Consciousness and Cognition 30 (2014) 62–72

Contents lists available at ScienceDirect

Consciousness and Cognition

journal homepage: www.elsevier .com/locate /concog

Perceptual hysteresis as a marker of perceptual inflexibilityin schizophrenia

http://dx.doi.org/10.1016/j.concog.2014.07.0141053-8100/� 2014 Elsevier Inc. All rights reserved.

⇑ Corresponding author at: Institut Jean Nicod – UMR 8129, École Normale Supérieure, 29, rue d’Ulm, F 75005 Paris, France.E-mail addresses: jeanremy08@gmail.com, Jean-Remy.Martin@ens.fr (J.-R. Martin).

Jean-Rémy Martin a,b,⇑, Guillaume Dezecache a,b,c, Daniel Pressnitzer d, Philippe Nuss a,e,Jérôme Dokic b,f, Nicolas Bruno e, Elisabeth Pacherie b, Nicolas Franck g,h

a Université Paris VI (UPMC), Paris, Franceb Institut Jean Nicod (IJN), UMR 8129 (CNRS, ENS, EHESS), Paris, Francec Laboratoire de Neurosciences Cognitives (LNC), U960 (INSERM, CNRS, ENS), Paris, Franced Laboratoire des Systèmes Perceptifs (LSP), UMR 8248 (CNRS, ENS), Paris, Francee Department of Psychiatry, St-Antoine Hospital, AP-HP, Paris, Francef École des Hautes Études en Sciences Sociales, Paris, Franceg Service Universitaire de Réhabilitation, Centre de Réhabilitation, Centre Hospitalier Le Vinatier, Lyon, Franceh Centre de Neurosciences Cognitives (CNC), UMR 5229 (CNRS), Bron Cedex, France

a r t i c l e i n f o a b s t r a c t

Article history:Received 10 April 2014

Keywords:SchizophreniaPerceptual inflexibilityBelief inflexibilityPerceptual hysteresisPredictive deficits

People with schizophrenia are known to exhibit difficulties in the updating of their currentbelief states even in the light of disconfirmatory evidence. In the present study we testedthe hypothesis that people with schizophrenia could also manifest perceptual inflexibility,or difficulties in the updating of their current sensory states. The presence of perceptualinflexibility might contribute both to the patients’ altered perception of reality and the for-mation of some delusions as well as to their social cognition deficits. Here, we addressedthis issue with a protocol of auditory hysteresis, a direct measure of sensory persistence,on a population of stabilized antipsychotic-treated schizophrenia patients and a sampleof control subjects. Trials consisted of emotional signals (i.e., screams) and neutral signals(i.e., spectrally-rotated versions of the emotional stimuli) progressively emerging fromwhite noise – Ascending Sequences – or progressively fading away in white noise –Descending Sequences. Results showed that patients presented significantly stronger hys-teresis effects than control subjects, as evidenced by a higher rate of perceptual reports inDescending Sequences. The present study thus provides direct evidence of perceptualinflexibility in schizophrenia.

� 2014 Elsevier Inc. All rights reserved.

1. Introduction

Clinical and empirical evidence show that people with schizophrenia exhibit belief inflexibility. In other words, they havedifficulties in the updating of their current belief states, even in the light of disconfirmatory evidence (Bruno, Sachs, Demily,Franck, & Pacherie, 2012; Speechley, Whitman, & Woodward, 2010; Woodward, Moritz, Cuttler, & Whitman, 2006;Woodward, Moritz, Menon, & Klinge, 2008). For example, in Garety, Hemsley, and Wessely (1991), schizophrenia patientswere presented with two urns, one containing 80% of blue balls and the other 80% of red balls. Patients had to decide, froma sequence of balls taken from one urn, which urn they were being confronted with. Results showed that patients ‘‘jumped to

J.-R. Martin et al. / Consciousness and Cognition 30 (2014) 62–72 63

conclusions’’ and did not easily update their beliefs in view of new data contradicting their initial conclusions (Garety et al.,1991; see also Warman, 2008). In other words, people with schizophrenia show a bias against disconfirmatory evidence (orBADE) once a belief has become sufficiently robust (e.g., Buchy, Woodward, & Liotti, 2007; Fletcher & Frith, 2009;Speechley et al., 2010; Woodward et al., 2006; see, however, Kaliuzhna, Chambon, Franck, Bérangère, & Van der Henst, 2012).

Indirect evidence also suggests that people with schizophrenia could exhibit perceptual inflexibility. By perceptual inflex-ibility, we mean difficulties in the online updating of current sensory states, even in the light of new sensory evidence. Forexample, in the study of Barbalat, Rouault, Bazargani, Shergill, and Blakemore (2012), participants were instructed to look ata target facial expression (e.g., fear or happiness) and were subsequently presented with a sequence of facial expressionscontaining faces with the target expression and faces with other expressions. Results showed that while both patientsand healthy subjects were slower and less accurate at identifying faces that did not match the target expression than facesmatching it, patients were even slower and less accurate at identifying faces that did not match the prior expectations gen-erated by the target expression. So, we hypothesized that, in addition to belief inflexibility, people with schizophrenia couldalso manifest online perceptual inflexibility.

Perceptual inflexibility could contribute to the altered perception of reality typical of schizophrenia as well as to somedelusions. For instance, it could play a role to delusions of reference (American Psychiatric Association, 2000), with patientsshowing abnormal persistence in the sensory interpretation that some person is looking at her (while in fact this person hasbeen looking at someone else for some time). In addition, perceptual inflexibility could also contribute to explaining socialcognition deficits that are present in schizophrenia (Cook, Barbalat, & Blakemore, 2012). People’s emotional expressionschange quite often over time and, consequently, updating processes of one’s current sensory interpretations of others’ emo-tional expressions have to be efficient to permit smooth social interactions (Sacharin, Sander, & Scherer, 2012). A deficit inthese updating processes could lead to altered perceptions of other emotional states in persisting to perceive a kind of emo-tional expression that, in fact, has changed to another kind.

To test the presence of perceptual inflexibility in schizophrenia we used a protocol of perceptual hysteresis (Buckthought,Kim, & Wilson, 2008; Gepshtein & Kubovy, 2005; Hock, Kelso, Schöner, 1993; Hock & Schoner, 2010; Kleinschmidt, Büchel,Hutton, Friston, & Frackowiak, 2002; Large, Aldcroft, & Vilis, 2005; Sacharin et al., 2012; Schwiedrzik et al., 2012). Perceptualhysteresis protocols have consistently shown that the content of one’s perception at time t depends on the recent history ofthe perceptual system. As such, perceptual hysteresis reflects the presence of normal sensory persistence biases in healthypeople. We predicted that if people with schizophrenia have perceptual inflexibility, they should manifest stronger hyster-esis effects in comparison to healthy people.

Perceptual hysteresis is usually investigated using designs comprising ‘‘ascending’’ and ‘‘descending’’ sequences, that is,sequences ordered in terms of a certain physical parameter. There is hysteresis if the results obtained for the same values aredifferent when these values are presented in ascending versus in descending sequences. As an illustration, it has been shownthat when people are presented with a letter progressively emerging from, or fading away in, a high-density random dotsfield, the mean level of contrast needed to report the letter is less important in the fading away trials than in the emergingtrials (Kleinschmidt et al., 2002). This means that people persist in the letter percept in the first case whereas they persist inthe high-density dots field percept in the second case.

In the present study we tested a group of schizophrenia patients and a control group of healthy participants with an audi-tory protocol validated in a previous experiment (Martin, Dezecache, Dokic, & Grèzes, in press). Each Ascending andDescending Sequence consisted of many steps with different signal-to-noise ratios between a target and a mask. The maskwas composed of bursts of white noise of constant intensity. The targets were short emotional or neutral auditory signals.The signal-to-noise ratio was progressively increased in Ascending Sequences or progressively decreased in DescendingSequences. We incorporated neutral but also emotional stimuli because, as mentioned above, perceptual inflexibility mightparticipate to the social cognition deficits present in schizophrenia (e.g., Chambon, Baudouin, & Franck, 2006 and see Cooket al., 2012 for a review). Therefore, we may expect patients to show perceptual inflexibility for emotional stimuli too.

Results showed that patients presented significantly stronger hysteresis effects than control subjects, as evidenced by ahigher rate of perceptual reports in Descending Sequences. The present study thus strongly suggests the presence of percep-tual inflexibility in schizophrenia.

2. Materials and method

2.1. Participants

Table 1 gives a description of patients PANSS scores as well as antipsychotic treatments. The examined psychiatric sampleconsisted in 17 stabilized chronic individuals with no known hearing disturbances (two patients were removed from statis-tical analysis, see Section 2.5). Exclusion criteria were the following: drug or alcohol dependence, any history of neurologicalillness or trauma, older than 60 years, illiteracy, and French as a nonnative language.

Patients were essentially recruited from the outpatient clinic the Service Universitaire de Réhabilitation (Lyon, France)and a few of them from the Saint Antoine Hospital (Paris, France). All studied patients met the diagnostic and statisticalmanual of mental disorders (DSM-IV) criteria for schizophrenia, with no other psychiatric diagnosis on DSM-IV-TR. A seniorpsychiatrist from each institution of recruitment established the psychiatric diagnosis and established the PANSS scores (NF

Table 1Characteristics of patients with schizophrenia and controls.

Patients (n = 17) Controls (n = 16)

Age 34 (7.53) 29 (3.74)PANSS positive 18.23 (6.15) –PANSS negative 18.70 (7.09) –PANSS general 42.47 (8.81) –PANSS total 79.41 (14.94) –Medication Quetiapine –

AripiprazoleAmisulprideRisperidoneOlanzapineClozapine

64 J.-R. Martin et al. / Consciousness and Cognition 30 (2014) 62–72

for the Service Universitaire de Réhabilitation and PN for the Saint Antoine Hospital). All patients but one were receivingantipsychotic medication (unchanged at the time of the study as well as during the month preceding the study) and wereclinically stable at the time of testing.

Control subjects were recruited from the database of the ‘Relais d’Information sur les Sciences de la Cognition’ (RISC, Paris,France). None of the control subjects presented a history of mental disorder. Finally, neither patients nor controls presented ahistory of auditory impairments.

Seventeen schizophrenia patients (2 females; mean age of 34 ± 7.53 SD) and sixteen control subjects (9 females; meanage of 29 ± 3.74 SD) took part in the present study. All participants gave their written informed consent for the study, whichwas approved by the local Ethical Committee (Comité de Protection des Personnes, no. AFSSAPS: 2011-A01581-40).

2.2. Experimental setup

All participants completed the experiment seated in a quiet room. Stimuli were presented through a pair of headphones(HD 250 linear II) and delivered by a MacBook Pro, processor 2,53 GHz, Intel Core i5 equipped with a professional externalsound card (One by APOGEE, A/D and D/A conversion, 44, 1/48 kHz, 24-bit). All stimuli were displayed using Matlab (Math-Works Inc. R2009b) with the Psychophysics toolbox extensions.

2.3. Stimuli

Participants were exposed to neutral stimuli as well as to emotional stimuli (Fig. 1). The emotional stimuli consisted offive fear stimuli (i.e., five different screams produced by two women) compiled by Sauter and colleagues (Sauter & Eimer,2010). For the stimuli to be of the same length we shortened each of them to 600 ms. From pilot studies, it appeared that600 ms was a good trade-off between recognition of the emotion, and the length of trials (in pilot experiments, trials thatwere too long induced inattention in subjects).

Fig. 1. Spectrograms of an original fear stimulus (top) and its spectrally-rotated version (bottom). The neutral stimuli consisted of spectrally-rotatedversions of the original emotional stimuli: low pass filtering at 3.8 kHz followed by spectral rotation around 2 kHz (Blesser, 1972; Green et al., 2013). Thisacoustic manipulation suppresses the emotional character of the original emotional signals but results in stimuli with a similar level of acoustic complexity.

J.-R. Martin et al. / Consciousness and Cognition 30 (2014) 62–72 65

To design the neutral stimuli we employed the technique of spectral rotation which consists in transforming the high-fre-quency components of the original spectrum in low-frequency components and vice versa around a specific frequency(Blesser, 1972). In particular, the neutral stimuli consisted of spectrally-rotated versions of the original emotional stimuli:low pass filtering at 3.8 kHz followed by spectral rotation around 2 kHz (Blesser, 1972; Green, Rosen, Faulkner, &Paterson, 2013; Sauter & Eimer, 2010). This resulted in five neutral stimuli (which sound like meaningless ‘‘robotic sound’’)matched to the fear stimuli, thus enabling fair comparisons between emotional and neutral stimuli. Indeed, from previousinvestigations, it is assumed that spectral rotation preserves some of the physical characteristics of the original stimuli (i.e.,duration, amplitude envelope) but radically alters their global configuration, intelligibility and, importantly, their emotionalsignificance (Sauter & Eimer, 2010; Scott, Blank, Rosen, & Wise, 2000; Warren et al., 2006). Fig. 1 shows the spectrogram of afear stimulus and its spectrally-rotated version.

In a previous study (Martin et al., in press), in order to make sure that neutral stimuli were actually perceived as affec-tively neutral and that emotional stimuli were well recognized as belonging to the fear category we conducted three pre-testexperiments. In the first pre-test experiment, participants (n = 11; age range = 23–33 y.o.) were presented with the five fearand five more amusement stimuli and the spectrally-rotated versions of these 10 stimuli in a random order and were asked,in a forced-choice task, to categorize each stimulus in one of the three following categories: fear, amusement or neutral. Par-ticipants categorized stimuli above chance-level (t10 = 5.724, p < .001 for fear stimuli; t10 = 13.895, p < .001 for spectrally-rotated fear stimuli; t10 = 9.682, p < .001 for amusement stimuli; t10 = 5.073; p < .001, for spectrally-rotated amusementstimuli). In the second pre-test experiment, to ensure that spectrally-rotated stimuli could not convey any fear or amuse-ment tones, another group of participants (n = 10; age range = 22–34 y.o.) was presented with only the spectrally-rotatedstimuli; they were asked to classify each stimulus in one of two categories, i.e., fear or amusement, in a forced-choice task.We found that participants classify stimuli at chance-level (t9 = 1.097, p > .1 for spectrally-rotated fear stimuli; t9 = 0.178,p > .1 for spectrally-rotated amusement stimuli), confirming that spectrally-rotated stimuli are perceived as affectivelyneutral. Finally, we found that, in 10 other participants (age range = 20–50 y.o.), emotional stimuli were perceived as morenegatively (for fear stimuli) and positively (for amusement stimuli) arousing that their spectrally-rotated counterparts (fearstimuli: t9 = �2.311, p < .05; amusement stimuli: t9 = 4.595, p = .001).

2.4. Procedure

To test the hypothesis that schizophrenia patients would have aberrant perceptual persistence phenomena in comparisonto control subjects, we designed a hysteresis protocol using the modified method of limits (Hock & Schoner, 2010). The variouscharacteristics of the different types of trial are indicated in Table 2. The protocol was specifically designed to avoid deci-sional biases or other potential artefacts present in the traditional method of limits (Fechner, 1860). In particular, it avoidsthe perseveration in response bias by randomizing ascending and descending trials as well as the inference production fromtrial duration by making all trials of the same length but with different starting and ending points. Finally, and importantly,in the modified method of limits, contrary to the traditional method, participants respond at the end of each trial and not

Table 2Sequences of SNR (signal-to-noise ratio) levels for the different conditions. Table shows the series of SNRs, measured in dB, from stimulus1 to 14 (stimulus 14 was the test stimulus for the detection judgments) across the different trial types. All Ascending Sequences startedwith a SNR of �30 dB. All Descending Sequences started with a SNR of �16 dB. The grey region in each row indicates the beginning ofincrement (in Ascending Sequences) or decrement (in Descending Sequences) of the signal-to-noise ratio.

Step 1 2 3 4 5 6 7 8 9 10 11 12 13 14

Ascending Sequences

(SNR-values) -30 -30 -30 -30 -30 -30 -30 -30 -30 -30 -30 -30 -30 -30

-30 -30 -30 -30 -30 -30 -30 -30 -30 -30 -30 -29 -28 -27

-30 -30 -30 -30 -30 -30 -30 -30 -30 -30 -29 -28 -27 -26

-30 -30 -30 -30 -30 -30 -30 -30 -30 -29 -28 -27 -26 -25

-30 -28 -27 -26 -25 -24 -23 -22 -21 -20 -19 -18 -17 -16

Descending Sequences

(SNR-values) -16 -16 -16 -16 -16 -16 -16 -16 -16 -16 -16 -16 -16 -16

-16 -16 -16 -16 -16 -17 -18 -19 -20 -21 -22 -23 -24 -25

-16 -16 -16 -16 -17 -18 -19 -20 -21 -22 -23 -24 -25 -26

-16 -16 -16 -17 -18 -19 -20 -21 -22 -23 -24 -25 -26 -27

-16 -18 -19 -20 -21 -22 -23 -24 -25 -26 -27 -28 -29 -30

66 J.-R. Martin et al. / Consciousness and Cognition 30 (2014) 62–72

during the current sequence of ascending or descending steps avoiding specific decision or response-time biases (Hock &Schoner, 2010). In sum, the modified method of limits has been specifically designed to show authentic persistence or hys-teresis effects, that is, persistence effects not attributable to certain artefacts but, instead, indicating the presence of specificprocesses (Hock et al., 1993; Hock & Schoner, 2010).

Patients and control subjects were presented with sequences composed of 14 steps (Inter-stimulus interval(ISI) = 100 ms); a step consisting of a signal (neutral or emotional) embedded in a burst of white noise with a given sig-nal-to-noise ratio.

Three main conditions or types of trial were used. First, in Ascending Sequences, the signal-to-noise ratio (SNR) measuredin dB was progressively increased, by steps of 1 dB. Second, in Descending Sequences, the SNR was progressively decreasedby steps of 1 dB. At the end of the Ascending and Descending Sequences, participants had to make a judgment about thepresence or absence of the signal in noise in the last step. Finally, in the Absolute Condition, a trial did not consist in a longsequence but in an isolated burst of white noise in which, once again, a signal (neutral or emotional) was embedded andparticipants had to report the presence or absence of the signal in noise. The SNR was varied across trials in this condition,in random order. The Absolute Condition was conducted in an independent short session ( around 7 minutes long) before themain hysteresis experiment composed of the Ascending and Descending Sequences. This condition is necessary to furtherspecify the nature of the potential hysteresis effects we might find. As described in the introduction, hysteresis shows thatthe content of one’s perception at time t depends on the recent past stimulations of the perceptual system. Therefore, a dif-ference in detection scores between Ascending and Descending Sequences is not sufficient to show that the influence of paststimulations played in both types of sequence. To this end, detection scores in Ascending and Descending Sequences havealso to be different from detection scores in the Absolute Condition where stimuli are presented in isolation.

In the Ascending and Descending Sequences, the progressive increase or decrease of SNR in a trial began at a specific stepwithin a trial according to the final value of intensity the stimulus should reach. 5 different final SNR-values were used (�30,�27, �26, �25, �16 dB) and repeated 20 times (10 times within Ascending Sequences and 10 times within DescendingSequences) for each kind of stimulus (fear and neutral stimuli). The same values of SNR were tested for the AbsoluteCondition.

Finally, Ascending and Descending Sequences as well as the different SNR-values were randomly presented. The exper-iment lasted about 1 h including three short breaks.

2.5. Statistical analysis

Hysteresis values were first computed by subtracting mean detection scores obtained in Ascending Sequences from meandetection scores obtained in Descending Sequences. To check whether patients and control participants showed a hysteresiseffect with the present protocol, we systematically compared hysteresis values to zero using t-tests. To further specify thenature of the putative hysteresis effect, both sequence types were also compared to the results in the Absolute Condition. Thedata from two patients were removed before analysis. The first patient was later found not to meet criteria for schizophreniaand the second hallucinated during the test.

The design of the present protocol makes possible the identification of at least three levels of inter-group persistence dif-ferences: (i) in Ascending Sequences, patients could have a lower detection rate than control subjects (i.e., persist with thenoise longer) with no difference in Descending Sequences; (ii) in Descending Sequences, i.e., patients could have a higherdetection rate than control subjects (i.e., persist with the signal longer) with no difference in Ascending Sequences; (iii)in both Ascending and Descending Sequences, i.e., patients would have a lower detection rate than control subjects inAscending Sequences and a higher detection rate than control subjects in Descending Sequences.

In order to explore differences in mean detection scores between patients and control participants across conditions, weran an ANOVA using Content (Emotion, Neutral), Direction (Ascending, Descending) and SNR (�30, �27, �26, �25 dB) aswithin-participant factors, and Group (Patients, Control participants) as an inter-participant factor, on mean detectionscores. The SNR-value �16 was excluded from data analysis because most of the participants showed a ceiling effect at thisvalue and so little information is provided with regard to our experimental manipulations. In particular, it may obscure thecomparison between groups.

Bonferroni corrections were also employed to account for multiple testing. Post-hoc comparisons (two-tailed t-tests)were performed for the analysis of simple main effects.

3. Results

3.1. Hysteresis in schizophrenia patients and control participants

The percentage of reported signal were obtained for the various conditions (Fig. 2). Results clearly show the presence of ahysteresis pattern for each group. Differences between detection scores in Descending and Ascending Sequences reveal thatpatients and control participants show persistence patterns (i.e., their mean detection scores were higher in Descending thanin Ascending Sequences). More precisely, t-tests showed that hysteresis was greater than zero in each group for both fear

Fig. 2. Hysteresis patterns for each group across the different SNR-values. The top panel shows the mean percentage of reported signal for controlparticipants for both neutral stimuli (left) and emotional stimuli (right) and for Descending Sequences (red curve), Ascending Sequences (blue curve) andthe Absolute Condition (green curve). The bottom panel of the figure shows the mean percentage of reported signal for schizophrenia participants for bothneutral stimuli (left) and emotional stimuli (right) and for Descending Sequences (red curve), Ascending Sequences (blue curve) and the Absolute Condition(green curve). The space between the red curve and the blue curve in each figure represents hysteresis. NS stands for neutral stimuli and ES stands foremotional stimuli. As can be seen, red curves are always higher for patients than for controls. Error bars indicate mean ± SEM. (For interpretation of thereferences to color in this figure legend, the reader is referred to the web version of this article.)

J.-R. Martin et al. / Consciousness and Cognition 30 (2014) 62–72 67

stimuli and neutral stimuli (Fear: t14 = 7.642, p < .001 and t15 = 5.457, p < .001 for schizophrenia patients and controls respec-tively; neutral: t14 = 6.935, p < .001 and t15 = 4.249, p = .001 for schizophrenia patients and controls respectively).

As a whole, data indicate that schizophrenia patients (as well as control subjects) were able to perform the task. In par-ticular, mean detection scores decreased as the relative signal intensity level decreased in both Ascending and DescendingSequences (see Fig. 2). In addition, we did not find any inter-group differences for the Absolute Condition for neutral stimuli(t28 = 1.873, p > .05) suggesting that patients had no decisional biases in the first place (i.e., they did not show a bias towardthe yes or no response, but see Section 4) relative to control subjects. However, patients’ detection rate was slightly higherfor emotional stimuli (t28 = 2.088, p = .046) coherent with the findings showing an abnormal processing of emotionalinformation in people with schizophrenia (Cook et al., 2012, see Section 4).

Fig. 3. Persistence differences between schizophrenia patients and control subjects. The figure shows the mean percentage of reported signal averaged overall SNR-values for each group (red for patients and blue for controls); for Descending (a) and Ascending (b) Sequences and, finally, for neutral (left) andemotional stimuli (right). Black lines indicate significant effects at *p < .05; **p � .01; ***p � .001. Error bars indicate mean ± SEM. (For interpretation of thereferences to color in this figure legend, the reader is referred to the web version of this article.)

68 J.-R. Martin et al. / Consciousness and Cognition 30 (2014) 62–72

3.2. A persistence effect in Descending but not in Ascending Sequences

In the control population, the comparison of the detection scores obtained in Ascending Sequences and DescendingSequences to the detection scores obtained in the Absolute Condition could identify that descending values were signifi-cantly different from absolute values for both neutral (t15 = �4.671, p < .0001) and emotional stimuli (t15 = �5.174,p < .0001) (Fig. 2). More specifically, detection scores were, on average, higher for Descending Sequences than for the Abso-lute Condition. In contrast, ascending values were not significantly different from absolute values for both neutral(t15 = 0.716, p > .05) and emotional stimuli (t15 = 1.852, p > .05). This indicates that, in the present hysteresis protocol, onlyDescending Sequences actually demonstrated a persistence effect for the studied stimuli.

The same pattern of results was observed for the patient population. Here also, the detection scores were, on average,higher for Descending Sequences than for the Absolute Condition for both neutral (t13 = �5.677, p < .0001) and emotionalstimuli (t13 = �4.497, p = .001). In contrast, ascending values were not significantly different from absolute values for bothneutral (t13 = 0.806, p > .05) and emotional stimuli (t13 = �1.457, p > .05). This indicates that schizophrenia patients didnot have an overall bias toward the yes response.

3.3. Persistence differences across groups and conditions

Fig. 3 shows the mean percentage of reported signal averaged overall all SNR-values in Ascending and DescendingSequences for both schizophrenia patients and control subjects. While the direction of persistence was following the samepattern in patients and controls (see Section 3.2), detection scores differed in the two populations, as revealed by a maineffect of Group (F1,29 = 7,207, p = .012). In particular, detection scores were higher for patients than for control participants.In addition, analyses could identify (i) a main effect of Content (F1,29 = 21.307, p < .0001) with detection scores higher foremotional than for neutral stimuli, (ii) a main effect of Direction (F1,29 = 89.116, p < .0001) with detection scores higherfor Descending than for Ascending Sequences and, finally, (iii) a main effect of SNR (F3,87 = 110.099, p < .0001) was found withdetection increasing as the relative signal intensity level increases (all ps < .001).

We also found interactions between the factors (i) Content and Direction (F1,29 = 6.416, p = .017) with a higher detectionrate in Descending and Ascending Sequences for emotional stimuli than for neutral stimuli (both ps < .05), (ii) Content andSNR (F3,87 = 3.031, p = .034), as well as (iii) between Direction and SNR (F3,87 = 16.305, p < .0001). Finally, the interactionbetween factors Content, Direction, SNR and Group (F3,87 = 2.592, p = .058) was marginally significant.

Our data are thus in favour of the existence of perceptual inflexibility in schizophrenia.

3.4. Interaction between content, direction, SNR and group

We now further explore the interaction between Content, Direction, SNR and Group. Using post-hoc tests, we found that,when considering every SNR-level, patients actually showed higher detection scores than control participants in DescendingSequences for emotional (t29 = 2.472, p = .020) as well as for neutral stimuli (t29 = 3.563, p = .001). In the case of neutral stim-uli, this pattern was observed for each SNR-value (all ps < .05). However, when considering emotional stimuli, the differencebetween patients and control participants was significant for the SNR-values of �30 and �27 (ps < .05) but disappeared forthe higher SNR-values of �26 and �25 (all ps > .1). No inter-group differences were found in Ascending Sequences (ps > .05).

3.5. Extrinsic factors: level of education, age, medication and PANSS scores

We finally looked at possible confounding factors as well as possible correlations between mean detection scores andclinical symptoms. Detection was not significantly correlated with the level of education and age in either control or schizo-phrenia participants (all ps < .05). Finally, detection was not significantly correlated with medication (using the equivalent-chlorpromazine) or with PANSS global or sub-scores (all ps < .05).

4. Discussion

4.1. Summary of findings

The present findings are indicative of the existence of perceptual inflexibility in schizophrenia patients relatively to con-trol subjects, at least in the auditory domain. In the studied protocol, patients showed perceptual inflexibility exclusively inthe part of the protocol that produced a persistence effect for the signal, i.e., in Descending Sequences. Therefore, we coulddemonstrate that in patients the described phenomenon is not a general tendency towards persistence but a specific aber-rant manifestation of a normal process observed in healthy people. Finally, this aberrant persistence effect was as much pres-ent in patients presenting an essentially positive clinical dimension as in patients presenting an essentially negative clinicaldimension.

J.-R. Martin et al. / Consciousness and Cognition 30 (2014) 62–72 69

4.2. Interpreting perceptual inflexibility in schizophrenia

Perceptual hysteresis has been argued to result from the influence of top-down signals or prior expectations on bottom-up sensory processing (Large et al., 2005). In particular, recent accounts of perceptual hysteresis suggest that the primed per-ception of one of the two possible perceptual alternative states, ‘‘adjusts the prior toward this interpretation’’ (Schwiedrziket al., 2012, p. 9; see also Kleinschmidt et al., 2002). In our protocol the two possible perceptual alternatives states are signalor noise (see also, Kleinschmidt et al., 2002). Descending Sequences start with a relatively high signal-to-noise ratio and,therefore, the signal is easily detectable at the beginning. In this context, the subject expects or predicts that a signal willbe present at the following step and as the sequence unfolds, each new step tends to predict the presence of a forthcomingsignal at the following step. As a consequence, the signal is more often selected over the noise state in comparison toAscending Sequences and to the Absolute Condition. Note that predictive/inferential processes are usually considered asoptimization mechanisms (Kording, 2007). Therefore, we can speculate that Ascending Sequences did not show persistenceeffects because there appears to be no utility in maintaining pure noise (this would even be harmful by preventing thesystem to detect relevant signals) compared to the prediction and maintenance of the signal.

Increasing evidence supports the presence of systematic predictive deficits in schizophrenia (Corlett & Fletcher, 2012;Corlett, Krystal, Taylor, & Fletcher, 2009; Corlett, Taylor, Wang, Fletcher, & Krystal, 2010; Corlett et al., 2007; Fletcher &Frith, 2009; Frith & Friston, 2013; Keefe & Kraus, 2009; Kraus, Keefe, & Krishnan, 2009; Krishnan, Kraus, & Keefe, 2011;Martin, 2013; Martin & Pacherie, 2013; Moore & Fletcher, 2011). These predictive deficits could account for perceptualinflexibility in schizophrenia by disturbing the balance between prior expectations and incoming sensory information(Corlett et al., 2009; Fletcher & Frith, 2009). In the present experiment, in Descending Sequences, two competing expecta-tions are likely to be activated. First, there is the expectation that a signal will be present at the next step but, second, thereis also the expectation that the signal will disappear in the noise as the sequence unfolds (i.e., expectation of reducing SNR).When the SNR becomes very low and the signal hypothesis produces prediction errors that are too large, the no-signalhypothesis is selected. In this context, at least two alternative explanations of the presence of perceptual inflexibility inpatients can be considered.

First, it has been shown that, in some conditions, patients put abnormal weight on their prior expectations (Aleman, Koen,Hijman, De Haan, & Kahn, 2003; Cook et al., 2012; Ilankovic et al., 2011). In line with these findings, perceptual inflexibilitymay result from an undue reliance on the prior adjusted toward the signal. In other words, the prior expectation that thenext step should present the participant with a signal aberrantly (in comparison to healthy people) overrides the currentavailable sensory information. As a consequence, patients’ perceptual states are not correctly updated and they have diffi-culties shifting from the signal hypothesis to the no-signal hypothesis.

The second alternative rests on the idea of abnormal prediction error estimation in patients (Adams, Stephan, Brown,Frith & Friston, 2013; Fletcher & Frith, 2009; Frith & Friston, 2013). In particular, it has been hypothesized that patients esti-mate non-pertinent prediction errors with too much precision. Now, when the no-signal hypothesis is selected this producesprediction errors because although the SNR is very low, it is not yet altogether null. In controls, these prediction errors areignored because they are estimated to be too imprecise given the situation (very weak SNR). In contrast, in patients, due toover-estimation biases of prediction errors, they are not ignored. As a result, signals take an aberrant salience (Corlett et al.,2007, 2010; Fletcher & Frith, 2009; McGhie & Chapman, 1961; Murray et al., 2008; Roiser et al., 2009) producing strongerpersistence effects in patients than in controls. Future studies should disentangle these two alternatives.

It is also conceivable that perceptual inflexibility is related to a weakness of the process of forgetting in patients. Anechoing-like condition would thus be created by the weakness of the memory extinguishing processes.

The presence of perceptual inflexibility in schizophrenia patients probably contributes to the observed reality distortionand may provide some basis for the formation of delusional ideas (see Keefe & Kraus, 2009; Kraus et al., 2009; Krishnan et al.,2011). In addition, it would be interesting to know whether perceptual inflexibility in schizophrenia is present in othermodalities than audition.

4.3. Emotional versus neutral persistence

We also found that schizophrenia patients maintained emotional stimuli more than control participants did (but for lessSNR-values, see Section 4.4). Our protocol actually mirrors noisy dynamic ecological situations where the intensity of signalsis progressively decreasing or increasing. In such dynamic and noisy situations it is likely advantageous to maintain persis-tently emotional signals because fear (for instance) may indicate a threat in the immediate environment (Sauter, Eisner,Ekman & Scott, 2010). However, an aberrant expression of this phenomenon would produce systematic false positives cuese.g., the impression one still hears a scream when in fact it is now a laughter bout, or when there is no signal at all anymore.It could be that the undue weight that patients put on prior expectations might be more pronounced when signals have anemotional content. Previous work has indeed shown that the abnormal influence of prior expectations on the processing ofincoming sensory information (specifically, on facial expression discrimination) in schizophrenia is stronger when priorshave a (negative) emotional content rather than a neutral one (Barbalat et al., 2012; Cook et al., 2012). Future studies shouldexplore this possibility in the context of hysteresis.

70 J.-R. Martin et al. / Consciousness and Cognition 30 (2014) 62–72

This abnormal persistence in emotional perception could have detrimental consequences for the social cognitive abilitiesof patients as it could be responsible for a failure to update their perceptual interpretations of others’ emotional states, acrucial capacity to maintain fluent social interactions (Sacharin et al., 2012).

4.4. Limitations of the study

A first concern is the nature of detection differences between patients and controls, whether they result from sensory ordecisional processes. Patients did not show a higher detection rate in Ascending Sequences but only in DescendingSequences – the only condition which produced true hysteresis effects. This suggests that patients did not have a generalbias toward the yes response. However, with the present protocol, we are not able to decide whether the detection differ-ences in Descending Sequences are purely sensory or are also decisional. Future work should further explore this issue bydesigning a protocol enabling d-prime measures. However, even if our results were due to decisional biases this should stillbe very interesting, as they might indicate that the ‘‘higher’’ perceptual decision processes in patients reflect a stronger biasby priors in comparison to healthy subjects. In addition, the results in the Absolute Condition could suggest a potential base-line difference. Even if we did not find significant inter-group differences for the Absolute Condition regarding neutral stim-uli the t-value (t = 1.873) is close to significance. In addition, patients’ detection rate was slightly higher for emotional stimuli(t28 = 2.088, p = .046). Future work should clarify this point, again by designing a protocol enabling d-prime measures, and bymatching patients and controls for basic auditory abilities with basic auditory tests.

A second limitation is the lack of significance for the interaction between Direction and Group. However, the interactionbetween Content, Direction, SNR and Group can potentially explain this lack of significance. In our post-hoc tests of thisinteraction (Section 3.4), we indeed found that patients had higher detection scores than control participants in DescendingSequences for neutral stimuli (t29 = 3.563, p = .001) and this was true for each SNR-value. Nonetheless, with respect to emo-tional stimuli, patients had still an overall higher detection score than control participants (t29 = 2.472, p = .020) but thiseffect held only for extreme SNR-values: �30 and �27. This is because, as we found in a previous study (Martin et al., inpress), healthy people maintain emotional sounds more than neutral sounds and, in the present study, healthy peoplehad indeed very strong persistence effects for emotional stimuli for the values �26 and �25. This most probably preventedthe disclosure of potential differences between patients and controls, explaining the lack of significance for the interactionbetween Direction and Group. We predict that, in future studies, the use of more values, especially of values near threshold(where priors probably have the greatest influence), will make the difference between patients and controls much moresalient.

A third potential limitation of this study is that our groups were not perfectly gender-matched and the gender ratio ofpatients was not perfectly balanced. As a preliminary clue, we ran the same ANOVA as described above (Section 2.5) butwe replaced the inter-participant factor Group (Patients, Controls) by Gender (Women, Men). We did not find any effectof Gender (F1,29 = 1.145, p = .293), suggesting that balanced participant samples for sex would not affect our main result.

Finally, detection scores do not show any correlation with PANSS scores. However, future hysteresis studies shouldexplore this issue in a more direct way with designs integrating more SNR-values and a larger sample of schizophreniapatients with clearly different clinical dimensions. In addition, the present experiment should be replicated with differenttypes of stimuli in order to explore whether perceptual inflexibility in schizophrenia is stimulus-specific or not.

4.5. Concluding remarks

Beyond providing a direct demonstration of perceptual inflexibility in schizophrenia patients, the use of such perceptualhysteresis paradigm could provide useful information to clinicians, giving them a better understanding of the cognitive pro-cesses at work in their patients: first, hysteresis may serve as a further objective measure to evaluate the level of self-mon-itoring disturbance as well as its evolution in the course of the disease. Measuring hysteresis may be a way of assessing towhat extent patients are able or not to distinguish between the actual presence of an (even weak) external signal and a self-generated persistent signal.

Second, in addition to its use as an evaluative instrument, hysteresis may also be used as a training tool by the clinicalstaff, especially in the context of cognitive remediation (e.g., Demily & Franck, 2008; Franck et al., 2013). In particular, hys-teresis could be used to help people with schizophrenia gain insight and recover, at least partially, self-monitoring abilities.To this end, the protocol at stake should be designed in such a way that the caregiver would know what the standard thresh-old for each tested value is. This would enable caregivers to measure more precisely to what extent patients are far from thestandard threshold, as well as to train patients at distinguishing the point where the signal is really present from the pointwhere the signal is aberrantly and internally maintained.

Acknowledgments

We would like to thank the clinical team of the Service Universitaire de Réhabilitation, 69006 Lyon, France as well as theCH le Vinatier (Bron, France). The research was supported by an ANR-11-0001-02 PSL*, an ANR-10-LABX-0087, a ED3C/UPMCscholarship (J.R.M.).

J.-R. Martin et al. / Consciousness and Cognition 30 (2014) 62–72 71

References

Adams, R. A., Stephan, K. E., Brown, H. R., Frith, C. D., & Friston, K. J. (2013). The computational anatomy of psychosis. Front Psychiatry, 4(47). http://dx.doi.org/10.3389/fpsyt.2013.00047.

Aleman, A., Koen, B. B. E., Hijman, R., De Haan, E. H. F., & Kahn, R. S. (2003). Cognitive basis of hallucinations in schizophrenia: Role of top-down informationprocessing. Schizophrenia Research, 64, 175–185.

American Psychiatric Association (2000). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: American Psychiatric Association.Barbalat, G., Rouault, M., Bazargani, N., Shergill, S., & Blakemore, S. J. (2012). The influence of prior expectations on facial expression discrimination in

schizophrenia. Psychological Medicine, 1–11.Blesser, B. (1972). Speech perception under conditions of spectral transformation: I. Phonetic characteristics. Journal of Speech, Language and Hearing

Research, 15(1), 5.Bruno, N., Sachs, N., Demily, C., Franck, N., & Pacherie, E. (2012). Delusions and metacognition in patients with schizophrenia. Cognitive Neuropsychiatry,

17(1), 1–18.Buchy, L., Woodward, T. S., & Liotti, M. (2007). A cognitive bias against disconfirmatory evidence (BADE) is associated with schizotypal traits. Schizophrenia

Research, 90, 334–337.Buckthought, A., Kim, J., & Wilson, H. R. (2008). Hysteresis effects in stereopsis and binocular rivalry. Vision Research, 48, 819–830.Chambon, V., Baudouin, J.-Y., & Franck, N. (2006). The role of configural information in facial emotion recognition in schizophrenia. Neuropsychologia, 44,

2437–2444.Cook, J., Barbalat, G., & Blakemore, S. J. (2012). Top-down modulation of the perception of other people in schizophrenia and autism. Frontiers in Human

Neuroscience, 6(175), 1–10.Corlett, P. R., & Fletcher, P. C. (2012). The neurobiology of schizotypy: Fronto-striatal prediction error signal correlates with delusion-like beliefs in healthy

people. Neuropsychologia, 50, 3612–3620.Corlett, P. R., Krystal, J. H., Taylor, J. R., & Fletcher, P. C. (2009). Why do delusions persist? Frontiers in Human Neuroscience, 3(12), 1–9.Corlett, P. R., Murray, G. K., Honey, G. D., Aitken, M. R. F., Shanks, D. R., Robbins, T. W., et al (2007). Disrupted prediction-error signal in psychosis: Evidence

for an associative account of delusions. Brain, 130(9), 2387–2400.Corlett, P. R., Taylor, J. R., Wang, X.-J., Fletcher, P. C., & Krystal, J. H. (2010). Toward a neurobiology of delusions. Progress in Neurobiology, 92, 345–369.Demily, C., & Franck, N. (2008). Cognitive remediation: A promising tool for the treatment of schizophrenia. Expert Review in Neurotherapeutics, 8,

1029–1036.Fechner, G. T. (1860). Elemente der psychophysik (Elements of psychophysics) (H. E. Adler, Trans.). New York, USA: Holt, Rinehart and Winston.Fletcher, P. C., & Frith, C. D. (2009). Perceiving is believing: A Bayesian approach to explaining the positive symptoms of schizophrenia. Nature Reviews.

Neuroscience, 10(1), 48–58.Franck, N., Duboc, C., Sundby, C., Amado, I., Wykes, T., Demily, C., et al (2013). Specific vs general cognitive remediation for executive functioning in

schizophrenia: A multicenter randomized trial. Schizophrenia Research, 147(1), 68–74.Frith, C., & Friston, K. J. (2013). False perceptions & false beliefs: Understanding schizophrenia. Neurosciences and the Human Person: New Perspectives on

Human Activities, 121, 1–15.Garety, P. A., Hemsley, D. R., & Wessely, S. (1991). Reasoning in deluded schizophrenic and paranoid patients. Biases in performance on a probabilistic

inference task. The Journal of Nervous and Mental Disease, 179, 194–201.Gepshtein, S., & Kubovy, M. (2005). Stability and change in perception: Spatial organization in temporal context. Experimental Brain Research, 160, 487–495.Green, T., Rosen, S., Faulkner, A., & Paterson, R. (2013). Adaptation to spectrally rotated speech top of form. Journal of the Acoustical Society of America,

134(1369). http://dx.doi.org/10.1121/1.4812759.Hock, H. S., Kelso, J. S., & Schöner, G. (1993). Bistability and hysteresis in the organization of apparent motion patterns. Journal of Experimental Psychology:

Human Perception and Performance, 19, 63–80.Hock, H. S., & Schoner, G. (2010). Measuring perceptual hysteresis with the modified method of limits: Dynamics at the threshold. Seeing and Perceiving, 23,

173–195.Ilankovic, L. M., Allen, P. P., Engel, R., Kambeitz, J., Riedel, M., Müller, N., et al (2011). Attentional modulation of external speech attribution in patients with

hallucinations and delusions. Neuropsychologia, 49, 805–812.Kaliuzhna, M., Chambon, V., Franck, N., Bérangère, T., & Van der Henst, J.-B. (2012). Belief revision and delusions: How do patients with schizophrenia take

advice. PloS One, 7(4), 1–10.Keefe, R. S. E., & Kraus, M. S. (2009). Measuring memory-prediction errors and their consequences in youth at risk for schizophrenia. Annals of the Academy of

Medicine, Singapore, 38(5), 414–416.Kleinschmidt, A., Büchel, C., Hutton, C., Friston, K. J., & Frackowiak, R. S. (2002). The neural structures expressing perceptual hysteresis in visual letter

recognition. Neuron, 34, 659–666.Kording, K. (2007). Decision theory: What ‘‘should’’ the nervous system do? Science, 318, 606–610.Kraus, M. S., Keefe, R. S. E., & Krishnan, R. K. R. (2009). Memory-prediction errors and their consequences in schizophrenia. Neuropsychology Review, 19(3),

336–352.Krishnan, R. K. R., Kraus, M. S., & Keefe, R. S. E. (2011). Comprehensive model of how reality distortion and symptoms occur in schizophrenia: Could

impairment in learning-dependent predictive perception account for the manifestations of schizophrenia. Psychiatry and Clinical Neurosciences, 65,305–317.

Large, M. E., Aldcroft, A., & Vilis, T. (2005). Perceptual continuity and the emergence of perceptual persistence in the ventral visual pathway. Journal ofNeurophysiology, 93, 3453–3462.

Martin, J.-R. (2013). Experiences of activity and causality in schizophrenia: When predictive deficits lead to a retrospective over-binding. Consciousness andCognition, 22(4), 1361–1374.

Martin, J. -R., Dezecache, G., Dokic, J., & Grèzes, J. (in press). Prioritization of emotional signals by the human auditory system: Evidence from a perceptualhysteresis protocol. Evolution and Human Behavior.

Martin, J.-R., & Pacherie, E. (2013). Out of nowhere: Thought insertion, ownership and context-integration. Consciousness and Cognition, 22, 111–122.McGhie, A., & Chapman, J. (1961). Disorders of attention and perception in early schizophrenia. British Journal of Psychiatry, 34, 103–116.Moore, J. W., & Fletcher, P. C. (2011). Sense of agency in health and disease: A review of cue integration approaches. Consciousness and Cognition, 21(1),

59–68.Murray, G. K., Corlett, P. R., Clark, L., Pessiglione, M., Blackwell, A. D., Honey, G., et al (2008). Substantia nigra/ventral tegmental reward prediction error

disruption in psychosis. Molecular Psychiatry, 13(3), 239, 267–276.Roiser, J. P., Stephan, K. E., den Ouden, H. E., Barnes, T. R., Friston, K. J., & Joyce, E. M. (2009). Do patients with schizophrenia exhibit aberrant salience?

Psychological Medicine, 39, 199–209.Sacharin, V., Sander, D., & Scherer, K. R. (2012). The perception of changing emotion expressions. Cognition and Emotion, 26(7), 1273–1300.Sauter, D. A., & Eimer, M. (2010). Rapid detection of emotion from human vocalizations. Journal of Cognitive Neuroscience, 22, 474–481.Sauter, D. A., Eisner, F., Ekman, P., & Scott, S. K. (2010). Cross-cultural recognition of basic emotions through nonverbal emotional vocalizations. Proceedings

of the National Academy of Sciences. http://dx.doi.org/10.1073/pnas.0908239106.Schwiedrzik, C. M., Ruff, C. C., Lazar, A., Leitner, F. C., Singer, W., & Melloni, L. (2012). Untangling perceptual memory: Hysteresis and adaptation map into

separate cortical networks. Cerebral Cortex. http://dx.doi.org/10.1093/cercor/bhs396.Scott, S. K., Blank, C. C., Rosen, S., & Wise, R. J. S. (2000). Identification of a pathway for intelligible speech in the left temporal lobe. Brain, 123, 2400–2406.

72 J.-R. Martin et al. / Consciousness and Cognition 30 (2014) 62–72

Speechley, W. J., Whitman, J. C., & Woodward, T. S. (2010). The contribution of hypersalience to the ‘‘jumping to conclusions’’ bias associated with delusionsin schizophrenia. Journal of Psychiatry and Neuroscience, 35, 7–17. /fgt’8790.

Warman, D. M. (2008). Reasoning and delusion proneness – Confidence in decisions. Journal of Nervous and Mental Disease, 196, 9–15.Warren, J. E., Sauter, D. A., Eisner, F., Wiland, J., Dresner, M. A., Wise, R. J. S., et al (2006). Positive emotions preferentially engage an auditory-motor « mirror »

system. The Journal of Neuroscience, 26, 13067–13075.Woodward, T. S., Moritz, S., Cuttler, C., & Whitman, J. (2006). The contribution of a cognitive bias against disconfirmatory evidence (BADE) to delusions in

schizophrenia. Journal of Clinical and Experimental Neuropsychology, 28, 605–617. http://dx.doi.org/10.1080/13803390590949511.Woodward, T. S., Moritz, S., Menon, M., & Klinge, R. (2008). Belief inflexibility in schizophrenia. Cognitive Neuropsychiatry, 13, 267–277.