Talk at Macquarie University

I'm in Sydney, Australia where I've just given a talk at the Australian Hearing Hub at Macquarie University. It's my first visit to Macquarie and it's been great. The recently-established Hesring Hub combines researchers, clinicians, and industry partners in a single building (which includes a nice cafe on the ground floor...I'm jealous). This multi-pronged approach to hearing science is exemplary and a model for interdisciplinary collaboration. I look forward to good things happening here over the coming years!

One of the special treats on my visit was a visit to the state-of-the-art anechoic chamber in the basement. Now THAT is a proper sound booth.

New paper: A role for the angular gyrus in combinatorial semantics (Price et al.)

We know what a "leaf" is, and we know what "wet" means. But combining these concepts together into a "wet leaf" yields a new and possibly more specific idea. Similarly, a "brown leaf" is qualitatively different than any old leaf. Our ability to flexibly and dynamically combine concepts enables us to represent and communicate an enormous set of ideas from a relatively small number of constituents. The question of what neural systems might support conceptual combination has been a focus of research for Amy Price at Penn. Combinatorial semantics is an especially timely topic as there are ongoing debates about the anatomical systems most strongly involved in semantic memory more generally (angular gyrus? anterior temporal lobes? ventral visual regions?), as well as the nature of the information being represented (to what degree do concepts rely on sensorimotor cortices?).

In a new paper out this week in the Journal of Neuroscience (Price et al., 2015), Amy presents data from both fMRI and patients with neurodegenerative disease suggesting that the angular gyrus plays an important role in conceptual combination. Amy designed a clever task in which participants read word pairs that varied in how easily they could be combined into a single concept. For example, you could imagine that "turnip rock" is difficult to combine, whereas a "wet rock" is easier. Amy used all adjective-noun pairs, but still found a considerable amount of variability (for example a "plaid apple" combines less easily than a "plaid jacket"). This "ease of combination" was initially quantified using subject ratings, but Amy found that lexical co-occurrence statistics for these word pairs strongly correlate with their degree of combination, and thus co-occurrence measures were used in all analyses. 

These findings are in good agreement with previous work emphasizing an important role for the angular gyrus in semantic representation (Binder & Desai 2011; Bonner et al. 2013).

References:

Binder JR, Desai RH (2011) The neurobiology of semantic memory. Trends in Cognitive Sciences 15:527-536. doi:10.1016/j.tics.2011.10.001

Bonner MF, Peelle JE, Cook PA, Grossman M (2013) Heteromodal conceptual processing in the angular gyrus. NeuroImage 71:175–186. doi:10.1016/j.neuroimage.2013.01.006 (PDF)

Price AR, Bonner MF, Peelle JE, Grossman M (2015) Converging evidence for the neuroanatomic basis of combinatorial semantics in the angular gyrus. Journal of Neuroscience 35:3276–3284. http://www.jneurosci.org./content/35/7/3276.short (PDF)

Talk at UCL

I gave a talk today at University College London, to the Speech Science Forum. UCL has a strong complement of speech, language, and cognitive scientists and it was a real pleasure to be here. There were a lot of interesting questions afterward, and several people helpfully pointed me towards some additional constraints or predictions that would be useful to consider. 

As a side note, London will be the location of the 2016 annual meeting of the Society for the Neurobiology of Language..it will be a great opportunity to visit this amazing city.

Talk at Oxford University

I gave a talk today at the Center for Neural Circuits and Behaviour at Oxford University. Though I spent over two years in Cambridge I was only in Oxford once before—and that was literally just to stop at the Eagle and Child. So, this is my first proper visit to Oxford, and I'm enjoying it very much.

The audience was diverse, as many people at CNBC are doing theoretical work or work in nonhuman systems (including both ferrets and *drosophila*—fruit flies). If I'm not mistaken there were so some folks from Experimental Psychology and FMRIB. I tried to give an overview of recent work on the role of ongoing oscillations in speech perception, and connect this with a somewhat separate line of research showing that degraded or noisy speech requires additional cognitive resources. The lines between these two bodies of research are tentative but also tantalizing. In any case, it's been a great visit and I'm already looking forward to returning!

New paper: Automatic analysis (aa) for neuroimaging analyes

I'm extra excited about this one! Out now in Frontiers in Neuroinformatics is our paper describing the automatic analysis (aa) processing pipeline (Cusak et al., 2015). aa started at the MRC Cognition and Brain Sciences Unit in Cambridge, spearheaded by Rhodri Cusack and aided by several other contributors. Recent years have seen aa mature into an extremely flexible processing environment. My own commitment to using aa was sealed at the CBU when working on our VBM comparison of 400+ subjects—with aa it was possible to run a full analysis in about a week (with 16-32 compute nodes running full time) (don't tell anyone—I think technically we weren't supposed to use more than 8...). And, because we were comparing different segmentation routines (among other things) we ran several of these analyses. Without aa I can't imagine ever doing the study. aa also played a key role in our winning HBM Hackathon entry from 2013 (or as we affectionally called it, the haackathon).

Based on my own experience I strongly recommend that all neuroimagers learn to use some form of imaging pipeline, and aa is a great choice. For most of us there is a significant upfront investment of time and frustration. However, the payoff is well worth it, both in terms of time (you will end up saving time in the long run) and scientific quality (reproducibility, openness, and fewer opportunities for point-and-click error).

The code for aa is freely available, hosted on github. Links, help, and more can be found on the main aa website: automaticanalysis.org. Comments and suggestions are very welcome, especially for the "getting started" portions (many of which are new).

By the way, several os the aa team will be at HBM this year, and we are submitting an aa poster as well. Please stop by and say hi!

Reference:

Cusack R, Vicente-Gravobetsky A, Mitchell DJ, Wild C, Auer T, Linke AC, Peelle JE (2015) Automatic analysis (aa): Efficient neuroimaging workflows and parallel processing using Matlab and XML. Frontiers in Neuroinformatics 8:90. http://journal.frontiersin.org/Journal/10.3389/fninf.2014.00090/abstract