15th Neural Computation and Psychology Workshop

NCPW15 – August 8-9, 2016 – Philadelphia, PA, USA

15^th Neural Computation and Psychology Workshop

Contemporary Neural Network Models:
Machine Learning, Artificial Intelligence, and Cognition

Reproducibility project: A front row seat

A recent paper in Science reports the results of a large-scale effort to test reproducibility in psychological science. The results have caused much discussion (as well they should) in both general public and science forums. I thought I would offer my perspective as the lead author of one of the studies that was included in the reproducibility analysis. I had heard about the project even before being contacted to participate and one of the things that appealed to me about it was that they were trying to be unbiased in their selection of studies for replication: all papers published in three prominent journals in 2008.  Jim Magnuson and I had published a paper in one of those journals (Journal of Experimental Psychology: Learning, Memory, & Cognition) in 2008 (Mirman & Magnuson, 2008), so I figured I would hear from them sooner or later. 

Aphasia factors vs. subtypes

One of the interesting things (to me anyway) that came out of our recent factor analysis project (Mirman et al., 2015, in press; see Part 1 and Part 2) is a way of reconsidering aphasia types in terms of psycholinguistic factors rather than the traditional clinical aphasia subtypes.

The traditional aphasia subtyping approach is to use a diagnostic test like the Western Aphasia Battery or the Boston Diagnostic Aphasia Examination to assign an individual with aphasia to one of several subtype categories: Anomic, Broca's, Wernicke's, Conduction, Transcortical Sensory, Transcortical Motor, or Global aphasia. This approach has several well-known problems (see, e.g., Caplan, 2011, in K. M. Heilman & E. Valenstein (eds) Clinical Neuropsychology, 5th Edition, Oxford Univ. Press, p. 22 - 41), including heterogeneous symptomology (e.g., Broca's aphasia is defined by co-occurrence of symptoms that can have different manifestations and multiple, possibly unrelated causes) and the relatively high proportion of "unclassifiable" or "mixed" aphasia cases that do not fit into a single subtype category. And although aphasia subtypes are thought to have clear lesion correlates (Broca's aphasia = lesion in Broca's area; Wernicke's aphasia = lesion in Wernicke's area), this correlation is weak at best (15-40% of patients have lesion locations that are not predictable from their aphasia subtype). 

Our factor analysis results provide a way to evaluate the classic aphasia syndromes with respect to data-driven performance clusters; that is, the factor scores. Our sample of 99 participants with aphasia had reasonable representation of four aphasia subtypes: Anomic (N=44), Broca's (N=27), Conduction (N=16), and Wernicke's (N=8); 1 Global and 3 TCM are not included here due to small sample size. The figure below shows, for each aphasia subtype group, the average (+/- SE) score on each of the four factors. Factor scores should be interpreted roughly like z-scores: positive means better-than-average performance, negative means poorer-than-average performance.

Credit: Mirman et al. (in press), Neuropsychologia

At first glance, the factor scores align with general descriptions of the aphasia subtypes: Anomic is a relatively mild aphasia so performance was generally better than average, participants with Broca's aphasia had production deficits (both phonological and semantic), participants with Conduction aphasia had phonological deficits (both speech recognition and speech production), and Wernicke's aphasia is a more severe aphasia so these participants had relatively impaired performance on all factors that was particularly pronounced for the semantic recognition factor. However, these central tendencies hide the tremendous amount of overlap among the four aphasia subtype groups for each factor. This can be seen in the density distributions of exactly the same data:

As one example, consider the top left panel: the Wernicke's aphasia group clearly had the highest proportion of participants with poor semantic recognition, but some participants in that group were in the moderate range, overlapping with the other groups. Similarly, the other panels show that it would be relatively easy to find an individual in each subtype group who violates the expected pattern for that group (e.g., a participant with Conduction aphasia who has good speech recognition). This means that the group label only provides rough, probabilistic information about an individual's language abilities and is probably not very useful in a research context where we can typically characterize each participant's profile in terms of detailed performance data on a variety of tests. Plus, as our papers report, unlike the aphasia subtypes, the factors have fairly clear and distinct lesion correlates.

In clinical contexts, one usually wants to maximize time spent on treatment, which often means trying to minimize time spent on assessment and a compact summary of an individual's language profile can be very useful. Even so, I wonder if continuous scores on cognitive-linguistic factors might provide more useful clinical guidance than an imperfect category label.

Mirman, D., Chen, Q., Zhang, Y., Wang, Z., Faseyitan, O.K., Coslett, H.B., & Schwartz, M.F. (2015). Neural Organization of Spoken Language Revealed by Lesion-Symptom Mapping. Nature Communications, 6 (6762), 1-9. DOI: 10.1038/ncomms7762.
Mirman, D., Zhang, Y., Wang, Z., Coslett, H.B., & Schwartz, M.F. (in press). The ins and outs of meaning: Behavioral and neuroanatomical dissociation of semantically-driven word retrieval and multimodal semantic recognition in aphasia. Neuropsychologia. DOI: 10.1016/j.neuropsychologia.2015.02.014.

Mapping the language system: Part 2

This is the second of a multi-part post about a pair of papers that just came out (Mirman et al., 2015, in press). Part 1 was about the behavioral data: we started with 17 behavioral measures from 99 participants with aphasia following left hemisphere stroke. Using factor analysis, we reduced those 17 measures to 4 underlying factors: Semantic Recognition, Speech Production, Speech Recognition, and Semantic Errors. For each of these factors, we then used voxel-based lesion-symptom mapping (VLSM) to identify the left hemisphere regions where stroke damage was associated with poorer performance. 

Mapping the language system: Part 1

My colleagues and I have a pair of papers coming out in Nature Communications and Neuropsychologia that I'm particularly excited about. The data came from Myrna Schwartz's long-running anatomical case series project in which behavioral and structural neuroimaging data were collected from a large sample of individuals with aphasia following left hemisphere stroke. We pulled together data from 17 measures of language-related performance for 99 participants, each of those participants was also able to provide high-quality structural neuroimaging data to localize their stroke lesion. The behavioral measures ranged from phonological processing (phoneme discrimination, production of phonological errors during picture naming, etc.) to verbal and nonverbal semantic processing (synonym judgments, Camel and Cactus Test, production of semantic errors during picture naming, etc.). I have a lot to say about our project, so there will be a few posts about it. This first post will focus on the behavioral data.

When lexical competition becomes lexical cooperation

Lexical neighborhood effects are one of the most robust findings in spoken word recognition: words with many similar-sounding words ("neighbors") are recognized more slowly and less accurately than words with few neighbors. About 10 years ago, when I was just starting my post-doc training with Jim Magnuson, we wondered about semantic neighborhood effects. We found that things were less straightforward in semantics: near semantic neighbors slowed down visual word recognition, but distant semantic neighbors sped up visual word recognition (Mirman & Magnuson, 2008). I later found that the same pattern in spoken word production (Mirman, 2011). Working with Whit Tabor, we developed a preliminary computational account. Later, when Qi Chen joined my lab at MRRI, we expanded this computational model to capture orthographic, phonological, and semantic neighborhood density effects in visual and spoken word recognition and spoken word production (Chen & Mirman, 2012). The key insight from our model was that neighbors exert both inhibitory and facilitative effects on target word processing with the inhibitory effect dominating for strongly active neighbors and the facilitative effect dominating for weakly active neighbors.

In a new paper soon to be published in Cognitive Science (Chen & Mirman, in press) we test a unique prediction from our model. The idea is that phonological neighborhood effects in spoken word recognition are so robust because phonological neighbors are consistently strongly activated during spoken word recognition. If we can reduce their activation by creating a context in which they are not among the likely targets, then their inhibitory effect will not just get smaller, it will become smaller than the facilitative effect, so the net result will be a flip to a facilitative effect. We tested this by using spoken word-to-picture matching with eye-tracking, more commonly known as the "visual world paradigm". When four (phonologically unrelated) pictures appear on the screen, they provide some semantic information about the likely target word. The longer they are on-screen before the spoken word begins, the more this semantic context will influence which lexical candidates will be activated. At one extreme, without any semantic context, we should see the standard inhibitory effect of phonological neighbors; at the other extreme, if only the pictured items are viable candidates, there should be no effect of phonological neighbors. Here is the cool part (if I may say so): at an intermediate point, the semantic context reduces phonological neighbor activation but doesn't eliminate it, so the neighbors will be weakly active and will produce a facilitative effect. 

We report simulations of our model concretely demonstrating this prediction and an experiment in which we manipulate the preview duration (how long the pictures are displayed before the spoken word starts) as a way of manipulating the strength of semantic context. The results were (mostly) consistent with this prediction. 

At 500ms preview (middle panel), there is a clear facilitative effect of neighborhood density: the target fixation proportions for high density targets (red line) rise faster than for the low density targets (blue line). This did not happen with either the shorter or longer preview duration and is not expected unless the preview provides semantic input that weakens activation of phonological neighbors, thus making their net effect facilitative rather than inhibitory.

I'm excited about this paper because "lexical competition" is such a core concept in spoken word recognition that it is hard to imagine neighborhood density having a facilitative effect, but that's what our model predicted and the eye-tracking results bore it out. This is one of those full-cycle cases where behavioral data led to a theory, which led to a computational model, which made new predictions, which were tested in a behavioral experiment. That's what I was trained to do and it feels good to have actually pulled it off.

As a final meta comment: we owe a big "Thank You" to Keith Apfelbaum, Sheila Blumstein, and Bob McMurray, whose 2011 paper was part of the inspiration for this study. Even more importantly, Keith and Bob shared first their data for our follow-up analyses, then their study materials to help us run our experiment. I think this kind of sharing is hugely important for having a science that truly builds and moves forward in a replicable way, but it is all too rare. Apfelbaum, Blumstein, and McMurray not only ran a good study, they also helped other people build on it, which multiplied their positive contribution to the field. I hope one day we can make this kind of sharing the standard in the field, but until then, I'll just appreciate the people who do it.

Apfelbaum K S, Blumstein S E, & McMurray B (2011). Semantic priming is affected by real-time phonological competition: evidence for continuous cascading systems. Psychonomic Bulletin & Review, 18 (1), 141-149 PMID: 21327343
Chen Q, & Mirman D (2012). Competition and cooperation among similar representations: toward a unified account of facilitative and inhibitory effects of lexical neighbors. Psychological Review, 119 (2), 417-430 PMID: 22352357
Chen Q, & Mirman D (2015). Interaction Between Phonological and Semantic Representations: Time Matters. Cognitive Science (in press) PMID: 25155249
Mirman D (2011). Effects of near and distant semantic neighbors on word production. Cognitive, Affective & Behavioral Neuroscience, 11 (1), 32-43 PMID: 21264640
Mirman D, & Magnuson J S (2008). Attractor dynamics and semantic neighborhood density: processing is slowed by near neighbors and speeded by distant neighbors. Journal of Experimental Psychology: Learning, Memory, and Cognition, 34 (1), 65-79 PMID: 18194055

The mind is not a (digital) computer

The "mind as computer" has been a dominant and powerful metaphor in cognitive science at least since the middle of the 20th century. Throughout this time, many of us have chafed against this metaphor because it has a tendency to be taken too literally. Framing mental and neural processes in terms of computation or information processing can be extremely useful, but this approach can turn into the extremely misleading notion that our minds work kind of like our desktop or laptop computers. There are two particular notions that have continued to hold sway despite mountains of evidence against them and I think their perseverance might be, at least in part, due to the computer analogy.

The first is modularity or autonomy: the idea that the mind/brain is made up of (semi-)independent components. Decades of research on interactive processing (including my own) and emergence have shown that this is not the case (e.g., McClelland, Mirman, & Holt, 2006; McClelland, 2010; Dixon, Holden, Mirman, & Stephen, 2012), but components remain a key part of the default description of cognitive systems, perhaps with some caveat that these components interact.

The second is the idea that the mind engages in symbolic or rule-based computation, much like the if-then procedures that form the core of computer programs. This idea is widely associated with the popular science writing of Steven Pinker and is a central feature of classic models of cognition, such as ACT-R. In a new paper just published in the journal Cognition, Gary Lupyan reports 13 experiments showing just how bad human minds are at executing simple rule-based algorithms (full disclosure: Gary and I are friends and have collaborated on a few projects). In particular, he tested parity judgments (is a number odd or even?), triangle judgments (is a figure a triangle?), and grandmother judgments (is a person a grandmother?). Each of these is a simple, rule-based judgment, and the participants knew the rule (last digit is even; polygon with three sides; has at least one grandchild), but they were nevertheless biased by typicality: numbers with more even digits were judged to be more even, equilateral triangles were judged to be more triangular, and older women with more grandchildren were judged to be more grandmotherly. A variety of control conditions and experiments ruled out various alternative explanations of these results. The bottom line is that, as he puts it, "human algorithms, unlike conventional computer algorithms, only approximate rule-based classification and never fully abstract from the specifics of the input."

It's probably too much to hope that this paper will end the misuse of the computer metaphor, but I think it will be a nice reminder of the limitations of this metaphor.

Dixon JA, Holden JG, Mirman D, & Stephen DG (2012). Multifractal dynamics in the emergence of cognitive structure. Topics in Cognitive Science, 4 (1), 51-62 PMID: 22253177
Lupyan, G. (2013). The difficulties of executing simple algorithms: Why brains make mistakes computers don’t. Cognition, 129(3), 615-636. DOI: 10.1016/j.cognition.2013.08.015
McClelland, J.L. (2010). Emergence in Cognitive Science. Topics in Cognitive Science, 2 (4), 751-770 DOI: 10.1111/j.1756-8765.2010.01116.x
McClelland JL, Mirman D, & Holt LL (2006). Are there interactive processes in speech perception? Trends in Cognitive Sciences, 10 (8), 363-369 PMID: 16843037

Graduate student positions available at Drexel University

The Applied Cognitive and Brain Sciences (ACBS) program at Drexel University invites applications for Ph.D. students to begin in the Fall of 2014. Faculty research interests in the ACBS program span the full range from basic to applied science in Cognitive Psychology, Cognitive Neuroscience, and Cognitive Engineering, with particular faculty expertise in computational modeling and electrophysiology. Accepted students will work closely with their mentor in a research-focused setting, housed in a newly-renovated, state-of-the-art facility featuring spacious graduate student offices and collaborative workspaces. Graduate students will also have the opportunity to collaborate with faculty in Clinical Psychology, the School of Biomedical Engineering and Health Sciences, the College of Computing and Informatics, the College of Engineering, the School of Medicine, and the University's new Expressive and Creative Interaction Technologies (ExCITe) Center.

Specific faculty members seeking graduate students, and possible research topics after the jump.