 [RBD The most recent available clinical CT (17) or MRI (five) performance being worse than LBD (Bayes factor ≥ 3)], (b) scan was obtained for each patient. All scans were aligned evidence for the null hypothesis of no performance difference to the anterior commissure in SPM8. Lesions were traced (Bayes factor ≤ 0.33) or whether the data were not sensitive manually in MRIcron (Rorden et al. 2007) and spatially nor- enough to confidently distinguish between the two alternatives malized using the Clinical Toolbox in SPM (Rorden et al. (Bayes factor > 0.33 and < 3). Our cut-offs used were those 2012). Common involvement of brain-damaged regions recommended by Dienes (2014). One advantage of Bayes fac- across different patient groups was identified by overlapping tors is their robustness to small sample sizes that are under- individual normalized brain lesions on a standard template powered for conventional analyses (Vadillo et al. 2016). Bayes (i.e., AICHA—An atlas of intrinsic connectivity of homo- factors were calculated using the online-calculator provided topic areas; Joliot et al. 2015) in MRIcron. A summary of the by Dienes http://www.lifesci.susse x.ac.uk/home/Zolt an_Diene location and size of participant lesions was obtained using s/infer ence/Bayes .htm and where effect sizes came from our the descriptive tool. Due to our small sample size, these data previous study (Stöttinger et al. 2014). are underpowered for statistical analyses of lesion location and performance scores [e.g., voxel-based lesion-symptom mapping; (VLSM; Bates et al. 2003)]. Thus, while acknowl- edging the exploratory nature of the data, we include them for comparison with prior reports, and for the purpose of generating structure–function hypotheses (Danckert et al. 2012b; Stöttinger et al. 2014, 2015). 1 3 1758 Experimental Brain Research (2018) 236:1749–1765 patient in the random condition was excluded from further Results analysis as the patient did not recognize one object (i.e., the patient failed to identify the spider in the spider–sun pic- Online study ture set even at the first picture position—100% spider/0% sun). See Table 3 for average performance in all three con- Seventy six of the 77 participants reported all catch trials ditions for all groups (note: smaller numbers reflect better correctly. Mean percentages of first object reports are dis- performance). played in Fig. 3. We refrained from comparing performance in the repeat condition with either the gradual or random Gradual condition conditions as morphing direction was not counterbalanced. A repeated measures ANOVA revealed a significant main Compared to the HCO-age-matched group, results dem- effect for picture position [F (14,1064) = 2543.29, p < .001, onstrated a significant main effect for image number η = .97]. As evident in Fig. 3, probability of first answer [F(14,392) = 353.04, p < .001, η = .93] due to a decrease reports decreased as a function of picture position. A sig- of first object reports as a function of picture position, nificant main effect for condition [F (1,76) = 31.35, p < .001, as well as a significant main effect for participant group η = .29] showed that participants reported significantly [F(2,28) = 9.47, p = .001, η = .40] (Fig.  4, left panel). A fewer first objects in the gradual (mean 7.36, SD 0.76) com- post hoc Bonferroni test showed that age-matched HCO and pared to the random condition (mean 8.00, SD 0.72). The LBD patients had a comparable percentage of first object interaction between condition x picture position showed that reports (p = .16). Age-matched HCO had a significantly participants reported the second object earlier in the grad- lower percentage of first object reports than RBD patients ual compared to the random context [F(14,1064) = 13.04, (p = .001). This analysis also demonstrated a significant p < .001, η = .15]: at picture #7, 8, 9 and 11 probability of interaction between image number and participant group first object reports was significantly lower in the gradual [F(14,392) = 3.53, p < .001, η = .20]. Independent samples t compared to the random condition (all p’s < .05). These find - test (Bonferroni corrected) revealed that both patient groups ings replicate Stöttinger et al. (2016). showed a significantly higher percentage of first object reports at picture image #9 and 10 compared to HCO-age Patients matched controls. The same repeated measures analysis restricted to patient Patient participants identified all catch trials correctly, with groups (i.e., RBD vs. LBD) demonstrated a significant main the exception of one patient on one occasion. One set for one effect for picture position [F (14,280) = 206.97, p < .001, η = .91] and a marginally significant main effect for par - ticipant group: RBD patients tended to have a higher pro- portion of first object reports compared to LBD patients [F(1,20) = 4.12, p = .056, η = .17]. To further evaluate the difference in LBD and RBD first object reports, we calcu- lated the Bayes factor (Dienes 2014). We found a Bayes fac- tor of 3.24, indicating support for a greater updating impair- ment after damage to the right side of the brain. The same analysis conducted with the HCO-cognitively matched group revealed similar results. There was a sig- nificant main effect for image number [F (14,392) = 301.09, p < .001, η = .92] and participant group [F(2,28) = 5.33, p = .011, η = .28]. Age-matched HCO had a comparable percentage of first object reports to LBD patients (p = .88), and a significantly lower percentage of first object reports than RBD patients (p = .01). Fig. 3 Average % of first object reports in the online study col- This analysis also showed a significant interac- lapsed across all picture sets and all participants in each condition. tion between picture position and participant group The x-axis represents picture position (100% first object to 0% first object). Participants stopped reporting the first object earlier, when [F(28,392) = 1.55, p = .038, η = .10]. HCO-cognitively it was presented in the gradual condition (blue line) compared to the matched controls showed a comparable percentage of first random condition (red line). At picture #8 (i.e., vertical dotted line), object reports to LBD patients at all picture positions (all the picture was composed equally of both pictures (for interpretation p’s > .05), and significantly lower percentage of first object of the references to color in this figure legend, the reader is referred to the web version of this article) reports than RBD patients at picture #7 (p = .048) (Fig. 4). 1 3 Experimental Brain Research (2018) 236:1749–1765 1759 Table 3 Average performance of patients (a) and HCO (b) ID Gradual Random Repeat Difference (a) Performance of patients  Left brain damaged patients   110 6.00 8.50 7.50 2.50   442 8.50 8.75 6.50 0.25   588 5.75 6.50 7.00 0.75   788 7.50 8.50 9.00 1.00   828 7.75 7.50 7.00 − 0.25   835 6.25 6.25 6.50 0.00   838 7.50 7.25 9.50 − 0.25   872 7.75 7.00 9.00 − 0.75   898 8.50 9.00 10.00 0.50   902 8.00 7.75 6.50 − 0.25  Right brain damaged patients   27 7.25 8.00 7.50 0.75   205 8.50 7.75 8.00 − 0.75   228 9.50 7.00 8.50 − 2.50   284 10.25 6.75 7.00 − 3.50   489 8.50 9.75 8.00 1.25   729 7.50 7.25 6.00 − 0.25   744 9.25 7.25 10.50 − 2.00   792 9.75 6.75 10.00 − 3.00   856 6.75 7.00 5.50 0.25   874 6.50 5.75 7.00 − 0.75   932 7.25 8.50 11.50 1.25   946 9.25 7.25 10.00 − 2.00 ID Gradual Random Random Difference (b) Performance of healthy controls  HCO (≤ 70 years) (age-matched)   1 6.25 7.25 6.00 1.00   46 5.75 7.00 5.50 1.25   110 6.75 7.25 9.50 0.50   143 6.75 7.50 7.00 0.75   193 6.25 7.50 5.50 1.25   230 6.25 8.00 7.50 1.75   249 6.25 8.50 6.00 2.25   351 6.00 7.75 7.00 1.75   408 7.50 7.25 6.00 − 0.25  HCO (>70 years) (cognitively-matched)   3 5.75 8.00 8.50 2.25   32 6.25 7.25 6.00 1.00   37 7.75 6.80 7.50 − 0.95   148 7.25 7.75 8.50 0.50   206 7.50 8.25 8.50 0.75   208 8.00 7.00 5.50 − 1.00   321 5.00 6.75 6.00 1.75   369 6.75 9.50 8.00 2.75   409 7.00 7.25 8.50 0.25 a b Neglect at time of screening ( and at time of testing): please note that there was no significant difference in any of the dependent measures between participants who showed neglect at screening (or at the time of this testing) and other RBD patients (all p’s > .05) Time since stroke was significantly longer than time since stroke for the other patients [t(20) = 15.58, p < .01]. This, however, did not result in better performance in this patient as evident in the table Italic values represent performance 2 SDs outside the range of controls 1 3 1760 Experimental Brain Research (2018) 236:1749–1765 Fig. 4 Average % of first object reports for gradual (left) random 70, gray lines = HCO-Over 70. Results from the online study are (middle) and repeat (right) conditions. The x-axis represents picture included for comparison purposes (blue dotted lines = HCO-online) position (100% first object to 0% first object). At the vertical dot- (for interpretation of the references to color in this figure legend, the ted line, the picture is composed of 50% first object and 50% second reader is referred to the web version of this article) object. Green lines = LBD, red lines = RBD, black lines = HCO-Under Random condition p < .001, η = .83]. No other main effect or interaction reached significance (all p’s > .05). While RBD and LBD The same analyses conducted for averaged percentage of patients demonstrated comparable performance, a Bayes fac- tor of 0.57 suggest that our data is not sensitive enough to first object reports in the random condition revealed a signif- icant main effect for image number for both analyses—with draw a definite conclusion of no patient group differences. The same repeated measures ANOVA conducted with age-matched HCO as controls [F(14,392) = 403.85, p < .001, η = .94] and with HCO matched on cognitive impairment HCO-cognitively-matched controls revealed no significant effect other than a significant main effect for image posi- [F(14,392) = 361.12, p < .001, η = .93]. Percentage of first object reports decreased as a function of picture position tion [F(14,392) = 141.84, p < .001, η = .84]. No significant difference was found for average percentage of first object in both analyses. No other effects or interactions were sig- nificant (all p ’s > .05) indicating that all groups performed reports between HCO-cognitively matched and LBD (p = 1) or RBD patients (p = .68). Results therefore demonstrate that at equivalent levels (Fig. 4). A Bayes factor of 0.13 for the difference between RBD and LBD patients confirmed the when the control group was matched for cognitive impair- ment, no difference was found between groups (Fig.  4). likelihood of equal performance between the two patient groups. Difference: random minus gradual Repeat condition Healthy individuals see the second object earlier when The percentage of first object reports in the repeat condi- the pictures are presented in a gradual morphing context (Fig. 3). We submitted the difference scores (average pro- tion was submitted to the same repeated measures ANOVA described above and revealed a significant main effect portion of first object reports in random–gradual conditions; Table 3) to two separate univariate ANOVAs with partici- for image number with HCO-age-matched as controls [F(14,392) = 140.03, p < .001, η = .83]: the percentage of pant group (RBD, LBD, and either HCO-age matched or HCO-cognitively matched) as the between subject factor. first object reports decreased as a function of picture posi- tion. We also found a trend towards a significant effect for Larger values for this difference score indicate greater ben- efit from the gradual condition. The analysis with HCO-age participant group [F(2,28) = 2.82, p = .077, η = .17]. HCO- age-matched controls showed a slight, but not significantly matched as a control group showed a significant main effect (Fig. 5) [F(2,28) = 7.74, p = .002, η = .36]. RBD patients lower percentage of first object reports compared to RBD patients (p = .075), and a comparable performance to LBD had significantly smaller difference scores compared to LBD patients [F(1,20) = 4.82, p = .04, η = .19] and HCO-age- patients (p = .76). Calculating the analysis for patient groups only (RBD vs. LBD) revealed a significant main effect for matched controls: F(1,19) = 12.25, p = .002, η = .39; LBD patients tended to have smaller difference compared to age- picture position, again showing a decrease of first object reports as a function of picture position [F(14, 280) = 99.94, matched HCO [F(1,17) = 4.12, p = .058, η = .20]. 1 3 Experimental Brain Research (2018) 236:1749–1765 1761 HCO-cognitively matched group in the repeat condition. Three of these patients had overlap in the parietal cortex (postcentral, supramarginal gyrus, angular gyrus, inferior parietal gyrus, intraparietal sulcus), and the posterior insula and superior temporal gyrus (Fig. 6). LBD patients demon- strated less overlap in their lesions than did the RBD group with only four of ten patients showing common involvement of the putamen. Discussion The aim of the present work was to explore why RBD Fig. 5 Average (95% CI) difference scores (random–gradual) for patients show a selective and differential impairment of LBD (green), RBD (red), HCO-age-matched and HCO-cognitively- mental model updating, and to replicate results of our ear- matched (gray). Results of healthy young participants from the online lier study (Stöttinger et al. 2014). As in our previous study, study are included for comparison purposes (blue) (for interpretation of the references to color in this figure legend, the reader is referred to RBD patients needed significantly more pictures than HCO the web version of this article) before they reported seeing the second object in the gradual morphing condition—even after controlling for general The same analysis done for HCO-cognitively-matched cognitive impairment (Fig. 4). Comparisons between LBD group as controls also showed a significant main effect and RBD groups found clear evidence for worse updating [F (2,28) = 4.85, p = .016, η = .26]. HCO-cognitively in RBD patients (Bayes Factor 3.24). Further, five of the 12 matched controls showed a significantly larger benefit for RBD patients, but none of the ten LBD patients, performed gradual vs. random presentation than did RBD patients well outside the average performance of the control group [F(1,19) = 6.88, p = .017, η = .27] and a comparable benefit (i.e., mean of HCO-cognitively-matched ± 2 SD; Table 3). to LBD patients [F(1,17) = 0.81, p = .381, η = .05]. To test whether the updating impairment in RBD patients reflects (1) an inability to explore new hypotheses, (2) a Healthy controls—age effects general impairment of proactive switching or (3) a more general perceptual and/or attentional impairment, we com- The two HCO groups differed on age in the gradual con- pared three conditions: gradual, random and repeat. All dition only. The HCO-age-matched (≤ 70  years) group groups performed at the same level in the random condition, stopped reporting the first object earlier compared to the indicating that basic perceptual, attentional, and language HCO-cognitively-matched (> 70 years) counterparts as evi- capacities were equivalent in all participant groups (Bayes dent in a significant interaction between image number x factor of 0.13). To test whether RBD performance was due participant group (age matched vs. cognitively matched) to impaired ability to explore alternative interpretations, we [F(14,224) = 1.92, p = .026, η = .11] (Fig.  4, left panel). compared first object reports in the random condition with None of the other conditions revealed a significant differ - the gradual condition. In healthy individuals, we consist- ence between the two control groups (all p’s > .05). ently find that participants report the second object earlier when presented in a gradual context (Stöttinger et al. 2016; Lesion overlay analysis Fig. 3), reflective of active exploration strategies (i.e., “It was a shark but it could also be a plane or a bird.”). In the Among the 12 RBD patients, nine had overlapping lesions online study, we replicated this benefit showing the robust- in the insula (anterior and posterior), rolandic operculum, ness of this effect. This is an important point given the ongo- and the putamen. Five RBD patients were considered poor ing discussion about failures to replicate in social science explorers based on average first object reports and difference (Bohannon 2015). Results in the patient study showed a sig- scores more than two SDs above the mean of the HCO-cog- nificant difference between participant groups in how much nitively matched group (highlighted by gray bars in Table 3). they benefited from the gradual presentation. The strong- Four of these five patients had common involvement in the est effect was seen in healthy age-matched controls. This inferior frontal cortex, precentral gyrus, insula (anterior and benefit decreased with age and was about the same in the posterior), rolandic operculum, superior temporal gyrus, older seniors (matched on general cognitive impairment) as and the putamen (Fig. 6). Four RBD patients showed a poor in LBD patients (Fig. 5). In contrast, RBD patients exhibited general switching performance as indicated by average first a perceptual hysteresis—becoming “stuck” on their initial object reports that were two SDs above the mean of the interpretation even as the pictures gradually morphed into 1 3 1762 Experimental Brain Research (2018) 236:1749–1765 Fig. 6 Lesion overlay maps for LBD and RBD patients (top panel). Bottom panel shows 4 of the 5 RBD patients who were considered poor explorers (left panel) and 3 of the 4 RBD patients, considered poor proactive updaters (right panel). Shading indicates the amount of overlap 1 3 Experimental Brain Research (2018) 236:1749–1765 1763 something else. Combined with our previous work (Danck- uncertain environments has been found by several groups ert et al. 2012b; Stöttinger et al. 2014), this indicates that (Blanchard and Gershman 2018; Ohira et al. 2013, 2014; RBD participants are aware that things are changing in the Laureiro-Martínez et al. 2015). When we presented healthy perceptual representation they are forming, but are defi- individuals with some of the gradually morphing picture sets cient in exploring alternative models to encapsulate those used here in an fMRI experiment, we found that the anterior changes. That is, our RBD patients’ verbal reports of the insula was active not only at the actual time point of reported morphing images indicate that they perceive changes. What object change but also about five seconds before—consistent differentiated them from controls and LBD patients was with the involvement of the anterior insula in the explora- that they interpreted those changes within the context of tion of alternative choices (Stöttinger et al. 2015). We have their initial perceptual representation. Indeed, if the task we recently replicated this finding (Stöttinger et al. 2018). employed depended heavily on detecting the small (~ 4%) By directly comparing HCO groups, we found that the changes from one image to the next, one might assume that ability to recognize the second object is sensitive to age, it would be LBD that would lead to the greatest impairment, with older seniors needing significantly longer than younger as damage to this hemisphere is known to affect local as seniors to identify the second object in the gradual morphing opposed to global image processing (Martinez et al. 1997; condition. This is consistent with findings showing slower Robertson and Lamb 1991). Therefore, we suggest that the switch rates in ambiguous figures as people age (Ukai et al. deficit seen here in RBD patients is indicative of a failure to 2003). Since these groups were selected to match either the adapt to changing circumstances as they struggle to explore age or the cognitive status of the patient groups, it cannot be alternate hypotheses to explain the change. conclusively stated that age affects updating behavior. This Data in the repeat condition are harder to interpret given effect could also reflect general cognitive decline. Indeed, the smaller number of observations (i.e., two instead of four six of our 18 HCO had a MoCA score below 26 with four sets). In the repeat condition, there is no need to explore of these participants in the HCO-cognitively matched group alternate hypotheses—participants know that the plane will (> 70 years). According to the cutoff point initially pub - morph into a shark; they have seen it before. All that remains lished, this would be suggestive of a mild cognitive impair- is for participants to determine when to report seeing the ment (MCI; Nasreddine et al. 2005; Smith et al. 2007). How- shark. Performance of RBD patients as a group was not dif- ever, more recent studies suggest cut-off scores of 22/23 to ferent from that of LBD patients or cognitively matched con- be more sensitive for MCI (Lee et al. 2008; Luis et al. 2009). trols in the repeat condition (Fig. 4). However, our rejection All of our HCO were well within this range. of the proactive switch account rests on the strength of a In summary, our study suggests that updating impair- null result, and our Bayes factor of 0.57 suggests that these ments seen after damage to the right side of the brain most negative data are not strong. Further comparisons of gradual likely represent an impairment in the ability to explore the and repeat conditions will be necessary to evaluate whether alternative interpretations within a slowly changing envi- updating impairments in RBD patients are due only to fail- ronment. A better understanding of this type of cognitive ures of hypothesis exploration, proactive switching, or both. impairment is relevant for developing an improved under- Although RBD patients performed more poorly than HCO standing of post-stroke behavioral deficits and prompting the and LBD patients, it is also evident that performance within study of specific rehabilitation procedures. the RBD patients was heterogeneous (Table 3). Some RBD Acknowledgements Open access funding provided by Austrian Sci- patients performed within the norm of cognitively-matched ence Fund (FWF). This work was supported by the FWF Austrian controls, while others were outside this range. There was Science Fund (#V480-B27) Grant (Eliese Richter Program) to E.S., also heterogeneity in the type of impairment observed, with the Natural Sciences and Engineering Research Council of Canada some patients only impaired in one regard (e.g., exploring) Discovery (#261628-07), Canada Research Chair grants, Heart and Stroke Foundation of Ontario #NA 6999 to J.D. and Canadian Institutes while others had broader impairments (Table 3). Given our of Health Research #219972 operating grant to J.D. and B.A.. The sample sizes the lesion overlay analyses should be regarded above-mentioned funding agencies had no role in the study design, as exploratory. Nevertheless, they are consistent with general data collection and analysis, decision to publish, or the preparation of ideas about the systems important for adapting to chang- the manuscript. ing and uncertain circumstances. A failure to explore was Open Access This article is distributed under the terms of the Crea- mostly associated with damage to frontal regions (frontal tive Commons Attribution 4.0 International License (http://creat iveco operculum, insula), while a general impairment in alternat- mmons.or g/licenses/b y/4.0/), which permits unrestricted use, distribu- ing between interpretations (as indexed by the repeat condi- tion, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the tion) was associated more prominently with parietal dam- Creative Commons license, and indicate if changes were made. age (Fig. 6). In healthy individuals, an association between the right insula—commonly damaged in our patients—and exploration of alternative interpretations in stochastic and 1 3 1764 Experimental Brain Research (2018) 236:1749–1765 Joliot M, Jobard G, Naveau M, Delcroix N et al (2015) AICHA: an References atlas of intrinsic connectivity of homotopic areas. J Neurosci Methods 254:46–59 Bates E, Wilson SM, Saygin AP, Dick F, Sereno MI, Knight RT, Laureiro-Martínez D, Brusoni S, Canessa N, Zollo M (2015) Under- Dronkers NF (2003) Voxel-based lesion–symptom mapping. Nat standing the exploration–exploitation dilemma: an fMRI study of Neurosci 6:448–450 attention control and decision-making performance. Strat Manag Blanchard TC, Gershman SJ (2018) Pure correlates of exploration and J 36:319–338 exploitation in the human brain. Cogn Affect Behav Neurosci Lee Y, Dong Woo L, Cho S et al (2008) Brief screening for mild cog- 18(1):117–126 nitive impairment in elderly outpatient clinic: validation of the Bohannon J (2015) Many psychology papers fail replication test. Sci- Korean version of the Montreal cognitive assessment. J Geriatr ence 349:910–911 Psychiatry Neurol 21:104–110 Britz J, Landis T, Michel CM (2009) Right parietal brain activity pre- Long GM, Toppino TC (2004) Enduring interest in perceptual cedes perceptual alternation of bistable stimuli. Cereb Cortex ambiguity: alternating views of reversible figures. Psychol Bull 19:55–65 130:748 Britz J, Pitts MA, Michel CM (2011) Right parietal brain activity pre- Luis CA, Keegan AP, Mullan M (2009) Cross validation of the Mon- cedes perceptual alternation during binocular rivalry. Hum Brain treal cognitive assessment in community dwelling older adults Mapp 32:1432–1442 residing in the Southeastern US. J Geriatr Psychiatry 24:197–201 Brownell HH, Michel D, Powelson J, Gardner H (1983) Surprise but Martinez A, Moses P, Frank L, Buxton R, Wong E, Stiles J (1997) not coherence: sensitivity to verbal humor in right-hemisphere Hemispheric asymmetries in global and local processing: evi- patients. Brain Lang 18:20–27 dence from fMRI. Neuroreport 8:1685–1689 Danckert J, Stöttinger E, Anderson B (2012a) Neglect as a disorder of McGuire JT, Nassar MR, Gold JI, Kable JW (2014) Functionally dis- representational updating. Psychology of neglect. NOVA Science sociable influences on learning rate in a dynamic environment. Publishers, New York, pp 1–28 Neuron 84:870–881 Danckert J, Stöttinger E, Quehl N, Anderson B (2012b) Right hemi- Meenan JP, Miller LA (1994) Perceptual flexibility after frontal or sphere brain damage impairs strategy updating. Cereb Cortex temporal lobectomy. Neuropsychologia 32:1145–1149 22:2745–2760 Nasreddine ZS, Phillips NA, Bédirian V, Charbonneau S, Whitehead V, Dienes Z (2014) Using Bayes to get the most out of non-significant Collin I, Cummings JL, Chertkow H (2005) The Montreal cogni- results. Front Psychol 5:781 tive assessment, MoCA: a brief screening tool for mild cognitive Egré P, De Gardelle V, Ripley D (2013) Vagueness and order effects in impairment. J Am Geriatr Soc 53:695–699 color categorization. J Logic Lang Inform 22(4):391–420 Ohira H, Matsunaga M, Murakami H, Osumi T, Fukuyama S, Shinoda Égré P, Ripley D, Verheyen S (2018) The sorites paradox in psychol- J, Yamada J (2013) Neural mechanisms mediating association of ogy. In Oms S, Zardini E (eds) The sorites paradox. Cambridge sympathetic activity and exploration in decision-making. Neuro- University Press, Cambridge science 246:362–374 Ferber S, Danckert J (2006) Lost in space: the fate of memory represen- Ohira H, Ichikawa N, Kimura K, Fukuyama S, Shinoda J, Yamada J tations for non-neglected stimuli. Neuropsychologia 44:320–325 (2014) Neural and sympathetic activity associated with explo- Filipowicz A, Anderson B, Danckert J (2016) Adapting to change: the ration in decision-making: further evidence for involvement of role of the right hemisphere in mental model building and updat- insula. Front Behav Neurosci 8:381 ing. Can J Exp Psychol 70:201 Piper BJ, Li V, Eiwaz MA et al (2012) Executive function on the psy- Freitas S, Simoes MR, Alves L, Vicente M, Santana I (2012) Mon- chology experiment building language tests. Behav Res Method treal cognitive assessment (MoCA): validation study for vascular 44:110–123 dementia. J Intern Neuropsychol Soc 18:1031–1040 Raffman D (2011) Vagueness and observationality. In: Ronzitti G (ed) Geng JJ, Vossel S (2013) Re-evaluating the role of TPJ in atten- Vagueness: a guide. Springer, Dordrecht, pp 107–121 tional control: contextual updating? Neurosci Biobehav Rev Robertson LC, Lamb MR (1991) Neuropsychological contributions to 37:2608–2620 theories of part/whole organization. Cogn Psychol 23:299–330 Grant DA, Berg E (1948) A behavioral analysis of degree of reinforce- Rorden C, Karnath HO, Bonilha L (2007) Improving lesion–symptom ment and ease of shifting to new responses in a Weigl-type card- mapping. J Cogn Neurosci 19:1081–1088 sorting problem. J Exp Psychol 38:404 Rorden C, Bonilha L, Fridriksson J, Bender B, Karnath HO (2012) Griffin R, Friedman O, Ween J, Winner E, Happé F, Brownell H (2006) Age specific CT and MRI templates for spatial normalization. Theory of mind and the right cerebral hemisphere: refining the Neuroimage 61:957–965 scope of impairment. Laterality 11:195–225 Sepahvand N, Stöttinger E, Danckert J, Anderson B (2014) Sequential Griffiths TL, Tenenbaum JB (2012) Optimal predictions in everyday decisions: a computational comparison of observational and rein- cognition. Psychol Sci 17:767–773 forcement accounts. PLoS One 9:e94308 Happé F, Brownell H, Winner E (1999) Acquired theory of mind Shaqiri A, Anderson B (2013) Priming and statistical learning in right ‘impairments following stroke. Cognition 70:211–240 brain damage patients. Neuropsychologia 51:2526–2533 Hepworth L, Rowe F, Walker M, Rockliffe J, Noonan C, Howard C, Shaqiri A, Anderson B, Danckert J (2013) Statistical learning as a tool Currie J (2015) Post-stroke visual impairment: a systematic litera- for rehabilitation in spatial neglect. Front Hum Neurosci 7:224 ture review of types and recovery of visual conditions. Phthalmol Smith T, Gildeh N, Holmes C (2007) The Montreal cognitive assess- Res An Intern J 5:1–43 ment: validity and utility in a memory clinic setting. Can J Psy- Husain M, Mannan S, Hodgson T, Wojciulik E, Driver J, Kennard C chiatry 52:329–332 (2001) Impaired spatial working memory across saccades contrib- Sterzer P, Kleinschmidt A (2007) A neural basis for inference in per- utes to abnormal search in parietal neglect. Brain 124:941–952 ceptual ambiguity. Proc Natl Acad Sci 104:323–328 Johnson-Laird PN (2004) The history of mental models. In: Mank- Stöttinger E, Filipowicz A, Marandi E, Quehl N, Danckert J, Anderson telow K, Chung MC (eds) Psychology of reasoning: theoretical B (2014) Statistical and perceptual updating: correlated impair- and historical perspectives. Psychology Press, Hove, pp 179–212 ments in right brain injury. Exp Brain Res 232:1971–1987 Stöttinger E, Filipowicz A, Valadao D, Culham JC, Goodale MA, Anderson B, Danckert J (2015) A cortical network that marks the 1 3 Experimental Brain Research (2018) 236:1749–1765 1765 moment when conscious representations are updated. Neuropsy- Vocat R, Saj A, Vuilleumier P (2012) The riddle of anosognosia: does chologia 79:113–122 unawareness of hemiplegia involve a failure to update beliefs? Stöttinger E, Sepahvand NM, Danckert J, Anderson B (2016) Assess- Cortex 49:1771–1781 ing perceptual change with an ambiguous figures task: normative Weed E, McGregor W, Nielsen JF, Roepstor ff A, Frith U (2010) Theory data for 40 standard picture sets. Behav Res Method 48:201–222 of mind in adults with right hemisphere damage: what’s the story? Stöttinger E, Aichhorn M, Anderson B, Danckert J (2018) The neural Brain Lang 113:65–72 systems for perceptual updating. Neuropsychologia. https ://doi. Weilnhammer VA, Ludwig K, Hesselmann G, Sterzer P (2013) Fron- org/10.1016/j.neuro psych ologi a.2018.03.017 toparietal cortex mediates perceptual transitions in bistable per- Tenenbaum JB, Kemp C, Griffiths TL, Goodman ND (2011) How ception. J Neurosci 33:16009–16015 to grow a mind: statistics, structure, and abstraction. Science Wilson B, Cockburn J, Halligan P (1987) Development of a behavioral 331:1279–1285 test of visuospatial neglect. Arch Phys Med Rehabil 68:98–102 Ukai K, Ando H, Kuze J (2003) Binocular rivalry alternation rate Zaretskaya N, Thielscher A, Logothetis NK, Bartels A (2010) Disrupt- declines with age. Percept Mot Skills 97:393–397 ing parietal function prolongs dominance durations in binocular Vadillo MA, Konstantinidis E, Shanks DR (2016) Underpowered sam- rivalry. Curr Biol 20:2106–2111 ples, false negatives, and unconscious learning. Psychol Bull Rev 23:87–102 1 3

Journal

Experimental Brain ResearchSpringer Journals

Published: Apr 12, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off