Cognitive Science Colloquia


Cognitive Science Colloquia are held* throughout the year and consist of seminars/talks given by invited speakers from various universities, as well as seminars by the METU staff or our own PhD students. The Colloquia aim to share information on cognitive science research, as well as promote academic research and collaboration. The seminars are open to anyone interested and provide an opportunity for our students to keep up with the recent research conducted in the field, allowing them to observe how research topics develop through academic exchange.

* Unless it is announced otherwise, seminars take place every Friday at 12:40, in the Informatics Institute (room II-03), METU.

For past years' colloquia, please visit here.




NEXT IN THE SERIES    >    TBA




This Year (2017-2018)


Fall Semester


Issues in the Semantics of Nominalization
by Umut Özge (Cognitive Science, METU)
on 5th of January

Abstract Introductory expositions of syntax(-semantics interface) usually start with a discussion of the parts of speech: nouns, verbs, adjectives, adverbs. It is remarkable that this pedagogical square-one of linguistic theory is also a fertile source for many recalcitrant puzzles in the field. To put it crudely, characterizing the difference between what is verbal and what is nominal is not a trivial task. Nominalization, the business of turning verb-like entities to noun-like entities as in 'refuse' to 'refusal' or 'John examined the papers' to 'John's examining the papers', sits at the center of the web of puzzles involving the very fundamental notions of linguistic theory. My primary aim in this talk is to raise interest in the semantics of various nominalization processes through an informal discussion of some data from Turkish, where nominalization interacts with argument structure, existence presuppositions, temporal reference, and event structure.


Neural Correlates of Decision Making Processes
by Murat Perit Çakır (Cognitive Science, METU)
on 29th of December

Abstract Existing work in decision making research has identified normative accounts to evaluate the decisions made by humans as well as specific limitations and heuristics shaping those decisions in the face of perceived risks and uncertainties. There is now a growing interest in cognitive science and neurosciences regarding the dynamical unfolding of the processes through which humans arrive at decisions. This talk will introduce some of the approaches employed in the field and provide an overview of some of the main findings regarding implicated brain networks, in relation to the empirical studies conducted with optical brain imaging technology at the COGS lab.


Silent Reading and Oral Reading in Turkish: An Investigation of Reading Modality and the Eye-Voice Span
by Ayşegül Özkan (Cognitive Science, METU)
on 22nd of December

Abstract Reading is a complex cognitive process which is an assembly of processes at different levels, from low-level visual perception to high-level discourse comprehension. Most of the research on reading heavily depend on eye movement data. Previous studies provide extensive evidence about the predictable influences of lexical and syntactic manipulations on eye movement measures in reading. (Warren, Reichle & Patson, 2011, Rayner, Pollatsek, Ashby & Clifton Jr., 2012; Radach & Kennedy, 2013; Rayner, 1998.) In oral reading eyes move ahead of the voice. The eye-voice span (i.e., EVS) is the distance between the articulation and viewing of a word. The goal of this study is to investigate eye movement patterns in silent reading and oral reading in Turkish. It is work in progress. We conducted a within-subject experiment, which consisted of a silent reading block, and an oral reading block. The full target word set is designed according to lexical characteristics of words, i.e., word length and frequency. The relationship between reading modality, lexical characteristics of words and eye movement measures, and the relationship between lexical characteristics of words and EVS are investigated. Additionally, the relationship between the eye-voice span and eye movement measures is investigated indirectly through the relationship between lexical characteristics of a target word and the fixation duration at the beginning of the articulation of that target word. The preliminary results suggest that lexical characteristics of words and reading modality influence the eye movement measures, and characteristics of words influence EVS, as well. The relationship between lexical characteristics of target words and the fixation durations at the beginning of the articulation of them imply a relationship between EVS and eye movement measures.


Extending Discourse Annotation to Multiple Languages: TED-MDB
by Deniz Zeyrek (Cognitive Science, METU)
on 15th of December

Abstract Annotation (adding information to texts) has become a science, bringing together linguists, cognitive scientists and computational linguists to share knowledge, ask new questions, and develop software. Annotation can be done on all levels of language; in this talk I will only be concerned with discourse-level annotation and the potential gains from this endeavor. My focus will be on discourse relations (contract, condition, expansion, etc.) that hold between text segments, which are usually clauses or groups of clauses. After introducing the basic concepts of discourse and annotation, I will introduce a new initiative where we annotate the transcripts of TED talks involving English, Turkish and several other languages. The corpus is called TED-Multilingual Discourse Bank, or TED-MDB. This initiative has grown within Textlink, a European COST project, bringing together a group of researchers interested in discourse annotation. I will explain the annotation process of TED-MDB, describe the current stage of the research, as well as the research questions that arise during this effort.


KelimeDenVektörE: What Do Distributed Representations of Turkish Words Tell Us About Morphology?
by Özkan Kılıç (Cognitive Science, METU)
on 8th of December

Abstract Mikolov et al. (2013a, 2013b) proposed a novel model architecture based on an Artificial Neural Network for computing continuous vector representations of words from very large data sets. The architecture is based on a skip-gram model for learning the representations that can capture semantic word relationships. In this ongoing project, 600 million Turkish words are used to train a model similar to Mikolov et al.'s. Initial findings suggest that such distributed representations can guide learning morphology. The findings also support the continuum approach to 'inflection-derivation' distinction rather than the dichotomous approach.


Denial of Expectation in Turkish and in General: New Insights and Prospects for a Logical Treatment
by Ceyhan Temürcü (Cognitive Science, METU)
on 17th of November

Abstract Apparently, all human languages have strategies for blocking possible inferences from a preceding clause or discourse segment. This sense, which take part in the semantic ranges in Turkish *ama *or *fakat*, and of English *but*, has often been dubbed 'denial of expectation' (DE) and subsumed under the more general family of 'contrastive' senses. In this talk, which reflects our joint work with Prof. Deniz Zeyrek, I will propose a general logical specification for the DE sense, arguing that DE denies a possible inference afforded by a counter-actual knowledge state. I will then compare this proposal to that of Winter and Rimon (1994), which is based on Veltman's (1986) data logic, and that of Toosarvandani (2014), which recruits Kratzer's (1981, 1991) analysis of modality.


Plans, Actions, and Sentences
by Cem Bozşahin (Cognitive Science, METU)
on 10th of November
[slides]

Abstract At the level of sentences, predicates bear thematic relations to their arguments, such as agent, patient, beneficiary. They show diversity in argument-taking, from intransitives to ditransitives. At the level of discourse we do not see such diversity, or sensitivity to thematic roles. Recursion is also one observable difference between discourse and sentence structure. The same differences appear to hold between plans, which are functions from states to states, and scripts, which are participant-taking functions with some affinity to thematic structure. Scripts are thematic-role sensitive whereas plans need not be. We summarise some thinking about linking planning and language, both cognitively, and evolutionarily.


Interaction between Language and Vision: Findings from Multimodal Comprehension Studies
by Cengiz Acartürk (Cognitive Science, METU)
on 3rd of November

Abstract The study of multimodal comprehension involves the investigation of linguistic and non-linguistic means of communication. The production of specific types of utterances, such as reference and deixis provide an appropriate testbed to study multimodal conceptualization of entities in the visual world. This talk will present findings in deictic cross-referencing in multimodal environments under various communication modalities.


Cognitive Science meets Social Science: Joint Action Paradigms for Adults and Children
by Annette Hohenberger (Cognitive Science, METU)
on 27th of October
[slides]

Abstract Within the last decades, cognitive science has seen a surge in joint (dual- and multiple-agent) task paradigms extending the classical individual paradigms. This development leads to an overlap with social science and promises some fruitful interdisciplinary integration across the two fields. On this path, similar yet different paradigms such as “social facilitation” and “joint action” need to be distinguished, though. In this talk I want to present some recent showcase studies conducted by our Cogs students which highlight the benefits of various joint paradigms as well as their challenges, studying adults and children: the role of joint action in time perception (Kerem Alp Usal, 2016); co-representation of a partner’s task in children (Esra Katircioglu Terzi, 2017); joint change-detection (Mustafa Akkuscu, 2017); and joint tool-making in children (Gökhan Gönül, ongoing). I will lay particular emphasis on the particular task structure used to study joint action, the possible mechanism(s) underlying it and the differential role of collaboration vs competition. The talk will conclude with an outlook on future research along the lines of this exciting new interdisciplinary research agenda.


Characterization of the Purkinje Cell to Nuclear Cell Connections in Mice Cerebellum
by Orkan Özcan (Neuroscience, CNRS)
on 20th of October
[slides TBA]

Abstract The cerebellum integrates motor commands with somatosensory, vestibular, visual and auditory information for motor learning and coordination functions. The final step of this integration is done in the deep cerebellar nuclei (DCN) that process inputs from Purkinje cells (PC) and the two main inputs of the cerebellum: mossy and climbing fibers. We investigated the properties of PC connections from the lobule IV/V of the cerebellar cortex to the different DCN cell types in the medial nuclei using optogenetic stimulation in L7-ChR2 mice combined with in vivo multi electrode extracellular recordings. We identified two groups of DCN cells with significant differences in their action potential waveforms and firing rates which matched with the properties of GABAergic and non-GABAergic cells discriminated in vitro. The discharge of the DCN cells was correlated to the local field potentials and we found that DCN cells were phase locked to oscillations in the beta, gamma and high frequency bands. Although optogenetic stimulation of PCs resulted in the inhibition of the two groups of DCN cells (rate coding), spike times were controlled only for non-GABAergic cells. Moreover, local inhibition onto non-GABAergic cells was not observed in our in vivo experiments. Our results suggest that synchronized PC inputs drive the output of cerebellum by temporally controlling only non-GABAergic cells. Also, the internal DCN circuitry supports this phenomenon since GABAergic cells are not temporally controlled and the local inhibition they provide does not alter the time-locked output of the DCN.