Professional Documents
Culture Documents
Ebook Electrophysiological Recording Techniques 2Nd Edition Robert P Vertes Editor Timothy Allen Editor Online PDF All Chapter
Ebook Electrophysiological Recording Techniques 2Nd Edition Robert P Vertes Editor Timothy Allen Editor Online PDF All Chapter
https://ebookmeta.com/product/california-slavic-studies-volume-
vi-robert-p-hughes-editor/
https://ebookmeta.com/product/practical-recording-techniques-the-
step-by-step-approach-to-professional-audio-recording-7th-
edition-bruce-bartlett/
https://ebookmeta.com/product/encyclopedia-of-modern-optics-2nd-
edition-duncan-g-steel-editor-robert-d-guenther-editor/
https://ebookmeta.com/product/nfpa-30-flammable-and-combustible-
liquids-code-handbook-9th-edition-robert-p-benedetti-editor/
Modern Techniques for Food Authentication 2nd Edition
Da-Wen Sun (Editor)
https://ebookmeta.com/product/modern-techniques-for-food-
authentication-2nd-edition-da-wen-sun-editor/
https://ebookmeta.com/product/social-selling-techniques-to-
influence-buyers-and-changemakers-2nd-edition-timothy-hughes/
https://ebookmeta.com/product/cardiology-procedures-a-clinical-
primer-2nd-edition-robert-hendel-editor/
https://ebookmeta.com/product/lipoproteins-in-diabetes-mellitus-
contemporary-diabetes-2nd-edition-alicia-j-jenkins-editor-peter-
p-toth-editor/
https://ebookmeta.com/product/woody-allen-interviews-revised-and-
updated-2nd-edition-robert-e-kapsis/
Neuromethods 192
Robert P. Vertes
Timothy A. Allen Editors
Electrophysiological
Recording
Techniques
Second Edition
NEUROMETHODS
Series Editor
Wolfgang Walz
University of Saskatchewan
Saskatoon, SK, Canada
Edited by
Robert P. Vertes
Ctr for Complex Systems & Brain Sci, Florida Atlantic University, Boca Raton, FL, USA
Timothy A. Allen
Department of Psychology, Florida International University, Miami, FL, USA
Editors
Robert P. Vertes Timothy A. Allen
Ctr for Complex Systems & Brain Sci Department of Psychology
Florida Atlantic University Florida International University
Boca Raton, FL, USA Miami, FL, USA
This Humana imprint is published by the registered company Springer Science+Business Media, LLC, part of Springer
Nature.
The registered company address is: 1 New York Plaza, New York, NY 10004, U.S.A.
Preface to the Series
Experimental life sciences have two basic foundations: concepts and tools. The Neuro-
methods series focuses on the tools and techniques unique to the investigation of the
nervous system and excitable cells. It will not, however, shortchange the concept side of
things as care has been taken to integrate these tools within the context of the concepts and
questions under investigation. In this way, the series is unique in that it not only collects
protocols but also includes theoretical background information and critiques which led to
the methods and their development. Thus, it gives the reader a better understanding of the
origin of the techniques and their potential future development. The Neuromethods
publishing program strikes a balance between recent and exciting developments like those
concerning new animal models of disease, imaging, in vivo methods, and more established
techniques, including, for example, immunocytochemistry and electrophysiological tech-
nologies. New trainees in neurosciences still need a sound footing in these older methods in
order to apply a critical approach to their results.
Under the guidance of its founders, Alan Boulton and Glen Baker, the Neuromethods
series has been a success since its first volume published through Humana Press in 1985. The
series continues to flourish through many changes over the years. It is now published under
the umbrella of Springer Protocols. While methods involving brain research have changed a
lot since the series started, the publishing environment and technology have changed even
more radically. Neuromethods has the distinct layout and style of the Springer Protocols
program, designed specifically for readability and ease of reference in a laboratory setting.
The careful application of methods is potentially the most important step in the process
of scientific inquiry. In the past, new methodologies led the way in developing new dis-
ciplines in the biological and medical sciences. For example, physiology emerged out of
anatomy in the nineteenth century by harnessing new methods based on the newly discov-
ered phenomenon of electricity. Nowadays, the relationships between disciplines and meth-
ods are more complex. Methods are now widely shared between disciplines and research
areas. New developments in electronic publishing make it possible for scientists that
encounter new methods to quickly find sources of information electronically. The design
of individual volumes and chapters in this series takes this new access technology into
account. Springer Protocols makes it possible to download single protocols separately. In
addition, Springer makes its print-on-demand technology available globally. A print copy
can therefore be acquired quickly and for a competitive price anywhere in the world.
v
Preface
vii
viii Preface
cells) to them. By recoding from “tagged” traced cells (in slices), they could determine the
neurochemical identity and electrophysiological characteristics of neurons projecting to
new-born granule cells. Further, by adjusting the time interval between the first and second
viral injections into starter cells (between 3 and 90 days), they were able to define inputs to
immature versus mature granule cells. Finally, as was pointed out, the method allows for an
examination of the possible effects of various conditions (exercise, learning) or pathological
states (epilepsy) on the organization of afferents to new-born granule cells.
The chapter by Vertes, Linley, and Viena examines the nucleus reuniens (RE) of the
ventral midline thalamus, describing RE circuitry, functional properties, electrophysiological
characteristics, and involvement in CNS disorders. With respect to function, nucleus
reuniens has been shown to serve a critical role in cognitive, affective, and executive
functions. Specifically, alterations of RE significantly disrupt spatial working memory, atten-
tion/attentional set, reversal learning, and fear conditioning/fear extinction. Regarding
electrophysiological characteristics, subsets of RE neurons display spatial properties (head
direction cells, place cells), exhibit trajectory-dependent firing (predicting directional
choices), and discharge at significantly higher rates in waking and/or REM sleep compared
to slow wave sleep. The latter finding suggests that RE neurons may exert arousal or
attentional effects on target structures during waking or REM sleep. Finally, RE has been
linked to the CNS disorders of schizophrenia (SZ) and epilepsy. Disruptions of nucleus
reuniens produce abnormal oscillations between the hippocampus and medial prefrontal
cortex, and associated SZ-like cognitive deficits, whereas the over activation of reuniens
produces pronounced epileptic seizure activity.
In an extension of his chapter in the original volume, Bressler provides a detailed
overview of the neural mechanism responsible for event-related potentials (ERPs). As was
described, ERPs are triggered by external sensory and motor events or by “internal”
spontaneous or self-generated states, and represent dynamical changes in the brain to
internal/external events. Various methods were described for analyzing cortical ERPs, or
extracting their non-varying features from the general field activity. These include time
domain analysis, frequency domain analysis, spatial analysis, and interdependency analy-
sis—with each reflecting distinct aspects of the triggered ERPs. The ERP was compared to
its magnetic correlate, the event-related field (ERF), and as pointed out, for some analyses,
both methods are preferable to imaging techniques, as the time course of changes more
closely mimic the underlying neural events.
The chapter by Pittman-Polletta and Kocsis examines various methods used to assess the
cross-frequency coupling of electrophysiological signals (EEG and LFPs) such as amplitude-
amplitude coupling (AAC) and phase-phase coupling (PPC), but it describes in detail phase-
amplitude coupling (PAC). The PAC involves fluctuations in the amplitude of fast rhythms
dependent on the phase of slower rhythms. The authors provide a detailed description of the
signal processing techniques used to examine and quantify PACs—from the preprocessing
of data through to statistically determining the strength of PAC coupling. As was pointed
out, PAC analysis is a valuable tool for quantifying the relationship between simultaneously
occurring events, reflecting functional interactions between structures (or networks) exhi-
biting oscillations at different frequencies. The authors give examples of the application of
PAC analysis from their work showing interactions between delta/theta and high-frequency
oscillations (HFO) of the cortex and hippocampus as well as between high gamma activity in
the hippocampus (driver) and theta in the cortex (receiver).
The new chapter by Jayachandran and Allen looks at the neurophysiological mechanisms
that encode time and approaches to studying this fundamental aspect of cognition and
Preface ix
intracranial recording approaches in non-human primates (SUA, LFP and iEEG) and
humans (SUA, LFP, and iEEG) to non-invasive approaches used in humans (EEG, MEG,
fMRI, and PET) highlighting that each provides information about particular levels of
analysis, and that the transfer functions between levels are largely unknown. Unlike most
intracranial recordings, iEEGs tend to cover large areas of the brain, and have exceptional
spatial and temporal resolution. The author gets into the details of the physiological basis of
iEEG signal and addresses other approaches including adding arrays for unit recordings. The
author then provides practical methodological and equipment information for using iEEG
in the clinic. The chapter presents analytical approaches such as frequency decomposition
and event-related potentials, and the use of various statistical approaches for drawing
meaningful conclusions. Lastly, Helfrich provides several notes for consideration of issues
that arise in iEEG work in the clinic.
In conclusion, as a continuation of the previous volume, the present edition introduces
and details several recently developed techniques currently used to examine the electrophys-
iological properties of several CNS systems, which affect behavior across a range of species
from rodents to humans. It is our expectation that the present edition would have wide
applicability for basic researchers as well as for clinicians in utilizing some of the principles
and research designs described herein in their research programs.
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
xi
Contributors
xiii
Chapter 1
Abstract
Electroencephalography (EEG), an electrophysiological monitoring method to record the electrical activity
of the brain, has been one of the most important noninvasive brain imaging tools in psychiatry, but
surprisingly little is known about how the neural correlates of various EEG features are linked to cognition.
Recent advances in neuroscience and related technologies make this an ideal time for new discoveries about
the origin and significance of the contents of EEG. In particular, understanding the molecular and cellular
mechanisms underlying diverse EEG features has been facilitated using mouse models under genetic,
pharmacological, or optogenetic manipulations. A core challenge in mouse EEG was to obtain topographi-
cal neuroimaging that can be compared to human EEG. To overcome this challenge, we have developed a
high-density EEG using a polyimide-based microarray that fits to the mouse skull and can be applied to
various studies with a high spatiotemporal accuracy in free behavioral states. The benefits of mouse high-
density EEG are not only that it provides cross-species neuroimaging data comparable to human EEG but
also that it helps in dissecting enigmatic brain activity by probing the neural substrates of cognition when
combined with optogenetics. The aim of this chapter is to introduce the methodological aspects of high-
density EEG in mice. We explain the electrodes, surgery, recording, and analysis procedures and present
applications in studying the origin of EEG signals. In addition, we point to potential areas where this
technique will provide mechanical insight into circuit dysfunction in major psychiatric conditions.
1 Introduction
Robert P. Vertes and Timothy A. Allen (eds.), Electrophysiological Recording Techniques, Neuromethods, vol. 192,
https://doi.org/10.1007/978-1-0716-2631-3_1, © Springer Science+Business Media, LLC, part of Springer Nature 2022
1
2 Jee Hyun Choi and Eunjin Hwang
2 Materials
2.1 Technical Acquiring high-density EEG from mice in freely moving states
Challenges in EEG poses a significant technical challenge since the small size of the
Electrodes for Freely mouse brain (~0.5 cm3) limits the space for implantation of many
Moving Mice electrodes and the agile movements of the animal produce signifi-
cant movement artifacts. In the last decades, tremendous advances
in chronic neural recordings have been achieved, particularly in the
fields of implantable silicon probe arrays [9], accelerated by devel-
opment of wireless recording systems. Expanding these technolo-
gies to acquire high-density EEG can take several technical factors
into account but is not straightforward due to the following dis-
tinct features of high-density EEG. First, high-density EEG
requests to collect information from a broad region of cortex.
This implies that the electrodes should be distributed over the
surface of a semi-infinite medium, and the interface condition for
the contact should be equivalent for all EEG electrodes. The most
appropriate surface to apply the electrodes in mice is the skull
surface. This involves building a flexible, dry electrode that can fit
onto the curved skull surface. Second, EEG activities often appear
as oscillations whose frequencies are arranged from low (~1 Hz) to
high (~150 Hz). The serious issue in this broadband signal is the
abundance in the nonneural signal sources – some from physiologi-
cal origins such as heartbeat, respiration, vasomotion, etc., and
some from environmental noise, such as ventilation, acoustic reso-
nance of buildings, line noise, etc. Moreover, 1/f noise, which is
inversely proportional to the frequency, is common in biological
systems affecting the low-frequency regime of EEG. Therefore,
consistent contact/attachment to the skull with high mechanical
stability and effective electrical shielding is needed.
4 Jee Hyun Choi and Eunjin Hwang
2.2 Polyimide-Based To resolve the above issues, we applied a polyimide as the carrier
Microarray for Mouse substrate and fabricated a microarray as depicted in Fig. 1
High-Density EEG [7, 8]. Polyimide has been shown to have superior biocompatibility
and good mechanical properties [10] and has been widely used as a
substrate for flexible neural probes [11]. As the skull surface is
ellipsoidal, the overall structure has a backbone line with multiple
branches on both sides. The thickness is crucial for stable adher-
ence. There is a trade-off in the thickness of microarray: in the
thinner, the yielding rate of production drops, resulting in more
cost per array, whereas in the thicker, it is more easily detached from
the skull. We tested different thicknesses and found that the most
appropriate thickness was 7–8 μm with an acceptable level of up to
10 μm. Figure 1b shows different electrode configurations for
40 and 64 channels. In most cases, the 40-channel configuration
is enough to generate topographical mapping. The 64-channel
configuration is appropriate for somatosensory research, presenting
14 channels on top of the somatosensory cortex. For the electrical
contacts and interconnection lines, we used Pt [7], but later studies
showed that an Ag/AgCl electrode presents similar levels of signal
quality. A series of tests revealed that the most relevant features in
the electrode configuration are the size of electrical contact, the
width of the interconnection line, and the interline spacing. We
found that the adequate impedance for EEG measurements was
obtained with an electrode contact size of ~0.2 mm2. The width of
interconnection lines and the interline spacing are key factors in the
design because they determine the overall size. We found that there
is no significant generation of induced noise or cross-talk for widths
>15 μm and interline spacing >20 μm. With the best configura-
tion, the impedance ranges from 10 to 300 kΩ at 30 Hz when
measured on the skull.
Fig. 1 Polyimide-based microarray for mouse high-density EEG. (a) Picture of a 40-channel microarray. (b)
Configuration of 40-channel (left) and 60-channel (right) microarrays. Inset images are actual microarrays
positioned on a mouse skull. (c) Electrode locations overlaid on a cortical region map
High-Density Electroencephalography in Freely Moving Mice 5
3 Methods
3.1 Fixation of For a favorable surgery outcome, we suggest using bone cement as
Microarray a fixation material. Different types of commercially available bone
cements are applicable, but an ideal cementing material should be
noncorrosive, tightly filling, and not affected by moisture and
should have long-term stability. A typical cause of failure is the
detachment of the cement from the skull. A couple of anchoring
microscrews on the skull helps in preventing this detachment. Also,
when one applies cement, any debris on the skull surface should be
cleaned. We use a tap-water-soaked cotton swab for the cleaning,
which is used for pressing the microarray as well. We make the
surface wet with tap water when we place and press the microarray
and wait until the surface is fully dry. Placing cement is extremely
important for preventing its detachment from the skull. We suggest
using low-viscosity cement and applying the cement with a brush.
One should check the cement is fully dried before placing the next
cement.
3.2 Recording The measurement principle and instrumentation for mouse EEG
are identical to those for human EEG. The basic principle of EEG
recording is to measure the potential difference between two points
as the direct consequence of the existing electric dipole created by a
synchronously generated postsynaptic potential in the cortex.
These potentials can be measured by a voltage amplifier. Typically,
the microarray electrode has sub-Ω-level resistance, but on the
skull, the impedance level is approximately 200 kΩ when measured
at 30 Hz. Determination of the interface impedance is not trivial
because the interface between the electrode and the skull constitu-
tes a significant resistance and capacitance. In human scalp EEG,
the interface impedance is at the level of ~50 kΩ, and in general, the
difference of impedance between channels does not affect the read-
ings. However, when variation of the interface impedances between
channels is high, we recommend normalizing the signal levels.
Typically, we perform normalization of EEG by dividing the spec-
tral power above 200 Hz in all channels using EEG acquired in the
quiescent moment of mouse. For comparisons between different
mice, normalization of EEG is required. Additionally, an electrode
with an impedance level above 500 kΩ should be excluded from
further analysis.
3.3 Artifacts and Recording extracranial EEG from freely moving mice is subject to
Noises more environmental noise than is present in other electrophysio-
logical modalities. The environmental noise may arise from a variety
of sources, such as the mouse’s own body, instrumentation, or the
experimenter. The waveforms of noises depend on their noise types,
and unfortunately, many noises have waveforms similar to the EEG
6 Jee Hyun Choi and Eunjin Hwang
3.4 Analyzing EEG The surface potential produced by a mouse brain has an order of a
few μV, which is a comparable level to the human brain. Basically,
EEG signals are nonstationary data and the operations are done in a
time domain or frequency domain. In the case of averaging multi-
ple events in the time domain, an ergodic property of the signals is
assumed, i.e., all events are equivalent, which may be violated in the
real world. Hence, prior to the data analysis, several experimental
considerations are critical to obtain consistent signals across trials.
For example, cognitively or behaviorally equivalent trials, precise
timing, and uncorrelated noise across different trials are all impor-
tant conditions for a good EEG. In the case of a study of spontane-
ous activities, spectral power or phase-amplitude coupling
properties are analyzed in the frequency domain. In either case,
preprocessing is critical to uncorrelate the noise from the signal.
3.4.1 Preprocessing One of the first steps in EEG analysis is to identify and remove
potentially contaminated signals or artifacts. Mice produce signifi-
cant levels of artifacts, especially in a freely mobile condition, and in
many situations, these artifacts are not stereotypical and are con-
fused with real brain activity, such as epileptic activity. During the
experiment, an accelerometer recording the movement [12] or
videotaping helps in distinguishing the movement artifacts from
the signals. In our experience, laboratory mice, even the wild type,
produce epileptic events more often than humans. In most cases,
the epileptic events occur during a quiescent state and do not occur
in all channels like movement artifacts do, and this characteristic
High-Density Electroencephalography in Freely Moving Mice 7
3.4.2 EEG Topography One of the advantages of high-density EEG in mice is the ability to
investigate the spatiotemporal change of cortical activation with
topographical analysis. EEG topography is a neuroimaging tech-
nique for mapping the spatial distributions of cortical activity.
Basically, EEG topography is a contour plot on the top surface of
the mouse brain. First, we create a mesh (set of imaginary points) to
represent a cortical surface. Typically, the stereotaxic resolution is
used for the mesh size so that any EEG channel can be correspon-
dent to a mesh. Second, we limit the surface to an ellipsoidal
boundary. Third, the potential value of the mesh is calculated
from an interpolation method. According to our tests, different
types of interpolation do not affect the results to a noticeable
degree. We normally use cubic spline interpolation. Figure 2
shows a cine-mode presentation of EEG topographies, while the
ventral posterior medial nucleus of the thalamus (VPM) of a Thy1-
ChR2-EYFP mouse was optogenetically stimulated. The outer sur-
face of the mouse neocortex was rendered by a spm_surf function in
the SPM8 open-source toolbox (Wellcome Trust Centre for Neu-
roimaging UCL, London, UK) based on the mouse MRI down-
loaded from the open database of the magnetic resonance
microimaging neurological atlas group [15].
8 Jee Hyun Choi and Eunjin Hwang
Fig. 2 Topographical mapping of cortical signal propagation in response to optogenetic stimulation of VPM. (a)
Schematics of the signal propagation pathway (upper) and site of stimulation (lower). (b) Cine-mode
presentation of EEG topography
4 Note on Applications
4.1 Simultaneous EEG is a widely used method to monitor human brain activities due
Recordings of High- to its noninvasiveness and high temporal resolution. However,
Density EEG and since human EEG measures the voltage changes from the scalp,
Subcortical Brain an investigation of subcortical origins or their interplay with corti-
Activities cal activity is highly limited in human EEG. In this section, we
introduce a method to combine high-density EEG with a conven-
tional local field potential (LFP) to simultaneously monitor brain
activities in the subcortical region and cortex in mice. By combin-
ing EEG and LFP, we are able to investigate the functional connec-
tivity between the subcortical region and cortex. Here, we show an
example of combined EEG and LFP in a study of spike-and-wave
discharge (SWD) in the thalamus and cortex, simultaneously. The
SWDs are known to be the EEG hallmark of absence seizure [16]
and are often associated with loss of consciousness [17]. We moni-
tored the thalamic activation with cortical activation and the func-
tional connectivity between the two regions when the mouse
experienced absence seizure. The origin of such connectivity is
still under investigation, but an emergent phenomenon of syn-
chrony between the thalamus and cortex has been observed previ-
ously. The synchrony analysis of LFP and EEG quantitatively
assesses the functional connectivity of the thalamus and cortex,
and a calculation of time lag of thalamic and cortical SWDs leads
us to infer the causation of the signal.
High-Density Electroencephalography in Freely Moving Mice 9
4.1.1 Methodological Most of the procedures of implementing LFP together with EEG
Considerations of are the same as implementation of EEG only. The only difference is
Simultaneous Recording of that making a hole for insertion of the LFP electrode and the
High-Density EEG and LFP fixation of the LFP electrode takes place after the EEG microarray
has been attached on the skull. In the case where the subcortical
region of interest is along the midline, where the interconnection
lines of EEG channels are located, the hole can be moved laterally
by x in the same coronal plane, and the depth of electrode insertion,
d, is calculated by the Pythagorean theorem (x2 + z2 ¼ d2), in which
z is the dorsoventral coordinate. The angle of the electrode inser-
tion equals the tangent of z/x. The position of the LFP electrode
should be confirmed by brain histology after all experiments are
conducted. Typically, it is recommended to match the impedances
of the reference and active electrode at similar levels [18], but
sharing the same reference and ground electrodes for the LFP
electrode delivers similar signal quality. Moreover, it is important
to reduce the number of electrodes for the well-being of the
animals in long-term chronic recordings.
4.1.2 Comparison of EEG One of the most interesting questions to explore about seizures is
and LFP Signals During the seizure initiation factors. Here, we exemplify the application of
SWD high-density EEG to pharmacologically induced absence seizure.
To produce SWDs in the mouse, we injected gamma-butyrolactone
(GBL, 100 mg/kg, i.p.) in awake mice and monitored them during
the freely moving state. A previous study of simultaneous record-
ings of intracellular potentials and EEG showed that the spike of
SWD in EEG corresponds to synchronous burst firings activated by
low-voltage activated channels [19], which consequently recruit
the neurons to activate synchronously. By comparing the time
lags between spike peaks, the initiation or propagation of seizure
can be studied. Figure 3a shows an exemplary EEG and LFP during
SWD. In each channel, the first positive spikes were identified and
marked with inverted triangles. The polarity of the depth electrode
is the opposite to that of the surface electrode; thus, it was inverted
in the plot. In the particular event, the first spike was monitored in
the frontal cortex including the prefrontal and motor cortices. The
level of seizure was calculated by total power in the period of SWD
(Fig. 3c), and the functional connectivity between different cortical
regions (Fig. 3d) was represented as thickness of the lines. In this
case, the functional connectivity was calculated by cross-correlation
coefficient, which assesses the similarity of two signals. Other vari-
ables, such as the phase synchronization index, could also be used
to gauge the synchronization level of the two regions [20]. Most of
the connections were between the two sites located symmetrically
with respect to the midline. Likewise, the functional connectivity
between the thalamus and cortex (Fig. 3e) was calculated by the
cross-correlation coefficient and mapped as a color map. The tem-
poral relationship between spikes was assessed with peak latency
and mapped for the first and second peaks, respectively (Fig. 3f, g).
10 Jee Hyun Choi and Eunjin Hwang
Fig. 3 Simultaneous recording of EEGs and LFPs during GBL-induced SWD. (a) Exemplary traces of SWD. (b)
Cortical high-density EEG montage and relative location of LFP electrode in VB (green dot). (c) Power map of
SWD. Total power from 1 Hz to 60 Hz was used for topographical representation. (d) Cortico-cortical functional
connectivity during SWD. Functional connectivity was measured by computing the cross-correlation between
channels. Connections stronger than 0.95 were shown as links between channels. (e) Functional connectivity
between the cortex and thalamus. (f) Latency map of first spikes with respect to the first spike of the FP1
channel. (g) Latency map of second spikes with respect to the second spike of the FP1 channel
References
1. Cohen MX (2017) Where does EEG come 3. Buzsaki G, Wang XJ (2012) Mechanisms of
from and what does it mean? Trends Neurosci gamma oscillations. Annu Rev Neurosci 35:
40(4):208–218. https://doi.org/10.1016/j. 203–225. https://doi.org/10.1146/annurev-
tins.2017.02.004 neuro-062111-150444
2. Bragin A, Jando G, Nadasdy Z, Hetke J, 4. O’Keefe J, Recce ML (1993) Phase relation-
Wise K, Buzsaki G (1995) Gamma (40-100 ship between hippocampal place units and the
Hz) oscillation in the hippocampus of the EEG theta rhythm. Hippocampus 3(3):
behaving rat. J Neurosci 15(1 Pt 1):47–60
12 Jee Hyun Choi and Eunjin Hwang
Abstract
High-density multichannel recordings are a powerful technique to investigate the information represented
in the activity of neuronal ensembles. Recent technological advancement has achieved the development of
small and low-weight recording devices, allowing for the use of hundreds of channels to monitor the neural
activity from freely behaving animals. Because of the substantial increase in the channel number of
recording devices, it is more common to record the neural activity from multiple brain regions simulta-
neously, allowing for researchers to investigate functional dependency and interactions5 between regions.
However, multiregional recordings often require different techniques from the ones used for single-region
studies. This chapter will review the methodological details of simultaneous recordings from multiple brain
regions as well as analysis techniques used for such datasets.
Key words Spatial navigation, Multiregional interactions, Ensemble analysis, Freely behaving record-
ings, Tetrodes
1 Introduction
Robert P. Vertes and Timothy A. Allen (eds.), Electrophysiological Recording Techniques, Neuromethods, vol. 192,
https://doi.org/10.1007/978-1-0716-2631-3_2, © Springer Science+Business Media, LLC, part of Springer Nature 2022
15
16 Hiroshi T. Ito
2.1 Tetrode and Each channel of recording electrodes picks up voltage signals
Microdrive derived from hundreds of neurons nearby. Voltage signals gener-
ated by neurons are usually associated with either spikes or synaptic
potentials. A high-pass filter is used to isolate signals derived from
spikes, because spike waveforms are mostly confined in a high-
frequency spectral range compared to those of synaptic potentials.
A typical range of passbands used for spike detection is between
300 and 6000 Hz. To investigate the activity of individual neurons,
it also requires to classify and assign detected spikes to individual
neurons. While spike waveforms are often used as a criterion for
spike classification, the waveforms of the same neurons are not
necessarily always the same, for example, during spike bursting,
leading to a possibility of misclassification. Here, a high-density
channel configuration of electrodes provides another criterion for
spike classification [21, 22]. When the electrode has multiple chan-
nels in small space, signals from the same neuron can be detected
across multiple channels that are placed at different positions
Multisite Recording for the Analysis of Information Flow Between. . . 17
and hanged over a magnetic stirrer using a magnetic bar to hold the
bottom twist of the wire (Fig. 1b). After rotating the magnetic bar
for a few minutes, multiple twists of four wires will be formed. This
tetrode is then heated up to 210 C with a heat gun for 2 min for
sealing.
A microdrive of tetrodes can be obtained commercially, for
example, from Neuralynx or Axona, but it is also possible to con-
struct with your own design using a 3D printer. A Harlan 28 drive
from Neuralynx, for example, can hold 28 independently movable
tetrodes (Fig. 1c), which is a good size for the recordings from the
hippocampus and the nucleus reuniens simultaneously in behaving
rats. The microdrive will be fixed on the animal’s skull using several
anchor screws (e.g., M1.6, 3 mm) during the surgery. The ground
screws of the drive are typically placed on the skull above the
cerebellum for recordings from the hippocampus, the thalamus,
or other neocortical regions, as the cerebellar signals are presum-
ably largely uncorrelated with ones in these recording areas. The
use of reference tetrode is useful to reduce noise from muscle
activity (e.g., chewing noise), which helps obtain better spike wave-
forms for classification.
2.3 Recording of the Several companies offer a complete set of recording systems with
Neural Activity software integrated with video tracking of the animal’s positions
Together with the and head directions, which is necessary for the investigation of
Animal’s Behaviors neural correlates to behaviors. If one wants to use an open-source
system (e.g., Open Ephys), it is necessary to combine the recording
system with a position tracking software to monitor the animal’s
Multisite Recording for the Analysis of Information Flow Between. . . 19
3.1 Spectral Spectral coherence analysis is a widely used technique to assess the
Coherence Analysis degree of synchrony of oscillatory activity between regions. This
technique is mathematically related to spectral power analysis; while
the spectral power analysis is based on an autocovariance function
of local field potentials (LFPs) from a single brain region, the
spectral coherence analysis is on a cross-covariance function of
LFPs between regions.
Spectral power density:
X
τ¼1
S xx ðωÞ ¼ γ xx ðτÞe 2πiωτ
τ¼1
Spectral coherence:
S xy ðωÞ2
C xy ðωÞ ¼
S xx ðωÞS yy ðωÞ
Therefore, spectral coherence is considered “normalized”
cross-spectrum. The calculation of spectral coherence does not
require spike data, and thus it can be performed if at least one
channel is located in individual brain regions.
While spectral coherence analysis provides the degree of coher-
ence across different frequencies, it does not give time-dependent
changes. When one wants to investigate behavior-dependent
changes of coherence, for example, during a goal-directed naviga-
tion task, a time-resolved strategy should be applied by using a
sliding time window. There are several choices of windows (e.g.,
Hanning or Hamming), which reduces the edge effects due to the
handling of short-duration datasets. It is also known that the multi-
taper method helps reduce estimation bias [23–25], and a useful
MATLAB toolbox, Chronux, is available for its
implementation [25].
In Fig. 2a, the spectral coherence analysis was applied when the
animal performed a continuous alternation task, in which the ani-
mal was required to choose either right or left turn at the
T-junction of the maze alternately to obtain a reward. The analysis
reveals a peak of coherence at the theta range of frequency
(6–12 Hz) between LFPs in CA1 and the nucleus reuniens (RE).
The RE has mutual anatomical connections with the prefrontal
cortex and also gives rise to inputs in the area CA1 and the sub-
iculum regions of the hippocampus [19, 26–28]. The RE has been
considered a key anatomical hub for the communication between
the prefrontal cortex and the hippocampus because virtually no
direct projection exists from the prefrontal cortex to the hippocam-
pus [27]. Therefore, synchrony at the theta rhythm implies func-
tional coupling between RE and CA1, mediating prefrontal-
hippocampal interactions. To investigate if coherence is modulated
during the task phase, time-resolved spectral coherence analysis was
implemented between RE and CA1 using a sliding time window of
512 ms. The analysis shows a prominent coherence in the theta
frequency range that changes from time to time during the behav-
ioral task (Fig. 2b). The degrees of theta coherence are further
mapped on spatial positions of the maze using a color code,
which revealed a clear spatial bias of theta coherence (Fig. 2c).
The theta coherence between CA1 and RE is higher at the stem
part of the maze in which the animals are required to make a
decision of the next trajectory. The enhanced synchrony between
CA1 and RE implies a change of information flow, suggesting a key
role of their functional coupling during trajectory decisions.
Multisite Recording for the Analysis of Information Flow Between. . . 21
Fig. 2 Spectral coherence analysis between CA1 and RE LFPs. (a) Top: LFPs from the hippocampus CA1 and
the nucleus reuniens (RE). Bottom: the left two plots showing spectral power of LFPs from CA1 and RE. The
right plot shows spectral coherence. Note a peak of coherence at the theta range of frequency (6–12 Hz). (b)
Time-resolved spectral coherence of LFPs from CA1 and RE. The degree of coherence is color-coded. A
prominent coherence peak at the theta rhythm is observed, but it varies over time. (c) The degree of spectral
coherence is plotted on spatial positions of the maze. Left. A rat performed a continuous alternation task in an
8-shaped maze. At the stem part of the maze, the animal was required to choose either right or left trajectory
based on the alternation rule (i.e., choosing the different direction from the previous trial). The right top panel
shows the spectral coherence between CA1 and RE at the theta rhythm on the maze. The right bottom panel
shows the phase coherence estimated by ISPC. In both plots, a prominent increase of theta coherence is
observed at the stem part of the maze
1 X iðθx ðt Þθy ðt ÞÞ
N
ISPC ¼j e j
N t¼1
22 Hiroshi T. Ito
3.2 Spike-Phase While coherence analysis based on local field potentials (LFPs) is a
Analysis Across Brain useful method to assess the degree of synchrony between brain
Regions regions, it is important to note that LFPs are not necessarily local
signals. A study found that LFPs can spread more than a centimeter
range in the brain of nonhuman primates [30]. It is thus difficult to
exclude a possibility that the measured spectral coherence is influ-
enced by other brain regions near the recording electrodes. This
problem can be solved by using the correlation analysis based on
spike activity, because spikes can be detected only within a small
distance range from the electrode (~50 μm) [31]. Therefore, spike-
based analysis is an important addition to confirm synchrony
between regions.
Fig. 3 Spike-field coherence analysis between RE spikes and CA1 LFP. (a) Procedure of spike-field coherence
method. For each spike of a given RE cell, segments are extracted from CA1 LFP, so that the spike times are
aligned at the center of the segments. After averaging, a clear theta rhythm is observed in the real data, which
results in a prominent peak at the theta rhythm in the spectral power (left). By contrast, with simulated random
spikes, the LFP average as well as its spectral power are almost flat. The spectral power of the averaged LFP
Multisite Recording for the Analysis of Information Flow Between. . . 25
Fig. 3 (continued) segment is then normalized, giving a spike-field coherence. A prominent peak of coherence
is observed at the theta range, indicating that spikes of this particular RE cell are phase-locked to the theta
rhythm in CA1 LFP, but not in other frequency ranges (e.g., beta, gamma). (b) Comparison of spike-field
coherence between the stem and the nonstem parts of the maze. The coherence in the theta range is largely
increased at the stem part of the maze, indicating enhanced spike-time modulation by the CA1 theta. (c)
Influence of the spike number used to compute the spike-field coherence. A small number of spikes generally
give a larger coherence value, and therefore, it is important to use the same number of spikes for the group
comparison
26 Hiroshi T. Ito
Fig. 4 Signal directionality analysis between RE spikes and CA1 LFP. (a) Illustrations showing an idea behind
the analysis. If CA1 influences RE spikes, the correlation will become stronger by temporally shifting CA1 LFP
forward, because it will cancel out the axonal conduction delay. The opposite is the case if RE influences CA1.
(b) The plot shows the mean vector lengths for all recorded RE cells relative to CA1 theta at different amounts
of time shifts (thick line, means; shaded, standard errors). The peak was observed by shifting CA1 LFP
backward, indicating the signal directionality from RE to CA1
4.1 Population While the analysis based on ANOVA and ANCOVA can classify
Vector-Based each cell as either trajectory dependent or independent, this analysis
Decoding does not necessarily give the idea of information represented in a
brain area as a neural ensemble. A key advantage of decoding
analysis is the ability to evaluate the predictability of a particular
behavioral parameter from ensemble neural activity, and therefore,
it serves as a powerful strategy to assess its behavioral relevance.
Decoding analysis is usually performed in two steps, training
and testing phases. The dataset should be divided into two parts,
training and test datasets, where the training dataset is used to
optimize the decoding parameters and the test dataset is used
only for the evaluation of the decoder’s performance (Fig. 5b).
This is because the decoding performance should always be evalu-
ated with samples that are not used for parameter optimization, and
especially in a machine-learning algorithm, one needs to be cau-
tious about the decoder’s overfitting to the training dataset, which
will result in poor performance in new datasets even if it has almost
perfect performance on the training dataset.
A relatively simple form of decoding can be performed by
constructing vectors of firing rates of a neuronal ensemble (i.e.,
population vector). First, firing rates of each neuron were normal-
ized to z scores, so that a neuronal bias due to the difference of
baseline firing rates should be reduced. Then the averaged rate
vectors on the stem were calculated separately for either right- or
left-trajectory trials from the training dataset that is constructed by
removing one trial as a test dataset (Fig. 5a). Finally, the difference
between dot products for a test rate vector and the individual
trajectory vectors gives the decoded trajectory as in the following
equation:
T ¼ sign v c right v c left
in which cright and cleft are the population vectors from the training
dataset for right- or left-trajectory trials and v is a vector of firing
rates of the same population of neurons on the test dataset. T is the
output value of the classifier; 1 or 1 represents right or left
Fig. 5 Decoding of the animal’s next trajectory from a population of neurons. (a) Illustration showing a
construction of a population vector of firing rates of neurons on the stem part of the maze. Each neuron shows
different levels of modulation in firing rates between right- and left-oriented trajectories, which is summarized
as two population rate vectors for each of the two trajectories. The decoder works based on the similarity of a
given vector to these two population vectors. (b) Cross-validation to test the decoder’s performance. The
dataset is divided into training and test datasets, and only the training dataset is used to construct the average
PVs for the two trajectories. The decoder’s performance is then tested with a test dataset by taking dot
products of a given vector with the PVs for the two trajectories. (c) Linear classifier in the machine-learning-
based approach. The firing rates of individual neurons are weighted and summated to give an output. The
weight adjustment allows for increasing or decreasing the contribution of particular cells for the best
classification performance. (d) Performance of the decoders predicting the animal’s next trajectory based
on neurons in RE and CA1. The chance level is 50%. The performance was in general better for the SVM
decoder compared with the PV decoder. The high decoding performance of CA1 cells was significantly
impaired by the lesions of RE, indicating the importance of RE for trajectory representation in CA1
30 Hiroshi T. Ito
4.2 Support Vector While a population vector decoder can achieve a good performance
Machine for Decoding when an ensemble of neurons represents one common feature such
of the Next Trajectory as head direction or spatial position, neurons often exhibit mixed
Choice representations of multiple modalities [46–50], and it is thus pref-
erable for a decoder to have a mechanism to adjust the contribution
of individual neurons. This can be achieved by a machine-learning-
based decoding strategy, such as a support vector machine. Unlike a
population vector method in which the decoding efficacy can be
deteriorated by the inclusion of neurons that does not contribute to
the decoding feature, a machine-learning algorithm is able to dis-
regard such irrelevant information.
A support vector machine is one of the supervised learning
algorithms, in which new data are classified based on the class labels
of the training datasets. The performance of decoder should always
be tested with a dataset different from the one used in training (i.e.,
cross-validation), because even a decoder with perfect classification
performance in the training dataset does not necessarily give good
performance in another dataset, due to overfitting of the decoder
that impairs the generalization ability. The support vector machine
is considered an algorithm with excellent generalization ability [51]
and is widely used for machine-learning-based classification
problems.
While the support vector machine can be implemented with
nonlinear function, for simplicity, we use a linear decoder,
expressed as in the following equation, to predict the next trajec-
tory based on the spike rates on the stem (Fig. 5c):
T ¼ b þ w1 F 1 þ w2 F 2 þ w3 F 3 . . . : ¼ b þ w T F
F ¼ ½ F 1 , F 2 , F 3 , . . .T , w ¼ ½w 1 , w 2 , w 3 , . . .T
where F is a vector of firing rates for each cell, w is a vector of
respective weights, b is a scalar offset, and T is an output value of the
classifier, either 1 for right turn or 1 for left turn. The optimal
weights of the decoder were determined by a support vector
machine algorithm to maximize the separation margin for better
generalization performance. Briefly, for a given number N of trials
of rate-trajectory pairs, (Fi, Ti), i ¼ 1,2,3,. . .N, we searched for
w that satisfies the following condition:
1 T X N
min w wþC ξi
w, b, ξ 2
i¼1
Subject to y i b þ w T F i 1 ξi , ξi , 0:
where C is a penalty parameter for misclassification, which we
usually set at 1, but changing the C value may improve the
Multisite Recording for the Analysis of Information Flow Between. . . 31
4.3 Neurons in CA1 To assess the trajectory information represented in a neural ensem-
and RE Represent the ble in CA1 and RE, we constructed a decoder to understand the
Next Trajectory in the information about the animal’s next trajectory represented in these
Alternation Task neurons. The mean firing rates on each spatial bin on the stem were
used as the inputs of the classifier that predicts the animal’s next
trajectory. We used a one-leave-out cross-validation approach to
assess the performance of two decoders based on either population
vector or support vector machine. Both classifiers are able to predict
the animal’s next trajectory choice significantly better than a chance
level of 50%, suggesting that neurons in both CA1 and RE repre-
sent the information about the next trajectory as a population
(Fig. 5d). We further confirmed that the support vector machine
classifier performed better than a population vector decoder, indi-
cating excellent performance of the machine-learning-based
approach. To further confirm the signal directionality of the trajec-
tory information, we recorded CA1 neurons from the animals with
lesions in RE. We then implemented the same decoding approach
based on a support vector machine and found that the decoding
performance of the animal’s next trajectory was significantly
impaired in these animals (Fig. 5d), suggesting that RE is necessary
for CA1 neurons to express the trajectory information.
Fig. 6 Optogenetic manipulation of the activity of RE neurons. (a) Top: Schema showing the design of optrode.
Two tetrodes are glued to the optic fiber so that the tetrodes should extend ~0.75 mm from the tip of the optic
fiber. Bottom. After the injection of AAV-Jaws, the optrode was inserted so that the tip of the optic fibers is
0.25–0.5 mm above the RE. (b) Silencing of a representative RE cell during laser application. The top plot
shows the spike raster and the bottom shows the spike-count histograms. The 635 nm laser was applied
between 0 and 5 sec in the time axis. The spike activity was almost completely silenced during the laser
application. (c) Trajectory-dependent firing of a CA1 cell during the silencing of RE. A representative CA1 place
cell showing the trajectory-dependent rate change between right- and left-oriented trajectories. When RE
neurons were silenced by the laser application, the same cell still exhibited position-selective firing at the
stem, but its trajectory-dependent rate change was largely diminished, indicating that RE cells dynamically
modulate the firing rates of CA1 cells depending on the animal’s trajectory plan of ongoing trials
References
1. Singer W (1993) Synchronization of cortical 12. O’Keefe J, Nadel L (1978) The hippocampus
activity and its putative role in information as a cognitive map. Oxford University Press,
processing and learning. Annu Rev Physiol Oxford, UK
55:349–374 13. Moser EI, Kropff E, Moser MB (2008) Place
2. Gregoriou GG, Gotts SJ, Zhou H, Desimone cells, grid cells, and the brain’s spatial represen-
R (2009) Long-range neural coupling through tation system. Annu Rev Neurosci 31:69–89
synchronization with attention. Prog Brain Res 14. Siapas AG, Lubenov EV, Wilson MA (2005)
176:35–45 Prefrontal phase locking to hippocampal theta
3. Fries P (2015) Rhythms for cognition: com- oscillations. Neuron 46:141–151
munication through coherence. Neuron 88: 15. Benchenane K, Peyrache A, Khamassi M, Tier-
220–235 ney PL, Gioanni Y, Battaglia FP, Wiener SI
4. Jones MW, Wilson MA (2005) Theta rhythms (2010) Coherent theta oscillations and reorga-
coordinate hippocampal-prefrontal interac- nization of spike timing in the hippocampal-
tions in a spatial memory task. PLoS Biol 3: prefrontal network upon learning. Neuron 66:
e402 921–936
5. Ji D, Wilson MA (2007) Coordinated memory 16. Ito HT, Zhang SJ, Witter MP, Moser EI,
replay in the visual cortex and hippocampus Moser MB (2015) A prefrontal-thalamo-hip-
during sleep. Nat Neurosci 10:100–107 pocampal circuit for goal-directed spatial navi-
6. Tort AB, Kramer MA, Thorn C, Gibson DJ, gation. Nature 522:50–55
Kubota Y, Graybiel AM, Kopell NJ (2008) 17. Hallock HL, Wang A, Griffin AL (2016) Ven-
Dynamic cross-frequency couplings of local tral midline thalamus is critical for
field potential oscillations in rat striatum and hippocampal-prefrontal synchrony and spatial
hippocampus during performance of a T-maze working memory. J Neurosci 36:8372–8389
task. Proc Natl Acad Sci U S A 105: 18. Ito HT, Moser EI, Moser MB (2018) Supra-
20517–20522 mammillary nucleus modulates spike-time
7. Colgin LL, Denninger T, Fyhn M, Hafting T, coordination in the prefrontal-Thalamo-hippo-
Bonnevie T, Jensen O, Moser MB, Moser EI campal circuit during navigation. Neuron 99:
(2009) Frequency of gamma oscillations routes 576–587.e5
flow of information in the hippocampus. 19. Herkenham M (1978) The connections of the
Nature 462:353–357 nucleus reuniens thalami: evidence for a direct
8. Ito HT (2017) Prefrontal-hippocampal inter- thalamo-hippocampal pathway in the rat. J
actions for spatial navigation. Neurosci Res Comp Neurol 177:589–610
129:2–7 20. Vertes RP, Linley SB, Groenewegen HJ, Witter
9. Rainer G, Rao SC, Miller EK (1999) Prospec- MP (2014) Thalamus. In: Paxinos G (ed) The
tive coding for objects in primate prefrontal rat nervous system. Elsevier Academic Press,
cortex. J Neurosci 19:5493–5505 San Diego
10. Gold JI, Shadlen MN (2007) The neural basis 21. McNaughton BL, O’Keefe J, Barnes CA
of decision making. Annu Rev Neurosci 30: (1983) The stereotrode: a new technique for
535–574 simultaneous isolation of several single units in
11. O’Keefe J, Dostrovsky J (1971) The hippo- the central nervous system from multiple unit
campus as a spatial map. Preliminary evidence records. J Neurosci Methods 8:391–397
from unit activity in the freely-moving rat. 22. Gray CM, Maldonado PE, Wilson M,
Brain Res 34:171–175 McNaughton B (1995) Tetrodes markedly
improve the reliability and yield of multiple
Another random document with
no related content on Scribd:
large party of Miaki’s men, all armed, and watching as outposts.
Some were for shooting us, but others hesitated. Every musket was,
however, raised and levelled at me. Faimungo poised his great spear
and said, “No, you shall not kill Missi to-day. He is with me.” Having
made this flourish, he strode off after his own men, and my
Aneityumese followed, leaving me face to face with a ring of levelled
muskets. Sirawia, who was in command of this party, and who once
like Nowar had been my friend, said to me, Judas like, “My love to
you, Missi.” But he also shouted after Faimungo, “Your conduct is
bad in taking the Missi away; leave him to us to be killed!”
I then turned upon him, saying, “Sirawia, I love you all. You must
know that I sought only your good. I gave you medicine and food
when you and your people were sick and dying under measles; I gave
you the very clothing you wear. Am I not your friend? Have we not
often drunk tea and eaten together in my house? Can you stand there
and see your friend shot? If you do, my God will punish you
severely.”
He then whispered something to his company which I did not
hear; and, though their muskets were still raised, I saw in their eyes
that he had restrained them. I therefore began gradually to move
backwards, still keeping my eyes fixed on them, till the bush hid
them from my view, whereon I turned and ran after my party, and
God kept the enemy from following. I would like to think that
Sirawia only uttered the cruel words which I heard as a blind to save
his own life; for at this time he was joined to Miaki’s party, his own
people having risen against him, and had to dissemble his friendly
feelings towards me. Poor Sirawia! Well I knew that Miaki would
only use him as a tool for selfish interests, and sacrifice him at last.
All this showed how dangers grew around our path. Still we trusted
in Jehovah Jesus, and pressed on in flight. A second hostile party
encountered us, and with great difficulty we also got away from
them. Soon thereafter a friendly company crossed our path. We
learned from them that the enemies had slaughtered other two of
Manuman’s men, and burned several villages with fire. Another
party of the enemy encountered us, and were eager for our lives. But
this time Faimungo withstood them firmly, his men encircled us, and
he said, “I am not afraid now, Missi; I am feeling stronger near my
own land!”
Hurrying still onwards, we came to that village on their high
ground called Aneai, i.e., Heaven. The sun was oppressively hot, the
path almost unshaded, and our whole party very exhausted,
especially Faimungo, carrying his load of stolen goods. So here he sat
down on the village dancing ground for a smoke, saying,—
“Missi, I am near my own land now. We can rest with safety.”
In a few minutes, however, he started up, he and his men, in wild
excitement. Over a mountain, behind the village and above it, there
came the shoutings, and anon the tramp, tramp of a multitude
making rapidly towards us. Faimungo got up and planted his back
against a tree. I stood beside him, and the Aneityumese woman and
the two men stood near me, while his men seemed prepared to flee.
At full speed a large body of the tallest and most powerful men that I
had seen on Tanna came rushing on and filled the dancing ground.
They were all armed, and flushed with their success in war. A
messenger had informed them of our escape, probably from Miaki,
and they had crossed the country to intercept us. Faimungo was
much afraid, and said,—
“Missi, go on in that path, you and your Aneityumese; and I will
follow when I have had a smoke and a talk with these men.”
I replied, “No, I will stand by your side till you go; and if I am
killed, it will be by your side. I will not leave you.”
He implored us to go on, but that I knew would be certain death.
They began urging one another to kill us, but I looked round them as
calmly as possible, saying, “My Jehovah God will punish you here
and hereafter, if you kill me or any of His servants.”
A killing-stone, thrown by one of the savages, grazed poor old
Abraham’s cheek, and the dear soul gave such a look at me, and then
upwards, as if to say, “Missi, I was nearly away to Jesus.” A club was
also raised to follow the blow of the killing-stone, but God baffled the
aim. They encircled us in a deadly ring, and one kept urging another
to strike the first blow or fire the first shot. My heart rose up to the
Lord Jesus; I saw Him watching all the scene. My peace came back to
me like a wave from God. I realized that I was immortal till my
Master’s work with me was done. The assurance came to me, as if a
voice out of Heaven had spoken, that not a musket would be fired to
wound us, not a club prevail to strike us, not a spear leave the hand
in which it was held vibrating to be thrown, not an arrow leave the
bow, or a killing-stone the fingers, without the permission of Jesus
Christ, whose is all power in Heaven and on Earth. He rules all
Nature, animate and inanimate, and restrains even the savage of the
South Seas. In that awful hour I saw His own words, as if carved in
letters of fire upon the clouds of Heaven: “Seek, and ye shall find.
Whatsoever ye shall ask in My name, that will I do, that the Father
may be glorified in the Son.” I could understand how Stephen and
John saw the glorified Saviour as they gazed up through suffering
and persecution to the Heavenly Throne! Yet I never could say that
on such occasions I was entirely without fear. Nay, I have felt my
reason reeling, my sight coming and going, and my knees smiting
together when thus brought close to a violent death, but mostly
under the solemn thought of being ushered into Eternity and
appearing before God. Still, I was never left without hearing that
promise in all its consoling and supporting power coming up through
the darkness and the anguish, “Lo, I am with you alway.” And with
Paul I could say, even in this dread moment and crisis of being, “I am
persuaded that neither death nor life ... nor any other creature, shall
be able to separate us from the love of God which is in Christ Jesus
our Lord.”
Faimungo and others now urged us to go on in the path. I said,
“Faimungo, why are we to leave you? My God heard your promise
not to betray me. He knows now what is in your heart and in mine. I
will not leave you; and if I am to die, I will die by your side.”
He replied, “Now, I go on before; Missi, keep close to me.”
His men had gone, and I persuaded my Aneityumese to follow
them. At last, with a bound, Faimungo started after them. I followed,
keeping as near him as I could, pleading with Jesus to protect me or
to take me home to Glory. The host of armed men also ran along on
each side with their weapons ready; but leaving everything to Jesus, I
ran on as if they were my escort, or as if I saw them not. If any reader
wonders how they were restrained, much more would I, unless I
believed that the same Hand that restrained the lions from touching
Daniel held back these savages from hurting me! We came to a
stream crossing our path. With a bound all my party cleared it, ran
up the bank opposite, and disappeared in the bush. “Faint yet
pursuing,” I also tried the leap, but I struck the bank and slid back on
my hands and knees towards the stream. At this moment I heard a
crash above my head amongst the branches of an overhanging tree,
and I knew that a killing-stone had been thrown, and that that
branch had saved me. Praising my God, I scrambled up on the other
side, and followed the track of my party into the bush. The savages
gazed after me for a little in silence, but no one crossed the stream;
and I saw them separate into two, one portion returning to the
village and another pressing inland. With what gratitude did I
recognise the Invisible One who brought their counsels to confusion!
I found my party resting in the bush, and amazed to see me
escaped alive from men who were thirsting for my blood. Faimungo
and his men received me with demonstrations of joy, perhaps feeling
a little ashamed of their own cowardice. He now ascended the
mountain and kept away from the common path to avoid other
Native bands. At every village enemies to the Worship were ready to
shoot us. But I kept close to our guide, knowing that the fear of
shooting him would prevent their shooting at me, as he was the most
influential Chief in all that section of the island.
One party said, “Miaki and Karewick said that Missi made the
sickness and the hurricanes, and we ought to kill him.”
Faimungo replied, “They lie about Missi! It is our own bad conduct
that makes us sick.”
They answered, “We don’t know who makes the sickness; but our
fathers have taught us to kill all foreign men.”
Faimungo, clutching club and spear, exclaimed, standing betwixt
them and us, “You won’t kill Missi to-day!”
In the flight we passed springs and streamlets, but though parched
with sickening thirst, not one of us durst stoop down to drink, as we
should have been almost certainly killed in the act. Faimungo now
sent his own men home by a near path, and guided us himself till we
were close upon the shore. There, sitting down he said,—
“Missi, I have now fulfilled my promise. I am so tired, I am so
afraid, I dare not go farther. My love to you all. Now go on quickly!
Three of my men will go with you to the next rocks. Go quickly!
Farewell.”
These men went on a little, and then said, “Missi, we dare not go!
Faimungo is at war with the people of the next land. You must keep
straight along this path.”
So they turned and ran back to their own village.
To us this district was especially perilous. Many years ago the
Aneityumese had joined in a war against the Tannese of this tribe,
and the thirst for revenge yet existed in their hearts, handed down
from sire to son. Besides, Miaki had incited the people here to
murder the Teachers and me if we attempted to escape this way.
Most providentially the men were absent on a war expedition, and
we saw only three lads and a great number of women and children,
who ran off to the bush in terror. In the evening the enraged savages
of another district assaulted the people of the shore villages for
allowing us to pass, and, though sparing their lives, broke in pieces
their weapons of war—a very grievous penalty. In the next district, as
we hasted along the shore, two young men came running after us,
poising their quivering spears. I took the useless revolver out of my
little native basket, and raising it cried,—
“Beware! Lay down your spears at once on the sand, and carry my
basket to the next landing at the black rocks.”
They threw their spears on the sand, lifted the bag, and ran on
before us to the rocks which formed the march betwixt them and
their enemies. Laying it down, they said appealingly, “Missi, let us
return to our home!” And how they did run, fearing the pursuit of
their foes.
In the next land we saw none. After that we saw crowds all along,
some friendly, others unfriendly, but they let us pass on, and with the
blessing of Almighty God we drew near to Mr. Mathieson’s Station in
safety. Here a man gave me a cocoa-nut for each of our party, which
we greatly required, having tasted nothing all that day, and very little
for several days before. We were so weak that only the struggle for
life enabled us to keep our feet; yet my poor Aneityumese never
complained and never halted, not even the woman. The danger and
excitement kept us up in the race for life, and by the blessing of God
we were now approaching the Mission House, praising God for His
wonderful deliverances.
Hearing of our coming, Mr. Mathieson came running to meet me.
They had heard of my leaving my own Station, and they thought I
was dead! They were themselves both very weak; their only child had
just been laid in the grave, and they were in great grief and in greater
peril. We praised the Lord for permitting us to meet; we prayed for
support, guidance, and protection; and resolved now, in all events, to
stand by each other till the last.
Before I left the Harbour I wrote and left with Nowar letters to be
given to the Captains of any vessels which called, for the first, and the
next, and the next, telling them of our great danger, that Mr.
Mathieson was almost without food, and that I would reward them
handsomely if they would call at the Station and remove any of us
who might be spared thence to Aneityum. Two or three vessels
called, and, as I afterwards learned, got my letters; but, while buying
my stolen property from the Natives for tobacco, powder, and balls,
they took no further notice of my appeals, and sailed past Mr.
Mathieson’s straight on to Aneityum. “The tender mercies of the
wicked are cruel!”
Let me now cull the leading events from my Journal, that
intervened betwixt this date and the final break-up of the Mission on
Tanna—at least for a season—though, blessed be God! I have lived to
see the light rekindled by my dear friends Mr. and Mrs. Watt, and
shining more brightly and hopefully than ever. The candle was
quenched, but the candle-stick was not removed!
On Wednesday, 22nd January, 1862, we heard that other three of
Manuman’s people had been killed and a district burned with fire.
Though this poor man was one of Nowar’s chief friends, yet I heard
him say before my flight, “When so many children are being killed,
why do they not send one for food to me and my family? They are as
tender and good as the young fowls!” A remark like this lets you see
deep into the heart of a Cannibal, and he a sort of half-converted one,
if I may use such an expression; certainly not one of the worst type
by any means.
On the 23rd January, Mr. Mathieson sent for Taura, Kati, and
Kapuku, his three principal Chiefs, to induce them to promise
protection till a vessel called to take us away. They appeared friendly,
and promised to do their best. Alas! the promises of the Tannese
Chiefs had too often proved to be vain.
On Friday, 24th January, report reached our Station that Miaki
and his party, hearing that a friendly Chief had concealed two of
Manuman’s young men, compelled him to produce them and club
them to death before their eyes. Also, that they surrounded
Manuman’s party on a mountain, and hemmed them in there, dying
of starvation and trying to survive on the carcases of the dead and on
bark and roots. Also, that Miaki had united all the Chiefs, friends and
foes alike, in a bond of blood, to kill every one pertaining to the
whole Mission on Tanna. Jesus rules.
SPRINGING FORWARD HE
CAUGHT THE CLUB.
For the present, my pen is here laid aside. I shall wait to see what
use the Lord makes of Part First of my autobiography, before I
prosecute the theme. If the Christian public seems not to find in it
the help and quickening that some friends think it likely to bestow on
those who read, the remainder need not be written. Part Second, if
called for, will contain a record, in many respects, an utter contrast
to all that has gone before, and yet directly springing therefrom, as
will be seen by all who look beneath the surface. I am penning these
words in 1887, and five-and-twenty years lie betwixt this date and
my farewell to Tanna. These years, if ever published, will tell the
story of my visiting all the Colonial Churches, and collecting the
purchase money of our white-winged Mission Ship, the Dayspring;
my return to Scotland, visiting all the home congregations in 1864,
and securing several new Missionaries to follow me to the New
Hebrides; my second marriage, and settlement on Aniwa, with her
whom the good Lord still spares to me, the mother of our happy
family, and my God-given helpmeet in all the work of the Gospel; the
conversion of that whole island of Aniwa from idolatry, and the
planting there of a Church and Congregation of Christ, from which
have since gone forth many Native Evangelists and Teachers. Then
there will fall to be recorded my call from the Islands in recent years
to revisit all the Colonial Presbyterian Congregations once again,
telling them the story of the Conversion of Aniwa—the sinking of the
well, and other incidents, which turned an entire people from idols
and from cannibalism to the service of the living and true God—
whereby the Churches, and especially the children, were led more
and more to make the New Hebrides their own very harvest field in
the Heathen world. And finally, I will have to tell how I was again
sent home to Scotland in 1884 to raise money for the purchase or
building of a steam-auxiliary Mission Ship, now urgently required in
the interests of the Mission, both because of the great increase in the
number of the Missionaries and the necessities of so many families;
and also and chiefly to avert the dreadful disappointments and loss
of time, and thereby sometimes of life itself, caused by the frequent
becalming of our little Dayspring in these thickly-islanded seas. That
part of the story will show the fruits of the education and perils and
experiences of a lifetime, in the marvellous impression produced by
the simple and unadorned recital of the story of Tanna and Aniwa,
amongst the Christian people of Scotland, Ireland, and England.
Multitudes were blessed in almost every town where a meeting was
granted me. Three Missionaries devoted themselves to the New
Hebrides, and are already labouring there; while others consecrated
themselves to several of the great seats of Foreign Mission enterprise
in Africa and Asia. I returned to my own Church of Victoria with a
sum of nearly £9,000, of which £6,000 was for the new Missionary
Steam-Auxiliary, and the remainder for the outfit and support of
more Missionaries for the Islands; and that money I handed over to
the Australian Church, where it awaits, at interest in the bank, the
arrangements being made by all the Colonies to take each their due
share in the future up-keep of the Ship. For this—for everything—for
all, praise be to the Lord! I never asked one subscription, except in
prayer and in my public appeals. The Lord sent in all freely to me
through the hands of His people; to Him be all the glory. I went back
to Aniwa, and found the work of the Lord going forward there as if in
a regularly settled Congregation at home, fostered and guided by an
occasional visit of my ever dear and genuine friends, Mr. and Mrs.