Articles in PresS. J Neurophysiol (March 12, 2014). doi:10.1152/jn.00308.2013
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Decoupling motor plans from perceptual decisions
to investigate when and whether decisions are embodied
Neuro Forum to Filimon et al. on
The Journal of Neuroscience, Jan 30; 33(5): 2121- 2136.
Stefano Sandrone
1 NATBRAINLAB - Neuroanatomy and Tractography Brain Laboratory, Department
of Forensic and Neurodevelopmental Sciences, Institute of Psychiatry, King’s
College, SE5 8AF London, UK.
15
Total number of words (text):
2455.
16
Total number of references:
17.
*Address correspondence to:
Stefano Sandrone
NATBRAINLAB
Institute of Psychiatry
16 De Crepsigny Park
London SE5 8AF
United Kingdom
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
Email: [email protected]
Running Title:
Motor plans and perceptual decisions
Keywords:
Decision-making
Motor plans
Perceptual decisions
Embodied cognition
Acknowledgments: Many thanks to Richard Hahnloser, David Bontrager and
Henrietta Howells for their comments on a previous version of the manuscript.
Decision-making is a crucial part of our life: we sense information from the
Copyright © 2014 by the American Physiological Society.
43
environment and perform our motor response. However, ‘whether’ and ‘when’
44
decisions are embodied still needs to be fully elucidated. Neuroimaging data obtained
45
by the disentanglement of perceptual decision from motor preparation revealed an
46
increase in connectivity between inferior frontal cortex and sensory regions, and the
47
important role played by intraparietal sulcus in motor decisions. The results obtained
48
as well as the new research questions prompted by this work are carefully discussed
49
herein.
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
Every day, from the moment the alarm clock sounds (and we are of two minds
73
about getting out of the bed or postponing for a couple of minutes) until night
74
when we choose whether to set it again or not, we make a huge number of
75
decisions, with a plethora of deeply intermingled cognitive, emotional, motor,
76
spatial, and temporal aspects. In a paper recently published in The Journal of
77
Neuroscience, Filimon et al. addressed the challenging topic of decision-making
78
using event-related functional magnetic resonance imaging (fMRI) and a well-
79
designed experimental strategy (Filimon et al., 2013). They disentangled motor
80
preparation from perceptual decision-making to investigate whether the
81
sensorimotor system implements perceptual choices.
82
Previous research on this topic had been confounded by motor planning due to
83
pre-assigned sensorimotor mappings (Shapiro, 2011), leaving the question open
84
of whether and when these decisions are embodied in brain circuits. Preceding
85
neurophysiological studies on monkeys postulated the existence of a “decisional
86
threshold” that can be reached by the accumulation of sensory information, with
87
greater firing rates for perceptual decisions based on high sensory evidence, than
88
those based on low evidence. However, increased firing patterns in frontal eye
89
fields (FEF) and the lateral intraparietal area (LIP) (see for example Kim and
90
Shadlen, 1999) are misleading, since they could be correlated to motor plan
91
preparation rather than to perceptual decisions or sensory integration. Evidence
92
of the fact that prior expectation can modulate the interaction between sensory
93
and prefrontal regions in the human brain has already been found (Rahnev et
94
al., 2011), but the contribution of effector-specific sensorimotor regions to
95
perceptual decisions is not so clear, given that effector-nonspecific areas also
96
contribute to eye and hand movement planning. This scenario is further
97
complicated by the fact that the human decisional threshold acts in multiple
98
ways, e.g. with sustained activity vs. non-sustained activity modalities.
99
In the light of that, Filimon et al. avoided a priori assumptions on the decisional
100
threshold. From an epistemological point of view, this bottom-up experimental
101
approach is probably the best solution when different theoretical frameworks
102
are available and none of which can be identified as ultimately correct. They
103
analysed the increase in effective connectivity between sensory evidence areas
104
and higher-level regions. In the process, they revealed target locations and the
105
effector type to the experimental subjects only after the decision stage. After a
106
training period, nineteen people undertook a task in which they had to decide
107
repeatedly whether greyscale noisy stimuli represented a face or a house,
108
individually adjusted for each subject and presenting high or low levels of
109
sensory evidence (i.e. with low or high levels of noise). All stimulus images had
110
the same spatial frequency, contrast, and luminance. After a variable delay of up
111
to 9 seconds, subjects were given a specific motor preparation instruction on how
112
to indicate their decision. Starting from a fixation point, they were given an
113
additional 1.5 seconds in which to respond either by saccading to the
114
remembered target location (up, down, left or right) or by pressing the
115
corresponding button on a diamond-shaped response pad (hand trials). Eight
116
possible motor plans were thus decoupled from perceptual decisions. During
117
each motor preparation period, a star or a square, respectively representing a
118
face or a house, appeared as a target at one of the aforementioned locations, and
119
subjects responded without receiving any feedback.
120
The first finding of the experiment was that the subjects recognised faces and
121
houses with high levels of accuracy, and even more so for increased sensory
122
evidence, as witnessed by behavioural analysis and cerebral activations for
123
presented stimuli. Faces activated face dominant regions in occipital and ventral
124
temporal cortices significantly more than did houses, which in turn mainly
125
activated the parahippocampal gyrus, thus confirming previous reports
126
(Kanwisher and Yovel, 2006). However, the variability among subjects on
127
cerebral shape and size and individual differences in ear canals as well as in
128
lateral ventricles produced different field distortion in ventral temporal cortex in
129
each subject. Therefore, registration of individual run to a standard space was
130
imprecise, and could to some extent smear out the activations in the fusiform and
131
parahippocampal gyri. Because of that, regions of interest were selected at the
132
single-subject level. Regardless of the sensory evidence level, all stimuli led to
133
widespread BOLD activations that involved, inter alia, the prefrontal cortex,
134
FEF, the orbitofrontal region, intraparietal and parietal areas, the precuneus,
135
the cingulate cortex, and subcortical structures. These activations probably
136
reflected other mental processes and factors involved in this paradigm, such as
137
attention, memory, difficulties and other inter-individual differences.
138
The second finding was that psychophysiological interaction (PPI) analysis
139
allowed the authors to identify the specific decision-making network and thus
140
satisfied a requirement for increased effective connectivity with faces and house
141
areas during the perceptual decisions. The combination of a PPI analysis with
142
the comparison of the total amount of BOLD activation for High- versus Low-
143
evidence conditions could be viewed as the fMRI equivalent of the postulated
144
sensory evidence comparator process and the accumulation to threshold (see also
145
Gold and Shadlen, 2007). The authors hypothesised and empirically confirmed
146
that, during the phase of sensory evidence integration, perceptual decision-
147
making regions should systematically display higher connectivity with the
148
absolute difference between time series from face- and house-dominant sensory
149
regions over the perceptual decision period than during the rest of the trial,
150
when motor planning is underway.
151
However, even with this approach, the aforementioned factors could still account
152
for an increase in connectivity. The authors accordingly set three further
153
stringent criteria: (i) that perceptual areas involved in sensory integration should
154
also be modulated by the amount of sensory evidence available during the
155
perceptual decision; (ii) that said areas should show greater activation during the
156
decision stage than the post-decision stage, and (iii) that they should be activated
157
above baseline. At the face/house decision stage, the effective connectivity
158
increased substantially in several regions, but the inferior frontal sulcus (IFS)
159
was the only region to meet the established criteria for sensory integration. The
160
IFS responded to the sensory evidence independently of effector-specific motor
161
preparation and reached a perceptual decision in the absence of a motor plan,
162
without the need for the decision to be embodied. Motor plans cannot be formed
163
in putative LIP. Although sensory evidence modulation over the perceptual
164
decision in the parietal PPI areas was predictable (and possibly too weak to be
165
detected with fMRI techniques), such modulation was recorded in prefrontal but
166
not in parietal cortices, whose neurons enshrined a finer representation of very
167
specific stimuli (Quiroga, 2012). A re-analysis of the data on the basis of this
168
specific point confirmed previous reports, therefore revealing a greater motor
169
preparatory activity for oculomotor decisions with a greater putative LIP
170
High>Low modulation (Tosoni et al., 2008). This motor IPS region, interestingly
171
also not overlapping with the PPI areas when a less stringent liberal threshold
172
was applied, was indeed involved in motor and eye movement decisions between
173
possible targets to be reached with a saccade, probably through its connectivity
174
with middle temporal regions. The final observation of this study was that of
175
higher IFS activation for low- than for high-evidence decisions. In contrast, LIP
176
displayed a High>Low pattern during motor decisions, which suggested for the
177
first time that different regions and types of decisions have different neural
178
signatures.
179
These results are striking, and some interesting and stimulating issues can be
180
raised. We are still far from having ultimately mapped the LIP areas in monkeys
181
and the equivalent ones in humans, and future investigations will have to fill this
182
conceptual
183
(cytoarchitectonic) and functional (connectivity) perspectives. Also, modifying
184
the paradigm so as to increase the number of the training trials would have
185
made this study even more analogous to those performed in monkeys, allowing
186
more meaningful comparison. Given that the experimental paradigm requires a
187
saccadic movement, the authors correctly decided to use an eye-tracker inside
188
the fMRI. This tool is very useful when used to quantify the movements of the
189
subjects before the final decision, to measure the point of gaze and the motion of
190
an eye related to the head and to analyse whether specific eye movements are
191
linked to specific (i.e. correct vs incorrect) decisional outcomes.
192
Thanks to the study by Filimon and colleagues, we have learnt that abstract
193
decisions in the absence of sensorimotor mappings are not embodied, but that
194
sensorimotor networks do contribute to motor decisions. Interestingly, some
195
variants of the experimental strategy used by Filimon et al. could be really useful
196
to gain further insights. The first of them could rely on the exploitation of visual
197
lexical choice. Only tangentially discussed by the authors, a psycholinguistic
198
version of their paradigm would enable modulation of the ambiguity inherent to
199
the material presented, and it would facilitate a focus on the perceptual decision
200
that involves the dichotomy between abstract and concrete (more embodied)
201
stimuli. With these modifications, in fact, it would be possible to assess the
202
contrast between eye-, hand- and also mouth-movements (Cappa and
203
Pulvermüller, 2012). Also, the use of multi-modal and cross-sensorial responses
gap
by
properly
addressing
it
from
both
structural
204
(e.g. audio-visual object responses) could expand the range of perceptual
205
decisions here considered.
206
Crucial insights could be also gained by modifying the experimental design with
207
the introduction of new factors as regressors. For example, the modulation of the
208
response movement can be achieved by means of pointing or bimanual responses
209
(Madlon-Kay et al., 2013) rather than of the less demanding button pressing.
210
Conversely, exclusion of the high- and low-response targets could reduce the
211
combinations of motor plans, and thus simplify the task.
212
Designing a well-controlled experiment where scientists can properly manipulate
213
individual variables is a fundamental methodological issue, and Filimon and
214
colleagues have elegantly done this. At the same time, however, we live in a
215
dynamic and lively world where static stimuli are extremely rare. Therefore, one
216
of the next challenging steps could be the use of a visual motion discrimination
217
task instead of recognition, which could further reduce the gap between the
218
experimental setting and our daily decision-making process. Moreover,
219
displaying coloured stimuli, in spite of a slight difficulty in individual
220
adjustment, would make the task even more ecological.
221
There are many different kinds of decisions, and to obtain a complete picture we
222
need to insert processes that are pivotal in our everyday decision-making, such
223
as punishment and reward. In the original paradigm used by Filimon and
224
colleagues, subjects responded without receiving any positive or negative
225
feedback. On one hand, the punishment level for incorrect responses could have
226
been parametrically manipulated across blocks of trials (as in Blank et al., 2013).
227
On the other hand, a tiny modification in the paradigm (i.e. visual signal on the
228
screen or auditory pleasant stimulus when the correct answer is given) can allow
229
the researchers to gain insights from the influence of positive rewards on the
230
decisional outcome. These results provide a mechanism of how prior
231
expectations may affect perceptual decision making, namely by changing neural
232
activity in, and sensory drive to, prefrontal areas.
233
Given the neural differences between young adults and elder people in decisional
234
processing (Crone and Ridderinkhof, 2011) and the relatively low age of the
235
experimental subjects tested by Filimon and colleagues (24.9 years, and a rather
236
dispersed range of 20 to 32 years), these results need to be replicated in an older,
237
bigger, and more clustered representative population sample. It would also be
238
productive to extend this study to adolescents, and thus to intercept motor plans
239
and perceptual decision throughout their evolutive trajectories (Koolschijn et al.,
240
2011).
241
Considering the social nature of decision-making, the introduction of a trial-by-
242
trial feedback to the experimental subjects could also be used to model real or
243
fictitious social interactions by making the subject consider the presence of a
244
potential competitor who plays the same paradigm, but with decisional outcomes
245
that should be independent for every subject. This further modification coupled
246
with the experimental timing used by Filimon et al. would also be useful to
247
monitor the performance of the subjects, and it will result in an increase in
248
response accuracy, mirroring an increase in their motivation aimed at making
249
the right decision trial by trial, and intercepting the social side of decisional
250
processes.
251
In real life, we are supposed to make decisions quickly and, quite often, in a
252
stressful social environment, different from the non-stressful experimental
253
timing context used by Filimon et al. Therefore, in order to further complete this
254
scenario, future research on decision-making will also have to take advantage of
255
the recent findings from the emerging research field of psychoneuroimmunology
256
(see Eisenberger, 2012). In this case, to further study the link between social
257
relationships and physical health while performing decision-making tasks, the
258
level of pro inflammatory cytokines IL-6 and the soluble receptor for tumor
259
necrosis factor-α (sTNFαRII) that are usually increased in competitive social
260
interactions (Chiang et al., 2012) can be assessed. Finally, this level of analysis
261
should be coupled with a more specific investigation of the oscillatory activity in
262
the motor cortex at different frequency ranges during the decision-making
263
process, in order to further exploit the dynamics of perceptual choice, and also
264
those relating to expectation, a crucial component in our daily decisions (de
265
Lange et al., 2013).
266
The accomplishment of a very detailed neurophysiological understanding of the
267
decision-making process is possible only with a stimulating convergence of
268
several levels of studies, with multiple experimental strategies linked to an
269
interplay between studies on monkeys and on humans as well as a tight coupling
270
of hypothesis-driven and bottom-up approaches. With the aim of depicting new
271
theories of decision- making, future research will not only have to record
272
neurophysiological correlates of decision-making and the interactions between
273
cerebral areas, but also the modulations underpinning neural circuitry and
274
oscillatory activity with magnetoencephalography (MEG) and transcranial
275
stimulation tools (i.e. TMS), putting hypotheses on functional brain architecture
276
models, including the brain deep nuclei (Ding and Gold, 2013), to the test.
277
In conclusion, real life decisions mix perceptual, social, and reward-based
278
decision processes. Remarkable works such as this by Filimon and colleagues,
279
through carefully studying specific elements of the bigger picture, are
280
incrementally extending the boundaries of our knowledge in this area. To
281
properly intercept the complexity of this topic, we need to multiply the
282
perspectives from which we can study decision-making, before integrating all the
283
newly acquired pieces of evidence into novel theories on this subject. Intriguing
284
research avenues stand in front of us: it is time for neuroscientists to make
285
challenging decisions on decision-making.
286
287
288
289
290
291
292
References
293
294
Blank H, Biele G, Heekeren HR, Philiastides MG. Temporal characteristics of
295
the influence of punishment on perceptual decision making in the human brain.
296
J Neurosci 33: 3939–3952, 2013.
297
298
Cappa SF, Pulvermüller F. Cortex special issue: language and the motor system.
299
Cortex 48: 785–787, 2012.
300
301
Chiang JJ, Eisenberger NI, Seeman TE, Taylor SE. Negative and competitive
302
social interactions are related to heightened proinflammatory cytokine activity.
303
Proc Natl Acad Sci U S A 109: 1878-82, 2012.
304
305
Crone EA, Ridderinkhof
KR. The developing brain: from theory to
306
neuroimaging and back. Dev Cogn Neurosci 1: 101–109, 2011.
307
308
de Lange FP, Rahnev DA, Donner TH, Lau H. Prestimulus oscillatory activity
309
over motor cortex reflects perceptual expectations. J Neurosci 33: 1400–1410,
310
2013.
311
312
Ding L, Gold JI. The basal ganglia's contributions to perceptual decision
313
making. Neuron 79: 640-649.
314
315
Eisenberger NI. The pain of social disconnection: examining the shared neural
316
underpinnings of physical and social pain. Nat Rev Neurosci 13: 421-434, 2012.
317
318
Gold JI, Shadlen MN. The neural basis of decision making. Annu Rev Neurosci
319
30: 535-574, 2007.
320
321
Filimon F, Philiastides MG, Nelson JD, Kloosterman NA, Heekeren HR. How
322
embodied is perceptual decision making? Evidence for separate processing of
323
perceptual and motor decisions. J Neurosci 33:2121–2136, 2013.
324
325
Kanwisher N, Yovel G. The fusiform face area: a cortical region specialized for
326
the perception of faces. Philos Trans R Soc Lond B Biol Sci 361: 2109–2128,
327
2006.
328
329
Kim JN, Shadlen MN. Neural correlates of a decision in the dorsolateral
330
prefrontal cortex of the macaque. Nat Neurosci 2: 176–185, 1999.
331
332
Koolschijn PC, Schel MA, de Rooij M, Rombouts SA, Crone EA. A three-year
333
longitudinal functional magnetic resonance imaging study of performance
334
monitoring and test-retest reliability from childhood to early adulthood. J
335
Neurosci 31:4204–12, 2011.
336
337
Madlon-Kay S, Pesaran B, Daw ND. Action selection in multi-effector decision
338
making. Neuroimage 70: 66–79, 2013.
339
340
Quiroga RQ. Concept cells: the building blocks of declarative memory functions.
341
Nat Rev Neurosci 13: 587-597, 2012.
342
343
Rahnev D, Lau H, de Lange FP. Prior expectation modulates the interaction
344
between sensory and prefrontal regions in the human brain. J Neurosci 31:
345
10741-10748, 2011.
346
347
Shapiro, LA. Embodied cognition. New problems of philosophy. New York:
348
Routledge, 2011.
349
350
Tosoni A, Galati G, Romani GL, Corbetta M. Sensory-motor mechanisms in
351
human parietal cortex underlie arbitrary visual decisions. Nat Neurosci 11:
352
1446–1453, 2008.
353
354
355