Inversion effects for faces and objects in developmental prosopagnosia: A case series analysis

Author’s Accepted Manuscript Inversion effects for faces and objects in developmental prosopagnosia: A case series analysis Solja K. Klargaard, Randi Starrfelt, Christian Gerlach www.elsevier.com/locate/neuropsychologia

PII: DOI: Reference:

S0028-3932(18)30119-2 https://doi.org/10.1016/j.neuropsychologia.2018.03.026 NSY6729

To appear in: Neuropsychologia Received date: 9 July 2017 Revised date: 2 March 2018 Accepted date: 21 March 2018 Cite this article as: Solja K. Klargaard, Randi Starrfelt and Christian Gerlach, Inversion effects for faces and objects in developmental prosopagnosia: A case series analysis, Neuropsychologia, https://doi.org/10.1016/j.neuropsychologia.2018.03.026 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Inversion effects for faces and objects in developmental prosopagnosia: A case series analysis.

Solja K. Klargaard*a, Randi Starrfeltb, Christian Gerlacha a

Department of Psychology, University of Southern Denmark, Denmark;

b

Department of Psychology, University of Copenhagen, Denmark

* Corresponding author: Solja K. Klargaard, Department of psychology, Campusvej 55, 5230 Odense M, Denmark; E-mail: [email protected]

Abstract The disproportionate face inversion effect (dFIE) concerns the finding that face recognition is more affected by inversion than recognition of non-face objects; an effect assumed to reflect that face recognition relies on special operations. Support for this notion comes from studies showing that face processing in developmental prosopagnosia (DP) is less affected by inversion than it is in normal subjects, and that DPs may even display face inversion superiority effects, i.e. better processing of inverted compared to upright faces. To date, however, there are no reports of direct comparisons between inversion effects for faces and objects, investigating whether the altered inversion effect in DP is specific to faces. We examined this question by comparing inversion effects for faces and cars in two otherwise identical recognition tasks in a group of DPs (N = 16) and a matched control group, using a case series design. Although both groups showed inversion

1

effects for both faces and cars, only the control group exhibited a significant dFIE, i.e. a larger inversion effect for faces than cars. In comparison, the DPs were not significantly more affected by inversion than the control group when assessed with a face processing task that did not require recognition. Importantly, in both settings the DPs are better with upright than with inverted faces, and on the individual level no DP was found to perform significantly better with inverted than with upright faces. In fact, the DPs are impaired relative to the control group with both upright and inverted faces and to a less extent also with upright and inverted cars. These results yield no evidence of inversion superiority in DP but rather suggest that their face recognition problem is not limited to operations specialized for upright faces. 1. Introduction The disproportionate Face Inversion Effect (dFIE) refers to the normal finding that recognition of faces is more disrupted when stimuli are turned upside down (i.e. inverted) than recognition of other objects, e.g. cars, landscapes, houses (Yin, 1969). The dFIE is central to the debate on the nature of face processing which broadly can be lined up in two accounts: One account explains the dFIE in terms of face specific processing (Valentine, 1988; Yin, 1969), and the other in terms of expertise (Diamond & Carey, 1986; Tanaka & Gauthier, 1997) suggesting that with similar levels of expertise, the inversion effect can be as strong with other stimulus types as with faces. Both accounts share the interpretation that the dFIE is a demonstration of how normal and efficient face recognition relies on holistic / configural processing to a greater degree than recognition of non-face objects. The holistic face processing account suggests that a substantial part of the inversion effect can be explained by difficulties in applying experience-driven holistic templates (or

2

gestalts) to the inverted face (Rossion & Boremanse, 2008). According to this account, processing becomes predominantly feature based when faces are inverted. In accord with this, some have demonstrated spared featural processing when faces are inverted (Freire, Lee, & Symons, 2000; Leder & Bruce, 2000). Others, however, have demonstrated that processing of isolated features [e.g. shape of nose and mouth] are also disrupted by inversion (Civile, McLaren, & McLaren, 2016; Rakover & Teucher, 1997; Yovel & Duchaine, 2006) though surface properties as color and brightness are not (Leder & Bruce, 2000; Leder & Carbon, 2006); for a review see McKone and Yovel (2009). However, not all configural and featural information is disrupted to the same degree by inversion. In normal subjects, the eye region in particular seems to be spared following inversion (Rakover, 2010); also when it comes to horizontal spacing between the eyes, i.e. relational information (Goffaux & Rossion, 2007). In this study we asked whether participants with abnormal face processing exhibit atypical face inversion effects or even atypical dFIEs. Note that we reserve the term “dFIE” for instances where inversion effects for faces have been compared with inversion effects for a nonface contrast category. Only in the presence of such non-face contrasts can the specificity of an inversion effect be established. In cases where face inversion effects are reported without reference to a non-face contrast, it cannot be determined whether a reduced inversion effect is general (i.e., might affect processing of all stimulus categories) or specific to/disproportionate for faces. In developmental prosopagnosia (DP), where the ability to recognize faces never fully develops, the most common finding is a reduced or absent face inversion effect. Reduced inversion effects with faces has been reported both in single cases (de Gelder & Rouw, 2000a) and group studies (N = 14 DPs: Avidan, Tanzer, & Behrmann, 2011; N = 26 DPs: Russell, 3

Duchaine, & Nakayama, 2009; N = 16 DPs: Shah, Gaule, Gaigg, Bird, & Cook, 2015), and some even report completely absent inversion effects with faces (N = 5 DPs: Behrmann, Avidan, Marotta, & Kimchi, 2005). However, to our knowledge there are no reports of studies examining whether this reduction is specific to faces compared with objects, that is, whether DPs also show a reduced or abolished dFIE in tasks where normal subjects do show a dFIE. Interestingly, evidence suggests increased face inversion effects in people with superior face recognition abilities (also referred to as super-recognizers) compared to controls on the Cambridge Face Perception Test (CFPT; Russell et al., 2009). This lends support to the view that the size of the inversion effect with face stimuli can be a reasonable indicator of the level of face recognition aptitude. However, the interpretation of the reduced or absent face inversion effect in DP is not so straight forward, especially seen in the light of some more mixed results from patients with acquired prosopagnosia exhibiting superior performance with inverted relative to upright faces (de Gelder & Rouw, 2000b; Farah, Wilson, Drain, & Tanaka, 1995; Schmalzl, Palermo, Harris, & Coltheart, 2009). Similar face inversion superiority has been reported in DP. Huis in ‘t Veld et al (2012) for example reported that six out of ten DPs showed face inversion superiority on a delayed face matching task, indicating that they were better at matching inverted than upright faces. Behrmann et al. (2005) even concluded that: “… most CP [congenital prosopagnosics] subjects are better at inverted than upright faces” (p. 1142). As noted by Watson et al. (2016), such paradoxical inversion effects would suggest that something more is going on than mere disruption of holistic/configural processing together with increased reliance on feature processing, as this would not explain why patients do better with inverted faces. This takes us to the question of how reliable and common the inversion superiority effect is in DP, as this would influence how the face inversion effect in DP can be interpreted and understood.

4

Another issue relates to the interpretation of the inversion effect across different tasks; that is, to what degree the inversion effect can be universally understood across different task paradigms? For example, previous reports of reduced inversion effects in DP have been based on different discrimination/perception tasks, e.g. two-alternative forced choice matching tasks (e.g. Avidan et al., 2011; Behrmann et al., 2005) or sorting by degree of face similarity as the CFPT (Russell et al., 2009). However, performance on perception tasks can vary considerably in DP; and such tasks have also been shown to be less sensitive to inversion than memory tasks (Ulrich et al., 2017). Hence, it may be that inversion effects differ across tasks depending on whether they tap memorial in addition to perceptual processes. Other issues regarding the interpretation of face inversion effects include atypical fixation patterns in DP. In particular, some DPs fail to encode eye information to the same degree as controls as they spend more time fixating on the mouth than on the eyes when looking at upright faces (Bobak, Parris, Gregory, Bennetts, & Bate, 2017). This could have a qualitative impact on how face processing is disrupted by inversion in DP. In the present study we examine inversion effects on a face memory task, the Cambridge Face Memory Test (CFMT; Duchaine & Nakayama, 2006) in 16 adults with DP, and a control group matched on gender and age. We also examine inversion effects with non-face stimuli (cars) in a task similar to the CFMT, the Cambridge Car Memory Test (CCMT; Dennett et al., 2012). This aspect is important as dFIE’s cannot, as mentioned above, be assessed without reference to a control category (Bruyer, 2011). In comparison to controls, DPs are expected to show a reduced inversion effect for faces but not cars, that is showing a diminished dFIE or no dFIE at all. Next, we examine on a single case level if there are reliable instances of inversion superiority on any of these tasks. 5

Our second question is to what degree reduced face inversion effects in DP are detectable on a face perception task, and if there are reliable instances of inversion superiority on this task. This is examined by means of the Cambridge Face Perception Test (CFPT; Duchaine, Germine, & Nakayama, 2007). If face perception is less sensitive to disruption from inversion, we would expect to see reduced inversion effects in the DP group on the CFPT relative to the CFMT. Again, we look for instances of reliable inversion superiority on the face perception task.

6

2. General Method 2.1. Participants Our dataset consists of 16 adults (6 males) with DP (age: M = 41, SD = 11; educational level: M = 16, SD = 2), and a group of 32 typically developing controls, matched on age (M = 38, SD = 10), gender (20 females, 12 males) and educational level: (M = 16, SD = 2). All participants had normal (or corrected-to-normal) vision, no learning disability, and no known history of neurological damage or psychiatric illness. The participants provided written informed consent according to the Helsinki declaration. The Regional Committee for Health Research Ethics of Southern Denmark has assessed the project, and found that it did not need formal registration.

2.1.1. Developmental prosopagnosics All participants with DP have independently contacted our research group with subjective concerns about their ability to recognize faces, and have completed structured interviews regarding everyday difficulty with facial identity recognition and possible family history of DP. They all report difficulties recognizing friends, colleagues, and sometimes even close family members and themselves by their faces, and that these problems have been present throughout their life. The inclusion criteria for DP was a deficit in learning to recognize novel upright faces on the CFMT (Duchaine & Nakayama, 2006), and this was initially determined in relation to age and gender adjusted norms provided by Bowles et al. (2009). The participants also performed the CFPT (Duchaine et al., 2007). Both tests were kindly provided by Brad Duchaine and translated into Danish. All included DP’s also report severe difficulties with face recognition in their everyday life, as evaluated by the first part of the Faces and Emotion Questionnaire (FEQ, 297

items) -- (Freeman, Palermo, & Brock, 2015). The FEQ was translated to Danish by the last author. The final inclusion criteria for DP were based on an abnormal score of 2 SDs below the control group mean on the CFMT (see table 1) and above 2 SD of the control mean on the FEQ (controls: M = 19.2, SD = 10.1; DPs: M = 59.8, SD = 6.8). There was no significant correlation between the first part of the FEQ and the CFMT (r(16) = -.2, p = .45) or between the FEQ and the CFPT (r(16) = .23, p = .4), suggesting that these measures tap different aspects of face recognition efficiency. In comparison, there was a marginally significant correlation between the CFMT and the CFPT (r(16) = -.48, p = .06). All DPs performed within the normal range (score of 32 or less) on The AutismSpectrum Quotient (AQ) (Baron-Cohen, Wheelwright, Skinner, Martin, & Clubley, 2001). The DPs did not receive remuneration for their participation in this study. The DPs (and controls) included in this study have participated in extensive testing in our lab, and we report here all tests in which inversion effects were measured.

2.1.2. Control subjects Participants in the control group underwent that same test protocol as participants in the DP group. All controls performed within the normal range (not below 2 SDs) on the CFPT and the CFMT, evaluated by the Bowles et al. age and gender adjusted norms (2009). They also performed within the normal range (score of 32 or less) on the Autism-Spectrum Quotient (AQ). Controls received gift certificates of ~120 DKK (~20 USD) per hour for their participation.

8

2.2. General Procedure All participants completed the upright version of each task before completing the inverted version. The order of tasks was the same in the DP and control groups, starting with the CFMT, CFPT and then the CCMT. The tasks are described below.

2.3. Data analysis Inversion effect scores were computed as simple difference scores between upright and inverted trials for each subject. Group level inversion effects were examined by 3-way or 2-way ANOVAs with the between subject factor “group” and the within subject factors “orientation” and “stimulus type” when appropriate. Partial eta squared (ηp2) is reported as effect size. Simple main effects are analyzed with t-tests. For single case analysis, we applied Bayesian hypothesis testing to examine whether the individual DPs’ scores were significantly different from the control sample (program: SingleBayes_ES) (Crawford & Garthwaite, 2007). Age was applied as a covariate when it correlated significantly with the task in question (program: BDT_Cov) (Crawford, Garthwaite, & Ryan, 2011). We also applied Bayesian criteria for examining dissociations in single-case studies as suggested by Crawford et al. (2003), with a putatively classical dissociation being defined by: (i) impaired performance on one task, (ii) performance within the normal range on the other, and (iii) a significant difference between the two standardized scores (program: DissocsBayes_ES) (Crawford, Garthwaite, & Porter, 2010).

9

3. Experiment 1: Memory tasks 3.1. The Cambridge Face Memory Test (CFMT). In the CFMT (Duchaine & Nakayama, 2006), participants are asked to memorize faces of different individuals and later to point out the to-be-remembered face out of a line-up of three faces; this is done by pressing 1, 2, or 3 on the keyboard. All faces are male, in black and white and cropped so that no hair is visible. The test consists of a practice phase and three test phases of increasing difficulty. In the first test phase (Introduction/same images), a study face is shown from three different angles (3 secs for each angle) and one of the same study images is later to be pointed out from two additional faces; this is repeated for six target faces. In the second phase (Novel images), participants are presented with a review of the six target faces (20 secs) after which they are presented with a lineup of three faces out of which one is a novel image of the target face. The participant is then to identify which face is one of the six studies target faces. This is repeated 30 times, where a novel image of each target face is presented 5 times. In this phase, both lighting and pose are varied. In the third phase (Novel images with noise), noise is added to avoid ceiling effects. Participants are presented with a review of the target faces again for 20 secs, and are again asked to point out the target face among two other faces in a lineup. This is done four times for each test face (24 items in total); but otherwise the presentation is the same as in the second phase. Throughout, we report the scores on the original scale, with number of correct items out of 72. The internal consistency of the CFMT is high for upright faces, with previous reports ranging from Cronbach’s alpha = .83 in a German sample, N = 153 (Herzmann, Danthiir, Schacht, Sommer, & Wilhelm, 2008), Cronbach’s alpha = .89 in an Australian sample, n = 224 (Bowles et al., 2009) and between .86 and .88 based on various samples online and in laboratory settings (Wilmer et al., 2010). To our knowledge, Cronbach´s alpha has not been 10

reported for inverted trials on the original CFMT. All participants completed two versions of this test, i.e. with upright and inverted faces. All phases are repeated for the inverted face task.

3.2. The Cambridge Car Memory Test (CCMT). The CCMT was developed by Dennett et al. (2012) in order to be able to compare the recognition of faces with another real-world, three-dimensional, object class. Similar to faces, cars share components that are in a fixed relationship to each other (e.g. body, doors, wheels, windscreen, headlights). The CCMT is designed to match the CFMT. It is administered by the same procedure as the CFMT, with the same test phases, but instead of faces it uses cars as stimuli. Like the CFMT, it includes an upright and an inverted test condition. Regarding internal reliability, Dennett at al. (2012) reported a Cronbach alpha of .84 on upright car stimuli.

3.3. Results Results from the ANOVA showed a main effect of Group (F(1, 46) = 40.7, p < .001, ηp2= .47), a main effect of Stimulus type (F(1, 46) = 11.2, p = .01, ηp2= .20), a main effect of Orientation (F(1, 46)= 89.5, p < .001, ηp2= .66), a Group*Stimulus type interaction (F(1, 46)= 15.5, p < .001, ηp2= .25), a Group*Orientation interaction ( F(1, 46)= 16.2, p < .001, ηp2= .26), a Stimulus type*Orientation interaction (F(1, 46)= 43.8, p < .001, ηp2= .49), and finally, a Group*Stimulus type*Orientation interaction (F(1, 46)= 16.6, p < .001, ηp2= .27). The three-way interaction was further analyzed by conducting two-way ANOVAs separately for the DP group and the control group.

11

In the control group, there was a significant main effect of Orientation (F(1, 31)= 154.4, p < .001, ηp2= .83), and a significant Stimulus type*Orientation interaction (F(1, 31)= 93.7, p < .001, ηp2= .83). The main effect of Stimulus type was not significant (F(1, 31) = .022, p = .64). Planned comparisons revealed that: (i) upright stimuli were recognized more efficiently than inverted stimuli, this concerned both faces (t31 = 14, p < .001) and cars (t31 = 3.35, p < .01), (ii) upright faces were recognized more efficiently than upright cars (t31 = 4.47, p < .001) whereas (iii) the reverse was the case for inverted stimuli, where cars were recognized more efficiently than faces (t31 = -5.06, p < .001). Hence, the controls exhibited the classical dFIE, with significantly larger inversion effects with faces than with cars (see Figure 1; left panel). In the DP group, there was a main effect of Orientation (F(1, 15)= 8.9, p < .01, ηp2= .37), with overall higher scores on upright than on inverted stimuli, and a main effect of Stimulus type (F(1, 15)= 28.9, p < .001, ηp2= .66), with overall higher accuracy scores with cars than with faces. The interaction between Stimulus type and Orientation was not significant (F(1, 15)= 2.1, p = .17). Hence, the DP group did not demonstrate the typical dFIE (see Figure 1; right panel).

12

Figure 1. Left panel: Number of correct responses for upright and inverted faces (CFMT) and upright and inverted cars (CCMT) for the control subjects. Right panel: Number of correct responses for upright and inverted faces (CFMT) and upright and inverted cars (CCMT) for the DPs.

3.3.1. Single subject analysis All DPs met the Bayesian criteria for a deficit in recognizing upright faces (CFMT upright) compared to control sample, which is expected as this was an inclusion criterion. Four DPs (PP04, PP17, PP19, PP105) also showed a significant deficit in recognizing faces turned upside down (CFMT inverted). Nine DPs showed reduced face inversion scores (up-inv) to the degree that it met the criteria for a deficit (PP05, PP07, PP18, PP19, PP101, PP102, PP106, PP112, PP116), see table 1. Thereof, three DPs met the Bayesian criteria (two-tailed) for a putatively classical dissociation with impaired performance on the upright condition, performance within the normal range on the inverted condition, and a significant difference between the upright and inverted standardized scores: PP18: p < .05, ZDCC = -2.48, 95% CI [-4.0, -1.7], PP101: p < .05, ZDCC = -2.3, 95% CI [-3.2, -1.5], and PP116: p < .01, ZDCC = -3.1, 95% CI [-4.1, -2.2]. Two of these DPs (PP18 and PP116) had negative inversion scores, suggesting better performance with inverted than upright faces, but a chi-square test showed that the scores were not reliably higher in the inverted than in the upright condition for any of these individuals (PP18: diff = -1; PP116: diff = -4, all p > .05). With upright car stimuli (CCMT upright), three DPs met the criteria for a deficit, with significantly lower score than the control sample (PP17, PP25, PP106) whereas one DP showed a significantly higher score than controls (PP105). With car stimuli turned upside down (CCMT inverted), four DPs met the criteria for a deficit (PP07, PP09, PP17, PP18). Two DPs met the Bayesian criteria (two-tailed) for a deficit with significantly reduced inversion scores (up-inv) 13

compared to controls (PP102, PP106), while three DPs showed the opposite pattern with higher inversion scores than controls (PP07, PP09, PP105). In all, six DPs showed negative inversion scores (i.e., better performance with inverted than upright cars), of which only one showed reliably better performance on inverted car trials than on upright car trials (PP106: diff = -16, X2 = 7.84, p = .008). This same individual met the Bayesian criteria (two-tailed) for a putatively classical dissociation with impaired performance on the upright condition but with performance within the normal range on the inverted condition and a significant difference between the standardized scores on the upright compared to inverted condition (PP106: p < .005, ZDCC = -3.5, 95% CI [-4.5, -2.6]). The disproportionate face inversion effect (dFIE) was computed for each individual as the difference between the inversion effect (up-inv) with faces and cars (dFIE: IECFMT – IECCMT). Six individual DPs showed a significantly reduced dFIE compared to controls (PP07, PP09, PP18, PP101, PP112, PP116). That is, they did not show the typical effect of larger inversion effects for face stimuli than for other objects. Three DPs even met the criteria for a putatively classical dissociation with significantly higher inversion effects for cars than for faces relative to the controls: PP07: p < .005, ZDCC = -3.6, 95% CI [-4.6, -2.7]; PP09: p < .01, ZDCC = -2.8, 95% CI [-3.6, -2.0]; PP18: p < .05, ZDCC = -2.5, 95% CI [-3.3, -1.8].

4. Experiment 2: Perception task In experiment 1, we demonstrated that the DP group did not show the dFIE, i.e. larger inversion effect with faces than with cars. In this second experiment we examine if a diminished inversion effect in DP is also evident on a face perception task. 14

4.1. Cambridge face perception test (CFPT) In this test, subjects are asked to judge the similarity of faces by ranking six faces in descending order by how similar they are to a target face (Duchaine et al., 2007). The face stimuli are created by morphing a different face with a target face and manipulate them so they differ in their degree of similarity by a standard scale. There are eight trials with upright faces and eight trials with inverted faces, and in each trial, participants are given one minute to complete the ranking of similarity to the best of their ability. The sorting of faces in done with a computer mouse. The resulting scores are error scores, derived by summing the deviations of the six faces from their correct position after each trial. The final score is the total deviation score, which ranges from zero up to 93.3 errors at chance level. The CFPT has a Cronbach’s alpha of .74 for upright trials and .50 for inverted trials (Bowles et al., 2009). The difference between error scores on the upright and inverted test conditions is used as a measure of the inversion effect. To ease comparison of inversion effects on the CFPT and the CFMT, we subtracted the CFPTUprightscore from

the CFPTInverted-score so that positive difference scores on both tests signify better

performance with upright than inverted stimuli.

4.2. Results Results from the ANOVA showed a main effect of Group (F(1, 46) = 29.2, p < .001, ηp2= .39), with overall higher error rates in the DP group than in control group, and a main effect of Orientation (F(1, 46) = 91.7, p < .001, ηp2= .67), with overall more errors in the inverted compared to the upright condition. The Group*Orientation interaction was not significant (F(1, 46) = 1.6, p = .21), reflecting that the size of the inversion effect was not reliably larger in the control than in the DP group. Note that although the group difference is not statistically 15

significant with our sample size, the data shows a tendency towards larger inversion effects on the perception test in the control group (M = 22.1, SD = 12.9) than in the DP group (M = 16.9, SD = 14) as illustrated in Figure 2.

4.2.1. Single subject analysis Age correlated with performance in the upright condition of the CFPT (r = .4); and with age as a covariate, seven DPs met the Bayesian criteria for a deficit, that is, showing significantly more errors on the upright trials than controls (PP04, PP09, PP17, PP18, PP27, PP106, PP118). On the inverted condition of the CFPT, seven DPs showed significantly more errors than controls (PP04, PP09, PP17, PP18, PP105, PP106, PP116). One DP (PP118) showed a significantly reduced inversion effect (Inv-up), see table 1. Although this case had a negative inversion effect, a chi-square test (one-tailed) showed the difference was not significant (PP118: p = .071). Nevertheless, the case met the Bayesian criteria for a putatively classical dissociation (twotailed) with impaired performance on the upright condition and performance within the normal range on the inverted condition as well as a significant difference between the standardized upright and inverted scores.

16

Figure 2. Error bars represent 95% CI of the mean by group (DP, controls) and scatter represents individual scores. (a) inversion effect scores on the Cambridge Car Memory Test (CCMT IE) and the Cambridge Face Memory Test (CFMT IE); (b) the disproportionate Face Inversion effect, i.e. dFIE (CFMT IE – CCMT IE); (c) The inversion effect scores on the Cambridge Face Perception Test (CFPT IE), scores reversed to ease comparisons.

17

A correlation analysis was applied to examine if and to what degree there was a relationship between the inversion effect measure on the face perception task (IECFPT) and the inversion effect measure on the face memory task (IECFMT). This was done both across groups and separately for each group. These analyses showed no reliable association between the inversion effects in face memory and face perception for neither the across group analysis (r(48) = 11, p = .45) nor for the within-group analyses (controls: r(32) = -.07, p = .7; DP group: r(16) = .14, p = .61). This reflects that the size of the inversion effect on the face perception task is unrelated to the size of the inversion effect on the face memory task.

4.2.2. Reliability of inversion effect measures To assess the reliability of the inversion effect measures, we computed the Spearman-Brown reliability coefficients (rsb) based on the correlation between odd and even trials. This was done for the upright and inverted condition of each test as well as for the inversion effect measures; and for the DP and control groups separately as well as combined. These coefficients are reported in Table 2. On the CFMT upright, the reliability was high in the control group (rsb = .91) but low in the DP group (rsb = .16). This can be expected as the range of scores in the DP group is restricted because of impaired performance on this test. This restriction in the DP group was also reflected in lower reliability estimates on the CFMT inversion measure and the dFIE when looking at the DP group only. When looking across groups, the reliability estimates of the CFMT, and especially the inversion measure (IE) were reasonable (upright: rsb = .94; inverted: rsb = .84; IE: rsb = .81). This same a priori restriction on scores is not an issue with the other two tests, the CCMT and the CFPT, as the outcome on these tests was not part of the inclusion or

18

exclusion criteria. However, the reliability estimates of the CFPT were surprisingly low with the inversion measure (IECFPT = inverted score – upright score) showing no reliability (controls: rsb = -.09; DPs: rsb = .19; overall: rsb = -.02). This in turn may explain the lack of association between the inversion measures based on the CFMT and the CFPT.

Table 2. Split half (Odd/Even) Spearman-Brown reliability coefficients (rsb) for upright and inverted condition of the CFMT, CCMT and the CFPT and their inversion effect measures. Reliability coefficients are reported within (Control, DP) and across groups (All).

Number of items

Spearman-Brown reliability coefficients (rsb) Control (n = DP (n=16) All (n = 48) 32)

CFMT upright

72

.91

.16

.94

inverted

72

.79

.79

.84

.74

.41

.81

Inversion Effect (IE) CCMT upright

72

.79

.51

.73

inverted

72

.72

.80

.77

.58

.58

.58

.54

.26

.58

Inversion Effect (IE) dFIE: IECFMT – IECCMT CFPT upright

8

.50

.62

.70

inverted

8

.25

.26

.41

-.09

.19

-.02

Inversion Effect (IE)

19

5. Discussion The aim of this study was twofold. In the first experiment, which was based on a delayed matching task (Cambridge Face/Car Memory Tests), we examined whether the magnitude of the classical disproportionate face inversion effect (dFIE) would differ between a group of 16 adults with developmental prosopagnosia (DP) and an age and gender matched control group (N = 32). In the second experiment, with the same participants, we examined whether group differences in inversion effects would also manifest on a perceptual task with faces only (CFPT; Duchaine et al., 2007). In both experiments we were particularly interested in whether any of the DPs exhibited inversion superiority, that is, better performance with inverted than with upright faces. In the first experiment we found that controls showed the typical dFIE, but not the DPs. That is, while the DP group did exhibit better performance with upright than inverted faces, this difference was not significantly larger for faces than for cars. These findings were corroborated by the results from the single-case analyses showing that six DPs met the Bayesian criteria for an abnormal dFIE compared to controls, indicating that the dFIE is abnormal at an individual level in a large proportion of our DPs. Three DPs furthermore met the criteria for a putatively classical dissociation with higher inversion effects for cars than for faces relative to the controls. The upright version of the CFMT was used as a criterion measure for assessing face recognition ability in this study, and as expected, the DP group demonstrated overall impaired performance compared to controls on recognizing upright faces. However, the DP group also demonstrated overall impaired performance compared to controls on recognizing inverted faces, albeit to a lesser degree. On an individual level, four DPs met the criteria for a deficit in recognizing inverted faces. Moreover, the DP group demonstrated impaired performance on both

20

upright and inverted cars. These findings were again corroborated by the results from the singlecase analyses showing that three DPs were impaired with upright car stimuli, and four DPs with inverted car stimuli. This is in line with a previous study demonstrating a more general impairment in recognizing objects in a subsample of the present DPs (Gerlach, Klargaard, & Starrfelt, 2016). Returning to the present data, it is also important to note that the DP group did not show inversion superiority on the face memory task; and on an individual level, none of the DPs showed a reliable inversion superiority effect on the face memory task. Thus, no DP performed better with inverted than upright faces. In the second experiment, we examined if the inversion effect for faces would also be reduced on a perception task for the DPs. The results from this experiment indicated overall impaired performance in the DP group compared to controls on both upright and inverted trials. However, both the DP and the control group showed significant inversion effects, and the size of the inversion effect did not differ significantly between the groups, although there was a tendency towards a larger inversion effect in the control than in the DP group. A possible reason why we did not find a significantly reduced face inversion effect in the DP group could be the limited sample size (N = 16), as others have found diminished inversion effect in DP on this task with larger samples (N = 26; Russell et al., 2009, but see below). However, in line with previous studies, our results nevertheless suggest that memory based tasks may be more sensitive to inversion effects than perception based tasks (Schwaninger, Ryf, & Hofer, 2003). Finally, singlecase analysis of the face perception test showed that only one case met the criteria for a putatively classical dissociation with performance in the normal range for inverted faces but abnormal performance with upright faces, and a difference between these conditions that was

21

significantly smaller than that of the controls. None of the 16 DPs showed reliable inversion superiority on the face perception task. No DP showed significantly better performance with inverted than upright faces in either task, which contrasts with previous reports of face inversion superiority effects. Evidence for inversion superiority effects in prosopagnosia has come from single case studies of the acquired prosopagnosia patient LH (Farah et al. (1995); and later de Gelder et al. (2000b)). However, as pointed out by Busigny and Rossion (2010) the inversion superiority effect in this patient may have been caused by her upper visual field defect. Whether a similar explanation may apply to the case reported by Schmalzl et al. (2009) cannot be determined as no data on this patient’s visual fields are reported. Turning to developmental prosopagnosia there are also studies reporting cases with inversion superiority effects. Huis in't Veld et al. (2012) reported such an effect in 6/10 DP subjects. However, the results from this study are difficult to interpret for the following reasons: (i) The control group also showed face inversion superiority (d = .19), (ii) the inversion superiority effect in the DP group was not much larger numerically for faces (141 ms) than for shoes (122 ms), and (iii) the reliability of the inversion superiority effect for faces was not assessed at the individual level. In comparison, we also find individuals with higher raw scores on inverted than upright face conditions in the current study, but none showed a significant inversion superiority effect for faces. One DP in our sample, however, did show a significant inversion superiority effect, but this was on the car memory task. The same reservation applies to the study by Behrmann et al. (2005). In this study one case (BE) was reported to show 8% better performance with inverted than with upright faces on an individual level, but it was not reported whether the difference was significant. With these reservations in mind, we do not think that there is any solid evidence for inversion superiority in DP; a 22

conclusion that echoes the one reached by Busigny and Rossion (2010) for acquired prosopagnosia. As a second objective we were interested in quantifying the degree of shared variance between the face inversion effects as measured by the CFMT and the CFPT. Interestingly, there was no relation between the size of the inversion effect on the face memory and face perception tests in the controls or the DP group. Such a relation would have been expected if the measures gauged the same phenomenon. In fact, even if the tests differed in sensitivity to inversion, we would still expect some relation between the outcome on the face memory task and the face perception task if we were to claim that they are measuring the same underlying construct. However, it became clear that the assessment of such a correlation might have been obscured by the low reliability of the perception test, and particularly the inversion measure of this test. The size of the inversion effect based on the CFPT has been reported and interpreted as a promising indicator of the level of face recognition aptitude (Russell et al., 2009). This begs the question of whether inversion effect measures on face memory and face perception tasks should be taken to assess the same underlying construct. In our case, it seems rather that the two inversion measures reflect disruption of different underlying face processing mechanisms. It would be of interest for future studies to look into this aspect by applying tasks with reliable measures of face inversion effects. One possible limitation regarding the present results is that we examined inversion effects using one of the tests that was also used to classify the DPs in the first place; the CFMT. This might cause the inversion effect measured with the CFMT to be underestimated in the DP group, due to restriction of performance range (all DPs performed poorly with upright faces, potentially leaving limited room for inversion effects). It is difficult, however, to see how lower scores on 23

CFMTUpright can be avoided because even if the DPs were classified by some other means that the CFMT, they would still be expected to perform poorly on this task. Nevertheless, it is still possible to examine whether the inversion effect on the CFMT was underestimated due to restriction of performance range. If this is the case, we would expect the difference between CFMTUpright and CFMTInverted to be larger when cases scoring low on CFMTInverted are removed (potential floor effects) than when they are present. The reason is that cases scoring at floor level might in theory have scored even lower had the test been more sensitive. This is not the case for cases scoring above floor level. Hence, estimation of the difference between upright and inverted trials without inclusion of cases performing at floor level on inverted trials should give rise to a less restricted, and potentially larger, inversion effect. Considering that chance performance on the CFMT is 24, we tested this by removing all cases scoring below 30 on CFMTInverted (PP04, PP17, PP19 & PP105). The difference between CFMTUpright and CFMTInverted with these four cases removed was 4.7 which is actually less than the difference for the whole group which was 6.1. Hence, the inversion effect on the CFMT does not seem to be underestimated in the DP group due to floor effects on inverted trials. In sum, even if using the CFMT as both DP classifier and experimental variable might introduce a bias, there is no indication that this has caused the estimate of the inversion effect to be reduced in the DP group due to restriction of performance range; there was still potential room for the effect to have been larger. Several aspects of the present results that are noteworthy: First, we find that DP is associated with a reduced face inversion effect; even when compared with a contrast category (cars). Hence, while inversion affects face recognition disproportionally more than car recognition in the control group, this is not the case for the DPs neither at the group level nor in individual DPs. We also find that previous reports of face inversion superiority are not 24

reproducible. This aspect is important given that such paradoxical effects would be difficult to account for should the effect of face inversion reflect impaired holistic/configural processing (Watson et al., 2016). Our results, however, also indicate that face inversion effects differ across task. Not only were they larger on a task requiring face recognition (CFMT) than on a task requiring face perception only (CFPT), but they also did not correlate across these tasks. Given that face inversion is taken as a marker of holistic/configural processing, and indeed has been shown to be one of the best (less bad) predictors of face perception (Rezlescu, Susilo, Wilmer, & Caramazza, 2017), this is disappointing because it adds to the uncertainty regarding what exactly different measures of “holistic” measure (Rezlescu et al., 2017; Richler, Palmeri, & Gauthier, 2012; Sunday, Richler, & Gauthier, 2017). Even though the present study does not make us wiser regarding what specific process(es) inversion interferences with, the results do suggest that the DPs’ difficulties are not limited to upright faces. They are also impaired with inverted faces relative to controls, albeit less so. This could suggest that inversion does not disrupt holistic processing per se but rather impairs perceptual judgements performed with impoverished stimuli. On this account, the input (faces) is already degraded for the DPs when it is presented upright, because of their recognition difficulties, causing any further loss due to inversion to impact less on their performance relative to controls. If so, inversion might partially smooth out the processing disadvantage that the DPs already experience with upright faces bringing the two groups more on par. Evidence supporting such an, admittedly speculative, interpretation comes from a recent study by Murphy and Cook (2017) with normal subjects. They used a dynamic aperture paradigm in which faces are displayed in an incremental fashion preventing simultaneous processing of the whole face. Despite this degradation, recognition of faces presented in the aperture condition still suffered 25

from being inverted, and most importantly, the inversion effect was just as large for faces presented in the aperture conditions as it was for faces presented in their entirety. Hence, it might not be whole face (holistic) processing per se that that is disrupted by face inversion but rather the ability to process information efficiently in general. While this hypothesis cannot be assessed directly here, we note that the DPs did exhibit an across the board impairment in that they also performed worse than the controls with both upright and inverted cars, and this is at least consistent with a general deficit in (efficient) processing of visual information in faces and objects (see also Gerlach et al., 2016 for additional evidence of a general perceptual deficit in a subsample of the present DPs). What this account does not readily explain, however, is why such a deficit should be more harmful for face than for car recognition. One possibility, raised by Rezlescu et al. (2016), is that disproportional face inversion effects may only be found when the target category (faces) are more visually similar than the contrast category. In support of this they found similar inversion effects for faces and cars when the cars where highly visually similar, that is, members of the same car brand (BMWs) and of the same type (e.g., sedans or hatchbacks). In comparison, the cars in the CCMT are not of the same brand, and they even differ within trials in that different car types are mixed. Hence, according to the Rezlescu et al.’s rationale, the CCMT may not engage holistic processing because it can be performed by a feature-based strategy alone. In the present context this explanation has the desirable asset that it also accounts for the fact that DPs were not more impaired with processing of upright cars than they were with inverted cars relative to the controls.

In conclusion, we found the typical dFIE in the control group but not in the DP group in a memory task. In comparison, the DPs were not significantly more affected by inversion than the 26

control group when assessed with a face perception task (CFPT). In both settings the DPs are better with upright than with inverted faces, and on the individual level no DP was found to perform significantly better with inverted than with upright faces yielding no evidence of inversion superiority in DP. However, a lack of association between inversion effects on face memory and face perception tasks suggests that the face inversion effect should be understood in the context of the specific task requirements. Interestingly, the deficit in the present group of DPs was not limited to upright faces, as inverted faces were also affected. This suggest that their face recognition problem is not limited to operations specialized for upright faces. Indeed, the DPs also performed worse than the controls with both upright and inverted cars, suggesting that their deficit is not even limited to faces.

27

Acknowledgements We wish to thank the participating subjects and the Friends of Fakutsi Foundation. This project was supported by a grant from the Danish Research Council for Independent Research | Humanities (grant no. DFF – 4001-00115), to C.G. and R.S.

References Avidan, G., Tanzer, M., & Behrmann, M. (2011). Impaired holistic processing in congenital prosopagnosia. Neuropsychologia, 49(9), 2541–2552. https://doi.org/10.1016/j.neuropsychologia.2011.05.002 Baron-Cohen, S., Wheelwright, S., Skinner, R., Martin, J., & Clubley, E. (2001). The autismspectrum quotient (AQ): Evidence from asperger syndrome/high-functioning autism, malesand females, scientists and mathematicians. Journal of Autism and Developmental Disorders, 31(1), 5–17. Behrmann, M., Avidan, G., Marotta, J. J., & Kimchi, R. (2005). Detailed exploration of facerelated processing in congenital prosopagnosia: 1. Behavioral findings. Journal of Cognitive Neuroscience, 17(7), 1130–1149. Bobak, A. K., Parris, B. A., Gregory, N. J., Bennetts, R. J., & Bate, S. (2017). Eye-movement strategies in developmental prosopagnosia and “super” face recognition. The Quarterly Journal of Experimental Psychology, 70(2), 201–217. https://doi.org/10.1080/17470218.2016.1161059 Bowles, D. C., McKone, E., Dawel, A., Duchaine, B., Palermo, R., Schmalzl, L., … Yovel, G. (2009). Diagnosing prosopagnosia: Effects of ageing, sex, and participant–stimulus ethnic match on the Cambridge Face Memory Test and Cambridge Face Perception Test.

28

Cognitive Neuropsychology, 26(5), 423–455. https://doi.org/10.1080/02643290903343149 Bruyer, R. (2011). Configural Face Processing: A Meta-Analytic Survey. Perception, 40(12), 1478–1490. https://doi.org/10.1068/p6928 Busigny, T., & Rossion, B. (2010). Acquired prosopagnosia abolishes the face inversion effect. Cortex, 46(8), 965–981. https://doi.org/10.1016/j.cortex.2009.07.004 Civile, C., McLaren, R., & McLaren, I. P. (2016). The Face Inversion Effect: Roles of First and Second-Order Configural Information. The American Journal of Psychology, 129(1), 23– 35. Crawford, J. R., & Garthwaite, P. H. (2007). Comparison of a single case to a control or normative sample in neuropsychology: Development of a Bayesian approach. Cognitive Neuropsychology, 24(4), 343–372. https://doi.org/10.1080/02643290701290146 Crawford, J. R., Garthwaite, P. H., & Porter, S. (2010). Point and interval estimates of effect sizes for the case-controls design in neuropsychology: rationale, methods, implementations, and proposed reporting standards. Cognitive Neuropsychology, 27(3), 245–260. Crawford, J. R., Garthwaite, P. H., & Ryan, K. (2011). Comparing a single case to a control sample: testing for neuropsychological deficits and dissociations in the presence of covariates. Cortex, 47(10), 1166–1178. de Gelder, B., & Rouw, R. (2000a). Configural face processes in acquired and developmental prosopagnosia: evidence for two separate face systems? Neuroreport, 11(14), 3145–3150. de Gelder, B., & Rouw, R. (2000b). Paradoxical configuration effects for faces and objects in prosopagnosia. Neuropsychologia, 38(9), 1271–1279. Dennett, H. W., McKone, E., Tavashmi, R., Hall, A., Pidcock, M., Edwards, M., & Duchaine, B. (2012). The Cambridge Car Memory Test: A task matched in format to the Cambridge Face Memory Test, with norms, reliability, sex differences, dissociations from face 29

memory, and expertise effects. Behavior Research Methods, 44(2), 587–605. https://doi.org/10.3758/s13428-011-0160-2 Diamond, R., & Carey, S. (1986). Why faces are and are not special: an effect of expertise. Journal of Experimental Psychology: General, 115(2), 107. Duchaine, B., Germine, L., & Nakayama, K. (2007). Family resemblance: Ten family members with prosopagnosia and within-class object agnosia. Cognitive Neuropsychology, 24(4), 419–430. https://doi.org/10.1080/02643290701380491 Duchaine, B., & Nakayama, K. (2006). The Cambridge Face Memory Test: Results for neurologically intact individuals and an investigation of its validity using inverted face stimuli and prosopagnosic participants. Neuropsychologia, 44(4), 576–585. https://doi.org/10.1016/j.neuropsychologia.2005.07.001 Farah, M. J., Wilson, K. D., Drain, H. M., & Tanaka, J. R. (1995). The inverted face inversion effect in prosopagnosia: Evidence for mandatory, face-specific perceptual mechanisms. Vision Research, 35(14), 2089–2093. Freeman, P., Palermo, R., & Brock, J. (2015, August 4). Faces and Emotion Questionnaire. https://doi.org/10.6084/m9.figshare.1501441.v1 Freire, A., Lee, K., & Symons, L. A. (2000). The face-inversion effect as a deficit in the encoding of configural information: Direct evidence. Perception, 29(2), 159–170. Gerlach, C., Klargaard, S. K., & Starrfelt, R. (2016). On the relation between face and object recognition in developmental prosopagnosia: No dissociation but a systematic association. PloS One, 11(10), e0165561. Goffaux, V., & Rossion, B. (2007). Face inversion disproportionately impairs the perception of vertical but not horizontal relations between features. Journal of Experimental Psychology: Human Perception and Performance, 33(4), 995.

30

Herzmann, G., Danthiir, V., Schacht, A., Sommer, W., & Wilhelm, O. (2008). Toward a comprehensive test battery for face cognition: Assessment of the tasks. Behavior Research Methods, 40(3), 840–857. Huis in ’t Veld, E., Van den Stock, J., & de Gelder, B. (2012). Configuration perception and face memory, and face context effects in developmental prosopagnosia. Cognitive Neuropsychology, 29(5–6), 464–481. https://doi.org/10.1080/02643294.2012.732051 Leder, H., & Bruce, V. (2000). When inverted faces are recognized: the role of configural information in face recognition. The Quarterly Journal of Experimental Psychology. A, Human Experimental Psychology, 53(2), 513–536. https://doi.org/10.1080/713755889 Leder, H., & Carbon, C.-C. (2006). Face-specific configural processing of relational information. British Journal of Psychology, 97(1), 19–29. https://doi.org/10.1348/000712605X54794 McKone, E., & Yovel, G. (2009). Why does picture-plane inversion sometimes dissociate perception of features and spacing in faces, and sometimes not? Toward a new theory of holistic processing. Psychonomic Bulletin & Review, 16(5), 778–797. Murphy, J., & Cook, R. (2017). Revealing the mechanisms of human face perception using dynamic apertures. Cognition, 169, 25–35. Rakover, S. S. (2010). Configural processing hypothesis and face-inversion effect. Advances in Face Image Analysis: Techniques and Technologies: Techniques and Technologies, 316. Rakover, S. S., & Teucher, B. (1997). Facial inversion effects: Parts and whole relationship. Perception & Psychophysics, 59(5), 752–761. https://doi.org/10.3758/BF03206021 Rezlescu, C., Chapman, A., Susilo, T., & Caramazza, A. (2016). Large inversion effects are not specific to faces and do not vary with object expertise. PsyArXiv. https://doi.org/10.17605/OSF.IO/XZBE5 Rezlescu, C., Susilo, T., Wilmer, J. B., & Caramazza, A. (2017). The Inversion, Part-Whole, and Composite Effects Reflect Distinct Perceptual Mechanisms With Varied Relationships to Face Recognition. Retrieved from http://psycnet.apa.org/psycinfo/2017-16729-001/ 31

Richler, J. J., Palmeri, T. J., & Gauthier, I. (2012). Meanings, Mechanisms, and Measures of Holistic Processing. Frontiers in Psychology, 3. https://doi.org/10.3389/fpsyg.2012.00553 Rossion, B., & Boremanse, A. (2008). Nonlinear relationship between holistic processing of individual faces and picture-plane rotation: Evidence from the face composite illusion. Journal of Vision, 8(4), 3. https://doi.org/10.1167/8.4.3 Russell, R., Duchaine, B., & Nakayama, K. (2009). Super-recognizers: People with extraordinary face recognition ability. Psychonomic Bulletin & Review, 16(2), 252–257. https://doi.org/10.3758/PBR.16.2.252 Schmalzl, L., Palermo, R., Harris, I. M., & Coltheart, M. (2009). Face inversion superiority in a case of prosopagnosia following congenital brain abnormalities: what can it tell us about the specificity and origin of face-processing mechanisms? Cognitive Neuropsychology, 26(3), 286–306. https://doi.org/10.1080/02643290903086904 Schwaninger, A., Ryf, S., & Hofer, F. (2003). Configural information is processed differently in perception and recognition of faces. Vision Research, 43(14), 1501–1505. https://doi.org/10.1016/S0042-6989(03)00171-8 Shah, P., Gaule, A., Gaigg, S. B., Bird, G., & Cook, R. (2015). Probing short-term face memory in developmental prosopagnosia. Cortex, 64, 115–122. https://doi.org/10.1016/j.cortex.2014.10.006 Sunday, M. A., Richler, J. J., & Gauthier, I. (2017). Limited evidence of individual differences in holistic processing in different versions of the part-whole paradigm. Attention, Perception, & Psychophysics, 1–13. Tanaka, J., & Gauthier, I. (1997). Expertise in object and face recognition. Psychology of Learning and Motivation, 36, 83–125. Ulrich, P. I. N., Wilkinson, D. T., Ferguson, H. J., Smith, L. J., Bindemann, M., Johnston, R. A., & Schmalzl, L. (2017). Perceptual and memorial contributions to developmental

32

prosopagnosia. The Quarterly Journal of Experimental Psychology, 70(2), 298–315. https://doi.org/10.1080/17470218.2016.1177101 Valentine, T. (1988). Upside-down faces: a review of the effect of inversion upon face recognition. British Journal of Psychology (London, England: 1953), 79 ( Pt 4), 471– 491. Watson, R., Huis in ’t Veld, E. M. J., & de Gelder, B. (2016). The Neural Basis of Individual Face and Object Perception. Frontiers in Human Neuroscience, 10. https://doi.org/10.3389/fnhum.2016.00066 Wilmer, J. B., Germine, L., Chabris, C. F., Chatterjee, G., Williams, M., Loken, E., … Duchaine, B. (2010). Human face recognition ability is specific and highly heritable. Proceedings of the National Academy of Sciences, 107(11), 5238–5241. https://doi.org/10.1073/pnas.0913053107 Yin, R. K. (1969). Looking at upside-down faces. Journal of Experimental Psychology, 81(1), 141–145. https://doi.org/10.1037/h0027474 Yovel, G., & Duchaine, B. (2006). Specialized face perception mechanisms extract both part and spacing information: Evidence from developmental prosopagnosia. Journal of Cognitive Neuroscience, 18(4), 580–593.

33

Highlights · · · · ·

16 adults with developmental prosopagnosia (DP) and 32 matched controls were tested. No disproportionate face inversion effect was found in DP on memory tasks. DPs showed similar face inversion effects as controls on a perception task. No DP exhibited a reliable paradoxical inversion effect. No association was found between inversion effects on memory and perception tasks.

34

Gender

-3.4 -2.5 -2.8 -2.5 -3.6 -4.4 -3.9 -2.8 -2.6 -2.9 -2.2 -2.5 -2.5 -2.9 -2.9 -2.9

-3.0 0.6

39.8 4.1 60.4 7.0

Z-CC

37 43 41 43 35 30 33 41 42 40 45 43 43 40 40 40

Acc

CFMT Up

33.7 5.6 42.3 7.4

28 39 36 34 26 31 29 31 30 38 41 24 38 36 44 34

Acc

-1.2 0.8

-1.9 -0.5 -0.9 -1.1 -2.2 -1.5 -1.8 -1.5 -1.7 -0.6 -0.2 -2.5 -0.6 -0.9 0.2 -1.1

Z-CC

CFMT Inv

6.1 5.4 18.1 7.3

9 4 5 9 9 -1 a 4 10 12 2a 4 19 5 4 -4 a 6

Up-Inv

-1.6 0.7

-1.2 -1.9 -1.8 -1.2 -1.2 -2.6 -1.9 -1.1 -0.8 -2.2 -1.9 0.1 -1.8 -1.9 -3.0 -1.7

Z-CC

IE CFMT

46.1 8 52.4 7.4

42 47 55 47 33 41 40 39 45 49 49 66 39 43 56 47

Acc

-0.8 1.1

-1.4 -0.7 0.4 -0.7 -2.6 -1.5 -1.7 -1.8 -1.0 -0.5 -0.5 1.8 -1.8 -1.3 0.5 -0.7

Z-CC

CCMT Up

43.6 9.1 49.1 7.8

41 50 35 30 34 34 39 36 50 43 57 52 55 36 56 49

Acc

-0.7 1.2

-1.0 0.1 -1.8 -2.4 -1.9 -1.9 -1.3 -1.7 0.1 -0.8 1.0 0.4 0.8 -1.7 0.9 0.0

Z-CC

CCMT Inv

2.6 9.3 3.3 5.6

1 -3 20 a 17 a -1 7 1 3 -5 6 -8 14 -16 a 7 0 -2

Up-Inv

-0.1 1.7

-0.4 -1.1 3.0 2.4 -0.8 0.7 -0.4 -0.1 -1.5 0.5 -2.0 1.9 -3.5 0.7 -0.6 -1.0

Z-CC

IE CCMT

3.5 9.8 14.8 8.6

8 7 -15 a -8 a 10 -8 a 3 7 17 -4 12 5 21 -3 -4 8 -1.3 1.1

-0.8 -0.9 -3.5 -2.6 -0.6 -2.6 -1.4 -0.9 0.3 -2.2 -0.3 -1.1 0.7 -2.1 -2.2 -0.8

Z-CC

dFIE IE CFMT IE CCMT

60.5 17.6 39.6 12.4

86 28 60 70 88 78 48 44 66 60 38 54 58 42 68 80

Dev.

-1.7 1.6

-3.2 1.6 -1.7 -2.6 -3.7 -3.4 -1.8 -0.3 -3.0 -1.3 0.4 -0.7 -2.3 -0.1 -1.7 -3.6

Z-CCC

CFPT Up

77.4 10.4 61.7 11.7

82 68 72 84 94 88 72 66 78 66 62 92 90 70 84 70

Dev.

-1.3 0.9

-1.7 -0.5 -0.9 -1.9 -2.8 -2.2 -0.9 -0.4 -1.4 -0.4 0.0 -2.6 -2.4 -0.7 -1.9 -0.7

Z-CC

CFPT Inv

16.9 14 22.1 12.9

-4 40 12 14 6 10 24 22 12 6 24 38 32 28 16 -10 a

Inv-Up

-0.3 1.2

-1.4 2.0 -0.8 -0.6 -0.9 -1.0 -0.8 0.1 -1.4 -0.9 0.4 1.8 0.3 0.6 0.2 -2.7

Z-CCC

IE CFPT

Keywords: Developmental prosopagnosia; face recognition; Face inversion effect; holistic face processing.

Cases’ pattern of performance fulfils the criteria for a dissociation, putatively CLASSICAL, for the difference between the case and controls (Crawford et al., 2010).

a

PP04 57 male PP05 49 male PP07 40 female PP09 40 female PP17 49 female PP18 38 female PP19 16 male PP25 40 female PP27 25 male PP101 48 male PP102 43 female PP105 49 male PP106 25 female PP112 40 female PP116 55 female PP118 37 female Group means and SD: DP mean, N=16: DP SD: Control mean, N=32: Control SD:

Individual DPs:

Age

Table 1 Age, gender and performance score of 16 DPs on upright and inverted conditions of the Cambridge memory test for faces (CFMT) and cars (CCMT), the inversion effect (Up - Inv) for CFMT (IE CFMT) and CCMT (IE CCMT) and the disproportionate face inversion effect, dFIE: CFMT IE – CCMT IE. The Cambridge face perception test with upright (CFPT Up) and inverted (CFPT Inv) conditions and the inversion effect (CFPT IE: Inv – Up). Note that the CFPT is based on an error score and therefore the sign of the inversion effect (CFPT IE) is reversed to ease comparisons with the accuracy measures. Values in bold face designate abnormal performance applying Bayesian test for deficit with single case statistics, SingleBayes_ES (Crawford et al., 2010). Effect sizes (Z-CC) for difference between score of case versus controls are also listed. Age is correlated with the outcome on CFPT upright and IE CFPT in control group, and so BTD_Cov was applied to control for age (Crawford et al., 2011), and the associated effect size (Z-CCC) is listed.