Mechanisms of adaptation in macaque inferior temporal cortex

In the current dissertation, we examined the phenomenon of neural adaptation. Neural adaptation can result in repetition suppression which refers to the phenomenon of a commonly observed decrease in response to a repeated stimulus presentation. The decrease in response with stimulus repetition has been observed at multiple spatial scales, from the level of spiking activity of single cortical neurons to the level of activity of a neural population as captured by a number of techniques (e.g., EEG, MEG, fMRI). In addition to this, the temporal scale of adaptation effects has been detected to vary from milliseconds (msec) to minutes and days. Neural adaptation occurs under an impressively high variety of experimental conditions and in a wide range of brain areas. Yet, despite having been thoroughly studied over decades, the underlying neural mechanisms of adaptation as well as its functional implications on perception and behavior are still poorly understood. Nevertheless, the latter fact did not preclude the use of neural adaptation in the fMRI-adaptation paradigm to infer neural selectivities in humans. Repetition suppression has been suggested to be a neural correlate of the behavioral phenomenon of repetition priming, that is, enhanced accuracy and speed in the detection and discrimination/identification of already experienced stimuli. This link between the two phenomena is based on the similarities in experimental conditions under which both phenomena take place. Moreover, it has been demonstrated that the phenomena share common properties such as stimulus specificity, existence of short- and long-lived components, modulation of the effect size by the number of stimulus repetitions and the rather automatic nature of either phenomenon.In the dissertation, we advanced our knowledge of the phenomenon of neural adaptation. Specifically, we addressed the question of the source of adaptation in inferior temporal cortex. We used macaque monkeys as an animal model for our recordings due to the invasiveness of the recording technique which precludes its common use in healthy human subjects. The choice of the inferior temporal cortex was based on the fact that this area codes high-order visual properties as well as on the fact that this area shows repetition suppression at the level of single neurons and fMRI. Recent studies suggested that repetition suppression is a consequence of the fulfillment of perceptual expectations or, in other words, of a reduced mismatch between expected and observed percepts, thus, stressing the role of top-down (feed-back) factors in generating adaptation. In accordance with these studies, the degree of neural adaptation is expected to be modulated by contextual factors that determine subject's perceptual expectations. This top-down view on the source of adaptation contradicts to bottom-up explanations of adaptation. According to the latter, adaptation is a consequence of an interplay of the feed-forward flow of sensory information throughout the cortex and local functioning of neural circuitry. The adaptation-evoked changes in functioning of neural circuitry were our next research question. We investigated whether in addition to response suppression with stimulus repetition, adaptation affected synchronization of neural activity. Apart from the changes in synchronization with adaptation, the latter question is of special interest. As has been suggested, better temporally coordinated (synchronized) activity with stimulus repetition may be that neural mechanism which compensates for the detrimental effect of response suppression and facilitates transfer of information throughout the cortex, thus, in this way producing repetition priming. In our next study, we addressed the question how stimulus repetition affects object representation accuracy (coding). It has been suggested before that adaptation results in increased accuracy of stimulus representation, which may be a neural correlate of repetition priming. To investigate this, we compared classification accuracies before and after short-term adaptation. If one believes that neural adaptation is the mechanism at work behind the phenomenon of repetition priming, then stimulus repetition is expected to result in better object coding.All research questions in the dissertation were addressed by analyzing spiking activity of either single neurons or small population(s) of neurons and local field potentials (LFPs) recorded simultaneously in the inferior temporal cortex of macaque monkeys (Macaca mulatta). In contrast to spikes which correspond to the output of neurons, LFPs appear to correspond to the input to neurons by representing a population measure of neural activity generated within the local cortical network. During electrophysiological recordings, monkeys were passively fixating in the center of a computer monitor while being sequentially presented pairs of visual stimuli (adapter and test) on a single trial. Depending on the research question, the duration of stimulus presentation and inter-stimulus interval (blank screen between the two stimuli) varied across experiments, but never exceeded 500 msec (short-term adaptation paradigm). To examine the effect of adaptation on neural synchronization and classification accuracy, we recorded spiking activity and LFPs throughout cortical layers using a 16-channel laminar electrode. In all our experiments, we performed spectral analysis of recorded LFPs.In a first study, we manipulated subject's expectations of stimulus repetitions by creating two contexts with a high (75%) and a low (25%) probability of encountering stimulus repetitions versus stimulus alternations. According to the top-down explanation of the source of neural adaptation, we expected to observe a high and a low degree of response suppression in these contexts, respectively, as has been observed in a human fMRI study by Summereld et al. (2008). As expected from previous studies, our single-cell spiking activity and LFP gamma band (60-100 Hz) power data indeed showed a decrease in response with stimulus repetition. However, the magnitude of response suppressionwith stimulus repetition did not significantly differ between the contexts. This lack of modulation of response suppression with contextual factors related to stimulus repetition probability conflicts with the top-down explanation of neural adaptation. On the contrary, our data agree well with simpler bottom-up mechanisms of adaptation.In order to assess the changes in synchronization of neural activity with adaptation, we simultaneously recorded LFPs and spiking activity of a small population of neurons, as reflected in multi-unit activity (MUA), from up to 16 channels throughout the cortex using a laminar electrode. We then computed LFP-LFP and MUA-LFP coherences for different pairs of channels for each recording session. Coherency is a quantitative measure of temporal coordination (synchronization) between two signals. Stimulus repetition in our study attenuated the response to a repeated stimulus of both MUA and LFP power for spectral frequencies above 60 Hz. Contrary to the gamma band frequencies, LFPs showed an increase in power with stimulus repetition for the frequencies between 15 and 40 Hz. In addition to the response modulation, stimulus repetition affected synchronization between two simultaneously recorded LFP signals of different channels as well as between MUA and LFP. Importantly, these changes in synchronization were frequency- and lamina-dependent. Stimulus repetition resulted in a decrease of both LFP-LFP and MUA-LFP coherences for frequencies above 60 Hz, and this was true across different layers. For the frequencies between 15 and 40 Hz, the effect of stimulus repetition on synchronization depended on the electrode depth. Both LFP-LFP and MUA-LFP synchronization was enhanced with stimulus repetition for the putative superficial layers, while the LFPs of the putative deep layers decreased their synchrony across layers. Thus, we demonstrated that short-term adaptation modulates not only the response strength but also synchronization of neural activity.To answer whether stimulus repetition affects object representation accuracy, we performed classifications of stimulus identity before and after adaptation with either the same or different stimulus. Stimuli were classified based on the responses which they evoked at the level of single neurons and MUA. Additionally, we classified stimuli based on the responses of a small population of simultaneously recorded MUA or LFPs across different frequency bands. Neurons were worse at discriminating a repeated stimulus compared to when the same stimulus was presented as adapter or preceded by a different stimulus. Interestingly, prior adaptation with a different stimulus in some instances led to enhanced discriminability of the test stimulus compared to that of the adapter. These observations held true for spiking activity of both single neurons and MUA as well as of small populations of MUA. Decoding stimulus identity from LFP power revealed a strong dependency of the classification effects on a frequency range, with the frequencies below 30 Hz producing greater classification accuracies than the gamma range. However, the effect of adaptation on classification accuracy was more pronounced at the gamma frequencies, showing a decrease in classification accuracy for repeated stimuli and a tendency for an improved object coding when the stimulus was preceded by a different stimulus. Based on these observations for spiking activity and LFPs, we concluded that adaptation impairs coding of repeated stimuli while supporting efficient coding of stimuli that differ from recently seen ones.status: publishe