5D Library › Interoceptive Decoupling, Salience-Network Switching and Autonomic Visceral Restoration · 中文
Chapter 2. Switching Between Networks: Dwell Time and Its Cost
Three networks
Evidence
Resting-state networks (reviews and primary studies). Raichle and colleagues proposed a “default mode”: a set of regions more active at rest and less active during outward-directed tasks, including posterior cingulate/precuneus, medial prefrontal cortex and bilateral inferior parietal cortex (Raichle et al. 2001; Raichle 2015). Seeley and colleagues used resting-state functional connectivity to separate two networks that had previously been conflated: a salience network centred on dorsal anterior cingulate and anterior insula, and a central executive network centred on dorsolateral frontal and parietal cortex (Seeley et al. 2007). Menon and Uddin proposed that the insula acts as a “switch” between them (Menon and Uddin 2010), and Menon developed the relation of the three into a “triple network model” applied to psychopathology (Menon 2011). Sridharan and colleagues argued from Granger causality analysis that right anterior insula plays a critical role in switching between the executive and default-mode networks (Sridharan, Levitin, and Menon 2008).
Note
The word “network” carries a great deal. The studies above use functional connectivity, that is, statistical dependence between the time series of voxels or regions: not anatomical connection, and not causation. Volume 25 discussed the interpretive limits of network measures; Volume 52, Chapter 1 gave a more extreme example, in which a co-occurrence network reflecting only prescribing habits still yielded a handsome clustering coefficient and a list of “core nodes”. When reading “network”, ask what its edges are.
The default mode network is associated with spontaneous thought and mind-wandering: Mason and colleagues linked mind-wandering to the default network (Mason et al. 2007), and Christoff and colleagues proposed writing spontaneous thought as a dynamic variation in constraint rather than as a state (Christoff et al. 2016). In meditation research, Brewer and colleagues reported lower activity in main default-network nodes and different connectivity patterns in experienced meditators (Brewer et al. 2011), and Garrison and colleagues reported that such differences persist beyond an active task (Garrison et al. 2015).
Note
Those meditation studies are small cross-sectional or quasi-experimental studies. Brewer and colleagues and Garrison and colleagues had participant numbers in the tens, and compared “experienced against novice” or “state against state”, not randomly allocated groups. A cross-sectional difference cannot distinguish “practice produced the difference” from “people who differ to begin with are more likely to keep practising”. Van Dam and colleagues discuss this class of problem systematically (Van Dam et al. 2018).
Writing switching as a Markov chain
The second link of the original claim is this: inner events — a thought, a bodily sensation, an image — stop recruiting the salience network, and so one remains in the default mode uninterrupted. That can be made precise: record the “dominant state” at each sampling instant as D (default), S (salience) or E (executive), and describe switching with a transition matrix. State-sequence models of this kind have precedent in the resting-state literature: Vidaurre, Smith and Woolrich used a hidden Markov model to characterise the temporal organisation of resting-state networks and found that state visits are themselves hierarchically organised (Vidaurre, Smith, and Woolrich 2017).
The transition matrix in this chapter is a schematic chosen by the author, not a fit to any data. The program’s job is not to estimate real occupancies but to compute what follows from a transition matrix, and that part is exact.
Calculation
# Chapter 2. The triple network as a three-state Markov chain over dominant
# states (D = default mode, S = salience, E = central executive), fitted to
# nothing: the transition matrix is a schematic the author chooses. What the
# program computes is what follows FROM a transition matrix, which is exact:
# the stationary distribution (power iteration), expected dwell time
# 1/(1-p_ii) in units of the sampling step, the entropy rate, and mean first
# passage time from D to E. It then varies one entry, the D->S probability,
# which is the author's reading of "the salience network stops being recruited
# by inner events", and shows what that does to the rest.
import math
STEP = 0.72 # s per TR, a common resting-state fMRI sampling interval
STATES = ("D", "S", "E")
def stationary(P, iters=20000):
v = [1.0 / len(P)] * len(P)
for _ in range(iters):
v = [
math.fsum(v[i] * P[i][j] for i in range(len(P)))
for j in range(len(P))
]
s = math.fsum(v)
v = [x / s for x in v]
return v
def entropy_rate(P, pi):
h = 0.0
for i, row in enumerate(P):
for p in row:
if p > 0:
h -= pi[i] * p * math.log(p, 2)
return h
def mfpt(P, src, dst):
"""Mean first passage time src -> dst, in steps, by solving (I-Q) t =
1."""
idx = [i for i in range(len(P)) if i != dst]
Q = [[P[i][j] for j in idx] for i in idx]
n = len(idx)
A = [
[(1.0 if i == j else 0.0) - Q[i][j] for j in range(n)] + [1.0]
for i in range(n)
]
for c in range(n): # Gauss-Jordan
p = max(range(c, n), key=lambda r: abs(A[r][c]))
A[c], A[p] = A[p], A[c]
d = A[c][c]
A[c] = [v / d for v in A[c]]
for r in range(n):
if r != c and A[r][c] != 0.0:
f = A[r][c]
A[r] = [a - f * b for a, b in zip(A[r], A[c])]
return A[idx.index(src)][n]
def build(p_ds):
"""Schematic chain; the D row is renormalised so only D->S is varied
against D->D."""
rest = 0.10 # D -> E held fixed
return [[1.0 - p_ds - rest, p_ds, rest],
[0.22, 0.55, 0.23],
[0.18, 0.12, 0.70]]
if __name__ == "__main__":
print(
"schematic transition matrix (rows sum to 1), sampling step {:.2f} s"
.format(STEP)
)
P0 = build(0.25)
print(" " + "".join("{:>8s}".format(s) for s in STATES))
for s, row in zip(STATES, P0):
print(
" {:5s}".format(s) + "".join("{:8.2f}".format(v) for v in row)
)
pi = stationary(P0)
print("\nstationary occupancy: " +
" ".join("{} {:.3f}".format(s, v) for s, v in zip(STATES, pi)))
print("expected dwell (s): " +
" ".join("{} {:.2f}".format(s, STEP / (1 - P0[i][i]))
for i, s in enumerate(STATES)))
print("entropy rate: {:.4f} bits/step = {:.4f} bits/s"
.format(entropy_rate(P0, pi), entropy_rate(P0, pi) / STEP))
print("mean first passage D -> E: {:.2f} steps = {:.2f} s"
.format(mfpt(P0, 0, 2), mfpt(P0, 0, 2) * STEP))
print("\nvarying only D -> S (D -> E fixed at 0.10):")
print(
" p(D->S) occ D occ S occ E dwell D (s) H (bits/s) MFPT"
" D->E (s)"
)
for p_ds in (0.40, 0.30, 0.25, 0.15, 0.08, 0.03):
P = build(p_ds)
q = stationary(P)
print(" {:7.2f} {:7.3f} {:7.3f} {:7.3f} {:12.2f} {:11.4f} {:14.2f}"
.format(p_ds, q[0], q[1], q[2], STEP / (1 - P[0][0]),
entropy_rate(P, q) / STEP, mfpt(P, 0, 2) * STEP))
print(
"\nLowering D -> S raises default-mode occupancy and lengthens its"
" dwell"
)
print(
"time; it also LOWERS the entropy rate and LENGTHENS the path to the"
)
print(
"executive state. A model in which inner events stop recruiting"
" salience"
)
print(
"is therefore also a model of becoming slower to engage a task."
" That is"
)
print("a cost, not a side effect.")
# schematic transition matrix (rows sum to 1), sampling step 0.72 s
# D S E
# D 0.65 0.25 0.10
# S 0.22 0.55 0.23
# E 0.18 0.12 0.70
#
# stationary occupancy: D 0.362 S 0.293 E 0.345
# expected dwell (s): D 2.06 S 1.60 E 2.40
# entropy rate: 1.2747 bits/step = 1.7704 bits/s
# mean first passage D -> E: 6.83 steps = 4.92 s
#
# varying only D -> S (D -> E fixed at 0.10):
# p(D->S) occ D occ S occ E dwell D (s) H (bits/s) MFPT D->E (s)
# 0.40 0.285 0.351 0.364 1.44 1.8348 4.47
# 0.30 0.332 0.315 0.353 1.80 1.8035 4.74
# 0.25 0.362 0.293 0.345 2.06 1.7704 4.92
# 0.15 0.440 0.234 0.326 2.88 1.6437 5.43
# 0.08 0.519 0.174 0.307 4.00 1.4673 6.02
# 0.03 0.596 0.116 0.288 5.54 1.2472 6.66
#
# Lowering D -> S raises default-mode occupancy and lengthens its dwell
# time; it also LOWERS the entropy rate and LENGTHENS the path to the
# executive state. A model in which inner events stop recruiting salience
# is therefore also a model of becoming slower to engage a task. That is
# a cost, not a side effect.Read the table in this order. Given the matrix, the stationary occupancies are D 0.362, S 0.293, E 0.345; the expected dwell times ( with a sampling step of 0.72 s) are 2.06, 1.60 and 2.40 s; the entropy rate is 1.2747 bits per step or 1.7704 bits per second; and the mean first passage time from D to E is 6.83 steps, that is 4.92 s.
Then one number is changed: , the probability that an inner event recruits the salience network, with held at 0.10. The result is monotone, and three things move together:
- default-mode occupancy rises from 0.285 to 0.596;
- default-mode dwell time rises from 1.44 s to 5.54 s;
- the entropy rate falls from 1.8348 to 1.2472 bits per second, and first passage to the executive state lengthens from 4.47 s to 6.66 s.
The first two are what the original claim wants; the last two are the cost it does not mention. That cost is not an empirical finding but an arithmetic consequence of the matrix: making one state stickier simultaneously makes the whole chain more predictable and makes leaving that state slower.
Hypothesis · not established
The author’s hypothesis (not established): a practice aimed at “not recruiting salience” will, alongside a longer default-mode dwell time, lengthen the latency of switching to an external task for some period after practice, appearing as longer reaction times or more omissions at task onset. What would show it to be wrong: in a pre-registered trial with task-switching cost at a fixed post-practice time as a pre-specified outcome (in a cued task-switching paradigm, say), finding that the practice group’s switching cost is no longer than a structurally matched control group’s, with a confidence interval excluding a pre-set smallest difference of interest, would refute it.
Note
A falling entropy rate is not “bad”, and a longer first passage is not “bad” either. They are costs. Someone whose work demands frequent switching and someone repeatedly interrupted by intrusive thoughts can afford different costs. This book does not weigh them for the reader; it asks only that the cost be written down. Chapter 8 implements this as a pre-specified negative-control outcome and a secondary outcome for switching cost.
The trouble with the phrase “soothes the default mode network”
The original claim includes the phrase “higher-order awareness soothes the DMN”. It needs correcting at two levels.
First, direction. Default-mode activity is higher during mind-wandering, but it is not the “tense” network; outward-directed, effortful tasks lower its activity. If a practice raises default-mode occupancy and dwell time (the previous section), then in terms of network measures the practice makes the default mode more active, not soothed. The reduced default-network activity reported in meditation research (Brewer et al. 2011; Garrison et al. 2015) and “remaining in the default mode” are two different descriptions, and whether both can hold at once depends on whether “default-mode activity” and “dominant state is D” are the same thing, which they are not.
Second, level. “Higher-order awareness” is, in the Chapter 1 model, the parameter : a scalar scaling the second-level pathway. Identifying it with a region or a network requires independent evidence. The “switching” role of anterior insula (Sridharan, Levitin, and Menon 2008; Menon and Uddin 2010) is a candidate, but that is evidence about switching, not evidence that one network modulates the precision of another.
Evidence
What can be stated. Three networks can be separated in resting-state data (Seeley et al. 2007); their activity has temporal structure describable by state sequences (Vidaurre, Smith, and Woolrich 2017); anterior insula and dorsal anterior cingulate are involved in the reorienting of attention (Menon and Uddin 2010; Uddin 2015); and default-mode activity correlates with spontaneous thought (Mason et al. 2007; Christoff et al. 2016). What cannot be stated is which network “modulates” the precision of which other, or whether such modulation is the mechanism of any practice.
What a state sequence can measure
Suppose a state sequence really were recorded, estimated from functional MRI or EEG with a hidden Markov model. What questions could it answer? The three quantities computed above — occupancy, dwell time, entropy rate — all follow from the transition matrix, which has to be estimated from a sequence of finite length. There is a statistical problem here identical to one in Chapter 3: entropy estimates on short sequences carry a systematic downward bias. Chapter 3, program p13 gives the magnitude and the correction; there the object is a sequence of self-reported imagery categories and here it is a sequence of network states, but the mathematics is the same.
This book therefore does not recommend an entropy rate as any primary outcome. Occupancy and dwell time are usable: they are a proportion and a reciprocal, with far smaller bias and a direct interpretation. Chapter 8 lists attention-related frontal indices as secondary outcomes and states explicitly that they do not change the conclusion.
Summary
Three networks can be separated in resting-state data and their temporal organisation can be characterised with state-sequence models; both are supported. Once “inner events no longer recruit the salience network” is written as a fall in a transition probability, three things happen together: default-mode occupancy rises from 0.285 to 0.596 and dwell time from 1.44 to 5.54 s, while the entropy rate falls from 1.8348 to 1.2472 bits per second and first passage to the executive state lengthens from 4.47 to 6.66 s. The first two are what the claim wants; the last two are its cost, and that cost is an arithmetic necessity rather than an empirical finding. The phrase “soothes the default mode network” needs correction in both direction and level: raising default-mode dwell time and lowering default-mode activity are different descriptions, and identifying “higher-order awareness” with the precision modulation of some network is not yet supported. Finally, entropy-rate estimates on short state sequences are systematically biased and are unsuitable as a primary outcome.
References
Fan, Y. (2026). Interoceptive Decoupling, Salience-Network Switching and Autonomic Visceral Restoration: From Precision-Weighting in the Default Mode Network and the Redistribution of Vagal Tone to the Neurophenomenology of a Non-Judgmental Imagery Stream, with a Textual Examination of neiguan in the Huangdi Neijing [内感受解离、突显网络切换与内脏自主神经修复] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24020418 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).