About me

Since Fall 2016, I am a postdoctoral fellow in the Psycholinguistics and Cognition Lab at Concordia University, Montreal, Canada. I work with Professor Alan Bale and Professor Roberto G. de Almeida in investigating the neural basis of semantic representation and processing. 

In September 2016, I graduated with a Ph.D. degree in linguistics from New York University, where I was advised by Professor Liina Pylkkänen

My research focuses on natural language semantics and pragmatics, its brain basis, and its computational modelling. I use findings from formal semantics and pragmatics to design psycho-/neuro-linguistic experiments, investigating how meanings are represented and processed by the human brain in real time. Moreover, I use computational methods to quantitatively study the factors that weigh into the semantic processing by the human brain and develop artificial intelligence models. 

The central theme of my research is to understand the mechanisms underlying semantic compositionality. Compositionality comes into play at every level of semantic processing. Humans combine words to form conceptually more complex and novel meanings, compose sentential semantics from the meanings of functional and lexical words, gather information from both linguistic and non-linguistic sources to form a coherent whole, and establish causal dependence among separate events, etc. My research aims to address both the neural mechanisms specifically supporting meaning composition (together with those specifically supporting the processing of basic linguistic elements that potentially reflect human-specific cognitive abilities, e.g., negation, coordination, measurement, quantification, presupposition, attitude reporting, etc) as well as how these mechanisms are related to various parts of our general cognition, such as attention and working memory. 

About my dissertation:

Composing conceptual and logical information: 
Semantic processing in left anterior temporal cortex

Dissertation committee: 

Liina Pylkkänen (chair), Chris Barker, Alec Marantz, Gregory Murphy, and Anna Szabolcsi.


The true power of human language lies in its potential to communicate infinite complex and new ideas by combining a finite inventory of lexical and functional words. Though the use of functional words (and various grammatical factors encoded in them) plays a crucial role in systematically building larger linguistic constructions from smaller ones, and might potentially reflect human-specific cognitive abilities, the neural basis of functional words is still mostly not understood as compared to the processing of lexical words.

This dissertation builds on previous findings on the contribution of the left anterior temporal lobe (LATL) to both semantic memory and linguistic composition. We investigate the computational details of this region in processing combinatorial expressions, and in particular, in processing noun phrases and simple sentences containing functional words that encode logical operators or anaphoricity.

With the use of magnetoencephalography (MEG) and a word-by-word presentation paradigm, we conducted three language comprehension experiments, investigating the temporal progression of LATL activity during the processing of (i) noun-noun compounds varying in conceptual specificity (e.g., vegetable soup, tomato dish), (ii) positive and negative event-describing sentences with their subject DP varying in specificity (e.g., a reptile is moving, no green lizard is sleeping), and (iii) noun phrases with anaphoric functional adjectives (e.g., same star, different star).

Overall, we found robust LATL effects for semantic composition  as opposed to syntactic structure building, and that it is information integration – as opposed to information access – that leads to most reliable LATL effects. More specifically, integrating more specific information leads to larger brain activation in the LATL: in processing noun-noun compounds, phrases containing more specific modifiers elicited larger LATL effects during head noun presentation; at the sentential level, sentences containing a more specific subject DP caused larger LATL effects during the presentation of sentence-final verbs; and under a given context, anaphoric functional adjectives such as same and different, despite their lack of featural information in their lexical meaning, elicited as robust LATL effects as lexical adjectives (e.g., green) do. These LATL effects typically peak at 200 – 300 ms after the onset of the critical word (i.e., head nouns in noun phrases, and verbal predicates in sentences), suggesting an early involvement of the LATL in semantic composition.

These studies bridge theoretical investigation of natural language semantics and neurobiological research on brain mechanisms underlying language processing. We have shown, for the first time, that the processing of logical operators (e.g., sentential negation) and functional words (e.g., same, different) interacts with the processing of conceptual knowledge in the LATL in forming mental representations varying in specificity, and all these factors modulate LATL activation in integrating information and facilitating interlocutors to focus their attention to specific situations or items from the vast set of possibilities in the environment. Together, the studies in this dissertation constitute a step toward a better understanding of neurobiological basis for semantic compositionality.


Linmin Zhang

SP-244 (Loyola Campus)
Department of Psychology
7141 Sherbrooke Street West
Concordia University
Montréal, QC, Canada H4B 1R6
Phone 514-848-2424 - ext. 2210




(Updated on October 7th, 2017)