Neurobiology of Learning and Memory xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Neurobiology of Learning and Memory

journal homepage: www.elsevier.com/locate/ynlme

Neural representations of time-linked memory

⁎
Maryna Pilkiwa, Kaori Takehara-Nishiuchia,b,c,
a
Department of Cell and Systems Biology, University of Toronto, Toronto M5S 3G3, Canada
b
Department of Psychology, University of Toronto, Toronto M5S 3G3, Canada
c
Neuroscience Program, University of Toronto, Toronto M5S 3G3, Canada

A R T I C LE I N FO A B S T R A C T

Keywords: Many cognitive processes, such as episodic memory and decision making, rely on the ability to form associations
Time between two events that occur separately in time. The formation of such temporal associations depends on
Episodic memory neural representations of three types of information: what has been presented (trace holding), what will follow
Temporal association (temporal expectation), and when the following event will occur (explicit timing). The present review seeks to
Neural code
link these representations with firing patterns of single neurons recorded while rodents and non-human primates
Single-unit activity
associate stimuli, outcomes, and motor responses over time intervals. Across these studies, two distinct firing
Ensemble decoding
patterns were observed in the hippocampus, neocortex, and striatum: some neurons change firing rates during or
shortly after the stimulus presentation and sustain the firing rate stably or sidlingly during the subsequent
intervals (tonic firings). Other neurons transiently change firing rates during a specific moment within the time
intervals (phasic firings), and as a group, they form a sequential firing pattern that covers the entire interval.
Clever task designs used in some of these studies collectively provide evidence that both tonic and phasic firing
responses represent trace holding, temporal expectation, and explicit timing. Subsequently, we applied machine-
learning based classification approaches to the two firing patterns within the same dataset collected from rat
medial prefrontal cortex during trace eyeblink conditioning. This quantitative analysis revealed that phasic-
firing patterns showed greater selectivity for stimulus identity and temporal position than tonic-firing patterns.
Our summary illuminates distributed neural representations of temporal association in the forebrain and gen-
erates several ideas for future investigations.

1. Introduction linked memories. Parallel neurophysiological studies reported several

When you hear a ringtone for email notification, you expect to see a
new email on the screen of your phone. When you hear the sound of
skidding tires on a busy street, you feel a sense of dread for an imminent
car crash. When you see the tail lights of the car in front of yours go off,
you prepare to release your break. All of these behaviors are the out-
comes of processes through which the brain formed association be-
tween two arbitrary events that took place separately within a short
time window. The formation of these time-linked memories depends on
neural representations of at least three types of information. First, the
brain must continue representing events after they have passed (trace
holding). Second, expectation signals need to be formed regarding
which event would follow the first event (temporal expectation). Third,
the brain needs to represent the exact timing at which the subsequent
event would take place (explicit timing).
Over the past decades, a considerable number of studies has been
conducted to identify brain regions necessary for the formation of time-
unique firing patterns of single neurons during intervals separating two
events. This review provides a concise summary of these studies with
the aim of linking specific neuron firing patterns in a given brain region
with the information types required for the formation of time-linked
memory. In the following section, we will first outline several beha-
vioral paradigms used to study the association formed between stimuli,
outcomes, and motor responses across time. A particular emphasis will
be placed on their differences in the relative reliance on trace holding,
temporal expectation, and explicit timing. Subsequently, we will review
neurophysiological studies that characterized neuron firing patterns
during time intervals separating two events. We will highlight the
studies whose clever design uncovers the selectivity of specific neural
firing patterns for the three information types required for time-linked
memory. The subsequent section will expand the discussion with the
result of quantitative decoding analysis applied to two largely non-
overlapping groups of neurons with distinct firing profiles in rat medial
prefrontal cortex. We believe that the thorough survey of available

⁎
Corresponding author at: Department of Psychology, University of Toronto, Toronto M5S 3G3, Canada
E-mail address: takehara@psych.utoronto.ca (K. Takehara-Nishiuchi).

https://doi.org/10.1016/j.nlm.2018.03.024
Received 29 October 2017; Received in revised form 29 March 2018; Accepted 30 March 2018
1074-7427/ © 2018 Elsevier Inc. All rights reserved.

Please cite this article as: Pilkiw, M., Neurobiology of Learning and Memory (2018), https://doi.org/10.1016/j.nlm.2018.03.024
M. Pilkiw, K. Takehara-Nishiuchi
literature will contribute to (1) uncovering a potential functional dif-
ference, or absence thereof, among the variety of neural firing patterns
observed during the interval between temporally discontinuous events
and (2) providing ideas for future experiments.
2. Behavioral paradigms used to study time-linked memory
Various behavioral paradigms have been used to study mechanisms
underlying time-linked memory. Based on their design, these paradigms
can be categorized into three groups. In some paradigms, like delayed
paired-associate tasks, associations are formed between sensory stimuli,
such as images, objects, or object-location pairs presented with a tem-
poral gap (stimulus-stimulus association). In others, like classical trace
conditioning, subjects are trained to associate a sensory stimulus with
reward or punishment that occurs after a temporal delay (stimulus-
outcome association). In another set of paradigms, such as conditional
visuomotor learning and interval timing tasks, animals need to form
association between a sensory stimulus and a particular motor response
over a fixed time interval (stimulus-response association). This section
summarizes the design of these behavioral paradigms with an emphasis
on their differences in the degree of dependence on trace holding,
temporal expectation, and explicit timing.
2.1. Stimulus-stimulus associations
One paradigm traditionally used to characterize neural firing pat-
terns underlying time-linked memory is a delayed paired-associate task.
In the learning phase of this task, animals are repeatedly presented with
the pairs of arbitrary neutral stimuli. After the subjects have reached a
behavioral criterion, their memory for associations is tested. First, one
stimulus from a pair is presented as a cue, followed by a delay period.
Then, the animal is presented with a choice of two stimuli, one of which
was previously paired with the cue and another one from a different
pair. To receive a reward, the subject must choose the stimulus which
completes the pair. To successfully perform this task, animals must not
only maintain the identity of the cue during the delay period (trace
holding) but also prospectively retrieve a stimulus that was paired with
the cue during the learning phase (temporal expectation). This latter
requirement makes the paired-associate tasks different from working
memory tasks, in which correct responses depend solely on the cue
maintenance over the delay period. The successful performance does
not require the subject to track the duration of the delay period (explicit
timing) because the end of the delay period is explicitly signaled by the
presentation of the choice stimuli.
In primates, pairs of visual stimuli, such as unrelated color images of
scenes and objects, or monochrome geometric patterns (Fourier de-
scriptors) are frequently used (Fujimichi et al., 2010; Higuchi &
Miyashita, 1996; Naya, Yoshida, & Miyashita, 2003, 2001; Rainer, Rao,
& Miller, 1999; Sakai & Miyashita, 1991). In these studies, a monkey
typically forms the paired association after a few hundred presentations
per stimulus pair (Higuchi & Miyashita, 1996; Sakai & Miyashita,
1991). The duration of a cue period is 0.5–1 sec, followed by a delay
period of 1–4 sec. Murray, Gaffan, and Mishkin (1993) showed that
combined bilateral removal of the perirhinal and entorhinal cortices
impairs the retrieval of previously acquired paired associates as well as
the acquisition of new associates.
In rodents, the delayed paired-associate task pairs two arbitrary
odors (Bunsey & Eichenbaum, 1996) or an object and an odor
(MacDonald, Lepage, Eden, & Eichenbaum, 2011). For example, when a
rat has placed its nose in a port, a cue odor is presented for 0.75 sec,
followed by a 0.5 sec delay with no stimuli, and another odor for
0.75 sec which is an associate of either the cue or a different odor. If the
second odor is paired with the cue, the rat has to approach a water port
to receive a reward. Combined lesions to the entorhinal and perirhinal
cortices before learning prevent rats from forming associations between
the odors (Bunsey & Eichenbaum, 1993), a result comparable to that
2
Neurobiology of Learning and Memory xxx (xxxx) xxx–xxx
with monkeys (Murray et al., 1993). In contrast, rats with damage to
the dorsal hippocampus are able to acquire odor-odor associations;
however, they are not able to correctly respond to an odor from a pair
when it is presented in a temporal order reversed from the one used
during training (Bunsey & Eichenbaum, 1996).
In a different version of the task, a cue object is presented for
1.2–3 sec, followed by a 10 sec delay and a presentation of a container
filled with scented sand (Kesner, Hunsaker, & Gilbert, 2005). If the
scent was previously associated with the cue object, the animal must
respond by digging in the sand to find a reward, or otherwise, it must
withhold the response. Notably, the duration of a delay period in this
task is ∼10 sec, which is longer than the primate version of the delayed
paired-associate task with visual stimuli (a few seconds). Rats with le-
sions of the CA1 region of the dorsal hippocampus cannot form the
association between an object and an odor stimulus over the 10-sec
delay while lesions of the CA3 region do not have any effect on the task
performance (Kesner et al., 2005). Notably, lesions of the CA3, but not
CA1, impair the task performance when the cue and choice phases are
not separated by the delay (Gilbert & Kesner, 2003), suggesting the
unique involvement of the CA1 region for trace holding, temporal ex-
pectation, or both.
2.2. Stimulus-outcome associations
Other commonly used paradigms to study time-linked memory are
the trace paradigms of classical conditioning, such as trace eyeblink
conditioning and trace fear conditioning. In these paradigms, a neutral
conditioned stimulus (CS) is followed by an aversive stimulus that is
presented after a temporal gap, called a trace interval. To associate the
CS with the US, the brain must maintain the “trace” of the CS during the
temporal gap (Graves & Solomon, 1985; Kamin & Schaub, 1963;
Pavlov, 1927). Before conditioning, the US induces a specific reflex
response (unconditioned response, UR), whereas the CS does not. Re-
peated pairings of the CS and US build the association between them,
resulting in the development of anticipatory behavioral responses that
precede the onset of US (conditioned response, CR). The CR reaches the
maximum intensity at the expected onset of the US, suggesting that
animals not only acquire the association between the CS and US but
also the timing of the US presentation. Therefore, for successful
learning in the trace paradigms, animals must maintain the re-
presentation of the CS during the trace interval (trace holding), expect
the upcoming US (temporal expectation), and estimate precisely when
the US will be presented (explicit timing).
Trace fear conditioning presents footshock as the US after a
20–30 sec delay from the offset of the CS (Marlin, 1981; McEchron,
Bouwmeester, Tseng, Weiss, & Disterhoft, 1998). Rats, mice or rabbits
acquire the conditioned response, such as freezing and heart rate in-
crease, in one session that contains 5–16 CS-US pairings. The dorsal
hippocampus plays a critical role in the acquisition and expression of
the association, but its involvement is limited to the paradigm with a
trace interval longer than 10 sec (Bangasser, Waxler, Santollo, & Shors,
2006; Burman, Starr, & Gewirtz, 2006; Chowdhury, Quinn, & Fanselow,
2005; Misane et al., 2005). In the hippocampus-dependent version of
this task, many additional brain regions are involved in acquisition and
retrieval. Those include the ventral hippocampus (Burman et al., 2006;
Yoon & Otto, 2007), prelimbic cortex (Gilmartin & Helmstetter, 2010;
Runyan, Moore, & Dash, 2004), anterior cingulate cortex (Han et al.,
2003), perirhinal cortex (Kholodar-Smith, Boguszewski, & Brown,
2008), entorhinal cortex (Esclassan, Coutureau, Di Scala, & Marchand,
2009), and amygdala (Gilmartin, Kwapis, & Helmstetter, 2012; Kwapis,
Jarome, Schiff, & Helmstetter, 2011; Selden, Everitt, Jarrard, &
Robbins, 1991 but see Raybuck & Lattal, 2011). Moreover, the acqui-
sition of trace fear memory is impaired by optogenetic inhibition of
projections from the medial entorhinal cortex (MEC) to the dorsal CA1
region during the trace interval, but not during the CS (Kitamura et al.,
2014). Similarly, photoinhibition of the prelimbic region of the medial
M. Pilkiw, K. Takehara-Nishiuchi
prefrontal cortex during the trace interval, but not during the CS, pre-
vents rats from forming the CS-US association (Gilmartin, Miyawaki,
Helmstetter, & Diba, 2013). These results suggest that the formation of
CS-US association depends on the proper activity of the MEC-CA1
projections and the prelimbic region during the trace interval.
In trace eyeblink conditioning, the US is applied as an air-puff to the
cornea or an electric shock near the eyelid. After 300–800 presentations
of the CS-US pairings over several days, animals form anticipatory
blinking responses whose intensity peaks at the expected timing of the
US (Moyer, Deyo, & Disterhoft, 1990; Weiss, et al., 1999). The duration
of trace interval is 250–500 msec in rats (Beylin et al., 2001; Weiss
et al., 1999) and mice (Kishimoto, Kawahara, Mori, Mishina, & Kirino,
2001; Takehara, Kawahara, Takatsuki, & Kirino, 2002; Tseng, Guan,
Disterhoft, & Weiss, 2004) and 250–750 msec in rabbits (Graves &
Solomon, 1985; Hoehler & Thompson, 1980; Solomon, Vander Schaaf,
Thompson, & Weisz, 1986). Most mice and rats cannot form the CS-US
association if the duration of the trace interval exceeds 250 and 500 ms,
respectively (Tseng et al., 2004; Volle et al., 2016). Memory acquisition
in trace eyeblink task requires an intact hippocampus (Beylin et al.,
2001; Moyer et al., 1990; Solomon et al., 1986), medial prefrontal
cortex (Kronforst-Collins & Disterhoft, 1998; Takehara-Nishiuchi,
Kawahara, & Kirino, 2005; Weible, McEchron, & Disterhoft, 2000),
entorhinal cortex (Ryou, Cho, & Kim, 2001; Tanninen et al., 2015),
perirhinal, postrhinal cortices (Suter, Weiss, & Disterhoft, 2013), med-
iodorsal thalamus (Powell & Churchwell, 2002), caudate nucleus
(Flores & Disterhoft, 2009), and cerebellum (Pakaprot, Kim, &
Thompson, 2009; Woodruff-Pak, Lavond, & Thompson, 1985). Notably,
a dysfunction of these forebrain regions does not impair delay eyeblink
conditioning in which the CS co-terminates with the US (Berger & Orr,
1983; Morrissey, Maal-Bared, Brady, & Takehara-Nishiuchi, 2012;
Schmaltz & Theios, 1972; Takehara-Nishiuchi et al., 2005). Also, within
trace eyeblink paradigms, lesions to the entorhinal cortex do not impair
acquisition of CS-US association in rabbits if the trace interval is 300 ms
(Ryou et al., 2001). The rats with the hippocampal lesion can learn the
association if the trace interval is reduced to 50 ms (Walker &
Steinmetz, 2008). These observations suggest that the forebrain regions
are engaged in learning only when the association needs to be formed
across a sufficiently long time interval.
As reviewed above, trace fear and trace eyeblink conditioning share
many common components in the underlying network structure. They,
however, differ in the number of CS-US pairings required for the ac-
quisition of CRs: animals are able to acquire CRs in trace fear con-
ditioning with a few CS-US pairings while CR acquisition in trace
eyeblink conditioning requires a few hundreds CS-US pairings. There
are two factors that may account for this difference. Firstly, the US in
trace fear conditioning (footshock) is more aversive than the US in trace
eyeblink conditioning (mild electric shock to the eyelid or corneal air-
puff). Secondly, CRs in trace fear conditioning, such as freezing re-
sponses and heart rate changes, depend on the periaqueductal grey and
lateral hypothalamus (LeDoux, Iwata, Cicchetti, & Reis, 1988) while
CRs in trace eyeblink conditioning, such as the movement of eyelid and
nictitating membrane depend on the cerebellum (Hu et al., 2010;
Pakaprot et al., 2009; Woodruff-Pak et al., 1985). Plasticity in these
brain regions may develop with different speed.
2.3. Stimulus-response associations
Another set of paradigms requires subjects to link a sensory stimulus
with a specific motor response over a temporal interval. In one version
of these tasks, primate conditional visuomotor learning tasks, subjects
are trained to associate an arbitrary visual stimulus with a movement of
the eye or hand in a particular direction. The subjects need to withhold
the motor response until the fixation point disappears from the screen
after the fixed duration of a few seconds. Successful performance in this
task depends on the maintenance of the visual stimulus during the
fixation period (trace holding) or the preparation for the correct motor
3
Neurobiology of Learning and Memory xxx (xxxx) xxx–xxx
response (temporal expectation); however, the subject does not need to
estimate the elapsed time from the stimulus offset (explicit timing). The
acquisition of new sensorimotor associations is impaired by combined
ablation of the hippocampus and subjacent cortex (Murray & Wise,
1996), the dorsolateral prefrontal cortex (Petrides, 1982), and the dis-
connection of these regions (Eacott & Gaffan, 1992; Gaffan & Harrison,
1988; Parker & Gaffan, 1998).
Another version of stimulus-response association tasks, the interval
timing task, requires subjects to learn the correct timing at which they
need to initiate a motor response. A subject is presented with a cue
stimulus, such as a light or sound, and it must make a motor response
no sooner than a determined time interval after the cue presentation
(usually 12–60 sec). With learning, the timing at which animals initiate
the response comes to center around the expected timing of the reward
delivery. The acquisition of the well-timed response is impaired with
lesions to the dorsal hippocampus (Tam & Bonardi, 2012; Yin & Meck,
2014), ventral hippocampus (Yin & Meck, 2014), medial frontal cortex
(Narayanan, Horst, & Laubach, 2006; Xu, Zhang, Dan, & Poo, 2014),
and caudate putamen (Meck, 2006). Furthermore, pharmacological
inhibition of the D1 receptor activity within the medial prefrontal
cortex makes the timing of behavioral response less accurate and more
variable (Narayanan, Land, Solder, Deisseroth, & DiLeone, 2012;
Parker, Chen, Kingyon, Cavanagh, & Narayanan, 2014). In parallel,
optogenetic inhibition of prefrontal neurons expressing D1-receptors
impairs well-timed behavioral responses, whereas optogenetic activa-
tion during the waiting period improves the temporal accuracy of the
response (Narayanan et al., 2012).
2.4. Difference and commonality in brain regions involved in paradigms
assessing time-linked memory
The survey of available behavioral studies highlights remarkable
similarity in the network structure underlying the formation of sti-
mulus-stimulus, stimulus-outcome, and stimulus-response associations
over time. The hippocampus is required for all three types of associa-
tions. In the case of trace conditioning, the necessity of the hippo-
campus for memory acquisition is contingent on the presence of suffi-
ciently long temporal gaps between two paired stimuli. These results
suggest that the hippocampus likely mediates trace holding, temporal
expectation, or both, which are required for all paradigms. The pre-
frontal cortex (dorsolateral parts in primates and medial parts in ro-
dents) is implicated in trace conditioning, conditional visuomotor
learning, and interval timing tasks, in which neutral stimuli gain
emotional or motivational salience. Therefore, the prefrontal cortex
may play a role in the assignment of emotional or motivational values
to the stimulus that was presented a moment ago, which is a case of
temporal expectation. Delayed paired-associate tasks, on the other
hand, involve the formation of multiple associations between arbitrary,
behaviourally irrelevant stimuli. This particular task feature demands
accurate discrimination between multiple perceptually similar visual
stimuli or objects and precise maintenance of the stimulus identity over
the temporal delay. This heightened demand on trace holding may re-
late to the unique involvement of the inferior temporal cortices in these
paradigms. Lastly, among the memory paradigms reviewed above, the
involvement of the cerebellum, caudate putamen, and dopamine system
is needed for trace eyeblink conditioning and interval timing tasks, but
not for paired associates tasks or conditional visuomotor learning tasks.
The difference between the two sets of the paradigms is the dependence
on the accurate estimation of the time elapsed after stimulus onset.
These regions, therefore, may be uniquely involved in representations
and computations of explicit timing, as was previously proposed (Coull
& Nobre, 2008; Ivry & Spencer, 2004; Petter, Lusk, Hesslow, & Meck,
2016).
M. Pilkiw, K. Takehara-Nishiuchi
3. Neural activity patterns observed during the paradigms of time-
linked memory
By using various behavioral paradigms reviewed in the previous
section, a considerable number of studies characterized firing patterns
of single neurons during intervals between sensory stimuli, outcomes,
and motor responses. Collectively, these studies reported two types of
firing patterns during the interval between events. One is the tonic
firing responses to the first event which are initiated during or after the
event, persisting into the subsequent interval until the presentation of
the second event. The other is phasic firing changes that occur during a
specific moment within the inter-event interval. Below we review stu-
dies separately for each of the two firing patterns and further divide
them based on the recording locations. A special focus is placed on the
studies that allow for inferring the types of information represented by
these firing patterns.
3.1. Tonic firings
3.1.1. Prefrontal cortex
During working memory paradigms some neurons in monkey dor-
solateral prefrontal cortex change firing rates in response to a stimulus
and sustain the firing rate during a subsequent interval until the sti-
mulus information is no longer needed (Fuster & Alexander, 1971).
Similar persistent firing patterns are observed in the lateral prefrontal
cortex during the interval between the cue and target stimulus in the
delayed paired-associate tasks (Rainer et al., 1999). The study utilizes a
unique design which interleaves delayed paired-associate (DPA) trials
with delayed-match-to-sample (DMS) trials. Both types of the trials
include different images as the cue but share the same image as the
target. About 25% of prefrontal neurons differentiate firing rates during
the cue-target interval between DPA and DMS trials, suggesting that
these neurons maintain the representation of the image used as the cue.
In parallel, ∼30% of neurons show a comparable firing rate between
the two trial types, suggesting that these neurons prospectively signal
the image to be presented as the target moments later. Toward the end
of the interval, ensemble firing patterns come to signal the upcoming
image more strongly than the preceding image, suggesting the priority
from trace holing to temporal expectation. Neurons selective for the
identity of the cue or target were also found in the lateral prefrontal
cortex during the conditional visuomotor learning task in which each
image is associated with a certain direction of the eye movement
(Asaad, Rainer, & Miller, 1998; Pasupathy & Miller, 2005). Collectively,
these findings suggest that neurons in the lateral prefrontal cortex re-
present trace holding as well as temporal expectation for impending
sensory stimuli and movement.
In rats and rabbits, persistent firings are observed in the medial
prefrontal cortex (mPFC) during trace eyeblink conditioning (Hattori,
Yoon, Disterhoft, & Weiss, 2014; Siegel, 2014, 2016; Siegel, Kalmbach,
Chitwood, & Mauk, 2012; Siegel & Mauk, 2013; Siegel et al., 2015;
Takehara-Nishiuchi and McNaughton, 2008) and trace fear con-
ditioning (Baeg et al., 2001; Gilmartin & McEchron, 2005). Typically,
neurons show abrupt changes in firing rate during the conditioned
stimulus (CS) and sustain the firing rates during subsequent intervals
until the unconditioned stimulus (US) is presented. These sustained
firing patterns are observed in animals that received paired presenta-
tions of the CS and US, but not animals that received unpaired pre-
sentations (Gilmartin & McEchron, 2005; Hattori et al., 2014). More-
over, in the trace eyeblink conditioning, ∼25% of neurons in the
prelimbic region of the mPFC maintain CS-evoked firing changes during
the subsequent interval only when the CS is predictive of the upcoming
US (Takehara-Nishiuchi & McNaughton, 2008). Furthermore, when rats
are trained to form the association between an auditory CS and eyelid
shock in parallel to the association between a visual CS and the same
eyelid shock, ∼40% of persistent-firing neurons show comparable
firing rates during CS-shock intervals regardless of the modality of the
4
Neurobiology of Learning and Memory xxx (xxxx) xxx–xxx
CS while the remaining 60% sustain CS-evoked firing rate only after
one of the two CS (Morrissey et al., 2017). Together, these results in-
dicate that persistent activity in the prelimbic region serves as a re-
presentation of trace holding and temporal expectation.
During interval timing tasks, on the other hand, gradual, ramping
changes in firing rates are reported in the anterior cingulate cortex
(Emmons et al., 2017; Matell, Meck, & Nicolelis, 2003; Parker,
Ruggiero, & Narayanan, 2015; Parker, Chen, Kingyon, Cavanagh, &
Narayanan, 2014) and the prelimbic region of the mPFC (Xu et al.,
2014) in rats. When the duration of the interval is modified, the slope of
ramping firing patterns is adjusted accordingly so that the firing rate of
each neuron reaches the maximum at the new expected end of the in-
terval (Emmons et al., 2017; Xu et al., 2014). Classification analysis
demonstrated that the exact timing within the interval can be decoded
from population activity of neurons with ramping firing patterns, but
not from neurons without the ramping patterns (Emmons et al., 2017),
providing evidence that the slope of ramping firing patterns represents
explicit timing.
3.1.2. Inferior temporal cortex
Persistent firings are also reported in the inferior temporal (IT)
cortex in primates during delayed paired-associate tasks. The IT cortex
can be divided into the Areas 35, 36 (a.k.a., perirhinal cortex) and Area
TE, all of which contain neurons that show selective firing patterns to
specific visual stimuli (for reviews, see Rolls, 2000; Tanaka, 1996).
During the delayed paired-associate task with 24 arbitrary visual sti-
muli, the majority of IT neurons show the strongest firing rate increase
to one of these stimuli (“preferred stimulus,” perceptual selectivity). In
well-trained monkeys, these neurons also show a weaker phasic re-
sponse to a stimulus paired with the preferred stimulus when it was
presented as a cue (pair selectivity, Naya et al., 2003, 2001; Sakai &
Miyashita, 1991). Following the initial phasic response, some of these
neurons gradually increase firing rates and maintain the stable rate
until their preferred stimulus is presented as a target stimulus (Naya
et al., 2003, 2001; Sakai & Miyashita, 1991). In Area 36, the difference
in persistent firing responses across 24 stimuli is correlated with the
perceptual and pair selectivity at the equal strength (Naya et al., 2003,
see also Erickson & Desimone, 1999). In contrast, in Area TE, the firing
differences are correlated with the pair selectivity more strongly than
the perceptual selectivity. These patterns suggest that while persistent
firings in Ares 36 represent the cue and target stimuli with comparable
strength, those in the TE are more selective for the impending target
stimulus than the cue stimulus. In addition, the pair, but not perceptual,
selectivity of TE neurons is attenuated by lesions of the perirhinal and
entorhinal cortex (Higuchi & Miyashita, 1996), suggesting that the pair
selectivity of TE neurons is driven by the activity of the rhinal regions
while their perceptual selectivity is likely driven by early visual areas
(see also, Naya, Yoshida, & Miyashita, 2001; Takeda, Naya, Fujimichi,
Takeuchi, & Miyashita, 2005).
An interesting contrast is found in the selectivity of neurons in the
perirhinal cortex during the conditional visuomotor learning task, in
which monkeys associate an image with one of four directions of sac-
cade responses (Yanike, Wirth, Smith, Brown, & Suzuki, 2009). About
68% of neurons with image-evoked firing responses are selective for the
identity of the image. In contrast, virtually no neurons show a com-
parable response to two images associated with the same rewarded
saccade direction. Collectively, these findings suggest that neurons in
the IT cortex represent trace holding and temporal expectation, but its
contribution to temporal expectation appears to be limited to im-
pending sensory stimuli.
3.1.3. Hippocampus
In primates conditional visuomotor learning tasks, a sizable pro-
portion of neurons in the hippocampus change firing rates during the
presentation of the cue, the subsequent cue-response interval, or both
(Wirth et al., 2003; Yanike, Wirth, & Suzuki, 2004). The majority of
M. Pilkiw, K. Takehara-Nishiuchi
these neurons differentiate cue-evoked firing rates depending on the
identity of the stimulus (∼70%) while the proportion of neurons dif-
ferentiating firing rates for the specific saccade direction is small
(∼10%; Wirth et al., 2003; Yanike et al., 2004). These patterns raise a
possibility that cue-evoked firing rates in the primate hippocampus play
a more important role for trace holding than temporal expectation. In
parallel to these cue-responding neurons, other neurons gradually
change firing rates during the cue-response interval regardless of the
identity of the cue or response (Sakon, Naya, Wirth, & Suzuki, 2014).
These observations, along with similar incremental firing patterns ob-
served during a temporal order paradigm (Naya & Suzuki, 2011), sug-
gest that the primate hippocampus contains a specialized class of
neurons that signal explicit timing and that these neurons exist in
parallel to neurons representing trace holding and temporal expecta-
tion.
Similar stimulus-evoked firing patterns are also reported in the
hippocampus of rabbits and cats during trace eyeblink conditioning
(Hattori, Chen, Weiss, & Disterhoft, 2015; McEchron & Disterhoft,
1997; McEchron, Weible, & Disterhoft, 2001; Munera, Gruart, Munoz,
Fernandez-Mas, & Delgado-Garcia, 2001). The proportion of CS-re-
sponding neurons is higher in animals which received pairings of the CS
with corneal air-puff (US) over a fixed interval than in animals which
received the presentation of the CS and US with a random interval.
Although these patterns suggest that some neurons in the hippocampus
signal temporal expectation, it is important to note that hippocampal
neurons show similar firing responses even in delay eyeblink con-
ditioning which does not include the temporal separation of the CS and
US. Neurons in the hippocampus of rabbits robustly increase firing rates
upon the CS when it co-terminates with the US but not when the CS is
presented alone (Berger, Alger, & Thompson, 1976; Berger, Laham, &
Thompson, 1980; Berger & Thompson, 1978). As a group of neurons,
the temporal profile of firing responses closely models the amplitude
and time course of conditioned responses. These findings suggest that
the CS-evoked firing rate changes are selective for CS-US associations,
thereby providing further support for the role of the stimulus-evoked
firings in the hippocampus in representations of temporal expectation.
The hippocampus, however, appears to play this role even when two
events are not separated by an intervening temporal gap.
3.1.4. Striatum
During the conditional visuomotor tasks, some neurons in the pri-
mate striatum (putamen and caudate nucleus) show phasic responses to
the cue while others show tonic responses that persist into the cue-
response intervals (Brasted & Wise, 2004; Pasupathy & Miller, 2005;
Tremblay, Hollerman, & Schultz, 1998). The majority of neurons with
persistent firings do not differentiate tonic firing responses depending
on the direction of reaching response until immediately before monkeys
execute the correct response (Brasted & Wise, 2004). With learning, this
direction-selectivity comes to appear in an earlier part of the delay
period (Pasupathy & Miller, 2005), suggesting that striatum neurons
develop the selectivity for temporal expectation with learning.
Similar stimulus-evoked firing patterns are also observed while
rabbits receive trace eyeblink conditioning. Some putative medium
spiny neurons (MSNs) in the caudate nucleus increase firing rates
during the CS, the subsequent interval, or both (Flores & Disterhoft,
2009, 2013). The proportion of these CS-responding neurons is higher
in rabbits which received paired presentations of the CS and US than
those which received unpaired presentations. Similarly, during an ap-
petitive version of classical trace conditioning, some putative MSNs
differentially respond to the CS depending on whether it is predictive of
reward or not (Bakhurin, Mac, Golshani, & Masmanidis, 2016). Col-
lectively, these findings provide additional evidence that links MSN
persistent firings with temporal expectation.
In contrast, during interval timing tasks, MSNs show a ramping
increase or decrease of firing rates over 3–40 sec intervals (Emmons
et al., 2017; Matell et al., 2003). Like neurons in the mPFC (Emmons
5
Neurobiology of Learning and Memory xxx (xxxx) xxx–xxx
et al., 2017; Xu et al., 2014), the slope of ramping pattern changes
depending on the interval duration. Notably, the ramping firing
changes in MSNs are attenuated when the mPFC is pharmacologically
inactivated (Emmons et al., 2017), suggesting that the explicit timing
signal in the striatum depends on the integrity of the mPFC.
3.2. Phasic firing responses
3.2.1. Hippocampus
Past neurophysiological investigations of the rodent hippocampus
primarily focused on the characterization of neurons which fire at a
particular location within the environment (Hartley, Lever, Burgess, &
O’Keefe, 2014; Moser, Kropff, & Moser, 2008). In the past ten years, a
growing number of studies has characterized the selectivity of hippo-
campal neurons for the temporal information. In rat working memory
paradigms, some neurons in the hippocampus phasically fire during a
specific moment within the delay period between the sample and re-
sponse phases (MacDonald, Carrow, Place, & Eichenbaum, 2013;
Pastalkova, Itskov, Amarasingham, & Buzsáki, 2008; Salz et al., 2016).
MacDonald et al. (2011) extended the finding to time-linked memory
by using a delayed paired associate task, in which rats associate objects
with odor stimuli over a 10-sec interval. About half of individual neu-
rons in the CA1 region of the dorsal hippocampus show a phasic in-
crease in firing rates during a specific moment within the interval
period. Because the exact timing at which each of these neurons
changes the firing rate varies, the group of these “time cells” forms a
sequence firing pattern spanning the entire interval. The selectivity for
a specific moment within the interval is preserved even after removing
the influence of head direction, speed, distance traveled and rat’s po-
sition analytically (MacDonald et al., 2011) or experimentally (Kraus,
Robinson, White, Eichenbaum, & Hasselmo, 2013). When a different
object is presented, ∼31% of time cells change the magnitude or
temporal pattern of firing rates during the subsequent interval, sug-
gesting that these neurons retrospectively signal the identity of the
object presented a moment ago (MacDonald et al., 2011). In parallel,
when the duration of the interval is abruptly altered, about half of time
cells cease to fire or become active when they were previously inactive
(MacDonald et al., 2011). In parallel, a small proportion of time cells
(5%) shifts the temporal firing position relative to the new duration of
the interval while the remaining time cells do not change firing pat-
terns, suggesting that these neurons signal relative and absolute time
elapsed from interval onset, respectively. These patterns suggest that
the sequential firings of time cells represent trace holding and explicit
timing.
A question remains open as to whether hippocampal sequential
firing patterns also signal temporal expectation. In a task similar to the
one used in MacDonald et al. (2011), rats’ performance is significantly
attenuated when the medial entorhinal cortex is optogenetically in-
hibited for several seconds in the middle of the inter-event interval
(Robinson et al., 2017). The same manipulation disorganizes the se-
quential firing patterns in the CA1 region during the interval, but it
does not affect CA1 neuron selectivity for objects or locations within the
environment. Further evidence for the link between the hippocampal
sequential firing pattern and temporal expectation comes from firing
sequences of hippocampal neurons during trace eyeblink conditioning
(Modi, Dhawale, & Bhalla, 2014). The two-photon calcium imaging of
CA1 neurons shows that as immobilized mice associate an auditory
stimulus (CS) and an air-puff to the eye (US) over a 250-msec interval,
individual neurons become active for ∼100 msec at a various temporal
position within the interval. Prior to learning, individual neurons show
calcium responses at apparently random moments within the interval.
After learning, however, neurons come to reliably show calcium re-
sponses at fixed time-points within the CS-US interval. The enhanced
reliability of calcium responses is related to the CS-US association be-
cause it is not observed in mice which receive unpaired presentations of
the CS and US or those which do not acquire the CS-US association.
M. Pilkiw, K. Takehara-Nishiuchi
Collectively, these findings provide evidence that along with their role
in trace holding and explicit timing, hippocampal sequential firing
patterns also serve as a representation of temporal expectation.
3.2.2. Striatum
Phasic firing changes are also reported in the striatum during an
interval timing task, in which a reward becomes available 2.5 sec after
the offset of an odor stimulus (Bakhurin et al., 2017). Like hippocampal
neurons, some putative medium spiny neurons change firing rates
during a specific moment within the stimulus-reward interval and as a
group, they form a sequential firing pattern that covers the entire in-
terval. With a group of these neurons, machine learning classification
algorithms are able to decode time elapsed from the onset of the odor
stimulus. They are also able to decode when mice would start antici-
patory licking responses to receive the reward. These findings suggest
that phasic-firing responses in the striatum encode the temporal in-
formation of stimuli as well as behavioral expression of memory. Si-
milar patterns are also observed in a paradigm in which rats need to
estimate when the reward becomes available after they have collected
the reward in a previous trial (Mello, Soares, & Paton, 2015). When the
duration of the interval is altered within a range from 12 to 60 sec,
∼68% of striatum neurons maintain the selectivity for the temporal
position relative to the duration of the interval, thereby maintaining
their ordinal position in time across the neuron population. This result
contrasts with the very small proportion of hippocampal neurons se-
lective for relative timing (∼5%; MacDonald et al., 2011), raising a
possibility that the sequential firing patterns in the two regions re-
present explicit timing differently. Specifically, in the striatum, each
neuron maintains a stable ordinal position in time and scales the firing
duration relative to the duration of the interval. In the hippocampus, on
the other hand, the ordinal position of each neuron is flexible, resulting
in orthogonal sequential patterns for different interval duration.
4. Distributed representations of associations between temporally
discontiguous events in prefrontal neural ensembles
The across-study comparisons conducted in the previous section
illuminate some similarities and differences in the selectivity of tonic
and phasic firing patterns. Yet, the source of these differences is difficult
to decipher because these studies were conducted in different para-
digms and brain regions. By using a decoding approach with machine
learning algorithms, we here compare the selectivity for trace holding,
temporal expectation, and explicit timing between tonic and phasic
firing patterns within the same dataset recorded from the prelimbic
regions of the medial prefrontal cortex (mPFC) while rats received trace
eyeblink conditioning (Morrissey et al., 2017). This quantitative ana-
lysis uncovers that although both firing patterns are selective for all
three types of information, phasic firing patterns outperform tonic
firing patterns in trace holding and explicit timing.
4.1. Data analysis
The present study analyzed action potentials of individual neurons
recorded from the prelimbic region of the mPFC while four male Long-
Evans rats received two epochs of trace eyeblink conditioning
(Morrissey et al., 2017). Details of the materials and methods for data
acquisition are found in our previous publication with this dataset
(Morrissey et al., 2017). Briefly, daily recording sessions included two
epochs of trace eyeblink conditioning in which a neutral conditioned
stimulus (CS, 100 msec, auditory or visual stimulus) was paired with a
mild electric shock near the eyelid (US, 100 msec) which was presented
500 msec after the offset of the CS. In each epoch, the CS was presented
by itself during the first 20 trials (CS alone trials) while the CS was
paired with the US during the latter 80 trials (CS-US paired trials;
Fig. 1A). Intervals between trials were pseudorandomized between 20
and 40 sec. Over the course of two weeks of conditioning sessions, the
6
Neurobiology of Learning and Memory xxx (xxxx) xxx–xxx
rats increased the frequency of anticipatory blinking responses that
occurred immediately before the expected onset of the US (CR%,
Fig. 1B) in the block of CS-US paired trials, but not in the block of CS
alone trials. By using chronically implanted array of tetrodes, we col-
lected firing patterns of total 2077 neurons across four rats.
To screen neurons which sustained CS-evoked firing responses to
the subsequent interval, the difference in averaged firing rates during
CS-US intervals and pre-CS periods (900–400 ms before CS onset) was
calculated in each neuron. The same procedures were repeated 1000
times by randomly assigning each firing rate to either the CS-US or the
pre-CS period to estimate the distribution of firing difference at chance.
The firing rate difference with the correct period label was considered
as significant when it fell within the 5% upper tail of the chance dis-
tribution. We separately applied this analysis to firing patterns during
the auditory CS-US pairings and those during the visual CS-US pairings
to screen “persistent-firing” neurons in each trial block.
The above mentioned criteria would overlook neurons which tran-
siently changed firing rates during a specific phase of the CS-US inter-
vals due to the averaging of firing rates across the entire duration of CS-
US intervals. To screen these “phasic-firing” neurons, we quantified the
degree of firing rate changes within the CS-US interval by using mutual
information as a measure. In each neuron, firing rates during a CS-US
interval were binned into five 100-msec time bins in each trial. The
firing rate in each time bin across all trials was binned into 10 bins to
describe the probability distribution. Mutual information was calcu-
lated as follows:
P (i,j ) ⎞
Mutual information = ∑ P (i,j ) ∗log ⎛⎜ ⎟
i,j ⎝ P (i) P (j ) ⎠
where P(i,j) is the joint probability distribution of time bin “i” and firing
rate “j”, P(j)is the marginal probability distribution of firing rates,
averaged across time bins, and P(i) is the marginal distribution of firing
rate in time bin “i”. To assess the significance of selectivity, permutation
tests were performed for each neuron using the exact same procedure as
above, after assigning randomized labels to each time bin. This proce-
dure, repeated 1000 times, yielded the distribution of the chance level
of mutual information values. An observed mutual information value
with the correct time bin labels was considered as significant when it
fell in the 5% upper tail of its corresponding chance distribution. We
separately applied this analysis to firing patterns during the auditory
CS-US pairings and those during the visual CS-US pairings to screen
“phasic-firing” neurons in each trial block.
To quantify the selectivity of neuron ensembles for the identity of
the CS (trace holding), the CS-US relationship (temporal expectation),
and the temporal position within the CS-US interval (explicit timing),
we used a machine learning algorithm, a Support Vector Machine
(SVM) classifier (Chang & Lin, 2011; Cortes & Vapnik, 1995). SVM
classification was performed in MATLAB (Mathworks, Natick, MA,
USA) with the algorithms from the open source LIBSVM library (Chang
& Lin, 2011). The classifier constructed a model with the Gaussian ra-
dial basis function kernels. To maximize the decoding accuracy, two
SVM parameters, cost and gamma, were identified by performing a grid
search over a range of values, using the firing rates from all available
trials for a given target value. Then, a new population firing rate matrix
was generated from a subset of trials in the original population firing
matrix. These trials were randomly drawn, without replacement, from
all trials in a given trial block. Then, in each neuron, the firing rate in
each trial was divided by the maximum firing rate of the neuron among
all selected trials. All of the trials except one trial were then used to
train the SVM classifier with the parameters selected by the grid search.
This was done by using a 5-fold cross-validation procedure to minimize
over-fitting. The one trial was then used to test the decoding accuracy
after training. The accuracy was defined as a proportion of correct
predictions out of all test trials. Note that the neurons were recorded in
separate sessions from four rats, and thus we ignored any correlated
M. Pilkiw, K. Takehara-Nishiuchi Neurobiology of Learning and Memory xxx (xxxx) xxx–xxx

Fig. 1. Behavior and single neuron firing patterns during trace eyeblink conditioning. A. Daily recording sessions included two epochs of trace eyeblink conditioning
with either an auditory or visual stimulus as the conditioned stimulus (CS). In the first 20 trials, the CS was presented alone while in the latter 80 trials, the CS was
paired with mild electric shock near the eyelid (US). Two epochs took place in an identical conditioning chamber. B. Over the course of two weeks of daily
conditioning, four rats gradually increased the frequency of anticipatory blinking responses (CR%, mean ± s.e.m.) in trials of paired presentations of the CS and US
(red, blue) but not in trials of CS alone presentations (pink, cyan). C–E. Representative raster plots and peri-stimulus time histograms (1-ms bins, smoothed with a 50-
ms Hanning window) of firing responses to the auditory CS (red) or the visual CS (blue) during CS-US paired trials. Some neurons changed firing rates during the CS
and stably sustained the firing rate until the US was presented (C). Other neurons show a transient increase in firing rates during a specific phase of the CS-US interval
(D). Other neurons show ramping increase/decrease of firing rates over the CS-US interval (E).

7
M. Pilkiw, K. Takehara-Nishiuchi
activity between neurons.
To quantify the selectivity of ensemble firing patterns for the tem-
poral position within CS-US intervals, we followed the strategy used in
Bakhurin et al. (2017). Each run of multiclass SVM analysis used a set of
200 randomly subsampled cells (total 20 runs). We chose this sample
size because classification accuracy reaches asymptote with 200 neu-
rons or greater (Morrissey et al., 2017). Population firing rate vectors
were constructed by concatenating five 100-msec binned firing rates of
each neuron during CS-US intervals (200 neuron × 5 time bins × 80
trials). All trials except one trial were used as training trials (200
neurons × 5 time bins × 79 trials), and classifier was tested on the
remaining trial (200 neurons × 5 time bins × 1 trial). This procedure
yielded 400 predictions (5 time bins × 80 trials). To quantify the ac-
curacy of predictions, we calculated Pearson correlation coefficient (r)
between 400 predictions and their corresponding true temporal posi-
tions. These procedures were applied separately to four different
neuron types, phasic-firing neurons in the auditory CS-US pairings,
phasic-firing neurons in the visual CS-US pairings, persistent-firing
neurons in the auditory CS-US pairings, and persistent-firing neurons in
the visual CS-US pairings. Because the r values distributed normally
(Lilliefors test), their difference between phasic- and persistent-firing
neurons were examined with t-test.
To quantify the selectivity of ensemble firing patterns for the CS-US
relationships, we randomly selected 20 trials out of 80 CS-US paired
trials and constructed a population firing rate matrix by concatenating
five 100-msec binned firing rates of each neuron during CS-US intervals
(200 neuron × 5 time bins × 20 trials). We also constructed a popu-
lation firing rate matrix of five 100-msec firing rates of the same set of
neurons during CS alone trials (200 neurons × 5 time bins × 20 trials).
To quantify the selectivity for CS-US relationships, SVM classifiers were
trained to discriminate CS-US paired trials from corresponding CS alone
trials. To quantify the selectivity of ensemble firing patterns for the
modality of the CS, SVM classifiers were trained to discriminate audi-
tory CS-US paired trials from visual CS-US paired trials. In both cases,
all trials except one trial were used as training trials (200 neurons × 5
time bins × 39 trials), and classifier was tested on the remaining trial
(200 neurons × 5 time bins × 1 trial). This procedure yielded 200
predictions (5 time bins × 40 trials). This procedure was repeated
twenty times, each of which used 200 sub-sampled neurons. Because
classification accuracy was not distributed normally in most cases
(Lilliefors test), we used the Wilcoxon signed-rank test to test whether
classification accuracy was better than chance (i.e., 50%). Also, the
Wilcoxon ranked sum test was used to test whether classification ac-
curacy significantly differed between persistent- and phasic-firing
neurons as well as between the CS identity and CS-US relationship.
4.2. Results
4.2.1. Firing profiles of single neurons
Among 2077 neurons recorded across four rats, 1248 neurons
(60.1%) were classified as having significant firing rate difference
during the CS and subsequent interval relative to those during the
period before CS onset (Table 1). The proportion of these “persistent-
Table 1
Number and percentage of CS-selective neurons with persistent and phasic
firing patterns. Table summarizes the number of neurons with persistent and
phasic firing responses in auditory (A), visual (V) CS-US paired trial, or both.
The values in parentheses show the percentage of neurons to the total number
of recorded neurons.
ACS-US VCS-US Both Total recorded
neurons
Persistent 1034 922 (44%) 708 (34%) 2077
neurons (50%)
Phasic neurons 243 (12%) 268 (13%) 62 (3%) 2077
8
Neurobiology of Learning and Memory xxx (xxxx) xxx–xxx
firing” neurons was comparable between the ACS-US paired block and
the VCS-US block, and ∼34% of them passed the criteria for both of the
two CS. These neurons maintained CS-evoked firing increase or de-
crease stably until US onset (Neurons 1–4; Fig. 1C).
In parallel, 449 neurons (21.6%) were classified as significantly
changing firing rates across five time bins during the CS-US interval
(Table 1). Most of these “phasic-firing” cells showed bi-phasic firing
rate changes, with an initial dip followed by a sharp, transient increase
of firing rates (Fig. 1D). Some neurons increased firing rates im-
mediately after the offset of the CS (Neurons 5, 9) while others in-
creased firing rates in the middle of CS-US intervals (Neurons 6, 10, 11).
Others showed a sharp increase of the firing rate toward the expected
onset of the US (Neurons 7, 8, 12). The proportion of these neurons with
temporally modulated firing patterns was comparable between the
ACS-US paired block and the VCS-US block; however, most of these
cells were judged to be selective during either one of the CS types
(Table 1).
A small proportion of neurons (∼7%) passed the criterion for the
persistent and phasic firings. These neurons showed a steep ramping of
firing rate decrease (Neurons 13, 14) or increase (Neurons 15, 16)
during CS-US intervals. In a subsequent decoding analysis, these neu-
rons were included to both persistent- and phasic-firing neurons. This
allowed for increasing the number of phasic-firing neurons usable for
the decoding analysis, which made it reach the minimum number of
neurons required for stable decoding performance (Morrissey et al.,
2017).
4.2.2. Selectivity of neuron ensembles for CS identity, CS-US association,
and temporal position within CS-US intervals
Similar to time cells in the hippocampus (MacDonald et al., 2011;
Modi et al., 2014), each phasic-firing cell reached the peak firing rate at
various time points spanning across the entire CS-US interval (Fig. 2A,
left). On the other hand, most persistent-firing cells reached the peak
firing rate either during the CS or the first 100-msec period immediately
after CS offset (Fig. 2B, left). In both cases, the order of the peak firing
timing became disorganized during the corresponding interval when
the same CS was presented alone (middle) and during the interval when
the other CS was paired with the US (right).
To test whether the differentiation of sequential ensemble firing
patterns between the trial types was robust at the single-trial level, we
used a Support Vector Machine (SVM) classifier to decode the temporal
position within the trace interval, the identity of the CS (ACS-US or
VCS-US paired trials), or stimulus relationships (CS-US paired or CS
alone trials) from binned ensemble firing patterns. With a group of
phasic-firing neurons, classifiers were able to decode the first bin almost
perfectly, whereas the accuracy dropped in the subsequent bins
(Fig. 2C, left). Most inaccurate classifications were due to mis-
classifications of an adjacent time bin, suggesting that ensemble firing
patterns were similar between two time points closest in time. In con-
trast, with a group of persistent-firing neurons, misclassifications spread
across time bins regardless of temporal positions of the bins (Fig. 2C,
right), suggesting weaker relationships between ensemble similarity
and the temporal proximity of the time bins. In both phasic- and per-
sistent-firing neurons, the predicted time bin was positively correlated
with the correct time bin (Fig. 2D). The degree of correlation, however,
was significantly higher in phasic-firing neurons than in persistent-
firing neurons (t test, ps < 0.001, in both ACS-US and VCS-US trials, 20
SVM runs with 200 randomly subsampled neurons). These results
suggest that phasic-firing neuron ensembles showed stronger selectivity
for the temporal position within the CS-US interval than persistent-
firing neuron ensembles.
We then compared the selectivity of two neuron types for CS
identity and CS-US relationship. In both phasic- and persistent-firing
neurons, classification accuracy for the stimulus identity and relation-
ship was significantly better than chance (i.e., 50%) in all five 100-msec
bins during the ACS-US and VCS-US intervals (Fig. 2E; Wilcoxon signed-
M. Pilkiw, K. Takehara-Nishiuchi Neurobiology of Learning and Memory xxx (xxxx) xxx–xxx

(caption on next page)

9
M. Pilkiw, K. Takehara-Nishiuchi Neurobiology of Learning and Memory xxx (xxxx) xxx–xxx

Fig. 2. Ensemble firing patterns and their selectivity for task variables and temporal positions within CS-US intervals. A. The firing patterns of all identified neurons
with phasic-firing responses during the visual CS-US paired trials. Each row represents the normalized firing rate of one neuron over the duration from the onset of
the visual CS to the onset of the US, and the order of neurons presented is determined by the timing at which each neuron reached the maximum firing rate (peak
time, left). The same order was used to depict firing patterns of the same set of neurons during visual CS alone trials (middle) and auditory CS-US paired trials (right).
B. The firing patterns of all identified neurons with persistent-firing responses during the visual CS-US paired trials, as shown in A. C. Confusion matrices visualizing
SVM classifiers’ performance on the decoding of the temporal position within the CS-US interval from ensembles of phasic- (left) and persistent-firing neurons (right).
Each row represents the proportion of predictions for firing patterns during the corresponding correct time bins (y-axis). D. Correlation coefficients (mean ± s.e.m.)
between predicted and correct time bins across 20 classification runs with 200 randomly subsampled neurons. ***, p < 0.001, t-test. E. Classification accuracy for
binary discrimination between CS-US paired trials versus CS along trials (red, blue) as well as between auditory versus visual CS trials (pink, cyan). Each dot displays
classification accuracy observed in one of 20 classification runs with 200 randomly subsampled neurons, and lines connect their median. ***, **, *, p < 0.001, 0.01,
0.05, between two classification types, Wilcoxon Rank-Sum test.

rank test, α = 0.05/15 comparisons, all ps < 0.001). In phasic-firing
neurons, classification accuracy for stimulus identity was greater than
that for stimulus relationship during the first 100-msec period after CS
offset (Wilcoxon Rank-Sum test, ACS-US and VCS-US paired trials,
ps < 0.002). The pattern became reversed in the middle of the CS-US
interval, and during the last bin, classification accuracy for the CS-US
relationship became greater than that for the CS identity in ACS-US
trials (p = 0.0004). A similar trend was found during VCS-US trials
(p = 0.027). Such time-dependent changes in the selectivity were not
observed in persistent-firing neurons (all ps > 0.003). The comparison
of the selectivity between two neuron types revealed that during the
first 100-msec bin, the selectivity for the CS modality was higher in
phasic-firing neurons than persistent-firing neurons (Wilcoxon Rank-
Sum test, p < 0.001 in ACS-US trials, p = 0.002 in VCS-US trials)
while the selectivity for the CS-US relationship was comparable be-
tween the two neuron types in all time bins (ps > 0.003). These results
suggest that phasic-firing neuron ensembles showed stronger selectivity
for the identity of the CS during the early phase of the trace interval
than persistent-firing neuron ensembles while the selectivity for the CS-
US relationship was comparable between the two groups of neurons.
4.3. Discussion
By applying two different screening criteria to the group of pre-
frontal neurons, we identified two largely non-overlapping populations
of neurons, one with persistent-firing responses and the other with
phasic-firing responses during the intervals between the CS and US
(trace intervals) in trace eyeblink conditioning. Similar to time cells in
the hippocampus (MacDonald et al., 2013, 2011; Modi et al., 2014;
Pastalkova et al., 2008; Salz et al., 2016), phasic-firing cells reached the
maximum firing rate at various time points during the trace interval,
thereby forming a sequential firing pattern that covered the entire
duration of the CS-US interval. These neurons formed a different se-
quential firing pattern when the same stimulus was presented by itself
and when a different stimulus was paired with the same punishment.
These findings suggest that in parallel to neurons showing the well-
established persistent firing patterns (Hattori et al., 2014; Siegel &
Mauk, 2013; Siegel et al., 2012; Takehara-Nishiuchi & McNaughton,
2008), other neurons in the mPFC show phasic firings selective for a
specific moment during the interval between temporally discontiguous
stimuli.
To compare the contribution of phasic- and persistent-firing neurons
to trace holding, temporal expectation, and explicit timing, we em-
ployed SVM classification analysis to quantify their selectivity for the
identity of the CS, the CS-US relationship, and the temporal position
within the trace interval. The strong selectivity for these features was
found in both persistent- and phasic-firing neurons, suggesting that
both of these neuron types contribute to trace holding, temporal ex-
pectation, and explicit timing. Phasic-firing neurons showed stronger
selectivity for the temporal position than persistent-firing neurons,
which was expected due to the criteria used for selecting phasic-firing
neurons (i.e., significant firing rate variations across five time bins
during the CS-US interval). This observation is also consistent with the
successful decoding of the elapsed time from sequential firing patterns
of neurons in the orbitofrontal cortex and striatum during the interval
timing task with a short interval (1.5 sec, Bakhurin et al., 2017). On the
other hand, another firing pattern associated with explicit timing
during interval timing tasks, ramping firings (Emmons et al., 2017;
Matell et al., 2003) were observed in less than 10% of neurons in our
dataset (i.e., neurons passing the criteria for both persistent and phasic
firings). This may be due to the difference in the duration of time in-
tervals between trace eyeblink conditioning (< 1 sec) and interval
timing tasks used in these studies (from 3 to 20 sec), raising a possibility
that the mPFC neurons may use different firing profiles to encode in-
tervals of milliseconds from those of seconds.
A strong selectivity for the CS identity and the upcoming US was
observed in both persistent-firing and phasic-firing neurons, and the
degree of selectivity was mostly comparable between the two groups of
neurons. Notably, phasic-firing neurons differed from persistent-firing
neurons in that they changed the relative strength of selectivity for
these features during the CS-US interval. Specifically, during the early
phase of the CS-US interval, phasic-firing neurons showed greater se-
lectivity for the identity of the CS than the upcoming US while their
selectivity for the upcoming US became stronger towards the end of
trace interval. Although a similar shift in the selectivity was reported in
persistent-firing patterns in the lateral prefrontal cortex in monkeys
during a delayed paired associate task (Rainer et al., 1999), persistent-
firing neurons in our dataset did not show this coding property. The two
studies differ in many aspects, including the difference in the duration
of the time interval between the task used (0.5 msec versus 1 sec), the
brain region, and species. Therefore, future studies are needed to fur-
ther examine the time-dependent shift in the selectivity of neuron ac-
tivity during inter-event intervals.
Collectively, these results suggest that neurons in the medial pre-
frontal cortex, regardless of their firing profiles, play a role in all three
critical processes underlying the association formation between sti-
mulus and punishment over a temporal gap. Phasic-firing patterns,
however, showed a better selectivity for the task features important for
trace holding and explicit timing, raising the possibility that for the
encoding of short time intervals (less than one second), population
coding in phasic-firing neuron ensembles may have a computational
advantage over rate coding in individual persistent-firing neurons.
5. Concluding remarks
Memories of daily experiences consist of complex spatial and tem-
poral sequences of events. Compared to well-documented neural re-
presentations of the spatial information (Hartley et al., 2014; Moser
et al., 2008), our understanding of neural representations of the tem-
poral information is still developing. The present review aimed to
provide a concise summary of single neuron firing patterns observed
while rodents or non-human primates associate stimuli, outcome, and
responses across a temporal gap. Our survey included careful con-
sideration of the experimental design used in each study which allowed
for deciphering the selectivity of each neuron firing pattern for the
content of preceding events (trace holding), the content of following
events (temporal expectation), and the temporal position within inter-
event intervals (explicit timing). Across the various behavioral

10
M. Pilkiw, K. Takehara-Nishiuchi Neurobiology of Learning and Memory xxx (xxxx) xxx–xxx

Table 2
Neuron firing patterns and brain regions supporting trace holding, temporal expectation and explicit timing in time-linked memory tasks. Task refers to a particular
paradigm used in each study. Duration refers to the delay between two stimuli or stimulus-response used in each task.
Pattern Task Duration (sec) Species Region Citation

Trace holding Persistent Visuomotor learning 0.5 Monkey Lateral prefrontal Asaad et al. (1998)
Phasic Paired associate 10 Rat Hippocampus MacDonald et al. (2011)
Persistent Trace eyeblink 0.5 Rat Medial prefrontal Morrissey et al. (2017)
Persistent Paired associate 2 Monkey Area 36 Naya et al. (2001)
Persistent Paired associate 2 Monkey Area TE Naya et al. (2001)
Persistent Paired associate 2 Monkey Area 36 Naya et al. (2003)
Persistent Paired associate 2 Monkey Area TE Naya et al. (2003)
Persistent Paired associate 1 Monkey Lateral prefrontal Rainer et al. (1999)
Persistent Paired associate 4 Monkey Inferior temporal Sakai and Miyashita (1991)
Persistent Visuomotor learning 0.7 Monkey Hippocampus Wirth et al. (2003)
Persistent Visuomotor learning 0.7 Monkey Hippocampus Yanike et al. (2004)
Persistent Visuomotor learning 0.7 Monkey Perirhinal Yanike et al. (2009)

Temporal expectation Persistent Visuomotor learning 0.5 Monkey Lateral prefrontal Asaad et al. (1998)
Persistent Trace conditioning 1.5 Mouse Striatum Bakhurin et al. (2016)
Persistent Trace eyeblink 0.5 Rabbit Striatum Flores and Disterhoft (2009)
Persistent Trace eyeblink 0.5 Rabbit Striatum Flores and Disterhoft (2013)
Persistent Trace fear 20 Rat Medial prefrontal Gilmartin and McEchron (2005)
Persistent Trace eyeblink 0.5 Rabbit Medial prefrontal Hattori et al. (2014)
Phasic Paired associate 10 Rat Hippocampus MacDonald et al. (2011)
Persistent Trace eyeblink 0.5 Rabbit Hippocampus McEchron and Disterhoft (1997)
Persistent Trace eyeblink 0.5 Rabbit Hippocampus McEchron et al. (2001)
Phasic Trace eyeblink 0.25 Mouse Hippocampus Modi et al. (2014)
Persistent Trace eyeblink 0.5 Rat Medial prefrontal Morrissey et al. (2017)
Persistent Paired associate 2 Monkey Area 36 Naya et al. (2001)
Persistent Paired associate 2 Monkey Area TE Naya et al. (2001)
Persistent Paired associate 2 Monkey Area 36 Naya et al. (2003)
Persistent Paired associate 2 Monkey Area TE Naya et al. (2003)
Persistent Visuomotor learning 1 Monkey Dorsolateral prefrontal Pasupathy and Miller (2005)
Persistent Visuomotor learning 1 Monkey Striatum Pasupathy and Miller (2005)
Persistent Paired associate 1 Monkey Lateral prefrontal Rainer et al. (1999)
Persistent Paired associate 4 Monkey Inferior temporal Sakai and Miyashita (1991)
Persistent Trace eyeblink 0.5 Rat Medial prefrontal Takehara-Nishiuchi and McNaughton (2008)

Explicit timing Phasic Interval timing 1.5 Mouse Striatum Bakhurin et al. (2017
Phasic Interval timing 1.5 Mouse Orbitofrontal Bakhurin et al. (2017)
Persistent Interval timing 3, 12 Rat Medial prefrontal Emmons et al. (2017)
Persistent Interval timing 3, 12 Rat Dorsal striatum Emmons et al. (2017)
Phasic Paired associate 10 Rat Hippocampus MacDonald et al. (2011)
Persistent Interval timing 10, 40 Rat Striatum Matell et al. (2003)
Persistent Interval timing 10, 40 Rat Medial prefrontal Matell et al. (2003)
Phasic Interval timing 12, 24, 36, 48, 60 Mouse Striatum Mello et al. (2015)
Phasic Trace eyeblink 0.25 Mouse Hippocampus Modi et al. (2014)
Persistent Visuomotor learning 0.7 Monkey Hippocampus Sakon et al. (2014)
Persistent Interval timing 1.5, 2.5 Rat Medial prefrontal Xu et al. (2014)

paradigms used in neurophysiological studies, two distinct firing pat-
terns, tonic and phasic firings, were observed during intervals separ-
ating two events (Table 2). Tonic firing patterns in the prefrontal, in-
ferior temporal cortices, and hippocampus are selective for the content
of preceding and following events, suggesting their role in representa-
tions of trace holding and temporal expectation. This coding property is
observed in paradigms in which the duration of the interval ranging
from 0.5 sec to a few seconds. In parallel, tonic firings in the primate
hippocampus, rodent medial prefrontal cortex, and striatum signal ex-
plicit timing by scaling the slope of firing rate changes over inter-event
intervals depending on their duration. The coding property was ob-
served during the interval ranging from 1.5 to 12 sec.
In parallel, a group of neurons with phasic firings form a sequential
firing pattern spanning the entire interval ranging from 0.25 to 10 sec.
Sequential firing patterns in the hippocampus form orthogonal patterns
selective for the identity of preceding and following events as well as
the duration of the interval. In contrast, sequential firing patterns in the
striatum encode the different duration of intervals by scaling each
neuron’s firing duration while maintaining the ordinal position of its
firing timing within the intervals. In the medial prefrontal cortex, the
sequential firing pattern is observed in parallel to persistent firings, and
they outperform persistent firings in the selectivity for explicit timing
and trace holding, but not for temporal expectation.
The survey highlights remarkable similarity among neural firing
patterns observed during the inter-event interval whose duration ranges
from milliseconds to seconds. In parallel, it also illuminates some dif-
ferences in the selectivity for trace holding, temporal expectation, and
explicit timing between firing patterns within each brain region as well
as between brain regions. Below we list several unresolved points that
need to be addressed in future studies.
1. Potential functional differences between tonic- and phasic-firing
patterns need further investigation. Our discussion regarding this
point is limited because most studies only report neurons with either
of these firing patterns. This is likely due to the typical step of neural
activity analysis, in which investigators construct a screening cri-
terion for neurons with a specific firing pattern of interest. Given the
co-existence of tonic- and phasic-firing neurons in some brain re-
gions (see, Sections 3 and 4), it is important to search for neurons
with various firing patterns by applying several screening criteria to
a single dataset. This allows for directly comparing the selectivity
for various task features between neurons with different firing
profiles, leading to a richer description of neural representations of
the temporal information (see Diehl, Hon, Leutgeb, & Leutgeb, 2017
for successful application of this approach to spatial representations
in the medial entorhinal cortex).

11
M. Pilkiw, K. Takehara-Nishiuchi
2. Another point that requires further investigation is the influence of
motor output on neuron firing patterns. In trace conditioning, ani-
mals initiate behavioral responses in the middle of intervals between
the stimulus and outcome. In fact, in some cases, the temporal
profile of firing patterns appears to model the temporal pattern of
behavior (e.g., Fig. 1D). Even in conditional visuomotor learning and
interval timing tasks, in which animals are trained to withhold any
behavioral responses during the interval, they may internally plan
the action (Namboodiri & Shuler, 2014) and even change their body
position to a very small degree which cannot be detected without
close monitoring (Cowen & McNaughton, 2007). Uncovering whe-
ther observed neural responses are cause or consequence of beha-
vioral manifestation of memory is notoriously difficult: neuron
firing patterns selective for memory should be correlated with
movement reporting memory expression. One case that successfully
tackled this challenge is an elegant study by Siegel and Mauk
(2013). By recording single neuron firings from the medial pre-
frontal cortex while rabbits received trace eyeblink conditioning,
they found that persistent firing patterns in some, but not all, pre-
frontal neurons are abolished with the inactivation of the cere-
bellum which supresses the motor output (i.e., conditioned eyelid
responses). Thus, although some persistent firing patterns are an
efferent copy of eyelid movement, others are likely related to tem-
poral stimulus association formed as a result of the conditioning.
Applying a similar approach to the other paradigms holds the pro-
mise of disentangling memory-related firing patterns from motor-
related firing patterns.
3. Lastly, potential differences in the roles that each brain region plays
for time-linked memory call for further investigations. As reviewed
in Section 2, all behavioral paradigms assessing time-linked memory
depend on an interconnected network including many brain regions.
By comparing neuron firing patterns between these brain regions
during the same memory paradigm would provide valuable insight
into the difference in the selectivity for task features between the
brain regions (e.g., Bakhurin et al., 2017; Emmons et al., 2017;
Hattori et al., 2014; Yanike et al., 2009). Moreover, the combination
of electrophysiological recordings from one brain region with ma-
nipulations of another brain region is essential for uncovering the
influence that the latter brain region has on neural firing patterns in
the former brain region. Reversible inactivation (e.g., Emmons et al.,
2017) or optogenetic inhibition during a specific phase of the task
(e.g., Robinson et al., 2017) would effectively identify the source of
specific information represented in the neural firing patterns.
In summary, convergent evidence from behavioral and neurophy-
siological studies points to distributed neural representations of tem-
poral organization of events that we encounter during everyday life. In
various brain regions involved in this important aspect of everyday
memory, neurons show tonic and phasic firing patterns during time
intervals separating two events. Depending on the brain regions, these
firing patterns are selective for the prior event, expectation, or timing of
the coming event with different strength. Future studies are necessary
to further characterize the difference in the selectivity between two
firing patterns within each brain region as well as between brain re-
gions. These investigations, along with those directed at neural re-
presentations of time in more flexible, dynamic settings, such as social
interaction, communication, and language (Finnerty, Shadlen, Jazayeri,
Nobre, & Buonomano, 2015; Goel & Buonomano, 2014), would further
expand our understanding of how the brain processes time.
Acknowledgements
This work was supported by NSERC Discovery Grant (RGPIN-2015-
05458), CIHR Operating Grant (MOP-133693), Ontario Research Fund
Early Researcher Award (ER 13-09-01), CFI Leaders Opportunity Fund
(25026) to KT, NSERC graduate fellowship (396157093) to MP. The
12
Neurobiology of Learning and Memory xxx (xxxx) xxx–xxx
authors thank Dr. Mark Morrissey for collecting single neurons firing
patterns used for the data analysis in Section 4.
References
Asaad, W. F., Rainer, G., & Miller, E. K. (1998). Neural activity in the primate prefrontal
cortex during associative learning. Neuron, 21, 1399–1407.
Baeg, E. H., Kim, Y. B., Jang, J., Kim, H. T., Mook-Jung, I., & Jung, M. W. (2001). Fast
spiking and regular spiking neural correlates of fear conditioning in the medial
prefrontal cortex of the rat. Cerebral Cortex, 11, 441–451.
Bakhurin, K. I., Goudar, V., Shobe, J. L., Claar, L. D., Buonomano, D. V., & Masmanidis, S.
C. (2017). Differential encoding of time by prefrontal and striatal network dynamics.
Journal of Neuroscience, 37(4), 854–870.
Bakhurin, K. I., Mac, V., Golshani, P., & Masmanidis, S. C. (2016). Temporal correlations
among functionally specialized striatal neural ensembles in reward-conditioned mice.
Journal of Neurophysiology, 115, 1521–1532.
Bangasser, D. A., Waxler, D. E., Santollo, J., & Shors, T. J. (2006). Trace conditioning and
the hippocampus: The importance of contiguity. Journal of Neuroscience, 26,
8702–8706.
Berger, T. W., Alger, B., & Thompson, R. F. (1976). Neuronal substrate of classical con-
ditioning in the hippocampus. Science, 192, 483–485.
Berger, T. W., Laham, R. I., & Thompson, R. F. (1980). Hippocampal unit-behavior cor-
relations during classical conditioning. Brain Research, 193, 229–248.
Berger, T. W., & Orr, W. B. (1983). Hippocampectomy selectively disrupts discrimination
reversal conditioning of the rabbit nictitating membrane response. Behavioural Brain
Research, 8, 49–68.
Berger, T. W., & Thompson, R. F. (1978). Neuronal plasticity in the limbic system during
classical conditioning of the rabbit nictitating membrane response. I. The hippo-
campus. Brain Research, 145, 323–346.
Beylin, A. V., Gandhi, C. C., Wood, G. E., Talk, A. C., Matzel, L. D., & Shors, T. J. (2001).
The role of the hippocampus in trace conditioning: Temporal discontinuity or task
difficulty? Neurobiology of Learning and Memory, 76, 447–461.
Brasted, P. J., & Wise, S. P. (2004). Comparison of learning-related neuronal activity in
the dorsal premotor cortex and striatum. European Journal of Neuroscience, 19,
721–740.
Bunsey, M., & Eichenbaum, H. (1993). Critical role of the parahippocampal region for
paired-associate learning in rats. Behavioral Neuroscience, 107, 740.
Bunsey, M., & Eichenbaum, H. (1996). Conservation of hippocampal memory function in
rats and humans. Nature, 379, 255–257.
Burman, M. A., Starr, M. J., & Gewirtz, J. C. (2006). Dissociable effects of hippocampus
lesions on expression of fear and trace fear conditioning memories in rats.
Hippocampus, 16, 103–113.
Chang, C.-C., & Lin, C.-J. (2011). LIBSVM: A Library for Support Vector Machines. ACM
Transactions on Intelligent Systems and Technology, 2 27:1–27:27.
Chowdhury, N., Quinn, J. J., & Fanselow, M. S. (2005). Dorsal hippocampus involvement
in trace fear conditioning with long, but not short, trace intervals in mice. Behavioral
Neuroscience, 119, 1396.
Cortes, C., & Vapnik, V. (1995). Support-vector networks. Machine Learning, 20, 273–297.
Coull, J. T., & Nobre, A. C. (2008). Dissociating explicit timing from temporal expectation
with fMRI. Current Opinion in Neurobiology, 18, 137–144.
Cowen, S. L., & McNaughton, B. L. (2007). Selective delay activity in the medial pre-
frontal cortex of the rat: Contribution of sensorimotor information and contingency.
Journal of Neurophysiology, 98, 303–316.
Diehl, G. W., Hon, O. J., Leutgeb, S., & Leutgeb, J. K. (2017). Grid and nongrid cells in
medial entorhinal cortex represent spatial location and environmental features with
complementary coding schemes. Neuron, 94, 83–92.
Eacott, M. J., & Gaffan, D. (1992). Inferotemporal-frontal disconnection: The uncinate
fascicle and visual associative learning in monkeys. European Journal of Neuroscience,
4, 1320–1332.
Emmons, E. B., De Corte, B. J., Kim, Y., Parker, K. L., Matell, M. S., & Narayanan, N. S.
(2017). Rodent medial frontal control of temporal processing in the dorsomedial
striatum. Journal of Neuroscience, 37, 8718–8733.
Erickson, C. A., & Desimone, R. (1999). Responses of macaque perirhinal neurons during
and after visual stimulus association learning. Journal of Neuroscience, 19,
10404–10416.
Esclassan, F., Coutureau, E., Di Scala, G., & Marchand, A. R. (2009). A cholinergic-de-
pendent role for the entorhinal cortex in trace fear conditioning. Journal of
Neuroscience, 29, 8087–8093.
Finnerty, G. T., Shadlen, M. N., Jazayeri, M., Nobre, A. C., & Buonomano, D. V. (2015).
Time in cortical circuits. Journal of Neuroscience, 35, 13912–13916.
Flores, L. C., & Disterhoft, J. F. (2009). Caudate nucleus is critically involved in trace
eyeblink conditioning. Journal of Neuroscience, 29, 14511–14520.
Flores, L. C., & Disterhoft, J. F. (2013). Caudate nucleus in retrieval of trace eyeblink
conditioning after consolidation. Journal of Neuroscience, 33, 2828–2836.
Fujimichi, R., Naya, Y., Koyano, K. W., Takeda, M., Takeuchi, D., & Miyashita, Y. (2010).
Unitized representation of paired objects in area 35 of the macaque perirhinal cortex.
European Journal of Neuroscience, 32, 659–667.
Fuster, J. M., & Alexander, G. E. (1971). Neuron activity related to short-term memory.
Science, 173, 652–654.
Gaffan, D., & Harrison, S. (1988). Inferotemporal-frontal disconnection and fornix
transection in visuomotor conditional learning by monkeys. Behavioural Brain
Research, 31, 149–163.
Gilbert, P. E., & Kesner, R. P. (2003). Localization of function within the dorsal hippo-
campus: The role of the CA3 subregion in paired-associate learning. Behavioral
M. Pilkiw, K. Takehara-Nishiuchi
Neuroscience, 117, 1385.
Gilmartin, M. R., & Helmstetter, F. J. (2010). Trace and contextual fear conditioning
require neural activity and NMDA receptor-dependent transmission in the medial
prefrontal cortex. Learning and Memory, 17, 289–296.
Gilmartin, M. R., Kwapis, J. L., & Helmstetter, F. J. (2012). Trace and contextual fear
conditioning are impaired following unilateral microinjection of muscimol in the
ventral hippocampus or amygdala, but not the medial prefrontal cortex. Neurobiology
of Learning and Memory, 97, 452–464.
Gilmartin, M. R., & McEchron, M. D. (2005). Single neurons in the medial prefrontal
cortex of the rat exhibit tonic and phasic coding during trace fear conditioning.
Behavioral Neuroscience, 119, 1496–1510.
Gilmartin, M. R., Miyawaki, H., Helmstetter, F. J., & Diba, K. (2013). Prefrontal activity
links nonoverlapping events in memory. Journal of Neuroscience, 33, 10910–10914.
Goel, A., & Buonomano, D. V. (2014). Timing as an intrinsic property of neural networks:
Evidence from in vivo and in vitro experiments. Philosophical Transactions of the Royal
Society B, 369, 20120460.
Graves, C. A., & Solomon, P. R. (1985). Age-related disruption of trace but not delay
classical conditioning of the rabbit’s nictitating membrane response. Behavioral
Neuroscience, 99, 88.
Han, C. J., O’Tuathaigh, C. M., van Trigt, L., Quinn, J. J., Fanselow, M. S., Mongeau, R., ...
Anderson, D. J. (2003). Trace but not delay fear conditioning requires attention and
the anterior cingulate cortex. Proceedings of the National Academy of Sciences, 100,
13087–13092.
Hartley, T., Lever, C., Burgess, N., & O’Keefe, J. (2014). Space in the brain: How the
hippocampal formation supports spatial cognition. Philosophical Transactions of the
Royal Society B, 369, 20120510.
Hattori, S., Chen, L., Weiss, C., & Disterhoft, J. F. (2015). Robust hippocampal re-
sponsivity during retrieval of consolidated associative memory: Hippocampal
Involvement after Memory Consolidation. Hippocampus, 25, 655–669.
Hattori, S., Yoon, T., Disterhoft, J. F., & Weiss, C. (2014). Functional reorganization of a
prefrontal cortical network mediating consolidation of trace eyeblink conditioning.
Journal of Neuroscience, 34, 1432–1445.
Higuchi, S.-I., & Miyashita, Y. (1996). Formation of mnemonic neuronal responses to
visual paired associates in inferotemporal cortex is impaired by perirhinal and en-
torhinal lesions. Proceedings of the National Academy of Sciences, 93, 739–743.
Hoehler, F. K., & Thompson, R. F. (1980). Effect of the interstimulus (CS–UCS) interval on
hippocampal unit activity during classical conditioning of the nictitating membrane
response of the rabbit (Oryctolagus cuniculus). Journal of Comparative and
Physiological Psychology, 94, 201.
Hu, B., Chen, H., Feng, H., Zeng, Y., Yang, L., Fan, Z. L., ... Sui, J. F. (2010). Disrupted
topography of the acquired trace-conditioned eyeblink responses in guinea pigs after
suppression of cerebellar cortical inhibition to the interpositus nucleus. Brain
Research, 1337, 41–55.
Ivry, R. B., & Spencer, R. M. (2004). The neural representation of time. Current Opinion in
Neurobiology, 14, 225–232.
Kamin, L. J., & Schaub, R. E. (1963). Effects of conditioned stimulus intensity on the
conditioned emotional response. Journal of Comparative and Physiological Psychology,
56, 502.
Kesner, R. P., Hunsaker, M. R., & Gilbert, P. E. (2005). The role of CA1 in the acquisition
of an object-trace-odor paired associate task. Behavioral Neuroscience, 119, 781.
Kholodar-Smith, D. B., Boguszewski, P., & Brown, T. H. (2008). Auditory trace fear
conditioning requires perirhinal cortex. Neurobiology of Learning and Memory, 90,
537–543.
Kishimoto, Y., Kawahara, S., Mori, H., Mishina, M., & Kirino, Y. (2001). Long-trace in-
terval eyeblink conditioning is impaired in mutant mice lacking the NMDA receptor
subunit ε1: Eyeblink conditioning in GluRε1 mutant mice. European Journal of
Neuroscience, 13, 1221–1227.
Kitamura, T., Pignatelli, M., Suh, J., Kohara, K., Yoshiki, A., Abe, K., & Tonegawa, S.
(2014). Island cells control temporal association memory. Science, 343, 896–901.
Kraus, B. J., Robinson, R. J., White, J. A., Eichenbaum, H., & Hasselmo, M. E. (2013).
Hippocampal “time cells”: Time versus path integration. Neuron, 78, 1090–1101.
Kronforst-Collins, M. A., & Disterhoft, J. F. (1998). Lesions of the caudal area of rabbit
medial prefrontal cortex impair trace eyeblink conditioning. Neurobiology of Learning
and Memory, 69, 147–162.
Kwapis, J. L., Jarome, T. J., Schiff, J. C., & Helmstetter, F. J. (2011). Memory con-
solidation in both trace and delay fear conditioning is disrupted by intra-amygdala
infusion of the protein synthesis inhibitor anisomycin. Learning and Memory, 18,
728–732.
LeDoux, J. E., Iwata, J., Cicchetti, P., & Reis, D. J. (1988). Different projections of the
central amygdaloid nucleus mediate autonomic and behavioral correlates of condi-
tioned fear. Journal of Neuroscience, 8, 2517–2529.
MacDonald, C. J., Carrow, S., Place, R., & Eichenbaum, H. (2013). Distinct hippocampal
time cell sequences represent odor memories in immobilized rats. Journal of
Neuroscience, 33, 14607–14616.
MacDonald, C. J., Lepage, K. Q., Eden, U. T., & Eichenbaum, H. (2011). Hippocampal
“Time Cells” bridge the gap in memory for discontiguous events. Neuron, 71,
737–749.
Marlin, N. A. (1981). Contextual associations in trace conditioning. Animal Learning &
Behavior, 9, 519–523.
Matell, M. S., Meck, W. H., & Nicolelis, M. A. (2003). Interval timing and the encoding of
signal duration by ensembles of cortical and striatal neurons. Behavioral Neuroscience,
117, 760.
McEchron, M. D., Bouwmeester, H., Tseng, W., Weiss, C., & Disterhoft, J. F. (1998).
Hippocampectomy disrupts auditory trace fear conditioning and contextual fear
conditioning in the rat. Hippocampus, 8, 638–646.
McEchron, M. D., & Disterhoft, J. F. (1997). Sequence of single neuron changes in CA1
13
Neurobiology of Learning and Memory xxx (xxxx) xxx–xxx
hippocampus of rabbits during acquisition of trace eyeblink conditioned responses.
Journal of Neurophysiology, 78, 1030–1044.
McEchron, M. D., Weible, A. P., & Disterhoft, J. F. (2001). Aging and learning-specific
changes in single-neuron activity in CA1 hippocampus during rabbit trace eyeblink
conditioning. Journal of Neurophysiology, 86, 1839–1857.
Meck, W. H. (2006). Neuroanatomical localization of an internal clock: A functional link
between mesolimbic, nigrostriatal, and mesocortical dopaminergic systems. Brain
Research, 1109, 93–107.
Mello, G. B., Soares, S., & Paton, J. J. (2015). A scalable population code for time in the
striatum. Current Biology, 25, 1113–1122.
Misane, I., Tovote, P., Meyer, M., Spiess, J., Ögren, S. O., & Stiedl, O. (2005). Time-
dependent involvement of the dorsal hippocampus in trace fear conditioning in mice.
Hippocampus, 15, 418–426.
Modi, M. N., Dhawale, A. K., & Bhalla, U. S. (2014). CA1 cell activity sequences emerge
after reorganization of network correlation structure during associative learning.
Elife, 3, e01982.
Morrissey, M. D., Insel, N., & Takehara-Nishiuchi, K. (2017). Generalizable knowledge
outweighs incidental details in prefrontal ensemble code over time. Elife, 6, e22177.
Morrissey, M. D., Maal-Bared, G., Brady, S., & Takehara-Nishiuchi, K. (2012). Functional
dissociation within the entorhinal cortex for memory retrieval of an association be-
tween temporally discontiguous stimuli. Journal of Neuroscience, 32(16), 5356–5361.
Moser, E. I., Kropff, E., & Moser, M.-B. (2008). Place cells, grid cells, and the brain’s
spatial representation system. Annual Review of Neuroscience, 31, 69–89.
Moyer, J. R., Jr., Deyo, R. A., & Disterhoft, J. F. (1990). Hippocampectomy disrupts trace
eye-blink conditioning in rabbits. Behavioral Neuroscience, 104, 243–252.
Munera, A., Gruart, A., Munoz, M. D., Fernandez-Mas, R., & Delgado-Garcia, J. M. (2001).
Hippocampal pyramidal cell activity encodes conditioned stimulus predictive value
during classical conditioning in alert cats. Journal of Neurophysiology, 86, 2571–2582.
Murray, E. A., Gaffan, D., & Mishkin, M. (1993). Neural substrates of visual stimulus-
stimulus association in rhesus monkeys. Journal of Neuroscience, 13, 4549–4561.
Murray, E. A., & Wise, S. P. (1996). Role of the hippocampus plus subjacent cortex but not
amygdala in visuomotor conditional learning in rhesus monkeys. Behavioral
Neuroscience, 110, 1261.
Namboodiri, V. M. K., & Shuler, M. G. H. (2014). Report of interval timing or action?
Proceedings of the National Academy of Sciences, 111 E2239–E2239.
Narayanan, N. S., Horst, N. K., & Laubach, M. (2006). Reversible inactivations of rat
medial prefrontal cortex impair the ability to wait for a stimulus. Neuroscience, 139,
865–876.
Narayanan, N. S., Land, B. B., Solder, J. E., Deisseroth, K., & DiLeone, R. J. (2012).
Prefrontal D1 dopamine signaling is required for temporal control. Proceedings of the
National Academy of Sciences, 109, 20726–20731.
Naya, Y., & Suzuki, W. A. (2011). Integrating what and when across the primate medial
temporal lobe. Science, 333, 773–776.
Naya, Y., Yoshida, M., & Miyashita, Y. (2001). Backward spreading of memory-retrieval
signal in the primate temporal cortex. Science, 291, 661–664.
Naya, Y., Yoshida, M., & Miyashita, Y. (2003). Forward processing of long-term asso-
ciative memory in monkey inferotemporal cortex. Journal of Neuroscience, 23,
2861–2871.
Pakaprot, N., Kim, S., & Thompson, R. F. (2009). The role of the cerebellar interpositus
nucleus in short and long term memory for trace eyeblink conditioning. Behavioral
Neuroscience, 123, 54–61.
Parker, A., & Gaffan, D. (1998). Memory after frontal/temporal disconnection in mon-
keys: Conditional and non-conditional tasks, unilateral and bilateral frontal lesions.
Neuropsychologia, 36, 259–271.
Parker, K. L., Chen, K.-H., Kingyon, J. R., Cavanagh, J. F., & Narayanan, N. S. (2014). D1-
dependent 4 Hz oscillations and ramping activity in rodent medial frontal cortex
during interval timing. Journal of Neuroscience, 34, 16774–16783.
Parker, K. L., Ruggiero, R. N., & Narayanan, N. S. (2015). Infusion of D1 dopamine re-
ceptor agonist into medial frontal cortex disrupts neural correlates of interval timing.
Frontiers in Behavioral Neuroscience, 9.
Pastalkova, E., Itskov, V., Amarasingham, A., & Buzsáki, G. (2008). Internally generated
cell assembly sequences in the rat hippocampus. Science, 321, 1322–1327.
Pasupathy, A., & Miller, E. K. (2005). Different time courses of learning-related activity in
the prefrontal cortex and striatum. Nature, 433, 873–876.
Pavlov, I. P. (1927). Conditional reflexes: An investigation of the physiological activity of
the cerebral cortex.
Petrides, M. (1982). Motor conditional associative-learning after selective prefrontal le-
sions in the monkey. Behavioural Brain Research, 5, 407–413.
Petter, E. A., Lusk, N. A., Hesslow, G., & Meck, W. H. (2016). Interactive roles of the
cerebellum and striatum in sub-second and supra-second timing: Support for an in-
itiation, continuation, adjustment, and termination (ICAT) model of temporal pro-
cessing. Neuroscience & Biobehavioral Reviews, 71, 739–755.
Powell, D. A., & Churchwell, J. (2002). Mediodorsal thalamic lesions impair trace eye-
blink conditioning in the rabbit. Learning and Memory, 9, 10–17.
Rainer, G., Rao, S. C., & Miller, E. K. (1999). Prospective coding for objects in primate
prefrontal cortex. Journal of Neuroscience, 19, 5493–5505.
Raybuck, J. D., & Lattal, K. M. (2011). Double dissociation of amygdala and hippocampal
contributions to trace and delay fear conditioning. PloS One, 6, e15982.
Robinson, N. T., Priestley, J. B., Rueckemann, J. W., Garcia, A. D., Smeglin, V. A., Marino,
F. A., & Eichenbaum, H. (2017). Medial entorhinal cortex selectively supports tem-
poral coding by hippocampal neurons. Neuron, 94, 677–688.
Rolls, E. T. (2000). Functions of the primate temporal lobe cortical visual areas in in-
variant visual object and face recognition. Neuron, 27, 205–218.
Runyan, J. D., Moore, A. N., & Dash, P. K. (2004). A role for prefrontal cortex in memory
storage for trace fear conditioning. Journal of Neuroscience, 24, 1288–1295.
Ryou, J. W., Cho, S. Y., & Kim, H. T. (2001). Lesions of the entorhinal cortex impair
M. Pilkiw, K. Takehara-Nishiuchi
acquisition of hippocampal-dependent trace conditioning. Neurobiology of Learning
and Memory, 75, 121–127.
Sakai, K., & Miyashita, Y. (1991). Neural organization for the long-term memory of paired
associates. Nature, 354, 152–155.
Sakon, J. J., Naya, Y., Wirth, S., & Suzuki, W. A. (2014). Context-dependent incremental
timing cells in the primate hippocampus. Proceedings of the National Academy of
Sciences, 111, 18351–18356.
Salz, D. M., Tiganj, Z., Khasnabish, S., Kohley, A., Sheehan, D., Howard, M. W., &
Eichenbaum, H. (2016). Time cells in hippocampal area CA3. Journal of Neuroscience,
36, 7476–7484.
Schmaltz, L. W., & Theios, J. (1972). Acquisition and extinction of a classically condi-
tioned response in hippocampectomized rabbits (Oryctolagus cuniculus). Journal of
Comparative and Physiological Psychology, 79(2), 328–333.
Selden, N. R. W., Everitt, B. J., Jarrard, L. E., & Robbins, T. W. (1991). Complementary
roles for the amygdala and hippocampus in aversive conditioning to explicit and
contextual cues. Neuroscience, 42, 335–350.
Siegel, J. J. (2014). Modification of persistent responses in medial prefrontal cortex
during learning in trace eyeblink conditioning. Journal of Neurophysiology, 112,
2123–2137.
Siegel, J. J. (2016). Prefrontal single-neuron responses after changes in task contingencies
during trace eyeblink conditioning in rabbits. Eneuro, 3 ENEURO–0057.
Siegel, J. J., Kalmbach, B., Chitwood, R. A., & Mauk, M. D. (2012). Persistent activity in a
cortical-to-subcortical circuit: Bridging the temporal gap in trace eyelid conditioning.
Journal of Neurophysiology, 107, 50–64.
Siegel, J. J., & Mauk, M. D. (2013). Persistent activity in prefrontal cortex during trace
eyelid conditioning: dissociating responses that reflect cerebellar output from those
that do not. Journal of Neuroscience, 33, 15272–15284.
Siegel, J. J., Taylor, W., Gray, R., Kalmbach, B., Zemelman, B. V., Desai, N. S., ...
Chitwood, R. A. (2015). Trace eyeblink conditioning in mice is dependent upon the
dorsal medial prefrontal cortex, cerebellum, and amygdala: behavioral character-
ization and functional circuitry. Eneuro, 2 ENEURO-0051.
Solomon, P. R., Vander Schaaf, E. R., Thompson, R. F., & Weisz, D. J. (1986).
Hippocampus and trace conditioning of the rabbit’s classically conditioned nictitating
membrane response. Behavioral Neuroscience, 100, 729.
Suter, E. E., Weiss, C., & Disterhoft, J. F. (2013). Perirhinal and postrhinal, but not lateral
entorhinal, cortices are essential for acquisition of trace eyeblink conditioning.
Learning and Memory, 20, 80–84.
Takeda, M., Naya, Y., Fujimichi, R., Takeuchi, D., & Miyashita, Y. (2005). Active main-
tenance of associative mnemonic signal in monkey inferior temporal cortex. Neuron,
48, 839–848.
Takehara, K., Kawahara, S., Takatsuki, K., & Kirino, Y. (2002). Time-limited role of the
hippocampus in the memory for trace eyeblink conditioning in mice. Brain Research,
951, 183–190.
Takehara-Nishiuchi, K., Kawahara, S., & Kirino, Y. (2005). NMDA receptor-dependent
processes in the medial prefrontal cortex are important for acquisition and the early
stage of consolidation during trace, but not delay eyeblink conditioning. Learning and
Memory, 12, 606–614.
14
Neurobiology of Learning and Memory xxx (xxxx) xxx–xxx
Takehara-Nishiuchi, K., & McNaughton, B. L. (2008). Spontaneous changes of neocortical
code for associative memory during consolidation. Science, 322, 960–963.
Tam, S. K., & Bonardi, C. (2012). Dorsal hippocampal involvement in appetitive trace
conditioning and interval timing. Behavioral Neuroscience, 126, 258.
Tanaka, K. (1996). Inferotemporal cortex and object vision. Annual Review of
Neuroscience, 19, 109–139.
Tanninen, S. E., Yu, X., Giritharan, T., Tran, L., Bakir, R., Volle, J., ... Takehara-Nishiuchi,
K. (2015). Cholinergic, but not NMDA, receptors in the lateral entorhinal cortex
mediate acquisition in trace eyeblink conditioning. Hippocampus, 25, 1456–1464.
Tremblay, L., Hollerman, J. R., & Schultz, W. (1998). Modifications of reward expecta-
tion-related neuronal activity during learning in primate striatum. Journal of
Neurophysiology, 80, 964–977.
Tseng, W., Guan, R., Disterhoft, J. F., & Weiss, C. (2004). Trace eyeblink conditioning is
hippocampally dependent in mice. Hippocampus, 14, 58–65.
Volle, J., Yu, X., Sun, H., Tanninen, S. E., Insel, N., & Takehara-Nishiuchi, K. (2016).
Enhancing prefrontal neuron activity enables associative learning of temporally
disparate events. Cell Reports, 15, 2400–2410.
Walker, A. G., & Steinmetz, J. E. (2008). Hippocampal lesions in rats differentially affect
long- and short-trace eyeblink conditioning. Physiology & Behavior, 93, 570–578.
Weible, A. P., McEchron, M. D., & Disterhoft, J. F. (2000). Cortical involvement in ac-
quisition and extinction of trace eyeblink conditioning. Behavioral Neuroscience, 114,
1058–1067.
Weiss, C., Knuttinen, M.-G., Power, J. M., Patel, R. I., O’connor, M. S., & Disterhoft, J. F.
(1999). Trace eyeblink conditioning in the freely moving rat: Optimizing the con-
ditioning parameters. Behavioral Neuroscience, 113, 1100.
Wirth, S., Yanike, M., Frank, L. M., Smith, A. C., Brown, E. N., & Suzuki, W. A. (2003).
Single neurons in the monkey hippocampus and learning of new associations. Science,
300, 1578–1581.
Woodruff-Pak, D. S., Lavond, D. G., & Thompson, R. F. (1985). Trace conditioning:
Abolished by cerebellar nuclear lesions but not lateral cerebellar cortex aspirations.
Brain Research, 348, 249–260.
Xu, M., Zhang, S., Dan, Y., & Poo, M. (2014). Representation of interval timing by tem-
porally scalable firing patterns in rat prefrontal cortex. Proceedings of the National
Academy of Sciences, 111, 480–485.
Yanike, M., Wirth, S., Smith, A. C., Brown, E. N., & Suzuki, W. A. (2009). Comparison of
associative learning-related signals in the macaque perirhinal cortex and hippo-
campus. Cerebral Cortex, 19(5), 1064–1078.
Yanike, M., Wirth, S., & Suzuki, W. A. (2004). Representation of well-learned information
in the monkey hippocampus. Neuron, 42, 477–487.
Yin, B., & Meck, W. H. (2014). Comparison of interval timing behaviour in mice following
dorsal or ventral hippocampal lesions with mice having δ-opioid receptor gene de-
letion. Philosophical Transactions of the Royal Society of London. Series B, Biological
sciences, 369, 20120466.
Yoon, T., & Otto, T. (2007). Differential contributions of dorsal vs. ventral hippocampus
to auditory trace fear conditioning. Neurobiology of Learning and Memory, 87,
464–475.