This Accepted Manuscript has not been copyedited and formatted. The final version may differ from this version.
Research Article: New Research | Cognition and Behavior
Large-Scale Network Coupling with the Fusiform Cortex Facilitates Future Social Motivation Network Coupling Facilitates Social Motivation 1,2
Amanda V. Utevsky
3
4
, David V. Smith , Jacob S. Young and Scott A. Huettel
1,2
1
Center for Cognitive Neuroscience, Duke University, Durham, NC 27708
2
Department of Psychology and Neuroscience, Duke University, Durham, NC 27708
3
Department of Psychology, Temple University, Philadelphia, PA 19122
4
Pritzker School of Medicine, University of Chicago, Chicago, IL 60637
DOI: 10.1523/ENEURO.0084-17.2017 Received: 15 March 2017 Revised: 20 September 2017 Accepted: 26 September 2017 Published: 3 October 2017
Author Contributions: AVU, DVS, JSY, and SAH designed research. AVU, DVS, JSY performed research. AVU, DVS, JYS analyzed data. AVU, DVS, SAH wrote the paper. Funding: HHS | NIH | National Institute of Mental Health (NIMH) 100000025 F31-MH105137
Funding: HHS | NIH | National Institute of Mental Health (NIMH) 100000025 F32-MH107175
Funding: HHS | National Institutes of Health (NIH) 100000002 NIMH RC1-88680
Funding: Duke Institute for Brain Sciences, Duke University (DIBS) 100005572
Conflict of Interest: Authors report no conflict of interest. This study was funded by a grant from the National Institutes of Health (NIMH RC1-88680), an Incubator Award from the Duke Institute for Brain Sciences (SAH), and by NIMH National Research Service Awards F31MH105137 (AVU) and F32-MH107175 (DVS). Correspondence should be addressed to Address for correspondence: Scott A. Huettel, Box 90999, Duke University, Durham, NC 27708, USA. E-mail:
[email protected] phone: (919) 668-5286, fax: (919) 681-0815 Cite as: eNeuro 2017; 10.1523/ENEURO.0084-17.2017 Alerts: Sign up at eneuro.org/alerts to receive customized email alerts when the fully formatted version of this article is published.
Accepted manuscripts are peer-reviewed but have not been through the copyediting, formatting, or proofreading process. Copyright © 2017 Utevsky et al. This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International license, which permits unrestricted use, distribution and reproduction in any medium provided that the original work is properly attributed.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42
Large-Scale Network Coupling with the Fusiform Cortex Facilitates Future Social Motivation Abbreviated Title: Network Coupling Facilitates Social Motivation Amanda V. Utevsky1,2, David V. Smith3, Jacob S. Young4, Scott A. Huettel1,2* 1 Center for Cognitive Neuroscience, Duke University, Durham, NC, 27708 2 Department of Psychology and Neuroscience, Duke University, Durham, NC, 27708 3 Department of Psychology, Temple University, Philadelphia, PA, 19122 4 Pritzker School of Medicine, University of Chicago, Chicago, IL, 60637 AVU, DVS, JSY, and SAH designed research. AVU, DVS, JSY performed research. AVU, DVS, JYS analyzed data. AVU, DVS, SAH wrote the paper. *Address for correspondence: Scott A. Huettel Box 90999, Duke University Durham, NC 27708, USA
[email protected] phone: (919) 668-5286 fax: (919) 681-0815 Number of Figures: 4 Number of Tables: 2 Number of words for Abstract: 233 Number of words for Significance Statement: 116 Number of words for Introduction: 637 Number of words for Discussion: 1759 Acknowledgments: This study was funded by a grant from the National Institutes of Health (NIMH RC1-88680), an Incubator Award from the Duke Institute for Brain Sciences (SAH), and by NIMH National Research Service Awards F31-MH105137 (AVU) and F32-MH107175 (DVS). The authors thank Christopher Coutlee and Cary Politzer for their assistance in task design and data collection. Conflict of Interest: Authors report no conflict of interest. Funding sources: National Institutes of Health (NIMH RC1-88680), the Duke Institute for Brain Sciences (SAH), NIMH National Research Service Awards F31-MH105137, F32MH107175 (DVS).
Page 1 of 38
43
Abstract
44
Large-scale functional networks – as identified through the coordinated activity of spatially
45
distributed brain regions – have become central objects of study in neuroscience because of
46
their contributions to many processing domains. Yet, it remains unclear how these domain-
47
general networks interact with focal brain regions to coordinate thought and action. Here,
48
we investigated how the default-mode network (DMN) and executive control network (ECN)
49
– two networks associated with goal-directed behavior – shape task performance through
50
their coupling with other cortical regions several seconds in advance of behavior. We
51
measured these networks’ connectivity during an adaptation of the monetary incentive
52
delay response-time task in which human participants viewed social and non-social images
53
(i.e., pictures of faces and landscapes, respectively) while brain activity was measured using
54
fMRI. We found that participants displayed slower response times (RTs) subsequent to
55
social trials relative to nonsocial trials. To examine the neural mechanisms driving this
56
subsequent-RT effect, we integrated independent components analysis (ICA) and a network-
57
based psychophysiological interaction (nPPI) analysis; this allowed us to investigate task-
58
related changes in network coupling that preceded the observed trial-to-trial variation in RT.
59
Strikingly, when subjects viewed social rewards, an area of the fusiform gyrus consistent
60
with the functionally-defined fusiform face area (FFA) exhibited increased coupling with the
61
ECN (relative to the DMN) – and the relative magnitude of coupling tracked the slowing of
62
RT on the following trial. These results demonstrate how large-scale, domain-general
63
networks can interact with focal, domain-specific cortical regions to orchestrate subsequent
64
behavior.
Page 2 of 38
65 66 67
Keywords: effective connectivity, executive control, default mode network, faces, goal-
68
directed behavior
69
70 71 72 73 74 75 76 77 78 79 80 81 82 83
Page 3 of 38
84
Significance
85
A diverse set of behaviors – from normal to pathological – has been linked to the responses
86
of large-scale functional networks. Yet, it remains unclear how these domain-general
87
networks shape subsequent processing in cortical regions with domain-specific function.
88
Here, we examine how two such networks — the default-mode network (DMN) and
89
executive control network (ECN) — connect functionally with other cortical regions to alter
90
performance in an incentive-compatible task. We found that differential coupling between a
91
prototypical face processing region and DMN and ECN tracked subsequent improvements in
92
performance to social stimuli. Our approach allowed us to examine direct coupling with
93
functional networks to future behavior, providing a significant step forward in
94
understanding how large-scale networks coordinate thought and action.
95 96 97
Page 4 of 38
98 99
Introduction Since the discovery of functionally correlated brain regions (Biswal et al., 1995),
100
large-scale functional networks have been considered fundamental features of brain activity
101
(Beckmann et al., 2005; Behrens and Sporns, 2012; Gusnard and Raichle, 2001). These
102
networks are highly reliable across large samples of participants (Biswal et al., 2010; Smith
103
et al., 2013) and are thought to reflect intrinsic properties of brain organization. Many of
104
these networks reflect sensory and perceptual processes instantiated within visual or
105
auditory regions (Smith et al., 2009). Still others contribute broadly to cognitive processing
106
across stimulus and task domains, including the default-mode (DMN), executive control
107
(ECN), and frontoparietal (FPN) networks (Dosenbach et al., 2007; Smith et al., 2009). These
108
functional networks have been linked to various aspects of behavior including demographic
109
variables (Cowdrey et al., 2012; Filippini et al., 2009), traits (Kucyi et al., 2014; Shannon et
110
al., 2011), and cognitive states (De Havas et al., 2012; Smith et al., 2009).
111
Yet, for effective behavior in a particular task, these large-scale domain-general
112
networks must alter ongoing task-specific processing in focal brain regions. To claim that
113
such interactions (i.e., between large-scale networks and specific brain regions) are critical
114
for behavior change, several conditions should be met. First, a large-scale network should be
115
identifiable during task performance independently of other concurrent activation; that is,
116
networks should be able to be extracted regardless of other processes occurring in focal
117
brain regions that might contribute to that task (i.e., without relying on seed-based analyses;
118
Cole et al., 2010; Smith et al., 2014b). Second, the coupling between a given network and a
119
given focal brain region should systematically vary across task conditions according to the
120
relative engagement of the task (Friston et al., 1997; O’Reilly et al., 2012). Third, those Page 5 of 38
121
changes in coupling (e.g., effective connectivity (Friston, 2011)) should predict the
122
characteristics of subsequent behavior, to provide evidence that the coupling contributes to
123
effective task performance. If these conditions are met, there would be strong evidence that
124
coupling between a functional network and a focal brain region contributes to a specific
125
cognitive process.
126
In the current study, we examined effective connectivity between large-scale
127
networks and focal brain regions while subjects played a response-time game to receive
128
social and nonsocial rewards in a modified monetary incentive delay (MID) task (Knutson et
129
al., 2000). We have previously used this task to produce meaningful variability in reaction
130
time behavior that reflects relative motivation (Clithero et al., 2011). We focused on the
131
DMN and ECN since both have been linked to task performance, engagement, and other
132
markers of executive function and preparatory behavior. Behaviorally, we found that
133
participants exhibited slower response times (RTs) subsequent to social relative to
134
nonsocial trials, reflecting a change in motivation according to social stimulus type. We
135
investigated whether coupling with DMN and ECN contributed to this subsequent-RT effect
136
by combining independent components (ICA) and psychophysiological interactions (PPI)
137
analyses. This network PPI (nPPI) pipeline allowed us to examine how DMN and ECN
138
contribute to other cortical regions to shape subsequent motivated behavior up to several
139
seconds later. Strikingly, we found that DMN and ECN differentially coupled with a region in
140
the fusiform gyrus (FG) when subjects viewed the social rewards, and that changes in this
141
coupling tracked the effect of stimulus type on subsequent RT. This region of the FG is
142
consistent with the functionally-defined fusiform face area (FFA) – a region classically
143
associated with face processing (Kanwisher et al., 1997; McCarthy et al., 1997).
Page 6 of 38
144
These findings highlight functional network interactions that guide subsequent
145
changes in reaction time behavior. While the cognitive mechanisms underlying the
146
behavioral effect of previous stimulus type on subsequent reaction time require future
147
research, we propose that this effect may be driven by increased attentional interference of
148
social rewards relative to nonsocial rewards or a subsequent change in task motivation.
149
Collectively, these results indicate that functional networks associated with goal-directed
150
behavior can interact with focal brain regions to support future motivated behavior.
151
Materials and Methods
152
Participants
153
A group of 50 self-reported heterosexual males completed the task (mean age: 23.8
154
years, range 18-32); this sample size was established prior to data collection. All participants
155
were screened prior to data collection to rule out prior or current psychiatric or neurological
156
illness. We excluded four participants because of data quality issues (see Preprocessing), and
157
we excluded five participants because of a malfunctioning button box in the scanner. These
158
exclusions left a final sample of 41 participants (mean age: 24.1 years, range: 18-32). All
159
participants gave written informed consent as part of a protocol approved by the
160
Institutional Review Board of Duke University Medical Center.
161
Stimuli and Tasks
162
The experiment consisted of four components: 1) a training session outside the
163
scanner; 2) a modified monetary incentive delay task (MID; Knutson et al., 2000, 2001); 3) a
164
passive viewing task with results previously published (Young et al., 2015); 4) two
Page 7 of 38
165
additional post-scan rating tasks. All participants received variable cash payment depending
166
on their performance.
167
In advance of the neuroimaging session, participants memorized the associations of
168
four fractal images with different potential rewards (see an example in Figure 1). On the day
169
of the scan, subjects were tested on the fractal images’ associations to ensure they
170
remembered the meaning of each image; if participants identified every association correctly
171
on the first attempt (n = 31), they received an extra $5 in addition to their study
172
compensation. All participants successfully learned the image associations before entering
173
the scanner.
174
Each trial of the modified MID task began with a fractal image presented as a cue for
175
1500 ms (Figure 1). The different cues indicated the different potential benefits of a fast
176
response on that trial: [1] reducing the delay until the presentation of a face image of a
177
variable attractiveness rating (Social Delay; SD), [2] increasing the attractiveness of a face
178
image presented after a variable duration (Social Attractiveness; SA), [3] reducing the delay
179
until the presentation of a landscape image of a variable attractiveness rating (Nonsocial
180
Delay; ND), [4] increasing the attractiveness of a landscape image presented after a variable
181
duration (Nonsocial Attractiveness; NA).
182
Following the cue, a fixation cross appeared for a variable duration (interstimulus
183
interval; 1500 – 4500 ms), followed by a target (a white triangle) that appeared on the
184
screen for 750 ms. Participants then responded by pressing a button box with their right
185
index finger before the target disappeared. A feedback shape then appeared for 750 ms; that
186
shape was colored green if the participant’s response was sufficiently fast (a win), and was
187
colored blue if the response was too slow (a loss). After the feedback image disappeared,
Page 8 of 38
188
participants waited a delay period that varied according to the cue and whether the subject
189
won or lost that trial. Lastly, the reward image (either a face or a landscape) was shown for
190
2000 ms, and there was an intertrial interval (ITI) that was jittered between 1500 and 4500
191
ms (with the 2000 ms reward image duration, the minimum and maximum time delays
192
between reward onset and subsequent cue onset was 3500 and 6500 ms, respectively).
193
Each run of the modified MID task consisted of 36 trials, nine images of each SD, SA,
194
ND, and NA in randomized order. Across all runs included in the analyses, the four types of
195
transitions (e.g., social followed by social, social followed by nonsocial, etc.) each occurred on
196
approximately 25±1% of trials, reflecting effective randomization of stimulus types across
197
trials. Social images were photographs of young adult women and were cropped to show
198
only the face (images were drawn from Smith et al. 2010). Nonsocial images were
199
photographs of landscapes, and did not contain any body or facial features. All images were
200
resized to uniform dimensions. Attractiveness (high/medium/low) of images was
201
determined by ratings by an independent group. Additionally, performance thresholds were
202
determined independently for each subject by an algorithm that adapted the task difficulty
203
so that accuracy would be approximately 60% (observed hit rate = 60.7%).
204
After completing the four runs of the modified MID task, participants completed an
205
unrelated passive-viewing task described elsewhere (Young et al., 2015). Following this
206
second scanner task, participants completed two tasks outside the scanner. In the first task,
207
participants rated the attractiveness of each image they viewed in the modified MID task on
208
a scale of 1 to 7. Each participant’s highest-rated face and landscape images were then used
209
in a second task involving forced choices between a landscape image and a face image
210
presented simultaneously.
Page 9 of 38
211
All tasks were programmed and displayed using the Psychophysics Toolbox (Version
212
3; Brainard, 1997) for MATLAB (Mathworks Inc.).
213
Behavioral Analysis
214
Behavioral data were analyzed using MATLAB and IBM SPSS Statistics 20. A response
215
time (RT) difference was calculated for each participant by subtracting their mean RT on the
216
trials subsequent to nonsocial trials from their mean RT on trials subsequent to social trials.
217
Consequently, positive RT differences indicate shorter RTs subsequent to nonsocial trials
218
(relative to social trials), whereas negative differences indicate shorter RTs subsequent to
219
social trials (relative to nonsocial trials).
220
To control for factors other than prior stimulus category – such as attractiveness of
221
the prior trial’s reward image, the prior trial’s RT, and the subsequent trial’s stimulus type –
222
we incorporated these factors into a general linear model (GLM) for each subject. The GLM
223
included the following as regressors: (1) stimulus type on previous (t-1) trial, (2) RT on
224
previous trial, (3) attractiveness rating of the reward image of the previous trial, and (4)
225
stimulus type on current trial t. This analysis indicates the specific influence of each
226
regressor on the current trial’s RT, and allowed us to examine the unique influence of the
227
previous trial’s stimulus category (social or nonsocial) on current RT. In our analyses,
228
positive stimulus type beta weights reflect slower RTs on or following social trials, whereas
229
negative stimulus type betas weights reflect faster RTs on or following social trials.
230
Image Acquisition
231 232
Neuroimaging data were collected using a General Electric MR750 3.0 Tesla scanner equipped with an 8-channel parallel imaging system. We used a T2*-weighted spiral-in
Page 10 of 38
233
sensitivity encoding sequence (SENSE factor = 2), with slices parallel to the axial plane
234
connecting the anterior and posterior commissures (repetition time (TR): 1580 ms; echo
235
time (TE): 30 ms; matrix: 64 x 64; field of view (FOV): 243 mm; voxel size: 3.8 x 3.8 x 3.8
236
mm; 37 axial slices; flip angle: 70 degrees). The first eight volumes of each run were
237
removed to allow for magnetic stabilization. We additionally acquired whole-brain high-
238
resolution anatomical scans (T1-weighted FSPGR sequence; TR: 7.58 ms; TE: 2.93 ms;
239
matrix: 256 x 256; FOV: 256 mm; voxel size: 1 x 1 x 1 mm; 206 axial slices; flip angle: 12
240
degrees) to allow for coregistration and normalization.
241
Preprocessing
242
Our preprocessing used tools from the FMRIB Software Library package (FSL Version
243
4.1.8; http://www.fmrib.ox.ac.uk/fsl/; Smith et al., 2004; Woolrich et al., 2009). We
244
corrected for head motion by realigning the time series to the middle volume (Jenkinson and
245
Smith, 2001), and then removed non-brain material using a brain extraction tool (Smith,
246
2002). We then corrected intravolume slice-timing differences using Fourier-space phase
247
shifting to align to the middle slice (Sladky et al., 2011). After spatially smoothing the image
248
using a 5-mm full-width-half-maximum isotropic Gaussian kernel, we applied a high-pass
249
temporal filter with a 100-second cutoff, and we normalized each 4-dimensional dataset to
250
the grand-mean intensity using a single multiplicative factor. Lastly, we spatially normalized
251
the functional data to the Montreal Neurological Institute (MNI) Template avg152 T1-
252
weighted template (3 mm isotropic resolution) using a 12-parameter affine transformation
253
implemented in FLIRT (Jenkinson and Smith, 2001).
254 255
As part of our preprocessing and quality control, we additionally examined three partially correlated measures of quality assurance: signal-to-fluctuation-noise ratio (SFNR; Page 11 of 38
256
Friedman and Glover, 2006), volume-to-volume head motion, and number of motion spikes
257
within the time series (motion spikes were identified by evaluating the root-mean-square-
258
error of each volume relative to the middle time point). Measures on each metric were
259
considered outliers if they exceeded the 75th percentile plus the value of 150% of the
260
interquartile range (i.e., a standard boxplot threshold); runs that were identified as outliers
261
were excluded from further analyses. Additionally, any participant who had fewer than two
262
good runs (out of four total runs) was excluded from further analyses. These criteria
263
eliminated four participants.
264
Neuroimaging Analysis
265
To best meet the conditions for identifying task-specific interactions between large-
266
scale networks and focal brain regions (see Introduction), our neuroimaging analyses
267
proceeded in two phases, each described in a separate section below. First, we used
268
independent components analysis (ICA; Beckmann & Smith, 2004) and spatial regression
269
(Filippini et al., 2009) to identify the large-scale neural networks of interests (DMN and ECN)
270
and to examine the networks’ levels of activation over the course of each run. Second, we
271
used generalized network psychophysiological interaction (nPPI) models (adapted from
272
McLaren et al., 2012) to identify brain regions whose coupling with the ECN and DMN
273
changed as a function of the effect of stimulus type on subsequent RT. Importantly, this ICA-
274
based nPPI approach follows the logic of ROI-based PPI analyses, with the critical difference
275
of examining connectivity with data-driven large-scale neural networks instead of a specific
276
seed region. Critically, we note that this nPPI pipeline allowed us to test the three necessary
277
conditions for inferences that functional networks alter task-specific processing in focal
278
brain regions. Page 12 of 38
279
Identifying Large-Scale Functional Networks
280
We used FSL’s Multivariate Exploratory Linear Decomposition into Independent
281
Components (MELODIC) Version 3.10 to identify large-scale functional networks in the
282
neuroimaging data (Beckmann and Smith, 2004). MELODIC ICA was implemented using
283
temporal concatenation, which looks for common spatial patterns of components across
284
participants’ data without assuming a specific or common time course across all
285
participants; we note that previous research has also used temporal concatenation-based
286
ICA on task-based fMRI data (Utevsky et al., 2014; Young et al., 2015). The preprocessed data
287
were whitened and projected into a 25-dimensional subspace (Ray et al., 2013; Utevsky et
288
al., 2014). The whitened data were decomposed into sets of vectors describing the temporal
289
and spatial signal variation, using a fixed-point iteration technique to optimize non-Gaussian
290
spatial source distribution (Hyvärinen, 1999). The estimated component maps were then
291
thresholded by dividing the maps by the standard deviation of the residual noise, then fitting
292
a Gaussian-Gamma mixture model to the histogram of the normalized intensity values
293
(Beckmann and Smith, 2004). This first step provided a data-driven means to identify
294
functional networks present during task performance; this allowed us to meet the first
295
condition for identifying task-specific interactions between large-scale networks and specific
296
brain regions.
297
All unthresholded spatial maps from the ICA were then submitted to a spatial
298
regression (part of FSL’s dual regression analysis) to estimate the time courses of each
299
network (Filippini et al., 2009; Leech et al., 2011). In this analysis, spatial maps are regressed
300
onto each participant’s functional data, resulting in a matrix of T (time points) x C
Page 13 of 38
301
(components) beta coefficients that characterize each subject’s time courses for each
302
network.
303
Characterizing Reward-Related Network Connectivity and Activation
304
We assessed task-dependent network coupling using a generalized nPPI model. The
305
generalized PPI, in contrast to a standard PPI, computes a separate PPI term for each task
306
condition. This approach has been shown to yield more accurate estimates of how
307
connectivity varies as a function of psychological context (McLaren et al., 2012). The
308
generalized nPPI analysis was carried out using FMRI Expert Analysis Tool (FEAT) Version
309
5.0.1.
310
The run-level model included six task regressors: social cue (duration = 1.5 – 4.5 s),
311
nonsocial cue (duration = 1.5 – 4.5 s), hits (duration = 0.75 s), misses (duration = 0.75 s),
312
social reward outcome (duration = 2 s), and nonsocial reward outcome (duration = 2 s). We
313
additionally included time courses of both the DMN and ECN that were produced by the
314
spatial regression. Because of the minimum ITI of 1500 ms and the potential confound of the
315
cue presentation occurring between the outcome and subsequent target phases of the task,
316
we examined whether there was any collinearity between the outcomes and cues across the
317
runs included in our analysis. For each run included in our analyses, we calculated the
318
correlation values between the face and land outcome regressors and the face and land cue
319
regressors – and then averaged these correlation values across all runs. The mean
320
correlation values between face or land cues and face or land outcomes ranged from r = -
321
0.25 to r = -0.17 (minimum r = -0.28; maximum r = -0.10), indicating that our run-level
322
analyses were able to isolate and model cues and outcomes independently, with minimal
323
collinearity between regressors. Page 14 of 38
324
For our network interaction analysis, nPPI regressors were formed by multiplying the
325
DMN and ECN time courses (zeroed to the mean of the time course), separately, by the social
326
outcome and nonsocial outcome regressors (zeroed to the minimum value of the task time
327
course) (McLaren et al., 2012); this yielded four nPPI regressors: (1) DMN*social, (2)
328
DMN*nonsocial, (3) ECN*social, (4) ECN*nonsocial. To control for motion in the scanner, we
329
additionally included motion spikes and motion parameters as regressors. Lastly, to control
330
for the influence of other networks and potential artifacts on our generalized nPPI, we
331
included the time courses of the remaining 23 components from the ICA. The nPPI analysis
332
allowed us to examine task-specific coupling between ECN and DMN with other regions in
333
the brain, fulfilling the second condition for identifying the task-specific interactions of
334
interest.
335
Subject-level analyses for the generalized nPPI were run using FEAT and
336
implementing FMRIB’s Local Analysis of Mixed Effects (FLAME 1), and examined activation
337
across runs within each participant. Group-level analyses included each subject’s demeaned
338
betas from the prior stimulus-category regressor (see Behavioral Analysis); this allowed us
339
to examine whether network coupling predicts the characteristics of subsequent RT, and
340
fulfill the third condition for identifying task-specific interactions of interest. The group-level
341
analysis additionally included the main effect of group and three motion-related parameters
342
(SFNR, volume-to-volume head motion, and number of motion spikes within the time
343
series). All resulting z-statistic images were thresholded using a cluster-forming threshold of
344
2.3 and a corrected cluster-significance threshold of p < 0.05. Although this threshold
345
combined with FSL’s FLAME 1 protects against false positives, we note that all of our results
346
also survived permutation-based testing (Eklund et al., 2016). In these supplemental tests,
Page 15 of 38
347
statistical significance was assessed in a nonparametric fashion via FSL’s randomise; this tool
348
uses Monte Carlo permutation-based testing with 10,000 permutations and D = 0.05,
349
corrected for multiple comparisons across the whole brain (Nichols and Holmes, 2002;
350
Winkler et al., 2014).
351
Brain images and activations are displayed using MRIcroGL
352
(http://www.mccauslandcenter.sc.edu/mricrogl/; Rorden et al., 2007). All coordinates are
353
reported in Montreal Neurological Institute (MNI) space.
354
Results
355
Previous stimulus category influences current behavior
356
Social rewards, such as images of individuals or interactions with others, provide a
357
useful tool for examining the effects of both social context and non-consumable rewards on
358
motivated behavior. However, relatively little is known about the effects of social and
359
nonsocial rewards on future motivated behavior, and how the brain orchestrates future
360
motivated action. To examine the effect of social rewards on subsequent motivated actions,
361
we calculated a subsequent-RT effect by subtracting RTs following nonsocial trials from RTs
362
following social trials. When averaging across value and delay trials, we found an overall
363
effect of previous trial stimulus type on current trial RT: participants exhibited increased
364
(slower) RTs subsequent to social trials (M = 0.313 s, SD = 0.008 s) compared to subsequent
365
to nonsocial trials (M = 0.306 s, SD = 0.008 s; t(40) = 2.63, p = 0.01, d = 0.41; Figure 2A). This
366
pattern replicated when examining hit trials only (social: M = 0.317 s, SD = 0.009 s;
Page 16 of 38
367
nonsocial: M = 0.309 s, SD = 0.008 s; t(40) = 2.63, p = 0.01, d = 0.41). Thus, participants were
368
slower after performing a social trial compared to a nonsocial trial.
369
We next ran a general linear model (GLM) to control for properties of the previous
370
and current trials (see Behavioral Analysis). Across participants, current stimulus category
371
was the strongest predictor of current RT (t(40) = -4.31, p < .0001, d = 0.67), reflecting that
372
participants were faster to respond during a social trial than during a nonsocial trial. The
373
next strongest predictor was the previous trial’s stimulus category (t(40) = 2.46, p = 0.01, d =
374
0.38), reflecting that participants were slower to respond following a social trial than
375
following a nonsocial trial. Neither previous trial’s RT nor previous trial’s reward
376
attractiveness rating significantly predicted the current trial’s RT (Figure 2B; Table 1). We
377
note that the effect of previous stimulus type on current RT was still significant even after
378
accounting for the large effect of current stimulus type, reflecting a distinct role of previous
379
stimulus type unaccounted for by the other variables included in the GLM. These results
380
indicate that of the measured properties of the previous trial, prior stimulus type had the
381
strongest effect on current RT.
382
Network coupling with fusiform gyrus tracks effect of stimulus category on subsequent
383
behavior
384
Our behavioral results indicated that current-trial stimulus type influences motivated
385
behavior on the subsequent trial (which occurred seven seconds or more later), such that
386
participants were slower to respond following social trials compared to following nonsocial
387
trials. However, it remains unclear how interactions in the brain during this previous
388
stimulus outcome affect the subsequent RT. To investigate this, we examined how effective
Page 17 of 38
389
connectivity with the ECN – a network implicated in cognitive control and goal-directed
390
behavior – and the DMN – a network linked with task engagement orchestrate this change in
391
future motivated behavior. We predicted that this subsequent-RT effect would be guided, in
392
part, by changes in the coupling between these large-scale functional networks and domain-
393
specific brain regions when viewing social images compared to when viewing nonsocial
394
images. In particular, we predicted that the subsequent-RT effect would be driven by
395
changes in the relative coupling of the ECN and the DMN during the reward outcome phase
396
of the previous trial. Our analysis pipeline allowed us to meet the criteria to claim that
397
interactions between large-scale networks and specific brain regions are critical for behavior
398
changes (see Introduction).
399
After running the ICA, we identified the DMN and ECN maps by running spatial
400
correlations between the unthresholded maps from our ICA and the DMN and ECN maps
401
from Smith and colleagues (2009). From our 25 components, we selected the maps that best
402
matched the DMN (R = 0.776; other components: Rmean = -0.006, Rmin = -0.179, Rmax = 0.124)
403
and ECN (R = 0.64; other components: Rmean = 0.019, Rmin = -0.092, Rmax = 0.296) maps from
404
Smith and colleagues (2009) (see Table 2). For ease of visualization, thresholded maps (Z >
405
4) are shown in Figure 3.
406
Following network identification, we ran a nPPI analysis that used participant-
407
specific DMN and ECN time courses (estimated by the spatial regression analysis) as the
408
physiological regressors and presentation of social and nonsocial images as the
409
psychological regressors. This nPPI identified regions that are influenced by the ECN and
410
DMN in a task-dependent manner. We then tested whether these influences on cortex
411
predicted the effect of prior stimulus type on current RT. Our nPPI analysis indicated that
Page 18 of 38
412
effective connectivity between the fusiform gyrus (FG; peak: x = 38, y = -64, z = -20, p
social cues. Notably, there
425
were no regions exhibiting changes in connectivity with the ECN relative to the DMN in any
426
of the three contrasts tested. To further investigate this, we used the FG region showing
427
changes in ECN-DMN connectivity during the previous outcome phase as a mask during the
428
current cue phase, and examined whether the ECN-DMN connectivity from this region of
429
interest (ROI) tracked the effect of prior stimulus on current RT. Across the three contrasts
430
tested (social, nonsocial, nonsocial > social), there was no correlation between the ROI
431
connectivity estimates and the effect of previous stimulus type on current RT (social: r = -
432
0.12, p = 0.45; nonsocial: r = 0.01, p = 0.94; nonsocial > social: r = 0.23, p = 0.15). Collectively,
433
these results support the claim that changes in subsequent RT are associated with
Page 19 of 38
434
differential network connectivity with the FG during the previous outcome phase, and
435
cannot be attributed to other aspects of task performance.
436
No regions showed increased coupling with ECN (relative to DMN) during nonsocial
437
rewards that tracked the effect on subsequent RT, nor did any regions show increased
438
coupling with DMN (relative to ECN) during social or nonsocial rewards that tracked the
439
effect on subsequent RT. Additionally, there were no regions that exhibited changes in
440
coupling with the DMN or ECN when comparing social and nonsocial rewards (e.g., DMN-
441
social > DMN-nonsocial; ECN-social > ECN-nonsocial; and the inverse contrasts) that tracked
442
the effect on subsequent RT. Lastly, we ran a whole-brain GLM using the RT betas as a
443
covariate to examine whether any regions’ activation tracked the effect of stimulus type on
444
subsequent RT; we found that no regions tracked this effect using this traditional GLM. Thus,
445
coupling between large-scale networks and the FG that tracked the effect on subsequent RT
446
was only observed during the viewing of social rewards.
447
Discussion
448
Recent neuroscience research has highlighted the relevance of large-scale functional
449
networks to various aspects of behavior (Anticevic et al., 2010; Eichele et al., 2008; Kelly et
450
al., 2008; Rosen et al., 2015). While many of these studies have linked network activation
451
and connectivity to behavior, the contribution of these networks to motivated behaviors via
452
focal cortical regions has been relatively understudied. For example, although previous work
453
has found correlations between DMN and working memory (Piccoli et al., 2015; Sambataro
454
et al., 2010) or sustained attention (Bonnelle et al., 2011; Gui et al., 2015), understanding
455
how these distributed functional networks influence other cortical regions to influence
Page 20 of 38
456
behavior has remained a significant challenge. Here, we found that participants were slower
457
in a response time task after having performed a trial for a social reward relative to a non-
458
social reward (after accounting for the influence of the current trial cue type), reflecting a
459
change in motivated behavior according to previous social stimulus type (Clithero et al.,
460
2011). We then examined the neural mechanisms underlying this effect using nPPI analysis –
461
an adaptation of generalized PPI analysis (i.e., effective connectivity) that examines effective
462
connectivity with large-scale functional networks. This analysis pipeline allowed us to
463
identify changes in coupling between large-scale networks and focal brain regions that
464
predicted subsequent motivated behavior. Our results demonstrate that two goal-relevant
465
networks, the DMN and ECN, interact with FG in a manner that predicts trial-to-trial
466
adjustments in response time.
467
Our analysis on our modified MID task identified that participants exhibit slower
468
reaction times on trials subsequent to social outcomes, relative to trials subsequent to
469
nonsocial outcomes. These results are consistent with a difference in motivational processes
470
due to the social nature of the previous reward outcome, analogous to effects on motivation
471
observed in similar tasks with nonsocial stimuli (Clithero et al., 2011). There are at least two
472
potential explanations for this subsequent reaction time effect. The first hypothesis is that
473
social images are more distracting than nonsocial images, and participants may still covertly
474
attend to a prior social reward more so than they do a prior nonsocial reward. Previous
475
research indicates that social images interfere with visual attention to a greater degree than
476
nonsocial images (Ebitz et al., 2013); thus, it is possible that this differential interference
477
may affect subsequent trials differently, as well. A second and related potential explanation
478
is that participants’ decreased motivation can be attributed to the satisfaction of receiving a
Page 21 of 38
479
motivating reward on the previous trial, much akin to results seen in “satisfaction of search”
480
(SOS) research. In SOS research, participants performing visual search tasks tend to
481
discontinue their search after finding an initial item, and either miss or display slower
482
reaction times for subsequent items (Berbaum et al., 1990; Fleck et al., 2010). As applied to
483
our MID task, a more motivating reward outcome (social images) may hinder the
484
performance on the subsequent trial. Future research will be needed to explore the cognitive
485
mechanisms driving the observed subsequent-RT effect and to determine whether it is
486
specific to social versus nonsocial stimuli, or may generalize to items that are more
487
motivating versus less motivating.
488
Our findings expand on recent research examining the relevance of DMN and ECN
489
activation to behavior, potentially in task-relevant contexts. Specifically, we found that
490
increased ECN (relative to DMN) coupling with the FG is associated with enhanced
491
subsequent task performance. These results are consistent with previous studies
492
demonstrating different relationships between the DMN and ECN with task behavior: while
493
ECN activation is often associated with heightened task performance and behavior
494
(Dosenbach et al., 2007; Seeley et al., 2007), DMN activation is frequently linked to
495
decrements in behavior and engagement in both humans (Eichele et al., 2008; Weissman et
496
al., 2006) and non-human primates (Hayden et al., 2009, 2010; Heilbronner and Platt, 2013).
497
Importantly, however, our findings extend these previous results by demonstrating direct
498
coupling of these networks with a prototypical social-perception processing region in a task-
499
dependent manner. Our results additionally highlight that the opposing effects of DMN and
500
ECN are not limited to concurrent behavior, but also affect subsequent behavior with effects
501
observed seven or more seconds later. These findings support the idea that large-scale
Page 22 of 38
502
networks interact with lower-level perceptual regions to contribute to motivational
503
processes (Clithero et al., 2011) and shape later behavior.
504
Unlike previous studies examining large-scale networks (Brewer et al., 2011;
505
Ossandon et al., 2011; Seeley et al., 2007; Utevsky et al., 2014), our experiment demonstrates
506
network coupling that directly shapes subsequent response time. While prior studies report
507
associations between functional connectivity and behavior, estimates of functional
508
connectivity solely report on correlations in activation (not coupling) between regions,
509
which can be the result of various phenomena (Friston, 2011). Specifically, reported changes
510
in functional connectivity can arise from numerous causes, including: changes in
511
connectivity with another region, changes in observation noise (or signal-to-noise ratios),
512
and changes in the degree of neuronal fluctuations (for a review, see Friston, 2011). Thus,
513
changes in functional connectivity may not reflect changes in coupling between cortical
514
regions. In contrast to functional connectivity, we implemented an adaptation of a traditional
515
PPI analysis to measure effective connectivity between functional networks and other
516
cortical regions. Psychophysiological interaction analyses measure whether a psychological
517
context (e.g., outcome stimulus category) influences how one brain region or network (the
518
“seed”) contributes to another (the “target”) by examining whether an interaction between
519
the psychological context and the seed is identified in the target (Smith et al., 2016). Thus, a
520
PPI analysis can eschew the potential confounds of functional connectivity analyses and
521
reflects a change in neural coupling (e.g., effective connectivity). We note that our novel
522
approach of generalized nPPI – applying generalized PPI analyses to large-scale networks –
523
allows us to examine task-dependent contributions between networks and other cortical
Page 23 of 38
524
regions (Friston et al., 1997; Friston, 2011). In this way, our study extends prior work by
525
demonstrating specific task-dependent coupling of the DMN/ECN.
526
Although recent meta-analytic work has demonstrated that PPI produces consistent
527
and specific patterns of connectivity (Smith et al., 2016; Smith and Delgado, 2017), it is
528
important to note that PPI results can be interpreted in two ways (Friston et al., 1997). First,
529
our effects could reflect a context-specific modulation of effective connectivity. In this case,
530
face presentations modulate the degree to which the DMN and ECN contribute to FG. Our
531
results focus on the difference between DMN and ECN contributions to FG which seem to
532
facilitate social motivation (Clithero et al., 2011). Alternatively, our effects could reflect a
533
modulation of stimulus-specific responses. In this case, the DMN and ECN influence how FG
534
responds to the presentation of the face; under this interpretation, our results suggest that
535
the degree to which DMN enhances face responses in FG is greater than that of ECN. In either
536
interpretation, the resulting effect on FG connectivity predicts behavior. A better
537
understanding of the mechanisms and causal relationship underlying our results may be
538
facilitated by other analytical approaches, such as dynamic causal modeling (DCM) (Friston
539
et al., 2003; Friston, 2011), although relatively less is understood regarding how the
540
biophysical models implemented in DCM apply to distributed functional networks (Buxton
541
et al., 1998; Friston et al., 2003). Additionally, other methodological approaches, such as
542
transcranial magnetic stimulation (Fox et al., 2012; Luber and Lisanby, 2014; Mueller et al.,
543
2014) or transcranial current stimulation (Hampstead et al., 2014; Keeser et al., 2011) may
544
also better inform the specific interactions between FG and the ECN and DMN.
545
One potential caveat to note is that our task design included a cue to indicate the current
546
trial’s stimulus type in between the previous trial’s outcome phase and the current trial’s
Page 24 of 38
547
response phase. Because the current trial’s stimulus type is the strongest predictor of the
548
current trial’s RT, there may be concerns over whether the neural coupling we observe can
549
be attributed to the cue, rather than the previous trial’s outcome. This concern is
550
ameliorated through both the randomization of trial types (i.e., the identity of the outcome
551
on the previous trial is uncorrelated with the trial type on the subsequent trial), as well as
552
the run-level models implemented in our PPI analysis that included separate regressors for
553
both the outcome phase (social and nonsocial outcomes, separately) and the cue phase
554
(social and nonsocial cues, separately). Including these separate regressors for each phase
555
allowed us to distinguish effects of the cue and outcome phases on connectivity. This
556
concern is additionally mitigated by the analyses examining connectivity during the cue
557
phase. A nPPI showed no changes in ECN-DMN connectivity during social cues, nonsocial
558
cues, or social > nonsocial cues when looking across the whole brain. A supplementary ROI
559
analysis using the FG region indicated that FG connectivity with the ECN (relative to the
560
DMN) during these three contrasts did not track the effect of the previous trial stimulus type
561
on subsequent RT (as illustrated by the lack of significant correlations between the
562
connectivity estimates and the effect of previous stimulus type on subsequent RT). However,
563
future research may benefit from a modified task design in which there are no informative
564
stimuli presented between a reward image and subsequent response.
565
As an additional caveat, we note that our paradigm and results leave room for
566
interpretive challenges. Because participants exhibited faster response times to view social
567
images compared to nonsocial images, we cannot discern whether our nPPI results from the
568
social-stimuli condition are due to the stimulus type itself (i.e. face images) or would
569
generalize to other highly motivating stimuli. While prior work linking FG to face processing
Page 25 of 38
570
supports the interpretation that our results are indeed due to that specific stimulus type
571
(Engell and McCarthy, 2014; Kanwisher et al., 1997; Kanwisher and Yovel, 2006; McCarthy et
572
al., 1997), future studies could compare within-domain images of varying attractiveness to
573
other categories of motivating stimuli (e.g., money). Such an analysis would speak to
574
whether these behavioral and neural results are unique to face image rewards, or are due to
575
participant differences in subjective value (Bos et al., 2013; Smith et al., 2014a).
576
Our results demonstrate task-dependent contributions between the DMN and ECN
577
and the FG that shape subsequent motivated behavior. These large-scale networks are
578
known to be disrupted in a variety of psychopathologies marked by impairments in
579
attention and reward processing, including autism spectrum disorder (Assaf et al., 2010;
580
Young et al., 2015), obsessive compulsive disorder (Stern et al., 2011, 2012), and major
581
depressive disorder (Grimm et al., 2011; Sheline et al., 2009), and so an improved
582
understanding of how they influence moment-to-moment behavior could have clinical
583
relevance and advance models of pathophysiology (Cuthbert and Insel, 2013; Insel et al.,
584
2010). Thus, this study marks a significant step toward better understanding and treatment
585
of disorders characterized by impaired social and reward processing.
Page 26 of 38
586
References
587 588 589
Anticevic, A., Repovs, G., Shulman, G.L., Barch, D.M., 2010. When less is more: TPJ and default network deactivation during encoding predicts working memory performance. NeuroImage 49, 2638–2648. doi:10.1016/j.neuroimage.2009.11.008
590 591 592 593
Assaf, M., Jagannathan, K., Calhoun, V.D., Miller, L., Stevens, M.C., Sahl, R., O’Boyle, J.G., Schultz, R.T., Pearlson, G.D., 2010. Abnormal functional connectivity of default mode sub-networks in autism spectrum disorder patients. NeuroImage 53, 247–256. doi:10.1016/j.neuroimage.2010.05.067
594 595 596
Beckmann, C.F., DeLuca, M., Devlin, J.T., Smith, S.M., 2005. Investigations into resting-state connectivity using independent component analysis. Philos. Trans. R. Soc. Lond. 360, 1001–1013. doi:10.1098/rstb.2005.1634
597 598 599
Beckmann, C.F., Smith, S.M., 2004. Probabilistic independent component analysis for functional magnetic resonance imaging. IEEE Trans. Med. Imaging 23, 137–52. doi:10.1109/TMI.2003.822821
600 601
Behrens, T.E., Sporns, O., 2012. Human connectomics. Curr. Opin. Neurobiol. 22, 144–53. doi:10.1016/j.conb.2011.08.005
602 603 604
Berbaum, K.S., Franken, E.A., Dorfman, D.D., Rooholamini, S.A., Kathol, M.H., Barloon, T.J., Behlke, F.M., Sato, Y., Lu, C.H., el-Khoury, G.Y., 1990. Satisfaction of search in diagnostic radiology. Invest. Radiol. 25, 133–140.
605 606 607
Biswal, B., Zerrin Yetkin, F., Haughton, V.M., Hyde, J.S., 1995. Functional connectivity in the motor cortex of resting human brain using echo-planar mri. Magn. Reson. Med. 34, 537– 541. doi:10.1002/mrm.1910340409
608 609 610 611 612 613 614 615 616 617
Biswal, B.B., Mennes, M., Zuo, X.-N., Gohel, S., Kelly, C., Smith, S.M., Beckmann, C.F., Adelstein, J.S., Buckner, R.L., Colcombe, S., Dogonowski, A.-M., Ernst, M., Fair, D., Hampson, M., Hoptman, M.J., Hyde, J.S., Kiviniemi, V.J., Kötter, R., Li, S.-J., Lin, C.-P., Lowe, M.J., Mackay, C., Madden, D.J., Madsen, K.H., Margulies, D.S., Mayberg, H.S., McMahon, K., Monk, C.S., Mostofsky, S.H., Nagel, B.J., Pekar, J.J., Peltier, S.J., Petersen, S.E., Riedl, V., Rombouts, S.A.R.B., Rypma, B., Schlaggar, B.L., Schmidt, S., Seidler, R.D., Siegle, G.J., Sorg, C., Teng, G.-J., Veijola, J., Villringer, A., Walter, M., Wang, L., Weng, X.-C., Whitfield-Gabrieli, S., Williamson, P., Windischberger, C., Zang, Y.-F., Zhang, H.-Y., Castellanos, F.X., Milham, M.P., 2010. Toward discovery science of human brain function. Proc. Natl. Acad. Sci. 107, 4734–4739. doi:10.1073/pnas.0911855107
618 619 620 621
Bonnelle, V., Leech, R., Kinnunen, K.M., Ham, T.E., Beckmann, C.F., Boissezon, X.D., Greenwood, R.J., Sharp, D.J., 2011. Default Mode Network Connectivity Predicts Sustained Attention Deficits after Traumatic Brain Injury. J. Neurosci. 31, 13442–13451. doi:10.1523/JNEUROSCI.1163-11.2011
Page 27 of 38
622 623 624
Bos, W. van den, Talwar, A., McClure, S.M., 2013. Neural Correlates of Reinforcement Learning and Social Preferences in Competitive Bidding. J. Neurosci. 33, 2137–2146. doi:10.1523/JNEUROSCI.3095-12.2013
625
Brainard, D.H., 1997. The Psychophysics Toolbox. Spat. Vis. 10, 433–436.
626 627 628
Brewer, J.A., Worhunsky, P.D., Gray, J.R., Tang, Y.-Y., Weber, J., Kober, H., 2011. Meditation experience is associated with differences in default mode network activity and connectivity. Proc. Natl. Acad. Sci. 108, 20254–9. doi:10.1073/pnas.1112029108
629 630
Buxton, R.B., Wong, E.C., Frank, L.R., 1998. Dynamics of blood flow and oxygenation changes during brain activation: the balloon model. Magn. Reson. Med. 39, 855–64.
631 632 633
Clithero, J.A., Reeck, C., Carter, R.M., Smith, D.V., Huettel, S.A., 2011. Nucleus accumbens mediates relative motivation for rewards in the absence of choice. Front. Hum. Neurosci. 5, 87. doi:10.3389/fnhum.2011.00087
634 635 636
Cole, D.M., Smith, S.M., Beckmann, C.F., 2010. Advances and pitfalls in the analysis and interpretation of resting-state FMRI data. Front. Syst. Neurosci. 4, 8. doi:10.3389/fnsys.2010.00008
637 638 639
Cowdrey, F.A., Filippini, N., Park, R.J., Smith, S.M., McCabe, C., 2012. Increased resting state functional connectivity in the default mode network in recovered anorexia nervosa. Hum. Brain Mapp. 0. doi:10.1002/hbm.22202
640 641
Cuthbert, B.N., Insel, T.R., 2013. Toward the future of psychiatric diagnosis: the seven pillars of RDoC. BMC Med. 11, 126. doi:10.1186/1741-7015-11-126
642 643 644
De Havas, J.A., Parimal, S., Soon, C.S., Chee, M.W.L., 2012. Sleep deprivation reduces default mode network connectivity and anti-correlation during rest and task performance. NeuroImage 59, 1745–1751. doi:10.1016/j.neuroimage.2011.08.026
645 646 647 648
Dosenbach, N.U.F., Fair, D.A., Miezin, F.M., Cohen, A.L., Wenger, K.K., Dosenbach, R. a T., Fox, M.D., Snyder, A.Z., Vincent, J.L., Raichle, M.E., Schlaggar, B.L., Petersen, S.E., 2007. Distinct brain networks for adaptive and stable task control in humans. Proc. Natl. Acad. Sci. 104, 11073–11078. doi:10.1073/pnas.0704320104
649 650
Ebitz, R.B., Watson, K.K., Platt, M.L., 2013. Oxytocin blunts social vigilance in the rhesus macaque. Proc. Natl. Acad. Sci. 110, 11630–11635. doi:10.1073/pnas.1305230110
651 652 653
Eichele, T., Debener, S., Calhoun, V.D., Specht, K., Engel, A.K., Hugdahl, K., von Cramon, D.Y., Ullsperger, M., 2008. Prediction of human errors by maladaptive changes in event-related brain networks. Proc. Natl. Acad. Sci. 105, 6173–8. doi:10.1073/pnas.0708965105
654 655 656
Eklund, A., Nichols, T.E., Knutsson, H., 2016. Cluster failure: Why fMRI inferences for spatial extent have inflated false-positive rates. Proc. Natl. Acad. Sci. 113, 7900–7905. doi:10.1073/pnas.1602413113
Page 28 of 38
657 658 659
Engell, A.D., McCarthy, G., 2014. Face, eye, and body selective responses in fusiform gyrus and adjacent cortex: an intracranial EEG study. Front. Hum. Neurosci. 8. doi:10.3389/fnhum.2014.00642
660 661 662 663
Filippini, N., MacIntosh, B.J., Hough, M.G., Goodwin, G.M., Frisoni, G.B., Smith, S.M., Matthews, P.M., Beckmann, C.F., Mackay, C.E., 2009. Distinct patterns of brain activity in young carriers of the APOE-ϵ4 allele. Proc. Natl. Acad. Sci. 106, 7209–7214. doi:10.1073/pnas.0811879106
664 665 666
Fleck, M.S., Samei, E., Mitroff, S.R., 2010. Generalized “satisfaction of search”: adverse influences on dual-target search accuracy. J. Exp. Psychol. 16, 60–71. doi:10.1037/a0018629
667 668 669 670
Fox, M.D., Halko, M.A., Eldaief, M.C., Pascual-Leone, A., 2012. Measuring and manipulating brain connectivity with resting state functional connectivity magnetic resonance imaging (fcMRI) and transcranial magnetic stimulation (TMS). NeuroImage 62, 2232–43. doi:10.1016/j.neuroimage.2012.03.035
671 672 673
Friedman, L., Glover, G.H., 2006. Reducing interscanner variability of activation in a multicenter fMRI study: controlling for signal-to-fluctuation-noise-ratio (SFNR) differences. NeuroImage 33, 471–81. doi:10.1016/j.neuroimage.2006.07.012
674 675
Friston, K.J., 2011. Functional and Effective Connectivity: A Review. Brain Connect. 1, 13–36. doi:10.1089/brain.2011.0008
676 677 678
Friston, K.J., Buechel, C., Fink, G.R., Morris, J., Rolls, E., Dolan, R.J., 1997. Psychophysiological and Modulatory Interactions in Neuroimaging. NeuroImage 6, 218–229. doi:10.1006/nimg.1997.0291
679 680
Friston, K.J., Harrison, L., Penny, W., 2003. Dynamic causal modelling. NeuroImage 19, 1273– 1302. doi:10.1016/S1053-8119(03)00202-7
681 682 683
Grimm, S., Ernst, J., Boesiger, P., Schuepbach, D., Boeker, H., Northoff, G., 2011. Reduced negative BOLD responses in the default-mode network and increased self-focus in depression. World J. Biol. Psychiatry 12, 627–637. doi:10.3109/15622975.2010.545145
684 685 686
Gui, D., Xu, S., Zhu, S., Fang, Z., Spaeth, A.M., Xin, Y., Feng, T., Rao, H., 2015. Resting spontaneous activity in the default mode network predicts performance decline during prolonged attention workload. NeuroImage. doi:10.1016/j.neuroimage.2015.07.030
687 688
Gusnard, D.A., Raichle, M.E., 2001. Searching for a baseline: Functional imaging and the resting human brain. Nat. Rev. Neurosci. 2, 685–694. doi:10.1038/35094500
689 690 691
Hampstead, B.M., Brown, G.S., Hartley, J.F., 2014. Transcranial Direct Current Stimulation Modulates Activation and Effective Connectivity During Spatial Navigation. Brain Stimulat. 7, 314–324. doi:10.1016/j.brs.2013.12.006
Page 29 of 38
692 693
Hayden, B.Y., Smith, D.V., Platt, M.L., 2010. Cognitive control signals in posterior cingulate cortex. Front. Hum. Neurosci. 4, 223. doi:10.3389/fnhum.2010.00223
694 695 696
Hayden, B.Y., Smith, D.V., Platt, M.L., 2009. Electrophysiological correlates of default-mode processing in macaque posterior cingulate cortex. Proc. Natl. Acad. Sci. 106, 5948–5953. doi:10.1073/pnas.0812035106
697 698 699
Heilbronner, S.R., Platt, M.L., 2013. Causal Evidence of Performance Monitoring by Neurons in Posterior Cingulate Cortex during Learning. Neuron 80, 1384–1391. doi:10.1016/j.neuron.2013.09.028
700 701
Hyvärinen, A., 1999. Fast and robust fixed-point algorithms for independent component analysis. IEEE Trans. Neural Netw. 10, 626–34. doi:10.1109/72.761722
702 703 704 705
Insel, T., Cuthbert, B., Garvey, M., Heinssen, R., Pine, D.S., Quinn, K., Sanislow, C., Wang, P., 2010. Research domain criteria (RDoC): Toward a new classification framework for research on mental disorders. Am. J. Psychiatry 167, 748–751. doi:10.1176/appi.ajp.2010.09091379
706 707
Jenkinson, M., Smith, S., 2001. A global optimisation method for robust affine registration of brain images. Med. Image Anal. 5, 143–56.
708 709
Kanwisher, N., McDermott, J., Chun, M.M., 1997. The fusiform face area: A module in human extrastriate cortex specialized for face perception. J. Neurosci. 17, 4302–4311.
710 711 712
Kanwisher, N., Yovel, G., 2006. The fusiform face area: a cortical region specialized for the perception of faces. Philos. Trans. R. Soc. B Biol. Sci. 361, 2109–2128. doi:10.1098/rstb.2006.1934
713 714 715 716
Keeser, D., Meindl, T., Bor, J., Palm, U., Pogarell, O., Mulert, C., Brunelin, J., Moller, H.-J., Reiser, M., Padberg, F., 2011. Prefrontal Transcranial Direct Current Stimulation Changes Connectivity of Resting-State Networks during fMRI. J. Neurosci. 31, 15284–15293. doi:10.1523/JNEUROSCI.0542-11.2011
717 718 719
Kelly, A., Uddin, L.Q., Biswal, B.B., Castellanos, F.X., Milham, M.P., 2008. Competition between functional brain networks mediates behavioral variability. NeuroImage 39, 527– 37. doi:10.1016/j.neuroimage.2007.08.008
720 721
Knutson, B., Adams, C.M., Fong, G.W., Hommer, D., 2001. Anticipation of Increasing Monetary Reward Selectively Recruits Nucleus Accumbens. J. Neurosci. 21, RC159-RC159.
722 723 724
Knutson, B., Westdorp, A., Kaiser, E., Hommer, D., 2000. FMRI visualization of brain activity during a monetary incentive delay task. NeuroImage 12, 20–7. doi:10.1006/nimg.2000.0593
725 726
Kucyi, A., Moayedi, M., Weissman-Fogel, I., Goldberg, M.B., Freeman, B.V., Tenenbaum, H.C., Davis, K.D., 2014. Enhanced Medial Prefrontal-Default Mode Network Functional
Page 30 of 38
727 728
Connectivity in Chronic Pain and Its Association with Pain Rumination. J. Neurosci. 34, 3969–3975. doi:10.1523/JNEUROSCI.5055-13.2014
729 730 731
Leech, R., Kamourieh, S., Beckmann, C.F., Sharp, D.J., 2011. Fractionating the default mode network: distinct contributions of the ventral and dorsal posterior cingulate cortex to cognitive control. J. Neurosci. 31, 3217–3224. doi:10.1523/JNEUROSCI.5626-10.2011
732 733 734
Luber, B., Lisanby, S.H., 2014. Enhancement of human cognitive performance using transcranial magnetic stimulation (TMS). NeuroImage 85, Part 3, 961–970. doi:10.1016/j.neuroimage.2013.06.007
735 736
McCarthy, G., Puce, A., Gore, J.C., Allison, T., 1997. Face-Specific Processing in the Human Fusiform Gyrus. J. Cogn. Neurosci. 9, 605–610. doi:10.1162/jocn.1997.9.5.605
737 738 739
McLaren, D.G., Ries, M.L., Xu, G., Johnson, S.C., 2012. A generalized form of context-dependent psychophysiological interactions (gPPI): A comparison to standard approaches. NeuroImage 61, 1277–1286. doi:10.1016/j.neuroimage.2012.03.068
740 741 742 743
Mueller, J.K., Grigsby, E.M., Prevosto, V., Petraglia Iii, F.W., Rao, H., Deng, Z.-D., Peterchev, A.V., Sommer, M.A., Egner, T., Platt, M.L., Grill, W.M., 2014. Simultaneous transcranial magnetic stimulation and single-neuron recording in alert non-human primates. Nat. Neurosci. advance online publication. doi:10.1038/nn.3751
744 745
Nichols, T.E., Holmes, A.P., 2002. Nonparametric permutation tests for functional neuroimaging: a primer with examples. Hum. Brain Mapp. 15, 1–25.
746 747 748
O’Reilly, J.X., Woolrich, M.W., Behrens, T.E., Smith, S.M., Johansen-Berg, H., 2012. Tools of the trade: psychophysiological interactions and functional connectivity. Soc. Cogn. Affect. Neurosci. 7, 604–9. doi:10.1093/scan/nss055
749 750 751 752
Ossandon, T., Jerbi, K., Vidal, J.R., Bayle, D.J., Henaff, M.-A., Jung, J., Minotti, L., Bertrand, O., Kahane, P., Lachaux, J.-P., 2011. Transient Suppression of Broadband Gamma Power in the Default-Mode Network Is Correlated with Task Complexity and Subject Performance. J. Neurosci. 31, 14521–14530. doi:10.1523/JNEUROSCI.2483-11.2011
753 754 755 756
Piccoli, T., Valente, G., Linden, D.E.J., Re, M., Esposito, F., Sack, A.T., Salle, F.D., 2015. The Default Mode Network and the Working Memory Network Are Not Anti-Correlated during All Phases of a Working Memory Task. PLoS ONE 10, e0123354. doi:10.1371/journal.pone.0123354
757 758 759
Ray, K.L., McKay, D.R., Fox, P.M., Riedel, M.C., Uecker, A.M., Beckmann, C.F., Smith, S.M., Fox, P.T., Laird, A.R., 2013. ICA model order selection of task co-activation networks. Front. Neurosci. 7. doi:10.3389/fnins.2013.00237
760 761
Rorden, C., Karnath, H.-O., Bonilha, L., 2007. Improving Lesion-Symptom Mapping. J. Cogn. Neurosci. 19, 1081–1088. doi:10.1162/jocn.2007.19.7.1081
Page 31 of 38
762 763 764
Rosen, M.L., Stern, C.E., Michalka, S.W., Devaney, K.J., Somers, D.C., 2015. Cognitive Control Network Contributions to Memory-Guided Visual Attention. Cereb. Cortex bhv028. doi:10.1093/cercor/bhv028
765 766 767
Sambataro, F., Murty, V.P., Callicott, J.H., Tan, H.-Y., Das, S., Weinberger, D.R., Mattay, V.S., 2010. Age-related alterations in default mode network: Impact on working memory performance. Neurobiol. Aging 31, 839–852. doi:10.1016/j.neurobiolaging.2008.05.022
768 769 770 771
Seeley, W.W., Menon, V., Schatzberg, A.F., Keller, J., Glover, G.H., Kenna, H., Reiss, A.L., Greicius, M.D., 2007. Dissociable intrinsic connectivity networks for salience processing and executive control. J. Neurosci. 27, 2349–2356. doi:10.1523/JNEUROSCI.558706.2007
772 773 774 775
Shannon, B.J., Raichle, M.E., Snyder, A.Z., Fair, D.A., Mills, K.L., Zhang, D., Bache, K., Calhoun, V.D., Nigg, J.T., Nagel, B.J., Stevens, A.A., Kiehl, K.A., 2011. Premotor functional connectivity predicts impulsivity in juvenile offenders. Proc. Natl. Acad. Sci. U. S. A. 108, 11241–11245. doi:10.1073/pnas.1108241108
776 777 778 779
Sheline, Y.I., Barch, D.M., Price, J.L., Rundle, M.M., Vaishnavi, S.N., Snyder, A.Z., Mintun, M.A., Wang, S., Coalson, R.S., Raichle, M.E., 2009. The default mode network and selfreferential processes in depression. Proc. Natl. Acad. Sci. 106, 1942–7. doi:10.1073/pnas.0812686106
780 781 782
Sladky, R., Friston, K.J., Tröstl, J., Cunnington, R., Moser, E., Windischberger, C., 2011. Slicetiming effects and their correction in functional MRI. NeuroImage 58, 588–94. doi:10.1016/j.neuroimage.2011.06.078
783 784 785
Smith, D.V., Clithero, J.A., Boltuck, S.E., Huettel, S.A., 2014a. Functional connectivity with ventromedial prefrontal cortex reflects subjective value for social rewards. Soc. Cogn. Affect. Neurosci. 9, 2017–2025. doi:10.1093/scan/nsu005
786 787 788
Smith, D.V., Delgado, M.R., 2017. Meta-analysis of psychophysiological interactions: Revisiting cluster-level thresholding and sample sizes. Hum. Brain Mapp. n/a-n/a. doi:10.1002/hbm.23354
789 790 791
Smith, D.V., Gseir, M., Speer, M.E., Delgado, M.R., 2016. Toward a cumulative science of functional integration: A meta-analysis of psychophysiological interactions. Hum. Brain Mapp.
792 793 794
Smith, D.V., Hayden, B.Y., Truong, T.-K., Song, A.W., Platt, M.L., Huettel, S.A., 2010. Distinct value signals in anterior and posterior ventromedial prefrontal cortex. J. Neurosci. 30, 2490–5. doi:10.1523/JNEUROSCI.3319-09.2010
795 796 797 798
Smith, D.V., Utevsky, A.V., Bland, A.R., Clement, N., Clithero, J.A., Harsch, A.E.W., McKell Carter, R., Huettel, S.A., 2014b. Characterizing individual differences in functional connectivity using dual-regression and seed-based approaches. NeuroImage 95, 1–12. doi:10.1016/j.neuroimage.2014.03.042
Page 32 of 38
799 800
Smith, S.M., 2002. Fast robust automated brain extraction. Hum. Brain Mapp. 17, 143–55. doi:10.1002/hbm.10062
801 802 803 804 805 806
Smith, S.M., Beckmann, C.F., Andersson, J., Auerbach, E.J., Bijsterbosch, J., Douaud, G., Duff, E., Feinberg, D.A., Griffanti, L., Harms, M.P., Kelly, M., Laumann, T., Miller, K.L., Moeller, S., Petersen, S., Power, J., Salimi-Khorshidi, G., Snyder, A.Z., Vu, A.T., Woolrich, M.W., Xu, J., Yacoub, E., Uğurbil, K., Van Essen, D.C., Glasser, M.F., 2013. Resting-state fMRI in the Human Connectome Project. NeuroImage, Mapping the Connectome 80, 144–168. doi:10.1016/j.neuroimage.2013.05.039
807 808 809 810
Smith, S.M., Fox, P.T., Miller, K.L., Glahn, D.C., Fox, P.M., Mackay, C.E., Filippini, N., Watkins, K.E., Toro, R., Laird, A.R., Beckmann, C.F., 2009. Correspondence of the brain’s functional architecture during activation and rest. Proc. Natl. Acad. Sci. 106, 13040–5. doi:10.1073/pnas.0905267106
811 812 813 814 815
Smith, S.M., Jenkinson, M., Woolrich, M.W., Beckmann, C.F., Behrens, T.E., Johansen-Berg, H., Bannister, P.R., De Luca, M., Drobnjak, I., Flitney, D.E., Niazy, R.K., Saunders, J., Vickers, J., Zhang, Y., De Stefano, N., Brady, J.M., Matthews, P.M., 2004. Advances in functional and structural MR image analysis and implementation as FSL. NeuroImage 23 Suppl 1, S208–19. doi:10.1016/j.neuroimage.2004.07.051
816 817 818
Stern, E.R., Fitzgerald, K.D., Welsh, R.C., Abelson, J.L., Taylor, S.F., 2012. Resting-State Functional Connectivity between Fronto-Parietal and Default Mode Networks in Obsessive-Compulsive Disorder. PLoS ONE 7, e36356. doi:10.1371/journal.pone.0036356
819 820 821 822
Stern, E.R., Welsh, R.C., Fitzgerald, K.D., Gehring, W.J., Lister, J.J., Himle, J.A., Abelson, J.L., Taylor, S.F., 2011. Hyperactive Error Responses and Altered Connectivity in Ventromedial and Frontoinsular Cortices in Obsessive-Compulsive Disorder. Biol. Psychiatry, Genes and Anxiety 69, 583–591. doi:10.1016/j.biopsych.2010.09.048
823 824
Utevsky, A.V., Smith, D.V., Huettel, S.A., 2014. Precuneus is a functional core of the defaultmode network. J. Neurosci. 34, 932–940. doi:10.1523/JNEUROSCI.4227-13.2014
825 826
Weissman, D.H., Roberts, K.C., Visscher, K.M., Woldorff, M.G., 2006. The neural bases of momentary lapses in attention. Nat. Neurosci. 9, 971–978. doi:10.1038/nn1727
827 828 829
Winkler, A.M., Ridgway, G.R., Webster, M.A., Smith, S.M., Nichols, T.E., 2014. Permutation inference for the general linear model. NeuroImage 92, 381–397. doi:10.1016/j.neuroimage.2014.01.060
830 831 832
Woolrich, M.W., Jbabdi, S., Patenaude, B., Chappell, M., Makni, S., Behrens, T.E., Beckmann, C.F., Jenkinson, M., Smith, S.M., 2009. Bayesian analysis of neuroimaging data in FSL. NeuroImage 45, S173–86. doi:10.1016/j.neuroimage.2008.10.055
833 834 835
Yarkoni, T., Poldrack, R.A., Nichols, T.E., Van Essen, D.C., Wager, T.D., 2011. Large-scale automated synthesis of human functional neuroimaging data. Nat. Methods 8, 665–670. doi:10.1038/nmeth.1635
Page 33 of 38
836 837 838
Young, J.S., Smith, D.V., Coutlee, C.G., Huettel, S.A., 2015. Synchrony between sensory and cognitive networks is associated with subclinical variation in autistic traits. Front. Hum. Neurosci. 146. doi:10.3389/fnhum.2015.00146
839 840
Page 34 of 38
841
Figure Captions
842
Figure 1: Task Design
843
Fifty heterosexual men performed a modified MID task which examined timed responses to
844
view social and nonsocial images as rewards (faces and landscapes, respectively). Each trial
845
of the task began with an abstract image presented as a cue for 1500 ms indicating the
846
potential reward for that trial. Following the cue, a fixation cross appeared for a variable
847
duration (interstimulus interval; 1500 – 4500 ms), followed by a target (a white triangle)
848
that appeared on the screen for 750 ms; participants responded by pressing a button box
849
with their right index finger before the target disappeared. After the target, a feedback image
850
(a colored triangle) appeared for 750 ms; the feedback image was colored green if the
851
participants’ response was sufficiently fast on the trial (a win), and was blue if the response
852
was too slow (a loss). After the feedback image disappeared, the participants waited a delay
853
period that varied according to the cue and whether the subject won or lost that trial. Lastly,
854
the reward image (either a face or a landscape) was shown for 2000 ms. Consent was
855
provided for the use of the image in this figure.
856
Figure 2: Response Times are Slower Following Face Trials Compared to Landscape
857
Trials
858
(A) Distribution of response time (RT) differences according to previous trial’s stimulus
859
type, calculated by subtracting RTs following nonsocial trials from RTs following social trials.
860
Response times on trials following social rewards were greater than those on trials following
861
nonsocial rewards, indicating an effect of previous reward stimulus type on subsequent
862
behavior. (B) Average beta weights (with S.E.M. plotted) across subjects from a behavioral Page 35 of 38
863
regression predicting current RT. We regressed current trial RT on a model including the
864
following regressors: stimulus category (social or nonsocial) on previous trial, RT on
865
previous trial, attractiveness rating of the previous trial’s reward, and stimulus category on
866
current trial. Of these four regressors, current stimulus category most strongly predicted
867
current RT; the negative beta weight indicates that participants are faster to respond during
868
a social trial compared to during a nonsocial trial. The next most predictive regressor was
869
previous trial’s stimulus category; the positive beta weight indicates that participants were
870
slower to respond following social trials compared to following nonsocial trials. Neither
871
previous trial’s RT nor previous trial’s reward attractiveness significantly predicted the
872
current trial’s RT.
873
Figure 3: Networks Identified from Independent Components Analysis
874
Using independent components analysis across our task runs, we identified large-scale
875
functional networks. This analysis produced 25 components. Our analyses focused on the
876
components that best matched default-mode network (highest-correlating component
877
illustrated on top) and the executive control network (highest-correlating component
878
illustrated on bottom) from Smith and colleagues (2009). For visualization purposes, maps
879
are thresholded at Z > 4.
880
Figure 4: Fusiform Gyrus Connectivity Tracks Effect of Stimulus Type on Subsequent
881
Behavior
882
(A) A network PPI indicated that an area in the fusiform gyrus (FG) exhibits heightened
883
effective connectivity with ECN (compared to DMN) during social reward outcomes;
884
thresholded at p < 0.05. We note that this result also held with permutation-based testing
Page 36 of 38
885
(Eklund et al., 2016). (B) Parameters estimates extracted from the FG connectivity track the
886
effect of stimulus type on subsequent behavior (stimulus-type beta weights estimated from
887
our behavioral GLM; see Behavioral Analysis): as FG-DMN connectivity increases relative to
888
FG-ECN connectivity, RTs are further slowed following social trials compared to following
889
nonsocial trials.
890
Page 37 of 38
891
Table Captions
892
Table 1. Behavioral Regression Results
893
To examine the effects of other trial characteristics on current trial RT, we regressed current
894
trial RT on a model including: stimulus category (social or nonsocial) on previous trial, RT on
895
previous trial, attractiveness rating of the previous trial’s reward, and stimulus category on
896
current trial. Our analysis indicated that current stimulus type had the strongest effect on
897
current RT; however, of all the characteristics from the prior trial, only prior stimulus type
898
had a significant effect on current RT. Regressor Current trial (t) stimulus type Prior trial (t-1) stimulus type Prior trial (t-1) attractiveness Prior trial (t-1) response time
Parameter Estimate (s.e.m) -0.055 (0.013) 0.033 (0.013) 0.019 (0.011) < 0.0001 (0.013)
t-stat -4.31 2.46 1.74 0.005
p-value < 0.0001 0.01 0.09 0.99
899
Table 2. Spatial Correspondence with Canonical Networks
900
To identify the DMN and ECN, we ran spatial correlations between canonical neural
901
networks (Smith et al., 2009) and the 25 components from our ICA. The highest-correlating
902
ICA component numbers for each network map and the correlation values are listed. Canonical Network Visual 1 Visual 2 Visual 3 Default-Mode Cerebellar Sensorimotor Auditory Executive Control R Frontoparietal L Frontoparietal
Independent Component No. IC10 IC01 IC01 IC06 IC18 IC13 IC14 IC04 IC03 IC07
Spatial Correlation (r) 0.82 0.66 0.45 0.78 0.32 0.59 0.66 0.64 0.64 0.77
903
Page 38 of 38
Outcome
Cue
ISI
Target
Feedback
+ 1500ms
1500-4500ms
Delay
+ 750ms
750ms
ITI 2000ms
1500-4500ms
5000-11000ms
2000ms
A
8 Faster following face trials
Number of Participants
7
Faster following landscape trials
6 5 4 3 2 1 0 -0.03
-0.02
-0.01
0
0.01
0.02
0.03
0.04
0.05
Face-Land RT Difference (s)
B
*
Previous trial attractiveness Previous trial stimulus type
0.04
Previous trial reaction time
Parameter Estimates (β)
Current trial stimulus type
0.02
0
-0.02
-0.04
-0.06 ** p < 0.001
* p < 0.05
**
Z = 21
Y = -61
X=0
Default-Mode Network
Z = 12
Y = 27
Executive Control Network
X=0
Z = -17 Network Coupling with FG (DMN - ECN PPI β)
Increased FG-ECN coupling
B
Increased FG-DMN coupling
A
Y = -65
X = 42
15
10
5
0
-5
-10 -0.1
0
0.1
0.2
Change ininBehavior (Face-Scene RT Effect β)β) Change Behavior (Current Stimulus
Faster after social trials
Slower after social trials
