YNIMG-11390; No. of pages: 9; 4C: 6, 8 NeuroImage xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

NeuroImage journal homepage: www.elsevier.com/locate/ynimg

a

7

a r t i c l e

8 9 10

Article history: Accepted 15 May 2014 Available online xxxx

11 12 13 14 15 16 17

Keywords: perceptual decision making post-decision confidence metacognition fMRI grating orientation task anterior prefrontal cortex

Hans Berger Department of Neurology, Jena University Hospital, Friedrich Schiller University, 07743 Jena, Germany Department of Pediatrics, HELIOS Children's Hospital Wuppertal, Witten/Herdecke University, 42283 Wuppertal, Germany c Department of Biological and Clinical Psychology, Friedrich Schiller University, 07743 Jena, Germany d Center for Sepsis Control and Care, Jena University Hospital, Friedrich Schiller University, 07743 Jena, Germany

i n f o

R O

b

O

3 4 5 6

F

Raphael Hilgenstock a,b,⁎, Thomas Weiss c, Otto W. Witte a,d

a b s t r a c t

P

Current findings suggest that confidence emerges only after decision making. However, the temporal and neural dynamics of the emergence of post-decision confidence – a metacognitive judgement – are not fully explored. To gain insight into the dynamics of post-decision confidence processing and to disentangle the processes underlying confidence judgements and decision making, we applied a tactile discrimination task during functional magnetic resonance imaging (fMRI). Our results revealed that reaction times to post-decision confidence depend on the level of confidence, suggesting that post-decision confidence in a perceptual choice is not processed in parallel to perceptual decision making. Moreover, we demonstrated by the parametric analysis of fMRI data that postdecisionally modelled confidence processing can be distinguished from processes related to decision making through anatomical location and through the pattern of neural activity. In contrast to perceptual decision making, post-decision confidence appears to be strictly allocated to a prefrontal network of brain regions, primarily the anterior and dorsolateral prefrontal cortex, areas that have been related to metacognition. Moreover, the processes underlying decision making and post-decision confidence may share recruitment of the dorsolateral prefrontal cortex, although the former probably has distinct functions with regard to processing of perceptual choices and post-decision confidence. Thus, this is the first fMRI study to disentangle the processes underlying post-decision confidence and decision making on behavioural, neuroanatomical, and neurofunctional levels. With regard to the temporal evolution of post-decision confidence, results of the present study provide strong support for the most recent theoretical models of human perceptual decision making, and thus provide implications for investigating confidence in perceptual paradigms. © 2014 Published by Elsevier Inc.

D

Q22Q1

You’d Better Think Twice: Post-Decision Perceptual Confidence

E

C

T

E

1

R

36

45 46 47 48 49 50

Much theoretical work has been carried out in the effort to understand human decision making. Choice, decision time, and confidence have emerged from these efforts as key concepts of understanding and of modelling the processes underlying decision making (Kepecs and Mainen, 2012; Pleskac and Busemeyer, 2010). Yet the temporal and neural dynamics of the processes underlying the emergence of confidence and a metacognitive judgement about confidence are not fully characterized. Recent findings indicate that the formation of confidence emerges only after primary (e.g., perceptual) decision making. For

N C O

43 44

Introduction

U

42

R

40 38 37 39 41

18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35

Abbreviations: APFC, anterior prefrontal cortex; BOLD, blood oxygen level dependent; CRT, confidence reaction time; DLPFC, dorsolateral prefrontal cortex; FCT, functional connectivity toolbox; fMRI, functional magnetic resonance imaging; GOT, gratings orientation task; HRT, hemodynamic response function; MNI, Montreal Neurological Institute; PSC, percent signal change; PRT, perceptual reaction time; ROI, region of interest; rWLS, robust weighted least-square regression; SMG, superior medial gyrus; SPM, statistical parametric map; 2DSD, two-stage dynamic signal detection. ⁎ Corresponding author at: Department of Pediatrics, HELIOS Children's Hospital Wuppertal, Witten/Herdecke University, 42283 Wuppertal, Germany. E-mail address: [email protected] (R. Hilgenstock).

example, while reaction times (RTs) related to perceptual decisions (PRTs) vary linearly with regard to confidence ratings, subsequent reaction times on the level of confidence (CRT) in the preceding perceptual decision vary non-linearly as a function of confidence ratings, which has been interpreted to indicate an ongoing accumulation of evidence (Petrusic and Baranski, 2003). Resulaj et al. (2009) investigated changes of mind after primary decision making in a motor paradigm; this study also indicated ongoing information processing that could be related to the emergence of confidence (Van Zandt and Maldonado-Molina, 2004). Based on these and similar findings, the two-stage dynamic signal detection (2DSD) model, a comprehensive model of human decision making, has been conceptualised (Pleskac and Busemeyer, 2010). The model postulates that perceptual choice (e.g., categorizing noisy images into distinct classes or assessing the orientation of a tactile grating) and confidence differ in terms of the amount of accumulated evidence. Accordingly, confidence and a metacognitive judgement about confidence only come into existence post-decisionally by the ongoing accumulation of information. This assumption, as implemented in the 2DSD model, offers the advantage of accounting for all three key aspects of human decision making, including confidence judgements or, within the framework of the model, post-decision confidence.

http://dx.doi.org/10.1016/j.neuroimage.2014.05.049 1053-8119/© 2014 Published by Elsevier Inc.

Please cite this article as: Hilgenstock, R., et al., You’d Better Think Twice: Post-Decision Perceptual Confidence, NeuroImage (2014), http:// dx.doi.org/10.1016/j.neuroimage.2014.05.049

51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71

113

Altogether, 26 subjects were recruited to participate in the study. Two subjects were excluded—one because of excessive motion during scanning, and the other because of technical problems with the device used for psychophysical testing during the scanning procedure. Consequently, the study had 24 participants (18 females; ages 18–27 years), all healthy, with no history of neurological or psychiatric disorders, trauma, or brain abnormalities. All subjects were right-handed, as

97 Q3 98 99 100 101 102 103 104 105 106 107 108

114 115 116 117 118 119

C

95 96

E

93 94

R

91 92

R

89 90

O

87 88

C

85 86

N

83 84

The Grating Orientation Task (GOT; van Boven and Johnson, 1994), which requires subjects to indicate the orientation of tactile gratings, was used to assess tactile acuity (Fig. 1). We modified the original GOT task. Our subjects were required to indicate the orientation of a stimulus pair according to four response alternatives (lengthwise– lengthwise, lengthwise–crosswise, crosswise–lengthwise and crosswise–crosswise; see Fig. 2). Additionally, subjects were requested to indicate their level of confidence in their preceding perceptual decision, using indicators analogous to Petrusic and Baranski (2003): "guess", "slightly certain", "moderately certain", and "certain". For the subsequent fMRI testing, we selected two stimuli, individual to each subject, one grating for which the orientation was easily identified ("easy pen," P(correct) = 75–100%), and a second grating for which the orientation was more difficult to identify ("difficult pen", P(correct) 30–60%). Gratings were chosen to individually maximize the distribution of confidence ratings over the four levels of choice confidence.

125 126

MRI: Pretesting

142

129 130 131 132 133 134 135 136 137 138 139 140 141

MRI: Main experiment

160

The main experiment consisted of 108 tactile task trials (Fig. 2). In total, participants were presented with 48 pairs of easy tactile gratings and 60 pairs of difficult tactile gratings. The study design used a larger number of difficult trials in response to the phenomenon

161 162

1

…

2

Trial 2

127 128

143

3

Trial 1

122 123

Two fMRI practice sessions were carried out to refine the choice of gratings for the main fMRI experiment and to acquaint subjects with the fMRI setting and design. Each trial was announced by three transverse white bars that were presented for 0.5 s. Subsequently, GOT-pens were presented twice, for two seconds each time, and with a two-second pause between pens. Subjects rated the orientation of a stimulus pair according to four response alternatives: lengthwise–lengthwise, lengthwise–crosswise, crosswise–lengthwise and crosswise–crosswise. Immediately afterwards subjects indicated their level of confidence associated with this preceding decision on stimulus-pair orientation (guess, slightly certain, moderately certain, and certain) within four seconds. A fixation cross was shown between trials (see Fig. 2). Altogether, 30 trials were presented in each practice session. A pneumatically driven, MRI-compatible stimulator was used for stimulus presentation. The sequence and duration of stimulus presentation were controlled using the software Presentation (Neurobehavioral Systems, Albany, CA).

U

81 82

124

F

Subjects

79 80

Psychophysical testing

O

112

78

R O

Method

76 77

120 121

P

111

74 75

assessed by the high-validity subset of the Edinburgh handedness inventory (Raczkowski et al., 1974). All subjects gave informed written consent after explanation of the experimental procedure. The study was approved by the local ethics committee.

T

109 110

The difference between post-decision confidence and perceptual choice may not only be quantitative – with regard to the amount of accumulated sensory evidence over time – but may also be qualitative. Kiani and Shadlen (2009) suggest that perceptual choices may simply result from the raw data provided by secondary cortices, especially the parietal cortex, involved in the accumulation of sensory evidence. Thus, at the time of perceptual choice only a preliminary conceptualisation of confidence may emerge that is not available for a metacognitive report (Fleming et al., 2012; Kepecs and Mainen, 2012; Middlebrooks and Sommer, 2012). However, these raw data may subsequently translate into higher order cognitive processing in prefrontal areas where metacognitive judgements about confidence emerge (Bode et al., 2012; Kepecs and Mainen, 2012; Kiani and Shadlen, 2009). Accordingly Fleming et al. (2010, 2012) and Yokoyama et al. (2010), for example, provide evidence to indicate that metacognitive ability, the accuracy of confidence, is located primarily in the anterior prefrontal cortex (APFC). Thus, perceptual decision making and the formation of postdecision confidence likely differ not only with regard to their temporal dynamics, but also differ qualitatively through their recruitment of higher order cognitive processes in a distinct network of brain areas. These differences may be linked to behavioral discrepancies between perceptual choice and post-decision confidence, as is indicated by the phenomena of over- and underconfidence. Overconfidence and underconfidence refer to discrepancies between objective task performance at the time of decision making and expected performance based on post-decision confidence ratings (Baranski and Petrusic, 1995; Juslin, Winman, and Olsson, 2000; Pleskac and Busemeyer, 2010). Despite these findings indicating that perceptual choice and postdecision confidence may differ on temporal, behavioral, and neural scales, to the best of the authors’ knowledge no study has rigorously attempted, using fMRI, to delineate the processes underlying perceptual decision making, confidence associated with the decision making process, and the emergence of post-decision confidence. Therefore, the present study was primarily intended (1) to provide insight into the temporal dynamics of post-decision confidence, (2) to achieve a better understanding of the underlying functional somatosensory network of post-decision confidence, and thus (3) to investigate differences between perceptual choice and confidence embedded in the assumption of a post-decisional emergence of confidence.

D

72 73

R. Hilgenstock et al. / NeuroImage xxx (2014) xxx–xxx

E

2

Trial 3

Trial 4

Fig. 1. The device used for psychophysiological testing and an exemplary sequence of gratings. Eight gratings, each with a surface of different ridge and groove widths (0.25; 0.5; 0.7; 1.0; 1.2; 1.5; 2.0; 3.0), were mounted on a rotatable disk inside the custom-made device depicted at the far left part of Fig. 1. GOT-gratings were presented by releasing a lever on side of the device (1) and changed by another control shifter on its front (2, indicated by white lines). The index finger was immobilized by hook-and-loop tape (3) to eliminate the potential for introduction of confounds. As an example, a GOT-grating with a large resolution is shown (dark grey corresponds to grooves, light gray to ridges). Testing was conducted in a stepwise block procedure, beginning with the pen having the largest "resolution" of 3 mm. Within each block, 20 stimulus pairs were presented. Each of the four possible combinations of pairs was presented five times within each block, in pseudo-randomized sequences that changed every block. The stepwise testing procedure was carried out twice.

Please cite this article as: Hilgenstock, R., et al., You’d Better Think Twice: Post-Decision Perceptual Confidence, NeuroImage (2014), http:// dx.doi.org/10.1016/j.neuroimage.2014.05.049

144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159

163 164

3

O

F

R. Hilgenstock et al. / NeuroImage xxx (2014) xxx–xxx

180 181

190

Images were acquired on a 3.0-T MR whole body scanner (TRIO, Siemens, Erlangen, Germany) using a standard head coil. Approximately 1300 echo-planar T2*-weighted images (EPI) were recorded for each subject. Each functional volume consisted of 44 contiguous, ascending scanned axial slices of 3 mm thickness (voxel size 3 mm × 3 mm × 3 mm). Parallel imaging (iPAT) with an acceleration factor of two was used to record images, with a time to repeat (TR) of 2.53 s, a time to echo (TE) of 30 ms, and a field of view (FOV) of 192 mm. High-resolution anatomical images were acquired using a threedimensional MPRAGE consisting of 192 slices with a spatial resolution of 1 mm × 1 mm × 1 mm (TR 2300 ms, TE 3.03 ms and FOV 256 mm).

191

Data preprocessing and analysis

192

Data quality was checked using ArtRepair (http://cibsr.stanford.edu/ tools/human-brain-project/artrepair-software.htmland) and tsdiffana (http://imaging.mrc-cbu.cam.ac.uk/imaging/DataDiagnostics). Bad slices were repaired by the ArtRepair implementation of an interpolationalgorithm that employs an adaptive threshold for each slice. For a discussion of the applied data quality check methods, please refer to Mazaika et al. (2009). Data analysis was performed using MATLAB (Mathworks, Natick, MA, USA) and SPM8 (Wellcome Department of Cognitive Neurology, Institute of Neurology, London, UK; http://www.fil.ion.ucl.ac.uk/spm/). The unwarp function as part of SPM8 was used to align all subject images to the first volume by a fourth-degree B-spline interpolation. The highresolution anatomical image of each subject was coregistered to its mean, realigned EPI-volume by a fourth-degree B-spline interpolation.

184 185 186 187 188 189

193 194 195 196 197 198 199 200 201 202 203 204 205

C

E

R

182 183

R

175 176

N C O

173 174

U

171 172

206 207

Data analysis: confidence reaction times

215

A repeated measure mixed-effect model was specified to test for the dependence of confidence reaction times (CRTs) (dependent variable) on task difficulty (independent variable (IV)1), confidence rating (IV2), and the interaction of task difficulty and confidence rating. Task difficulty and confidence rating were treated as fixed factors, while the subject factor was considered to be random. To account for the covariance structure of the repeated measures, a diagonal covariance term was chosen to model dependencies of both factors. An identity matrix was used to describe the covariance pattern of the random subject factor. A full fixed-effect model was set up for main effects and interactions, as well as for a random intercept model.

216

Coding the MRI paradigm

227

A general linear model (GLM) approach was used to obtain statistical maps. Twelve regressors were specified by convolving SPM’s canonical hemodynamic response function (HRF) with either a stick or a boxcar input for each condition that was modelled. Boxcar inputs were used to model higher order cognitive functions such as the decision on the orientation of stimulus pairs. Stick functions were used to model all other events. Modelling each trial with its duration as a boxcar input, known as a variable epoch approach (Grinband et al., 2008), has been shown to generate a more physiologically plausible model of neural activity that accounts for the fact that brain regions involved in the processing of higher cognitive functions show sustained activity between stimulus presentation and subsequent response (e.g., Kiani and Shadlen, 2009, Ratcliff et al., 2007; Shadlen and Newsome, 2001). Moreover, the variable epoch approach is suitable for increasing the statistical power in detecting activations (and thus consistency) and interpretability of results (Grinband et al., 2006; Grinband et al., 2008). Four regressors coded the onset of the first stimulus of each stimulus pair, representing a 2 × 2 matrix of task difficulty (easy vs. difficult) and correctness of answer (correct vs. incorrect). Thus, the first regressor

228 229

P

fMRI recordings

169 170

Acquisition differences between slices were accounted for by slice time correction (with reference to the 22nd slice). The "new segment" option that has been implemented in SPM8 was used for segmentation of the structural image from the first MRI measurement (Ashburner and Friston, 2005). Deformation fields were derived and subsequently used to normalize each subject’s data to the Montreal Neurological Institute (MNI) standard space. Data were smoothed using a 9 mm, isotropic, three-dimensional full-width-at-half-maximum Gaussian kernel.

D

179

167 168

T

177 178

of underconfidence—corresponding to the observation that in easy choice sets people tend to display disproportionate levels of confidence (Baranski and Petrussic, 1994). In order to account for motor activation during task trials, a second trial type required subjects to press only one of two numbers ("3" or "4") with either an index or middle finger, respectively (e.g., Zhang et al., 2005). Overall, subjects were presented with 22 of these trials. Another 40 null trials, during which a fixation cross was shown for 13 s, were intermixed with task trials. These null trials were intended to prevent the formation of expectance based on trial length, as each trial’s length had to be adapted to the number of rotations that were necessary for the disk to move to the next position for the presentation of a new stimulus pair (14 s at longest). Moreover, null trials were intended to establish baseline activity between trials. Therefore, the overall fMRI paradigm for each subject took about 55 min to run.

E

165 166

R O

Fig. 2. Schematic representation of the experimental fMRI design. Trials were announced by three transverse white bars presented for 0.5 s. Subsequently, GOT-pens were presented twice, for 2 s each time, with a 2-s pause between. Subjects rated the orientation (lw = lengthwise; cw = crosswise) of a stimulus pair, and immediately thereafter indicated their confidence associated with the preceding decision (guess, slightly = slightly certain, moderate = moderately certain, certain) within 4 s. A fixation cross was shown between trials. The long interstimulus intervals were due to technical reasons. Text and symbols are enlarged here for improved illustration.

Please cite this article as: Hilgenstock, R., et al., You’d Better Think Twice: Post-Decision Perceptual Confidence, NeuroImage (2014), http:// dx.doi.org/10.1016/j.neuroimage.2014.05.049

208 209 210 Q4 211 212 213 214

217 218 219 220 221 222 223 224 225 226

230 231 232 233 234 235 236 237 238 239 240 Q5 241 242 243 244 245 246

262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283

Model estimation

285

292 293

The β-weights for each regressor outlined above were estimated using a robust weighted least-square regression (rWLS) approach that has been shown to successfully account for even severe noise in fMRI time series, independent of the source (Diedrichsen and Shadmehr, 2005). A restricted maximum likelihood approach is employed to estimate the variance of noise in each image of the time series specified. These estimates are then used to weight each observation by the inverse of its noise. High-pass filtering, as part of the standard SPM model specification at 1/128 Hz, was performed to remove slowly varying trends.

294

Statistical inference and mapping of activation

295 296

Statistical parametric maps (SPMs) were thresholded using the height as well as the spatial extent of activation. Thus, an SPM was initially thresholded at p ≤ .001 on the voxel level, and was subsequently thresholded at p ≤ .05 on a cluster level, FWE-corrected for multiple comparisons (Poline et al., 1997; Smith et al., 2010). Activation was localized using the SPM Anatomy toolbox (Eickhoff et al., 2005; http://www2.fz-juelich.de/inm/index.php?index=194) and the WFU_PickAtlas (Maldjian et al., 2003; http://fmri.wfubmc. edu/software/PickAtlas). Because the assignment of activation to cortical areas by the SPM Anatomy toolbox and the WFU_PickAtlas is only probabilistic, assignments were visually verified. Both MRIcron (http://www. mccauslandcenter.sc.edu/mricro/mricron/index.html) and Caret visualization software (Van Essen et al., 2001; http://brainvis.wustl.edu/wiki/

297 298 299 300 301 302 303 304 305 306 307

R

O

C

N

290 291

U

288 289

R

284

286 287

Post-decision confidence

318

F

260 261

311 312

The regressor modulated by the indicated level of confidence that was thus representing activation specifically related to the formation of post-decision confidence (see also 2.7) was tested as the regressor coding GOT-related activation as outlined above.

O

258 259

We tested for the suitability of the paradigm in order to show activations for the GOT that have been reported by other studies (van Boven et al., 2005; Zhang et al., 2005). The regressor coding the first stimulus presentation of correctly identified easy stimuli, as well as correctly identified difficult stimuli, was tested against the regressor that coded trials requiring only a motor response. The resulting first-level contrast was taken to the second level by a one-sample t-test.

R O

256 257

310

313 314 315 316 317

319 320 321 322

Controlling for the effects of reaction time, task difficulty and task 323 performance 324 Varying reaction times and task difficulty (easy/difficult) have been shown to correlate with confidence reports (e.g., Grinband et al., 2006; Petrusic and Baranski, 2003). Thus, additional analyses were conducted to remove their confounding effects on the relationship between postdecision confidence and the magnitude of the BOLD response at the time confidence was evaluated (second decision). The following analysis steps were conducted. First, we (re-)estimated the first-level analysis for those subjects showing a significant correlation between confidence reaction times (CRTs) and post-decision confidence reports. To do so, we used a subset of trials sampled from a time window of 400 ms of the initial and individual CRT distribution (e.g., 1400–1800 ms) in accordance with Grinband et al. (2008). Thus, we accounted for the confounding effects of varying reaction times on the relation between post-decision confidence reports and the BOLD response. Second, in order to account for confounding effects of task difficulty, we conducted a region of interest (ROI) analysis based on this timerestricted data set. MarsBaR (MARSeille Boîte À Région d’Intérêt; http://marsbar.sourceforge.net/) was used to define spherical ROIs of 7 mm diameter located at the first maximum of all clusters within the anterior prefrontal cortex (APFC), the left lateral and dorsolateral prefrontal cortex (DLPFC), and the right and left superior medial gyrus (SMG), as were identified by the parametric analysis of the unrestricted data. The mean time course of each ROI and subject of the CRT-restricted first-level analysis was extracted. Data were high-pass filtered and (pre-) whitened using the implementation of an autoregressive model as part of the Functional Connectivity Toolbox (FCT; Zhou, Thompson and Siegle, 2009). Head movement parameters were not included as covariates because unwarping was used for the preprocessing of the functional data. To account for confounding effects of task difficulty, these data were subjected to a mediation analysis—which is used to identify an indirect effect that partially or fully explains the effect of an independent variable (postdecision confidence) on a dependent variable (BOLD-signal change) by a third variable (task difficulty). Mediation was assessed by computing the indirect effect of post-decision confidence on brain activity through task difficulty that is the product of the unstandardized regression coefficients of the regressions R(Difficulty|Confidence) and R(Brain Activity|Difficulty). The indirect effect and its significance were assessed by confidence intervals generated by nonparametric re-sampling (bootstrapping). Mediation analysis was performed by custom-written MATLAB code. 1000 bootstrap samples were performed to generate confidence intervals for the indirect effect of post-decision confidence on brain activity in the APFC, DLPFC, and

P

254 255

GOT: fMRI map

D

253

index.php/Main_Page) were used as a means of visualizing the results. 308 Statistical maps were projected on an averaged group template. 309

T

251 252

C

249 250

was used to model the first-stimulus onset of a combination of an easy trial that was answered correctly. The second was used to model an easy–incorrect combination, the third was used to model a difficult– correct combination, and a fourth regressor was used to model the onset of a difficult–incorrect combination regarding each stimulus pair. The onset of the decision for the stimulus pair orientation was aligned to the presentation of the second stimulus. The duration of each boxcar input was modelled with each trial's duration between the onset of the second stimulus and the decision on the stimulus pair orientation. Regressors were specified according to the four regressors representing the first stimulus presentation, i.e., task difficulty (easy vs. difficult) and correctness of answer (correct vs. incorrect). A crucial assumption of the 2DSD model is the evaluation and formation of confidence after a decision has been made—for example, a decision made on certain stimulus characteristics (Pleskac and Busemeyer, 2010). Consequently, an additional regressor for the formation of confidence was modelled with a boxcar input function of the duration between the perceptual decision on stimulus-pair orientation and the confidence rating, independent of task difficulty or correctness of the behavioral choice. To identify brain regions that encode post-decision confidence in a linear or quadratic way, this regressor was parametrically modulated by the confidence rating of each trial. In order to minimize the risk of misattribution of shared variance among the unmodulated and modulated regressors, regressors were orthogonalized. Orthogonalization was performed as implemented in SPM8 by a Gram–Schmidt approach—that is, the parametric modulated regressors were orthogonalized with respect to the regressor that coded the second decision unmodulated by post-decision confidence. Therefore, the modulated regressors represented only those brain regions that were responsive to a linear or quadratic dependency on the BOLD signal for post-decision confidence. This resulted in two additional regressors. Note that the modulated regressor coding a quadratic trend was orthogonalized with respect to the parametric modulator coding a linear trend. An additional regressor was used to represent those trials in which subjects made a solely motor response, with no tactile stimulation.

E

247 248

R. Hilgenstock et al. / NeuroImage xxx (2014) xxx–xxx

E

4

Please cite this article as: Hilgenstock, R., et al., You’d Better Think Twice: Post-Decision Perceptual Confidence, NeuroImage (2014), http:// dx.doi.org/10.1016/j.neuroimage.2014.05.049

325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367

Results

415

Confidence reaction times

416 417

We investigated confidence reaction times (CRTs) by using a version of the grating orientation task (GOT; see Fig. 1) (Van Boven and Johnson, 1994) that required subjects to indicate the orientation of a pair of tactile gratings, as well as to articulate their level of confidence in the perceptual choices (Fig. 2 gives an overview of the fMRI paradigm). Two sets of GOT gratings with varying difficulty (easy/difficult) were individually chosen during pretesting. A repeated mixed-effect model was specified to test for the dependency of CRTs on task difficulty and on post-decision confidence reports. There was a significant main effect of post-decision confidence on CRT (F(3, 43.26) = 27.82, p ≤ .001; Fig. 3). CRTs varied systematically as a function of post-decision confidence category, and increased with less confidence. However, there was no main effect of task difficulty (F(1, 27.32) = 2.361, p = .136),

396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411

418 419 420 421 422 423 424 425 426 427 428

C

394 395

E

387 388

R

385 386

R

383 384

N C O

381 382

U

379 380

easy difficult

O

414

377 378

R O

412 413

To verify the proposed difference in the temporal evolution of perceptual decision making and post-decisional emerging confidence, we also investigated the BOLD response in those brain regions parametrically dependent on post-decision confidence at the time of perceptual decision making (and not at the time of the metacognitive rating described in Section 2.11). In order to do so, we extracted the PSC of these regions at the time of perceptual decision making, and used MATLAB to test for either a linear or quadratic dependence of activity on post-decision confidence ratings. In addition, we employed RFXplot to extract the time course of activation in the APFC and SMG separately for low confidence trials (trials rated “1 = guess” or “2 = slightly certain”) and high confidence trials (trials rated “3 = moderately certain” or “4 = certain”) to investigate differences in the temporal evolution of the BOLD signal within different brain regions, as suggested by the literature investigating perceptual decision making (e.g. Grinband et al., 2006; Hernández et al., 2010; Kiani and Shadlen, 2009; Lemus et al., 2007). Therefore, we placed a sphere with a diameter of 7 mm at the first maximum in the APFC and SMG, centered at the individual peak of activation, and extracted the event-related BOLD time course from one second prior to the trial until 21.5 s after the trial began.

375 376

Fig. 3. Reaction times on post-decision certainty. Reaction times (in seconds) are reported in dependence on decisional confidence and task difficulty (easy/difficult). Certainty ratings correspond to 1 = “guess”, 2 = “slightly certain”, 3 = “moderately certain”, 4 = “certain”. Error bars correspond to standard error of mean (SEM).

P

393

374

D

Testing the temporal evolution of decision making and post-decision confidence

372 373

nor was there a significant interaction (F(3, 43.32) = 0.496, p = .687) 429 between task difficulty and post-decision confidence. 430

E

391 392

370 371

T

389 390

SMG. This procedure was carried out for each subject. An α-level of 30% was chosen to increase the chance of refusing H0 (no mediation) and instead accepting H1 (mediation by task difficulty). Thus, we also accounted for multiple testing. Eventually, percent signal change (PSC) from the restricted second-level data set was extracted using RFXplot (Glascher, 2009). Data were removed for those subjects showing a mediation effect for task difficulty for one or more of the preselected ROIs. MATLAB was used to test the resulting data set for a linear or quadratic relationship between PSC and post-decision confidence. Third, results from the preceding steps of analysis with regard to the confounding effect of task difficulty were verified by conducting an analysis contrasting the regressors for the presentation of the first tactile stimulus of easy trials against the presentation of the first tactile stimulus of difficult trials at the time of perceptual decision making. The analysis was performed as described in Sections 2.9 and 2.10. Similarly, we intended to exclude a confounding effect of task performance (correct or incorrect answer) at the time of perceptual decision making by contrasting the regressors for the presentation of the first tactile stimulus of correct trials against the presentation of the first tactile stimulus of incorrect trials. Data were analyzed according to the procedure described in Sections 2.9 and 2.10.

368 369

5

F

R. Hilgenstock et al. / NeuroImage xxx (2014) xxx–xxx

GOT manipulation check

431

We proved that our task and the method of stimulation were suitable for eliciting brain responses that are commonly reported on the GOT paradigm. A one-sample t-test was employed to test for areas responsive to passive tactile stimulation independent of task difficulty (in contrast to a single-button press without concomitant tactile stimulation). We found bilateral activation in primary (BA 1, BA 2, BA 3) and secondary somatosensory systems (opercular structures) (Eickhoff et al., 2006, 2010), in motor and supplementary motor structures (BA 4, BA 6), and in inferior parietal (BA 39, BA 40) and superior parietal (BA 5) areas (see Fig. 4). These activations are largely in line with previous studies employing the GOT (van Boven et al., 2005; Zhang et al., 2005).

432

Post-decision confidence

444

A crucial assumption of the 2DSD model is the evaluation and formation of confidence after the initial perceptual decision (Pleskac and Busemeyer, 2010). Accordingly, a regressor was modelled with a boxcar input function of the duration between the response to the perceptual decision on the stimulus pair's orientation and the confidence report. To identify brain regions encoding decisional confidence in a linear or quadratic fashion, this regressor was parametrically modulated by the confidence report of each trial. Testing for this modulated regressor against trials that only required a motor response by a one-sample t-test revealed significant activation that decreased linearly with increasing post-decision confidence (see Fig. 5 and Table 1). Activation dependent on post-decision confidence was found within the superior medial gyrus bilateral (SMG, BA 8), the anterior prefrontal cortex (APFC, BA 10), and the dorsolateral prefrontal cortex (DLPFC, BA 9, BA 46). No region was found to show a positive or quadratic dependency on decisional confidence.

445

Please cite this article as: Hilgenstock, R., et al., You’d Better Think Twice: Post-Decision Perceptual Confidence, NeuroImage (2014), http:// dx.doi.org/10.1016/j.neuroimage.2014.05.049

433 434 435 436 437 438 439 440 441 442 443

446 447 448 449 450 451 452 453 454 455 456 457 458 459 460

R. Hilgenstock et al. / NeuroImage xxx (2014) xxx–xxx

Reaction times and task difficulty

461

Additional analyses were carried out to exclude a confounding effect of confidence reaction times (CRT) on the relation between confidence ratings and the magnitude of activation (e.g., Grinband et al., 2006, 2008). Thus, we de-correlated the relation between confidence reports and CRT by choosing an individualized reaction time window from the initial distribution of CRTs from which trials for subsequent fMRI analysis were sampled. A window length of 400 ms proved suitable for breaking the correlation between confidence ratings and CRT (r corrected = − .086). An additional mediation analysis was conducted to exclude confounding effects of task difficulty. Altogether, data from six subjects showed a significant indirect effect of confidence ratings on brain activity through task difficulty at the time when confidence was assessed in at least one brain region (see Table 2). Accordingly, data from these subjects were excluded from subsequent analysis. When accounting for effects of CRT and task difficulty, however, the pattern of fMRI activation reported for the modulated regressor (see above discussion) did not change. There was a significant linear relationship between activation of prefrontal brain regions and the degree of post-decision confidence (APFC: r = − 0.267, p = .047; DLPFC (BA 9): r = − .388, p = .004; DLPFC (BA 46): r = − .275, p = .035; SMG left: r = − .296, p = .023; SMG right: r = − .424, p = .001) (see Fig. 5). This analysis was substantiated by testing for brain regions in which activation was dependent on task difficulty. We found only one area within the right prefrontal gyrus, close to BA 9 (uncorrected for multiple comparisons, p b .001), that was responsive to differences in task difficulty that did not overlap with activation found when testing for the modulated regressor. Moreover, at the time of perceptual decision making there was no region, even uncorrected for multiple comparisons

462 463

U

N

C

O

R

R

E

C

T

E

Fig. 4. fMRI results for grating orientation task. Random effect group analysis (n = 24) of activations and deactivations elicited by passive tactile stimulation of the right index finger in contrast to a button press. The map is FWE-corrected for multiple comparisons on a cluster-level at p ≤ .05. Data are projected on the inflated standard MRI-mapping brain provided by CARET. Yellow–red tones encode activations.

P

14.3

D

0

R O

O

F

6

Fig. 5. fMRI results for post-decision confidence. Random effect group analysis (n = 24) of confidence-related areas. Those regions within a prefrontal network (yellow = anterior prefrontal cortex (APFC, BA 10); bright red = lateral prefrontal cortex (LPFC, BA 9); dark red = dorsolateral prefrontal cortex (DLPFC, BA 46); light blue = right superior frontal gyrus (SFG, BA 8); dark blue = left SFG, BA 8) showing an increase of the BOLD response with increasing uncertainty (1 = guess, 4 = certain) are projected on a rendered averaged group template (FWE-corrected for multiple comparison, p ≤ .05). The percent signal change (PSC) for the subset of trials accounting for the confounding effect of reaction time as well as task difficulty during the emergence of post-decision confidence is shown for each region in the respective color of the region (error bars correspond to SEM). In addition, the PSC for each region during perceptual decision making is depicted in white. Note: Only in parts of the superior frontal gyrus do both activity patterns overlap and follow a (significant) linear trend.

Please cite this article as: Hilgenstock, R., et al., You’d Better Think Twice: Post-Decision Perceptual Confidence, NeuroImage (2014), http:// dx.doi.org/10.1016/j.neuroimage.2014.05.049

464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491

R. Hilgenstock et al. / NeuroImage xxx (2014) xxx–xxx t1:1 t1:2

7

Table 1 Regions involved in decisional confidence processing.

t1:3

Cluster

Label

Laterality

Area

Peak voxel MNI coordinates

t1:4 t1:5 t1:6 t1:7 t1:8 t1:9 t1:10

1

MFG MFG MFG SMG SMG SFG SFG

R R R L R R R

BA 9 BA 46

38 44 38 −3 6 20 21

2 3

BA 8 BA 10

24 29 21 32 35 57 50

36 33 48 43 43 30 37

Peak T-statistic

p-value

Cluster Size

7.72 5.22 4.52 7.20 5.71 5.82 4.62

b0.001

901

b0.001

1480

0.012

341

Analysis is FWE-corrected for multiple comparisons on a cluster-level at p ≤ .05. The order of the table follows the magnitude of the T-statistics, starting with the cluster with the largest T-statistic. Labels are provided by the SPM Anatomy toolbox and the WFU_PickAtlas: BA = Brodmann Area, MFG = medial frontal gyrus, SMG = superior medial gyrus, SFG = superior frontal gyrus. Images are resliced to 1.5 × 1.5 × 1.5 mm.

492 493 494 495

at p b 0.001, showing significant activation when comparing correct and incorrect trials. This analysis was intended to rule out confounding effects of task performance that should have resulted in brain activity at least similar to the findings presented in section 3.3.

496

The temporal evolution of perceptual choice and post-decision confidence

497

507 508

Importantly, there was no dependence on confidence ratings in either the APFC or in the DLPFC at the time of perceptual decision making. However, activity in the bilateral SMG was linearly dependent on confidence ratings, and increased with less confidence after the initial decision regarding the stimulus pair's orientation (1st decision) was made (Fig. 5). As can be seen from Fig. 6, there is a difference in the latency as well as in the peak latency at which activation in the APFC and SMG begins to discriminate between low and high confidence with activation in the SMG occurring earlier than in the APFC. Thus, both brain regions differ with regard to the temporal evolution of the processing of post-decision confidence in terms of discriminating between low and high confidence trials.

509

Discussion

510

513

The present study investigated the network underlying the formation of post-decisionally modelled somatosensory confidence, as well as differences from perceptual decision making, such as perceptual choice.

514

Perceptual choice versus post-decision confidence

515

524

Results of the present study provide three lines of evidence – on temporal, neuroanatomical, and neurofunctional levels – in support of the idea that different processes underlie post-decision confidence and perceptual choice. First, our study indicates that reaction times on postdecision confidence are not dependent upon the preceding perceptual choice, but vary in degree of dependence on the indicated level of postdecision confidence. If the evaluation of post-decision confidence had been finished at the time of perceptual choice, then the reaction times on post-decision confidence ratings should not vary as a function of the level of indicated post-decision confidence—that is, confidence reaction

t2:1 t2:2

Table 2 Results of the mediation analysis.

511 512

516 517 518 519 520 521 522 523

D

P

R O

O

times should remain constant across the different levels of confidence. In line with Petrusic and Baranski (2003), we conclude that there is a clear indication for post-decisional processing of confidence. Thus, perceptual choice and post-decision confidence differ with regard to their temporal evolution. Second, the notion that perceptual choice and post-decision confidence may also differ on a neural level is supported by the parametrical analysis of fMRI. The superior frontal gyrus (SMG, BA 8), the dorsolateral prefrontal cortex (DLPFC, BA 9/46), and the anterior prefrontal cortex (APFC, BA 10) exhibited a linear relation between BOLD response and post-decisionally modelled confidence. Importantly, at the time of perceptual decision making (i.e., prior to the emergence of confidence), neither the APFC nor the DLPFC showed activity indicting (linear or quadratic) dependence on the level of post-decision confidence; that is, activity in these regions was constant over all levels of post-decision confidence. However, there was a linear dependence of the BOLD response on post-decision confidence in the bilateral SMG at the time of perceptual choice that is in line with Grinband et al. (2006), who conclude that BA 8 is implicated in the detection of lack of confidence (uncertainty) (also see Daniel et al., 2011). To summarize, post-decision confidencedependent activity does not emerge in the APFC and DLPFC until a decision on confidence is required, which provides two important implications. First, this finding supports the idea that perceptual choice and post-decision confidence differ with regard to their temporal evolution, as has been indicated by the analysis of confidence reaction times. Second, the processing of perceptual choice is unrelated to the processing of post-decision confidence in the APFC and DLPFC (but not the bilateral BA 8), and thus perceptual choice and post-decision confidence differ with regard to their primary site of processing. Moreover, the finding of differences in the temporal evolution of the BOLD time course in the SMG and APFC implies that the processing of post-decision confidence may come about in a successive fashion, starting as early as with making a perceptual decision as shown for the SMG. The latter implication is supported by studies investigating perceptual decision making. Some of the brain regions associated with perceptual decision making, particularly the MFG and the DLPFC (Heekeren et al., 2008), overlap with those areas we found to be dependent on decisional confidence. However, perceptual decision making does not seem to depend categorically on the APFC (Heekeren et al., 2008).

E

T

C

E

505 506

R

503 504

R

502

N C O

500 501

U

498 499

F

t1:11 t1:12 t1:13

t2:3

Vpn

BA 10

BA 9

BA 46

BA 8 right

BA 8 left

t2:4 t2:5 t2:6 t2:7 t2:8 t2:9

5 7 10 11 13 21

3.90 [1.19; 6.75] n.s. −3.11 [−6.07; −1.77] −6,80 [−11.09; −3.96] −13.36 [−19.62; −9,54] 1.58 [0.16; 3.16]

4.05 [1.39; 6.80] n.s. −2.68 [−5.23; −1.44] −4.63 [−7.53; −2.36] −12.02 [−17.09; −8.82] 1.17 [0.19; 2.31]

4.70 [1.86; 8.47] 0.78 [0.15; 1.65] −2.70 [−5.09; −1.46] −7.01 [−11.20; −3.30] −12.02 [−17.47; −8.47] 1.21 [0.36; 2.67]

5.09 [1.89; 9.16] n.s. −2.81 [−5.32; −1.28] −7.25 [−12.07; −4.23] −12.36 [−17.67; −9.02] n.s.

4.91 [2.00; 8.92] 0.70 [0.10; 1.58] −2.83 [−5,43; 1.45] −7.21 [−11.38; −4.05] −13.36 [−18.41; −9.51] n.s.

t2:10 t2:11

Cells display significant indirect effects of confidence on brain activity through task difficulty in those brain regions parametrically dependent on post-decision confidence in the data set. Numbers in brackets correspond to the lower and upper bound of the indirect effects confidence intervals at α = 30%. N.s. = not significant.

Please cite this article as: Hilgenstock, R., et al., You’d Better Think Twice: Post-Decision Perceptual Confidence, NeuroImage (2014), http:// dx.doi.org/10.1016/j.neuroimage.2014.05.049

525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 Q6 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563

R. Hilgenstock et al. / NeuroImage xxx (2014) xxx–xxx

not only distinguishes post-decision confidence from perceptual decision making, but also links post-decision confidence to metacognition. In this study, the second region to be exclusively dependent on post-decision confidence, the DLPFC, has previously been related to the processing of perceptual choices, such that easy perceptual choices are associated with a higher BOLD response than are difficult perceptual choices (Heekeren et al., 2004; Pleger et al., 2003). However, during post-decision confidence processing we observed an opposite pattern of activation in the DLPFC. With increasing uncertainty, there was a post-decisional increase of the BOLD response, which has been previously reported by Fleck et al. (2006), who employed a similar paradigm. Perceptual choice and post-decision confidence, consequently, may share recruitment of the DLPFC, but the DLPFC likely has distinct functions related to perceptual choice and post-decision confidence processing. With regard to post-decision confidence processing, the DLFPC has been implicated in the processing of awareness, as well as metacognition (Crick and Koch, 2003; Dehaene et al., 2003; Lau and Passingham, 2006). For example, a study conducted by Rounis et al. (2010) reveals that manipulation of the DLPFC by TMS compromises subjective reports of awareness and metacognition, which implies that the DLPFC is involved in the processing of post-decision confidence as a result of the close interrelationship of post-decision confidence and metacognition (Fleming et al., 2010; Pleskac and Busemeyer, 2010).

Low confidence High confidence

Conclusions Low confidence

565 566 567 568 569 570 571 572 573 574 575

Thus, the recruitment of the APFC for post-decision confidence processing may distinguish the formation of post-decision confidence from the formation of perceptual choices. In fact, the APFC may have a prominent function for the formation of post-decision confidence by integrating input from other, especially secondary, areas (Petrides and Pandya, 2007; Ramnani and Owen, 2004). This idea has recently been supported by volumetric and functional studies addressing the significance of the APFC for interindividual differences in metacognitive abilities (Chua et al., 2006; Fleming et al., 2010, 2012; Yokoyama et al., 2010). Metacognitive ability is simply the accuracy of post-decision confidence (Fleming et al., 2010; Pleskac and Busemeyer, 2010). Along this line of reasoning, the recruitment of the APFC for post-decision confidence

580 581 582 Q7 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599

Conflict of Interest

632

E 564

578 579

600

T C E R R O

U

N

C

Fig. 6. fMRI time course of activation. Time course of the effect size (a.u. = arbitrary unit) of activation in the APFC (yellow/orange) and SMG (blue/light blue, bottom panel) separately plotted for low (trials rated “1 = guess” or “2 = slightly certain”) and high (trials rated “3 = moderately certain” or “4 = certain”) confidence trials. First decision = decision on the stimulus pair orientation; second decision = decision on confidence associated with the first decision (see also Fig. 2 for sequence of a single trial). Mean response time for the first and second decision is chosen to mark the respective point in time. Error bars correspond to standard error of mean (SEM).

576 577

Using behavioral and imaging data, results of the present study provide support for decision making models, primarily the 2DSD model, that postulate differences between perceptual choice and post-decision confidence with regard to their temporal evolution (Pleskac and Busemeyer, 2010). However, perceptual choice and post-decision confidence differ not only on a temporal scale. The processing of perceptual choice and post-decision confidence appears to dissociate along the dimension of time in distinct networks of processing. Evidence indicates that perceptual choice and concomitant confidence emerge along a processing stream in a network of secondary sensory cortices and prefrontal regions that are implicated in the evaluation and monitoring of task difficulty and sensory certainty (e.g., Grinband et al., 2006; Heekeren et al., 2008; Kiani and Shadlen, 2009). Thus, perceptual choice primarily seems to represent task difficulty in a graded way (Fleming et al., 2012). Post-decision confidence, in contrast, appears to be strictly allocated to a prefrontal network that has been related to conscious processing and metacognition (e.g., Crick and Koch, 2003; Del Cul et al., 2009; Fleming et al., 2010, 2012; Lau and Passingham, 2006; Rounis et al., 2010). Therefore, post-decision confidence indicates a subjective representation of task performance amenable to consciousness for a metacognitive report (Nelson, 1996). Both networks may evidence partial anatomical overlap but still differ with regard to the functional processing of both entities, for example, in the DLPFC—as shown both in previous studies (Fleck et al., 2006) and in the present study. This is the first study to disentangle perceptual decision making and the emergence of post-decision confidence along several important dimensions – specifically on behavioral, neuroanatomical, and neurofunctional dimensions – that strongly support a view of postdecision confidence and perceptual choice as distinct entities. Moreover, this observation indicates that perceptual decision making as a decisional report alone does not provide the specificity to reveal the level of post-decision confidence.

D

High confidence

P

R O

O

F

8

The authors declare no competing financial interests.

601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631

633

Uncited references

634 Q8

Ashburner, 2007

635

Please cite this article as: Hilgenstock, R., et al., You’d Better Think Twice: Post-Decision Perceptual Confidence, NeuroImage (2014), http:// dx.doi.org/10.1016/j.neuroimage.2014.05.049

R. Hilgenstock et al. / NeuroImage xxx (2014) xxx–xxx

Ashburner, J., 2007. A fast diffeomorphic image registration algorithm. Neuroimage 38, 95–113. Baranski, J.V., Petrusic, W.M., 1994. The calibration resolution of confidence in perceptual judgments. Percept. Psychophys. 55, 412–428. Bates, J.F., Goldman-Rakic, P.S., 1993. Prefrontal connections of medial motor areas in the rhesus monkey. J. Comp. Neurol. 336, 211–228. Benjamini, Y., Krieger, A.M., Yekutieli, D., 2006. Adaptive linear step-up procedures that control the false discovery rate. Biometrika 93, 491–507. Bode, S., Bogler, C., Soon, C.S., Haynes, J.D., 2012. The neural encoding of guesses in the human brain. Neuroimage 59, 1924–1931. Brett, M., Anton, J.L., Valabregue, R., Poline, J.-B., 2002. Region of interest analysis using an SPM toolbox. 8th International Conference on Functional Mapping of the Human Brain. Chua, E.F., Schacter, D.L., Rand-Giovannetti, E., Sperling, R.A., 2006. Understanding metamemory: Neural correlates of the cognitive process and subjective level of confidence in recognition memory. Neuroimage 29, 1150–1160. Cools, R., Clark, L., Owen, A.M., Robbins, T.W., 2002. Defining the neural mechanisms of probabilistic reversal learning using event-related functional magnetic resonance imaging. J. Neurosci. 22, 4563–4567. Crick, F., Koch, C., 2003. A framework for consciousness. Nat. Neurosci. 6, 119–126. Daniel, R., Pollmann, S., 2010. Comparing the neural basis of monetary reward cognitive feedback during information-integration category learning. J. Neurosci. 30, 47–55. Dehaene, S., Sergent, C., Changeux, J.P., 2003. A neuronal network model linking subjective reports and objective physiological data during conscious perception. Proc. Natl. Acad. Sci. U. S. A. 100, 8520–8525. Del Cul, A., Dehaene, S., Reyes, P., Bravo, E., Slachevsky, A., 2009. Causal role of prefrontal cortex in the threshold for access to consciousness. Brain 132, 2531–2540. Diedrichsen, J., Shadmehr, R., 2005. Detecting and adjusting for artifacts in fMRI time series data. Neuroimage 27, 624–634. Eickhoff, S.B., Stephan, K.E., Mohlberg, H., Grefkes, C., Fink, G.R., Amunts, K., 2005. A new SPM toolbox for combining probabilistic cytoarchitectonic maps and functional imaging data. Neuroimage 25, 1325–1335. Eickhoff, S.B., Schleicher, A., Zilles, K., Amunts, K., 2006. The human parietal operculum. Cytoarchitectonic mapping of subdivisions. Cereb. Cortex 16, 254–267. Eickhoff, S.B., Jbabdi, S., Caspers, S., Laird, A.R., Fox, P.T., Zilles, K., Behrens, T.E., 2010. Anatomical functional connectivity of cytoarchitectonic areas within the human parietal operculum. J. Neurosci. 30, 6409–6421. Fleck, M.S., Daselaar, S.M., Dobbins, I.G., Cabeza, R., 2006. Role of prefrontal and anterior cingulate regions in decision-making processes shared by memory and nonmemory tasks. Cereb. Cortex 16, 1623–1630. Fleming, S.M., Weil, R.S., Nagy, Z., Dolan, R.J., Rees, G., 2010. Relating introspective accuracy to individual differences in brain structure. Science 329, 1541–1543. Fleming, S.M., Huijgen, J., Dolan, R.J., 2012. Prefrontal contributions to metacognition in perceptual decision making. J. Neurosci. 32, 6117–6125. Glascher, J., 2009. Visualization of group inference data in functional neuroimaging. Neuroinformatics 7, 73–82. Grinband, J., Hirsch, J., Ferrera, V.P., 2006. A neural representation of categorization uncertainty in the human brain. Neuron 49, 757–763. Grinband, J., Wager, T.D., Lindquist, M., Ferrera, V.P., Hirsch, J., 2008. Detection of timevarying signals in event-related fMRI designs. Neuroimage 43, 509–520. Heekeren, H.R., Marrett, S., Bettini, P.A., Ungerleider, L.G., 2004. A general mechanism for perceptual decision-making in the human brain. Nature 431, 859–862. Heekeren, H.R., Marrett, S., Ungerleider, L.G., 2008. The neural systems that mediate human perceptual decision making. Nat. Rev. Neurosci. 9, 467–479.

C

E

R

R

N C O

U

766

F

644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697

O

References

T

643

R O

642

P

640 641

Hernández, A., Nácher, V., Luna, R., Zainos, A., Lemus, L., Alvarez, M., Vázquez, Y., Camarillo, L., Romo, R., 2010. Decoding a perceptual decision process across cortex. Neuron 66, 300–314. Juslin, P., Winman, A., Olsson, H., 2000. Naive empiricism and dogmatism in confidence research: A critical examination of the hard-easy effect. Psychol. Rev. 107, 384–396. Kepecs, A., Mainen, Z., 2012. A computational framework for the study of confidence in humans and animals. Philos. Trans. R. Soc. B 367, 1322–1337. Kiani, R., Shadlen, M.N., 2009. Representation of confidence associated with a decision by neurons in the parietal cortex. Science 324, 759–764. Lau, H.C., Passingham, R.E., 2006. Relative blindsight in normal observers and the neural correlates of visual consciousness. Proc. Natl. Acad. Sci. U. S. A. 103, 18763–18768. Lemus, L., Hernández, A., Luna, R., Zainos, A., Nácher, V., Romo, R., 2007. Neural correlates of a postponed decision report. Proc. Natl. Acad. Sci. U. S. A. 104, 17174–17179. Maldjian, J.A., Laurienti, P.J., Kraft, R.A., Burdette, J.H., 2003. An automated method for neuroanatomic and cytoarchitectonic atlas-based interrogation of fMRI data sets. Neuroimage 19, 1233–1239. Mazaika, P., Hoeft, F., Glover, H., Reiss, A.L., 2009. Methods and software for fMRI analysis of clinical subjects. Human brain mapping conference. Middlebrooks, P.G., Sommer, M.A., 2012. Neural correlates of metacognition in primate frontal cortex. Neuron 75, 517–530. Nelson, T.O., 1996. Consciousness and metacognition. Am. Psychol. 51, 102–116. Petrides, M., Pya, D.N., 2007. Efferent association pathways from the rostral prefrontal cortex in the macaque monkey. J. Neurosci. 27, 11573–11586. Petrusic, W.M., Baranski, J.V., 2003. Judging confidence influences decision processing in comparative judgments. Psychon. Bull. Rev. 10, 177–183. Pleger, B., Foerster, A.F., Ragert, P., Dinse, H.R., Schwenkreis, P., Malin, J.P., et al., 2003. Functional imaging of perceptual learning in human primary and secondary somatosensory cortex. Neuron 40, 643–653. Pleskac, T.J., Busemeyer, J.R., 2010. Two-stage dynamic signal detection: A theory of choice, decision time, and confidence. Psychol. Rev. 117, 864–901. Poline, J.B., Worsley, K.J., Evans, A.C., Friston, K.J., 1997. Combining spatial extent and peak intensity to test for activations in functional imaging. Neuroimage 5, 83–96. Preacher, K.J., Hayes, A.F., 2008. Asymptotic resampling strategies for assessing and comparing indirect effects in multiple mediator models. Behav. Res. Methods 40, 879–891. Raczkowski, D., Kalat, J.W., Nebes, R., 1974. Reliability and validity of some handedness questionnaire items. Neuropsychologia 12, 43–47. Ramnani, N., Owen, A.M., 2004. Anterior prefrontal cortex: Insights into function from anatomy and neuroimaging. Nat. Rev. Neurosci. 5, 184–194. Resulaj, A., Kiani, R., Wolpert, D.M., Shadlen, M.N., 2009. Changes of mind in decisionmaking. Nature 461, 263–266. Ridderinkhof, K.R., Ullsperger, M., Crone, E.A., Nieuwenhuis, S., 2004. The role of the medial frontal cortex in cognitive control. Science 306, 443–447. Rounis, E., Maniscalco, B., Rothwell, J.C., Passingham, R.E., Lau, H., 2010. Theta-burst transcranial magnetic stimulation to the prefrontal cortex impairs metacognitive visual awareness. Cogn. Neurosci. 1, 165–175. Shadlen, M.N., Newsome, W.T., 2001. Neural basis of a perceptual decision in the parietal cortex (area LIP) of the rhesus monkey. J. Neurophysiol. 86, 1916–1936. Smith, J.F., Alexer, G.E., Chen, K., Husain, F.T., Kim, J., Pajor, N., et al., 2010. Imaging systems level consolidation of novel associate memories: A longitudinal neuroimaging study. Neuroimage 50, 826–836. Van Boven, R.W., Johnson, K.O., 1994. A psychophysical study of the mechanisms of sensory recovery following nerve injury in humans. Brain 117, 149–167. Van Boven, R.W., Ingeholm, J.E., Beauchamp, M.S., Bikle, P.C., Ungerleider, L.G., 2005. Tactile form and location processing in the human brain. Proc. Natl. Acad. Sci. U. S. A. 102, 12601–12605. Van Essen, D.C., Drury, H.A., Dickson, J., Harwell, J., Hanlon, D., Person, C.H., 2001. An integrated software suite for surface-based analyses of cerebral cortex. J. Am. Med. Inform. Assoc. 8, 443–459. Van, Z.T., Maldonado-Molina, M.M., 2004. Response reversals in recognition memory. J. Exp. Psychol. Learn. Mem. Cogn. 30, 1147–1166. Yokoyama, O., Miura, N., Watanabe, J., Takemoto, A., Uchida, S., Sugiura, M., Horie, K., Sato, S., Kawashima, R., Nakamura, K., 2010. Right frontopolar cortex activity correlates with reliability of retrospective rating of confidence in short-term recognition memory performance. Neurosci. Res. 68, 199–206. Zhang, M., Mariola, E., Stilla, R., Stoesz, M., Mao, H., Hu, X., et al., 2005. Tactile discrimination of grating orientation: fMRI activation patterns. Hum. Brain Mapp. 25, 370–377. Zhou, D., Thompson, W.K., Siegle, G., 2009. MATLAB toolbox for functional connectivity. Neuroimage 47, 1590–1607.

D

638 639

Bates and Goldman-Rakic, 1993 Benjamini et al., 2006 Brett et al., 2002 Cools et al., 2002 Daniel and Pollmann, 2010 Preacher and Hayes, 2008 Ridderinkhof et al., 2004

E

636 637

9

Please cite this article as: Hilgenstock, R., et al., You’d Better Think Twice: Post-Decision Perceptual Confidence, NeuroImage (2014), http:// dx.doi.org/10.1016/j.neuroimage.2014.05.049

698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765