TY - JOUR
AU - Lee, Inah
AB - Abstract Despite its anatomical positioning as an interface between the perceptual and memory systems, the perirhinal cortex (PER) has long been considered dedicated for object recognition memory. Whether the PER is also involved in object perception has been intensely debated in recent studies, but physiological evidence has been lacking. We recorded single units from the PER while the rat made categorical responses immediately after sampling a visual object as the originally learned objects were ambiguously morphed to varying degrees. Some neurons in the PER changed their firing rates monotonically following the gradual changes across the morphed objects as if they coded perceptual changes of the object stimuli. However, other neurons abruptly changed their firing rates according to the response categories associated with the morphed objects, seemingly responding to the learned relationships between the stimulus and its associated choice response. The gradual and abrupt changes in object-tuning properties were also found at the neural population level. Furthermore, the object-associated tuning characteristics of neurons in the PER were more readily observable in correct trials than in error trials. Our findings suggest that neurons in the PER represent perceptual details of an object in addition to its mnemonic identity. object perception, object recognition, pattern completion, pattern separation, perirhinal cortex Introduction Animals including humans recognize an object despite constant variations in its physical features (i.e., invariant object recognition), and it has long been suggested that the perirhinal cortex (PER) plays critical roles in such object recognition memory. Some studies have shown that damaging the PER results in performance deficits in tasks in which an object must be recognized when it reappears against a novel one (Meunier et al. 1993; Ennaceur et al. 1996; Ennaceur and Aggleton 1997; Winters and Bussey 2005). Neurons in the PER in primates encode arbitrary associations between visual objects (Liu et al. 2000; Naya et al. 2001), and represent the relative familiarity of an object by subsequently decrementing the responses toward the repeated presentation of the same object (Miller et al. 1991; Fahy et al. 1993; Xiang and Brown 1998). In rodents, PER neurons significantly alter their firing rates near an object stimulus, forming a so-called “object field” (Burke et al. 2012; Deshmukh et al. 2012; Burke et al. 2014). A recent study also reported that the information about familiarity and novelty of an object is conveyed at different frequency bands in the PER (Ho et al. 2015). While the traditional theories have investigated the roles of the PER under the theme of recognition memory as described above, a relatively recent theory states that the PER is important not only for object recognition memory but also for object perception especially when there is a significant amount of “feature ambiguity” between objects (Bussey et al. 2002, 2003, 2006; Bartko et al. 2007; Murray et al. 2007). Prior studies that examined this perceptual theory for the PER have used a concurrent discrimination task, in which animals were trained to discriminate between different sets of visual objects sharing common features (e.g., AB+ vs. AC−, AB+ vs. BD−) (Bussey et al. 2002; Bartko et al. 2007), or objects that underwent physical morphing so that they become perceptually similar to each other (Bussey et al. 2003, 2006; Clark et al. 2011). In those studies, more severe deficits were found following PER lesions in high ambiguity conditions, compared with low or intermediate ambiguity conditions (Bussey et al. 2002, 2003, 2006; Bartko et al. 2007; Murray et al. 2007). This perceptual–mnemonic theory posits that a conjunctive representation of an object is formed and stored in the PER as its individual sensory features are processed hierarchically from lower to higher cortical areas (Kravitz et al. 2013), and the resulting object representation in the PER resolves potential feature ambiguity. However, the debate on the mnemonic versus perceptual roles of the PER in object information processing remains unresolved (Baxter 2009; Suzuki 2009; Suzuki and Baxter 2009; McTighe et al. 2010; Clark et al. 2011), resulting in mixed results in behavioral studies. For example, McTighe et al. (2010) showed that rats with PER damage could recognize novel objects as normally as controls when perceptual interference was minimal during the delay before a recognition test. The authors interpreted the results as suggesting that the PER lesions made the rats more susceptible to perceptual interference from individual features of objects and that the blackout period during the delay removed such interference. In contrast, Clark et al. (2011) demonstrated that rats with PER lesions were not different from controls in discriminating visually morphed objects presented on a computer monitor. These conflicting results in prior studies may stem from various sources including different experimental protocols for making lesions and behaviorally testing rats. For example, permanent lesions in the PER may allow other areas to compensate for the missing PER network in an abnormal way especially when the lesions are made before the acquisition of the task (Clark et al. 2011). Therefore, to obtain a complete picture of the neural mechanisms of object information processing in the PER, physiologically monitoring neuronal activity in the normal brain while an animal performs in an object recognition memory task is critical. However, despite the ongoing theoretical debate with conflicting behavioral results, it is surprising that there has been no physiological study to address the issue with a proper experimental design. In the current study, we tested rats in an object memory task in which both perceptual and mnemonic information must be processed while recording single units in the PER. Unlike the previous concurrent discrimination paradigm where behavioral decision was made simultaneously with stimulus sampling (Bussey et al. 2002, 2003, 2006; Clark et al. 2011), the object stimulus in the current task disappeared immediately after the rat sampled the object, and its subsequent choice was made in the absence of the object. In this way, we could minimize possible confounds between perceptual and mnemonic components that might have been present in prior, concurrent discrimination paradigms. Also, the object stimuli originally used for training were visually morphed to different degrees in our study because feature ambiguity has been one of the key factors to recruit the PER in perception (Bussey et al. 2002, 2003, 2006). We showed in our previous study that inactivating the PER significantly impaired performance in rats when 2 distinct visual objects must be associated with different choice responses (Ahn and Lee 2015). In the current study, by recording the firing patterns of single units in the PER, we examined whether neurons in the PER exhibited differential firing patterns between when the rat perceived physically different objects (object-sampling period), and when the animal was required to make categorical responses to those objects (choice period). Materials and Methods Subjects Four male Long-Evans rats (350–420 g at the time of surgery) were housed in individual cages and maintained under a 12 h light/dark cycle. All experiments were performed in the light portion of the cycle. Rats were maintained at 85% of the free-feeding weight throughout the experiment with ad libitum access to water. All experimental protocols were in compliance with the NIH Guide for the Care and Use of Laboratory Animals and the guidelines set by the Institutional Animal Care and Use Committee of the Seoul National University. Behavioral Apparatus Rats performed the task on an elevated linear track (7 × 44.5 cm; 84 cm above the floor) with a food tray attached at its end. At the other end of the track was a guillotine door-operated start box (16 × 22.5 × 30 cm). An array of 3 LCD monitors was positioned right above the food tray. The center monitor was only used, and it was equipped with an infrared touchscreen panel (Elo Touch Systems) to detect animal's touch responses. A transparent acrylic panel with 3 round holes (diameter = 3 cm) was overlaid on the monitor to define the animal's response areas. Two opaque acrylic dividers were attached in between the round holes to prevent the animal from making indiscreet choices. Four fiber optic sensors (Autonics) were installed along the linear track to record movement. The breakage of the third sensor (installed in front of the monitor) triggered the onset of a gray-scaled object stimulus (4 × 6 cm) in the monitor. When the animal nose-poked the object image, it disappeared with a sound feedback (2.25 kH, 83 dB) and 2 gray response disks (diameter = 3 cm) immediately appeared on the sides. The rat was required to touch one of the disks associated with the cueing object for obtaining food reward (Cocoballs, Kellogg's). The apparatus was located in a curtained area and was dimly lit by a halogen light (0.2 lux) on the ceiling. White noise (80 dB) was provided throughout the experiment. A custom-written Matlab software (using PsychToolbox) was used to control the onset and offset of experimental stimuli. Experimental Paradigm Rats were trained in a touchscreen-based object memory task (Ahn and Lee 2015) that required to form an associative memory between 2 perceptually distinct object images (a Toy figure [T] and an Egg [E]) and the response disks that subsequently appeared on both sides (Fig. 1, see Supplementary Fig. S1). Specifically, rats were trained to sample an object by touching an object image on a touchscreen monitor. Once the rat touched the object, the object cue disappeared, and 2 disks appeared on both sides of the location where the object cue had been presented. The rat was required to touch one of the choice disks to obtain a reward. Transparent acrylic dividers were installed on both sides of the object to prevent the rats from using a mediating response strategy (Chudasama and Muir 1997). Besides the initial training for nose-poking, rats were trained for 19.75 ± 1.89 (mean ± SD) days to reach the presurgery performance criterion (≥70% correct responses for 2 consecutive sessions) with 2 standard objects. Afterward, a hyperdrive carrying 16 tetrodes was implanted targeting the PER and a week of recovery period was given. Once most tetrodes recorded single units, the rat was retested in the same object memory task while the firing patterns of single units were monitored at the same time. Unlike the original task, however, during the postsurgical testing, the 2 original images were digitally morphed into each other (Bussey et al. 2003, 2006; Clark et al. 2011) to create 10 different, feature-ambiguous object stimuli (T5 to E5 in Fig. 2A). Along the morphed object dimension, half of object stimuli (T1–T5) closer to the original Toy figure were associated with the left disk and the other half (E1–E5) closer to the Egg figure were associated with the right disk. Such task demands required rats to make discrete categorical choices to obtain a reward (Fig. 2A). The order of presenting the morphed objects across trials was pseudo-randomized with the restriction that the same object did not appear for more than 3 consecutive trials, and the rat experienced all objects equally. Object-choice/reward contingencies were counterbalanced among rats. Figure 1. Open in new tabDownload slide Behavioral paradigm. A cartoon of the rat performing the task with a schematic illustration of event epochs (object-sampling period and choice period) in the object recognition task. When the rat touched the object image on the touchscreen, the object cue disappeared with a 2.25 kHz tone and was immediately replaced by 2 response disks. The rat subsequently touched the object-associated disk and the trial terminated with an auditory feedback (1.5 kHz for correct and 0.15 kHz for incorrect choices). The object-sampling period (pink) was defined from the object onset to object touch and the choice period (green) from object touch to disk choice. Figure 1. Open in new tabDownload slide Behavioral paradigm. A cartoon of the rat performing the task with a schematic illustration of event epochs (object-sampling period and choice period) in the object recognition task. When the rat touched the object image on the touchscreen, the object cue disappeared with a 2.25 kHz tone and was immediately replaced by 2 response disks. The rat subsequently touched the object-associated disk and the trial terminated with an auditory feedback (1.5 kHz for correct and 0.15 kHz for incorrect choices). The object-sampling period (pink) was defined from the object onset to object touch and the choice period (green) from object touch to disk choice. Figure 2. Open in new tabDownload slide Feature-ambiguous object stimuli and behavioral performance. (A) The images of 2 original objects were morphed into each other to yield 10 object images differing in the amount of feature overlap. Each object was named by combining the object category (initialed as T for Toy and E for Egg) with the serial position (1–5) along the object-morphing spectrum. The choice border was set in the middle of the morphing continuum (between T1 and E1), and all objects in the same category required the same behavioral choices (denoted by “+” for reward and “−” for no reward for illustration purposes only). (B) The level of physical object dissimilarity calculated between the original object and one of the morphed objects (including the original ones). (C) Behavioral performance as a function of increasing feature ambiguity. (D) The performance from the first session was grouped into a high ambiguity category (ambiguity levels 3–4), and a low ambiguity category (levels 0–2). Note that the performance in AMB-hi increased, whereas the performance in AMB-lo decreased in the second half of the session. Dotted lines are regression lines. (E) Performance curves for individual rats across 10 morphed object stimuli. The dotted line indicates the choice category border. (F) An averaged object categorical bias plotted across sessions. Figure 2. Open in new tabDownload slide Feature-ambiguous object stimuli and behavioral performance. (A) The images of 2 original objects were morphed into each other to yield 10 object images differing in the amount of feature overlap. Each object was named by combining the object category (initialed as T for Toy and E for Egg) with the serial position (1–5) along the object-morphing spectrum. The choice border was set in the middle of the morphing continuum (between T1 and E1), and all objects in the same category required the same behavioral choices (denoted by “+” for reward and “−” for no reward for illustration purposes only). (B) The level of physical object dissimilarity calculated between the original object and one of the morphed objects (including the original ones). (C) Behavioral performance as a function of increasing feature ambiguity. (D) The performance from the first session was grouped into a high ambiguity category (ambiguity levels 3–4), and a low ambiguity category (levels 0–2). Note that the performance in AMB-hi increased, whereas the performance in AMB-lo decreased in the second half of the session. Dotted lines are regression lines. (E) Performance curves for individual rats across 10 morphed object stimuli. The dotted line indicates the choice category border. (F) An averaged object categorical bias plotted across sessions. Data Analysis Unit-Screening Criteria The following set of criteria was applied to screen neurons for final analyses: (1) units with sufficient cluster separation (isolation distance ≥10) (Harris et al. 2001) and signal-to-noise ratio (L-ratio <0.3) (Schmitzer-Torbert et al. 2005), (2) average firing rate during at least one event period >0.5 Hz, and (3) cells with the mean firing rate exceeding 10 Hz in a session were considered interneurons and eliminated from the analyses. Unit Classification The units that met the above screening criteria have been further classified into (1) bursting, (2) regular spiking, and (3) unclassified neurons based on the spike autocorrelogram characteristics (±500 ms window, bin size = 1 ms) (Bartho et al. 2004; Ahn and Lee 2015). In autocorrelogram, bursting neurons exhibited a substantial peak between 3 and 6 ms, followed by an exponential decay afterward. For regular-spiking neurons, the maximum peak was found at >35 ms. The neurons that did not belong to any of those categories were labeled as unclassified neurons. Physical Object Dissimilarity Index Physical similarity between object stimuli was quantified by calculating the pixel-by-pixel Euclidean distance between one standard image and other images across object images. The resulting values were mapped between 0 (e.g., the original Toy image being compared with itself) and 1 (e.g., the original Egg image being compared with the original Toy image). Event Periods for Analysis Unlike our prior study (Ahn and Lee 2015), a trial was divided into 2 event periods for analytical purposes in the current study as follows: (1) “object-sampling period” (from object onset to object touch) during which the object stimulus was visible on the screen and touched, and (2) “choice period” (from object touch to choice) during which 2 disks appeared on both sides, and the rat was required to choose one of them in the absence of the cueing object. Thus, our paradigm imposed no delay between the 2 event periods. Curve-Fitting Procedures for Object-Tuning Curve Discharge rates for 10 morphed objects were averaged across the correct trials for an event period in which a given neuron exhibited sufficient spikes (>0.5 Hz). The firing rates were then normalized from 0 to 1 and oriented to show the object category with the higher firing rates on the right-hand side. The data were fit with the following set of model equations: Quadratic model: α+β*obj+γ*obj2 Four-parameter sigmoid model: γmin+(γmax−γmin)(1+e(−α*(obj−β))), where γmax and γmin denote upper and lower asymptotes, respectively, and α and β indicate a growth rate and inflection point, respectively. Obj denotes object stimuli. Bayesian information criteria (BIC) were applied to the 2 models, and the model that yielded a lower BIC value was selected as the best-fit model for a given neuron. Only the neurons that showed sufficient fitting to the model (R2 ≥ 0.3) were used for further analyses. The neurons best fit with the sigmoidal model were discarded if the inflection point fell outside the object range (from 1 to 10). The neurons fit with the quadratic model were named “Q-cells,” and those fit with the sigmoidal model were called “S-cells” throughout the current study. Population object-tuning curves were obtained separately for Q-cells and S-cells by averaging across the individual tuning curves and fitting the quadratic and sigmoidal models to the data, respectively. Results Behavioral Performance In our study, the physical differences between the ambiguous stimuli (object dissimilarity index, see Materials and Methods section) gradually increased in a curvilinear fashion along the morphing dimension (Fig. 2B). Nonetheless, rats made correct choices in approximately 71% of the trials in a given session. Rats were affected by the physical ambiguity in object stimuli, showing a significant decrease in performance as feature ambiguity increased (F4,64 = 29.68, P < 0.0001, repeated-measures ANOVA) (Fig. 2C). Importantly, however, performance stayed significantly higher than chance across all ambiguity levels (t’s > 4.41, P’s < 0.0001, one-sample t-test) (Fig. 2C), suggesting that rats were able to disregard some changes made in the original objects to make correct choices. This was also the case when the rats encountered the ambiguous objects for the first time on day 1 (67.5 ± 0.05%; mean ± SEM) (t(3) = 3.47, P < 0.05, one-sample t-test). To examine if the rat learned the object-response paired associations within the first session, we grouped performance for ambiguity levels 0–2 (T3–T5 and E3–E5 in Fig. 2A) as a low ambiguity condition (AMB-lo), and that from ambiguity levels 3–4 (T1–T2 and E1–E2 in Fig. 2A) as a high ambiguity condition (AMB-hi). The cumulative performance from AMB-lo decreased linearly across trials (P < 0.0001), whereas performance from AMB-hi showed the opposite pattern showing a significant increase across trials (P < 0.001; linear regression) (Fig. 2D). The slopes of the regression lines differed significantly between AMB-lo and AMB-hi (F1,60 = 53.31, P < 0.0001; ANCOVA). The results suggest that some learning took place during the first session for the rat to acquire the object-response paired-associative relationships for significantly altered objects (AMB-hi). It is also interesting that performance for AMB-lo also decreased gradually as the rat learned AMB-hi (Fig. 2D). It is possible that introducing ambiguous objects in a session interfered with the normal recognition of the original objects. Across the sessions, however, performance did not change significantly at all ambiguity levels (see Supplementary Fig. S2). A 2-way repeated-measures ANOVA with ambiguity level and session as within-subject factors revealed a significant effect of ambiguity level (F4,12 = 25.81, P < 0.0001), but the effect of neither session (F3,9 = 1.85, P = 0.21) nor the interaction between the 2 factors (F12,36 = 1.17, P = 0.34) was significant. We found that except for one rat, all rats tended to perform slightly better in one object category than the other (Fig. 2E). The direction of performance bias varied across rats, suggesting that rats did not innately prefer a particular object category. Also, the bias cannot be attributed to the rat preferring a particular motor response to one side because, for example, although rat 144 was run with the opposite object-response reward contingency compared with rat 141, both rats showed the performance bias in the same direction (Fig. 2E). A categorical object bias was measured by calculating the absolute performance difference between the Egg category (E1–E5) and the Toy category (T1–T5). The averaged bias was not statistically significant in all recording sessions (P’s = 0.13, one-sample sign test) (Fig. 2F). We also measured the response latency in the sampling (from object onset to object touch) and choice period (from object touch to disk choice). Rats exhibited a higher response latency in the sampling period (1.74 ± 0.28; mean ± SEM) than in the choice period (0.91 ± 0.05; mean ± SEM) (t(16) = 3.28, P < 0.01; paired t-test). Continuous Versus Categorical Coding of Feature Ambiguity in the PER We recorded spiking activities of single units in the PER while the rat performed the ambiguous object recognition task. Only the units that met isolation criteria (n = 128; Materials and Methods section) were used in final analyses. Most units (67.2%, n = 86/128) were regular-spiking cells, whereas some units were classified as bursting cells (21.1%, n = 27/128) and others as unclassified cells (11.7%, n = 15/128) based on autocorrelograms (Bartho et al. 2004; Ahn and Lee 2015) (Fig. 3A). Unit recordings were made from the dorsoventral extent (A35 and A36 subregions), and from both deep and superficial layers in the PER (Fig. 3B). Figure 3. Open in new tabDownload slide Single-unit recorded from the PER. (A) Representative examples of bursting (left), regular-spiking (middle), and unclassified (right) neurons in the PER. Inset: Average waveform of the neuron with the number below indicating its mean firing rate and spike width. (B) Tip locations of individual tetrodes marked on histological sections obtained from an online rat atlas (http://www.rbwb.org). Different rats were marked with different colors. Lines demarcate the boundaries of the PER. The number indicates the relative distance from bregma. Figure 3. Open in new tabDownload slide Single-unit recorded from the PER. (A) Representative examples of bursting (left), regular-spiking (middle), and unclassified (right) neurons in the PER. Inset: Average waveform of the neuron with the number below indicating its mean firing rate and spike width. (B) Tip locations of individual tetrodes marked on histological sections obtained from an online rat atlas (http://www.rbwb.org). Different rats were marked with different colors. Lines demarcate the boundaries of the PER. The number indicates the relative distance from bregma. To examine how single units in the PER responded to the gradual changes in object features across the morphed objects, for each unit, the firing rates recorded across trials (only correct trials used) for each object stimulus were averaged and normalized (from 0 to 1) separately for the object-sampling period and the choice period (Materials and Methods section). The normalized firing rates associated with the morphed objects were then oriented in a scatter plot in such a way that the object category (defined by the choice response) associated with higher firing rates (“preferred category,” or P) was positioned on the right-hand side, and the “non-preferred category” (NP) on the left-hand side (Fig. 4A). Figure 4. Open in new tabDownload slide Q-cell and S-cell in the PER. (A) For drawing an object-tuning curve, for each neuron, the object category associated with higher firing rates (“preferred category” or P) was positioned on the right-hand side and the “non-preferred category” (or NP) on the left-hand side. (B) Representative examples of curve fitting for a Q-cell (fit by a quadratic model) and S-cell (fit by a sigmoid model). Filled circles denote the normalized firing rates for each morphed object. Figure 4. Open in new tabDownload slide Q-cell and S-cell in the PER. (A) For drawing an object-tuning curve, for each neuron, the object category associated with higher firing rates (“preferred category” or P) was positioned on the right-hand side and the “non-preferred category” (or NP) on the left-hand side. (B) Representative examples of curve fitting for a Q-cell (fit by a quadratic model) and S-cell (fit by a sigmoid model). Filled circles denote the normalized firing rates for each morphed object. For neurons that exhibited significant discharges (>0.5 Hz) during either the object-sampling period or the choice period, the tuning curve that best described the relationships between the morphed objects and the corresponding firing rates was determined by fitting either a quadratic or sigmoidal function to the data. Then, the best fitting model was selected based on BIC (Fig. 4B) (Materials and Methods section). The neurons whose response profiles seemed suitable for perceptually coding continuous changes in sensory features were best fit by the quadratic model (“Q-cells”). In contrast, the neurons whose response profiles captured the response contingencies associated with the morphed objects, but not necessarily the physical changes in object features, were best fit by the sigmoidal model (“S-cells”). We found both classes of neurons in both the object-sampling period (Fig. 5A) and the choice period (Fig. 5B). There were approximately similar numbers of Q-cells in the sampling period (n = 25; cells 1–5 in Fig. 5A) and the choice period (n = 28; cells 9–11 in Fig. 5B,C, P = 0.78). However, the number of S-cells almost doubled during the choice period (n = 33; cells 12–16 in Fig. 5B), compared with the sampling period (n = 16; cells 6–8 in Fig. 5A,C), as if reflecting a significant mnemonic task demand in that period. We found a statistical trend for a proportional difference between the Q-cells and S-cells across the sampling and choice periods (F = 2.23, P = 0.14, Chi-square test). No significant proportional difference was found between the neurons showing preference (i.e., higher firing rates) for the Toy category (54.7% for Q-cells and 55.1% for S-cells) and for the Egg category (45.2% for Q-cells and 44.9% for S-cells, P’s > 0.57; binomial test), suggesting that neither object category was innately preferred by the PER neurons. Figure 5. Open in new tabDownload slide Distributions of Q-cells and S-cells in the object-sampling period and the choice period. (A–B) Representative examples of object-tuning curves obtained from single units during the object-sampling period (A) and the choice period (B). Each neuron was fit with a quadratic (Q) or sigmoidal (S) model, and the optimal model was selected based on the BIC value. R2 measured the tightness of curve fitting. (C) The number of Q-cells and S-cells in the sampling and choice period. (D) A Venn diagram showing the proportions of Q-cells and S-cells. The intersection of the 2 circles represents the neurons (Q- and S-cells) fit by both quadratic and sigmoidal functions across the event periods. The number in the parenthesis indicates the number of units. Figure 5. Open in new tabDownload slide Distributions of Q-cells and S-cells in the object-sampling period and the choice period. (A–B) Representative examples of object-tuning curves obtained from single units during the object-sampling period (A) and the choice period (B). Each neuron was fit with a quadratic (Q) or sigmoidal (S) model, and the optimal model was selected based on the BIC value. R2 measured the tightness of curve fitting. (C) The number of Q-cells and S-cells in the sampling and choice period. (D) A Venn diagram showing the proportions of Q-cells and S-cells. The intersection of the 2 circles represents the neurons (Q- and S-cells) fit by both quadratic and sigmoidal functions across the event periods. The number in the parenthesis indicates the number of units. The majority of PER neurons (87.5%, n = 70/80) were best fit with either a quadratic or sigmoidal model at least in one event period. (Fig. 5D and see Supplementary Fig. S3A–F). Interestingly, some neurons dynamically switched their response profiles across the event periods (Fig. 5D; 12.5%, n = 10/80), showing a quadratic response during the sampling period and a sigmoidal response during the choice period (see Supplementary Fig. S3G), or vice versa (see Supplementary Fig. S3H). We could not find layer-specific differences in the proportion of Q-cells and S-cells (F = 0.83, P = 0.66, Chi-square test) (Q-cells: 42% [n = 24/57] in deep layers and 52% [n = 12/23] in superficial layers; S-cells: 44% [n = 25/57] in deep layers and 39% [n = 9/23] in superficial layers; Q- and S-cells: 14% [n = 8/57] in deep layers and 9% [n = 2/23] in superficial layers). These results indicate that the population of Q-cells and S-cells in the PER, while functionally heterogeneous, may not be mutually exclusive, and some PER neurons might perform dual cognitive functions across the event periods. Performance-Related Object-Turning Curves of PER Neurons To examine whether the firing properties of Q-cells and S-cells were correlated with the animal's performance in the task, we obtained the object-tuning curves separately for correct and error trials. Fitting the firing rates from the incorrect trials with the same model chosen as best for correct trials for the cell markedly disrupted the tuning properties. Specifically, data points were not as tightly fit to the tuning curve as in correct trials when the rat made errors in both the object-sampling (Fig. 6A) and choice periods (Fig. 6B). Furthermore, the tuning curves of S-cells became noticeably flattened near the choice border (e.g., cell 3 in Fig. 6A; cells 7, 8, 9, and 12 in Fig. 6B) and/or were off-tuned such that the inflection points were found off of the choice border (e.g., cell 4 in Fig. 6A; cells 10 and 11 in Fig. 6B). Figure 6. Open in new tabDownload slide Comparison of object-tuning curves between correct and error trials. (A–B) Normalized firing rates from error trials (open circles) were fit with the same fitting model obtained from the correct trials (filled circles). The 2 curves obtained from correct (solid curves) and error trials (dashed curves) were overlaid to illustrate how the tuning characteristics were affected by performance. For S-cells (cells 3–4 in [A] and cells 7–12 in [B]), the inflection points of the curves were indicated by arrowheads on the abscissa. The inflection points from the incorrect trials were marked with gray arrowheads on the same axis (omitted if the inflection point fell out of the morphing range). (C) Cumulative proportions of R2 values that quantify the goodness-of-fit of the tuning curves of Q-cells from the sampling (dark blue) and choice (light blue) period, respectively. Dashed lines indicate incorrect trials. (D) The relationships between the inflection points (abscissa) and goodness-of-fit (R2, ordinate) of S-cells from the sampling (pink) and choice period (green). Solid and dashed lines denote correct and incorrect trials, respectively. C.B. stands for the choice border. (E) The slope of the individual S-cells recorded from both event periods was calculated at the choice border. Note that the S-cells from the choice period exhibited the sharpest slope at the choice border when the rats made correct choices compared with other conditions. ***P < 0.001. Figure 6. Open in new tabDownload slide Comparison of object-tuning curves between correct and error trials. (A–B) Normalized firing rates from error trials (open circles) were fit with the same fitting model obtained from the correct trials (filled circles). The 2 curves obtained from correct (solid curves) and error trials (dashed curves) were overlaid to illustrate how the tuning characteristics were affected by performance. For S-cells (cells 3–4 in [A] and cells 7–12 in [B]), the inflection points of the curves were indicated by arrowheads on the abscissa. The inflection points from the incorrect trials were marked with gray arrowheads on the same axis (omitted if the inflection point fell out of the morphing range). (C) Cumulative proportions of R2 values that quantify the goodness-of-fit of the tuning curves of Q-cells from the sampling (dark blue) and choice (light blue) period, respectively. Dashed lines indicate incorrect trials. (D) The relationships between the inflection points (abscissa) and goodness-of-fit (R2, ordinate) of S-cells from the sampling (pink) and choice period (green). Solid and dashed lines denote correct and incorrect trials, respectively. C.B. stands for the choice border. (E) The slope of the individual S-cells recorded from both event periods was calculated at the choice border. Note that the S-cells from the choice period exhibited the sharpest slope at the choice border when the rats made correct choices compared with other conditions. ***P < 0.001. We subsequently tested whether the fitting parameters of a tuning curve including the goodness-of-fit (measured by R2) and inflection point were related to the animal's performance in the task. Because a spurious relationship may arise from the discrepant number of samples between correct and incorrect trials, the fitting parameters were obtained by using a downsampling method. Specifically, we downsampled the correct trials to match the number of error trials for individual objects, and the same fitting model was applied to the firing rates obtained from those subsampled correct trials. We repeated this step for 100 times, and the resulting R2 values associated with the individual curves were averaged across iterations, and compared with that from error trials. The inflection point of S-cells was obtained by applying the fitting model to the firing rates from subsampled correct trials. In Q-cells, R2 values decreased significantly in error trials, compared with correct trials in both event periods (Z’s < −2.13, P’s < 0.05, signed-rank test; Fig. 6C). The tuning characteristics of S-cells were examined by plotting the R2 values against the inflection points (Fig. 6D). In correct trials, S-cells showed a tighter fit to the curve (mean R2 = 0.52) relative to the error trials (mean R2 = 0.44, Z = −2.10, P < 0.05, signed-rank test; Fig. 6D). The inflection points of the curves were clustered near the choice categorical border in correct trials (mean = 5.55, SD = 1.57), whereas the same neurons exhibited relatively scattered inflection points in error trials (mean = 4.14, SD = 2.41) (Fig. 6D). Moreover, the proportion of cells whose inflection points fell within the choice categorical boundary (between 5 and 6 in Fig. 6D) was significantly larger in correct trials (51.0%, n = 25/49) compared with error trials (20.4%, n = 10/49; F = 10, P < 0.01, Chi-square test). Also, the direction of neuronal bias (the deviation of the inflection point toward a particular object category) tended to match that of the behavioral bias (rats performing higher in one object category than the other) in correct trials. That is, of the 12 S-cells that exhibited neuronal bias with an inflection point shifted toward one object category in the choice period, 10 neurons showed the matching behavioral bias for the same object category (P < 0.05, binomial test). The results strongly indicate that the response characteristics of S-cells may be associated with the categorical representation of the morphed objects, but not simply with the motoric behavior of the animals. Furthermore, we measured the slope of the individual tuning curves of S-cells at the choice border by taking the derivative of the curve at the border. The slopes of the individual curves were averaged and subjected to a 2-way mixed ANOVA with the event period as a between-subject factor, and correctness as a within-subject factor. There were significant effects of event period (F1,47 = 9.34, P < 0.01), correctness (F1,47 = 6.72, P < 0.05), and the interaction between the 2 factors (F1,47 = 4.37, P < 0.05) (Fig. 6E). A post hoc test revealed a significant difference in correctness in the choice period (t(32) = 3.89, P < 0.001; t-test), but not in the sampling period (t(15) = 0.35, P = 0.73). These results strongly indicate that individual visual features of objects and their associated mnemonic responses were more reliably represented at the single neuronal level in the PER when the rat made correct choices than when making errors. Monotonic VersusStepwise Coding of the Continuously Morphed Objects by the Neuronal Populations in the PER We further examined whether the perceptual–mnemonic coding properties observed at the single-unit level were also observed at the neural population level in the PER. For this purpose, the individual object-tuning curves associated with the 2 neuronal classes, that is, Q-cells and S-cells, were averaged for each class and fitted with the quadratic model for Q-cells and with the sigmoidal model for S-cells. During the object-sampling period, the population object-tuning curve for Q-cells increased monotonically in a curvilinear fashion across the morphed objects as if to reflect the increasing physical dissimilarity of the objects (Pearson's R between the 2 curves = 0.97; Fig. 7A). More importantly, in the choice period, the population tuning curve for S-cells increased in a stepwise fashion, exhibiting a sharp nonlinear transition from one state to the other across the choice border (Fig. 7A). By contrast, the population tuning curves of Q-cells (cyan) and S-cells (orange) obtained from the event periods that might not be optimal for task demands did not reflect such task demands. Specifically, the population object-tuning curve of Q-cells in the choice period did not follow the physical object dissimilarity curve as closely as in the object-sampling period (Pearson's R = 0.93, Fig. 7B). Furthermore, the population tuning curve of S-cells during the sampling period exhibited a relatively linear transition along the choice border (Fig. 7B), showing a more flattened curve (growth rate = 0.34), compared with the choice period (growth rate = 2.26). This could be attributable to the relatively large variability in the inflection points of the individual tuning curves (Fig. 6D). It is also important to note that the inflection point of the turning curve for S-cells was found within the choice border only during the choice period (see Supplementary Table S1). Our findings suggest that the firing characteristics of Q-cells and S-cells at the population level are optimized for the perceptual coding of a feature-ambiguous object and the subsequent associative recognition of the object, respectively. Figure 7. Open in new tabDownload slide Population object-tuning curves of Q-cells and S-cells. (A) The population tuning curves of Q-cells (blue) and S-cells (red) from the object-sampling and choice periods. The curve representing the physical object dissimilarity (black dash) was overlaid. (B) The population tuning curves of Q-cells (cyan) and S-cells (orange) obtained from the event periods that might not be optimal for task demands (i.e., Q-cells in the choice period and S-cells in the sampling period). The inverted arrowheads denote the choice border. Figure 7. Open in new tabDownload slide Population object-tuning curves of Q-cells and S-cells. (A) The population tuning curves of Q-cells (blue) and S-cells (red) from the object-sampling and choice periods. The curve representing the physical object dissimilarity (black dash) was overlaid. (B) The population tuning curves of Q-cells (cyan) and S-cells (orange) obtained from the event periods that might not be optimal for task demands (i.e., Q-cells in the choice period and S-cells in the sampling period). The inverted arrowheads denote the choice border. We also examined how the differences in firing characteristics of neurons between correct and incorrect trials (Fig. 6) influenced the object-tuning curves at the population level, and found that the performance-related differences in tuning curves were more readily observable for S-cells than for Q-cells at the population level (Fig. 8). Specifically, the Q-cells in the sampling period (Fig. 8A) showed a more flattened (almost linear) fitting curve in incorrect trials compared with that in correct trials. The Q-cells from both correct and incorrect trials in the choice period (Fig. 8B) showed similar fitting patterns with a higher growth rate in the preferred object category. The S-cells in the sampling period showed nonoptimal fitting patterns with the inflection point located off from the choice border both in correct and incorrect trials (Fig. 8C). As for the S-cells in the choice period, the inflection points fell within the choice border in both correct and incorrect trials (Fig. 8D). However, the contrast in firing rates (the difference between upper and lower asymptote) and growth rate across the choice border were larger in correct trials than in incorrect trials (Fig. 8D and see Supplementary Table S1). We also calculated the slope of the population tuning curve of S-cells similarly by taking the derivative of the curve at the choice border and found that the tuning curve from the correct trials exhibited a steeper slope (0.3) compared with that from the error trials (0.07) in the choice period. (Fig. 8D). However, this was not the case in the object-sampling period (Fig. 8C), suggesting that the representations of the morphed objects in S-cells were more sharply orthogonalized according to behavioral choices in correct trials than in error trials. Figure 8. Open in new tabDownload slide Population object-tuning curves drawn separately for correct and incorrect trials. The tuning curves from incorrect trials (dashed lines) were superimposed on those from correct trials to facilitate comparisons between the 2 curves according to performance (A–B). For Q-cells, the physical object dissimilarity curve (gray dashed line) was overlaid to show how the perceptual object coding of Q-cells in the sampling (A) and choice periods (B) reflected the physical changes along the morphing dimension. (C–D). For S-cells, the inflection point of each curve was marked with colored arrowheads beneath the abscissa. The filled arrowheads indicate the inflection points of the tuning curve from correct trials and the empty ones from error trials. The black inverted triangle on the abscissa marks the choice border. The number on the right-hand side of each curve of S-cells denotes the slope of the tuning curve at the choice border, calculated by getting the derivative of the curve at the border. Figure 8. Open in new tabDownload slide Population object-tuning curves drawn separately for correct and incorrect trials. The tuning curves from incorrect trials (dashed lines) were superimposed on those from correct trials to facilitate comparisons between the 2 curves according to performance (A–B). For Q-cells, the physical object dissimilarity curve (gray dashed line) was overlaid to show how the perceptual object coding of Q-cells in the sampling (A) and choice periods (B) reflected the physical changes along the morphing dimension. (C–D). For S-cells, the inflection point of each curve was marked with colored arrowheads beneath the abscissa. The filled arrowheads indicate the inflection points of the tuning curve from correct trials and the empty ones from error trials. The black inverted triangle on the abscissa marks the choice border. The number on the right-hand side of each curve of S-cells denotes the slope of the tuning curve at the choice border, calculated by getting the derivative of the curve at the border. Taken together, these results suggest that the neural populations in the PER may carry out dual functions associated with different task demands, one being monotonic perceptual discrimination of object features and the other being nonlinear orthogonalization by taking object memory-guided choices into account. Discussion In the current study, we recorded single units from the PER while the rat made a categorical response after sampling one of the continuously morphed objects. Neuronal firing in the PER monotonically changed when viewing a morphed object cue but exhibited stepwise shift according to its paired-associate response category during choice. The monotonic or stepwise neuronal tuning for objects was visible at both the single-unit level and the population level and was strongly correlated with the behavioral performance. Our findings provide the evidence that the PER is involved in both perceiving feature-ambiguous objects and recognizing the objects by using their associative properties. Our experimental design has several advantages in testing the perceptual–mnemonic hypothesis for the PER. Specifically, we used a morphing technique (Bussey et al. 2003, 2006; Clark et al. 2011) to provide the rats with perceptually similar, yet physically different visual stimuli. Prior behavioral studies operationally defined “feature ambiguity” by making animals perform concurrent discriminations between different sets of objects (e.g., making some of the objects common to both groups [Bussey et al. 2002, 2005]). However, we preferred the morphing procedure for the following reasons: First, a possible influence of learning on the perceptual process could be reduced because new learning would be minimally required when objects were morphed from well-learned objects. Second, it was unclear whether the brain would process a conjunctive combination of objects (e.g., AB+ vs. AC−, AB+ vs. BD−) used in prior studies (Bussey et al. 2002; Bartko et al. 2007) as an individual object, scene, or both. Also, it was easier to quantify the relative similarities among objects that underwent digital morphing. Furthermore, unlike the previous studies that focused primarily on the animal's perceptual capability (Bussey et al. 2002, 2003, 2006; Clark et al. 2011), we pitted the perceptual component against the mnemonic component within the same task by subsequently requiring the rats to make discrete choices in the absence of the object. That is, the rat was required to know the similarities and differences among the objects (i.e., perceptual differences) in the object-sampling period to make an object-paired-associative response (i.e., left or right disk touch) during the choice period. In the PER, we found 2 classes of neurons, Q-cells and S-cells, whose firing patterns were related to task demands. Q-cells exhibited incremental firing patterns, reflecting the physical differences among the morphed objects, whereas S-cells showed the nonlinear response profile as if to reflect the paired relationships between the object stimulus and its associated choice response. Based on the response profiles, it is conceivable that Q-cells may encode the perceptual features of an object and S-cells represent the mnemonic aspects of the objects. We previously reported that, before and after making a choice response, the PER neurons fired differentially according to the choice response associated with the object, but not for object identity per se (Ahn and Lee 2015). Such findings are most likely attributable to the fact that, in our previous study, the object firing rates were obtained from the prechoice period (from object onset to choice) that did not differentiate the perceptual sampling period from the choice period. Also, firing rates for individual objects were averaged for each object category in our previous study, which may have made it difficult to capture the subtle, yet significant response changes of the PER neurons along the morphed objects reported in the current study. An ensemble of neurons in the hippocampus is known to perform computations such as “pattern completion” and “pattern separation” for generalizing similar memory representations into a common representation and orthogonalizing dissimilar memories into separate representations, respectively (Marr 1971; Lee et al. 2004; Leutgeb et al. 2005, 2007). The stepwise object-tuning curves of S-cells reported in the current study are reminiscent of the nonlinear changes in the neural activity of hippocampal neurons (Leutgeb et al. 2005, 2007; Wills et al. 2005) observed when rats explored continuously morphed geometric environments. A canonical computational theory views pattern separation as an input–output function where the output representation becomes less correlated than its original inputs (O'Reilly and McClelland 1994; Guzowski et al. 2004). The current results may satisfy this criterion because the neural output (mnemonic tuning curves of S-cells) became more orthogonalized than the original curvilinear inputs (object stimuli varying over the morphing continuum). According to a recently proposed “representational–hierarchical view” of pattern separation (Kent et al. 2016), pattern separation may be present across multiple regions upstream of the hippocampus including the PER. To the best of our knowledge, our study provides the first physiological support for this claim that the PER may perform pattern separation (and pattern completion) for object recognition. However, there are some notable differences between our results and the physiological phenomenon reported in the hippocampal literature. One of the main differences is that we were able to observe pattern separation (shown by the mnemonic tuning characteristics) in the PER not only at the neural population level but also at the individual single-unit level. This contrasts with the previous hippocampal findings because, to our knowledge, the physiological evidence for pattern separation or pattern completion has been observable only at the neural population levels (Lee et al. 2004; Leutgeb et al. 2005, 2007; Wills et al. 2005), but not at the individual neuronal level in the hippocampus. Another difference is on performance correlates. It is important to note that the putative pattern separation performed by S-cells, especially those from the choice period was significantly related to successful behavioral performance in our study. That is, at the individual neuronal level, the S-cells in the choice period exhibited a more precise point of inflection at the choice border, compared with other conditions. Also, at the population level, only S-cells in the choice period showed task-related tuning profiles with sharper orthogonalization across the choice border in correct trials, compared with error trials. However, such relationships between neural activity and behavioral performance have been scarcely reported in prior hippocampal studies, possibly because most hippocampal data were recorded in a foraging paradigm. Such foraging paradigm typically imposed no paired-associative requirement between the environmental stimulus and behavior (Lee et al. 2004; Leutgeb et al. 2005, 2007; Wills et al. 2005). One may argue that the stepwise response profile of S-cells might be derived from a motor response component largely because the objects in the same category were always coupled with a specific choice response. Although it is difficult to rule out such a possibility, the following aspects of our results make this possibility less likely. First, the parameters used to measure the object-tuning characteristics (e.g., goodness-of-fit and the inflection point of the object-turning curve) were significantly disrupted when the rats made errors. If the stepwise responses of S-cells were driven mostly by motor signals associated with touching the left or right disk, such motor components should also be observable in the curve-fitting parameters in error trials because rats made equally decisive motor responses during incorrect trials as they did during correct trials. Second, the orthogonalization of object category by S-cells in the PER was sharper in correct trials in the choice period compared with error trials although the same behavioral responses were made in those trials. Based on these findings, it is likely that the mnemonic firing properties of S-cells in our study were largely driven by the learned associative relationships between the object stimulus categories and choice responses. Despite our attempt to separate the perceptual and mnemonic components in the task, the 2 coding schemes might not be strictly discretized in the PER because we were able to identify S-cells during the sampling period and Q-cells during the choice period, albeit to different degrees. The presence of the above neuronal classes in the event period that may not be compatible with task demands might reflect the nature of the representations in the PER. It could also be attributable to the fact that there was no explicit delay in our task to separate the object-sampling and the choice period, resulting in a potential gray transition period between 2 event periods. It is possible that the choice-related representation of an object might have taken place at the time of object-sampling because the choice response could be specified based on the sample object alone in the current task. Also, there might have been some lingering activities of perceptual signals that persisted through the choice period. In the presence of a delay, one may expect to see a more explicit memory signal in the choice period that is less confounded with the perceptual signals carrying over from the sampling period, and the number of neurons showing nonoptimal responses may be minimized. However, we decided to start our experiment without delay first because putting a delay also may entail several confounding issues (e.g., mediating strategy). Nonetheless, the absence of delay should not undermine the main results of the current study because the population of Q-cells in the sampling period reflected the physical change along the morphed objects more faithfully than that in the choice period. Likewise, the population of S-cells exhibited sharper and more accurate orthogonalization across the choice border in the choice period than in the sampling period. The perceptual and mnemonic firing correlates of neurons in the PER thus appear to be optimized for information processing required by corresponding task demands associated with object-sampling and behavioral choice in the current task. Prior studies typically used delayed-matching/nonmatching-to-sample paradigms (Meunier et al. 1993; Ennaceur et al. 1996; Ennaceur and Aggleton 1997; Winters and Bussey 2005) and concurrent discrimination tasks (Bussey et al. 2002, 2003, 2006; Clark et al. 2011). Compared with those paradigms, the current task required the animal to form an explicit paired-associative relationship between an object and its associated response disk, and thus could not be solved merely based on relative familiarity, or recency information of the stimulus. It is conceivable that the current paradigm imposed sufficient mnemonic demand on the animals even though the delay between the sample and choice was minimal. The functional heterogeneity of the PER observed in the current study may be linked to its anatomical position as an interface between the MTL memory system and the ventral visual pathway (Bussey and Saksida 2002; Naya et al. 2003; Murray et al. 2007; Clark et al. 2011; Kravitz et al. 2013). This unique position of the PER may accommodate the presence of these functionally distinct, if not mutually exclusive, classes of neurons, that is, Q-cells and S-cells, in the same area. The present findings provide the first physiological evidence that the rodent PER may serve functionally as a hybrid network for supporting the dual coding schemes by encoding perceptual differences between objects along the morphing continuum, and also by processing object recognition signal using object-associated mnemonic information. Supplementary Material Supplementary data is available at Cerebral Cortex online. Funding The National Research Foundation of Korea (NRF-2015M3C7A1031969, NRF-2016R1A2B4008692, SRC-2014051826, BK21+ program) Notes J.R.A. contributed to designing and executing the experiment, analyzing the data, writing the manuscript; I.L. contributed designing the study, analyzing the data, and writing the manuscript. Conflict of Interest: None declared. References Ahn JR , Lee I. 2015 . Neural correlates of object-associated choice behavior in the perirhinal cortex of rats . J Neurosci . 35 : 1692 – 1705 . Google Scholar Crossref Search ADS PubMed WorldCat Bartho P , Hirase H, Monconduit L, Zugaro M, Harris KD, Buzsaki G. 2004 . Characterization of neocortical principal cells and interneurons by network interactions and extracellular features . J Neurophysiol . 92 : 600 – 608 . Google Scholar Crossref Search ADS PubMed WorldCat Bartko SJ , Winters BD, Cowell RA, Saksida LM, Bussey TJ. 2007 . Perirhinal cortex resolves feature ambiguity in configural object recognition and perceptual oddity tasks . Learn Mem . 14 : 821 – 832 . Google Scholar Crossref Search ADS PubMed WorldCat Baxter MG . 2009 . Involvement of medial temporal lobe structures in memory and perception . Neuron . 61 : 667 – 677 . Google Scholar Crossref Search ADS PubMed WorldCat Burke SN , Maurer AP, Hartzell AL, Nematollahi S, Uprety A, Wallace JL, Barnes CA. 2012 . Representation of three-dimensional objects by the rat perirhinal cortex . Hippocampus . 22 : 2032 – 2044 . Google Scholar Crossref Search ADS PubMed WorldCat Burke SN , Maurer AP, Nematollahi S, Uprety A, Wallace JL, Barnes CA. 2014 . Advanced age dissociates dual functions of the perirhinal cortex . J Neurosci . 34 : 467 – 480 . Google Scholar Crossref Search ADS PubMed WorldCat Bussey TJ , Saksida LM. 2002 . The organization of visual object representations: a connectionist model of effects of lesions in perirhinal cortex . Eur J Neurosci . 15 : 355 – 364 . Google Scholar Crossref Search ADS PubMed WorldCat Bussey TJ , Saksida LM, Murray EA. 2002 . Perirhinal cortex resolves feature ambiguity in complex visual discriminations . Eur J Neurosci . 15 : 365 – 374 . Google Scholar Crossref Search ADS PubMed WorldCat Bussey TJ , Saksida LM, Murray EA. 2003 . Impairments in visual discrimination after perirhinal cortex lesions: testing ‘declarative’ vs. ‘perceptual-mnemonic’ views of perirhinal cortex function . Eur J Neurosci . 17 : 649 – 660 . Google Scholar Crossref Search ADS PubMed WorldCat Bussey TJ , Saksida LM, Murray EA. 2005 . The perceptual-mnemonic/feature conjunction model of perirhinal cortex function . Q J Exp Psychol B . 58 : 269 – 282 . Google Scholar Crossref Search ADS PubMed WorldCat Bussey TJ , Saksida LM, Murray EA. 2006 . Perirhinal cortex and feature-ambiguous discriminations . Learn Mem . 13 : 103 – 105 . Google Scholar Crossref Search ADS PubMed WorldCat Chudasama Y , Muir JL. 1997 . A behavioural analysis of the delayed non-matching to position task: the effects of scopolamine, lesions of the fornix and of the prelimbic region on mediating behaviours by rats . Psychopharmacology (Berl) . 134 : 73 – 82 . Google Scholar Crossref Search ADS PubMed WorldCat Clark RE , Reinagel P, Broadbent NJ, Flister ED, Squire LR. 2011 . Intact performance on feature-ambiguous discriminations in rats with lesions of the perirhinal cortex . Neuron . 70 : 132 – 140 . Google Scholar Crossref Search ADS PubMed WorldCat Deshmukh SS , Johnson JL, Knierim JJ. 2012 . Perirhinal cortex represents nonspatial, but not spatial, information in rats foraging in the presence of objects: comparison with lateral entorhinal cortex . Hippocampus . 22 : 2045 – 2058 . Google Scholar Crossref Search ADS PubMed WorldCat Ennaceur A , Aggleton JP. 1997 . The effects of neurotoxic lesions of the perirhinal cortex combined to fornix transection on object recognition memory in the rat . Behav Brain Res . 88 : 181 – 193 . Google Scholar Crossref Search ADS PubMed WorldCat Ennaceur A , Neave N, Aggleton JP. 1996 . Neurotoxic lesions of the perirhinal cortex do not mimic the behavioural effects of fornix transection in the rat . Behav Brain Res . 80 : 9 – 25 . Google Scholar Crossref Search ADS PubMed WorldCat Fahy FL , Riches IP, Brown MW. 1993 . Neuronal activity related to visual recognition memory: long-term memory and the encoding of recency and familiarity information in the primate anterior and medial inferior temporal and rhinal cortex . Exp Brain Res . 96 : 457 – 472 . Google Scholar Crossref Search ADS PubMed WorldCat Guzowski JF , Knierim JJ, Moser EI. 2004 . Ensemble dynamics of hippocampal regions CA3 and CA1 . Neuron . 44 : 581 – 584 . Google Scholar Crossref Search ADS PubMed WorldCat Harris KD , Hirase H, Leinekugel X, Henze DA, Buzsaki G. 2001 . Temporal interaction between single spikes and complex spike bursts in hippocampal pyramidal cells . Neuron . 32 : 141 – 149 . Google Scholar Crossref Search ADS PubMed WorldCat Ho JW , Poeta DL, Jacobson TK, Zolnik TA, Neske GT, Connors BW, Burwell RD. 2015 . Bidirectional modulation of recognition memory . J Neurosci . 35 : 13323 – 13335 . Google Scholar Crossref Search ADS PubMed WorldCat Kent BA , Hvoslef-Eide M, Saksida LM, Bussey TJ. 2016 . The representational-hierarchical view of pattern separation: Not just hippocampus, not just space, not just memory . Neurobiol Learn Mem . 129 : 99 – 106 . Google Scholar Crossref Search ADS PubMed WorldCat Kravitz DJ , Saleem KS, Baker CI, Ungerleider LG, Mishkin M. 2013 . The ventral visual pathway: an expanded neural framework for the processing of object quality . Trends Cogn Sci . 17 : 26 – 49 . Google Scholar Crossref Search ADS PubMed WorldCat Lee I , Yoganarasimha D, Rao G, Knierim JJ. 2004 . Comparison of population coherence of place cells in hippocampal subfields CA1 and CA3 . Nature . 430 : 456 – 459 . Google Scholar Crossref Search ADS PubMed WorldCat Leutgeb JK , Leutgeb S, Moser MB, Moser EI. 2007 . Pattern separation in the dentate gyrus and CA3 of the hippocampus . Science . 315 : 961 – 966 . Google Scholar Crossref Search ADS PubMed WorldCat Leutgeb JK , Leutgeb S, Treves A, Meyer R, Barnes CA, McNaughton BL, Moser MB, Moser EI. 2005 . Progressive transformation of hippocampal neuronal representations in “morphed” environments . Neuron . 48 : 345 – 358 . Google Scholar Crossref Search ADS PubMed WorldCat Liu Z , Murray EA, Richmond BJ. 2000 . Learning motivational significance of visual cues for reward schedules requires rhinal cortex . Nat Neurosci . 3 : 1307 – 1315 . Google Scholar Crossref Search ADS PubMed WorldCat Marr D . 1971 . Simple memory: a theory for archicortex . Philos Trans R Soc Lond B Biol Sci . 262 : 23 – 81 . Google Scholar Crossref Search ADS PubMed WorldCat McTighe SM , Cowell RA, Winters BD, Bussey TJ, Saksida LM. 2010 . Paradoxical false memory for objects after brain damage . Science . 330 : 1408 – 1410 . Google Scholar Crossref Search ADS PubMed WorldCat Meunier M , Bachevalier J, Mishkin M, Murray EA. 1993 . Effects on visual recognition of combined and separate ablations of the entorhinal and perirhinal cortex in rhesus monkeys . J Neurosci . 13 : 5418 – 5432 . Google Scholar PubMed OpenURL Placeholder Text WorldCat Miller EK , Li L, Desimone R. 1991 . A neural mechanism for working and recognition memory in inferior temporal cortex . Science . 254 : 1377 – 1379 . Google Scholar Crossref Search ADS PubMed WorldCat Murray EA , Bussey TJ, Saksida LM. 2007 . Visual perception and memory: a new view of medial temporal lobe function in primates and rodents . Annu Rev Neurosci . 30 : 99 – 122 . Google Scholar Crossref Search ADS PubMed WorldCat Naya Y , Yoshida M, Miyashita Y. 2001 . Backward spreading of memory-retrieval signal in the primate temporal cortex . Science . 291 : 661 – 664 . Google Scholar Crossref Search ADS PubMed WorldCat Naya Y , Yoshida M, Miyashita Y. 2003 . Forward processing of long-term associative memory in monkey inferotemporal cortex . J Neurosci . 23 : 2861 – 2871 . Google Scholar PubMed OpenURL Placeholder Text WorldCat O'Reilly RC , McClelland JL. 1994 . Hippocampal conjunctive encoding, storage, and recall: avoiding a trade-off . Hippocampus . 4 : 661 – 682 . Google Scholar Crossref Search ADS PubMed WorldCat Schmitzer-Torbert N , Jackson J, Henze D, Harris K, Redish AD. 2005 . Quantitative measures of cluster quality for use in extracellular recordings . Neuroscience . 131 : 1 – 11 . Google Scholar Crossref Search ADS PubMed WorldCat Suzuki WA . 2009 . Perception and the medial temporal lobe: evaluating the current evidence . Neuron . 61 : 657 – 666 . Google Scholar Crossref Search ADS PubMed WorldCat Suzuki WA , Baxter MG. 2009 . Memory, perception, and the medial temporal lobe: a synthesis of opinions . Neuron . 61 : 678 – 679 . Google Scholar Crossref Search ADS PubMed WorldCat Wills TJ , Lever C, Cacucci F, Burgess N, O'Keefe J. 2005 . Attractor dynamics in the hippocampal representation of the local environment . Science . 308 : 873 – 876 . Google Scholar Crossref Search ADS PubMed WorldCat Winters BD , Bussey TJ. 2005 . Glutamate receptors in perirhinal cortex mediate encoding, retrieval, and consolidation of object recognition memory . J Neurosci . 25 : 4243 – 4251 . Google Scholar Crossref Search ADS PubMed WorldCat Xiang JZ , Brown MW. 1998 . Differential neuronal encoding of novelty, familiarity and recency in regions of the anterior temporal lobe . Neuropharmacology . 37 : 657 – 676 . Google Scholar Crossref Search ADS PubMed WorldCat © The Author 2017. Published by Oxford University Press. This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com © The Author 2017. Published by Oxford University Press.
TI - Neural Correlates of Both Perception and Memory for Objects in the Rodent Perirhinal Cortex
JF - Cerebral Cortex
DO - 10.1093/cercor/bhx093
DA - 2017-07-01
UR - https://www.deepdyve.com/lp/oxford-university-press/neural-correlates-of-both-perception-and-memory-for-objects-in-the-T0BSxVQkbM
SP - 3856
EP - 3868
VL - 27
IS - 7
DP - DeepDyve
ER -