 
 

 
 

 

 

 

 

 

 

 

 

 

 

 

. ; .
. V . .. .
V . . . .
v V .. .
. V . . .
V p .‘ . V
. . . .
r .
m . .
. . . . V . V
v . . . .
. V V . . . . .
‘V . V. V .
V . . . .
. . . V V
, . . . . V
. .V. V.
. . V . . . ‘. V
a .
. . .. V. . . .. .
V . . .
.
. . . .. V .x . ‘
A . .
. . .
‘ . . . V
,V‘ . V . .
,4 V V .
. V
. . . .
. . . . . V
. v .. . .
, .
. .
. V .
r . .
.. . .
. . .
.. . .. V .. .
«V .. V V . V .
. . V
. V ‘ . .
V . 1. . . . .

’.vu."

-.- nu-

...-..-..

m; _ h. . . .i mmahﬁh

. V . . V V. . V V (tat.

 

This is to certify that the
dissertation entitled

NEURODYNAMICS OF EPISODIC MEMORY
CONSOLIDATION

presented by

JEREMY LEE SMITH

has been accepted towards fulﬁllment
of the requirements for the

PHD. degree in NEUROSCIENCE

 

 

Major Professor’s Signature
/ MM DPQ
Date

MSU is an afﬁnnalive-action, equal-opportunity employer

 

o.-.-v-.-.-.-.-.-.-.-.-.--.-.—.-u-.-.-.—-.--.-.-.—.—.—.-.-—-—-n-.-.—.—.-.--.---._.-—._._ _.--.-.-.-.--.-.—._._

 

 

 

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

DATE DUE DATE DUE DATE DUE

 

 

NEURODYNAMICS OF EPISODIC MEMORY CONSOLIDATION

By

Jeremy Lee Smith

A DISSERTATION

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Neuroscience

2008

ABSTRACT
NEURODYNAMICS OF EPISODIC MEMORY CONSOLIDATION
By

Jeremy Lee Smith

The present study endeavored to implement conventional FMRI analyses and two novel
approaches to BOLD signal characterization with consolidation — Shannon entropy and
voxel-response cross-correlation with task (VTCC) — to assess the brain's ability to
optimize memory network assemblies with consolidation. In this dissertation the current
state of the episodic consolidation literature is reviewed. These concepts are then related
to the anatomy and ﬁmctional architecture of episodic memory systems. Past and present
FMRI and electrophysiological methodologies are also discussed and related to the study

of physiological changes in the memory system with consolidation.

A paired-associate task was used to examine changes in ﬁmctional modularity and
connectivity with episodic memory consolidation. This task required subjects to learn to
associate visually presented words with abstract pictures, then identify correctly-matched,
incorrectly-matched, and control pairings during a testing session. A hypothesized
improvement subject recall performance with consolidation was not substantiated, but
instead remained level between pre- and post-consolidation subjects, suggesting that
access to the memory traces remained approximately constant over the consolidation
interval despite network changes. Improvements in performance were correlated with

decreased recruitment of premotor cortex and visual areas V2, V3, and V4, but not

 

lllfl I.\ III. {ll III I. Ill .1!

medial temporal, primary motor, or primary visual cortices. Changes in activation
volume, BOLD signal magnitude, the mean correlation between activated voxels and the
paired-associate task, and BOLD signal Shannon entropy, were also assessed. The
present ﬁndings indicate that neurons still participatory in trace retrieval were
individually more active than their pre-consolidation counterparts, that BOLD signal
entropy increased with consolidation in perirhinal cortex and in components of the
ventral visual stream, and, moreover, that BOLD signal response rates increased
signiﬁcantly with consolidation in all areas except left hemisphere primary motor cortex.
These ﬁndings are discussed in the context of basic scientiﬁc research in memory

network dynamics, as well as the methodological contributions to neuroscience.

Copyright by
JEREMY LEE SMITH
2007

ACKNOWLEDGEMENTS

The author wishes to thank Mrs. Katey Smith, Mr. William Smith, Ms. Barbara Parker
Smith, Mr. and Mrs. Donald and Marcia Brown, Mr. Kenneth M. Cobb, and Dr. Craig
Branch for their ongoing support during the writing of this dissertation; Drs. Alessandra
M. Passarotti, Laura Symonds, Sue Barman, Christine Larson, David Zhu, Jerry Gebber,
and Cheryl Sisk for indispensable advice and guidance; Dr. David Zhu, Mr. Jeremy
Grounds, and Mr. Dave McFarlane for unremitting patience and assistance with daily
technical issues; Dr. Matt Hoptman, for critical feedback on this manuscript; and the
students and faculty of the Neuroscience Program, without whose encouragement and

helpful comments this work would likely have suffered exceedingly.

TABLE OF CONTENTS

LIST OF TABLES ..................................................................................................... VIII
LIST OF FIGURES ........................................................................................................ X
LIST OF FIGURES ........................................................................................................ X
EPISODIC MEMORY AND MEMORY CONSOLIDATION ........................................ 1
Connectivity and Functional Architecture in Episodic Retrieval ................................... 4
Declarative memory: the semantic and episodic systems .......................................... 5
Role of other memory systems in episodic retrieval ................................................ 10
REGIONS OF INTEREST ............................................................................................ 13
Premotor cortex as a visuospatial attention area ......................................................... 17
Visual areas ............................................................................................................... 18
Primary motor cortex: a control area .......................................................................... 20
FMRl SPATIAL AND TEMPORAL SPECIFICITY ..................................................... 21
BOLD signal physiology ........................................................................................... 21
Spatiotemporal resolution of F MRI ............................................................................ 23
Modeling the hemodynamic response functions ......................................................... 26

AIMS AND THEORETICAL TREATMENT OF CONSOLIDATION

NEURODYNAMICS .................................................................................................... 27
Consolidation as an optimizing process ...................................................................... 28
Evidence for interim physiological changes in retrieval substrates ............................. 3O
Computational efﬁciency and cortical demands: activation volume and entropy ........ 32
Cortical reinstatement and retrieval orientation effects in visual and medial-temporal
areas .......................................................................................................................... 35

HYPOTHESES ............................................................................................................. 38

METHODOLOGY ........................................................................................................ 43
Subject pool and participants ..................................................................................... 43
Paradigm ................................................................................................................... 44
Data acquisition ......................................................................................................... 47
Behavioral analysis .................................................................................................... 48
Modular analysis ....................................................................................................... 49

vi

Deconvolution ....................................................................................................... 49

Percent BOLD signal change from baseline ........................................................... 52
Activation volume .................................................................................................. 53
Voxel-task cross-correlation ................................................................................... 54
Shannon entropy of the voer time series ............................................................... 55

MEMORY CONSOLIDATION AND RETRIEVAL PERFORMANCE: EFFECTS OF

STIMULUS MODALITY AND PAIR CONCORDANCE ............................................ 59
Behavioral results ...................................................................................................... 61
Reaction time ......................................................................................................... 62
Accuracy ................................................................................................................ 63
Changes in BOLD Physiology with Consolidation ..................................................... 71
R01 analysis Of metabolic demand ......................................................................... 71
R01 analysis of cortical recruitment ....................................................................... 73
R01 analysis of BOLD response rate ...................................................................... 75
Retrieval orientation and cortical reinstatement: Interpretation of Physiological
Findings in the Context of Signal Entropy .................................................................. 78
NEURODYNAMICS OF EPISODIC MEMORY CONSOLIDATION: CONCLUSIONS
AND DISCUSSION ...................................................................................................... 85
Effects of stimulus modality and testing interval on task performance ........................ 87
Disparity between behavioral predictions and ﬁndings ........................................... 87
Theoretical reconsolidative basis for the non-Kamin Ebbinghaus exception ........... 89
Interpretation of Metabolic Demand, Recruitment, and Signal Response Rate ............ 93
Disparity between physiological predictions and ﬁndings ...................................... 93
Caveats of the entropy metric ............................................................................... 100
Theory: Cortical reinstatement and Retrieval Orientation ......................................... 101
Possible cortical reinstatement effects .................................................................. 103
Closing remarks ....................................................................................................... 105
APPENDIX A: ANOVA TABLES FOR BEHAVIORAL DATA ................................ 107

APPENDIX B: MAIN-EFFECT ANOVA AND POST-HOC RESULTS FOR
PHYSIOLOGICAL STATISTICS ............................................................................... 109

APPENDIX C: DECONVOLUTION COMMAND LINE ........................................... 134

vii

LIST OF TABLES

Table 1 Projections to the perihippocampal regions (Lavenex and Amaral 2000).
Abbreviations: PPC, posterior parietal [association] cortex; SSC, somatosensory
cortex; OFC, orbitofrontal cortex; PFC, prefrontal cortex; ST S, superior temporal
sulcus; TE, V4, visual areas TE and V4. ................................................................... 8

Table 2 Regions typically involved in episodic retrieval tasks. LH, left hemisphere; RH,
right hemisphere; BL, bilateral. Several of these regions will be used as target ROIs
in the present study (see next chapter). ................................................................... 12

Table 3 Summary of the hypotheses for the present study. ............................................ 41

Table 4 Summary of the predictions for the present study with respect to subject
performance (reaction time and accuracy), as well as physiological descriptions of
the shape and temporal characteristics of the BOLD signal on the delayed match—to-
sample portions of the paired associates task. Both behavioral and physiological
ﬁndings are represented as being greater (+), less (—), or equal (no change, N/C) in
post-consolidation subjects relative to pre-consolidation subjects. Abbreviations:
HC, hippocampus, ERC, entorhinal cortex, EcRC, ectorhinal cortex, PRC, perirhinal
cortex, V1, V2, V3, visual areas 1-3, F US], fusiform gyrus, M1, primary motor
cortex, PMC, premotor cortex; RT, subject reaction time, Acc, accuracy; BOLD
mag, percent BOLD signal change from baseline, recruitment, cortical recruitment
(activation volume), BOLD rate, BOLD signal response rate (evaluated by the
voxel-task cross-correlation, VTCC). Hypothesis 4 (H4), which predicts a
predominance of right hemisphere activation in post-consolidation subjects, is not
included in the table. .............................................................................................. 41

Table 5 A qualitative review of the results presented in the present chapter and the
conclusions drawn ﬁom them. “Per-neuron activity” was determined by dividing
the metabolic demand PC) of an ROI by the extent of its recruitment (AV) to obtain
a rudimentary AP/mm estimate. Abbreviations: N/C, no change; fusi., fusiform
gyrus; Vx, visual area x. ......................................................................................... 84

Table 6 Summary of the predictions (“Predicted”) and ﬁndings (“Actual”) for the present
study with respect to subject performance (reaction time and accuracy), as well as
physiological descriptions of the shape and temporal characteristics of the BOLD
signal on the delayed match-tO-sample portions of the paired associates task. Both
behavioral and physiological ﬁndings are represented as being greater (+), less (—),
or equal (no change, N/C) in post-consolidation subjects relative to pre-
consolidation subjects. Abbreviations: HC, hippocampus, ERC, entorhinal cortex,
EcRC, ectorhinal cortex, PRC, perirhinal cortex, V l , V2, V3, visual areas 1-3, F USI,
fusiform gyrus, MI, primary motor cortex, PMC, premotor cortex; RT, subject

viii

reaction time, Acc, accuracy; BOLD mag, percent BOLD signal change from
baseline, recruitment, cortical recruitment (activation volume), BOLD rate, BOLD
signal response rate (voxel—task cross-correlation) .................................................. 86

LIST OF FIGURES

Figure 1 Connections of the rhinal and perihippocampal cortices. PHC, perihippocampal
cortex; PRC, perirhinal cortex; EcRC, ectorhinal cortex; ERC, entorhinal cortex;
HC, hippocampus. Adapted from (Lavenex and Amara12000) and (Save, Nerad and
Poucet 2000). ........................................................................................................... 7

Figure 2 (A) Interconnections among constituent regions of the MTL and other areas of
interest in the present study. Blue arrows indicate neuronal projections between
areas as determined by anatomical studies of the cortex and temporal lobe. Arrow
locations do not accurately reﬂect the anatomical location of interregional
projections (Sources: Deacon, Eichenbaum, Rosenberg and Eckmann 1983;
Schacter, Buckner, Koutstaal, Dale and Rosen; Burwell and Amaral; Burwell and
Amaral; Lavenex and Amaral; Save and Poucet 2000a; Save and Poucet 2000b). (B)
The anatomical locations of MTL areas in the Talairach-Tourneaux template brain,
obtained from the AFNI suite’s Talairach Daemon. Abbreviations: LH, left
hemisphere; RH, right hemisphere; ctx., cortex; BA, Brodmann area. ..................... 16

Figure 3 The training/testing paradigm for the present study. Stimuli consisted of 20 non-
rehearsable, non-generalizable abstract pictures paired with animal nouns. Word
stimuli were presented either audibly (computer-based training/testing) or visually
(FMRI training/testing). Following a training session, during which the correct
pairings of stimuli were learned, subjects were tested on their recall either
immediately (pre-consolidation time point; 50% of subjects) or 7 days after training
(post-consolidation time point; 50% of subjects). For the purposes of testing,
subjects were presented with 33 correctly-matched, 33 incorrectly-matched, and 34
control (scrambled) stimulus pairs, and subjects were required to identify each type
correctly by button press. ....................................................................................... 44

Figure 4 (A) Example of a paired associate as presented for the FMRI version of the
training/testing paradigm: note that the word component of FMRI training and
testing associates was presented below the abstract visual component. The centroid
of the presented paired associate was positioned in the center of the subjects' ﬁeld of
View. (B) A schematic of a single run ﬁ'om the FMRI testing paradigm: match,
mismatch, and control-condition stimuli were presented with a constant
interstimulus interval of 12.5 seconds. Presentation order was randomized across
the three types of stimuli; x-axis represents TR (1 TR = 2.5 seconds). .................... 46

Figure 5 Mean reaction times for the paired-associate task by testing group (computer-
based, CBT, versus FMRI subjects), pair concordance (matched, mismatched, or
control), and time point (pre- versus post-consolidation). FMRI subjects, who were
given unimodal (visual) training and testing stimulus pairs, were slower at
identifying both correctly-matched and mismatched stimulus pairs than CBT
subjects, who were given dual-modal (visual and auditory) training and testing

stimulus pairs. This effect was seen at both pre- and post-consolidation time points.
However, neither FMRI nor CBT subjects were signiﬁcantly faster at identifying
correctly-matched stimulus pairs after one week. Neither group improved with
respect to reaction time with consolidation for control stimuli, nor did CBT and
F MRI subjects differ signiﬁcantly with respect to reaction time for control stimuli at
either time point. Error bars represent standard errors of the means. *p < .05. ........ 62

Figure 6 Mean subject accuracy for the paired-associate task by testing group (computer-
based, CBT, versus FMRI subjects), pair concordance (matched, mismatched, or
control), and time point (pre- versus post-consolidation). F MRI subjects, who were
given unimodal (visual) training and testing stimulus pairs, were more likely to
identify mismatched associates as matched associates (exhibited more false
positives) than CBT subjects, who were given dual-modal (visual and auditory)
training and testing stimulus pairs. However, neither FMRI nor CBT subjects
improved with respect to the number of false positives over the putative memory
consolidation period, nor did either group improve with respect to accuracy with
consolidation for matched associates (false negatives) or control stimuli. CBT and
FMRI subjects did not differ signiﬁcantly with respect to accuracy for correctly-
matched associates or control stimuli at either time point. Symbols (light grey)
indicate individual data points; error bars represent standard errors of the means. *p
< .05; **p < 0.01 .................................................................................................... 65

Figure 7 Group statistical parametric maps (SPMs) of voxel BOLD signal correlation
with presentation of match, mismatch, and control-condition stimuli by region of
interest. Voxel BOLD signal changes shown are signiﬁcant at a full-model F =
3.786 (p = 0.001, corrected); colors represent whether changes are positive (yellow)
or negative (blue) relative to the baseline. Group SPMs were obtained by
studentizing individual BOLD datasets and registering them to a common Talairach-
Tourneaux coordinate system (Talairach and Tourneaux, 1988). Datasets were then
subjected to cluster analysis using a minimal 5mm, 10u1 Gaussian kernel (to correct
for type 11 error: see Salrnond, Ashburner, et al. 2002; Geissler, Lanzenberger, et a1.
2005) and averaged by condition. The location of the calcarine ﬁssure (visual area
V1) is shown on an axial slice in each case for reference (yellow boxes).
Hemodynamic response functions for representative subjects were also obtained for
ectorhinal area 36 (red box) and visual area V3/19 (green box) and are shown in
Figure 9. Abbreviations: Fus, fusiform gyrus; Ins, AIC, anterior insular complex;
ST G, superior temporal gyrus; IPL, inferior parietal lobule; SPL, superior parietal
lobule; DG, dentate gyrus; pul, pulvinar nucleus of the thalamus, Ver, cerebellar
vermis, Cal, calcarine ﬁssure. Numbers represent Brodmann areas: 3a, digit
somatosensory area; 4, digit motor area; 31, posterior cingulate cortex ................... 67

Figure 8 Estimated event-related average hemodynamic response functions (HRFs) for
pre- (dotted line) and post- (solid line) consolidation right hemisphere ectorhinal
cortex (EcRC, Brodmann area 36) and visual area V3 (Brodmann area 19) seed
locations indicated in Figure 7. HRFs were obtained from the same studentized
respresentative subject data as in Figure 7. Talairach-Toumeaux coordinates for
each region are also given (see Talairach and Toumeaux, 1988) ............................. 69

xi

Figure 9 Percent BOLD signal change ﬁ'om baseline (BOLD signal magnitude) in
signiﬁcantly task-correlated voxels constituting regions of interest at pre- and post-
consolidation for matched associates. Analysis by main effect ANOVA indicated a
signiﬁcant difference in BOLD magnitude between pre- (T1) and post- (T2)
consolidation subjects (p < 0.001) overall, however, post-hoe analysis by Tamhane
T2 failed to indicate any differences in per-region signal change with consolidation.
These ﬁndings suggest that the BOLD response magnitude is not affected by
consolidation in signiﬁcantly task-correlated voxels in these regions. Error bars
indicate standard errors. ......................................................................................... 72

Figure 10 Activation volume of signiﬁcantly task-correlated voxels at pre— and post-
consolidation for matched associates. Analysis by main effect ANOVA indicated a
signiﬁcant difference in activation volume between pre- and post-consolidation
subjects overall (p < 0.001). Post-hoe analysis by Tamhane T2 further indicated
signiﬁcant reductions in cortical recruitment in the fusiform gyrus, premotor cortex,
and visual areas V2 and V3. These ﬁndings suggest that the number of signiﬁcantly
task-correlated voxels is affected by consolidation in premotor areas, area V2, and
ventral visual areas, but not in medial temporal, primary motor, or primary visual
areas. Error bars indicate standard errors; *indicates a signiﬁcant difference in
activation volume, p < 0.05. ROI signiﬁcances computed by post-hoc Tamhane T2.
.............................................................................................................................. 74

Figure 11 Mean voxel BOLD response-stimulus presentation covariance (voxel-task
cross-correlation, VTCC) in signiﬁcantly task-correlated voxels constituting regions
of interest at pre- (T1) and post- (T2) consolidation for matched associates. Analysis
by main effect ANOVA indicated a signiﬁcant difference in VTCC between pre-
and post-consolidation subjects overall (p < 0.001), and post-hoe analysis by Tukey
HSD revealed signiﬁcant differences in VTCC in all areas (p < 0.05). These
ﬁndings suggest that the BOLD-task covariance is affected by consolidation in
signiﬁcantly task-correlated voxels in medial temporal, motor, premotor, and visual
regions. Error bars indicate standard errors; *indicates a signiﬁcant difference in
VTCC, p < 0.05. ROI signiﬁcances computed by post-hoc Tukey HSD unequal-N
test. ........................................................................................................................ 76

Figure 12 Mean BOLD signal Shannon entropy (TSE) in signiﬁcantly task-correlated
voxels constituting regions of interest at pre- and post-consolidation for matched
associates. Analysis by main effect ANOVA indicated a signiﬁcant difference in
BOLD TSE between pre- and post-consolidation subjects overall (p < 0.001). Post-
hoc analysis by Tamhane T2 ﬁrrther indicated signiﬁcantly greater signal entropy in
the ﬁrsiform gyrus and visual areas V2 and V3 (p < 0.01). These ﬁndings suggest
that the BOLD signal entropy is affected by consolidation in signiﬁcantly task-
correlated voxels in extrastriate ventral visual areas, but not in medial temporal,
premotor, or primary motor or visual cortex. Error bars indicate standard errors;
*indicates a signiﬁcant difference in TSE, p < 0.01. ROI signiﬁcances computed by
Tamhane T2. .......................................................................................................... 81

xii

KEY TO SYMBOLS AND ABBREVIATIONS

Acc Accuracy

ANOVA Analysis of variance

AV Activation volume

BA Brodmann area

BOLD Blood oxygenation level dependent
CBF Cerebral blood ﬂow

CBT Computer-based testing

CMRO; Cerebral metabolic rate of oxygen
DMTS Delayed match to sample

DNMS Delayed nonmatch to sample
dVS Dorsal visual stream

EcRC Ectorhinal cortex

EEG Electroencephalography

EPI Echo-planar imaging

ERC Entorhinal cortex

EWS Extrastriate ventral visual stream
FMRI Functional magnetic resonance

imaging

HbIdI-lb Hemoglobin/deoxyhemoglobin

HC
HRF
IRF
LFP
LH
LTD
LTP
M1
MTNS
MTL

Hippocampal formation
Hemodynamic response function
Impulse response function

Local ﬁeld potential

Left hemisphere

Long-term depression

Long-term potentiation

Primary motor cortex

Visual area V5/MT

Medial temporal lobe

PC I AP Percent BOLD signal change over

PET
PHC
PHG
PMC
PRC
PRS
R

baseline

Positron emission tomography
Parahippocampal gyrus
Perihippocampal cortex
Premotor cortex

Perirhinal cortex

Perceptual representation subsystem
Pearson product-moment correlation

RH
ROI

RT
Sh

SNR
SPM

TSE
V1
VTCC

vVS

xiii

Right hemisphere
Region of interest

Response time
Shannon bits

Signal-to-noise ratio

Statistical parametric map

Tesla

Shannon entropy of the time series
Visual area V1 (etc.)

Voxel-to-task cross-correlation
(indicative of BOLD signal rate)
Ventral visual stream

Denoised BOLD signal response vector
Raw BOLD signal (EPI) response vector

EPISODIC MEMORY AND MEMORY CONSOLIDATION

Consolidation refers to the process by which short-term memories are gradually indexed,
transferred to cortex, and made permanent (Abel and Lattal 2001; Eichenbaum 2001;
Haist, Bowden Gore and Mao 2001). Episodic memories are quite labile until stabilized
through this process, which results in stronger connections among the elements (nodes)
that are relevant to the memory and weaker connections among elements that are
irrelevant to the memory. Neuroimaging and neuropsychological studies have provided
reasonable evidence that the consolidation mechanism proceeds in two stages. In the
ﬁrst stage, short-term memory traces are encoded in the hippocampal CA3 region and
linked by long-term potentiation to form stable networks. In the second stage, the medial
temporal lobe (MTL) and association cortex backprojections, which are Hebb-modiﬁable
by entorhinal 0 rhythm entrainment (ﬁring rate coherence: Treves and Rolls 1994;
Chrobak and Buzsaki 1998b) are reactivated and reinforced. It is believed that the
backprojections themselves drive the “repositioning” of the index of a memory from the
hippocampal complex to the frontal lobe over a period of several years (Treves and Rolls
1994; Rolls 2000). The present study encompasses seven days between training on an
association task and subsequent testing. Consequently, only Stage 1 consolidation effects
— long-term potentiation of the short-term memory traces, evinced by increased
association strength (retrieval efficiency, discussed in a later chapter) between the

hippocampus and primary, secondary, and association areas — are expected.

 

Electrophysiological studies have produced reasonable evidence that particular
spectral (frequency-related) changes in the interactions of neural populations occur at this
stage, particularly with respect to coherence (the correlation between the timing and
amplitude of population responses) and spectral power (Bodizs, Bekesy, Szucs, Barsi and
I-Ialasz 2002; Cantero, Atienza, Salas and Dominguez-Marin 2002b). At least one
neuroimaging study has also demonstrated an overall decrease in the recruitment of brain
areas after a period of consolidation (Buchel, Coull and Friston 1999). Both ﬁndings
corroborate the intuitive prediction of changes due to Hebbian processes: brain regions
necessary for learning, such as executive, attentional, and working memory regions, are
recruited at the pre—consolidation (encoding) time point, as are regions involved in the
perception of peripheral (distractor) stimuli; connections between these regions and the
memory network undergo long-term depression, however, and are not activated at the
post-consolidation recall time point (Alvarez and Squire 1994; Chrobak and Buzsaki
1998b; Chrobak and Buzsaki 1998a). Electrophysiological studies cannot unambiguously
resolve signal sources, however, and thus cannot be used to create a ﬁnal, systems-level
interpretation of mnemonic binding; nor has any functional magnetic resonance (FMRI)
study, to the knowledge of the author, attempted to operationalize changes in retrieval
efficiency with these measures. The present study therefore endeavors to assess changes
in cortical recruitment, metabolic demands, and FMRI signal complexity1 and rate of

change as correlates of memory consolidation.

 

‘ The terms complexity, entropy, and chaoticity are used interchangeably in the present dissertation, but it
should be pointed out that all three terms are used in a technical (as opposed to lay) sense to refer to the
Shannon (information-theoretic) entropy of a signal. More speciﬁcally, the Shannon entropy is a
quantiﬁcation for the degree of a vector’s unstable aperiodicity. The BOLD time series Shannon entropy
(TSE) is described in depth in the “Shannon entropy of the voxel time series (TSE)” section of the
Methodology chapter; the reader is also referred to Shannon’s seminal 1948 paper, A Mathematical Theory
of Communication (Shaman, C. E. (1948). A mathematical theory of communication. Bell Sys Tech J 27:

The consolidation period has been shown to result in the decreased lability of
memory traces, increased long-term potentiation (LTP) and ﬁring rate coherence among
mnemonic network constituents, and increased long-term depression (LTD) with
decreased recruitement among peripheral networks (Bailey, Montarolo, Chen, Kandel
and Schacher 1992; Alvarez and Squire 1994; Zhang, Endo, Cleary, Eskin and Byrne
1997; Chrobak and Buzsaki 1998b; Chrobak and Buzsaki 1998a; Buchel, Coull et al.
1999; Eichenbaum 2001; Bodizs, Bekesy et al. 2002; Cantero, Atienza et al. 2002b).
Thus, one might expect the memory consolidation interval to serve as a temporal window
during which binding and synaptic integration among memory traces are “tuned” to
optimal efficiency (see McClelland, McNaughton and O'Reilly 1995; Squire and Alvarez
1995; Chrobak and Buzsaki 1998b; Abel and Lattal 2001; Eichenbaum 2001; Haist,
Bowden Gore et al. 2001; Ross and Eichenbaum 2006; Wang, Hu and Tsien 2006). This
does in fact appear to be the case, as previously-stored memory traces in medial temporal
cortex and isocortex are reactivated and entrained to so-called sharp wave-ripple complex
oscillations during the presumptive consolidation period of sleep, thus reenforcing
synaptic weights (Buzsaki 1998; Siapas and Wilson 1998; Stickgold, Scott, Rittenhouse
and Hobson 1999; Datta 2000; Maier, Nimmrich and Draguhn 2003; Gais and Born
2004; Walker and Stickgold 2004).

Conventional wisdom holds that, in spite of this putative consolidation-related
mnemonic network optimization, recall performance declines with time, except when the
items to be encoded are repeatedly presented; yet research into the neurodynamic
consequences of the consolidation phenomenon at the whole-brain level have only

recently been undertaken, and many of the published studies on the subject have focused

 

379—423, 623-656.)

on the hippocampal, rather than cortical and medial-temporal, contributions to recall.
Furthermore, recent experimental manipulations of encoding depth have indicated that
subjects who are well-trained to remember speciﬁc memory items may exhibit an
improved proﬁciency in cortical reinstatement, the reactivation of previously-encoded
memory traces, and an improved capacity, or retrieval orientation, for processing
stimulus features so as to better match them with stored representations, which suggests
that the strength of sensory and retrieval-mode binding processes may differentially
inﬂuence recall performance (Weldon, Roediger and Challis 1989; Rugg and Wilding
2000; Robb and Rugg 2002; Wheeler and Buckner 2003). However, these phenomena
have not yet been studied in the context of a between-subjects paired associates task, nor

in the context of episodic memory consolidation.

CONNECTIVITY AND FUNCTIONAL ARCHITECTURE IN EPISODIC RETRIEVAL

Demarcations of the various memory systems have classically been established through
neuropsychological observation of human patients, notably through experiments with the
amnesic patient H.M., and lesion studies in animals (Eichenbaum 1996; Eichenbaum,
Dusek, Young and Bunsey 1996; Eichenbaum, Schoenbaum, Young and Bunsey 1996).
These data have helped to establish the idea that memory can be dissociated into several
cognitive networks, each of which is responsible for a particular type of mental
representation of the past. At least ﬁve such networks are commonly referred to in the
literature: two declarative memory systems (episodic and semantic memory) and the

procedural, perceptual, and working memory systems.

Schacter and Tulving have proposed three criteria that a network should meet in
order to be considered an independent and complete memory system (Schacter and
Tulving 1994). First, the network should be composed of regions that perform
interrelated tasks that permit the storage and retrieval of a particular class of information.
Second, the network can be characterized in terms of operational qualities that are, as a
whole, unique to that network. And third, the operations of the network can be
dissociated from those of other networks through ablative or neuropsychological studies.

Based on these criteria, at least ﬁve major memory systems have been identiﬁed.
These include semantic memory (long-term memory for facts and concepts, including
associative memory for objects and words, independent of an episode in memory), the
perceptual representation system (long-term, non-associative memory for Objects and
words), procedural memory (long-term habit learning, classiﬁcation skills, grammatical
rules, and sequence memory), working memory (short-term retention of information for
cognitive problem solving), and episodic memory (long-term retention of events:
Eichenbaum 1996; long-term retention of events: Eichenbaum, Dusek et al. 1996; long-
terrn retention of events: Eichenbaum, Schoenbaum et al. 1996; Gabrieli 1998). Many of

these ﬁve systems can be further dissociated into subsystems, as discussed below.

Declarative memory: the semantic and episodic systems

Semantic memory is memory for information in the abstract: factual memory, such as
memory about historical events in which the subject did not participate. It is dependent
on the medial temporal lobes (MTL; including rhinal cortex and the hippocampal and

perihippocampal areas), and the fact that this anatomical division is shared with episodic

memory (discussed below) incorporates semantic and episodic memory into the
descriptive heading “declarative memory.” In addition, semantic memory seems to
recruit anterior and lateral regions of the temporal lobe in the left hemisphere
(Eichenbaum et al., 1996; Gabrieli, 1998).

Episodic memory itself is perhaps the most complex and best understood of the
ﬁve memory systems. It is involved in memory for events that are the personal
experience of the subject, and which are imbued with “warmth and intimacy,” as
characterized by William James (James 1893). Like semantic memory, episodic memory
has been linked with the MTL, which is comprised by the entorhinal cortex, perirhinal
cortex, perihippocampal cortex, perihippocampal gyrus, and the hippocampal formation.
The connections of the MTL are extensive; the hippocampus is connected to various
unimodal and polymodal association areas through the perihippocampal gyrus and to the
thalamus (see Figure 1). Furthermore, preﬁontal areas (particularly the ventrolateral
prefrontal cortex, precentral sulcus and inferior frontal gyrus; the dorsolateral preﬁontal
cortex, posterior middle frontal gyrus; and the left anterior preﬁ'ontal cortex are engaged
during the recall effort (ecphory: Buckner, Koutstaal, Schacter, Dale, Rotte and Rosen
l998a; Buckner, Koutstaal, Schacter, Wagner and Rosen l998b; Ranganath, Johnson and
D'Esposito 2003), whereas the hippocampal formation itself is recruited during actual
recall (Eichenbaum et al., 1996). Right prefrontal cortex, which is involved in attention,
has also been disambiguated from episodic retrieval (Cabeza, Locantore and Anderson
2003).

The medial temporal areas have traditionally been classiﬁed as subserving novelty

detection (recognition: perirhinal and entorhinal cortices and the posterior

Figure 1 Connections of the rhinal and perihippocampal cortices. PHC, perihippocampal
cortex; PRC, perirhinal cortex; EcRC, ectorhinal cortex; ERC, entorhinal cortex; HC,
hippocampus. Adapted from (Lavenex and Amaral 2000) and (Save, Nerad and Poucet
2000)

 

Isocortex

Visual (object) cortex Visuospctial, polysensory, and auditory
association cortices

 

 

 

 

 

PRC : = PHC

 

ll

EcRC . . ERC
\

 

 

 

Long-term memory acquisition,
. consolidation, and recall;
I"SUIa HC responds to saliencies across

temporally»dispersed stimuli

Visual long-term memory,
exact function unknown

Encodes stimulus saliencies and
stores or “indexes” memory
traces in abstracted, or non-
sensory, form

hippocampusz), recency judgements (posterior hippocampus), visuospatial learning
(medial entorhinal, anterior hippocampus), as well as the association of salient

simultaneous or temporally-dispersed stimulus features (declarative associative learning
and contextual learning: perirhinal, perihippocampal, entorhinal and ectorhinal cortices

and posterior hippocampus) (Winters and Bussey 2005c; Witter and Moser 2006;

 

2 The ﬁmctional neuroarchitectural divisions in question differ from quadripeds to bipeds. The ﬁmctional
differentiation of the hippocampus, originally cited as following a “dorsoventral” axis, has more recently
been retooled as following a “septotemporal” axis, since the rodent dorsal hippocampus corresponds to the
primate posterior hippocampus, whereas the rodent ventral hippocampus corresponds to the primate
anterior hippocampus. See Moser, M. B. and E. I. Moser (1998). Functional differentiation in the
hippocampus. Hippocampus 8(6): 608-19.

Dudukovic and Wagner 2007; Kerr, Agster, Furtak and Burwell 2007; Kumaran and
Maguire 2007). The individual medial temporal areas elude ascription to particular
computational operations, however, due not only to the difficulty of constructing tasks
that dissociate suppositional functions from each other, but also by the intricate
interconnectivity, and therefore interdependence, of these areas (Preston and Gabrieli
2002).

The structure of the perihippocampal region consists of the entorhinal cortex,
Brodmann areas 28 and 34; medial (area 35) and lateral (area 36) perirhinal cortex; and
the parahippocampal cortex. Information ﬂows from isocortex to the perirhinal and
parahippocampal cortex, to the ERC, dentate gyrus, CA3 field, and subiculurn, and back
to isocortex. Information is initially processed in isocortex and then passed through the
entorhinal (ERC), perirhinal (PRC), and parahippocampal (PHC) cortices to the
hippocampal complex, where it is presumably optimized. After this Optimization, the
information is returned to the isocortical area of origin, resulting in long-term memory
consolidation. Parahippocampal lesions cause deﬁcits as severe as hippocampal lesions

alone, yet the role of areas surrounding the hippocampus, such as the entorhinal cortex, is

 

Table l Projections to the perihippocampal regions (Lavenex and Amaral 2000).
Abbreviations: PPC, posterior parietal [association] cortex; SSC, somatosensory cortex;
OF C, orbitoﬁontal cortex; PFC, prefrontal cortex; ST S, superior temporal sulcus; TE, V4,
visual areas TE and V4.

 

Modality Projection
TE (visual object information) PRC
PPC and V4 (visuospatial) PHC
880 Both
Ventrolateral cortex, OFC, cingulate, Both

retrosplenial cortex, PFC, dorsal STS
(polysensory cortices)
Auditory association cortex PHC

 

distinct from the role of the hippocampus itself. For example, ERC+PRC lesions prevent
the formation of mnemonic representations of visually paired associates in the
inferotemporal cortex. However, different areas of isocortex have different projections
to the perihippocampal region as well, resulting in sensory speciﬁcity in many areas of
the MTL (see Figure 1 and Table 1). Feedback, feedforward, auto- and reciprocal
connections result in a conversion Of unimodal information to polymodal, supermodal, or
amodal information, and allow abstractions in the HC complex before being returned to
isocortex.

MTL afferents to unimodal sensory cortex can interact with lower level
information processed or stored in unimodal cortex. Efferent projections ﬁom entorhinal
and perirhinal cortex are necessary for maintenance and consolidation of pair codings in
auditory area TE and may be responsible for short-term memory maintenance (since
these areas activate during the delay phase of a delayed nonmatch to sample, or DNMS,
task). Perirhinal neurons are multimodal and probably encode more than sensory
attributes. The entorhinal cortex is less involved in object/sensory analysis and coding
than is the perirhinal cortex. These neurons may respond only to simultaneous activation
from multiple (multimodal) areas; the ERC, for example, is thought to respond to
“compound” stimuli. ERC and PRC 0 patterns are coherent during REM sleep, which is
thought to participate in memory consolidation. The 0 origin is probably driven by the
ERC and/or subiculum. The PRC does not get input from the medial septum, the
HC/ERC pacemaker, but it does receive inputs ﬁom the neurons in the deep ERC, which

have intrinsic 0 ﬁring patterns (Lavenex and Amaral 2000). The hippocampus, in

contrast, responds only to supermodal or amodal information and returns abstracted

representations to the originating areas of isocortex.

Role of other memory systems in episodic retrieval

Three non-declarative memory systems — the perceptual representation system,
procedural memory, and working memory — may also interact with the declarative
memory system, depending on task demands (see Table 2).

The perceptual representation system (PRS), memory for word and Object
identiﬁcation, can be subdivided into three subsystems: the visual word form subsystem
(extrastriate visual cortex and temporal and frontal cortex), which appears to process
orthography, the auditory word form subsystem (superior temporal and ﬁ'ontal cortex),
which processes auditory information, and the structural description subsystem
(occiptotemporal regions, e. g. inferior temporal and fusiform gyri), which processes parts
of objects and matches them to global structure. Although the PRS has been functionally
and anatomically dissociated from declarative memory through PET and lesion studies, it
is possible, if not likely, that a declarative memory task involving word or visual
recognition (such as the task proposed in the present dissertation) would also engage the
PRS (Eichenbaum et al., 1996; Gabrieli, 1998).

Procedural memory is memory for habits and skills. Procedural memory does not
recruit the MTL and is almost therefore anatomically independent of declarative
(semantic and episodic) memory. Neuropsychological and pathological studies of

amnesic patients have revealed that procedural memory is dependent upon a

10

corticostriatal system, as well as the motor cortex and basal ganglia (Eichenbaum et al.,
1996; Gabrieli, 1998).

Working memory enables short-term retention of information over several seconds
(Eichenbaum et al., 1996; Gabrieli, 1998). Like the PRS, working memory can be
anatomically and functionally categorized into two slave subsystems: the phonological
loop (auditory and parietal cortex), which contains auditory information for working
memory, and the visuospatial sketch pad, which is anatomically composed of the visual
association cortex, inferior parietal lobule, and inferior preﬁ'ontal cortex (Eichenbaum et

al. , 1996).

ll

 

Table 2 Regions typically involved in episodic retrieval tasks. LH, left hemisphere; RH,
right hemisphere; BL, bilateral. Several of these regions will be used as target ROIs in the
present study (see next chapter).

Working Maintenance Precentral sulcus (BL)
memory VISUOSPATIAL SKETCH PADZ:
Visual association cortex
Inferior parietal lobule
Inferior prefrontal cortex (area 45)
Inferior frontal gyrus (BL)
Anterior frontal gyrus (LH)
Medial frontal gyrus (RH)
Phonological loop.
Parietal cortex
Auditory cortex

Recognition Precentral sulcus (BL)
Inferior frontal gyrus (BL)
Posterior middle frontal gyrus (BL)
Middle frontal gyrus (BL)
Anterior prefrontal cortex (LH)

Long-term Retrieval Entorhinal cortex (9 source)3
memory Medial septum (9 source)
Perihippocampal cortex3
Perirhinal cortex3
Dorsal hippocampus5
Right frontal cortex6

PRS Recognition AUDITORY WORD FORM SUBSYSTEMZ:
Superior temporal gyrus
Right frontal cortex
STRUCTURAL DESCRIPTION SYSTEM:
Inferior temporal gyrus
Fusiforrn gyrus

;,(Ranganath Johnson etal. 2003)

:(Eichenbaum, Schoenbaum etal. 1996)

:(Lavenex and Amaral 2000)

:(Chrobak and Buzsaki 1998b)

Z,(Bannerman Rawlins, McHugh, Deacon, Yee, Bast, Zhang, Pothuizen and Feldon 2004)
:‘,(Buckner Koutstaaletal1998b)

7(Haxby, Honrvitz,UJeI1eider,Maisog, Pietrini and Grady 1994)

 

12

REGIONS OF INTEREST

Regions of interest (ROIs) for the present study were delineated according to the AFNI
implementation of the Talairach-Tourneaux human brain atlas (Talairach and Tournoux
1988) by warping the individual datasets to Talairach space. In this section, the
signiﬁcance of the various ROls, according to recently published literature, will be
discussed in order to place them into the context of the previously-stated hypotheses.

Ten ROls were selected for the present study based on their functional
signiﬁcance as long-term memory, visual, or motor regions. Four, including
hippocampus and entorhinal, ectorhinal, and perirhinal cortex, constitute the majority of
the MTL with the exception of the periamygdaloid area (also known as the semilunar
gyrus), posterior PHG, the prepyriform area (also called piriform cortex; laterally
bounded by the retrosubicular area 48), and the presubiculum, which are not known to
play a pivotal role in declarative memory storage or recall. An additional four ROIs,
including visual areas V1, V2, and V3 and the fusiform gyrus, were included in order to
facilitate sampling from lower (V1 and V2) and higher-order (V3 and fusiform) visual
areas, particularly in the ventral visual stream. Primary motor (M1) cortex and premotor
(PMC) cortex were also included as control and motor attention areas (see below),

respectively.

MEDIAL TEMPORAL AREAS: LONG-TERM MEMORY R018

The hippocampus (HC), entorhinal cortex (ERC: Brodmann areas 28 and 34), ectorhinal

cortex (EcRC: Brodmann area 36), and perirhinal cortex (PRC: Brodmann area 35) were

13

identiﬁed as regions of interest for the present study based largely on the
neuroanatomical and neuropsychological studies of Lavanex and Amaral (2000),
Eichenbaum and colleagues (1996), Gabrieli (1998), and Nadel and colleagues (2000).
The functional neuroanatomy of these long-term memory regions is further described in
the introductory chapters. To review: information in conveyed from isocortex to the
perirhinal and parahippocampal regions, depending on the modality of the stimulus; PRC
appears to be specialized for visual object information, whereas PHG seems to be related
to the storage and conveyance of visuospatial information. It has been suggested that
PRC and PHG are necessary for the maintenance and consolidation of pair codings, and
as such are of particular interest in the present study. Furthermore, current anatomical
models suggest that information exchange takes place between the PRC/PHG and HC,
parietal cortex, anterior cingulate, and the insula (Lavenex and Amara12000). The PRC is
located along the banks of the rhinal sulcus and is bounded medially by entorhinal area
28 and laterally by ectorhinal area 36; the PHG is deﬁned for the purposes of this
dissertation as consisting of the periamygdaloid area (also known as the semilunar gyrus),
posterior PHG, the prepyriform area (also called piriform cortex; laterally bounded by the
retrosubicular area 48), and the presubiculum, and is not included as an ROI in the

present dissertation.

Output from the PHG/PRC region is sent to the ERC (Lavenex and Amaral,
2000), which also interacts with parietal association cortex (PAC) for spatial perception
and cognition. The ERC in turn sends its output to the dentate gyrus (in the
hippocampus). The function of the ERC is not well understood, however, except that it

plays a role in all phases of long-term memory acquisition, consolidation, and recall, and

14

seems to function in the pairing of stimuli that are temporally dispersed (Egorov,
Hamam, Fransen, Hasselmo and Alonso 2002; Frank and Brown 2003). The ERC is
divided into Brodmann areas 28 and 34; both areas are located on the medial-caudal

aspect of the temporal lobe.

Information is sent from the hippocampal dentate gyrus to the CA3 ﬁeld and
ﬁnally to the subiculum (also in the HC), which returns abstracted information to
isocortex. The nature of abstraction in the hippocampal complex is not well understood;
however, it is believed that the HC responds only to supramodal or amodal information,
such as salient uniﬁed features of a memory, and it is this supramodal and amodal
information that is returned to cortex. As discussed previously, the HC itself has been
shown to activate during acquisition, consolidation, and recall, although the time course
of its inﬂuence upon these phenomena is a matter of controversy (Eichenbaum,
Schoenbaum et al. 1996; Nadel and Land 2000; Nadel, Samsonovich, Ryan and

Moscovitch 2000).

The ectorhinal cortex (EcRC: area 36) has been shown to have backprojections
onto V1 and is heavily interconnected with area TE in simians (Naya, Yoshida and
Miyashita 2003a; Naya, Yoshida, Takeda, Fujimichi and Miyashita 2003b; Clavagnier,
Falchier and Kennedy 2004). It appears to be part of a visual long-term memory system
(along with other visual areas) in these higher primates, but its role in humans has not
been decisively determined. Extrapolation from animal data, however, suggests that area
36 may also participate in long-term visual memory in humans. EcRC is located

primarily in the fusiform gyrus (this area is excluded from the fusiform gyrus ROI in the

15

(A) Unimodal association cortex,

   
   
 

polymodal association
cortex, frontal cortex,
cingulate, V4

Polymodal association cortex

I mugt,‘r
ERC(34) I? ' ‘ ‘
AA...
' 3,,

   

Orbitofrontal ,
primary motor (M1),
primary visual (V1),
primary
somatosensory
(S1), supplementary
somatosensory
(SSC), and visual
association cortex

  

Unimodal association cortex,

polymodal association cortex ,
frontal cortex

(V3)

(B)
FRONTAL POLE
Entorhinal ctx.

   

Entorhinal ctx. (BA 28)
(BA 28 and 34). Ectorhinal ctx.
(BA 36)

Hippocampus

Peri r‘h1' nal ctx.

(BA 35) Ectorh1nal ctx.

(BA 36)
LH RH
OCCIPITAL POLE

Figure 2 (A) Interconnections among constituent regions of the MTL and other areas of
interest in the present study. Blue arrows indicate neuronal projections between areas as
determined by anatomical studies of the cortex and temporal lobe. Arrow locations do not
accurately reﬂect the anatomical location of interregional projections (Sources: Deacon,
Eichenbaum, Rosenberg and Eckmann 1983; Schacter, Buckner, Koutstaal, Dale and
Rosen; Burwell and Amaral; Burwell and Amaral; Lavenex and Amaral; Save and Poucet
2000a; Save and Poucet 2000b). (B) The anatomical locations of MTL areas in the
Talairach-Toumeaux template brain, obtained ﬁom the AFNI suite’s Talairach Daemon.
Abbreviations: LH, leﬁ hemisphere; RH, right hemisphere; ctx., cortex; BA, Brodmann
area.

16

present study) and is bounded laterally and caudally by inferotemporal area 20, medially

by PRC area 35, and rostally by temporopolar area 38.

MTL areas, which, again, serve to associate salient features Of encoded
declarative stimuli, also appear to be affected by consolidation processes. Intuitively, this
would mean that changes in plasticity would drive the amalgamation of pre-consolidation
medial temporal interneurons serving to link memory traces stored in isocortex, such that

fewer interneurons are required for the linkage at post-consolidation (see Figure 2).

PREMOTOR CORTEX AS A VISUOSPATIAL ATTENTION AREA

Brodmann area 6 has been cited as an attentional area since its activation is strongly
correlated with tasks involving covert shifts of attention, and particularly in tasks that
demand shifts in visuospatial attention (Nobre, Sebestyen, Gitelman, Mesulam,
Frackowiak and Frith 1997). Furthermore, area 6 has been subdivided into medial and
lateral regions, which are differentially activated by verbal and spatial memory
sequencing tasks, respectively. This ﬁnding was also supported by a repetitive
transcranial magnetic stimulation (TMS) inference study, in which TM stimulation to the
medial Brodmann area 6 selectively disrupted verbal but not spatial cognitive processing,
whereas TM stimulation to the lateral Brodmann area 6 disrupted spatial but not verbal
cognitive processing (Tanaka, Honda and Sadato 2005). Area 6 comprises parts of the
superior and middle ﬁ'ontal gyri as well as the rostral precentral gyrus and is bounded

medially by the cingulate sulcus and laterally by the lateral sulcus.

l7

VISUAL AREAS

It is beyond the scope of this dissertation to discuss in depth the ﬁmction and history of
research into visual cortex. Here, the ﬁmctional architecture of the visual streams from

V1 to V5 will be highlighted (with the exception of V4/1PL, discussed above). For a

more thorough review of human visual neuroanatomy, the reader is referred to the recent

review by Wichmann and Muller-Forell (Wichmann and Muller-Forell 2004); the
seminal paper on the dorsal and ventral streams by Ungerleider and Mishkin (Ungerleider

and Mishkin 1982) is also a good reference.

Visual area VI , located in Brodmann area 17, gets input from the thalamic lateral
geniculate nucleus (LGN) and responds selectively to (i.e., its cells are tuned to)
orientation-speciﬁc contours and (in some cases) respond to color (Hubel and Wiesel
1959; Hubel and Wiesel 1962). V1 exhibits complex within-region circuitry, as well as
interactions with the LGN, in addition to visual areas V2 and MT/VS, through
unidirectional and reciprocal projections, which may serve simple attentional ﬁrnctions in
visual selection (Livingstone and Hubel 1984b; Livingstone and Hubel 1984a; Albright
1993). V1 is bounded by Brodmann area 18 and surrounds the calcarine ﬁssure. It is

anatomically identiﬁable by the presence of the bands of Gennari.

Visual areas V2 and V3, located in Brodmann areas 18 and 19, respectively, are
the ﬁrst of many so—called extrastriate visual areas. V2 shares both feedback connections
with area V1 and feedforward connections with V3, V4/IPL, and V5/MT. Neurons

comprising area V2 are, like area V1, tuned to feature orientation and color, but

18

additionally have ﬁgure-ground tuning (Qiu and von der Heydt 2005). V2 is also
interconnected with areas V4 and MT, and exhibits some attentional modulation
(Albright 1993; Rao 2005). Visual information is relayed from V2 to V3 as part Of a
ventral visual stream (vVS), which has a role in object recognition, and from V2 to
V5/MT as part of a dorsal visual stream (dVS), which functions in spatiotemporal tasks
(Ungerleider and Mishkin 1982). V3 (Brodmann area 19), part of the vVS, is tuned to
aspects of visual motion (Braddick, O'Brien, Wattam-Bell, Atkinson, Hartley and Turner
2001). Area V3 has strong interconnections with both area V2 and area V4/IPL (Albright
1993). Brodmann area l8/V2 is bounded by area l7/V1 and area 19/V3; area 19/V3 is

further bounded by areas 37 and 39.

The ﬁrsiform gyrus is a functionally heterogenous area most recently shown to be
involved in face processing (Kanwisher, McDermott and Chun 1997; Haxby, Gobbini,
Furey, Ishai, Schouten and Pietrini 2001; Haxby, Hoffman and Gobbini 2002; Rossion,
Schiltz and Crommelinck 2003; Yovel and Kanwisher 2005). However, it also seems to
be involved in other aspects of visual processing and computation, including visual
working memory, lexical tasks, and mental manipulation of orientation of visual stimuli
(Orban and Vogels 1998; Pammer, Lavis and Comelissen 2004; Uncapher and Rugg
2005a; Uncapher and Rugg 2005b). The fusiform gyrus consists of ectorhinal area 36
(which was included as the EcRC ROI but excluded from the fusiform ROI in the present
study) as well as area 37. Area 37 is bounded caudally by area 19/V 3, rostrally by area

20/V 5 and area 21, and dorsally by area 39.

19

PRIMARY MOTOR CORTEX: A CONTROL AREA

Finally, primary motor cortex (Ml, Brodmann area 4) was included in the analyses in
order to assess its connectivity with cortical and subcortical areas during our episodic
retrieval task. It has been shown that increased within-M1 neural ﬁring rate coherence is
related to corticospinal (descending voluntary) control of muscle movement, and,
furthermore, that this coherence is further correlated with coherence between M1 and
motor neurons in the beta frequency range (Gerloff, Braun, Staudt, Hegner, Dichgans and
Krageloh-Mann 2006). Area 4/Ml is bounded rostrally by premotor area 6 and caudally

by area 3.

20

F MRI SPATIAL AND TEMPORAL SPECIFICITY

The present study endeavors to assess changes in cortical recruitment, metabolic
demands, and FMRI signal entropy and rate of change as correlates of memory
consolidation. In order to formulate a proper methodology to perform this assessment,
however, the arguments must be made that (a) the BOLD signal is substantially coupled
to, and co-registered with, neural activity; (b) that BOLD-FMRI is satisfactorily
colocalized to neural activity, so one may infer details about the underlying neural
activity; and (c) that one may extrapolate, based on certain features of the F MRI dataset,
where in the brain the T2*—weighted signal change originates (i. e., at which locations the
BOLD signal arises from increased metabolic activity as opposed to passive diffusion of
deoxyhemoglobin). This chapter, then, will turn to the principal methodology of the
present dissertation, functional magnetic resonance imaging (FMRI). The underlying
physiology of the F MRI Signal will be examined and the implications of this physiology
for investigations of ﬁrnctional neural architecture, as well as the spatial and temporal
resolution of the FMRI method and its colocalization with aspects of neural

electrophysiological phenomena, will be highlighted.

BOLD SIGNAL PHYSIOLOGY

FMRI is an extension of anatomical MRI that relies on a T2*-weighted, blood

oxygenation-level dependent (BOLD) signal (Ogawa, Lee, Kay and Tank 1990) that

21

arises indirectly from the metabolic sequelae of postsynaptic potentials, particularly local
ﬁeld potentials (LFPS), which correspond to cooperative population activity
(synchronized somatodendritic potentials: see Logothetis, Pauls, Augath, Trinath and
Oeltermann 2001; Logothetis 2002; Logothetis 2003; Logothetis and Pfeuffer 2004;
Logothetis and Wandell 2004). LFPS ﬂuctuate relatively slowly; the magnitude of this
ﬂuctuation is not related to cell size but with the geometry and extent of dendritic
arborization. Thus, the parallel arrangement of apical dendrites heavily inﬂuences LFP
measurements due to their “open ﬁeld” arrangement. LFPS reﬂect a net effect of cellular
events — synaptic events, voltage responsive membrane oscillations,
afterhyperpolarizations, and aﬁerdepolarizations due to Kca channels (Logothetis, Pauls
et al. 2001). Anatomically, the density of brain vasculature seems to be related to the
location of perisynaptic elements, rather than the location of neural somata, which
suggests that cerebral blood ﬂow (CBF) is related to neurotransmitter release rather than
perikaryal metabolic demands. This is further veriﬁed by increases in CBF only with
orthodromic, not antidromic, microstimulation. Logothetis (2001) has noted that glucose
uptake, the postsynaptic effects of glutamate, and the restoration of ion gradients after an
action potential — all of which can be categorized under the heading “synaptic events” —
are the most energy-consuming processes in the brain and drive CBF changes. Of these,
presynaptic and postsynaptic currents drive the energy demand and are the dominant
elements in LFPS (Logothetis et al., 2001).

Just as cortical areas are delineated anatomically by differences in cyto- and
myeloarchitecture, they can be delineated functionally by differential physiological

properties (electrophysiology) and connectivity (neuroimaging, tract tracing).

22

Neuroimaging methods like FMRI have the advantage over invasive methods in that they
permit longitudinal studies (e.g. on learning, plasticity, reorganization). The BOLD
contrast signal used in FMRI is derived from magnetic ﬁeld inhomogeneities caused by
differential deoxyhaemoglobin (de) concentrations. de is conﬁned to the intracellular
Space of erythrocytes, which are in turn conﬁned to the vasculature. Thus, the ﬁeld
gradient responsible for the BOLD signal is a result of the susceptibility differences
between the de-containing compartments and the surrounding space. FMRI pulse
sequences are sensitive to these differences and alter the signal when de concentration
changes. Upon neural activation the increase in de concentration would be expected to
enhance the difference in magnetic susceptibility, thereby decreasing the BOLD signal;
however, the exact Opposite occurs. This is due to an increase in CBF that
overcompensates for the increase in oxygen uptake: an oversupply of oxygenated blood is
delivered to the site (Raichle 1998). Thus, changes in CBF corresponding to neural

processes are reﬂected by an increase in the BOLD signal.

SPATIOTEMPORAL RESOLUTION OF F MRI

The development in the late 19803 of high-ﬁeld (1.5T) MR imaging techniques (echo-
planar imaging and spiral imaging) and discovery of two endogenous “contrast”
mechanisms — BOLD (Ogawa, Lee et al. 1990) and differential net longitudinal
magnetization with changes in tissue perfusion — initiated a revolution in ﬁrnctional
neuroimaging, largely replacing paramagnetic tracer technology for in-depth studies of

cognition. The temporal relationship of neural and vascular events (neurovascular

23

coupling) was actually ﬁrst described by Angelo Mosso in 1881, but the nature of this
relationship has only recently been deﬁned (Logothetis et al., 2001). Whisker barrel
stimulation in murine models demonstrated that CBF increases within approximately 2
sec of stimulation onset and peaks within 5-7 sec. However, paramagnetic tracer studies
have revealed that gross cerebral volume changes actually lag behind changes in CBF,
likely due to a “capacitance system” in the capillaries and post-capillary beds, and
I-Ib/HbOz changes even precede CBF changes. Thus, changes in hemoglobin
concentrations, local CBF, and gross CBF are related temporally but perhaps not
causally. The implication of the BOLD signal technique is that the block design, which
was necessary for single-photon and positron emission tomography (since they require a
quasi-equilibrium state of > 60 sec), became superﬂuous for MRI, and event-related
designs, such as the one proposed in the present dissertation, were made possible.
Consequently, the short history of fast (event-related or ER)—ﬂV1RI has been characterized
by investigations of ever-smaller sampling windows. Blamire et al. demonstrated
detectable signal changes after stimulus durations as short as 2 seconds; Bandettini et al.
decreased this window to as little as 500 msec (Bandettini, Wong, Hinks, Tikofsky and
Hyde 1992; Blamire, Ogawa, Ugurbil, Rothman, McCarthy, Ellermann, Hyder, Rattner
and Shulman 1992). More recently, Savoy et al. established that the BOLD signal could
be detected after a 34 msec-duration visual stimulus and Boynton et al. showed that
responses can be summated linearly over time (with some limitations) (Savoy 2005). The
computability of a hemodynamic response function (HRF) is directly related to the
strength of the ﬁeld; as ﬁeld strength increases (e.g. 1.5T to 3T) the signal—to-noise ratio

(SNR) increases. In the lower magnetic ﬁelds, time series can be reconstructed with (a)

24

overlap correction methods similar to those used in ERP research and (b) averaging
across trials, which brings out the modal tendencies in the HRF. Rosen et al. note that
temporal sensitivity is a ﬁlnction of the SNR (and by implication, ﬁeld strength) as well
as the haemodynamic response (HRF) variance and the methods used to characterize it
(Rosen, Buckner and Dale 1998). They also report that variance in HRF latency within
and between brain regions is not merely a matter of vasculature diameter. Although
larger vessels do exhibit longer latencies (a signiﬁcant factor within regions such as V1)
the longest delays are seen in anterior PFC, an area with a minimum of large vessels.
Savoy et al. (1995) have reported latencies of 500 msec in V1; Buckner et al. report
latencies of 500 to 1000 msec between visual and prefrontal areas; and at the other
extreme, Schacter et al. note delays of 2 to 4 sec between various preﬁ'ontal regions
(Buckner, Bandettini, O'Craven, Savoy, Petersen, Raichle and Rosen 1996; Schacter,
Buckner et al. 1997; Savoy 2005).

The spatial resolution of F MRI has been examined most closely with manganese
contrast, which is paramagnetic and emphasizes the calcium-dependent sites of neural
activity. Manganese enters voltage-operated calcium channels and mimic calcium’s role
in neurotransmitter release. In a recent study with manganese MRI contrast in rats,
Duong et al. demonstrated a co-registration of < 200 um between synaptic activity and
BOLD contrast, suggesting that the BOLD signal (at high ﬁelds [9.4T]) resolves synaptic
activity to a satisfactory degree (Duong, Silva, Lee and Kim 2000). It is therefore

reasonable to conclude that the BOLD signal satisfactorily exempliﬁes the localization of

[post]synaptic activity.

25

MODELING THE HEMODYNAMIC RESPONSE FUNCTIONS

Aguirre et al have investigated the variability in HRF shape and timing within subjects
(different brain regions), across subjects, and at different times of the same day or on
different days (Aguirre, Zarahn and D'Esposito 1998). Formerly the best estimates for
the HRF were (I) a Poisson distribution (Friston et al, 1994) or (2) a gamma distribution
with two free parameters and a phase delay (Boynton et al, 1996). It was determined that
the mean tirne-to-peak in subjects was 4.7:t1.1 sec and ranged from 2.7 to 6.2 sec; HRFS
across subjects were signiﬁcantly more variable than those within a subject, even across
several days. Additionally, the Poisson ﬁinction accounted for only 25% of the HRF
variances; the gamma function accounted for 70%, which was a signiﬁcant difference (p
< 0.001). When a single subject-speciﬁc HRF estimate was used (generated empirically
from the single-subject data) approximately 96% of variance was accounted for.
Therefore, the authors conclude that an empirical estimate of the HRF is much more
desirable than an a priori estimate. For blocked ﬂVIRI paradigms, in which the HRF is
oversampled, an a priori estimate will not decrease statistical sensitivity signiﬁcantly;
however, the authors note that an empirical HRF lends to the analysis a measurable
increase in sensitivity — as it will for event-related designs (which are much more
sensitive to the statistical decrement imposed by the a priori ﬁrnction). Therefore, the
HRF varies within and between subjects and a deconvolution analysis (discussed in the
Methodology chapter) will be utilized to overcome this variance in the present

dissertation.

26

AIMS AND THEORETICAL TREATMENT OF CONSOLIDATION
NEURODYNAMICS

A paired-associate task (described in the Methodology chapter) was utilized to examine
changes in task performance and BOLD physiology with episodic memory consolidation.
This task required subjects to learn, via a computer-based training session, to associate
visually presented words with abstract pictures, then identify correctly matched,
incorrectly-matched, and control condition pairings during a testing session, which took
place within an MRI magnet. Two FMRI time points were established by dividing the
subject pool into a pre—consolidation time point group (TPI), whose constituents were
tested in the MRI magnet immediately after the training session, and a post-consolidation
time point group (TPz), who were tested in the MRI magnet seven days after the initial

training session.

This paradigm, coupled with a non-invasive, systems-level approach, ﬁrnctional
MRI (FMRI), enabled the assessment of changes in subject memory performance and
physiology with consolidation. Although many FMRI investigations of mnemonic
functions have been performed, a majority has dealt with modular cortical activity
correlated with a given memory function; fewer are the studies of such memory ﬁinctions
as a system (e.g., a recent analysis of cortical connectivity during episodic encoding by
Summerﬁeld, Greene, Wager, Egner, Hirsch and Mangels 2006), and none, to the
knowledge of the author, have attempted to analyze differences in the FMRI signal

entropy or signal response rate in episodic memory areas with consolidation.

27

This scarcity of neuroimaging-based mnemonic neurodynamics research is likely
due to the indirect relationship between neural activity and the F MRI signal: the BOLD
signal, upon which most FMRI studies are founded, is an epiphenomenon of glycolytic
activity concomitant to synaptodendritic metabolic demands (Ogawa, Lee et al. 1990).
As such, neural activity and the BOLD signal are said to be related through a black box
or “transfer function” of unknown parameters, and these parameters vary both across
brain areas within a given subject as well as across subjects and task demands.
Nevertheless, the correlation between the BOLD signal and neural activity (more
Speciﬁcally, the local ﬁeld potential phenomenon) has been shown to exhibit a rough
linearity across low to moderate metabolic magnitudes, such that neuroimaging research
may generate probative, albeit admittedly indirect, neurodynamics data (Logothetis, Pauls

et al. 2001 ).

CONSOLIDATION AS AN OPTIMIZING PROCESS

A discussion of the theoretical basis of the hypotheses of the present dissertation iS best
approached through a conceptualization of the episodic memory system as a self-tuning
network that aims to minimize its energy expenditure relative to the storage and recall of
a given memory trace.

For the purposes of the present dissertation, retrieval efﬁciency was deﬁned as the
Speed, recruitment, and retrieval success or eﬂicacy of the mnemonic (long-term
memory) and supporting (working memory, attentional and executive, and visual and
motor) networks recruited during trace recall. Retrieval “speed” and “success” were

characterized behaviorally by subject response time and accuracy, respectively, on the

28

paired associates task, whereas retrieval-concomitant “recruitment” was characterized
physiologically by the degree of cortical recruitment and metabolic demand relative to
subject retrieval speed and success. It should be noted that this deﬁnition of“eff1ciency”
is closely related to the use of the term in thermodynamics in that it is conceptually a
ratio of useful output (behavioral trace retrieval performance) versus energy expenditure

(metabolic intensity per unit recruited volume):

 

 

performance or performance
recruitment PC /AV '

where PC represents the percent BOLD signal change from baseline and A V represents
the degree of cortical recruitment (activation volume). Although it is beyond the scope of
the present work to attempt to typify this ratio as a quantitative index of neural systems
efﬁciency (although such attempts have been made: see Chance 2007), this
conceptualization of retrieval efﬁciency was used exhaustively in the origination of the
theory and hypotheses in the current study.

It is imperative to note that, by this deﬁnition, increased efficiency over a
consolidation interval could be argued to have occurred in any of several cases. Namely,
the output-to-energy expenditure ratio would increase if

(a) subject performance on a recall task improved without a corresponding decrease
in metabolic demand or cortical recruitment;

(b) metabolic demand and cortical recruitment decreased without a corresponding
improvement in subject performance;

(c) metabolic demand decreased without corresponding improvements in cortical

recruitment or subject performance; or

29

(d) cortical recruitment decreased without corresponding improvements in metabolic

demand or subject performance.

EVIDENCE FOR INTERIM PHYSIOLOGICAL CHANGES IN RETRIEVAL SUBSTRATES

Prior behavioral, electrophysiological, and neuroimaging studies of memory performance
in well-trained versus poorly-trained subjects have in fact provided some evidence that
performance on episodic memory tasks are dependent upon the “attunement” of
perceptual and trace retrieval circuits toward task Speciﬁcity, i.e., the identiﬁcation of
previously learned stimuli, and, therefore, do exhibit some hallmarks of optimization
(Tulving 1983; Weldon, Roediger et al. 1989; Rugg and Wilding 2000; Robb and Rugg
2002; Wheeler and Buckner 2003). In particular, well-trained subjects appear to exhibit
an enhanced capacity for quickly identifying, through covert attentional mechanisms,
distinctive characteristics of presented stimuli in order to expedite retrieval of those
features from memory; and, ﬁthhermore, such subjects also exhibit an enhanced capacity
for reactivating, upon recall, the cortical networks recruited during encoding. These
phenomena have become known as retrieval orientation and cortical reinstatement (or
‘transfer-appropriate processing”) effects, respectively (Lockhart 2002; Vaidya, Zhao,
Desmond and Gabrieli 2002; Herron and Rugg 2003b; Herron and Wilding 2004;
Homberger, Morcom and Rugg 2004; Herron and Wilding 2006a; Homberger, Rugg and
Henson 2006b; Mulligan and Lozito 2006; Stenberg, Johansson and Rosen 2006;
Woodruff, Uncapher and Rugg 2006). Clearly, such changes in perceptual and retrieval

circuits, if they exist, are relevant to an understanding of pre- and post-consolidation

30

retrieval efﬁciency; however, these phenomena have not yet been examined in a time-
dependent or consolidation context.

It is logical to assume that the shifts in recall neurophysiology predicted by the
retrieval orientation and cortical reinstatement phenomena, as well as the Hebbian
changes in neuronal efficacy that would ostensibly cause them, would likely occur both
within the trace network and the visual cortices recruited for future recognition of the
stimuli the traces represent. To the knowledge of the author, the ideal test of this
assumption — the presence or absence of retrieval orientation and/or cortical reinstatement
effects in the context of a MTL lesion — has not been performed. Given the robust
evidence for close connectivity between episodic and attentional brain regions, however,
it is unlikely that these phenomena are asymmetric across the regions that mediate
attention to remembered stimuli and the memory traces which drive them (Chun and
Turk-Browne 2007); therefore, correlates of retrieval orientation and cortical
reinstatement were expected in medial temporal, attentional, and visual areas, except
where direct descending attentional input is lacking (e. g. Vl: Rao 2005).

It was of particular interest in the current work to substantiate or refute retrieval
orientation effects in the right hemisphere fusiform gyrus and ventral visual stream,
where such effects have been reported (Vaidya, Zhao et al. 2002; Woodruff, Uncapher et
al. 2006). The ventral, or object identiﬁcation, visual stream and fusiform gyri,
especially in the right cortical hemisphere, are believed to subserve more holistic or
“global” feature processing (Brown and Kosslyn 1993; Johannes, Wieringa, Matzke and
Munte 1996; Yamaguchi, Yamagata and Kobayashi 2000). Therefore, three predictions

were made:

31

I) that increased synaptic weights within visuospatial networks concomitant to
the cortical reinstatement phenomenon would increase the efﬁcacy of
neuronal ﬁring, such that the BOLD signal rates of all areas (except primary
visual and primary motor cortices) would be accelerated (H. 3);

2) that post-consolidation subjects would exhibit shifts in the laterality of
metabolic demands in extrastriate areas relative tO pro-consolidation subjects,
thereby evincing changes in visual attention with consolidation (H. 4); and

3) that increased synaptic weights within visuospatial networks concomitant to
the cortical reinstatement phenomenon, as well as the differential modulation
of the activity of those networks by descending attentional neurons due to
retrieval orientation effects, would moderate the requisite amount of
information processing in all extrastriate visual areas, such that both left- and
right-hemisphere post-consolidation visual cortices would exhibit lower

metabolic demands than at pre-consolidation (H. 4).

COMPUTATIONAL EFFICIENCY AND CORTICAL DEMANDS: ACTIVATION VOLUME AND
ENTROPY
A fundamental assumption in the present study is that the number of neurons required to
accomplish an encoding or computational function decreases as the encoding or
computational efficiency of a neuron or population of neurons increases, an assumption
that is supported by molecular, electrophysiological, and neuroanatomical evidence that
repeated neuronal activity drives changes in proximal and distal dendritic morphology,

which in turn effect changes in synaptic integration and therefore the computational

32

functions of the neuron (Holmes and Levy 1990; Zador, Koch and Brown 1990; Bailey,
Chen, Keller and Kandel 1992; Bailey, Montarolo et al. 1992; Yuste and Denk 1995;
Mallot and Giannakopoulos 1996; Shepherd 1996; Zhang, Endo et al. 1997; Hausser,
Spruston and Stuart 2000; Verzi 2004; Frith 2005; Segev 2006). Thus, if memory
consolidation is a process of mnemonic network optimization, the number of recruited
neurons (or voxels, in the present case) would be expected to decrease. Two important
exceptions in the present study are the premotor cortex, which appears to subserve
attentional functions, and primary motor cortex, which subserves the motor response.
Neither ROI was expected to differ with respect to recruitment nor metabolic demands

with consolidation, since no changes in their computational ﬁinctions were expected.

In the present study, therefore, the spatial extent of BOLD signal activation is
quantitatively assessed through statistical analysis of the activation volume (in mm3) per
region of interest. However, activation volume measures are susceptible to noise in the
MR magnet environment and intersubject variability (Saad, Ropella, DeYoe and
Bandettini 2003), and furthermore, it is not informative with respect to the nature or
degree of information processing in signals ﬁ'om the neural populations (remnant voxels)
that remain correlated with the task after consolidation. Moreover, interpretation of
observed decreases in activation volume is further complicated by the indirect
relationship between neuronal activity and the BOLD signal which underlies the
deﬁnition of “activated” voxels. A change in activation volume may, however, be due to
any number of factors — internalization of salient stimulus features, circuit multiplexing,
or increased neuronal computational efﬁciency — but in the absence of data indicating

that these changes in recruitment are accompanied by increases in (or stability of) the

33

task-speciﬁcity and magnitude of neuronal activity, it is equally justiﬁable to conclude
that such decreased recruitment is a consequence of, for example, subject inattentiveness
or task habituation. If however, a region of interest was shown to exhibit a higher
information load or entropy, suggesting an increase in computational efﬁciency, and/or a
higher covariance between the regional BOLD signal and stimulus presentation,
suggesting task attunement, a stronger argument could be made that network Optimization
had occurred in the region. In the present study, therefore, computational efﬁciency and
task attunement are quantitatively assessed through statistical analysis of the entropy
(voxel Shannon time series entropy, or voxel TSE, a novel application to FMRI) and
voxel/task cross-correlation (VTCC, or the BOLD signal response rate), respectively. The

particulars of both metrics are discussed in the Methodology chapter.

An analysis of task-related signal change and voxel TSE will also serve to support
or refute an inference that remnant voxels exhibiting a post-consolidation increase in
entropy are as metabolically active as (or more metabolically active than) signiﬁcantly
task-correlated voxels at pre-consolidation. This issue is of interest because, again, the
assessment of BOLD signal entropy is a novel application for an FMRI-based study of
human neurodynamics, and it is as yet unclear how faithfully BOLD signal entropy
corresponds to neuronal signal entropy; a statistical concurrence of task-correlated
increases in metabolic intensity and voxel TSE would suggest that remnant voxels both
carry more information and are more metabolically active after consolidation, thereby
supporting the assumption that the metrics discussed above are in actuality assessing the
degree of synaptic integration (and therefore the computational intensity) within the

region of interest.

34

CORTICAL REINSTATEMENT AND RETRIEVAL ORIENTATION EFFECTS IN VISUAL AND
MEDIAL-TEMPORAL AREAS

Cortical reinstatement and retrieval orientation effects should be evinced by a decrease in
cortical recruitment in long-term memory (LTM) and visual substrates, consequent to
Hebbian competition over the consolidation interval, as well as an increase in BOLD
response rates subsequent to the onset of neural activity, as correlates of enhanced
synaptic weights of neurons still participatory in the LTM network after consolidation
(Kida, Josselyn, de Ortiz, Kogan, Chevere, Masushige and Silva 2002; Bozon, Davis and
Laroche 2003; Bozon, Kelly, Josselyn, Silva, Davis and Laroche 2003). Furthermore,
LTM and visual circuit attunement (reinstatement) and retrieval orientation effects should
be at least correlated with behavioral performance improvement, consequent to the
efﬁcacy Of the post-consolidation network. However, one would not expect
consolidation to have the same effect magnitude on all retrieval-related networks:
differential retrieval efﬁciency concomitant to memory consolidation was expected to
manifest differently in attentional areas, which support the trace retrieval process, than in
medial temporal and visual networks, which actually store and retrieve previously-stored
memory traces.

Shifts in the orienting Of attention from local features to global features with
consolidation, corresponding to retrieval orientation effects, were predicted to be evinced
by bilateral activation in pre-consolidation subjects and right-hemisphere dominance
(fewer left—hemisphere and more right-hemisphere clusters: Brown and Kosslyn 1993;
Johannes, Wieringa et al. 1996; Yamaguchi, Yamagata et al. 2000), particularly in

extrastriate ventral. visual stream (EVVS) components in post-consolidation subjects

35

(H.3-4: Knierim and Van Essen 1992b; Knierim and van Essen 1992a; Polat and Norcia
1998; Haxby, Gobbini et al. 2001; Lerner, Hendler, Ben-Bashat, Hare] and Malach
2001). lntemalization, a consolidation-driven shift in attention from external lexical and
visual features to comparison of external features with stored representations of salient
stimulus features (Tulving 1983) and a possible consequence of cortical reinstatement,
was also predicted to be evinced by a decrease in activation volume as well as metabolic
demand in striate visual cortex (V1) and EVVS components with consolidation.

In contrast, “mnemonic” areas of the MTL are believed to function as a “linkage”
Site between traces stored in isocortex, as discussed in the Regions of Interest chapter (see
also Sakai and Miyashita 1991; Miyashita, Morita, Naya, Yoshida and Tomita 1998;
Egorov, Hamam et al. 2002; Egorov, Heinemann and Muller 2002; Frank and Brown
2003; Naya, Yoshida et al. 2003a; Naya, Yoshida et al. 2003b; Clavagnier, Falchier et al.
2004). Thus, recruitment of the ectorhinal, perirhinal and entorhinal cortices was
predicted to decrease while metabolic demand remained constant between pre- and post-
consolidation time points. Intracellular competition among trace neurons over the
consolidation interval should result in an amalgamation of, and diminution in, the number
of these trace neurons (Alvarez and Squire 1994; Chrobak and Buzsaki l998b; Chrobak
and Buzsaki I998a; lKida, Josselyn et al. 2002; Bozon, Davis et al. 2003; Bozon, Kelly et
al. 2003). This diminution should be reﬂected in a decrease in recruitment of
“mnemonic” medial temporal areas (ectorhinal, perirhinal, and perihippocampal
cortices). Furthermore, retrieval orientation effects were assumed to depend on both
visuospatial information (subserved by entorhinal cortex and anterior hippocampus) and

visual recognition (subserved by the majority of the MTL and posterior hippocampus).

36

However, it is worthy of note that, as discussed in the introductory chapters of this
dissertation, there exists as yet no clear consensus as to the computational operations of
the individual MTL areas; instead, most research has focused on a “visuospatial” versus
“mnemonic” functional categorization scheme (see, for example, Preston and Gabrieli
2002). Moreover, the simple DMTS task utilized in the present study was not designed

to disambiguate these visuospatial and mnemonic ﬁrnctions.

37

HYPOTHESES

In the present dissertation, mnemonic neurodynamics are investigated with respect to
differences in behavior, with respect to differential static, or modular, brain activity, and
with respect to differences in metabolic demands, between pre-consolidation and post-
consolidation subjects. In particular, these differences were evaluated in two phases. In
the ﬁrst phase, conventional FMRI statistics — percent signal change ﬁom baseline and
activation volume (task-correlated cortical recruitment) were applied in order to calculate
changes in regional recruitment and the correlation between these changes and
improvements in subject performance on a forced-choice delayed match-to-sample
variant (paired-associate) task. In the second phase, two novel statistical assessments, the
cross-correlation or covariance (VTCC) between stimulus presentation timing and
subsequent BOLD responses and the BOLD signal complexity (the Shannon entropy of a
voxel time series, or voxel TSE) — were employed to aid in the interpretation of the data

returned by the conventional metrics.

It should be noted that both behavioral and physiological results are interpreted in
the present study in the absence of direct (e.g., molecular or cellular) evidence that
memory consolidation occurred over the 7-day interval. In this context, a MATCH
condition accuracy signiﬁcantly greater than chance (i.e., a normalized accuracy Of 2
66%) was taken as evidence that consolidation had indeed occurred, Since (ostensibly)

neither successﬁil retrieval nor recognition effects can occur after intervals longer than 6

38

hours in the absence of stage IA protein synthesis (Abel and Lattal 2001.; Walker 2004;

Walker and Stickgold 2004; Wang, Hu et al. 2006).

The present study attempted to quantitatively describe differences in
neurophysiology between pre- and post- stage I consolidation subject groups and
correlate such differences, if present, with subject performance on the recall of previously
learned (correctly-matched) paired associates, in order to test the working hypothesis
that episodic memory consolidation is a process of optimization of the memory trace
network for recall. More precisely, it was hypothesized that, based on recent evidence of
consolidation-interval competition among neurons for inclusion in the ﬁnal memory trace
(Alvarez and Squire I994; Chrobak and Buzsaki 1998b; Chrobak and Buzsaki 1998a;
Kida, Josselyn et al. 2002; Bozon, Davis et al. 2003; Bozon, Kelly et al. 2003),
consolidation would effect decreased cortical recruitment in anatomical substrates of
long-term memory storage and recall, but that neurons surviving competition would
“compensate” computationally for their excluded counterparts. Thus, metabolic demands
were expected to remain constant in mnemonic substrates (1.6., the medial temporal lobe)
across pre- and post-consolidation subjects. Furthermore, consolidation was
hypothesized to result in “attunement” of retrieval networks to recognition of previously-
leamed stimuli, such that post-consolidation subjects would exhibit faster visual,
mnemonic, and motor attention (i.e., premotor cortex) BOLD signal response rates
relative to the presentation of previously-leamed associates, increased BOLD signal

entropy, and better overall subject memory performance.

39

Thus, consolidation-related optimization was hypothesized, ﬁrst, (H.1) to
correlate positively with recall performance (corresponding to plasticity changes among
memory traces), resulting in decreased subject response time and increased subject
accuracy in identifying previously learned associates. Second (H. 2), anatomical
substrates of pre- and post-consolidation memory storage and recall were hypothesized to
exhibit equivalent metabolic demands, evinced by the BOLD signal magnitude, with
decreased cortical recruitment, evinced by the spatial extent of task-correlated activation
(activation volume), relative to the presentation of previously learned associates, in
medial temporal (hippocampus, entorhinal, ectorhinal, and perirhinal) and attentional
(premotor) ROIS, corresponding to greater activity at the level of the individual neuron.
Extrastriate visual (VI-V3 and fusiform) ROIS, in contrast, were predicted to exhibit
decreased metabolic demand and decreased cortical recruitment, corresponding to
diminished activity at the level of the individual neuron as a consequence of retrieval
orientation effects (see below). Third (H. 3), all substrates, with the exception of primary
motor cortex, were expected to exhibit faster rates of change, evinced by faster voxel
BOLD signal rates, as a correlate of cortical reinstatement, as discussed previously.

And ﬁnally (H. 4), consolidation was hypothesized to drive changes in synaptic
integration, thereby effecting differential sensory processing (retrieval orientation) of the
learned associates, as evinced by right-hemisphere dominance in the ventral visual stream
(Robb and Rugg 2002).

These hypotheses are summarized in Table 3; related predictions are summarized

in Table 4.

40

 

Table 3 Summary of the hypotheses for the present study.

See Table 4:
H. 1 Consolidation correlates positively with recall performance for RT
previously learned associates due to plasticity changes among ACC

memory traces.

H. 2 Consolidation correlates negatively with task-oriented recruitment BOLD mag./
(i.e., activation volume) of recall substrates. However, post- Recruitment
consolidation trace neurons are individually more active than pre-
consolidation trace neurons, such that net metabolic demand
remains constant over the consolidation interval, except in visual
areas, which are less recruited overall at post-consolidation.

H. 3 Consolidation correlates positively with BOLD signal response VTCC
rates (cortical reinstatement) in mnemonic, visual, and motor
attention substrates.

H. 4 Consolidation correlates positively with metabolic demands in (N/A)
right-hemisphere components of the ventral visual stream,
including area V3 and the fusiform gyrus, corresponding to
differential retrieval orientation.

Table 4 Summary of the predictions for the present study with respect to subject
performance (reaction time and accuracy), as well as physiological descriptions of the
shape and temporal characteristics of the BOLD signal on the delayed match-to-sample
portions of the paired associates task. Both behavioral and physiological ﬁndings are
represented as being greater (+), less (—), or equal (no change, N/C) in post-consolidation
subjects relative to pre-consolidation subjects. Abbreviations: HC, hippocampus, ERC,
entorhinal cortex, EcRC, ectorhinal cortex, PRC, perirhinal cortex, V1, V2, V3, visual
areas 1-3, F USI, fusiform gyrus, M1, primary motor cortex, PMC, premotor cortex; RT,
subject reaction time, Acc, accuracy; BOLD mag, percent BOLD signal change from
baseline, recruitment, cortical recruitment (activation volume), BOLD rate, BOLD signal
response rate (evaluated by the voxel-task cross-correlation, VTCC). Hypothesis 4 (H4),
which predicts a predominance of right hemisphere activation in post-consolidation
subjects, is not included in the table.

RTIH ACC/H BOLD RECRUITMENT/H

 

1 1 M AG. [H2 2 BOLD RATE/H3
Eric ”"3 — +
EcR ”’0 " +
c N/C — +
PRC ”/0 _ *
V1 -- + — N/C +
V2 — — +
V3 — — +
FUSI _ .. +
M1 N/C N/C N/C
PMC N/C N/C +

 

41

All four of the above hypotheses (see Tables 3, 4) follow from the assumption that the
phenomenon of consolidation is, in essence, a process of Optimization of the mnemonic
net. However, it should be noted that activation volume, regional signal change from
baseline, voxel TSE, and VTCC will only be assessed and discussed for brain activity
that was signiﬁcantly correlated with the presentation of correctly-matched stimulus pairs
(matched associates) in order to minimize the effect of BOLD Signals corresponding to
stable task-related sensory and lexical networks (i.e., responses to control stimuli) and
task-related but generalized attentional, executive, and search-and-retrieval mechanisms
(i. e., responses to mismatched stimulus pairs). Thus, the ﬁndings discussed in the present
dissertation will be related exclusively to the recognition and processing of matched

associates and the related memory traces putatively stored in the subject's cortex.

42

METHODOLOGY

Stage I episodic memory consolidation was assessed using a three-phase training and
testing paradigm. In the training phase, subjects were given 20 pairs of abstract pictures
and animal words (tiger, bear, jellyﬁsh, etc.) to memorize. These 20 pairs were repeated
10 times in order to stabilize the pairings in memory; subjects were then immediately
tested on their retention using a computer-based testing script. During testing, subjects
were presented with 100 total stimulus pairs (33 correctly matched words and pictures, 33
incorrectly matched words and pictures, and 34 control stimuli: scrambled pictures and
backwards words) in a slow event-related paradigm. Only metrics concerned with
correctly-matched stimuli will be assessed. in this dissertation. Subjects were then
randomly assigned to one of two groups. Group A proceeded directly to the FMRI
magnet and underwent the same testing paradigm encountered previously; Group B
underwent the testing paradigm 7 days later. Thus, Group A represented a presumptive
pre-consolidation cohort, whereas Group B represented a presumptive post-consolidation

cohort (see Figure 4).

SUBJECT POOL AND PARTICIPANTS

Subjects for the present study were obtained from undergraduate and graduate students at
Michigan State University and from the surrounding community. 93 subjects (43 female;
mean age 19.45 yrs) were subjected to a computer-based (non—FMRI) version of the task,

and 13 subjects (6 female, age = 26.08 i- 10.57 yrs) comprised the ﬁnal testing (F MRI)

43

 

 

 

group. Of these, 6 (2 female) were assigned to the pro-consolidation group and 7 were
assigned to the post-consolidation group (4 female). lnforrned consents were obtained in
accordance with the Michigan State University Committee on Research Involving Human
Subjects (UCRIHS). The study was also approved by UCRIHS (IRB #02-634).
Demographically, the subject pool consisted of 1 Hispanic, 0 African-American, 1 Indo-
European, 1 Asian, and 10 white subjects. Participants denied claustrophobia and
signiﬁcant neurological history, and were screened for handedness using the Edinburgh
Handedness Inventory (Oldﬁeld 1971). Only right-handed subjects were asked to

participate in the study.

PARADIGM
Stimulus presentation scripts were designed, prepared, and tested using E-Studio software

(Psychology Software Tools, Pittsburgh, Pennsylvania). Subjects were given a computer-

 

 

30 min.
(50% of
4 subjects)

L
‘ r

 

7 days
(50% of
subjects)
A k
‘

      

bee
jellyﬁsh hsifyllej hsifyllei

 

 

 

 

 

 

 

 

 

 

Figure 3 The training/testing paradigm for the present study. Stimuli consisted of 20 non-
rehearsable, non-generalizable abstract pictures paired with animal nouns. Word Stimuli
were presented either audibly (computer-based training/testing) or visually (FMRI
training/testing). Following a training session, during which the correct pairings of stimuli
were learned, subjects were tested on their recall either immediately (pre—consolidation
time point; 50% of subjects) or 7 days after training (post-consolidation time point; 50%
of subjects). For the purposes of testing, subjects were presented with 33 correctly-
matched, 33 incorrectly-matched, and 34 control (scrambled) stimulus pairs, and subjects
were required to identify each type correctly by button press.

44

based training session, during which they learned 20 pairs of presumably non-
rehearsable, non-generalizable abstract pictures and simultaneously-presented animal
nouns, such as rabbit, monkey, koala, jellyﬁsh, and parrot. The 20 pairs were each
repeated ten times to ensure thorough learning, stipulatively deﬁned as a 66% or better
accuracy on the MATCH condition (DMTS) task. Following the training session, 50% of
the participants were randomly selected to return after a 30-minute interval for the testing
session; the other 50% were asked to return in seven days for testing (see Figure 3). In
either case, subjects were instructed not to think about the task, nor attempt to remember

or rehearse the presented stimuli.

The paradigm was tested with 93 subjects (43 female) using a computer-based
version of the task. For these pilot subjects, the “animal words” were spoken by the
computer simultaneously with the abstract picture presentation; thus, this version of the
task was “dual modality.” However, technical difﬁculties in incorporating the audio
system inside the MRI magnet necessitated the substitution of lexical stimuli for the
auditory components for training as well as testing sessions for MRI subjects, such that
the MRI version of the task was constituted wholly of visual stimuli (see Figure 4).
Stimuli did not differ in any other respects. Differences in task performance between pilot
and MR1 groups were evaluated and are discussed in the “Changes in BOLD Physiology
with Consolidation” section in the Memory Consolidation and Retrieval Performance

chapter.

45

FMRI Paradigm. With the exception of stimulus modality, training and testing

were consistent across computer-based and FMRI groups. During the testing phase,

(A)

    

l
jellyﬁsh

(B) -lH—l—I—l——I——H———H—

II I II I llllll
MM
ISATC” II I II I lllll

CONTROL +I—HHH1—H—H—l—H—l—Hl-

Figure 4 (A) Example Of a paired associate as presented for the FMRI version of the
training/testing paradigm: note that the word component of FMRI training and testing
associates was presented below the abstract visual component. The centroid of the
presented paired associate was positioned in the center of the subjects' ﬁeld of view. (B)
A schematic of a single run ﬁ'om the FMRI testing paradigm: match, mismatch, and
control-condition stimuli were presented with a constant interstimulus interval of 12.5
seconds. Presentation order was randomized across the three types of stimuli; x-axis
represents TR (1 TR = 2.5 seconds).

 

subjects were asked to determine, based on their training session, whether presented
stimulus pairs were correctly or incorrectly matched, and to respond accordingly by
button press (thumb for match, index for mismatch). Additionally, control stimuli,
consisting of heavily pixelated versions of the testing stimuli and either backwards audio

(CBT subjects) or scrambled letters (for FMRI subjects), were presented at random

46

intervals. Subjects were asked to respond to these control stimuli by pressing both

response buttons simultaneously.

MATCH, MISMATCH, and CONTROL condition stimuli were presented in random
order3 with a constant interstimulus interval (181) of 12.5 seconds (Figure 4B). Subjects
were given four runs of 25 stimuli and each stimulus was presented for one second, for a
total run time of seven minutes in addition to 8 seconds of longitudinal magnetization

equilibration time (total, 7 minutes 8 seconds).

DATA ACQUISITION

Behavioral measures acquisition. Correctly matched and mismatched paired associates
and control stimulus pairs were presented to subjects within the MRI magnet, as
described above, on an LCD display. Stimulus presentation was controlled, with
millisecond accuracy, through an implementation of E-Prime and Integrated Functional
Imaging Systems (IFIS) software (Psychology Software Tools, Pittsburgh, Pennsylvania).
Subject response time and accuracy data were acquired through an electronic button

response system tethered to the subject's right hand.

MR acquisition. Anatomical and functional data acquisition was carried out on a
GE 3 Tesla MRI magnet (General Electric Medical Systems, Wisconsin, United States)
with an 8-channel head coil. Passive control of sagittal and lateral subject head motion

was established by using foam supports and medical tape secured to the subject's

 

3 Random ordering of stimuli was performed by assigned a given stimulus an integer between 1 and 100
using a random seed generation script in the Linux Boume Again shell.

47

forehead, respectively. Single-shot gradient echo-planar image (EPI) functional slices
(27.7 msec TE, 22 cm FOV, 3 mm slice thickness x 34 inferior-to-superior slices, 80° ﬂip
angle) were co-registered to high-resolution Tl-weighted sagittal anatomical volumes (5
msec TE, 24 cm FOV, 1.5 mm Slice thickness x 120 slices, £15.63 kHz bandwidth, 8°
ﬂip angle) for each subject. Anatomical and functional data was subjected to a 6-
parameter (roll, pitch, yaw, and axial, sagittal, and coronal shift) iterated linear least
squares motion correction algorithm using the programs ZdImReg and 3dvol reg, part
of the AFNI FMRI analysis suite (NIMH), and also temporally realigned to correct for
slice acquisition offset. A low-frequency bandstop was applied over the (-°°, 0.020] H2
interval to remove low-frequency artifacts in the data. A high-frequency bandstop was
also applied on the [0.080,+°°) Hz interval in order to remove pulsatile cardiac (40-60
bpm) and respiratory-related (12-16 bpm) motion components. Although stimulus
presentation order was randomized, stimuli of the same condition were presented at an
average of 0.057 H2, well within the data pass range of (0.020, 0.080) Hz. The bandstop
ﬁlters were not expected to affect the entropy calculation (described below) since entropy
is primarily inﬂuenced by moderate-probability events in a signal (e.g.random—walk-type
phenomena), rather than pulsatile components, which are higher-probability (more

predictable) Signal components.

BEHAVIORAL ANALYSIS

Reaction time and behavioral data were analyzed with Minitab 14.0; performance
statistics-related ﬁgures presented in the Results chapters were generated with SYSTAT

10.2 using the statistical results from Minitab. Inclusion criteria were established as

48

control stimulus response accuracy of at least 75%, in order to differentiate between high-
and low-compliance subjects, and a paired-associates response accuracy (i.e., the mean of
matched- and mismatched-associates accuracies) of at least 66%, in order to differentiate
between relatively low-performers and moderately-high to high-performers. Low-
performer data was not incorporated into the dataset for the present study; MRI data
would also have been excluded for low-perforrners, however, none of the 13 MRI

subjects exhibited data outside of the inclusion range.

For the purposes of analysis, both behavioral and MRI data were categorized as
responses to correctly-matched stimuli (MATCH condition), incorrectly-matched stimuli
(MISMATCH condition), or control stimuli (CONTROL condition). Reaction times and
accuracies were then calculated for each condition and are reported for each condition
separately, but only MATCH condition performance will be explored further in the present

study.

MODULAR ANALYSIS
Deconvolution
In order to differentiate between BOLD/neural responses to MATCH, MISMATCH, and
CONTROL stimuli, as well as responses to ﬁxation crosses embedded in the paradigm, a
deconvolution analysis was performed on the data using the AFNI program
3dDeconvolve. Deconvolution analysis de-overlaps contributors to the raw MR signal
such that subject motion, arteriovenous signal, baseline trends, and task-related Signal

components can be isolated and analysed as statistical parametric maps (SPMS) (see

Ward 1998).

49

Mathematically, given a raw voxel BOLD Signal vector y(t), the “true
haemodynamic response vector x(t), an Dirac delta function denoted 5(t), an a priori
polynomial vector y, a Gaussian noise component 8, and k so-called orthogonal functions
I//(t) , representing subject motion and signal drift components (the error term for the
linear regression of signal components) respectively, it is assumed that the raw voxel
BOLD signal is a linear composition, or convolution, of the Dirac delta ﬁinction,

polynomial, and orthogonal vectors:

y(t) =x<t>® 7(t)®I//(t)+€(t)-
The objective of deconvolution analysis, then, is to remove the motion, noise, and drift

components by linear least-squares estimation of x(t). Thus, given
Z(t)=Zy(t)5(t-r)=x(t)[1|r|5|W1+8(t),
where 8 represents the Dirac delta function, the sum-square error, SSE, can be

minimized,

 

 

2 r T T
min SSE = Eli—(Z - x”, 2)] [Z - ﬁx;— 2)].
dt x x

x x
Using 3dDeconvolve, voxels that activated in response to matched associates, but not
mismatched associates, control stimuli, or ﬁxation crosses, were isolated by a subtraction
method to create a MATCH condition SPM. Motion components were also deconvolved
against the raw signal, and correlation coefﬁcients between the voxel response vectors
and an ideal hemodynamic response waveform (composed by convolving the binary

stimulus onset vector with an empirically-detennined hemodynamic response model

waveform r(t): R.W. Cox, waver FA Q, 2005):

50

0for all {x, |—oo<x,. <0};
r(t) = 0.502 tanh[tan(0.5007r - 1.600x — 0.800)] + 0.996 for all {xi [0 S xi <1};
lforall {xi |l<x,. <oo}

where x, represents the voxel response vector.

In order to control for type II errors, bias alpha correction, and prepare individual
data for registration to the common Talairach-Toumeaux grid (see, for example,
Salmond, Ashburner, Vargha-Khadem, Connelly, Gadian and Friston 2002; Breakspear,
Brammer, Bullmore, Das and Williams 2004; Geissler, Lanzenberger, Barth, Tahamtan,
Milakara, Gartus and Beisteiner 2005; Kriegeskorte, Goebel and Bandettini 2006;
Scouten, Papademetris and Constable 2006), cluster analysis was performed on the
deconvolved functional data using minimal-volume thresholds computed with the AFNI
program AlphaSim. AlphaSim estimates the or level (and therefore the p-value)
signiﬁcance of random noise via Monte Carlo simulations within a virtual volume with
the same dimensions as the dataset voxels, using the empirical ﬁlter width at half
maximum (FWHM) of the data. Using this technique, it was found that a cluster size of
646 voxels (6379 LIL), corresponding to or z 0.955, would yield a corrected p-value of

0.045 for the dataset (minimal connection radius = 5.7 mm).

Percent BOLD signal change from baseline, activation voume, voxel-task cross-
correlation (VTCC) and entropy (TSE) statistics were obtained per region of interest
(ROI) ﬁ‘om these SPMS. ROIS (with the exception of the hippocampus) were deﬁned by

Brodmann area (Brodmann area assignments are described in the Regions of Interest

51

chapter) using the “Talairach” daemon in the AFNI analysis suite (see also Talairach and

Tournoux 1988).

Percent BOLD signal change from baseline (see also Ward 1998)

Brain activation is often characterized in FMRI experiments by calculating the statistical
signiﬁcance of a regional BOLD signal, the percent BOLD signal change from baseline.
In the present study, BOLD signal change was quantiﬁed for comparison with
neurodynamic metrics, such as voxel entropy and voxel-task cross-correlation, and in the
interest of assessing these latter statistics in the context of voxel recruitment (activation
volume) and glycolytic intensity, from which the BOLD Signal is largely derived
(Raichle 1998; Logothetis, Pauls et al. 2001). Mathematically, given a reference vector
r(t), a voxel time series vector x(t) having it data points x,, an a priori polynomial vector
y,- and k so-called orthogonal ﬁinctions 1m, representing subject motion and signal drift
components (the error term for the linear regression of signal components) respectively,

the percent BOLD signal change ﬁom baseline is calculated

AP : IOOX “(rmax —rmin)
B(xi)

 

9

where it represents the voxel time series-to-reference waveform ﬁt coefﬁcient, rmax
and rm,“ represent the maximum and minimum values of the reference vector deﬁned

above, and B(x,) represents the signal baseline, deﬁned in turn as

52

n A n k n n
B(x,.) = 23.17-] 2x. + 2 Z fljn‘] (pj + otrmin
i=1 i=1 j=l

j =1
meaning that AP is simply the echo-planar image intensity, weighted by the peak-to-peak
variation of the reference time series, and divided by the linear combination of the “true"

(neural activity-derived) voxel response and all “noise” contributors to the signal (B.D.

Ward, 3df im+ Reference Manual, 2000).

Percent signal change was computed for individual subjects' MATCH condition
SPMS using the AFNI program 3dfim+. For the purposes of establishing a baseline,
signal drift (7) was assumed to be a ﬁrst-order polynomial (i. e., a linear trend) and motion
components (III) were assumed to be negligible, since these components were removed
from the raw EPI signal waveform during deconvolution. Only percent signal changes
with an internal AFNI threshold of R _>_ 0.23, corresponding to p _<_ 0.01 (corrected), were

considered to be statistically signiﬁcant.
Activation volume (see also Ward 1998)

Activation volume (AV) is deﬁned as the number of voxels signiﬁcantly correlated with
a reference waveform representing a task of interest. Speciﬁcally, given a constant
rectangular prism voxel volume L WH over N voxels in a deconvolved subject dataset that
exhibits an F -statistic (following the variable naming conventions in the “Percent Signal

Change” section):

53

= de+x[SSE(B) —SSE(B +x)]
(de —de+x)[SSE(B + x)]

 

F

where SSE and df denote the residual sum of squares and degrees of freedom used in the

deconvolution analysis, respectively, then
A V a N(L . W . H)

Activation volume has been utilized in FMRI studies of auditory and motor cortex
recruitment, as well as comparisons of neurological patients and control subjects
following cortical recovery from stroke (Cramer, Weisskoff, Schaechter, Nelles, Foley,
Finklestein and Rosen 2002; Lasota, Ulrner, Firszt, Biswal, Daniels and Prost 2003);
however, it has been cautioned that activation volume should not be interpreted in the
absence of other measures of brain function, since it is susceptible to intersubject
variability (Saad, Ropella et al. 2003). Therefore, AV was utilized in the present study
only in conjunction with voxel—task cross-correlation, percent signal change, and voxel

entropy (see below).
Voxel-task cross-correlation (VTCC: see also Ward 1998)

The voxel-to-task cross-correlation (VTCC) is deﬁned as the Pearson product-moment
correlation coefﬁcient between a voxel BOLD signal time series and a reference
waveform (deﬁned above). Following the variable naming conventions ﬁ'om the “Percent

Signal Change” section,

54

[X(t) — ”x(t) :IZ [r(t) _ﬂr(t)]
JZ[X(t)-ﬂ,(,)]22[r(t)-l1..(,)]2

 

p

such that, when Signal drift (7) is accounted for,

 

 

_x(t) TTr r _r(t) TTr r

p: 3&0) 7 y 22;‘(t) 7T 7
.110) TTr 7 J0) TTr r

x(t) 7T 7 Z r(t) 7T 7

 

 

 

Note that motion components are not considered in the above equation since they were

removed from the raw signal during deconvolution (B.D. Ward, 3dfim+ Reference

Manual, 2000).

VTCC was computed for individual subjects' MATCH condition SPMS using the
AFNI program 3df im+. For the purposes of establishing a baseline, signal drift (7) was

assumed to be a ﬁrst-order polynomial (i. e., a linear trend). Only VTCCS with an internal
AFNI threshold of R _>_ 0.23, corresponding to p 5 0.01 (corrected), were considered to be

statistically signiﬁcant.

Shannon entropy of the voxel time series (T SE)

In contrast to metrics based on BOLD signal intensity, many estimates of the amount of

information in a time series based on information theory (also known as

55

“communications theory”) make no assumptions about the shape, origin, or stationarity
of the Signal under inspection, meaning that the conclusions based on such metrics are
not biased due to underlying models of neurovascular coupling (Gonzalez Andino, Grave
de Peralta Menendez, Thut, Spinelli, Blanke, Michel, Seeck and Landis 2000; de Araujo,
Tedeschi, Santos, Elias, Neves and Baffa 2003). Of interest in the present study is an
information theory-based metric known as the Shannon entropy of a time series (TSE),
which was employed in order to quantify the “chaoticity” — that is, the disorder, or,
more speciﬁcally, unstable aperiodicity (Shannon 1948) — of the BOLD response in
discrete brain regions. This quantiﬁcation of time series entropy is of interest in the
present study because, by deﬁnition, the disorder of a time series is directly and
positively related to the amount of information it carries, much as the instability in an
audio signal is related to the complexity of the sound it represents (Shannon, 1948).
Moreover, it has been suggested by electrophysiological studies that entropy is a direct
aid to synchrony among cell populations since signal instability facilitates changes in

neural ﬁring rate oscillation (Fell, Fernandez and Elger 2003).

In the case of a voxel or group of voxels, such as F MRI regions of interest (ROIS),
the Shannon entropy may be related to “chaotic” coupling between neural activity and the
BOLD Signal, or to chaos in the neural responses themselves, due to a surplus of neural
connections prior to, for example, synaptic pruning. With respect to the former case,
several instances exist in which neural activity and the BOLD signal become decoupled;
for example, in patients with a history of stroke or vascular disorders, or subjects who
have had caffeine or nicotine prior to their FMRI scan, or at very high and very low

levels of neural activity (Logothetis, Pauls et al. 2001; Jacobsen, Gore, Skudlarski,

56

Lacadie, Jatlow and Krystal 2002; Laurienti, Field, Burdette, Maldjian, Yen and Moody
2002; Hamzei, Knab, Weiller and Rother 2003; Liu, Behzadi, Restom, Uludag, Lu,
Buracas, Dubowitz and Buxton 2004; Krainik, Hund-Georgiadis, Zysset and von Cramon
2005). There is also some evidence that infants and toddlers have immature
neurovascular coupling that may complicate interpretation of fMRI results as compared
to adult fMRI data (Martin, Joeri, Loenneker, Ekatodrarrris, Vitacco, Hennig and Marcar
1999; Poldrack 2000). By age 8, however, the hemodynamic responses Of children and
adults are comparable (Kang, Burgund, Lugar, Petersen and Schlaggar 2003). In healthy,
compliant adults and in children over three years of age, therefore, neural activity and the
BOLD signal are believed to be tightly coupled (Martin, Joeri et al. 1999; Logothetis
2002). For this reason, it was inferred that, in the majority of subjects, voxel Shannon
time series entropy (voxel TSE) arises from disorder in the underlying neural activity,
and, given that the BOLD signal arises primarily from metabolic events at the synapse
(Logothetis, 2002), this disorder likely occurs at synaptic connections. The precise
relationship between BOLD signal chaos and neural activity, however, is as yet
unknown. It was surrnized that voxel TSE is directly correlated with the stability of
synaptic integration, which is operationalized here as the ease with which multiple inputs
undergo spatial or temporal summation at the synapse to create “coherent” output.
Intuitively, spatial and temporal summation could be complicated by (a) an
overabundance of inputs into a single output (morphological complexity), or (b)
inefﬁciency of the connection at the synaptic junction (connectivistic complexity — due

to, for example, a lack of Hebbian entrainment). This metric is therefore being applied in

57

the present study in order to determine whether TSE differs Signiﬁcantly in particular

ROIS between pre- and post-consolidation groups.

The Shannon time series entropy (TSE) was calculated using the wentropy
function in the MATLAB Wavelet Toolbox (The MathWorks, Inc., Natick,
Massachusetts). AS implemented in the went ropy program, the Shannon entropy H(t)ROI,
in Shannon bits (Sh), of the average time series across all voxels n meeting the threshold

criterion (see above) in a region of interest was calculated as

l n
H(t)ROI E -;Zin210g(xi2),
x=l i

where xi represents a data point in the time series Of a voxel x (x 6 R01). The Shannon
entropy of the voxel time series is calculated directly from the bandpass-ﬁltered voxel
response vector as described in de Araujo et al. (de Araujo, Tedeschi et al. 2003). The
contribution of response rates to the entropy therefore has a Poisson distribution, such
that very low- and high-probability responses contribute little to the overall TSE, and the
entropy calculation is biased toward x, values that appear with moderate frequency. What
this means conceptually, then, is that the voxel TSE is a quantiﬁcation of how disordered
a voxel response is, and consequently, it is also a measure of the amount of time- or
frequency-encoded information that is potentially contained within the time series. With
regard to the hypotheses outlined in the previous chapter, this means that the amount of
disorder in voxel response can be used as an indirect measure of the amount of

information carried in the BOLD response.

58

MEMORY CONSOLIDATION AND RETRIEVAL PERFORMANCE: EFFECTS
OF STIMULUS MODALITY AND PAIR CONCORDANCE

Experience and intuition suggest that episodic memory recall declines with the passage of
time. Yet episodic memory networks undergo a period of consolidation, during which
neuronal connections are assumed to be optimized - meaning that the mnemonic circuits
involved in the memory are indexed, transferred to isocortex, and made permanent. In
vivo as well as in vitro studies in humans and higher primates have conﬁrmed that
decreased memory and memory trace lability, increased intemeuronal ﬁring coherence,
decreased recruitment of memory-irrelevant (peripheral) brain areas, and heightened
long-term depression (LTD) of ﬁring coherence between these peripheral areas and
mnemonic networks are hallmarks of episodic consolidation (Alvarez and Squire 1994;
Eichenbaum I996; Eichenbaurrr, Dusek et al. 1996; Eichenbaum, Schoenbaum et al.
1996; Chrobak and Buzsaki l998b; Buchel, Coull et al. 1999; Abel and Lattal 2001;
Eichenbaum 2001; Haist, Bowden Gore et al. 2001; Bodizs, Bekesy et al. 2002; Cantero,
Atienza and Salas 20023; Cantero, Atienza et al. 2002b). The “consolidation as
optimization” hypothesis is further supported by the striking changes in proximal and
distal dendritic morphology that have been observed in a consolidation-concomitant long-
term potention (LTP) context (Bailey, Chen et al. 1992; Bailey, Montarolo et al. 1992;
Zhang, Endo et al. 1997; Verzi 2004; Verzi, Rheuben and Baer 2005). Such changes in
morphology have been shown to modulate synaptic integration, thereby modifying the
computational functions of the neuron (Shepherd and Brayton 1987; Holmes and Levy

1990; Zador, Koch et al. 1990; Yuste and Denk 1995; Shepherd 1996; Hausser, Spruston

59

et al. 2000; London and Hausser 2005; Segev 2006). It would therefore be reasonable to
expect that, as a consequence of this modiﬁed neuronal computation, post—consolidation
networks would be comprised of fewer neurons that nevertheless individually exhibit
more complex ﬁring rate patterns. This expectation could be substantiated or falsiﬁed by
evaluating pre- and post-consolidation glycolytic magnitude, ﬁring rate entropy and
magnitude, and the number of neural populations recruited for recall.

Conventional wisdom holds that, in Spite of this putative consolidation-related
mnemonic network optimization, recall performance declines with time, except when the
items to be encoded are repeatedly presented; yet research into the neurodynamic
consequences of the consolidation phenomenon at the whole—brain level have only
recently been undertaken, and many of the published studies on the subject have focused

on the hippocampal, rather than cortical and medial-temporal, contributions to recall.

The primary aim of the present FMRI study was to test the hypothesis that
Optimization does indeed enhance retrieval efﬁciency, where an “enhanced retrieval
efficiency” is deﬁned, for the purposes of this study, as (a) improvements in behavioral
performance — subject response time and accuracy — and (b) decreased overall
metabolic demand. The working hypothesis was tested in 93 subjects on a computer-
based bimodal paired associates (delayed match-to-sample or DMTS) task, and in 13

other subjects on an FMRI-based unimodal version of the same task.

It Should be noted that both behavioral and physiological results are interpreted in
the present study in the absence of direct (e. g, molecular or cellular) evidence that

memory consolidation occurred over the 7-day interval. In this context, a MATCH

60

condition accuracy signiﬁcantly greater than chance (i.e., a normalized accuracy Of 2
66%) was taken as evidence that consolidation had indeed occurred, Since (ostensibly)
neither successful retrieval nor recognition effects can occur after intervals longer than 6
hours in the absence of stage IA protein synthesis (Abel and Lattal 2001; Walker 2004;

Walker and Stickgold 2004; Wang, Hu et al. 2006)

Please refer to the Methodology chapter for details of the data acquisition and

analysis procedures.

BEHAVIORAL RESULTS

Subject performance, operationalized as reaction time and accuracy, were assessed with
two ANOVAs. Whereas FMRI and CBT training sessions differed only in the modality
of stimulus presentation, testing sessions differed with respect to both environment and
presentation modality. Therefore, the ﬁrst MANOVA tested the effect of paradigm type
(FMRI versus computer-based testing, CBT) and time point (pre- versus post-
consolidation) on reaction time and accuracy for previously learned (MATCH or delayed
match to sample, DMTS) stimuli, mismatched stimuli, and control stimuli. Whereas
MATCH and MISMATCH reaction times and accuracy were expected to differ between the
two paradigm types due to attentional demands, CONTROL performance was predicted to
remain constant across paradigm types and time points, and serve instead as a measure of
subject compliance. Furthermore, it was predicted that accuracy would increase whereas
reaction time would decrease as a function of time point but not paradigm type,
consistent with the stated hypothesis that task performance would improve with memory

consolidation.

6]

 

 

1500‘

Iii
fr

\

§
‘R\\\\\\\\\\\\\\

1000‘

RESPONSE TIME (ms)

0..

9..
Q
Q
0.
9.
9.
9.
9.

e
To.

9’.
O...

 

 

 

 

 

 

 

 

 

o l l . .
Time point Pre Post Pre Post Pre Post Pre Post Pre Post

Pre POst
Testing grp. CBT FMRI CBT FMRI CBT FMRI
Match Mismatch Control

 

 

 

Figure 5 Mean reaction times for the paired-associate task by testing group (computer-
based, CBT, versus FMRI subjects), pair concordance (matched, mismatched, or control),
and time point (pre- versus post-consolidation). FMRI subjects, who were given
unimodal (visual) training and testing stimulus pairs, were slower at identifying both
correctly-matched and mismatched stimulus pairs than CBT subjects, who were given
dual—modal (visual and auditory) training and testing stimulus pairs. This effect was seen
at both pre- and post-consolidation time points. However, neither FMRI nor CBT
subjects were signiﬁcantly faster at identifying correctly—matched stimulus pairs after one
week. Neither group improved with respect to reaction time with consolidation for
control stimuli, nor did CBT and FMRI subjects differ signiﬁcantly with respect to
reaction time for control stimuli at either time point. Error bars represent standard errors
of the means. *p < .05.

Reaction time (H. 1)

Reaction time was deﬁned as the response time, in msec, between the presentation of
stimulus pairs and button press response to matched (MATCH reaction time), mismatched
(MISMATCH reaction time), and control (CONTROL reaction time) stimulus presentations

separately. A two-way (2x2) ANOVA using paradigm type (FMRI versus computer-

62

based training/testing, CBT) and time point (pre- versus post-consolidation) as
independent factors indicated that neither F MRI nor CBT subjects were faster at
identifying previously-learned (correctly-matched) stimulus pairs at post-consolidation
than at pre-consolidation (F MRI pre-consolidation matched associate reaction time p. i 0'
= 1701.8 i 120.4 msec, post-consolidation reaction time p. i O' = 1615.7 3: 207.3 msec;
CBT pre-consolidation matched-associate reaction time u i 0' = 1288.9 :1: 174.8 msec,
post-consolidation reaction time u i O' = 1201.4 i 112.6 msec; F(1, 54) = 2.22, MSE =
23399, p = 0.119), although pre- and post-consolidation FMRI subjects were slower at
identifying correctly-matched associates than their CBT counterparts (F(1, 54) = 67.07,
MSE = 23399, p < .001). Moreover, FMRI subjects were slower at identifying
mismatched stimuli than were CBT subjects at both pre- and post-consolidation time
points (FMRI pre-consolidation mismatched associate reaction time u i O = 1376.0 i
739.0 msec, post-consolidation reaction time u i 0' = 1599.6 i 253.2 msec; CBT pre-
consolidation mismatched-associate reaction time p. i” 0' = 1315.6 i 182.3 msec, post-
consolidation reaction time p. i 0' = 1262.3 i 164.2 msec; F(1, 54) = 4.87, MSE = 74470,
p = 0.032), although there was no signiﬁcant difference in reaction time for mismatched
stimuli as a function of time (consolidation: F(1, 54) = 1.18, MSE = 74470, p = 0.314).
The reaction time ﬁndings thus contradicted the predictions following from H. I. (See

Figure 5 and Appendix A.)

Accuracy (H. 1)

Accuracy was deﬁned as the number of correct identiﬁcations of matched (MATCH

accuracy), mismatched (MISMATCH accuracy), and control (CONTROL accuracy) stimulus

63

presentations separately. Using a two—way 2x2 ANOVA, FMRI and computer-based
training/testing (CBT) subjects were found to exhibit comparable accuracy on correctly-
matched stimulus pairs (FMRI pre-consolidation accuracy 11 i o = 76.79 i- 9.13%, post-
consolidation accuracy It i O = 71.29 i 15.35%; CBT pre-consolidation accuracy [.1 i O' =
73.74 i 11.79%, post-consolidation accuracy p. i O' = 79.06 1- 11.56%), indicating that
neither FMRI nor CBT subjects differed with respect to the number of false negatives
(F(1, 54) = 0.350, MSE = 143.80, p = 0.554), nor did the number of false negatives
increase or decrease in either group as a function of time (F(2, 54) = 1.370, MSE = 143.8,
p = 0.264). FMRI subjects were found to exhibit lower accuracies, however, for
mismatched stimulus pairs (pre-consolidation accuracy 11 i 0' = 72.39 i 10.74%, post-
consolidation accuracy u i 0' = 66.54 i 15.06%) than their CBT counterparts (pre-
consolidation accuracy u i O' = 79.22 i 13.60%, post-consolidation accuracy It i o =
82.78 3: 13.30%), indicating that the number of false positives were higher in FMRI
subjects than in CBT subjects (F(1, 54) = 6.74, MSE = 181.1, p = 0.012), although neither
group exhibited a signiﬁcant change in accuracy for mismatched stimulus pairs as a
function of time (F(2, 54) = 0.65, MSE = 181.1, p = 0.526). These ﬁndings suggest that
both F MRI and CBT subjects identiﬁed the correctly matched associates learned during
training with statistically equal accuracy, but FMRI subjects were more likely to

incorrectly identify mismatched stimuli as a previously learned paired associate (See

Figure 6).

Moreover, the putatively post-consolidation subjects in both CBT and FMRI
groups identiﬁed the previously-learned paired associates as accurately as pre-

consolidation subjects, but, conversely, subjects neither improved at correct rejections of

64

 

100 <

%%

'V
1;:

.0. .

904

80 4 r11

70-

O

o
e
To.

v
0

3i?

.0

O

 

v
0
.0

Q
.0...

.'.' v
9:"?

0J5
90¢
90%? ,

‘ 66%

o
o
.0?

v v.3 v
. 6.6

%

,...V
O 9
92¢
.'
c
’0;

60-

0

.y.
0
A
'

3
.9.

ACCURACY (% Correct)
EiEiEiE:
2:535 35

50.

O
O

vrvvv
.000 0
92020202320
RV???
rooooo
ﬂﬁﬁ£§b

 

 

  

 

 

 

 

 

 

Pre Post Pre Post
CBT FMRI CBT FMRI
Match Mismatch Control

40 I I
Time point Pre Post Pre Post
Testing grp. FMRI

Figure 6 Mean subject accuracy for the paired-associate task by testing group (computer-
based, CBT, versus FMRI subjects), pair concordance (matched, mismatched, or control),
and time point (pre- versus post—consolidation). FMRI subjects, who were given
unimodal (visual) training and testing stimulus pairs, were more likely to identify
mismatched associates as matched associates (exhibited more false positives) than CBT
subjects, who were given dual-modal (visual and auditory) training and testing stimulus
pairs. However, neither FMRI nor CBT subjects improved with respect to the number of
false positives over the putative memory consolidation period, nor did either group
improve with respect to accuracy with consolidation for matched associates (false
negatives) or control stimuli. CBT and FMRI subjects did not differ signiﬁcantly with
respect to accuracy for correctly—matched associates or control stimuli at either time
point. Symbols (light grey) indicate individual data points; error bars represent standard
errors ofthe means. *p < .05; **p < 0.01.

mismatched stimuli nor improved at accurately identifying learned associates (correctly-
matched stimulus pairs) as a function of memory consolidation. An absence of
statistically signiﬁcant effects of time (consolidation: F(1, 54) = 1.43, MSE = 277.6, p =

0.238) or testing group (CBT versus FMRI: F(1, 54) = 1.03, MSE = 277.6, p = 0.316) on

accuracy for control stimuli indicated equivalent subject task compliance in all four

65

conditions (FMRI pre-consolidation accuracy 11 i 0' = 97.50 i 4.07%, post-consolidation
accuracy u i O’ = 94.79 i 11.13%; CBT pre-consolidation accuracy u i 0' = 86.13 i
24.01%, post-consolidation accuracy p. i O' = 93.72 i 9.40%). Thus, accuracy did not

appear to reﬂect memory consolidation effects, contradicting H. 1 (see Figure 6).

Thus, the multivariate ANOVA performed on CBT and F MRI subject response
time and accuracy failed to conﬁrm the stated hypothesis (H. 1) that the putative period
of memory consolidation was positively correlated with subject recall performance,
although the observed differences in recall performance between the CBT (bimodal) and
FMRI (unimodal) paradigms may also indicate that such changes are sensitive to the
number and type of modalities used to present the stimulus pairs to be learned.
Interestingly, however, the conservation of subject response time and accuracy across
pre- and post-consolidation subject groups suggests the possibility that access to memory
traces was not reduced over the consolidation period, over and above a reduction in trace
neurons as a consequence of the intraneuronal Hebbian competition that is believed to
occur within hours of episodic memory encoding (Kida, Josselyn et al. 2002; Bozon,

Davis et al. 2003; Bozon, Kelly et al. 2003).

66

Figure 7 Group statistical parametric maps (SPMS) of voxel BOLD signal correlation
with presentation Of match, mismatch, and control-condition stimuli by region of interest.
Voxel BOLD signal changes shown are signiﬁcant at a ﬁIll-model F = 3.786 (p = 0.001,
corrected); colors represent whether changes are positive (yellow) or negative (blue)
relative to the baseline. Group SPMS were obtained by studentizing individual BOLD
datasets and registering them to a common Talairach-Tourneaux coordinate system
(Talairach and Toumeaux, 1988). Datasets were then subjected to cluster analysis using
a minimal 5mm, 10p] Gaussian kernel (to correct for type 11 error: see Salmond,
Ashburner, et a1. 2002; Geissler, Lanzenberger, et al. 2005) and averaged by condition.
The location of the calcarine ﬁssure (visual area V1) is shown on an axial slice in each
case for reference (yellow boxes). Hemodynamic response functions for representative
subjects were also obtained for ectorhinal area 36 (red box) and visual area V3/19 (green
box) and are shown in Figure 9. Abbreviations: Fus, ﬁisiform gyrus; Ins, AIC, anterior
insular complex; ST G, superior temporal gyrus; IPL, inferior parietal lobule; SPL,
superior parietal lobule; DG, dentate gyrus; pul, pulvinar nucleus of the thalamus, Ver,
cerebellar vermis, Cal, calcarine ﬁssure. Numbers represent Brodmann areas: 3a, digit
somatosensory area; 4, digit motor area; 31, posterior cingulate cortex.

67

3 786. p = 0.001

t-statistic at F

 

3.786, p = 0 001

I-sratistic at F

 

.nuo

boo

68

Figure 8 Estimated event-related average hemodynamic response functions (HRFS) for
pre- (dotted line) and post- (solid line) consolidation right hemisphere ectorhinal cortex
(EcRC, Brodmann area 36) and visual area V3 (Brodmann area 19) seed locations
indicated in Figure 7. HRFS were obtained from the same studentized respresentative
subject data as in Figure 7. Talairach-Tourneaux coordinates for each region are also
given (see Talairach and Toumeaux, 1988).

69

 

 

NORMAUZED
MRI SIGNAL INTENSITY (A.U.)

NORMAUZED
MRI SIGNAL INTENSITY (A.U.)

Ectorhi nal Cortex: (41 . -33 . -6)

1430
1300
1200
1130

1000

800

TIME (TR: 1 TR = 2.5 sec)

visual Area v3: (36.-83,7)

2103

1903

1700

1500

1300

1100

3 4

900

TIME (TR: 1 TR = 2.5 sec)

- - Pre/M
—-Post/M
- - 'Pre/MM
—Post/MM
- - -Pre/C
-—Post/C

 

- - Pre/M
—Post/M
- - Pre/MM
—Post/MM
- - Pre/C
—Post/C

 

 

70

 

CHANGES IN BOLD PHYSIOLOGY WITH CONSOLIDATION

Initial three-way 2x11(2) ANOVAS were performed only on F MRI subject data to test
the effects of the training-testing interval (“time point”: day 0 versus day 7), region of
interest (ROI), and ROI hemisphere (left versus right) on various physiological measures

— metabolic demand (percent BOLD signal change from baseline, PC), cortical
recruitment (activation volume, AV), BOLD Signal rate (voxel-task cross-
correlation,VTCC), and BOLD signal entropy (time series Shannon entropy, TSE) —
collapsed across all ROIs, using hemisphere as a nested variable within ROI. Inspection
of the three-way ANOVA results indicated that only the “time point” factor had had
Signiﬁcant effects; therefore, the effect of the putative consolidation period on regional
metabolic demand, cortical recruitment, and BOLD response rates were subsequently
assessed with a main-effect ANOVA, using time point as the independent factor.
Physiological measures were collapsed across cerebral hemispheres and subjected to t-

tests using “time point” as the grouping variable. (See also Appendix B.)

Group and representative subject percent BOLD signal change SPMS Obtained
from studentized data are shown for the x, y, and 2 planes crossing through ectorhinal
area 36 in Figure 7, and estimated studentized hemodynamic response ﬁinctions from
selected voxels ﬁ‘om entorhinal cortex and area V3 are shown in Figure 8. The reader
may refer to these ﬁgures to supplement the discussion of results in the following

sections.

R01 analysis of metabolic demand (H. 2)

71

The main-effect ANOVA indicated a signiﬁcant effect of time point on regional

metabolic demand over all ROIs (F(2,286) = 81.068, MSE = 2.003, p < 0.001). However,

post-hoe analysis by Tamhane’s T2, applied due to heteroscedasticity between pre- and

post-consolidation subjects (Levene’s test), failed to return any signiﬁcant differences at

 

% BOLD Signal A from Baseline

 

3.0.

2.5-

2.0‘

1.51

1.0-

0.5 '

 

 

 

 

0.0 -
Time ct.
ROI

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

?

N
>

HC
PRC
V

r2 T1T2 T1 T2 T1 T2 T1 T2 Ti T2 T1 T2

co
>

 

_— I

 

 

Figure 9 Percent BOLD signal change from baseline (BOLD signal magnitude) in
signiﬁcantly task-correlated voxels constituting regions of interest at pre- and post-
consolidation for matched associates. Analysis by main effect ANOVA indicated a
signiﬁcant difference in BOLD magnitude between pre- (T1) and post- (T2)
consolidation subjects (p < 0.001) overall, however, post-hoc analysis by Tamhane T2
failed to indicate any differences in per-region signal change with consolidation. These
ﬁndings suggest that the BOLD response magnitude is not affected by consolidation in
signiﬁcantly task-correlated voxels in these regions. Error bars indicate standard errors.

72

the level of the individual ROIs. These ﬁndings may therefore suggest that decreases in
metabolic demand (pre-consolidation BOLD change = 1.145zt0. 123%, post-consolidation
BOLD change = 0.993zlz0.114%) were only apparent when all regions were pooled,

possibly due to the large per-ROI variance in the BOLD signal change data. The present

data thus failed to indicate any changes in metabolic demand with consolidation despite

the Observed decrease in per-region recruitment, contradicting some of the predictions

following from H. 2 (Figure 9).

Interestingly, the fusiform gyrus was found to exhibit signiﬁcant differences
between pre- (6532.292i538.623 uL) and post- (5340.595i498.668 ILL) consolidation
activation volumes (3 difference of 18.24%, p 5 0.05) in the right hemisphere. The left-
hemisphere fusiform gyrus did not differ signiﬁcantly with respect to activation volume

with respect to time point.

R01 analysis of cortical recruitment (H. 2)

The ANOVA also indicated a signiﬁcant effect of time point on cortical recruitment in
response to the presentation of previously-learned (matched) stimulus pairs
(3885.042i114.835 ILL, post = 3148.1262t106.3l6 uL, an average decrease of 18.97%;
F (2,286) = 1010.692, MSE = 1.76 X 109, p < 0.001). Post-hoe analysis by Tamhane’s T2,
applied due to heteroscedasticity between pre- and post-consolidation subjects (Levene’s
test), further indicated that post-consolidation subjects used a smaller cortical volume
concomitant to recall than pre-consolidation subjects in the fusiform gyrus (pre =

6314.938i380.864 uL, post = 5408.875i352.611 uL, a 14.35% reduction in recruitment;

73

 

 

14000«

12000 «

10000‘

Activation Volume (pL)

 

 

 

 

 

 

   

T2 T1 T2 T1 T2

9 9

T2 T1 T2 T1 T2 T1 T2 T1 T2 T1 T2 T1 T2 T1 T2 T
._- o
:I:

PRC
V1

 

 

 

Figure 10 Activation volume of signiﬁcantly task-correlated voxels at pre- and post-
consolidation for matched associates. Analysis by main effect ANOVA indicated a
signiﬁcant difference in activation volume between pre- and post-consolidation subjects
overall (p < 0.001). Post-hoc analysis by Tamhane T2 ﬁirther indicated signiﬁcant
reductions in cortical recruitment in the ﬁrsiform gyrus, premotor cortex, and visual areas
V2 and V3. These ﬁndings suggest that the number of signiﬁcantly task-correlated
voxels is affected by consolidation in premotor areas, area V2, and ventral visual areas,
but not in medial temporal, primary motor, or primary visual areas. Error bars indicate
standard errors; *indicates a signiﬁcant difference in activation volume, p < 0.05. ROI
signiﬁcances computed by post-hoe Tamhane T2.

p 5 0.05), premotor cortex (pre = 11717.083:I:380.864 ILL, post = 96170421352611 uL,
a 17.92% reduction; p 5 0.05), and visual areas V2 (pre = 8528.167ﬂ80864 ILL, post =
6586.702i352.611 uL, 22.77% reduction; p 5 0.05) and V3 (pre = 8356.250:I:380.864

pL, post = 6791.256i352.611 uL, 18.73% reduction; p 5 0.05). (See Figure 10.)

74

ROI analysis of BOLD response rate (VTCC; H. 3)

A main-effects ANOVA was also applied to the cross-correlation, or covariance, between
the voxel BOLD response and a binary reference waveform representing the presentation
of previously-learned (correctly-matched) stimulus pairs. This cross-correlation is
denoted the VTCC for the present study, and represents the “goodness-of—ﬁt” of the voxel
response vector to the binary waveform, such that larger VTCC values correspond to a
better ﬁt between the two vectors, while smaller VTCC values correspond to a poorer ﬁt.
The VTCC therefore approximates the rate of the BOLD response relative to the
presentation of correctly-matched stimuli, both in terms of time-to-peak (BOLD Signal

rising phase) and peak-tO-baseline (falling phase).

The main-effect ANOVA indicated a Signiﬁcant difference in overall VTCC
between pre- and post-consolidation subjects, such that post-consolidation VTCC (mean
R = 0.330:t0.012) was greater than pre-consolidation VTCC (mean R = 0.475:l:0.011)
when all ROIS were pooled (F(2,286) = 1365.942, MSE = 0.018, p < 0.001). VTCC data
was homoscedastic (Levene’s test), and post-hoe analysis by the Tukey HSD test for
unequal sample sizes conﬁrmed that regional VTCCS were indeed greater in post-
consolidation subjects than in pre-consolidation subjects by an average of 30% (p $0.05)
in all ROIs studied, including primary motor cortex. These ﬁndings conﬁrmed the
prediction (H. 3) that mean regional BOLD signal response rates would increase

signiﬁcantly over the putative consolidation interval (Figure l 1).

75

 

 

1.0

0.8 ‘

0.6 .

0.4 -

 

0.2 J

   

Task-Voxel Vector Cross-Correlation (R)
i

 

 

 

 

 

 

 

 

 

 

 

0.0

 

 

 

Time p1 : T1 T2 T1 T2 T1 T2 T1 T2 T1 T2 T1 T2 T1 T2 T1T2 T1 T2 T1 T2 T1 T2
[ROI ‘- 0 00 V --' O O O ‘—
5 o: “I ‘1’ B :I: 2 a: > 3‘ 9
qu 0 O u. o_ D.
II LI
LU uJ

 

 

 

Figure 11 Mean voxel BOLD response-stirnulus presentation covariance (voxel-task
cross-correlation, VTCC) in signiﬁcantly task-correlated voxels constituting regions of
interest at pre— (T1) and post- (T2) consolidation for matched associates. Analysis by
main effect AN OVA indicated a signiﬁcant difference in VTCC between pre- and post-
consolidation subjects overall (p < 0.001), and post-hoe analysis by Tukey HSD revealed
signiﬁcant differences in VTCC in all areas (p < 0.05). These ﬁndings suggest that the
BOLD-task covariance is affected by consolidation in signiﬁcantly task-correlated voxels
in medial temporal, motor, premotor, and visual regions. Error bars indicate standard
errors; *indicates a signiﬁcant difference in VTCC, p < 0.05. ROI signiﬁcances
computed by post-hoe Tukey HSD unequal-N test.

Interestingly, primary motor cortex was observed to exhibit signiﬁcant differences
between pre- and post-consolidation VTCC in the right, but not left, hemisphere (right-
hemisphere pre-consolidation Ml VTCC = 0.437i0.055, right-hemisphere post-
consolidation M1 VTCC = 0.556i0.051, p 5 0.05). As this was the only unilateral effect
in an otherwise globally-affected VTCC dataset, and subjects responded only with their

right hand, this surprising ﬁnding may suggest that left hemisphere primary motor cortex

76

may be exempt from the purported effects of the consolidation interval, thereby acting as
a “negative control” area as previously proposed. The involvement of right hemisphere
M1 in right-handed tasks has been reported anecdotally in the F MRI literature, but there

is as yet no clear consensus on its anatomical basis or its ﬁrnction.

77

RETRIEVAL ORIENTATION AND CORTICAL REINSTATEMENT: INTERPRETATION OF
PHYSIOLOGICAL FINDINGS IN THE CONTEXT OF SIGNAL ENTROPY
The above results suggest that episodic consolidation correlates with a generalized
decrease in voxel recruitment in response to matched associates in premotor areas, area
V2, and ventral visual areas, but not in medial temporal, primary motor, or primary visual
areas, although the BOLD signal in all ROIS (with the exception of left-hemisphere Ml)
exhibited a faster response rate with respect to matched stimulus presentation in post-
consolidation subjects relative to pre-consolidation subjects. Furthermore, per-ROI
BOLD Signal magnitudes in response to matched associates, corresponding to metabolic
demand, did not appear to change signiﬁcantly with consolidation in spite of this

decreased recruitment.

One interpretation of these ﬁndings (see H. 4) is that the instantiation of
differential visual processing strategies with consolidation — particularly, covert shifts in
attention from local visual stimulus features to more global features (“retrieval
orientation”) — and facilitation of network reinstatement through consolidation-related
increases in the synaptic strength among trace neurons (“cortical reinstatement”) enhance
the ability of trace neurons and ancillary retrieval networks to exchange and process
information through synaptic integration, such that fewer neurons are required for trace
encoding and/or retrieval (Rugg and Wilding 2000; Lockhart 2002; Robb and Rugg 2002;
Vaidya, Zhao et al. 2002; Mulligan and Lozito 2006). This hypothesis is testable in that
(a) it predicts the engagement and disengagement of particular brain regions upon recall,
which may be revealed through parametric manipulation of memory task parameters, and

(b) it predicts that the neural networks involved in the sensory processing of learned

78

mnemonic stimuli will differ with respect to the volume of information carried by the
neurons constituting those networks. The former prediction has been tested successfully
with ﬁinctional MRI, but, surprisingly, the latter prediction has not (e.g., Rugg and

Wilding 2000; Mulligan and Lozito 2006).

In the present study, the regions involved in sensory processing of matched
associates were the fusiform gyrus and visual areas V1, V2, and V3. However, BOLD
signal entropy in area VI was not expected to differ signiﬁcantly between time point
groups due to the absence of corticocortical or corticothalamic afferents from attentional
areas onto V1, which would ostensibly control differential sensory processing of
presented associates (Rao 2005). Interestingly, the present data indicated that recruitment
of the fusiform gyrus and visual areas V2 and V3, as well as premotor cortex — but not
area V1 — differed substantially between pre- and post-consolidation subjects. Given
that neither metabolic demand nor VTCC were observed to differ from region to region,
it was Of interest in the present study to attempt to quantify the entropy of BOLD signals
in ROIs that exhibited signiﬁcantly reduced cortical recruitment as a ﬁrnction of the time

point factor.

A main-effect ANOVA on the Shannon entropy of the mean regional BOLD
signals (time series entropy, TSE; see the Methodology chapter), a complexity metric,
using time point as the independent factor, indicated a signiﬁcant effect of time point on
TSE overall (F(l,242) = 10.355, MSE < 0.001, p 5 0.001), such that regional post-
consolidation BOLD signal entropy (see the Methodology chapter) was greater than pre-

consolidation signal entropy overall (pre-consolidation TSE = 0.0322t0.001 Sh, post-

79

consolidation TSE = 0.034:I:0.001 Sh, a 5.88% increase). Post-hoe analysis by Tamhane
T2, employed to compensate for heteroscedasticity in the TSE data, ﬁrrther indicated
signiﬁcant effects of time point on TSE at the ROI level in the fusiform gyrus (pre-
consolidation TSE = 0.048:t0.002 Sh, post-consolidation TSE = 0.055:l:0.002 Sh, a
12.73% increase) and visual areas V2 (pre-consolidation TSE = 0.066zt0.002 Sh, post-
consolidation TSE = 0.073:I:0.002 Sh, a 9.59% increase) and V3 (pre-consolidation TSE
= 0.069i0.002 Sh, post-consolidation TSE = 0.073i0.002 Sh, a 5.48% increase), but not
premotor cortex (pre-consolidation TSE = 0.103:t0.002 Sh, post-consolidation TSE =
0.104i0.002 Sh), which had also exhibited differential cortical recruitment between the
time point groups. Further post-hoe analysis by Tamhane T2 indicated a signiﬁcant
difference between pre- and post-consolidation TSE in the right (pre = 0.0692t0.002 Sh,
post = 0.0750002 Sh, +8.00%, p 5 0.05), but not left (pre = 0.068:t0.002, post =

0.072i0.002, +5.56%), hemisphere visual area V3. (See Figure 12.)

80

Thus, of the four ROIS —— fusiform gyrus, premotor cortex, and visual areas V2
and V3 —— that had previously been found to exhibit decreased cortical recruitment
between pre- and post-consolidation subjects, three — fusiform gyrus, V2, and V3 —
were also observed to exhibit increased signal entropy. However, these observations are
likely to be mediated through a third explanatory variable, since premotor cortex, which

exhibited differential activation volume as a function of time point, did not exhibit

 

 

 

 

 

 

 

 

 

 

 

 

E

(D

ui‘ 0.10‘

In

t:

a
0008‘ *

g a

C

m

g 0.064 t

C

C

(U

.C

(I) 0.041

E

C

.9

0’ 0.024

o

_l

O

In
0.00- .

Time pt_ r1r2 T1 r2 r1 T2 r1r2 T1 T2 T1 T2 T1 T2 r1r2 T1 T2 r172 r1 r2

ROI 0 o o ; gt ;

I II
E o.

Fusi.

‘- O
2 n:
o
LIJ

ERC-28
ERC-34

 

 

 

Figure 12 Mean BOLD signal Shannon entropy (TSE) in signiﬁcantly task-correlated
voxels constituting regions of interest at pre- and post-consolidation for matched
associates. Analysis by main effect ANOVA indicated a signiﬁcant difference in BOLD
TSE between pre- and post-consolidation subjects overall (p 5 0.001). Post-hoe analysis
by Tamhane T2 further indicated Signiﬁcantly greater signal entropy in the ﬁrsiform gyrus
and visual areas V2 and V3 (p < 0.01). These ﬁndings suggest that the BOLD signal
entropy is affected by consolidation in signiﬁcantly task-correlated voxels in extrastriate
ventral visual areas, but not in medial temporal, premotor, or primary motor or visual
cortex. Error bars indicate standard errors; *indicates a signiﬁcant difference in TSE, p <
0.01. ROI signiﬁcances computed by Tamhane T2.

81

differential signal entropy, whereas the converse was true for perirhinal cortex. It should
be noted, however, that ROIs that are comparatively more active at the per-neuron level
need not exhibit comparatively greater signal entropy. For example, the BOLD signal
may increase in amplitude while maintaining its spectral characteristics; in this case,
since no additional frequency components are introduced into the vector spectrum, its
Shannon entropy will be equivalent. This could be the case with an attentional area such
as the premotor COrtex, which might exhibit an increased post-consolidation spectal
power concomitant to “driving” descending (attentional) input to, for example, visual

cortex, but would not necessarily exhibit an increase in its time series entropy.

It is also possible that one or both of these variables are modulated by different
physiological effects. For example, premotor cortex, which appears to ﬁmction as an
attentional as well as motor planning area (Nobre, Sebestyen et al. 1997), and therefore a
possible source of modulation of sensory processing, would only exhibit appreciable
differences in ﬁring rate entropy in the synaptodendritic compartments of visual cortices.
As another example, perirhinal cortex, which functions in the maintenance and
consolidation of pair codings (Eichenbaum, Schoenbaum et al. 1996; Nadel,
Samsonovich et a1. 2000), may not be subject to the putatively consolidation-related
cortical recruitment effects reported for premotor, visual, and ventral visual areas, since it
likely serves as a “linkage site” between trace networks for paired stimuli. Moreover, the
present data are not sufﬁcient to make a conclusive case for the hypothesis that increases
in BOLD signal entropy and decreases in cortical recruitment are directly related, or even
that the entropy of an indirect measure of neural activity such as the BOLD signal

corresponds to the entropy of neural spike train vectors. Therefore, although the present

82

fmdings are probative, more studies are needed both to reproduce a correlation between
BOLD signal entropy and cortical recruitment with consolidation, and to directly test the
correspondence between BOLD and electrophysiological entropy. If this link is
established by future research, however, BOLD signal entropy might be utilized to draw
inferences about changes in the nature and degree of neural activity by noninvasive

means.

Thus effects of both retrieval orientation and cortical reinstatement — two means
by which episodic systems could become optimized with consolidation — may have been
substantiated in the present dissertation. Retrieval orientation effects may be argued to
have been evinced in that extrastriate visual areas were signiﬁcantly less recruited, and
their BOLD signals more complex, in post-consolidation subjects, phenomena that may
be related to more “efﬁcient” (e. g., holistic) stimulus processing and differential neuronal
ﬁring rate responses in post-consolidation subjects. However, the more robust ﬁnding
predicted in the context of H. 4 — that right-hemisphere (global) ventral visual stream
regions would exhibit higher metabolic demands in post-consolidation subjects, was not
boume out. Potential novel correlates of cortical reinstatement were also observed, in
that BOLD signal response rates (VTCC) increased globally; again, however, future
studies are required to demonstrate that this effect is not due to, for example,

angiogenesis over the consolidation interval.

These ﬁndings are summarized in Table 5.

83

 

Table 5 A qualitative review of the results presented in the present chapter and the
conclusions drawn from them. “Per-neuron activity” was determined by dividing the
metabolic demand (PC) of an ROI by the extent of its recruitment (AV) to obtain a
rudimentary AP/mm3 estimate. Abbreviations: N/C, no change; fusi., fusiform gyrus; Vx,
visual area x.

 

 

Demand Recruitment Entropy Resp. Rate Conclusion

Medial- N/C N/C N/C increased -

temporal

Primary N/C N/C N/C increased

visual Consolidation
increases speed of
recruitment only

Primary N/C N/C N/C increased

motor

Premotor” N/C decreased N/C increased Consolidation
correlates with higher
per-neuron activity but
not greater signal
entropy

V2, V3, fusi. N/C decreased increased increased Consolidation

correlates with higher
per-neuron activity
and greater signal
entropy

 

*It should be noted that, although Brodmann area 6 is referred to in this table and the rest of the dissertation
as premotor cortex or PMC to follow convention, it does in fact appear to be an attentional (speciﬁcally,
“motor-attentional”) region. See the Regions of Interest chapter for the relevant discussion.

 

84

NEURODYNAMICS OF EPISODIC MEMORY CONSOLIDATION:
CONCLUSIONS AND DISCUSSION

In the introduction to the present dissertation, it was proposed that an interval of episodic
memory consolidation may function as a temporal window during which memory traces
are optimized and the memory retrieval process made more efﬁcient. This “efﬁciency”
was operationalized schematically, but not quantitatively, as an increase in, or
conservation of, subject retrieval performance (useﬁil output) — reaction time and
accuracy — on a delayed-match-to-sample (DMTS) task and a decrease in regional
metabolic intensity (energy expenditure), characterized by both regional BOLD
magnitude and the degree of cortical recruitment, or the number of voxels signiﬁcantly
correlated with the DMTS task. By this stipulative defmition, increased efficiency over a
consolidation interval could be argued to have occurred in any of several cases. Namely,
the useful output-to-energy expenditure ratio used to illustrate the present work’s
approach to retrieval efﬁciency would increase had (a) subject performance on the
DMTS task improved without a corresponding decrease in metabolic demand or cortical
recruitment; (b) metabolic demand and cortical recruitment decreased without a
corresponding improvement in subject performance; (c) metabolic demand decreased
without corresponding improvements in cortical recruitment or subject performance; or
(d) cortical recruitment decreased while subject performance and cerebral metabolic
demand was conserved. The results provided in the present work indicate that, over a
seven day interval putatively corresponding to a Stage I consolidation period, retrieval

efﬁciency did in fact improve as in case (d). That is, neither subject response

85

 

Emu 85 cosmiomno .8 88065 80: :+.. 2:. doumE—omaoo 3 Bacchus: on 9 333% .3025 3ch m> 88 obiﬂﬁo: «3

£8.33 833:8:0905 9 038—2 amen E 5388 0.8.: AXE haemogcmu 053 Emcwa Odom m> 88 82335:

98.33. 89385063 9 038—2 -58 E baa 3..an 3.585an oh? 822 38%“: 940mm :2 Bonn—mas: 3m: 85 2233030

an 8:865 20: :+.. 2:. dose—58:8 E Bacoﬁag on o. @28ng 5:8 88.: >3th obﬁﬂﬁo: a2 5 88.. 8.8%“: :EwE Odom

.3833 FEES—8:00-05 3 9:28 amon— E 3:38.. $2 3% 3 33355.6 33 mg .5863 Bosnian: 2w:

 

 

35 nous/.530 ca 88065 ob: :I: 2:. .cosauzogoo .3 Bacobacd on 9 333% macaw 88.38 obﬁﬂﬁo: c2 05 .«o u5§_aoo.m _
92 + + I 92 02 92 05¢
92 5+ 92 92 92 92 92 :2

+ + + El I WWZ I _ma
A V+ + + I I Z I n>
n+ + + | I 22 I 92 92 ~>
22 + + 92 I 22 I 9>
92 + + 92 I 02 92 um“.
22 + + 92 I 92 92 cam.
22 + + 92 | 92 22 emu
22 + + 22 I 92 92 o:

.2506. umuoﬁmhn. _m30< UQOEQQ $30.4. 8869a _m30< _m30<

mm... m._.<¢ Odom ksztDmomm .0<E 040m

63* £8.50 355830 .onm £8.80 €52.85 URN £38889; .0: ”QSBEEEV

Agog—otoncémob v—waioxoi BE omcommoc 3&6 DAOm .85 QQQQ Abs—20> :23363 Eoﬁzenx:
.858 .Emﬁzfomx 62:33 80¢ owcmno Rama 940m :8ch ..mos QQOM S2333 dew 6:5 8:88 803%
Km 38:8 88an GEE .xotoo SSE basin <3 £33 8.5.53 55K .m-_ 32¢ :25; .m; .N; .S .528 3:553

438.33 838288905

8 0338 $8.33. comaE—oéooamoq 5 A92 .owcmso 05 :53 co .AIV 32 At 58on wEB 8 328858 0.8 mwﬁwzm
_mawo—ommxam 95 $8323 Eom .v—mﬁ 8368mm 3:8 2: mo £5an ogcmfornoaﬁ Benn—ow 05 so 18% 940mm 2:
no 8558820 .8888 28 33.0. 05 («o meosmtomoc Enema—063a mm :95 mm .3083on was 2:: E508: oocgotoa
80.33 8 80%»: 53> beam :5on of com €953»; mwﬁwcm can Fuouavoisv mcocomooa 2: mo $885 w 935.

 

86

time nor subject accuracy, the numerator portion of the illustrative efﬁciency ratio, nor
metabolic demand (BOLD signal magnitude), a measure of energy expenditure, differed
signiﬁcantly between putatively, pre- and post-consolidation subjects, whereas cortical
recruitment (activation volume), did decrease signiﬁcantly at the ROI level, particularly
in premotor, lower visual, and ventral visual areas. It is therefore argued that retrieval
efficiency did in fact increase, at least in these particular ROIs, between pre- and post-

consolidation testing groups.

Here, the results (Tables 5 and 6) and discussion portions of previous chapters
will be elaborated, and their potential contributions to existing mnemonic theory will be

discussed.

EFFECTS OF STIMULUS MODALITY AND TESTING INTERVAL ON TASK PERFORMANCE

Disparity between behavioral predictions and ﬁndings

Upon analysis, the computer-based testing (CBT) group was found to differ signiﬁcantly
from the F MRI group in that FMRI subjects were slower at identifying correctly-matched
and mismatched associates than their CBT counterparts, and exhibited more false
positives than CBT subjects overall, but neither CBT nor FMRI subjects differed with
respect to accuracy for previously learned associates at either time point. Neither group
was faster or more accurate at identifying learned associates after consolidation, and
neither group improved with respect to the number of false positives or false negatives

with time. These differences were not likely due to differences in training or testing

87

environments or subject compliance, as both F MRI and CBT subjects were trained in the
same environment and exhibited the same reaction times and accuracy for control stimuli
at both time points. Instead, the performance differences could be ascribed to the
requisite differences in attentional demands between testing groups: FMRI subjects, who
were given lexical rather than auditory stimuli, were required to divide their visual
attention between two simultaneously-presented stimuli that were wholly within the
visual modality, in contrast to CBT subjects, who were required to divide their attention
between two simultaneously-presented stimuli in two modalities. Such divided-attention
effects have been well documented in both behavioral and neuroimaging literature, and it
is generally agreed that bimodal attentional demands consistently have an effect on
reaction time during encoding but not retrieval, unless the task stimuli being retrieved,
such as simultaneous retrieval of different (“discordant”) lexical and auditory words,
require access to the same representational areas of cortex (e.g., Grunwald, Boutros,
Pezer, von Oertzen, Fernandez, Schaller and Elger 2003; Naveh—Benjamin, Kilb and
Fisher 2006). In addition to these attentional effects, saccade times were likely to have a
greater inﬂuence on FMRI subject reaction time than in CBT subjects, for whom saccade

times were relatively minimized.

Neither factor appeared to contribute to decreased accuracy for matched
associates, however, as evinced by the absence of signiﬁcant differences between the two
groups. This is likely due to the indexing facility of the hippocampal formation, which
appears to be independent of memory trace modality (Eichenbaum 2001; Ross and
Eichenbaum 2006). More puzzling, however, was the observed consistency in reaction

time accuracy across time points, contradicting the well-known Ebbinghaus retention

88

curve and Wickelgren power law, which predicts a power law relationship between
memory performance and the delay between encoding and retrieval, such that retention
declines with time (see, for example, Ebbinghaus 1913; Wickelgren 1974; Anderson and
Schooler 1991; Rubin, Hinton and Wenzel 1999; Wixted 2004). Here, a possible
explanation of this “non-Kamin” Ebbinghaus exception is presented based on the current

literature in episodic reconsolidation.

Theoretical reconsolidative basis for the non-Kamin Ebbinghaus exception

Although improvements in memory performance —— the so-called Kamin effect, an
exception to the monotonicity of the Ebbinghaus curve — have been demonstrated in
rodents, the mollusk Sepia oﬁicianalis and honeybees, these studies have been done,
without exception, in the context of affective stimuli, such as avoidance responses,
olfactory memory, and appetitive responses, and it is generally agreed that the Kamin
effect is neuroendocrinologically mediated through the amygdalar body (Kamin 1957;
Kamin 1963; Messenger 1971; Sanders and Barlow 1971; Woodruff and Kantor 1983;
Rudy and Morledge 1994; Belcadi-Abbassi and Destrade 1995; Gerber and Menze12000;
Rudy and Matus-Amat 2005). The present study, however, utilized stimuli that, being
abstract, should have had neither affective inﬂuences nor rehearsability, eliminating the

possibility of deep encoding.

Thus a larger question remains: how could subjects maintain retrieval accuracy
and comparable response or retrieval times on a paired-associates task with no emotional
content? A possible answer may lie in the physiology of episodic consolidation. As

discussed in the introductory chapters of this dissertation, episodic memory consolidation

89

appears to occur in two major stages (Chrobak and Buzsaki l998b; McGaugh 2000; Abel
and Lattal 2001; Wang, Hu et al. 2006). In the ﬁrst stage, memory traces are encoded in
' hippocampal area CA3 and interlinked through Hebbian processes over a period of hours
to months to form metastable mnemonic networks that, in the second stage of
consolidation, are gradually re-encoded into cortical, hippocampus-independent networks
over a period of months to years (McClelland, McNaughton et al. 1995; Squire and

Alvarez 1995; Ross and Eichenbaum 2006).

Stage I consolidation, with which the present study is concerned, has been ﬁrrther
subdivided into three phases. In Stage I phase 0, a short-term facilitation phase, recently-
encoded memory traces are associated through a protein synthesis inhibitor-insensitive
AMPAR-NMDAR-PKC pathway, which appears to last up to six hours (Abel and Lattal
2001; Walker and Stickgold 2004; Wang, Hu et al. 2006). In Stage I phase IA, a
stabilization phase, the memory traces undergo protein synthesis-dependent long-term
potentiation through a PKA-tPA-BDNF pathway (Abel and Lattal 2001; Walker and
Stickgold 2004; Wang, Hu et al. 2006). Memory traces can remain in this state for 6-12
hours before proceeding to Stage IB, an enhancement phase which appears to be
primarily if not solely dependent on sleep (Stickgold, Scott et al. 1999; Datta 2000; Gais
and Born 2004; Walker and Stickgold 2004). The proﬁle of protein synthesis shifts in
Stage [B to include several plasticity-related proteins, such as TrkB and the glutamate-
aspartate transporter GLAST, probably in preparation for increases in dendritic
arborization and other structural changes that are the hallmarks of this phase (Cirelli and

Tononi 2000b; Cirelli and Tononi 2000a; Abel and Latta12001).

90

Stages IA and IB are also characterized by competition among neurons
constituting a memory trace through a calcium-dependent calmodulin/CREB pathway,
such that neurons exhibiting a higher inﬂux of calcium will upregulate transcription,
thereby outcompeting neighboring cells for inclusion in the ﬁnal memory trace (Kida,

Josselyn et al. 2002; Bozon, Davis et al. 2003).

Thus it is likely that trace networks are destabilized prior to the onset of
cytostructural changes in consolidation phase IB due to factors modulating short-term
facilitation, such as AMPA and NMDA receptor turnover and differential activity-
dependent calcium inﬂux at encoding, as well as cellular competition within the network
in stage IA (Abel and Lattal 2001; Bozon, Davis et al. 2003). Consequently, the number
of cells constituting a trace would be smaller at the onset of phase IB than immediately
after encoding. The consistency of retrieval accuracy and retrieval time across time points
in the present study therefore indicates persistence in trace network accessibility over and
above this loss of trace neurons: that is, the trace is retrieved as easily after consolidation

than after encoding despite a decrease in the number of neurons in the memory trace.

One possible explanation for this persistence in trace accessibility is that the sum
of synaptic weights remains approximately equal across pre- and post-consolidation time
points, both within the trace network and between the trace network and other retrieval-
associated brain areas, while the activation thresholds of the neurons constituting the
trace network decrease. This conjecture would necessitate an increase in the synaptic
weight, and decrease in the activation threshold, per trace neuron across the consolidation

interval. These two parameters are not entirely independent, since the effective ﬁring

91

rate, and therefore the probability of activation, of a trace neuron is a ﬁtnction of both the
presynaptic neuron's ﬁring rate and synaptic weight and the activation threshold of the
trace neuron. Extending this logic to a population of neurons, then, the sum synaptic
weight of a trace network could be kept constant by an increase in the postsynaptic cell's
ﬁring rate through a decrease in the postsynaptic cell activation threshold, or by an
increase in the presynaptic cell's ﬁring rate through a decrease in the presynaptic cell

activation threshold.

Although this model is, at present, speculative, it would indicate that the trace
retrieval time and accuracy could theoretically be maintained while also maintaining a
relatively constant level of accessibility. Consequently, the neural populations involved in
long-term memory storage and retrieval should both activate more quickly and produce a
more complex signal after consolidation. These predictions were tested in the present
study through measurements of the voxel BOLD signal/task cross-correlation (VTCC)
and Shannon entropy of the voxel BOLD signal time series vector (TSE), respectively.
Furthermore, the number of recruited voxels (activation volume) and degree of metabolic
activity (percent BOLD signal change from baseline) were assessed in order to support or
reﬁite the assumption that the number of neural structures required for trace retrieval

decreased with consolidation.

92

INTERPRETATION OF METABOLIC DEMAND, RECRUITMENT, AND SIGNAL RESPONSE RATE

Disparity between physiological predictions and ﬁndings

As discussed in the previous chapter, the present results suggest that metabolic demand
was conserved across all observed R015 and was not accompanied by a decrease in
recruitment of the MTL, as was predicted. In fact, excluding increased BOLD signal
response rates, hippocampal and perihippocampal areas, as well as primary visual and
motor cortices, appeared to be unaﬂected across the conservation interval, even when
signal entropy was taken into account. Moreover, components of the extrastriate ventral
visual stream (EWS), which were predicted to demonstrate both decreased metabolic
demand and decreased recruitment over the consolidation interval —thereby reﬂecting a
diminution of per-neuron activity — instead exhibited no change in metabolic demand
but signiﬁcant reductions in recruitment (except the left hemisphere ﬁIsiform area, which
was unaffected), indicating that EWS neurons participating in trace retrieval after the
consolidation interval were individually more active than their pre-consolidation
counterparts. BOLD signal entropy data further revealed that, of all ROIs inspected in
the present study, only EWS areas exhibited an increase in entropy with consolidation.
It must therefore be concluded that the consolidation-concomitant Hebbian changes in
episodic retrieval networks (and particularly in the MTL) described in the literature are
not necessarily reﬂected in changes in cortical recruitment, metabolic demand, or BOLD
signal entropy, which may derive from cellular and/or network phenomena that are

unrelated, or secondarily or more distantly related, to Hebbian effects.

93

In the introductory chapters of the present dissertation, it was hypothesized that
areas comprising the MTL, which serve to associate salient features of encoded episodic
stimuli, would also be affected by the proposed optimizing processes of consolidation.
Intuitively, this would mean that changes in plasticity would drive the amalgamation of
pre-consolidation medial temporal interneurons serving to link memory traces stored in
isocortex, such that fewer interneurons are required for the linkage at post-consolidation.
That the entorhinal, perirhinal, perihippocampal, and ectorhinal cortices and the
hippocampus are both involved in and affected by stage I visual episodic memory
consolidation has been reasonably well established: short-term glutamate receptor
blockades, trauma, or neurotoxic lesions in the MTL disrupt visual long-term memory
(Zola-Morgan, Squire, Amaral and Suzuki 1989; Meunier, Bachevalier, Mishkin and
Murray 1993; Burwell, Bucci, Sanbom and Jutras 2004; Winters, Forwood, Cowell,
Saksida and Bussey 2004; Winters and Bussey 20050). Because day 7 subject accuracy
was signiﬁcantly greater than chance — which would not be the case if stage IA protein
synthesis had not taken place — it must also be assumed that at least stage IA
consolidation, and therefore plasticity effects, had also occurred (Abel and Lattal 2001;
Walker and Stickgold 2004; Wang, Hu et al. 2006). Therefore, if these physiological
changes do in fact occur as hypothesized, it is possible that they may do so on a spatial
scale much smaller than an EPI voxel, or may not be correlated with changes in BOLD
signal features. In such cases, conﬁrmation of the present hypotheses in the medial

temporal lobe will require direct electrophysiological observation in vivo.

Resolution issues, however, do not account for the absence of entropy effects in

MTL areas. Therefore, if the theoretical reasoning underlying the entropy metric (see the

94

Methodology chapter, the “Retrieval Orientation and Cortical Reinstatement” section in
the Memory Consolidation and Retrieval Performance chapter, and “Caveats” below) is
correct, and BOLD signal entropy does in fact primarily reﬂect neuronal spike train
information load, it must be concluded that these ﬁring rates are no more complex after
the consolidation interval than before, and, by deductive reasoning, the same (or very
similar) computational operations must be taking place in the pre- and post-consolidation
MTL. This conclusion contradicts the prediction that recruitment of, and demand from,
the MTL areas would suggest increased (compensatory) per-neuron activity, as well as
the underlying hypothesis that trace retrieval efﬁciency increases in the MTL over the
consolidation interval. Thus the question arises: are current models of MTL function
consistent with a position that the pre- and post-consolidation computational operations
of the MTL are identical, in spite of well-documented changes in plasticity in the MTL

with consolidation?

As discussed previously, neither the processing functions nor the temporal course
of involvement of particular medial temporal areas relative to memory consolidation or
storage are well understood; however, tetanic stimulation of hippocampus and frontal
(primary motor), entorhinal, and perirhinal cortices in vivo indicates that these areas are
not only plastic but also capable of effecting, and being affected by, long-term
potentiation (LTP) via neuronal afferents and efferents (Ivanco and Racine 2000).
Moreover, it has been proposed that the characteristic latency between direct stimulation
and steady-state LTP in each of these areas may reﬂect their hierarchical (yet
symmetrical) or time-sensitive ﬁInctions. For example, the hippocampus, which

potentiates quickly and sensitively to novel stimuli, may serve to stabilize memory traces

95

for the duration of stage I, and most of stage II, consolidation; isocortex, which
potentiates slowly but exhibits enduring LTP once it is effected, likely serves as the long-
terrn storage site for memory traces. In this view, components of the parahippocampal
region, which exhibit plasticity latencies and stabilities intermediate between those of the
hippocampus and isocortex, may perform their computational functions on a temporal
interval between the effective periods of the hippocampus (stage I consolidation) and
isocortex (stage II consolidation) (McClelland, McNaughton et al. 1995; Ivanco and
Racine 2000; Eichenbaum 2001). However, parahippocampal activation is frequently
reported during all phases of memory tasks (Eichenbaum, Schoenbaum et al. 1996;
Gabrieli, Brewer and Poldrack I998; Lavenex and Amaral 2000; Nadel and Land 2000;
Egorov, Hamam et al. 2002; Frank and Brown 2003; Naya, Yoshida et al. 2003a; Naya,
Yoshida et al. 2003b; Clavagnier, Falchier et al. 2004). Furthermore, lidocaine
inactivation of rodent perirhinal cortex, localized to Brodmann areas 35 and 36, has been
shown to preclude performance on DNMS tasks when administered over intervals
corresponding to episodic memory encoding, consolidation, and retrieval, which suggests
that parahippocampal areas are vital to the functions of the episodic system in all three
phases of the memory trace lifetime (Tassoni, Lorenzini, Baldi, Sacchetti and Bucherelli
1999; Winters and Bussey 2005b; Winters and Bussey 2005a; Winters and Bussey
2005c). Theories that address the problem of computational dissociations among
parahippocampal areas by positing phasic or temporally hierarchical functions must
therefore do so in light of the seemingly constant involvement of these areas in the trace

network.

96

Despite the current controversy regarding the temporal upper bound of medial
temporal involvement in trace retrieval, most theories of MTL function in declarative
memory concur with respect to its involvement in trace storage and conclude, with some
degree of consensus,4 that both the physiological characteristics of the medial temporal
structures and the protracted nature of consolidation itself subserve the merging of the
memory trace into the myriad of other isocortical memory traces, while simultaneously
precluding, upon trace retrieval, a debilitating cascade of memory recall precipitated by
the overlapping representations of similar, but non-identical, experiences known as
catastrophic hypermnesia (McNaughton and Wickens 2003; Tse, Langston, Kakeyama,
Bethus, Spooner, Wood, Witter and Morris 2007). In this context, then, the ostensible
role of the parahippocampal region is to modulate information transfer to the
hippocampus based on the novelty of the information, such that novel and salient
stimulus features, but not features similar to those found in previously-encoded traces,
will elicit hippocampal activation (see, for example, Fernandez and Tendolkar 2006).
This theoretical function of the parahippocampal region is in line with behavioral,
electrophysiological and neuroimaging data on correlates of repetition suppression,
recognition, and recency effects in the MTL (Henson and Rugg 2003; Weis, Klaver,

Reul, Elger and Fernandez 2004; Gonsalves, Kahn, Curran, Norman and Wagner 2005;

 

4 The most conspicuous exception to this consensus is the multiple memory trace theory propounded by
Lynn Nadel. Nadel’s theory maintains that trace retrieval never becomes completely independent of the
hippocampal formation and, therefore, the MTL is always required for trace recall. It could be argued,
however, that the hippocampal synaptic weight upon trace retrieval approaches, but would never
physiologically attain, zero; by this argument, the upper bound of MTL involvement proposed by Nadel
(inﬁnity) and that proposed by other theorists (possibly, the lifetime of the organism) may differ
mathematically, but in practical terms, the upper bound will always be the lifetime of the organism. Cf, for
example, Moscovitch Moscovitch, M., L. Nadel, G. Winocur, A. Gilboa and R S. Rosenbaum (2006). The
cognitive neuroscience of remote episodic, semantic and spatial memory. Current opinion in neurobiology
16(2): 179-90. and Nadel Nadel, L., A. Samsonovich, L. Ryan and M. Moscovitch (2000). Multiple trace
theory of human memory. computational, neuroimaging, and neuropsychological results. Hippocampus
10(4): 352—68..

97

Montaldi, Spencer, Roberts and Mayes 2006). Moreover, it is 'of particular relevance to
the present discussion because it minimizes the distinction between encoding- and
retrieval-concomitant roles for the parahippocampal areas and emphasizes instead a so-
called gatekeeper ﬁInction, by which neurons engaged in encoding are potentiated toward
novel stimuli (Fernandez and Tendolkar 2006). The gatekeeper theory thus renders
computational distinctions among the various MTL components (where such distinctions
cannot be ascribed to the sensory speciﬁcity or aspeciﬁcity of their projections: cf. Kerr,
Agster et al. 2007) less impedimentary and sustains the contention that the
parahippocampal areas might retain uniform computational ﬁrnctions over and above

changes in plasticity.

EWS areas, with the exceptions of left hemisphere visual area V3, which
exhibited greater putative per-neuron activity but not greater signal entropy, and left
hemisphere fusiform gyrus, which exhibited greater signal entropy but no change in
metabolic demand or recruitment, were observed to exhibit greater per-neuron activity
and BOLD signal entropy at the post-consolidation time point. Based on the theory
proposed in this dissertation (cf. “Caveats” below), this ﬁnding would indicate that —- in
contrast to predicted decreases in per-neuron activity in the extrastriate ventral visual
stream (EWS) consequent to differential retrieval orientation with consolidation — post-
consolidation EVVS neurons exhibit more complex ﬁring rates, a possible reﬂection of
greater information load, than their pre-consolidation counterparts. Interestingly, primary
visual cortex (visual area V1) appeared to be unaffected by consolidation (except for
possible correlates of cortical reinstatement, described below), a ﬁnding which may

indicate that the consolidation-correlated effects observed in the other visual ROIs may

98

be attentionally modulated (see the Regions of Interest chapter). Although this
hypothesis remains speculative pending direct measurement of neuronal ﬁring rate
entropy during a DMTS paradigm, it is interesting to note that premotor cortex — an
attentional area — exhibited greater per-neuron activity but, like parahippocampal areas,
conserved entropy (and, per the discussion above, conserved computational functions) at
post-consolidation. These characteristics are consistent with a hypothetical attentional

region that modulates retrieval orientation in post-, but not pre-consolidation subjects.

Thus, the ﬁrst component of the hypothesis stated above — that the neural
populations involved in long-term memory storage and retrieval should activate more
quickly following consolidation — may hold true for gatekeeper, ecphoric, and trace
networks, whereas the second component —— that neural populations involved in storage
and retrieval should produce a more complex signal after consolidation — may hold true
only for the trace networks themselves, that is, the EWS. Although this model has not
yet, to the knowledge of the author, been proposed in the literature, it is highly testable by
ﬁIture DMTS studies that directly examine ﬁring rate entropy. The present model
predicts that previously-learned visual DMTS stimuli will elicit greater post-
consolidation than pre-consolidation ﬁring rate entropy in visual cortices, where the
traces are stored, but not in attentional or executive cortices, which subserve ecphory, or
medial temporal structures, which appear only to link trace nodes and encode stimulus
novelty. Furthermore, entropy effects should not occur in primary visual cortex. In
contrast, DNMS stimuli should elicit greater ﬁring rate entropy in medial temporal, and

perhaps visual, but not attentional or executive, cortices. It would also be informative to

99

perform these future experiments in an FMRI context in order to attempt the replication

of metabolic demand ﬁndings.

Caveats of the entropy metric

It should be noted that the BOLD signal has been shown to be primarily determined by
the magnitude of perivascular oxidative glycolysis as a result of cooperative, or
synchronized, somatodendritic potentials (Boynton, Engel, Glover and Heeger 1996;
Logothetis 2002; Buxton, Uludag, Dubowitz and Liu 2004; Behzadi and Liu 2005), and
appears to be more closely associated with energy demand concomitant to
neurotransmitter release than with perikaryal metabolic demands (Logothetis 2002;
Logothetis 2003; Logothetis and Pfeuffer 2004; Logothetis and Wandell 2004). The
persistence with consolidation of BOLD signal magnitude within the observed ROIs
would therefore suggest that the degree of perisynaptic activity, including pre- and post-
synaptic currents, neurotransmitter release, and, consequently, neural ﬁring rates,
remained approximately constant across pre- and post-consolidation subjects. It could
therefore plausibly be proposed that ﬁring rates increased, relative to the post-encoding
state, in the recall-associated neural populations of the post-consolidation EWS, and,
consequently, BOLD signal entropy would be expected to increase with consolidation

within these areas.

Again, the hypothesis that ﬁring rates increased in recall areas with consolidation
was tested by computation of the Shannon, or information-theoretic, entropy of the
BOLD response from each Brodmann area. This voxel BOLD signal time series Shannon

entropy (TSE) was found to increase in three of the four regions that exhibited decreased

100

cortical recruitment between the two subject groups, indicating that the BOLD signal per
Brodmann area was more complex in still-participatory neural populations after the
consolidation interval than in the neural populations recruited for recall immediately after
encoding in those Brodmann areas. Although this entropy arguably arises ﬁom increased
neural ﬁring rates, the TSE results could also be the result of nonlinearities in
neurovascular coupling or due to other physiological variables, such as cytogeometry, as
discussed in the Methodology chapter. However, as was argued previously, in healthy,
compliant adults and in children over three years of age, neural activity and the BOLD
signal are believed to be somewhat linearly coupled (Martin, Joeri et al. 1999; Logothetis
2002) and it was inferred that, in the majority of subjects, voxel TSE arises from entropy
in neurotransmitter release in the perisynaptic compartment (Logothetis 2002). The
precise relationship between BOLD signal chaos and neural activity, however, is as yet
unknown. Thus, while the ﬁndings of the present study are probative, they must be
interpreted with caution, since more investigation is indicated as to the nature and origin

of the entropy in the BOLD signal that the voxel TSE represents.

THEORY: CORTICAL REINSTATEMENT AND RETRIEVAL ORIENTATION

An implicit assumption underlying the generation of the hypotheses of the present
dissertation was that subjects would exhibit non-conscious (covert) shifts in attention
from local to global presented visual stimulus features after the consolidation interval, a
phenomenon called retrieval orientation (Rugg and Wilding 2000; Robb and Rugg
2002). In the Aims and Hypotheses chapter, it was proposed that changes in retrieval

orientation would be evinced by a shift ﬁom bilateral visual cortex activation in both

101

dorsal and ventral visual streams at pre-consolidation to right-hemisphere dominance in
the ventral visual stream at post-consolidation (H. 4). This prediction follows ﬁom
divergent activation proﬁles observed in functional MRI studies in which mnemonic
demands were parametrically manipulated (Herron and Rugg 2003a; Homberger,
Morcom et al. 2004; Herron and Wilding 2006a; Herron and Wilding 2006b; Homberger,
Rugg et al. 2006b; Homberger, Rugg and Henson 2006a; Stenberg, Johansson et al.
2006; Woodruff, Uncapher et al. 2006). However, in contrast to those studies, which
utilized the conventional BOLD magnitude metric, no diﬂerences in signal intensity were
observed in the ventral visual stream in the present study, contradicting the predictions

following from H. 4.

Interestingly, however, increases in BOLD signal entropy, which may correspond
to the complexity of neural ﬁring rates, were observed in visual area V2, as well as area
V3 and the fusiform gyrus (V4), two components of the ventral visual stream. This
observation was particularly surprising in area V3, in that right, but not left, hemisphere
V3 differed signiﬁcantly with respect to signal entropy as a function of time point: area
V3 is considered the ﬁrst region that comprises the ventral visual stream (Ungerleider
and Mishkin 1982), and, ﬁIrthermore, the right hemisphere ventral visual stream is
believed to function in the processing of global visual stimulus features (Knierim and
Van Essen 1992b; Knierim and van Essen 1992a; Brown and Kosslyn 1993; Johannes,
Wieringa et al. 1996; Polat and Norcia 1998; Yamaguchi, Yamagata et al. 2000; Haxby,
Gobbini et a1. 2001; Lerner, Hendler et al. 2001). Area V2 and the ﬁIsiform gyrus, visual
area V4, are also known to be heavily interconnected (Albright 1993), and, although V2

is not considered to be a ventral visual component, the ﬁnding that both V2 and V4

102

exhibited greater entropy in post-consolidation subjects may indicate that increases in
BOLD signal entropy may indeed correspond to a ventral visual network effect. Notable,
too, is the absence of a signiﬁcant effect of time point on either recruitment or signal
entropy in area Vl, since, as discussed previously, V1 activity has not been shown to be
attentionally modulated, and it is the attentional network that would likely modulate
differential retrieval effects with consolidation, if they exist. Admittedly, however, the
present data are insufﬁcient to substantiate the previously stated hypothesis that retrieval
orientation effects occurred in the DMTS task; yet these ﬁndings may nevertheless be
seen as seminal in that differential sensory cortex effects were observed in domains other

than the conventional BOLD magnitude statistic.

Possible cortical reinstatement effects

A second assumption in the dissertation was that post-consolidation subjects would
exhibit faster reactivation of retrieval networks, and therefore BOLD response rates
relative to the presentation of previously-learned stimuli, as a consequence of increased
synaptic weighting after the consolidation interval. This effect has been referred to as
‘transfer-appropriate processing” or cortical reinstatement (Lockhart 2002; Vaidya, Zhao
et al. 2002; Mulligan and Lozito 2006). BOLD response rates (VTCC) were in fact
greater in post-consolidation subjects in all observed ROIs with the exception of left
hemisphere primary motor cortex, which did not differ with respect to VTCC as a
function of time point. Thus, the present data did substantiate the hypothesized time

point effect on network reinstatement.

103

Notably, regional BOLD signals in all Brodmann areas except left hemisphere
primary motor cortex were also found to exhibit a signiﬁcant increase in covariance
(voxel-task cross-correlation, VTCC) with a reference vector representing the temporal
presentation of previously learned (matched) stimuli with consolidation, meaning that
both the rising and falling phases of the BOLD response occurred at a faster rate relative
to the onset and offset of correctly-matched stimulus presentation in post-consolidation
subjects. This result could feasibly be interpreted as a consolidation-related decrease in
the activation threshold of the neurons constituting the voxels of interest, such that the
roughly linear relationship between neural activity and the hemodynamic response
remained unchanged with consolidation but increased presynaptic weighting or decreased
postsynaptic neuron activation thresholds facilitated a faster response onset and offset in
the neural activity underlying the BOLD signal. In this case, the altered BOLD signal
trajectory could reﬂect heightened efﬁciency of synaptic integration, such that neural
populations were activated more quickly (lower neuronal time constants facilitate higher
presynaptic ﬁring rates or vice versa) and maintained activity for shorter durations.
Alternatively, however, a change in the metabolic components of neurovascular coupling,
such as increased receptor or ion channel expression, neurogenesis, differential enzyme
kinetics, and alterations to the pathways of signal transduction or time course of the
relevant neurochemical cascades, could also explain the accelerated trajectory of the
hemodynamic response. Additionally, although the positive VTCC-TSE relationship
appears to be preserved in visual areas V2 and V3 and the fusiform gyrus, it should also
be noted that this relationship was not preserved in perirhinal or premotor cortex. If,

therefore, the TSE and VTCC metrics do in fact correspond to neuronal spike train

104

complexity and synaptic efﬁcacy respectively, the conclusion must also be drawn that
one does not necessarily predict the other, and may in fact be mediated through a third

explanatory variable.CLOSlNG REMARKS

The present study yielded probative, albeit preliminary, neurodynamics data that have not
yet been reported in prior FMRI-based studies, particularly in the context of episodic
memory consolidation. Notably, unique physiological proﬁles of recruitment, metabolic
demand, BOLD signal rates, and BOLD signal entropy were reported that suggest that
post-consolidation trace recall and retrieval support networks are characterized by
increased trace retrieval efﬁciency at the regional (rather than behavioral) level, and
differential sensory processing, which may reﬂect changes in synaptic integration both
within the episodic memory system and supporting networks. The reader should be
cautioned, however, that two of the methods used in the present dissertation — BOLD
signal entropy and VTC C -— have not been replicated or interpreted by other researchers,
and their reliability must be established by repeated investigations in the ﬁIture.

It is hoped that the present study will be utilized as groundwork for ﬁrture

neurodynamics research, especially with respect to neuroimaging investigations.

105

APPENDICES

106

APPENDIX A: ANOVA TABLES FOR BEHAVIORAL DATA

A two-way 2x2 ANOVA was utilized to test for the effect of computer-based (CBT)
versus FMRI testing scenarios and time point (immediate versus day-7 testing), the
independent factors, on task performance (reaction time and accuracy, the dependent
variables). Abbreviations: m, match; mm, mismatch; c, control; Acc, Accuracy; RT,
reaction time; FMRI, FMRI/unimodal testing group; CBT, computer-based
(pilot/bimodal) testing group.

Analysis of Variance for match-condition accuracy
Adjusted SS for Tests

 

Source DF Seq ss Adj ss Adj MS F P

FMRI v CBT 1 77.8 51.0 51.0 0.35 0.554

TimePointCFMRI v CBT) 2 393.0 393.0 196.5 1.37 0.264

Error 51 7334.3 7334.3 143.8

Total 54 7805.1

s = 11.9921 R-Sq = 6.03% R-SqCadj) = 0.56%

Analysis of Variance for mismatch-condition accuracy

Adjusted SS for Tests

Source DF Seq SS Ad' SS Ad' MS F P
FMRI v CBT 1 1366.0 12 1.5 12 1.5 6.74 0.012
TimePointCFMRI v CBT) 2 235.9 235.9 118.0 0.65 0.526
Error 51 9236.4 9236.4 181.1

Total 54 10838.3

 

9.77%

Analysis of Variance for control-condition accuracy
Adjusted SS for Tests

S = 13.4576 R-Sq = 14.78% R-SqCadj)

 

Source DF Seq SS Adj SS Adj MS F P
FMRI v CBT 1 327.1 354.8 354.8 1.28 0.264
TimePointCFMRI v CBT) 2 639.2 639.2 319.6 1.15 0.325
Error 51 14190.8 14190.8 278.3

Total 54 15157.0

5 = . 08 R-Sq = 6.37%’ R-Sq(adj) = 0.87%

Analysis of Variance for match-condition response time
Adjusted SS for Tests

Source DF Seq SS Adj SS Adj MS F P
FMRI v CBT 1 1557152 1569349 1569349 67.07 0.000
TimePointCFMRI v CBT) 2 103726 103726 51863 2.22 0.119
Error 51 1193331 1193331 23399

Total 54 2854209

 

 

s = 152.966 R-Sq = 58719% R-Sq(adj) = 55.75%

107

Analysis of Variance for mismatch-condition response time
Adjusted SS for Tests

Source DF Seq SS Adj SS Ad' MS F P
FMRI v CBT 1 446193 36 692 36 692 4.87 0.032
TimePointCFMRI v CBT) 2 176495 176495 88247 1.18 0.314
Error 51 3797985 3797985 74470

Total 54 4420672

 

s = 272.892 R-Sq = 14.09% R-SqCadj) = 9.03%

Analysis of Variance for control-condition response time
Adjusted SS for Tests

Source DF Seq SS Ad' SS Ad' MS F P
FMRI v CBT 1 37284 2 640 2 640 0.23 0.631
TimePointCFMRI v CBT) 2 190481 190481 95240 0.75 0.478
Error 51 6479807 6479807 127055

Total 54 6707572

5 = 356.448 R-Sq = 3.40% R-Sq(adj) = 0.00%

108

APPENDIX B: MAIN-EF F ECT ANOVA AND POST-HOC RESULTS FOR

PHYSIOLOGICAL STATISTICS

Initial three-way 2x1 1x2 AN OVAs were performed on F MR1 subject data only to test the
effects of the training-testing interval (“testing group”: day 0 versus day 7), region of
interest (ROI), and ROI hemisphere (leﬁ versus right) on various physiological measures
— metabolic demand (percent BOLD signal change from baseline, PC), cortical
recruitment (activation volume, AV), BOLD signal rate (voxel-task cross-correlation,
VTCC), and BOLD signal entropy (time series Shannon entropy, TSE) — collapsed across
all ROIs. Inspection of the three-way ANOVA results indicated that only the “testing
group” factor had had signiﬁcant effects; therefore, physiological measures were
collapsed across cerebral hemispheres and subjected to t-tests using “testing group” as the

grouping variable.

" TSE

Testing group

' ROI

Hemi.
1 Testing group
-1: ._1_1,_______ _,, a-
I Pre-Con.

Results of the initial 2x11x2 three-way ANOVA, the main-effect
ANOVA on “testing group,” and the post-hoe t-tests are given below.

Between-Subjects Factors

 

 

 

 

 

 

 

 

 

 

Value Label N I

1 Pre-Con. W _ * *———'3_2I

2 Post-Con. _ _ ‘_ 1541'

1 'Ml _ 26]

2_ MC , __ 391'

3 ERC-28 _ _ _ _26I

4 ERC-34 26I

'5 Fusi. _ "1 ~—— ‘26]

6 HC _ _ _ __2§I

:7 :PMC _ 11 W 26]

s PRC i _ _ _ 26J

9 v1 26I

.105 :v2 _——_ ‘26}

11 v3 . _ _2—61

1 S LH L — — T M _“—143I

1-11 ﬂ __ _ .'_.2.___ R11 1- _ 143I
Descriptive Statistics I
gHemi. I Mean I Std. Deviation I N I
lin— ” I ”.02251’1683Im .0002462143I 6I
IRH i .022656133IWI 0009594639775I
@6611, ’ i J 02253400467" “.0006720651I 12]

 

109

, Post-Con.

=V3

Total

EcRC
ERC-28
ERC-34
F usi.
HC
PMC
PRC

V1 I991

V2

1907849723I

 

 

llO

 

 

 

 

 

 

 

 

 

1 6I

 

 

 

 

 

 

 

 

 

0032996832
010768713071 1 11111 00821210482811 6»
.9977695151 .9991997299I 12;
.995074413 .0021863221f 6i
.9059787711 1111199219970931 16:
.095976592' 119929692969I‘ 1123
.994996960? 11 11 .9019542099I 6I
1.0046709261 11 11 .0018783719I 6.
.9047639431 1 .0018300495I 12!
.9474995851 1 1 0199516547I 6I
.948122593% 1  0200966724I 6;
1.947778964? 1 111;U10190958069I 12!
1.0952229941 1 9021049161I1
.904649119. 1 11 9018678021I 116I
19049929571 11 .0019212903I 112i
102977950j 119921694997I 6.
11913121618199? 11 11 19944204021 I 6;
1.109122976' 11 1 11.0033206321I 1121
.008478885I 1 0013976280I1“ 61
.903396942I111.0014128557I 116I
.9995972881 9013402017I 112I
.914195912I __ 90501047091 6I
.0144525981 .9051148192I 6i
.914929995I 11111 111 00482917291 12I
.965361417 1 1 .0169169178I 16I
.066979599I 119171777214I 6I
.965715968I .9162586746I 12;
1968396898; 11 111.9110750137I 116I
11969499193I 11 11111.0110375615I1116I
196991799871111 9105557898I 12I
19915779931 0338500076I1166I
.031783942 9341854976I 661
.091699938 11111111 0338882348I 132I
1.022697929‘ 1 111 110001433439~ 7I
.022986486 9009733493: 71
“-52.231571 ___.°0.£2533§8' 91
.009000927 0012446179 7;
.008701497I 11 1 9912300932 1117}

 

 

’ 0089151212

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

IT10ta1l I I ' 0011989426 14;

711.111 119951949276 1.0013244864I 7I

ERC-28 IRH 11 i 9091757986 I 1111 90127786021I 7II

_ 7 __ : '_jf;_i:i_¢ “”1? 1‘41
9:11.00553365390034912375 7}

ERC-34' iRH 1 1 199513195117 1 11 19903253545 7|

I161” I 99572161939 11111 .10—003428892 I 11413

31:11 1 9&18319‘ 111 _ (51370275251 _7I

Fusi. RH 054964086.‘ .0038352458I 7I

rT1otal ”9&786111413: . 9936264012 II 1411

IL? _. - 3909918633-; I _ - 0000473126I _ 7_I

_HC {RH 905387920 _9.000446476I 7}

I99! 093123261 9935535331 111

11.11 .1037317141 0006860585I 7;

PMC IRH1104786289 0008732983I 7g

IT6t1a1 11104259000: 11 0009320111I 141

111111 1 - 1098964633 1_ 0003591451T17‘I

PRC 1R9” _903995128 __ _ .0003618858I ___7I

3116131 ___ 903979879 .0003467357I 14I

ILH 916223814 .0000610444I 71

:v1 IRli 1 1 01189559161 .0000243196I_I1_7I

» ITotal .016364700I .0001528675I MI
I _ III1L111 1 11.10719774141I 11 111 11 00.02680646I 7I
va IRl-11111.(I)7—3?1-'a(1)10 999229072137 1171I

IT61311 _ __ 9726102571 _ 2-- (10069772181-111I

. [111 _ ___;0724360431;__ 1 V_ .0003597663I 7]

IV3 IRH I .0745170219IEI 11 .0002322144I 7I

1 - I161? :_ 1 __ .07347653_6__11 .0011182709I 14I
I 1 I111 _22 _ _933818210I _ ﬂ 9340363952I1171'1/I
‘ ITotal IRH_ .I_ _ _._034l9666§I___ﬁ .0346751205I 77I
I. - 4311.1 - - _ .034007437I 0342453058I 154 J
I I131“ .022563115I _ .0001948588I 13I
I iM1 IR_H____ - -d __ __._()22834015_I ___ .0006447040I___3 I
I i IT61a1 22-- .022698565I - _ .0004866363I 26I
ITotal 1511908469602? - .0023807497I 13I
I IEcRC [111111; 1 11.008233409I1111111 .0023140481I 13I
I I E61717, 11 11 1 0083515061 11 .0023033547I 216:1
' IERC-zé I911 11 1 911151911018 11 .0017406991I 13I

 

lll

..._.. _. if ‘_. ._ -_ —_ _ _. __ ﬂ

[RH 2 22 2 2005444340 00168291201131

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

                 

(Total 2 .005467724 00167762841 26!

1LH 0045222092128 2 22 ._ 222—2 0032—58331 E321

ERC-34 .1111 005019214 .00127916661 131

1Tot8l2 * 2 005120066, _ 2222 00125884612261

11.112 22 05129681612 22 2 22 02123659280712222131

Fusi. 1211112 I 2' 2 05180264747 22.0218749E1E1 13;

(Total 2 2 2 .051551645 2 01341559552128

11.11 f 2 005678234 1111 2 00142822741 131

HC 2RH 2 2 005044161 2 ' 06566502121321

Total * 0053611982 22 ‘2' 2.2003614574471226

1L}! 2 2 2 21203383408 22 4 0015275990121 31

PMC 1RH .104085585 .00302382531 131

[Total .103734496 — 2 2 222.06232422220212261

'L11 2 2 22 0037402802 _2 2 22 00097057261 131

PRC 2RH * 0037832242122 43.520927226270321“ 13!

[Total 22 0037617602 2 22 - 200095422461 261

11.1122 2 22 015287444 .003401524521222213I

VI ERH f 2 015558053 22 '2 2033469229821 131

{Total ' 22 015422248 222222.003896271261

LB 1 2 068923877 _ 2 .01144827891 2131

V2 11111 22 069932669222 _2222 2 2201169712191 131

fTotal 22 .0654221322734—222422 2 01135120041 261

11.11 2 070521792 2 2 22 2.002174541803122 131

v3 11111 2 7. . 1 .0722127—229461 2 “10075982398 131

1Total 1 '07‘3123991____,.- 1‘ __ 0074195146 261

11.11 . 032784236 0338494023 1431

Total 11111 2 22 03%011 2‘ 0343499673 1431

Total 22 03293—32668 — 22 034041054212861

1LH 2 f 2396. 761666666667661—343. 7640694090449001 61

1 Ml 11111 2 1 26277083333383ng 270. 7955140261128001 61
3 . 1Tota1 _; 2512.250000000000001 318 7314700496328001 121
I Iui 1152 375000000000001 36.8 9429410491546001 61
2AV sPre-Con. EcRC 111211 _1 2 121402560600000000001 347.4163568400313001 61
i 12Tot:l ' 7221146 43450860600601 341.7213299306109001 121
2 115121 2 780. 583333333333001 2‘ 2216.6233289991331001 61

ERC-28 1112211 2 . 7366666666662666—00l 22 22— 2" ‘ 621

__{Tota1 .2 2 758. 62500000000000 2 219434634759350001 1221

 

 

112

m "7'

'ERC-34 IRH

I Total

111
'Fusi. IRH _
Total

1LH

HC 31111

1 Total

ILH
PMC IRH

‘ Total

'LH
PRC IRH

rﬁiﬁ

Total

v1 IRH
'vz inn
v3 ‘ RH
Total I RH

I Total

‘Ml IRH

I 5 5 11.11

IERC-ZB IRH— " '

iTotal
. 5 1
ERC-34 % ~
: IRH

"'1911291666666667001 480. 778376130965000I 14I

Post-C on IT'otal' 3'

566. 79166666666600 237. 8'81032065750600 6 I

5 555. 50000000000000 ' 226. 227374559313600 I - 6

561.14583333333300 I 221 .403284300939000 I 12 I

6097. 583333333330001" 2661727280670705000I W6;

-.ﬁ __7 __ _-_ ___— __. _—__— ___.-- _-__.._..J

6532 29166666666000 I. 2492. 086492362708000I 6 I

 

63 14. 93750000000000 I 2468. 775000161 103000 I1
629.16666666666600 I 257.032763800000100I 6I

628 91666666666600‘ '7 210 8727499385984001'_6I

..7 _ _ V . _T ___.

629. 04166666666600 I 224.148079943648500 12 ‘

 

 

 

11183.166666666660001 2085.183167413996000I 6';

12251.00000000000000: 52359.040207372481000I 6I

1 1717. 08333333333000 I 21.94 747900032477000 I 12I '

463. 29166666666600 I“ 157. 056949596847400 I 6' I

468. 95833333333300 I 149. 098576172499500 I :6 1

466.12500000000000 I ' 146.0332] 1509013800 I

1689.16666666666700. 631. 852230878918000I '6I

_ _ __ [__2_.- __ ___—___—

18016.2500000000000I 568. 038395489248000I 6I

1745. 39583333333300T1 575. 83654—0276339000I 12' I

8454. 000000000000001 3109.152597573815000I' 6I

8602. 33333333333000' 1 2939.190290992855000Im 6 E

___m ___m -.,1

8528. 16666666666000 15‘2885 612516381395000I121

 

 

 

 

 

8185 70833333333000I 2290 067299320408000I 6I

__.T__.._

8526. 79166666666000 I 2735.842500] l 1559000I 6 I

A ,2, -l,__—_.—_._._ —,w____._ __-L_ H—

8356. 25000000000000 I 2412. 000784429994000 I 12I

3781 .69318181818300I 4048.169774201730000I 66I
3988 39015151515100I 4309. 866734091935000I 66I

3885 04166666666600I—4l66 369184832200000I 132 I

 

 

 

 

 

 

“51771.88095238095200I 583 713556355354000I7 7i

 

2050. 70238095238100' 338. 747214889271200 I ‘_I

 

 

_ 980 98809523809500I 171099632744774900I 7I

937 916666666666001 200. 789535334887400I 7I

959. 45238095238100 I 180. 606275725684200 I I4 I

W...”—

' 63.9 73809523809500| 186. 390311481808800I _ 7I

640. 90476190476100I 181. 309954506570400 I 7 I

640. 32142857142800 I 176. 655390475714000I I 1451

2m- - __ _ J

528.17857142857100 I 125. 600450693003900 l 7 i

' 494. 21428571428500 104. 505325678828900I ~47 I

 

 

 

 

 

 

 

F usi.
HC
; PMC
PRC
V1
V2
: V3
1 Total
1..

EcRC

Total

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Totai 51119642857142800 1112. 3930337019393001 1141
1LH 5477.1547619047601001 129256611173258130001 7i
13H 5 1 5340 5555869E§3i6901 1156092512226161000111171
1Total 5408 87499999999000 1180. 2519988188160001 14g
1} [HA 1 f 1 507. 6666666666660011 1 97.855051535989200 1 7
1RH 465 26190476119004001 '1 11128967123468622001 1715
17611111 1 486,464—281571385001 1 103. 8568711966121001 1:141
1”! 1 9458. 726190476190001 11896. 560—1987022680001 71
1RH 1 9775. 357142857140001— 1829. 605962001238000 111—71
1Total 9617 04166666666000 1797. 8038833954350001 141
1LH 375. 25000000000000 1 1 80. 4054828147102001 71
{RH 346. 73809523809520 82. 9430922236387001 71
7Tot1a1 1 360 99404761904760 1 79 8618228720206001 1141
.FLHh 133148809523809500 480. .4003910430190001___7J
1RH 1380. 083333333333001 3198.0434440275751001 71
1T0tal 1 I 1355. 785714285714001 1111424. 5906584686622001 11141
1L1H“ 1 1 1 6487. 77380933435001 “21104297206324278300011M71
[RH—1 6685.6310952380950001 1690.2528067112440001 71
[T6011 6586. 702380952380001 1804. 293010532696000 1 1411
mu 6615 04761904761000' 1486 3162570169970001 71
SR}! 1 6967 464285714280001 1867. 0766500767720001 :71
1To¢1al 6791 .255952380950001 16315485488218540001 141
1LH _ 310671753246753200111 3278. 0672960445610001 771
1RH 1 .1 11 3189. 533549783549001113368 2376713690550001 _771
{Total 31148. 125541125540001 331—28400112044670001 1541
1LH11 11 2060. 301112820512182001 569.8579100620220001 131
1112* _ 317012820512820001 421.3854201033076001 131
1Total 2188. 657051282051001 508.1731913470500001 261
11.111 11060.089743581974—3001 281.5350601923699001 131
113111: __111_110314.166666661666_001 1., 285.4791354995239001 131
-_ 1:93" _ J 1045. 753205128205001 _ 278.170298039777100 261
11.11 704. 743589743589001 205.581780782596600 131
1 RH ” 685.1025M4m1 1207.0975547146885001 131
1 17691 : __ 694 923076923076001 202. 4200279988192001 _261
1LH 545. 999999999999001 178 5139467007176001 1131
1181111 1 1 I 522. 500000000000001 ”11116627225427169343001 131
176115—111 534. 250000000000001 1 169.6529693226735001 1261
11 In 1 1 1 5763. 506411E641—0001 1 1972.5607045597040001 131

 

114

 

IPC

Pre-Con.

IRH

I -_.--.., _

I Total
I ILIH

HC R]! I

v1 ,HH '

v2 RH I

v3 'RH I

'Total I l-I
I LHI

Total IRH I

r ,2 “.33

ﬁﬁ

:LH

Ml [RH I

I ___-- - -
I Total

:ILH

EcRC IRH i

I ____. __.. __ - . -

I Total

ERC-28

Total '

IRH I

I 5890 60897435897000

5827. 05769230769000 I

 

l907. 439834005929000 IIIIl
l902.l76409737l95000 I3 26 I

 

 

563. 74358974358900 I

 

190.498534131919800I 13I

 

 

 

 

 

415. 88461538461530 I I I

 

I 403.14743589743580 I III

I 409. 516025641025701

 

1496.5705128I2051200 I

gr‘w

1574. 64102564102500 I

1535. 60576923076900 I II
7395. 26282051282000 I_

7570. 26282051282000 I

 

 

 

540 79487179487100I 179.197260357568300I 13:

552. 26923076923000 I 181 575260349366300II I26 I
Ii0254.621794871731§50I 2100.176776238846000I 13I

I 10917 96153846153000I 2375 409863419525000I 13I
I I” I I10I58—6I2I9T66I6I66666000I 2222 610191900255000I 26:

 

1.248891014470694OOI 13I
129.321887710571400I I133
124 725585496121900I 26
562 303189691070000I 13I
5511376313727428000I 13‘:
528. 087298421646000I I261

 

 

 

 

 

 

 

2674.990017255117000I 13 I
2452. 973657833618000 I

 

7482.76282051282000 I

2516.115091177523000 I 26I

 

II 7339. 96794871794000 I

1988.447561372168000III13I

 

* 7687.15384615384000 I

2348.681255731743000 I I13; I

 

7513. 56089743589000 I

3418. 24475524475606I II

3558. 23659673659500I:
3488. 24067599067600 I

.85089145 I

.38727509 I
.61908327 I

.6I8080263II

_._ _ ___

I I1.17851141I
.92965702I

 

 

 

174269839 I II

2139. 403893 164543000 III 26' I
3656. 043869491610000I 143I
3838. 530913552827000 I E3
3742. 473090933987000 I 286 I

1.537649016I 6II
IIIIIIII5558IIII982I74I I6I
1I.1_2I862I5I7I67I 1I2I
III I IIII7I9911136I7' 6'
1.422076321 I 6;
1.130066059I I12I

2 259973998 I 6I

 

 

 

 

 

 

 

 

 

 

I.90825018I
II I32547428I
III77425600II
L06495|08II

.91960354IIII

 

 

 

1.67876346

r-_ __ _

.891324727I 6I

___—__4

1.694875235 I 12 I

 

 574329137I 6I
1 .686967330I 6
1 .2110128951 12:

 

I2I. 149966979 II

 

198403296 i

I 1831398211

I
.J
I
I
II
I

6
2463441070I II6
2.210184443' I123

 

 

115

E Post-Con.

HC

, F usi.

.lﬂC

 

PMC

PRC

V1 .8313

V2 RH

.V3 RH

041 31111

Total
I LH
RH

Total

.3, __M ..

'LH

EcRC

ERC-28 LIB

Total 3

ILH

ERC-34 ERH

I Total

I111
éRH

Erouu Q

.811

“EL“ _

108606636 3

.69826655

3.39216644;

.81440086

".93062219

.87251152

3.88558190.
.63907800:
.76232995;

124357595

.95398377

109877986 3 3

2.07361138
167178188

'1. 87269663

1. 89407285

3 2 05014999

 

1972111421

1 .15679284

'L13335483

114507383

31.28997136;

.59479899
.94238517

85655908

 

58876108:

.72266008E

.74865004 I

 

 

            

 

 

 

 

 

 

1.28373235

1.24;345362 ‘

.99605 183;r

1. 10803735

 

1 .45942736

 

.92817579

131251708

 

____l 12034643 ‘1

.52380035

32784621 "—

 

 

 

 

 

 

 

 

 

‘0983872273' 6;

'33 .8686665263 61

3 33' 67051042333_32j
1 .323246837I ' 6:

7 16837757023 3363

_ _ 136647656393 123
A 1395247788I—338

" 9452760993-6i

_ 1143503746I '12I
‘1'6486988363d"_63
___"___—__ M63

3 3 1.395487637; 123
3 _ 233827217373 3 6g
___-3 187269666233_~6E

_- '22366609648I 123
2.333836180I 6;

3 “—33322442076943I 6I
2- ' 2278870050I” 12
3‘ 1 669374766I37661

_ 1546187388I 66

'1 602850797I 132I
21306085183 71

‘- .6628718893— 7;
33' ‘3 _ 1.558223273333743

' .936790008I 37E
3.532664040E 71
.74518166OE 14E

3373333 .53593001333 73
-3961349252I 73

733 .790588253E 14I
.961041938I 7I
L483619664E 71

7'33 3 L214671079E 14E
___33;:::m.909673342E3 73
1.7089ISOSSE 7:

33 3'1 .330253114El:143
___”"3_“_:601616115 ' 7E
373m_33ﬂ_ 367622419 __7E

 

116

Total

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

     

 

 

 

 

 

 

 

 

 

     

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Total 4253232811_ 4701746081841
1LH 1.12738879 1.6059947371 71
PMC 11111 _1.29061159.j_ 22492676861—7I
11:61:11 126966107191 8 -_‘ 18795210081 14;
11 4739348 _ if; “‘7‘
PRC 11111 _ 113793967 _ 11371829171 _71
_1T61a1 1056:4842 1.1489490821 141
1111 #:6812513 3157196931 __71
VI 11‘” 119914236; :774550423_[_ .71
1Total 89427785 .609590648 141
1L11 I. 1.189081461 8- __ 14480037661 —71
1v2 11111 1267383601 71389250583851 71
Tota1 - 122132561 1136387037171 1_41
1LH *' 1 987883201 “5287—4202471nj
v3 11111 1 1090301591 _ 1.0012938981 71
1Total 1038814901 .8609635881 141
1L11 __ 946869361 _ 1.1013870111
Total 1111 103818028 __ _ ‘4 “12020372651 771
@6611 99252482 _ 1. 1499501781 1154
LH 108732101 _ _ _ 18184618541 :31
Ml 1111149901873: 600044.8701 131l
176ml ' ”— 783868—911 — 13601752971 261
. 1L1: .77544072 1“ 1*.8444992361-1—31
EcRC 111118603541 7 10383308051 .13“;
1T'oka1 81817713, 8 19282941071 261
1 L11 ' _- 18207441591" "1.593036019118181
ERC-28 18117 1088744341 “___ 9073986051 __
176m 1.114809296 7 1.2716164611 261
“HH— _— 853984421 7932433101 131
ERC-34 11111 _ ___—1277361381 1 m 1.5258550241 131
1fota1 ' ~7‘1.115672901 - “—1.2028190951 261
81:11” 1- 1 1274600871 _ 1.5784216891 131
1Fusi. 11111 . 1“ 1.622447481 1 20273541471 131
' 176611" 7 1448524181 _ _ 17889152091 261
. 1L11 ' 321769271-— .4863903181 131
'HC 111111. 1 .498809441m .631396776 131
. 112.31 7 ”741386861_M_“ "559520513 261
1PMC L11 1 .982932821

 

”1430230644 1 131

 

1 VTCC ; Pre-Con.

PRC

V1

V2

'V3

Total

‘Ml

; EcRC

; ERC-28

iji'

1 Total

—._.r _..._ __.

1L1!
1R]!

i11na1 1

1 105369773

910769582 ;

92070144?

 

.988663391

1.597326011”

11145402896311111

 

w 1.525677481

1.4058257311

11 1.53330854112111

1 4.6956714

1 11.04375712ﬁ

1 108210699§1

1 1 .062932061

.43743954 1
44125479111 1

 

 

.247305871 11

.258324071 11 11
.252814971

 

 

23754938

.230128715

233839071
211'12212227619421

223253831
2254366371
.268864961
321416881

.29514092

 

 

 

 

 

 

112446264. 1

.93370705[ 112

94098075i11
103634603§

___. 7.22 -b ,_2__ _

445070041 11

.2691267111
.307905131

 

r _. ___.

1.9351852741'113:

1668726705 1 261

..... ___.)

1 .2605686421 13}

1 0420677851 131

1.1332018471 26:
1125444544111131
.9688335751 131

1 .0299957581 261

1 201492455511 1 1311

1—12571614437111 111131
1.7719163201 26,

1.6701785851 131
1.7983967961 131

-_.., . ,_ _. . ..___-____ ..___J

1. 701650583 261
1 3913741961 1431
1.3674511871 143;
1 1 .377176277i 2861
.1305505531 61
.0915168951 61

7,_ _____-.. “__.—m, __.—___;

1107563351 12

__2_‘r__ __.1

.1334320421 61

.1349967471 1161
.1280998761 121
.1368883031 61

.1160805211_6J
__.—___ 1 1
J
1

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.12106747611 12
1194239771 61
115075707111 61
.1118357961 12;
.1354434321 61
.1799594141 6!
.1543127521 121
157984561316}
1709309991 61

___4'

 

 

 

 

 

 

 

 

1.288515921

.158227091 1 121

 

2439341621

 

.45319019

 

1467038757 2111

1322243231 6%

 

 

122111173154911 6}
.1200501311 121

 

118

' Post-Con.

'PRC IRH

V1

'IL11
v2 IRE I I

Ir6699

7191 _ .

‘v3 Inu1

Total . RH
511
EcRC RH

ERC-28 IRH I

.27174938 _"

236126992

.28650965

'.29852083'

.357278877 "

.31789985

.397507711_'
.49445642]
.49098206I
.418660991

.44236813;
.430514067

.32011963 7

.339170937 3

152964528

51595200i

.55604798:

.53599999I

I37605029I'

137929010.
.37767919.
.346955711

. 4 q

'138931757f '“

'.37997376"_

'.44337374; ”1

.460852183

.45211296{

.371134611I7I3I
.359045161

.370629957 '3 3

2-_. ..__._ PM

.153988830 6 1

1.69674973 1 6 '

 

__155249116 '121

. -MW»—.V »___ J

I 157580325I 6I

.195611962I 6|

‘1170556456I 121

'1992134955I 61

.9958994621 6I

.6897332601 121
. 102052433 I 61

.087301008161

5691386816I 12I

.148107329I 661

I

 

“7.152711849I 66I
.150156719I 1321
.1453165751 71

.160584513I 71

 

_ .148596682I 141

_—_'__I

.1139486251 7I

__- 1

 

147133333 I 71

126439946I 14I

 

.112547768 I 7 I

 

1152905761 7I

 

16786200021 7!

.110174571I 141

_ __.—J

 

079476069I' 7I

 

 

076564501I 14I
.182738848I 71
1378154971 71

 

__.155758398I_ 141

 

;43954893‘

120255395I 71

 

.49183727' 9
.46569310§
_.58688586I

 

 

 

57030632
578596091

 

.42188486 I

42773914; I

 

165582236I 7I
141650238I 141
 1366171701 7I

 

.119373759 I 7 '

 

.12355259IIWWE4I
.131923089I 7I

__ mum!“ __r..f___.__1

.155150921! 71

 

. Total

Vl

1V2

V3

Total

Ml

EcRC

ERC-28

ERC-34

Fusi.

1R}!

[1 __
{Total} I

:Lll

1 RH
1 Total

-r _ __.

LB

1
1

ﬁnal

. Total l

.___,

6 4248075676

.138389686l8 146'

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

53505787. 1620122921 71

_ 54571616966666 6 61661660055617
5403879368 .1357201931 1461
66543528551'66666 —.l3500265l16766i
54444573766666 “61105914992166_71
54398714666666 66 61165745561 141

6 .5590749376 666 .l0582248l1 71

6 571131937 ' 6666*66" .12157069016 671
.56510343: 66: 966—"16967637016141

.46887548 .1464307811 771
64808293116u66 66666 661461585491mm771
474852401 66 6 1459391011 1541
483237251 1378860871 131
.501385621 _ 66 .1420244591 131
49227143; 666 76 1374517031 261
31662975: 67 6:1355472511 6131
33:59:21 1321333171131
.320044711.1398663571 261
296460481 66 __‘666M_Ti3177174616 131
.366854981.1327156101 131
.3612577316”6 129664963 261
6.314687655 12677726516 131

66 6 ;382610215666666m__ .120513543 1 131
6'6666 .36864893. 66 6-6721213424891— 261
6.36283122 66 .18038689016131

6666 6396497421 666—661688827621 13%

6 6 37966432 6 .1716437291 6261
66 .360892521 66 _ 66215952987416'131
406945511 .18705885016 131

66666U6 383919621 '6666 1719389701 261
___“51'§1§Z§§§? __”-221“___1499259401 131
522644371 _m_2322 1254221971 131

2.22 52071644 ___nm__1554398491 261
.352591561 .157005490| 131

966666 6§36936389166666 16833644516 131
666 _ ”.36097772T 66166 6 .1597101261 26.5
66m66 42588693 666666 66666196386471666131

 

12f)

11111111 44951435 1 18573388 I 131

 

1Total 1 1.1433700614 .1875413401 261
1111:1131 11 3 11.417161133143111. 11 13353511111131
'v2 ‘11111— f 1 4398351218111 1 1 ” 1213139361 131
=Total . 47798479 1.260760351 261
1.111.111 1 f 11 ' 1494361810141 11 1334—7393—6414 1131
V3 119i-" *1 _ 33376313 __. 1%..151
TTotal : .50298526 .1208782961 261
. 1111.11 1 1 46012118931 ’ 1 1614411343911 1431
Total 1RH .4154485216471377611431
#6141 1 11407333331 ' 111644—87034 2861

 

 

Box' 5 Test of Equality of Covariance Matrices(a) 1

 

 

1Box1'sM 11m 11111 1 2091. 6891
F 1 1 I 1 1 34031
an 430
on 11 1 11 11—1 1358317791
181g: 1 . 11111 1 11311 1 1.0001

Tests the null hypothesis that the observed covariance hiatrices of the dependeht variables are 111eq1ual across "
groups. 1

a Design: COHoRT+Co110RT* Ro1+Co_1101RT* ROI * HEM] 1' 1

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

_.___,,_H,_ T. __T__“u 1
Multivariate Tests(d) 1
_____ 7 A

1 Partial ‘
El‘fec t Value F 1Hypothesis1 Error 1 Sig. 1 Eta 11 Noncent. Observed
(If 1 dl’ 1 1 Squared 1Parameter Power(a) 1

' 1 1
1 #11 11 1 11 Yﬂd_ﬁ1—*MWT W 1 1
15.12:? 1.332 119.678 8.000T480000T .0001 .6661 957.427‘ 1.0001
Wilks' '1 1 1 11_ 111 1 1*ﬂ1_—_|1ﬁ 11 1 1111311 1 11111 H 11 11311141 1
| .006 708.227(b) 8. 000 478. 000 .0001 .922 ? 5665. 814 1.000
1L8Mbd8 l - 1 1 T

.COHORT “T ” " W“ ‘”'" 1 1 '
131113111111 S 1108 3331 3222.910; 8. 000 476 0001 000 982 25783. 277 1.000
1. 2-2- _ --- . . __.-.. i i __ _ _.
Roy's '1 1 , 1T 1 1 1 1 l
Largest T107815 6468.893(c)! 4.000 240 000! 0.00; 9911258755721 1.000
TT .. __T ___- ___________ T T ___. .T_.________.,___.-______
1 ma” 1 1.329' 6.022? 80 000 T.968 000 0001 .3321 481.766 1.000

Trace 1 1 1 1

i Wilks' i i T T 1
ICOHORT Lambda 1 .005 32.573. 80 000 945. .2461 000 .729i 2541.679. 1.000
1*ROI I __ -ﬂTi _ T 2 __.__;__.______.._TT .--. ._ 111
1 1%:311131151 127. 988 379.964 80 00011950 0001 000 970130397132 10001
1 1116;; 1.127.592 I 1454313616161 210. 000 242 0001 .000 11 9.92 331817131311 11 1.00011

4__._. r.._— __.—__.. __-_ ___—__.- ____ ___—__.— ______ -. ___— ____- ____.__ _. _ __4

121

; Largest

 

 

 

 

 

 

 

 

 

 

 

 

’Root
""3"“ .077 .217 88.000 968.000 1.0003 .019 3 19.086 1 .363 '-
Trace 1 | . ' i
. * 1T“ ‘5'“ ' ‘ .“ "‘1
WW“ .924 .216 88.000 947.548 1.000 .0191 18.789 .356I
COHORT Lambda 1 1
*ROI* :7 ~7 2', I I T T A“. WT ‘_ -__ A M _ _ _i-
HEM. #3:?“ s .080 .215 88.000 950.000 1.000 .020 . 18.944 . .359 ;
‘Roy S — i m" I — _ ' " I I m T f T 'T __u T
Largest .041 .456(C) 22.000 242.000 .984; .040 . 10.037 3 .3581?
.Root - '
a Computed using alpha = .05 i
b Exact statistic ‘
3c The statistic is an upper bound on F that yields a lower bound on the signiﬁcance level. y
d Design. COHORT+C0110RT * R01+coHORT * ROl * HEMl I
Levene' s Test of Equality of Error Variances(a) 1
1 . -2." ._ _ _ __ .1
F dfl : an Sig.
TSE 5.014 43 242 1 .000}
AV 4.497 43 242 : 000 f
PC 3.011 . 43 242 ' .000:
VTCC .491 f 43 j 242 ' .997!
Tests the null hypothesis that the error variance of the dependent variable IS equal across group. i
a Design: COHORT+COHORT ROI+COHORT ROI * HEM] i
Tests of Between-Subjects Effects
1. ’ "' 7 i I ‘ A ‘T _ _ _ A H I
I 1 I Partial T
Source iDependent . Type III Sum of ; df _ Mean Square i F ! Sig. Eta Noncent. Observed.
‘Variable . Squares . T Parameter Power(a)
1 1 Squared |
{1513.63105) 44 .014 386. 273T .000T .986 I 16996. 012T 1.000!
Mod 1 'Av 705049098413016) 160238431 457 92. 055T .000 .9441 4050.407 3 1 .000!
e - --—--- 1; .
PC 378. 903(d)1 44 8. 611 1 4. 299; .0001 .439 I 189153 I 1.000I
EVTCC 50.93416) g 44 1.158 64. 449 i 000 I .921 g 2835. 752 1 1 .000|
gTSE ‘ .311 . 2 ' ..155 4179. 749T .0001 .972 i 8359.498i 1 .000!
in 3518595376. 323 2 1759297688 161 T1010 692T“ .000 .893 i 2021.383i 1.0001
COHORT . ----—— -- -— --—- — : - 1:
IPC 324. 784. 2 162. 392' 81 .068T T000. .401 I 162.137i 1.000;
iVTCC 49. 069 2 24. 534 1365. 942 | .000 .919 2731. 884 1 .000;
TTSE .321 i 20 .016 431. 7771 000 j .973 i 8635. 548 1 .0001
$33?” 7v . 3526000367.240I 20 176300018. 362. 101.282T000; .8931 2025. 637 1.000;
' il’C 43.192 20 2.160 1 078 .373 ' .082 .I 21.562 .7821

122

___—l

 

.091 '

 

.000 I >

' .295 __

.000 '

 

 

 

 

VTCC 1.819 20 5.064E 9101.286ij1
TSE 35913-005; 22 16313-006 E 704411.000; .004. _ "1967 __ _ .068.
5333331” 5895240. Q7; 22' 12679863480 7 .154i 1.000 : 014 1 ﬂ 3 387 NEE-”.126...
HEM] ‘PC 10.927 22 .497 248 1.000E .022; 5.455» .189.
VTCC .046: 22' _ .002' .117 : 1.000 _ .011 2.582“ W .104;
TSE .009242 1‘ 317213-005 _ E E 1— ‘ E E ‘
Error AV 42124621170134EZ42 1740682091 1 1 ﬂ— 1 W _ E ___ ﬂ 5
PC 484 762 242 2. 003 1
EVTCC 4 347 242 .018 ' 8 1 1 E
TSE .640 286 7
Total AV 74717372608264EZ86 7 2 - - - -
PC 863.665 286
VTCC 55.281 286 1 1 ﬂ 1
aComputed using alpha: .05 1 — _ - ___ — ‘8 - H
bRSquared= 986 (AdjustedRSquared- .983) if _- _ _ _ _ 1
c R Squared— — .944 (Adjusted R Squared= .933) L i 1 E
d R Squared: 439 (Acljusted R Squared = .337) 1 1 i 1 hm -_
eRSquared= .92l (AdjustedRSquared= 907) — — F 71 8 if -- 1
CUSTOM HYPOTHESIS TESTS
Contrast Results (K Matrix) E
ETesting group Simple E E Dependent Variable . _ E
Contrast“) .TSEE AV E PC EVTch
JContrast Estimate E002! -7369161 .153'5 .145E
HVpanesgeﬁaue 10" E __ 0939—910, 2- 0
EDifference (Estimate- Hypothesized) E 002E -736 9161 .153? 145E
Levelsz.Levell Eskin—m _ __ _ ___ W 001E 156493E .168E: 0161
31;“ 4 EH _ _ E001 “000' 364 .000E
1 ELEM" “00" 1045.179I 4.83. “4E
Easimliglclzidence Interval for ll}f’fg‘Lﬁ . _m ___ ___ ;
i 355:; 004 -428. 653E 1.78E .1772

a Reference eategory— — l

 

 

123

Multivariate Test Results

___—”___ ___rj— __ _.___ ___—___

 

Observed ‘ 1
1

 

 

 

 

 

I a Computed using alpha— = .05

ESTIMATED MARGINAL MEANS
1. Grand Mean

 

 

7 1
Dependent Variable' r

Mean

 

 

Hypothesis 1 Error 1 . 1 Partial Eta 1 —I\_loncent.
Value f F ‘1 df 1 df 1 Slg' 1 Squared 1 Parameter . Power(a)
Pillai's trace .353 32.581(b) r 4. 000 239. 0001 .000 ‘ .353 130.323 1 1.0001
, 1 . :
wm“ .647 32.581(b) 40002390001000 .353 130.323: 1.0001
lambda 1
1122:1111.“ .545 32.581(b) 4000 239 000 000' .353 i 130. 323 1.0001
fogs ""3"“ 7 .545 32.58116) 4000 , 239000.000 .353 _ 130.323 1.0001
a Computed using alpha — .05 r _ _—h .___._ _-_ M
b Exact statistic
Univariate Test Results '
_ 22.3.HF__ 3. 3__ﬁ. 177 _1 _
1 1 1 Partial
Source Dependent Sum of df Mean Square F Sig. . Eta 1 Noncent. Observed 1
, Variable Squares - 1 Parameter Power(a)
. , _ .Squared 1 1
TSP: .000 1 .000 10.355 001 1 .041 1 10.3551 .8931
rAV 38597994. 423 ' 1 38597994. 423 .22. 1741 .000 i .084 1 22.174 ' .997?
Contrast; - -- . - —- —~~—— -J_
1PC 1.654 1 l . l .654 826; .364 .003 .826- .148
.vrcc 1.499 1 1.499 83.4381 :000 .256 1 83. 438 1 0001
.TSE 0097242 .0007 '
IAv 421246276134 242: 1740687. 091 F , i 1
Error ~77 - W- -f—wu—--
. PC 484 762 242 2.003 . J
1VTCC 4.347 242.018 ' 1 1

l ._ ___.

1 Std. Error 1 95% Confidence Interval—"1

 

1 ' Lower Bound 1 Upper Bound

 

1 Lower Bound 1 Upper Bound 1

 

TSE ‘ .033 1 .000 ' 032 1.0341
EAv- 357167.584 .' 78.247, __ 3362—452 H 36%7—151
115C — 1369 "Vi—.084. * V 0903 ; 1 2341
'9'ch 7.402. __ "H.008 _ .387 24181

 

 

124

Dependent Variable Testing group I + w» A Hi 44

2. TESTING GROUP

Pre-Con.
.Post-Con

I Pre-Con.

Post-Con

Pre-Con.
. Post-Con

Pre-Con.

TSE
AV
PC
VTCC
Dependent (1) Testing
. Variable group
Pre-Con.
TSE . - -
‘Post-Con.
Pre-Con
AV
Post-Con.
IPre-Con.
PC
:Post-Con.
Pre-Con.
VTCC
IPost-Con.

I Post-Con.

I
I(J) Testing I
group I
I Post-Con.
Pre-Con.

IPost-Con.

[Pre-Con.

Post-Con.

Pre-Con.
IPost-Con.

 

Pre-Con.

Based on estimated marginal means

Estimates

I22
I Mean

Lower Bound

3.885 042

31418126;
1.145:

i .993:
.330 M
8.475" *'

 

I Std. Error I 95% Confidence Interval

I
J

 

 

IUpper Bound Lower Bound I Upper Bound I

 

I
I
1

 

Pairwise Comparisons

 

Mean
IDifference (I-
J)

1 Lower Bound I

- .002(*):

002(*)

736. 916(*)I

-736916(*)I
.153 :

—.153I 8
i -1450)?
" 5457-61"

 

 

_WhT"

 

 

 

 

 

 

001 .031I 033I
8.000 ”“0337 _ _ 8 035;
881914.835} 8 "8658838 i ”-7171‘245 I
_ 106.316 I 2538-7821—833375488'
‘ _ 123i ' “8802': __ __.—1.388I
114g 8 8.768%"— __ i217:
__.0i82 _ — 830—7— — __. 38534
i .011" h .454. _ _— 8436AI
J
m, _
Std. Sig.(a) I 95% Conﬁdence Interval
I Error I for Difference(a)
Upper Lower I Upper Lower
Bound Bound Bound . Bound
w— 8:0013'I _BOII -.004I -.001I
_ .001; 8.001885" 6.01I W—‘88004I
”#156493 I 8.000} 8 $428858}? TogﬁéI
‘156—493 "—10005? 71045. 179 — 348883?
.168 8647—88—11—78 ' .483!
.168; 8‘ 364—I‘m“ -.483E t— .i78I
.016: .000] — 1778 -_ “ti—144
DWI—“66°.“ m—114I AWE:
I -__,_ _ _ ___ WI

'* The mean difference IS signiﬁcant at the .05 level.

a Adjustment for multiple comparisons: Bonferroni.

125

Multivariate Tests

 

 

 

 

Value F Hg4pothesisT : Error IS—i— Partial Eta I None—git. I Observed 4
df ’ df I g : Squared Parameter , Power(a) I
Pillaistrace .353 32.581(b) 4000 239.000 000; .353? 130323 1-0004
Wilks' I a
.647 32.581(b) 4.000I 239.000I.000 .353 130.323 1.0004

lambda
3334-4“ .545 32.58l(b) 4.000 239. 000.000 .353 130.323 1.000.
:‘o‘zyt‘mgm .545 32. 5811b), 4000: 239.000I 0.00, .353: 130.323i 1.000:

I Each F tests the multivariate effect of Testing group. These tests are based on the linearly independent pairwise
comparisons among the estimated marginal means.

a Computed using alpha- —- .05

b Exact statistic

Univariate Tests

 

 

_. _T __ __ ___ T _
' Partial 4

Dependent Sum of I I 4 Noncent. 4 Observed I

Variable Squares df MC“ Square ‘ F 98'g Eta Parameter I Power(a)
I Squared I

, ‘_. . - _ ..- __ I _ ___ 3__ .__ _,_ __

TSE Contrast .000 I 1 000 10. 355 001 .041 I 10.355 ' .893
Error 4 .009I 1242 3. 7215-005I I I . .
AV Contrast 38597994.423-I1 38597994. 423 I22.174 .000 ' .084 I 22.174 , .997 I
IError 421246276.134:IZ42 1740687. 091I I I
PC Contrast 4 1 6541 1 1 .654: .826 .364 3 .003 1 .826 .148!
I Error , 484.762 I242 2.003 I 4 ' I . j
Contrast 1.I499 I l 1 .499'183. 438 .000 .256 83.438 ' 1.000I
VTCC . . 9 . ;m—-— , mm._— -—9
I Error I 4.347 I242 .018? I 4 I i

The F tests the effect of Testing group. This test is based on the linearly independent pairwise comparisons I
among the estimated marginal means. I
I I

a Computed using alpha: — .05"

 

3. Testing group * ROI

I I I Mean I Std. Error I I 995% Confidence Interval —— I
Dependent Variable I Testing group R0] §~ - - II
I I Lower Bound Upper Bound I Lower BoundI Upper Bound I

 

 

 

 

 

h I II ______ ””7 M71” .023I .002I .019I .026I
IEcRC .008 : .002 : .004 i .011 I
. IERC-ZBI .005 - .002 I .002 ; .009:
TSE Pre-Con. IAN ~~~~~~ 9 - --~ -—— —- -9 -— —- --—~9—-———————- 5
ERC-34 4. .005 , .002 .001 . .008 ;

IFusn 048 .002 9 044 051
HC 005 .002 001 008 4

126

Post-Con.

- Pre-Con.

.AV

1 Post-Con.
1

 

1 131MC
1PRC
1 v1

v2
v3 1

1 M1
1EcRC
“11213028 ;
ERC-34 1
1 F681.
1H1C
PMC
1PRC 11
1v1

1 v2

v3

~ Ml

'1 E6RC
{ERC-1128 3

ERC-3417

31 Fusi.
1111C 11
1864C
113RC11
L V1

1 V21

{V3

3M1

EcRC
1ERC-28 3

1ER1C-34 1

 

I Fusi.

1-- , __
133C 3
{ 9617.042

1PMC
11ka
31311 1
11812
1_V13 _

1 6314.938

11.103
.004
.014
.066 1
.069 11

.0231 1

.009

.006 1 1
.005

.055 1
.006

.1104
.0041

.016

.073 1 11 11

.073 1
2512.250 .

1111 146. 438 1

758. 625
56l.l46.1

629.042

' 11717083

466.125
1745. 396

8528.167
8356.250 .

1911292111

959. 452.

1 640 321

51ll96

1 5408. 8715 ' 1 1

486. 464

360. 994

111111355. 786

6586. 702

67912561 11

127

 

 

 

 

 

 

 

 

 

 

‘ 11.06211 1 _ 100111 1‘1 11.107:

11 1.1002 114.1128005 ___577
1.002 11_1.10111 _-__ .0181:
.0021 .0621 #1- 1.1061911;
1.010211 ﬂ 0651—” 107211
.0027 “1111.020; 1 110121611
11.002 11 11-11——.006 _____ 11101123
.0021 1 11.003131 11111116001911:-
111.0021 1“ 20023 11 _ 1.010911;
.002 .052: .058;
1.002 11.100321 _ 1.00911

1 11.0102 _ 110111 1_ 11.111107;
.0021 1 _0011. “—— 00.71;
1.00231 111 — .0113’11—6—602013
1.002j 11.061911 ___ .0761
1.0021111 1 .070161_ .0777
380.8643 1 1762. 0191; 1‘3262481;
380.8643 396. 2061 18966691
3888641 8394 116311111150—881561
380.8641 389—0861 131131 T 3773

1 380.8641 ' 151516467061—_ 71065116911
1 :64 _:__ __"“
380. 8641 1.0966 852; 12467 3153
380.8641 11 -284 106 1 1216.356;
3801.864 1 1991511641 2495. 627;
380. 8641 1 17777—9353 1 9278. 3981
Q 38:88—95 .‘_‘__'_.Q7__68€°19r__.91£:§13
352.611 1216 7123 2605. 871;

' 1352.6113111112648731 1111116546321
11352116111, 111 111—54. 12518 '1 1“ 1334.901 1
113526111 1 -183. 3833 12057761
__ 3512611 1 54713—2961 111—6111184541;
352611 -208.1151 1181044;
335371" 1.3283562 __193133311
352.611 -333. 585 1055. 573.

1 11352—6111 1111 _661—206 #20501 3615 3
3521.611 1_ 581912423171 __ 1728112182 ‘
352.6111 —16096.677; _ 74858351

 

PC

7 VTCC

Pre-Con.

Post-Con.

Pre-Con.

 

Ml

 

1E RC3;

 

 

”.6197 1" .409; @1861“ 1.4241

__ .9301 .409 1 .125 1 1.734
1.325 1 .4091 .521 2.130

1 1.8311 F .4091 11.0271— 1' 2.6301

 

 

__ _. _ ‘ ..__._.. _ _ _ ._ _... _ -.____ _T. ._
--—— r- - ——-—— - - —— 'T" —————-— —— —— -———-——-——---ﬁ

.920 .409 ; .1151 1.724

 

 

——"_-’"Y‘ --— *—.— ——-— __. ,7

.392 .409 1 44133 1.197

__ __.. -' .-_ --. _-__ . . ___.

.873 i .409 l .068 g 1.677
.762 ' .409 —.042 1 1.5671

 

 

_ 1.099; _ .4091 "12—941“ "19041

1.873 '1 2409.1“ 1.068?

10561“ ’ .3381” ".311—[___1862“

__ ___;_-..1
2.6781

1.972 ”.4691" _' 11.1671" ”—2.717?
9“? __- " @117 7]:Fféi2_§_:§_6__§71
723 1 378 -.022 1 4681
535.3: _ 537Q81’l’ji'j g1"'—j1§_1‘1

1.284 . .378 g .539 2.029

1.1201 1 .378} ' " ‘ 3'7—51“"_1.8654

 

.426; .378' 43191 1.171

__,. __ . ___-.. . __ _-._._ .. _ ___

__-- , _1
1.209 1 .378 ’ .464 1 1.954

 

.894 1".3781' i '.149

 

1328 _37‘81 .483 : G73
1 039 3781 .2941 1 784

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1V2 __1 — 401 _ 039 __ .3251 477

1V3 1 431 039 1 .354; 507
1 — 1M1 I 5361 ' 036| .____ ”.4651 1 607
1 EcRC ' 378 ' 036 1 .307 g 448
fPost-Con. 1ERC-281 _359 1 ”0361 — “288 1 430
' ERC-341 - 1313—01 #036 ”1359 1 4511

IFusi - 452 1 036 .382 f 523 1

 

 

 

128

4. Testing group * ROI * Hemi.

 

.466111 ‘1036. _ __ .395 _: -—i:"_§g61
.579 .036 .508 .6491
.425. .6361 _ .354 __ 14951
.540" .036f .4701'__ 16111
.544' .036 _ '_ .473'___ 61§1
.565" .036' _ _ .495' __.6361

129

 

 

__ _ _.___.._--__1

 

 

A ' 1 _ ' _.-. 3 ' 1
. 1 1 Mean 1 Std. Error | 95% Conﬁdence Interval 1
[Dependent 1Testing ROI 1H i 12- 1 ~~-4 - - 1+“— 1'r A -—— ————
Variable 'group 1 cm ' , Lower Upper Lower Upper 1
1 1 Bound 1 Bound 1 Bound 1 Bound 1
“4] L11 3 .0231 .0021 .0181 .0271
1 1RH .023 - .002 ' .018 ' .0281
1lJi .008 .002 .0032 .0131
EcRC 1—-- ~- 4 -—+- n 1-“. _--
1 1 RH .008 .002 ' .003 . .0131
1 ‘1—‘ ‘ '*7‘ r "" f --“_:
ERG E1.11 ‘ .005 .002, ,_ .0001 __.0_1_01
123 11m .005 .002 F .000 g .0101
Hanc- 1w 1 995' .0025 4.521180% 1 _ -9191
134 1 RH .005 .002 i .000 g .010 1'
11J1 .047 .0021 .0431 .0521
Fusi. ’r- . - . - - r - - -—---
11m 1 .048 .002 1 .043 . .0531
llJl j .005 .002? .000 .0101
.Pre-Con. HC 1 g— ; - - -1 — -- 1- _ ---—
. 3 RH 1 .005 r .002 . .000; .0101
1 1L}! 1 .103- .002 .098 .1081
PthI 1 : - —w 1 - __.- ._ __.»
TSE 1 1RH 1 .103 .002 . .098 , .1081
1 1111 i .003: .002. 40011 .0081
PRC 1 n». » - . w— . —--w1-—————»,
inn 1 .mm .mn1 -00u .mm1
:VW. 11J1 1 .014' .0021 .0095 .0191
- 11111 f .014- .0021 .010; 0191
. _—7 __ .-', - _ . -__._. 1 __. ., __.—__J
1 5 1lJi 1 .065- .002: .0601 0701
.V2 . » --1 - — -—; — ~—
' 1RH .066; .0021 .061 1 0711
1L}! 1 .068 .002: .063 .0731
1V3 11— 1 __- 1 __- 1 ~~—— —-_— —~—
' ! 1nu1 1 .0691 .0021 .0651 0741
1 3 1L}! 1 .023: .002! .018? 0271
1 .841 1——-—; __,-_ —— m ,. — -» - .1— —4 ~4—
1 1 rRH 1 .023 1 .002 1 .018 1 0281
1Post-Con. 1 RC 11.11 i .0091 .002? .004? 0141
1 l c _"'1 _.-7i -, .-_- , . .L_. __
1 1 1RH 1 .009 .002 .004 1 013 1
58110 11.11 1 .006: .002 .001 1 .010!
__ 22,, A, W; 22, -,2_ _____ _ -_-_ - _'_ “_____'__._ ______ _..'_. ._._____._ .____ _. --___._J

28 RH '
ERC— «lfﬂﬂ
34 [R11
rL11
Fusi. 1 _-_ 33
.RH
LH
HC 13+-
1RH
,L11
PMC 1 -:
:RH
‘ L11
PRC . 1
1RH
FL11
v1 1: -
RH
1L11:
v2 1
l
11:"-
v3 1 — —-
1R11
1L" 1
M1 1-
1R11
1 rL11 »
EcRC 1
1R]! .
L r--- .
ERC- 1U}
:28 1R“
. _ -1 .
ERC- 1'1”“
34 1RHI
1LH :
Fusi 1-3-—-
1 11
AV Pre-Con. 1L}! l
11C ;—---———
IRH 1
TLH 1
=PMC

r____- ___“
. L11
1RH .
1m 1
'VI 1‘“—
1R“
11.11 1
1R11 <
v3 L11

 

I
1
l_—.-.——— I.
11111 1

 

 

 

 

 

 

 

 

 

 

 

         

.006—7' _ 4.0027 ' .90171_——010_1
.0063 h 2.002 __. .001-1 “— 0701
”.005 7 .6027 7— 7 ".007 [___01—01
.055 - .002 1 .050. .0591

2 .055 7.002 7— ___.0505”_ — .9603
.006" M .002" _— __ ”.0027 __-0111

.005 ”.00—2W 001~ 6101
7.10477 2.0027277“ —.0991'—__77081
.105 .002 .100; .109;
.0047 7 7' .002— 7 -.0011 7__'.0_091

.004 2 .002? '7” "27-001 1___ __.0691

.017 .002 .012: 0211

.072 .002 .067 : .077 1

' .0723 ".0027 _ ___—.0697; ___—“.6781
.072 .002 ; .0685 0771

.075: .0021 _____ “97.91“ ___.(291
2396 792 538. 623: 1335. 8041 3457. 779:
2627. 708 7 538 623 7 156687211 ”73688—6961
13313333 333633: __77133333' [3.3333631
1140. 500: 538.623. 79.5131 2201487.
' 333333? ._ 333 633' 33.33.33 ..33333331
736.667; 538. 623 -324. 3211 1797. 6541
566 792—I” 573—8 6213 __ 494.1961—7627—7791

- 555500 7 5386—23 505.4871 1616487
6097.583; 538.623 ' 5036 5961 7158. 5711
653272927 __ 538.623 5471 .3041 7593. 2791
629.7671 57387623; —-431 821: ””176E7541
628. 917 538. 6231 —432. 071. 1689.904!
11183767 7 538.6231 ‘101221-79 "722441541
12257696777 5386231 111900131 133119871
463292" "—538. 6231 -597. 69671 1524. 2791

7 468 9581— 538623 :7 :592. 0291 __:E@1
1689.167 538. 623 628.1791 2750.1541
18016251 538.623I_—-—_7747()_;3-81 28626121
8454000: ”7538—623: 7 7393. 01731 7779574987;
' 8602.333? “538.623 7 75413467799663 3211
' 8185.708 _ 538.623 . 7124.721j _ 779246 6961

 

130

Post-Con.

Pre-Con.

 

 

 

 

 

 

 

 

8526. 7921 “538. 6231" 7465. 8941 IIII9I58I7II.I77I91
1771.881 . 498.6683 II 789 5981 I 2752164?
2059702IIII_I498.-1568I 1968419? II 39329861
980.988 I I 498.668 -1 .2951 19632711
937 917II “7496.668 II 44.367: II 19292091
639 738 I 49846681 I -342. 5457 1622—02—11
640. 905 "498.668 I 434379 IIIII1623988
528.179 I7 498.668 454.1051 15494621
494.274 II I498. 668 I 4889691” 14764981
5477.155 II—419I866I81 II 44948711 64594381
53407595 ' 498 6681 4358.312 ' 63228791
597.667 7 4986681 I 474.6171IIII14I899591
I 465.262 498668 7-5779211 14475451
9458—72617III4986681 84764431 II10441.0101
I 9775.357 4986681 I 8793. 07417767576411
375.250 74236581 697—0331 13575331
7139-71? _ ’ 19:69.8 " 33335.51 1188797531
1331.488: 498.668 349. 2051 2313 7711
"1380.083 I49§66IS17I 397 8001 III23II62367I:
648777477" 498.6681 II 551754901 III7479957?
6663763117“ 4986681 IIII5793348 II 7667914‘
6615 0481 “4986681 :5632764 7597—3311
6967 464 498—6681 II 59851811 II794I9.74I8
.851I;—_ .578; -III.287I1 I IIIIII179I89I1
II.387I1 .578I'I I— "—4751; I IIIIII15251
681‘ 5781 -.457-. 1.819:
1:179? I —.5781 —— .040; 2.3171
1.743; “7578; "___—605: 2.8811
998, __ 3.5.7.8: ___ _5 231°] __ I_2I:i°‘I”_6_1

.774 1 .578 ' -.3641 1.9121
1_.065II7_I— 578 ___-.0731 “22.2631
1.679 .578 .541 2. 8171
‘
_ .9813; _ .578 _ ___-1052. 1.2241
.698 . 578 , -.440 1 836?

.8147 Ts78IIu— —;3_24II1I——I1.9531
.931g — _ .578 I 1208 2.0691
886 .578 1 2253 1 2.0241

 

'VTCC

Post-C on.

; Pre-Con.

V1

|
v2 g
v3 ?
Ml

EcRC }
'ERC- 11‘

{_W
‘23 :RH
gERC- “J1

34 {RH .

Fhﬂ. [—-—-

‘ HC
PMC 'r
PRC
V1 r *7
V2

;V3
m‘i ELH
wn

._. '”—"“f_ _.

ILH

EcRC “-
; [RH

1
1
|

r "__—— _TI ' '
'ERC- ﬂJi
L34 LRH
'ng'i. TLH

[Rji"?"

r fV~~—~w~
'ERC- ﬂJi .
E fa" ﬂ

w 1‘ .__ .l ___ .
l

:Aiﬂ‘
J05?

"bxn

.534 5

lih9j
_”1ﬁ4f
’Iiiif
‘"1£3;

_ _Eﬁﬂ
4M6;
oyﬂ

N.
N
-——1'

.Iéi'“

"_xmé7m”

“153?

.337 E

lﬁfﬂ
”'“ESéi
12mg
zsnﬂ
"3:032;

“_Lioé
L&B!
Lani

_Zﬁdﬂ
—"31éﬂ
"““£§n}
‘ _ 11$2E

' _"_f§m§
lagn
___ .* ___. _ J
2J8|

' 234%

_"”—§an
Hus;

___ __- _ J

53in

 

33] 366i

‘ Post-Con.

ﬂ .055"

7200 416F

— A ”.332: _ T5471
_ __.—3597 _ __.—575:
__ _'.164§ _ "380—1
__ .193 ___ E91

__ 191‘; _ “ 7106—1
ﬂ — 2301 _ __.445'"i

 

 

 

RH .404 .055 _ .297 — 5121

3 1LH 419 .055 311 ' .526:
V W 7 7 3 .7. __ -7 . _ _ ___ __ __- ___J
3 RH .442 .055 .335 .5501

.M TLH 516. .051, .416 6161
1 __ .- —— — —— 4 —
1RH .556 . 051 ﬁ .456 .656;

, ‘H__. _ _ . _ _ _ _____;_ _ __ . * 2____..__-J

1LH .376 .051 2 276 476!

EcRC .44— . — — —— - ———----@
1 1RH .379 ' .051 5 280 .4791
[ERG 1LH . .347? .051 1 .247. .4471
{23 1RH .371 .051 f .271 : .471 1
:m- it," 339., _ __ 051 _ 23° _ 5581
134 1RH ' 371 9 .051 , .271 . .4701
' T1LH 1 .443 : 051 .3443 543 1
Fusi — .:. - — — 2-, — — ~ ———-— ------ ——— ———-1
. gRH . .4611 .0511 3611 5611
FHC 1w .440 .051 1 340 .5391
1 1RH ' .4921 .051 . 3921 5921
T 11. .587 1 .051 3 487 . .6871
PMC . ..'. . 1 _ _ .3. ___~ _ __-5J
1 1 .570 .051 1 471 . 6701
1 1 1 .422 .051 .322 1 5221
PRC F” — <1 ———— — - ——— — ——'
1 1 428 051 1 .328 1 5281
iv1 TLH .535 051 1 .435 1 .635!
g 1RH .546 .051 5 .4461 .6451
ivz iLH 1 .5443 .051 g .444? .6431
1 :----——— +—#——-———-- ——— — — ___ -----—
:. 1RH .544 .051 1 .445 .6441

133

.051—1—

"____ ' —' _ Y—

.4511

APPENDIX C: DECONVOLUTION COMMAND LINE

The following command line templates were used to invoke the AFNI program
3dDeconvolve for Figure 7 of the present dissertation. 3dDeconvolve was run on
individual, motion-corrected (“3 dreg”) subject datasets which had been concatenated
and detrended across the four runs (“allruns”). The baseline was specified to consist
of six motion parameters — roll, pitch, yaw, and translations along the x, y, and z axes - as
well as a linear trend. Computation of voxel impulse response functions (IRFs: see
Figure 8) was found to be optimal at a maximum time lag of 4 TRs, or 10 seconds.

3dDeconvolve \
-input 3dreg $subj.allruns+orig \
—automask —GOFORIT 2 -jobs 4 —num_stimts 9 \
-stim_file l match.1D -stim_label l M \
—stim_file 2 mismatch.1D -stim_labe1 2 MM \
-stim_file 3 control.lD -stim_label 3 C \
-stim_file 4 '$subj.allruns.lD[O]' —stim_1abel 4 ROLL \
—stim_label PITCH -stim_file S '$subj.allruns 1D[l]' \
-stim_label RAW —stim_file 6 '$subj.allruns.1D[2]' \
-stim_label dP -stim_file 7 '$subj.allruns.lD[3]' \
—stim_label dI -stim_file 8 '$subj allruns.lD[4]' \
~stim_label dS —stim_file 9 '$subj.allruns.1D[5]' \
~stim_base 4 —stim_base 5 —stim_base 6 ~stim_base 7 \
-stim_base 8 —stim_base 9 \
—iresp l $subj.match_IRF_maxlag_4 \
~iresp 2 $subj.mismatch_IRF_maxlag_4 \
-iresp 3 $subj.control_IRF_maxlag_4 \
-sresp 1 $subj.match_sdIRF_maxlag_4 \
-sresp 2 $subj.mismatch_sdIRF_maxlag_4 \
—sresp 3 ssubj.control_sdIRF_maxlag_4 \
-stim_maxlag 1 4 -stim_maxlag 2 4 —stim_maxlag 3 4 \
-fout -rout -tout -vout -nobout -full_first \
—bucket 3dd.$subj.allruns_maxlag_4

\OUJQONU'I

134

REFERENCES

Abel, T. and K. M. Lattal (2001). Molecular mechanisms of memory acquisition,
consolidation and retrieval. Current opinion in neurobiologz 11(2): 180-7.

Aguirre, G. K., E. Zarahn and M. D'Esposito (1998). The variability of human, BOLD
hemodynamic responses. Neuroimage 8(4): 360-9.

Albright, T. D. (1993). Cortical processing of visual motion. Reviews of oculomotor
research 5: 177-201.

Alvarez, P. and L. R. Squire (1994). Memory consolidation and the medial temporal lobe:
a simple network model. Proceedings of the National Academy of Sciences of the
United States of America 91(15): 7041-5.

Anderson, J. R. and L. J. Schooler (1991). Reﬂections of the environment in memory.
Psychol Sci 2(6): 396-408.

Bailey, C. H., M. Chen, F. Keller and E. R. Kandel (1992). Serotonin-mediated
endocytosis of apCAM: an early step of learning-related synaptic growth in
Aplysia. Science (New York, N. Y 256(5057): 645-9.

Bailey, C. H., P. Montarolo, M. Chen, E. R. Kandel and S. Schacher (1992). Inhibitors of
protein and RNA synthesis block structural changes that accompany long-term
heterosynaptic plasticity in Aplysia. Neuron 9(4): 749-58.

Bandettini, P. A., E. C. Wong, R. S. Hinks, R. S. Tikofsky and J. S. Hyde (1992). Time
course EPI of human brain function during task activation. Magn Reson Med
25(2): 390-7.

Bannerman, D. M., J. N. Rawlins, S. B. McHugh, R. M. Deacon, B. K Yee, T. Bast, W.
N. Zhang, H. H. Pothuizen and J. Feldon (2004). Regional dissociations within
the hippocampus--memory and anxiety. Neuroscience and biobehavioral reviews
28(3): 273-83.

Behzadi, Y. and T. T. Liu (2005). An arteriolar compliance model of the cerebral blood
ﬂow response to neural stimulus. Neuroimage 25(4): 1100-11.

Belcadi-Abbassi, W. and C. Destrade (1995). Post-test apamin injection suppresses a
Kamin-like effect following a learning session in mice. Neuroreport 6(9): 1293-6.

135

Blamire, A. M., S. Ogawa, K. Ugurbil, D. Rothman, G. McCarthy, J. M. Ellermann, F.
Hyder, Z. Rattner and R. G. Shulman (1992). Dynamic mapping of the human

visual cortex by high-speed magnetic resonance imaging. Proceedings of the
National Academy of Sciences of the United States of America 89(22): 11069-73.

Bodizs, R., M. Bekesy, A. Szucs, P. Barsi and P. Halasz (2002). Sleep-dependent
hippocampal slow activity correlates with waking memory performance in
humans. Neurobiology of learning and memory 78(2): 441-57.

Boynton, G. M., S. A. Engel, G. H. Glover and D. J. Heeger (1996). Linear systems
analysis of functional magnetic resonance imaging in human V1. J Neurosci
16(13): 4207-21.

Bozon, B., S. Davis and S. Laroche (2003). A requirement for the immediate early gene
zit268 in reconsolidation of recognition memory aﬁer retrieval. Neuron 40(4):
695-701.

Bozon, B., A. Kelly, S. A. Josselyn, A. J. Silva, S. Davis and S. Laroche (2003). MAPK,
CREB and zif268 are all required for the consolidation of recognition memory.
Philosophical transactions of the Royal Society of London 358(1432): 805-14.

Braddick, O. J., J. M. O'Brien, J. Wattam-Bell, J. Atkinson, T. Hartley and R. Turner
(2001 ). Brain areas sensitive to coherent visual motion. Perception 30(1): 61-72.

Breakspear, M., M. J. Brammer, E. T. Bullmore, P. Das and L. M. Williams (2004).
Spatiotemporal wavelet resampling for functional neuroimaging data. Hum Brain
Mapp 23(1): l-25.

Brown, H. D. and S. M. Kosslyn (1993). Cerebral lateralization. Current opinion in
neurobiology 3(2): 183-6.

Buchel, C., J. T. Coull and K. J. Friston (1999). The predictive value of changes in
effective connectivity for human learning. Science (New York, NY 283(5407):
1538-41.

Buckner, R. L., P. A. Bandettini, K M. O'Craven, R. L. Savoy, S. E. Petersen, M. E.
Raichle and B. R. Rosen (1996). Detection of cortical activation during averaged
single trials of a cognitive task using ﬁmctional magnetic resonance imaging.
Proceedings of the National Academy of Sciences of the United States of America

93(25): 14878-83.

Buckner, R. L., W. Koutstaal, D. L. Schacter, A. M. Dale, M. Rotte and B. R. Rosen
(1998a). Functional-anatomic study of episodic retrieval. II. Selective averaging

136

of event-related ﬂVIRI trials to test the retrieval success hypothesis. Neuroimage
7(3): 163-75.

Buckner, R. L., W. Koutstaal, D. L. Schacter, A. D. Wagner and B. R. Rosen (1998b).
Functional-anatomic study of episodic retrieval using MRI. 1. Retrieval effort
versus retrieval success. Neuroimage 7(3): 151-62.

Burwell, R. D. and D. G. Amaral (l998a). Cortical afferents of the perirhinal, postrhinal,

and entorhinal cortices of the rat. The Journal of comparative neurology 398(2):
179-205.

Burwell, R. D. and D. G. Amaral (l998b). Perirhinal and postrhinal cortices of the rat:
interconnectivity and connections with the entorhinal cortex. The Journal of
comparative neurology 391(3): 293-321.

Burwell, R. D., D. J. Bucci, M. R. Sanborn and M. J. Jutras (2004). Perirhinal and
postrhinal contributions to remote memory for context. J Neurosci 24(49): 11023-
8.

Buxton, R. B., K Uludag, D. J. Dubowitz and T. T. Liu (2004). Modeling the
hemodynamic response to brain activation. Neuroimage 23 Suppl 1: S220-33.

Buzsaki, G. (1998). Memory consolidation during sleep: a neurophysiological
perspective. Journal of sleep research 7 Suppl 1: 17-23.

Cabeza, R., J. K. Locantore and N. D. Anderson (2003). Lateralization of prefrontal
activity during episodic memory retrieval: evidence for the production-monitoring
hypothesis. Journal of cognitive neuroscience 15(2): 249-59.

Cantero, J. L., M. Atienza and R. M. Salas (2002a). Effects of waking-auditory
stimulation on human sleep architecture. Behavioural brain research 128(1): 53-
9.

Cantero, J. L., M. Atienza, R. M. Salas and E. Dominguez-Marin (2002b). Effects of
prolonged waking-auditory stimulation on electroencephalogram synchronization

and cortical coherence during subsequent slow-wave sleep. J Neurosci 22(11):
4702-8.

Chance, F. S. (2007). Receiver operating characteristic (ROC) analysis for characterizing
synaptic efficacy. Journal of neurophysiology 97(2): 1799-808.

Chrobak, J. J. and G. Buzsaki (l998a). Gamma oscillations in the entorhinal cortex of the
freely behaving rat. J Neurosci 18(1): 388-98.

137

Chrobak, J. J. and G. Buzsaki (l998b). Operational dynamics in the hippocampal-
entorhinal axis. Neuroscience and biobehavioral reviews 22(2): 303-10.

Chun, M. M. and N. B. Turk-Browne (2007). Interactions between attention and memory.
Current opinion in neurobiology 17(2): 177-84.

Cirelli, C. and G. Tononi (2000a). Gene expression in the brain across the sleep-waking
cycle. Brain research 885(2): 303-21.

Cirelli, C. and G. Tononi (2000b). On the functional signiﬁcance of c-fos induction
during the sleep-waking cycle. Sleep 23(4): 453-69.

Clavagnier, S., A. Falchier and H. Kennedy (2004). Long-distance feedback projections
to area V1: implications for multisensory integration, spatial awareness, and

visual consciousness. Cognitive, aﬂective & behavioral neuroscience 4(2): 117-
26.

Cramer, S. C., R. M. Weisskoff, J. D. Schaechter, G. Nelles, M. Foley, S. P. Finklestein
and B. R. Rosen (2002). Motor cortex activation is related to force of squeezing.
Hum Brain Mapp 16(4): 197-205.

Datta, S. (2000). Avoidance task training potentiates phasic pontine-wave density in the
rat: A mechanism for sleep-dependent plasticity. J Neurosci 20(22): 8607-13.

de Araujo, D. B., W. Tedeschi, A. C. Santos, J. Elias, Jr., U. P. Neves and O. Baffa
(2003). Shannon entropy applied to the analysis of event-related ﬂVIRI time series.
Neuroimage 20(1): 31 1-7.

Deacon, T. W., H. Eichenbaum, P. Rosenberg and K W. Eckmann (1983). Afferent
connections of the perirhinal cortex in the rat. The Journal of comparative
neurology 220(2): 1168-90.

Dudukovic, N. M. and A. D. Wagner (2007). Goal-dependent modulation of declarative
memory: Neural correlates of temporal recency decisions and novelty detection.
Neuropsychologia 45(11): 2608-20.

Duong, T. Q., A. C. Silva, S. P. Lee and S. G. Kim (2000). Functional MRI of calcium-
dependent synaptic activity: cross correlation with CBF and BOLD
measurements. Magn Reson Med 43(3): 383-92.

Ebbinghaus, H. (1913). Memory. A Contribution to Experimental Psychology. New
York, Teachers College, Columbia University.

138

Egorov, A. V., B. N. Hamam, E. Fransen, M. E. Hasselmo and A. A. Alonso (2002).
Graded persistent activity in entorhinal cortex neurons. Nature 420(6912): 173-8.

Egorov, A. V., U. Heinemann and W. Muller (2002). Differential excitability and
voltage-dependent Ca2+ signalling in two types of medial entorhinal cortex layer
V neurons. The European journal of neuroscience 16(7): 1305-12.

Eichenbaum, H. (1996). Learning from LTP: a comment on recent attempts to identify
cellular and molecular mechanisms of memory. Learning & memory (Cold Spring
Harbor, N. Y 3(2-3): 61-73.

Eichenbaum, H. (2001). The long and winding road to memory consolidation. Nature
neuroscience 4(1 1): 1057-8.

Eichenbaum, H., J. Dusek, B. Young and M. Bunsey (1996). Neural mechanisms of
declarative memory. Cold Spring Harbor symposia on quantitative biology 61:
197-206.

Eichenbaum, H., G. Schoenbaum, B. Young and M. Bunsey (1996). Functional
organization of the hippocampal memory system Proceedings of the National
Academy of Sciences of the United States of America 93(24): 13500-7.

Fell, J ., G. Fernandez and C. E. E1ger(2003). More than synchrony: EEG chaoticity may
be necessary for conscious brain functioning. Medical hypotheses 61(1): 158-60.

Fernandez, G. and I. Tendolkar (2006). The rhinal cortex: 'gatekeeper' of the declarative
memory system. Trends Cogn Sci 10(8): 358-62.

Frank, L. M. and E. N. Brown (2003). Persistent activity and memory in the entorhinal
cortex. Trends in neurosciences 26(8): 400-1.

Frith, C. (2005). The neural basis of hallucinations and delusions. Comptes rendus
biologies 328(2): 169-75.

Gabrieli, J. D. (1998). Cognitive neuroscience of human memory. Annual review of
psychology 49: 87-115.

Gabrieli, J. D., J. B. Brewer and R. A. Poldrack (1998). Images of medial temporal lobe
ﬁmctions in human learning and memory. Neurobiology of learning and memory
70(1-2): 275-83.

139

Gais, S. and J. Born (2004). Declarative memory consolidation: mechanisms acting
during human sleep. Learning & memory (Cold Spring Harbor, N. Y 11(6): 679-
85.

Geissler, A., R. Lanzenberger, M. Barth, A. R. Tahamtan, D. Milakara, A. Gartus and R.
Beisteiner (2005). Inﬂuence of fMRI smoothing procedures on replicability of
ﬁne scale motor localization. Neuroimage 24(2): 323-31.

Gerber, B. and R. Menzel (2000). Contextual modulation of memory consolidation.
Learning & memory (Cold Spring Harbor, N. Y 7: 151-158.

Gerloff, C., C. Braun, M. Staudt, Y. L. Hegner, J. Dichgans and I. Krageloh—Mann
(2006). Coherent corticomuscular oscillations originate from primary motor

cortex: evidence ﬁom patients with early brain lesions. Hum Brain Mapp 27(10):
789-98.

Gonsalves, B. D., I. Kahn, T. Curran, K. A. Norman and A. D. Wagner (2005). Memory
strength and repetition suppression: multimodal imaging of medial temporal
cortical contributions to recognition. Neuron 47(5): 751-61.

Gonzalez Andino, S. L., R. Grave de Peralta Menendez, G. Thut, L. Spinelli, O. Blanke,
C. M. Michel, M. Seeck and T. Landis (2000). Measuring the complexity of time
series: an application to neurophysiological signals. Hum Brain Mapp 11(1): 46-
57.

Grunwald, T., N. N. Boutros, N. Pezer, J. von Oertzen, G. Fernandez, C. Schaller and C.
E. Elger (2003). Neuronal substrates of sensory gating within the human brain.
Biological psychiatry 53(6): 51 1-9.

Haist, F ., J. Bowden Gore and H. Mao (2001). Consolidation of human memory over
decades revealed by functional magnetic resonance imaging. Nature neuroscience
4(11): 1139-45.

Hamzei, F., R. Knab, C. Weiller and J. Rother (2003). The inﬂuence of extra- and
intracranial artery disease on the BOLD signal in FMRI. Neuroimage 20(2):
1393-9.

Hausser, M., N. Spruston and G. J. Stuart (2000). Diversity and dynamics of dendritic
signaling. Science (New York, N. Y 290(5492): 739-44.

Haxby, J. V., M. I. Gobbini, M. L. Furey, A. Ishai, J. L. Schouten and P. Pietrini (2001).
Distributed and overlapping representations of faces and objects in ventral
temporal cortex. Science (New York, N. Y 293(5539): 2425-30.

140

Haxby, J. V., E. A. Hofﬁnan and M. I. Gobbini (2002). Human neural systems for face
recognition and social communication. Biological psychiatry 51(1): 59-67.

Haxby, J. V., B. Horwitz, L. G. Ungerleider, J. M. Maisog, P. Pietrini and C. L. Grady
(1994). The functional organization of human extrastriate cortex: a PET-rCBF

study of selective attention to faces and locations. J Neurosci 14(11 Pt 1): 6336-
53.

Henson, R. N. and M. D. Rugg (2003). Neural response suppression, haemodynamic
repetition effects, and behavioural priming. Neuropsychologia 41(3): 263-70.

Herron, J. E. and M. D. Rugg (2003a). Retrieval orientation and the control of
recollection. Journal of cognitive neuroscience 15(6): 843-54.

Herron, J. E. and M. D. Rugg (2003b). Strategic inﬂuences on recollection in the
exclusion task: electrophysiological evidence. Psychonomic bulletin & review
10(3): 703-10.

Herron, J. E. and E. L. Wilding (2004). An electrophysiological dissociation of retrieval
mode and retrieval orientation. Neuroimage 22(4): 1554-62.

Herron, J. E. and E. L. Wilding (2006a). Brain and behavioral indices of retrieval mode.
Neuroimage 32(2): 863-70.

Herron, J. E. and E. L. Wilding (2006b). Neural correlates of control processes engaged
before and during recovery of information from episodic memory. Neuroimage
30(2): 634-44.

Holmes, W. R. and W. B. Levy (1990). Insights into associative long-term potentiation
from computational models of NMDA receptor-mediated calcium inﬂux and

intracellular calcium concentration changes. Journal of neurophysiology 63(5):
1148-68.

Homberger, M., A. M. Morcom and M. D. Rugg (2004). Neural correlates of retrieval
orientation: effects of study-test similarity. Journal of cognitive neuroscience
16(7): 1196-210.

Homberger, M., M. D. Rugg and R. N. Henson (2006a). ERP correlates of retrieval
orientation: direct versus indirect memory tasks. Brain research 1071(1): 124-36.

Homberger, M., M. D. Rugg and R. N. Henson (2006b). fMRI correlates of retrieval
orientation. Neuropsychologia 44(8): 1425-36.

141

Hubel, D. H. and T. N. Wiesel (1959). Receptive ﬁelds of single neurones in the cat's
striate cortex. The Journal of physiology 148: 574-91.

Hubel, D. H. and T. N. Wiesel (1962). Receptive ﬁelds, binocular interaction and
functional architecture in the cat's visual cortex. The Journal of physiology 160:
106-54.

Ivanco, T. L. and R. J. Racine (2000). Long-term potentiation in the reciprocal
corticohippocampal and corticocortical pathways in the chronically implanted,
freely moving rat. Hippocampus 10(2): 143-52.

Jacobsen, L. K, J. C. Gore, P. Skudlarski, C. M. Lacadie, P. Jatlow and J. H. Krystal
(2002). Impact of intravenous nicotine on BOLD signal response to photic
stimulation. Magnetic resonance imaging 20(2): 141-5.

James, W. (1893). The Stream of Consciousness. Psychology. W. James. New York,
Holt.

Johannes, S., B. M. Wieringa, M. Matzke and T. F. Munte (1996). Hierarchical visual
stimuli: electrophysiological evidence for separate left hemispheric global and
local processing mechanisms in humans. Neuroscience letters 210(2): 111-4.

Kamin, L. J. (1957). The retention of an incompletely learned avoidance response. J
Comp Physiol Psychol 50: 457-460.

.Kamin, L. J. (1963). The retention of an incompletely learned avoidance response: some
further analyses. J Comp Physiol Psychol 56: 719-722.

Kang, H. C., E. D. Burgund, H. M. Lugar, S. E. Petersen and B. L. Schlaggar (2003).
Comparison of functional activation foci in children and adults using a common
stereotactic space. Neuroimage 19(1): 16-28.

Kanwisher, N., J. McDermott and M. M. Chun (1997). The fusiform face area: a module
in human extrastriate cortex specialized for face perception. J Neurosci 17(11):
4302-11.

Kerr, K. M., K L. Agster, S. C. Furtak and R. D. Burwell (2007). Functional
neuroanatomy of the parahippocampal region: The lateral and medial entorhinal
areas. Hippocampus.

Kida, S., S. A. Josselyn, S. P. de Ortiz, J. H. Kogan, I. Chevere, S. Masushige and A. J.
Silva (2002). CREB required for the stability of new and reactivated fear
memories. Nature neuroscience 5(4): 348-55.

142

Knierim, J. J. and D. C. van Essen (1992a). Neuronal responses to static texture patterns
in area V1 of the alert macaque monkey. Journal of neurophysiology 67(4): 961-
80.

Knierim, J. J. and D. C. Van Essen (1992b). Visual cortex: cartography, connectivity, and
concurrent processing. Current opinion in neurobiology 2(2): 150-5.

Krainik, A., M. Hund-Georgiadis, S. Zysset and D. Y. von Cramon (2005). Regional
impairment of cerebrovascular reactivity and BOLD signal in adults after stroke.
Stroke; a journal of cerebral circulation 36(6): 1146-52.

Kriegeskorte, N., R. Goebel and P. Bandettini (2006). Information-based ﬁmctional brain
mapping. Proceedings of the National Academy of Sciences of the United States
of America 103(10): 3863-8.

Kumaran, D. and E. A. Maguire (2007). Which computational mechanisms operate in the
hippocampus during novelty detection? Hippocampus.

Lasota, K. J., J. L. Ulmer, J. B. Firszt, B. B. Biswal, D. L. Daniels and R. W. Prost
(2003). Intensity-dependent activation of the primary auditory cortex in functional
magnetic resonance imaging. Journal of computer assisted tomography 27(2):

213-8.

Laurienti, P. J., A. S. Field, J. H. Burdette, J. A. Maldjian, Y. F. Yen and D. M. Moody
(2002). Dietary caffeine consumption modulates fMRI measures. Neuroimage
17(2): 751-7.

Lavenex, P. and D. G. Amaral (2000). Hippocampal-neocortical interaction: a hierarchy
of associativity. Hippocampus 10(4): 420-30.

Lerner, Y., T. Hendler, D. Ben-Bashat, M. Hare] and R. Malach (2001). A hierarchical
axis of object processing stages in the human visual cortex. Cereb Cortex 11(4):
287-97.

Liu, T. T., Y. Behzadi, K. Restom, K Uludag, K Lu, G. T. Buracas, D. J. Dubowitz and
R. B. Buxton (2004). Caffeine alters the temporal dynamics of the visual BOLD
response. Neuroimage 23(4): 1402-13.

Livingstone, M. S. and D. H. Hubel (1984a). Anatomy and physiology of a color system
in the primate visual cortex. J Neurosci 4(1): 309-56.

Livingstone, M. S. and D. H. Hubel (1984b). Speciﬁcity of intrinsic connections in
primate primary visual cortex. J Neurosci 4(1 1): 2830-5.

143

Lockhart, R S. (2002). Levels of processing, transfer-appropriate processing, and the
concept of robust encoding. Memory (Hove, England) 10(5—6): 397-403.

Logothetis, N. K (2002). The neural basis of the blood-oxygen-level-dependent
functional magnetic resonance imaging signal. Philosophical transactions of the
Royal Society of London 357(1424): 1003-37.

Logothetis, N. K. (2003). The underpinnings of the BOLD functional magnetic resonance
imaging signal. J Neurosci 23(10): 3963-71.

Logothetis, N. K., J. Pauls, M. Augath, T. Trinath and A. Oeltermann (2001).
Neurophysiological investigation of the basis of the ﬂVIRI signal. Nature
412(6843): 150-7.

Logothetis, N. K and J. Pfeuffer (2004). On the nature of the BOLD ﬂVIRI contrast
mechanism. Magnetic resonance imaging 22(10): 1517-31.

Logothetis, N. K and B. A. Wandell (2004). Interpreting the BOLD signal. Annual
review of physiology 66: 735-69.

London, M. and M. Hausser (2005). Dendritic computation. Annual review of
neuroscience 28: 503-32.

Maier, N., V. Nimmrich and A. Draguhn (2003). Cellular and network mechanisms
underlying spontaneous sharp wave-ripple complexes in mouse hippocampal
slices. T he Journal of physiology 550(Pt 3): 873-87.

Mallot, H. A. and F. Giannakopoulos (1996). Population networks: a large-scale

framework for modelling cortical neural networks. Biological cybemetics 75(6):
441-52.

Martin, E., P. Joeri, T. Loenneker, D. Ekatodramis, D. Vitacco, J. Hennig and V. L.
Marcar (1999). Visual processing in infants and children studied using ﬁJnctional
MRI. Pediatric research 46(2): 135-40.

McClelland, J. L., B. L. McNaughton and R. C. O'Reilly (1995). Why there are
complementary learning systems in the hippocampus and neocortex: insights from
the successes and failures of connectionist models of learning and memory.
Psychological review 102(3): 419-57.

McGaugh, J. L. (2000). Memory--a century of consolidation. Science (New York, NY
287(5451): 248-51.

144

McNaughton, N. and J. Wickens (2003). Hebb, pandemonium and catastrophic
hypermnesia: the hippocampus as a suppressor of inappropriate associations.

Cortex; a journal devoted to the stuafv of the nervous system and behavior 39(4-
5): 1139-63.

Messenger, J. B. (1971). Two-stage recovery of a response in Sepia. Nature 232(5307):
202-3.

Meunier, M., J. Bachevalier, M. Mishkin and E. A. Murray (1993). Effects on visual
recognition of combined and separate ablations of the entorhinal and perirhinal
cortex in rhesus monkeys. J Neurosci 13(12): 5418-32.

Miyashita, Y., M. Morita, Y. Naya, M. Yoshida and H. Tomita (1998). Backward signal
from medial temporal lobe in neural circuit reorganization of primate

inferotemporal cortex. Comptes rendus de l'Academie des sciences 321(2-3): 185-
92.

Montaldi, D., T. J. Spencer, N. Roberts and A. R. Mayes (2006). The neural system that
mediates familiarity memory. Hippocampus 16(5): 504-20.

Moscovitch, M., L. Nadel, G. Winocur, A. Gilboa and R. S. Rosenbaum (2006). The
cognitive neuroscience of remote episodic, semantic and spatial memory. Current
opinion in neurobiology 16(2): 179-90.

Moser, M. B. and E. 1. Moser (1998). Functional differentiation in the hippocampus.
Hippocampus 8(6): 608-19.

Mulligan, N. W. and J. P. Lozito (2006). An asymmetry between memory encoding and
retrieval. Revelation, generation, and transfer-appropriate processing. Psychol Sci
17(1): 7-11.

Nadel, L. and C. Land (2000). Memory traces revisited. Nature reviews 1(3): 209-12.

Nadel, L., A. Samsonovich, L. Ryan and M. Moscovitch (2000). Multiple trace theory of
human memory: computational, neuroimaging, and neuropsychological results.
Hippocampus 10(4): 352-68.

Naveh-Benjamin, M., A. Kilb and T. Fisher (2006). Concurrent task effects on memory

encoding and retrieval: further support for an asymmetry. Memory & cognition
34(1): 90-101.

Naya, Y., M. Yoshida and Y. Miyashita (2003a). Forward processing of long-term
associative memory in monkey inferotemporal cortex. J Neurosci 23(7): 2861-71.

145

Naya, Y., M. Yoshida, M. Takeda, R. Fujimichi and Y. Miyashita (2003b). Delay-period
activities in two subdivisions of monkey inferotemporal cortex during pair
association memory task. The European journal of neuroscience 18(10): 2915-8.

Nobre, A. C., G. N. Sebestyen, D. R. Gitelman, M. M. Mesulam, R. S. Frackowiak and
C. D. Frith (1997). Functional localization of the system for visuospatial attention
using positron emission tomography. Brain 120 ( Pt 3): 515-33.

Ogawa, S., T. M. Lee, A. R. Kay and D. W. Tank (1990). Brain magnetic resonance
imaging with contrast dependent on blood oxygenation. Proceedings of the
National Academy of Sciences of the United States of America 87(24): 9868-72.

Oldﬁeld, R. C. (1971). The assessment and analysis of handedness: the Edinburgh
inventory. Neuropsychologia 9(1): 97-1 13.

Orban, G. A. and R. Vogels (1998). The neuronal machinery involved in successive
orientation discrimination. Progress in neurobiology 55(2): 117-47.

Pammer, K, R. Lavis and P. Comelissen (2004). Visual encoding mechanisms and their
relationship to text presentation preference. Dyslexia (Chichester, England) 10(2):
77-94.

Polat, U. and A. M. Norcia (1998). Elongated physiological summation pools in the
human visual cortex. Vision research 38(23): 3735-41.

Poldrack, R. A. (2000). Imaging brain plasticity: conceptual and methodological issues--a
theoretical review. Neuroimage 12( l ): 1-13.

Preston, A. R. and J. D. Gabrieli (2002). Different ﬁanctions for different medial temporal
lobe structures? Learning & memory (Cold Spring Harbor, N. Y 9(5): 215-7.

Qiu, F. T. and R. von der Heydt (2005). Figure and ground in the visual cortex: v2
combines stereoscopic cues with gestalt rules. Neuron 47(1): 155-66.

Raichle, M. E. (1998). Behind the scenes of functional brain imaging: a historical and
physiological perspective. Proceedings of the National Academy of Sciences of
the United States of America 95(3): 765-72.

Ranganath, C., M. K. Johnson and M. D'Esposito (2003). Prefrontal activity associated
with working memory and episodic long-term memory. Neuropsychologia 41(3):
378-89.

146

Rao, R. P. (2005). Bayesian inference and attentional modulation in the visual cortex.
Neuroreport 16(16): 1843—8.

Robb, W. G. and M. D. Rugg (2002). Electrophysiological dissociation of retrieval
orientation and retrieval effort. Psychonomic bulletin & review 9(3): 583-9.

Rolls, E. T. (2000). Memory systems in the brain. Annual review of psychology 51: 599-
630.

Rosen, B. R., R. L. Buckner and A. M. Dale (1998). Event-related ﬁJnctional MRI: past,
present, and future. Proceedings of the National Academy of Sciences of the
United States of America 95(3): 773-80.

Ross, R. S. and H. Eichenbaum (2006). Dynamics of hippocampal and cortical activation
during consolidation of a nonspatial memory. J Neurosci 26(18): 4852-9.

Rossion, B., C. Schiltz and M. Crommelinck (2003). The ﬁmctionally deﬁned right
occipital and fusiform "face areas" discriminate novel from visually familiar
faces. Neuroimage 19(3): 877-83.

Rubin, D. C., S. Hinton and A. Wenzel (1999). The precise time course of retention. J
Exp Psychol Learn Mem Cogn 25(5): 1161-1176.

Rudy, J. W. and P. Matus-Amat (2005). The ventral hippocampus supports a memory
representation of context and contextual fear conditioning: implications for a
unitary ﬁinction of the hippocampus. Behav Neurosci 119(1): 154-63.

Rudy, J. W. and P. Morledge (1994). Ontogeny of contextual fear conditioning in rats:
implications for consolidation, infantile amnesia, and hippocampal system
function. Behav Neurosci 108: 227-234.

Rugg, M. D. and E. L. Wilding (2000). Retrieval processing and episodic memory.
Trends Cogn Sci 4(3): 108-115.

Saad, Z. 8., K. M. Ropella, E. A. DeYoe and P. A. Bandettini (2003). The spatial extent
of the BOLD response. Neuroimage 19(1): 132-44.

Sakai, K and Y. Miyashita (1991). Neural organization for the long-term memory of
paired associates. Nature 354(6349): 152-5.

Sahnond, C. H., J. Ashburner, F. Vargha-Khadem, A. Connelly, D. G. Gadian and K J.
Friston (2002). Distributional assumptions in voxel-based morphometry.
Neuroimage 17(2): 1027-30.

147

Sanders, G. D. and J. J. Barlow (1971). Variations in retention performance during long
term memory formation. Nature 232(5307): 203-4.

Save, E., L. Nerad and B. Poucet (2000). Contribution of multiple sensory information to
place ﬁeld stability in hippocampal place cells. Hippocampus 10(1): 64-76.

Save, E. and B. Poucet (2000a). Hippocampal-parietal cortical interactions in spatial
cognition. Hippocampus 10(4): 491-9.

Save, E. and B. Poucet (2000b). Involvement of the hippocampus and associative parietal
cortex in the use of proximal and distal landmarks for navigation. Behavioural
brain research 109(2): 195-206.

Savoy, R. L. (2005). Experimental design in brain activation MRI: cautionary tales. Brain
research bulletin 67(5): 361-7.

Schacter, D. L., R. L. Buckner, W. Koutstaal, A. M. Dale and B. R. Rosen (1997). Late
onset of anterior prefrontal activity during true and false recognition: an event-
related fMRI study. Neuroimage 6(4): 259-69.

Schacter, D. L. and E. Tulving (1994). What are the memory systems of 1994? Memory
Systems 1994. D. L. Schacter and E. Tulving. Cambridge, MIT Press: 1-38.

Scouten, A., X. Papademetris and R. T. Constable (2006). Spatial resolution, signal-to-
noise ratio, and smoothing in multi-subject ﬁinctional MRI studies. Neuroimage
30(3): 787-93.

Segev, I. (2006). What do dendrites and their synapses tell the neuron? Journal of
neurophysiology 95(3): 1295-7.

Shannon, C. E. (1948). A mathematical theory of communication. Bell Sys Tech J 27:
379-423, 623-656.

Shepherd, G. M. (1996). The dendritic spine: a multifunctional integrative unit. Journal
of neurophysiology 75(6): 2197-210.

Shepherd, G. M. and R. K Brayton (1987). Logic operations are properties of computer-
simulated interactions between excitable dendritic spines. Neuroscience 21(1):
151-65.

Siapas, A. G. and M. A. Wilson (1998). Coordinated interactions between hippocampal
ripples and cortical spindles during slow-wave sleep. Neuron 21(5): 1123-8.

148

Squire, L. R. and P. Alvarez (1995). Retrograde amnesia and memory consolidation: a
neurobiological perspective. Current opinion in neurobiology 5(2): 169-77.

Stenberg, G., M. Johansson and I. Rosen (2006). Conceptual and perceptual memory:
retrieval orientations reﬂected in event-related potentials. Acta psychologica
122(2): 174-205.

Stickgold, R., L. Scott, C. Rittenhouse and J. A. Hobson (1999). Sleep-induced changes
in associative memory. Journal of cognitive neuroscience 11(2): 182-93.

Summerﬁeld, C., M. Greene, T. Wager, T. Egner, J. Hirsch and J. Mangels (2006).
Neocortical connectivity during episodic memory formation. PLoS biology 4(5):
6128.

Talairach, J. and P. Tournoux (1988). Co-planar Stereotaxic Atlas of the Human Brain.
New York, Thieme.

Tanaka, S., M. Honda and N. Sadato (2005). Modality-speciﬁc cognitive function of
medial and lateral human Brodmann area 6. J Neurosci 25(2): 496-501.

Tassoni, G., C. A. Lorenzini, E. Baldi, B. Sacchetti and C. Bucherelli (1999). A peculiar
pattern of temporal involvement of rat perirhinal cortex in memory processing.
Behav Neurosci 113(6): 1161-9.

Treves, A. and E. T. Rolls (1994). Computational analysis of the role of the hippocampus
in memory. Hippocampus 4(3): 374-91.

Tse, D., R. F. Langston, M. Kakeyama, 1. Bethus, P. A. Spooner, E. R. Wood, M. P.
Witter and R. G. Morris (2007). Schemas and memory consolidation. Science
(New York, N. Y 316(5821): 76-82.

Tulving, E. (1983). Elements of Episodic Memory. Oxford, Clarendon Press.

Uncapher, M. R. and M. D. Rugg (2005a). Effects of divided attention on MRI
correlates of memory encoding. Journal of cognitive neuroscience 17(12): 1923-
35.

Uncapher, M. R. and M. D. Rugg (2005b). Encoding and the durability of episodic
memory: a functional magnetic resonance imaging study. J Neurosci 25(31):
7260-7.

149

Ungerleider, L. G. and M. Mishkin (1982). Two cortical visual systems. Analysis of
Visual Behavior. D. J. Ingle, M. A. Goodale and R. J. W. Mansﬁeld. Cabridge,
MA, MIT Press: 549-586.

Vaidya, C. J., M. Zhao, J. E. Desmond and J. D. Gabrieli (2002). Evidence for cortical
encoding speciﬁcity in episodic memory: memory-induced re-activation of
picture processing areas. Neuropsychologia 40(12): 2136-43.

Verzi, D. W. (2004). Modeling activity-dependent synapse restructuring. Bulletin of
mathematical biology 66(4): 745-62.

Verzi, D. W., M. B. Rheuben and S. M. Baer (2005). Impact of time-dependent changes
in spine density and spine shape on the input-output properties of a dendritic
branch: a computational study. Journal of neurophysiology 93(4): 2073-89.

Walker, M. P. (2004). Issues surrounding sleep-dependent memory consolidation and
plasticity. Cell Mol Life Sci 61(24): 3009-15.

Walker, M. P. and R. Stickgold (2004). Sleep-dependent learning and memory
consolidation. Neuron 44( 1): 121-33.

Wang, H., Y. Hu and J. Z. Tsien (2006). Molecular and systems mechanisms of memory
consolidation and storage. Progress in neurobiology 79(3): 123-35.

Ward, B. D. (1998). 3dDeconvolve Reference Manual. At_1as of Fugctional Neuroimaging
(AFNI) Suite. NIMI-I.

Weis, S., P. Klaver, J. Reul, C. E. Elger and G. Fernandez (2004). Neural correlates of
successful declarative memory formation and retrieval: the anatomical overlap.

Cortex; a journal devoted to the stuay of the nervous system and behavior 40(1):
200-2.

Weldon, M. S., H. L. Roediger, 3rd and B. H. Challis (1989). The properties of retrieval
cues constrain the picture superiority effect. Memory & cognition 17(1): 95-105.

Wheeler, M. E. and R. L. Buckner (2003). Functional dissociation among components of
remembering: control, perceived oldness, and content. J Neurosci 23(9): 3869-80.

Wichmann, W. and W. Muller-Forell (2004). Anatomy of the visual system. European
journal of radiology 49(1): 8-30.

Wickelgren, W. A. (1974). Single-trace fragility theory of memory dynamics. Memory &
cognition 2: 775—780.

150

Winters, B. D. and T. J. Bussey (2005a). Glutamate receptors in perirhinal cortex mediate
encoding, retrieval, and consolidation of object recognition memory. J Neurosci
25(17): 4243-51.

Winters, B. D. and T. J. Bussey (2005b). Removal of cholinergic input to perirhinal
cortex disrupts object recognition but not spatial working memory in the rat. The
European journal of neuroscience 21(8): 2263-70.

Winters, B. D. and T. J. Bussey (2005c). Transient inactivation of perirhinal cortex
disrupts encoding, retrieval, and consolidation of object recognition memory. J
Neurosci 25(1): 52-61.

Winters, B. D., S. E. Forwood, R. A. Cowell, L. M. Saksida and T. J. Bussey (2004).
Double dissociation between the effects of peri-postrhinal cortex and
hippocampal lesions on tests of object recognition and spatial memory:
heterogeneity of ﬁrnction within the temporal lobe. J Neurosci 24(26): 5901-8.

Witter, M. P. and E. I. Moser (2006). Spatial representation and the architecture of the
entorhinal cortex. Trends in neurosciences 29(12): 671—8.

Wixted, J. T. (2004). The psychology and neuroscience of forgetting. Ann Rev Psychol
55: 235-269.

Woodruff, C. C., M. R. Uncapher and M. D. Rugg (2006). Neural correlates of
differential retrieval orientation: Sustained and item-related components.
Neuropsychologia 44(14): 3000-10.

Woodruff, M. L. and H. M. Kantor (1983). Fomix lesions, plasma ACTH levels, and
shuttle box avoidance in rats. Behav Neurosci 97(6): 897-907.

Yamaguchi, S., S. Yamagata and S. Kobayashi (2000). Cerebral asymmetry of the "top-
down" allocation of attention to global and local features. J Neurosci 20(9):
RC72.

Yovel, G. and N. Kanwisher (2005). The neural basis of the behavioral face-inversion
effect. Curr Biol 15(24): 2256-62.

Yuste, R. and W. Denk (1995). Dendritic spines as basic ﬁrnctional units of neuronal
integration. Nature 375(6533): 682-4.

Zador, A., C. Koch and T. H. Brown (1990). Biophysical model of a Hebbian synapse.
Proceedings of the National Academy of Sciences of the United States of America
87(17): 6718-22.

151

Zhang, F., S. Endo, L. J. Cleary, A. Eskin and J. H. Byrne (1997). Role of transforming

growth factor-beta in long-term synaptic facilitation in Aplysia. Science (New
York, NY 275(5304): 1318-20.

Zola-Morgan, S., L. R. Squire, D. G. Amaral and W. A. Suzuki (1989). Lesions of

perirhinal and parahippocampal cortex that spare the amygdala and hippocampal
formation produce severe memory impairment. J Neurosci 9( 12): 4355-70.

152

   

Itilllllﬂllliljﬁﬂilﬂlt1111111171