Thursday, January 31, 2008

Auditory dorsal stream: Pa & Hickok (2008) in print

In an entry several weeks back, I argued that the auditory dorsal stream may not be auditory-specific. (See "The auditory dorsal stream may not be auditory.") I previewed an experiment by a former TalkingBrains West grad student Judy Pa (now at UCSF) in which skilled pianists were asked to listen to novel melodies and either covertly hum or imagine playing the melody during fMRI scanning. What we found was that area Spt -- previously thought to be an auditory-motor integration area -- showed a reduction in amplitude during the listen-play condition relative to the listen-hum condition. An inferior parietal region showed the reverse pattern. We concluded that Spt is not auditory-motor, but sensory-vocal tract. That is, it is a sensory-motor integration area for the vocal tract that happens to receive a lot of auditory input, but could just as well receive input from other modalities (e.g., perception of visual speech, lip-reading, activates this general area).

The paper describing this work is now in print in Neuropsychologia, 2008, 46:362-8. Check it out!

Tuesday, January 29, 2008

Semantics and Brain - 2 more problems for the ATL = semantic hub hypothesis

A previous post pointed out that the available data from studies of cortical atrophy and metabolic measures do not support the idea that semantic dementia results from focal damage to the anterior temporal lobe.

Caveat: for researchers interested in differentiating the pathologies associated with the various forms of frontotemporal dementia, AD, and related diseases, the ATL pathology in SD may very well be considered to be "focal" in the sense that the bulk of the atrophy and hypometabolism, at least in early stages of the disease, is in the anterior half of the temporal lobes. So if claims about the ATL being a semantic hub (or some similar concept) are willing to include in their definition of ATL a number of different anatomical structures and cytoarchitectonic fields, including both neocortex and limbic structures (hippocampus, amygdala), with posterior involvement including roughly half of the temporal lobe ventro-laterally, then I think SD can provide reasonable support for this idea. My own interest in the link between SD and the ATL came out of claims that used SD to argue that the lateral anterior temporal lobe (i.e., those regions corresponding to "sentence-specific" activations, and "intelligible speech" activations) was the projection site for the auditory/speech ventral stream, in contrast to the position David and I have put forward which argues for the posterior temporal lobe instead (e.g., see Scott & Wise and Hickok & Poeppel in the Special Issue of Cognition that David and I edited in 2004). The fact that SD has substantial MTL involvement and has ventro-lateral involvement that appears to extend quite a ways posterior, undermines the SD argument for an anterior-only auditory ventral stream. This is the sense in which I say SD pathology is not focal.

Caveat aside, there are still some remaining problems with associating SD and ATL dysfunction that became clear in the readings from last week.

Problem #1: Atrophy/hypometabolism is not always correlated with function.

For example, SD patients have both atrophy and hypometabolism in MTL structures -- the same MTL structures implicated in episodic memory deficits in AD, Case H.M., etc. -- yet SD patients don't have significant episodic memory impairments. Nestor, et al. (2006; NeuroImage, 30:1010-20) write, "Bilateral MTL hypometabolism in SD is, however, paradoxical since this deficit ought to be associated with episodic memory impairment." p. 1013. And regarding volumetric studies they say, "...the present observation suggests that studies aiming to corrlate MRI-derived volume loss with a given neuropsychological profile are at greater risk of producing false-positive and false negative results than previously thought." p. 1017. These authors provide reasonable explanations for these paradoxes (interaction with other damaged networks, specific pathologies involved, etc.), but the point is that these gross measures of brain structure/function are not necessarily straightforward correlates of cognitive function/dysfunction.

Other examples of this same sort of problem come from comparisons of HSVE and SD, both of which have substantial ATL involvement, but with different behavioral manifestations.

Problem #2: Case G.T.

Levy et al. (2004, PNAS, 101:6710-15) report three cases of amnesia with rather extensive bilateral damage to medial temporal lobe structures, and compare findings from these three patients with published data from SD patients on the same tests. One case in particular, G.T., is particularly interesting in light of the extent of involvement of the ATL bilaterally: "The damage involves the anterior 7 cm [!] of the left temporal lobe, [and] the anterior 5 cm of the right temporal lobe..." p. 6711. The extent of the damage can be seen clearly in the structural MRIs below shown in radiological format (left=right).

This patient should behave like an SD patient, worse even, on all the standard tests. However, G.T.'s performance was well above the standard error of the group of SD patients on tests of word recognition, picture naming, naming to description, subordinate category sorting, the Pyramids and Palm Trees test, and real/non-real object judgment. This is not to say that the patient was unimpaired on these tests relative to controls (s/he was), or that G.T. didn't have SD-like trouble on some semantic tests (e.g., category fluency for living things, yes/no questions regarding semantic features), but it is quite clear that extensive bilateral ATL+MTL damage does not produce semantic dementia.

The other two amnesic patients, whose lesions did not extend so far laterally, performed very well on all these "semantic" tests, and generally better than G.T. suggesting some role for the lateral ATL in semantic abilities. But again, the question is, why are SD patients so much worse off than G.T. who had severe damage to all the "classic" SD areas? Perhaps it is the more posterior extent of the SD pathology, together with the more anterior disruption that explains the severity of the deficit.

Overall summary from this week's readings:

1. Pathology in SD is not restricted to the ATL.
2. Structural and metabolic correlates of cognitive dysfunction are not definitive, and must be interpreted with extreme caution.
3. Medial ATL involvement in SD does not appear to account for semantic deficits (see Levy et al.).
4. Antero-lateral involvement of the ATL may account for some of the semantic impairment, but certainly not all.

My tentative conclusion, then, is that the pathology has to interrupt function in more posterior ventro-lateral portions of the temporal lobe to produce the full blown deficits in SD.

Semantics and Brain course - reading set #4

Readings for this week focus on understanding the nature of the semantic deficit in semantic dementia, starting with Elizabeth Warrington's original 1975 description. Should be an interesting set. I have more than this, but thought 7 papers was enough for a week's worth of reading. They are listed in chronological order, so subsequent readings (for next week) will be more recent. As always, please let me know of any critical omissions.

Warrington EK.
Q J Exp Psychol. 1975 Nov;27(4):635-57.
The selective impairment of semantic memory.

Hodges JR, Bozeat S, Lambon Ralph MA, Patterson K, Spatt J.
The role of conceptual knowledge in object use evidence from semantic dementia.
Brain. 2000 Sep;123 ( Pt 9):1913-25.

Bozeat S, Lambon Ralph MA, Patterson K, Garrard P, Hodges JR.
Non-verbal semantic impairment in semantic dementia.
Neuropsychologia. 2000;38(9):1207-15.

Lambon Ralph MA, McClelland JL, Patterson K, Galton CJ, Hodges JR.
No right to speak? The relationship between object naming and semantic impairment: neuropsychological evidence and a computational model.
J Cogn Neurosci. 2001 Apr 1;13(3):341-56.

Passmore MJ, Ingles JL, Fisk JD, Darvesh S.
Disconnection of language and memory in semantic dementia: a comparative and theoretical analysis.
Curr Alzheimer Res. 2005 Oct;2(4):435-48. Review.

Patterson K, Lambon Ralph MA, Jefferies E, Woollams A, Jones R, Hodges JR,
Rogers TT.
"Presemantic" cognition in semantic dementia: six deficits in search of an explanation.
J Cogn Neurosci. 2006 Feb;18(2):169-83.

Benedet M, Patterson K, Gomez-Pastor I, Luisa Garcia de la Rocha M.
'Non-semantic' aspects of language in semantic dementia: as normal as they're said to be?
Neurocase. 2006 Feb;12(1):15-26.

Monday, January 28, 2008

Semantics and Brain - Comment from Matt Lambon Ralph

The following is a "guest post" from Matt Lambon Ralph. Sounds like some cool stuff Matt. Thanks for the info! I'll respond in the "comments" box soon.


Thanks for your thoughts and queries about our recent study, utilising rTMS to probe the function of the anterior temporal lobe, particularly with regard to its role in semantic memory.

Of course our jumping off point with this line of investigation was the neuropsychological studies of semantic dementia (SD) that a number of groups, including our own, have conducted. While the exact interpretation of the SD patients is a topic of current debate (see other blog entries here, Alex Martin’s 2007 Annual Review of Psychology, Karalyn Patterson et al’s NRN paper), the simplest explanation is that damage to the ATL leads to a specific semantic impairment – by which I mean one that affects any task that requires or involves semantic memory (in reception and expressive, verbal and nonverbal domains) but not other aspects of language, perception, problem-solving, etc. The selective nature of their impairment is really striking clinically as well as in formal testing. However, given the debate about the location of damage in SD – we have embarked upon a series of rTMS studies to provide converging evidence for what the SD data seem to be indicating. To license direct comparisons we have, wherever possible, tried to utilise the same tasks and test materials that we use in SD patients but in normal participants (collecting both RT and accuracy).

A quick summary of our findings: 10 mins of 1Hz rTMS over the left temporal pole significantly slowed synonym judgement times, slowed naming at the subordinate level (“Dalmatian”) but did not affect two “control” tasks (number judgement and number naming tasks – each of which was harder than the respective semantic tasks in terms of baseline reaction times – in order to rule out any findings being due to general difficulty). Basic level naming also slowed numerically but this change was not statistically significant. This pattern of findings is pretty much what one would expect from an SD perspective – as noted above, they exhibit receptive difficulties (they are impaired on the same synonym task) and expressive impairments (they are profoundly anomic) but not other taxing, non-semantic tasks. Indeed they tend to be rather good at number judgements (involving quantity judgements).

Greg makes a couple of interesting points about the design of the study and expresses a craving for more information with regard to TMS effects on nonverbal semantic tasks. So here’s some more information that might just help:

1. Control task vs. control area: the most common form of control method in TMS studies is the “control site” method – i.e. stimulate a site of non-interest to ensure that same results do not following from any form of non-specific brain stimulation. This is quite useful if one is trying to demonstrate that a new specific region (e.g., one that is identified in an imaging study) is required in a task. For our study, however, the motivation and research question was a bit different to the standard one underpinning most rTMS experiments. Given the SD data – we were specifically interested in probing the function of the ATL itself in normal participants: i.e., if we stimulate the same region that is maximally damaged in SD does it produce the same neuropsychological pattern as the patient? If the ATL atrophy in SD is a “red herring” - i.e., is not responsible for their semantic impairment then stimulating this same region in normal participants would not be expected to have any semantic effects. (Of course we are interested in which other brain regions are, and are not involved in semantic cognition but that is another story.) Because we were interesting in how closely ATL rTMS would mimic SD, we adopted the less common “control task” method. The “logic” of this is not that dissimilar to the idea of subtractions in neuroimaging – find a pair or selection of tasks, matched for general levels of difficult, which differ in terms of the representations/process of interest. We alighted upon numbers because we know SD patients tend to be good at them and also we were able to fashion a task in which the RTs were slower than the semantic tasks. By picking number judgement and number naming as two control tasks we mirrored the receptive and expressive aspects of the two semantic tasks. By gradually building up the range of semantic and control tasks used across studies, we should be able to construct a rather comprehensive picture of which aspects of cognitive behaviour involve this region.

Of course, it is possible that some of you may not be that interested in mimicking SD data as we were and consequently would like to see a control site as well. I’m pleased to say, therefore, that we have applied the “belt-and-braces” approach to more recent experiments by including both types of control method as well as other types of control tasks – and the findings are unchanged: ATL rTMS slows semantic tasks but not number or visual matching tests, whereas stimulation to a non-ATL site does not affect any of these tasks.

2. As Greg rightly points out, the results from SD (e.g., Bozeat et al., 2000, Neuropsychologia) suggest that the ATL is the basis for amodal semantic representations. Accordingly, we would really expect to see slowing on nonverbal as well as verbal semantic tasks after ATL stimulation. He suggests using a task like the Pyramids and Palm Trees Test (in which semantic associative knowledge is probed either with picture or word materials), which we commonly use in patient testing. As it happens, this is exactly the study we have just completed – and the results are just as predicted: rTMS to the ATL slows this task in both its verbal and its picture-based versions, and to the same degree.

So by combining the data from our published study with these new results, we can mirror the combination of problems observed in SD – i.e., the ATL seems to be involved in expressive and receptive semantic tasks for verbal and nonverbal domains. This is, of course, consistent with the idea that this region develops amodal representations that combine with information in other modality-specific regions to support semantic processing (Rogers et al, 2004, Psych Review).

Matt LR

Semantics and Brain - 3rd meeting: Specificity of ATL involvement in SD

This week we read over a bunch of papers looking at the distribution of atrophy and metabolic changes in semantic dementia, as well as correlation of these measures with semantic and naming tasks. Our goal was to determine whether the semantic deficits in this syndrome can be linked to ATL dysfunction, as some have claimed. As Patterson et al. (2007, NRN, 8:976-88) put it, the strength of this view "...hinges, however, on claims about the relatively focal nature of the pathology in SD." p. 980.

So how focal is it? Not so focal. Here's a few quotes and info from the papers we read this week:

"Patients with SD as a group, compared with controls, showed hypometabolism over the whole left temporal lobe... and in the right temporal pole...." Diehl et al. 2004, Neurobiology of Aging, 25:1051-6, p. 1053 (see figure below)

"...even patients in the mild SD subgroup were found to have bilateral temporal polar and amygdalar atrophy bilaterally, plus left-sided atrophy of the hippocampus, parahippocampal, fusiform, and inferior and middle temporal gyri. In the moderate SD group, the atrophy also involved the right parahippocampal, fusiform, and inferior and middle temporal gyri." Galton, et al., 2001, Neurology, 57:216-25, p. 221

Grossman et al (2004, Brain, 127:628-49) reports that, "...significant grey matter atrophy was evident in several areas of the left emporal cortex in SD." p. 635. These included (from Table 4, p. 637), Left ventral temporal, Left anterior temporal, Left posterolateral temporal, & Left parahippocampal.

Correlational analyses are any more focal:

Grossman et al. (2004, same as above) report significant correlation between naming performance and grey matter atrophy in two left ventral temporal sites (one quite posterior involving BA 19/37), right posterolateral temporal, and bilateral occipital.

The paper by Williams et al. (2005, NeuroImage, 24:1042-51) which included SD patients as well as the frontal variant FTD patients and correlated semantic scores across all patients, reports, "The analysis of correlation with the semantic test scores yielded extensive cluster in the left temporal lobe that included temporal pole... parahippocampal gyrus and fusiform gyrus ... superior and middle temporal gyrus... and inferior temporo-occipital region." p. 1046 (see figure below)
And Galton et al. (2001, same as above) report correlations between semantic tasks and volume measurements in the left fusiform.

Conclusion: data from semantic dementia does not conclusively implicate the ATL, unless by "ATL" one is willing to include minimally (e.g., see Nestor et al. 2006, NeuroImage, 30:1010-20 for the most "focal" results of all the papers we've read), limbic structures in the medial temporal lobe (hippocampus and surrounding cortices), and ventro-lateral temporal lobe structures extending (on the left) posteriorly to include about half the entire temporal lobe.

Friday, January 25, 2008

Special issue on speech/language

There is a Special Issue of the Philosophical Transactions of the Royal Society that is now available on their web site. The issue was edited by Brian Moore, Lorraine Tyler, and William Marslen-Wilson.

The Table of Contents is copied here. There are tons of useful papers in this issue. Our group also has a paper (Poeppel, Idsardi & van Wassenhove, Speech perception at the interface of neurobiology and linguistics.)

Volume 363, Number 1493 / March 12, 2008 of Philosophical Transactions of the Royal Society B: Biological Sciences is now available on the website at

This issue contains:

Introduction. The perception of speech: from sound to meaning p. 917
Brian C.J. Moore, Lorraine K. Tyler, William Marslen-Wilson
DOI: 10.1098/rstb.2007.2195

Neural representation of spectral and temporal information in speech p. 923
Eric D. Young
DOI: 10.1098/rstb.2007.2151

Basic auditory processes involved in the analysis of speech sounds p. 947
Brian C.J. Moore
DOI: 10.1098/rstb.2007.2152

Acoustic and auditory phonetics: the adaptive design of speech sound systems p. 965
Randy L. Diehl
DOI: 10.1098/rstb.2007.2153

Phonetic learning as a pathway to language: new data and native language magnet theory expanded (NLM-e) p. 979
Patricia K. Kuhl, Barbara T. Conboy, Sharon Coffey-Corina, et al.
DOI: 10.1098/rstb.2007.2154

The processing of audio-visual speech: empirical and neural bases p. 1001
Ruth Campbell
DOI: 10.1098/rstb.2007.2155

Listening to speech in the presence of other sounds p. 1011
C.J. Darwin
DOI: 10.1098/rstb.2007.2156

Functional imaging of the auditory processing applied to speech sounds p. 1023
Roy D. Patterson, Ingrid S. Johnsrude
DOI: 10.1098/rstb.2007.2157

Fronto-temporal brain systems supporting spoken language comprehension p. 1037
Lorraine K. Tyler, William Marslen-Wilson
DOI: 10.1098/rstb.2007.2158

The fractionation of spoken language understanding by measuring electrical and magnetic brain signals p. 1055
Peter Hagoort
DOI: 10.1098/rstb.2007.2159

Speech perception at the interface of neurobiology and linguistics p. 1071
David Poeppel, William J. Idsardi, Virginie van Wassenhove
DOI: 10.1098/rstb.2007.2160

Neural specializations for speech and pitch: moving beyond the dichotomies p. 1087
Robert J. Zatorre, Jackson T. Gandour
DOI: 10.1098/rstb.2007.2161

Language processing in the natural world p. 1105
Michael K. Tanenhaus, Sarah Brown-Schmidt
DOI: 10.1098/rstb.2007.2162

Thursday, January 24, 2008

Semantics and Brain - Comment on Pobric et al. 2007

More summary of class discussion during our last meeting...

Pobric, Jefferies, & Lambon Ralph. 2007. PNAS, 104:20137-41

I got an email from Matt recently with some interesting comments and suggestions on our semantic memory posts and reading selection, so hopefully if I get anything wrong in my summary here, he will correct me.

This paper took a novel approach to determining whether the ATL is involved in semantic knowledge. They used TMS to suppress function in the left ATL. They targeted a region in the middle temporal gyrus, 10mm posterior to the tip of the temporal pole. Subjects were then asked to perform three different kinds of naming tasks, (1) basic category picture naming (dog), (2) subordinate category picture naming (poodle), and (3) number naming. They also used two kinds of judgment tasks, (1) synonym judgment, and (2) number magnitude similarity (pick the number closest in value to the target).

TMS to left ATL was found to affect subordinate category naming (slower RTs), but not basic category or number naming, and it also affected synonym but not number judgment (again showing up in RTs). It is concluded that "...the ATL plays a necessary role in semantic cognition in healthy participants." p. 20139.

This is an interesting paper. I like the approach, and there is some interesting theoretical discussion. But I'm not convinced it shows that the ATL plays a necessary role in semantic cognition. Here's why:

1. Both tasks were verbal: picture naming and synonym judgment (on sets of words). The claim being put forward is that the ATL is involved in semantic cognition generally. It would have been nice to see an effect in a non-verbal task, for example, some version of the Pyramids and Palm Trees task, which is used regularly in SD studies. (Of course, given the claims that David and I have put out there regarding the role of posterior temporal areas and lexical-semantic access, I would not expect their results to be restricted to verbal material, but we need to see the data.)

2. A TMS control condition was not utilized. It would have been nice to see TMS delivered to some other area (and one with the same amount of discomfort) to ensure that the observed effects are not some non-specific effect of TMS interacting with the particular stimuli/tasks chosen. Pobric et al. did make an explicit argument for using stimulus controls rather than a TMS control, but I didn't buy it frankly. Relatedly, I'm also not thrilled with the use of a single control task. So number-related behaviors are affected, but does this generalized to ALL non-conceptual semantic abilities? Because of the design, the interpretation of the whole study relies on how well the number tasks generalize to all non-conceptual semantic abilities, and this worries me.

Overall though, I think this study is a great start. I'm looking forward to the next one from this group.

Tuesday, January 22, 2008

New Program in Cognitive Neuroscience - UC Irvine

A new doctoral-level program in Cognitive Neuroscience has recently been established here at UC Irvine. Although it is housed in the Department of Cognitive Sciences, it is a multidisciplinary program with participation from faculty members with primary appointments in departments ranging from Neurobiology and Behavior to Radiology. The program is approved to commence with the 2008-2009 academic year. A formal announcement with links to program details will follow. In the meantime, check out the list of participating faculty:

Alyssa Brewer - Human vision, fMRI, neurology
Lawrence Cahill - Memory, Emotion, functional imaging
Nicole Gage - Development, autism, language, MEG
Emily Grossman - Biological motion, fMRI, TMS
Gregory Hickok - Speech/language, fMRI, neuropsychology
Donald Hoffman - Visual perception, EEG, fMRI
Mary-Louise Kean - Language processing, fMRI, neuropsychology
Leonard Kitzes - Mammalian auditory system
Jeffery Krichmar - Memory, vision, Computational neuroscience
David Lyon - Primate visual system
James McGaugh - Neurobiology of memory
Tugan Muftuler - fMRI, cognition
Michael Rugg - Memory, fMRI, EEG
Kourosh Saberi - Hearing, fMRI
John Serences - Attention, vision, fMRI
George Sperling - Vision, memory, attention, fMRI
Ramesh Srinivasan - Consciousness, sensory systems, EEG
Norman Weinberger - Auditory cortex physiology, plasticity, learning, and memory
Fan-Gang Zeng - Hearing, clinical audiology

Thursday, January 17, 2008

Semantics and Brain course - reading set #3: more on semantic dementia

The goal of this set of readings is to hopefully answer the following two questions:

1. How focal to the ATL is the atrophy in semantic dementia?
2. Do semantic deficits correlate specifically with atrophy or hypoperfusion in the ATL?

The answer to these questions will determine how strong the evidence is regarding the link between semantic deficits and ATL dysfunction.

I've included a bunch of papers because I'd like to get as complete a picture as we can. But no need to rush through all of them this weekend, as our next class meeting isn't until Jan. 28th. In the meantime, if I've left off any critical papers, please send me a note, either on- or off-line. I've already heard from Murray Grossman's group (via Jonathan Peelle) in this respect, and this led me to some interesting papers. Any suggestions from the London Group? I noticed a couple of hits from UCL, so I know someone over there is watching... Would very much like input from all the SD experts!

Davies RR, Graham KS, Xuereb JH, Williams GB, Hodges JR.
The human perirhinal cortex and semantic memory.
Eur J Neurosci. 2004 Nov;20(9):2441-6.

Galton CJ, Patterson K, Graham K, Lambon-Ralph MA, Williams G, Antoun N, Sahakian BJ, Hodges JR.
Differing patterns of temporal atrophy in Alzheimer's disease and semantic dementia.
Neurology. 2001 Jul 24;57(2):216-25.

Williams GB, Nestor PJ, Hodges JR.
Neural correlates of semantic and behavioural deficits in frontotemporal dementia.
Neuroimage. 2005 Feb 15;24(4):1042-51. Epub 2004 Dec 19.

Grossman M, McMillan C, Moore P, Ding L, Glosser G, Work M, Gee J.
What's in a name: voxel-based morphometric analyses of MRI and naming difficulty in Alzheimer's disease, frontotemporal dementia and corticobasal degeneration.
Brain. 2004 Mar;127(Pt 3):628-49.

Levy DA, Bayley PJ, Squire LR.
The anatomy of semantic knowledge: medial vs. lateral temporal lobe.
Proc Natl Acad Sci U S A. 2004 Apr 27;101(17):6710-5.

Seeley WW, Bauer AM, Miller BL, Gorno-Tempini ML, Kramer JH, Weiner M, Rosen HJ.
The natural history of temporal variant frontotemporal dementia.
Neurology. 2005 Apr 26;64(8):1384-90.

Diehl-Schmid J, Grimmer T, Drzezga A, Bornschein S, Perneczky R, Forstl H, Schwaiger M, Kurz A.
Longitudinal changes of cerebral glucose metabolism in semantic dementia.
Dement Geriatr Cogn Disord. 2006;22(4):346-51.

Diehl J, Grimmer T, Drzezga A, Riemenschneider M, Förstl H, Kurz A.
Cerebral metabolic patterns at early stages of frontotemporal dementia and semantic dementia. A PET study.
Neurobiol Aging. 2004 Sep;25(8):1051-6.

Nestor PJ, Fryer TD, Hodges JR.
Declarative memory impairments in Alzheimer's disease and semantic dementia.
Neuroimage. 2006 Apr 15;30(3):1010-20.

Wednesday, January 16, 2008

Mis-represented positions on semantic organization in the brain

I noted this in previous posts, but thought, for emphasis sake, it would be worth dedicating a single entry to the issue of what our (Hickok and Poeppel) claims actually are regarding "semantics." In fact, our claims regarding the role of the posterior temporal lobes are rather circumscribed. Here's some quotes:

"This [ventral] pathway appears to be important for interfacing sound-based representations of speech with widely distributed conceptual representations, and therefore is involved in tasks that require access to the mental lexicon." HP 2000, p. 131

"The ventral stream projects ventro-laterally... These pITL [posterior inferior temporal lobe] structures serve as an interface between sound-based representations of speech in STG and widely distributed conceptual representations..." HP 2004, p. 72

"Our claim is simply that there exists a cortical network which performs a mapping between (or binds) acoustic-phonetic representations of the one hand, and conceptual-semantic representations on the other." HP 2004, p. 81

"The more posterior regions of the ventral stream, posterior middle and inferior portions of the temporal lobes correspond to the lexical interface, which links phonological and semantic information..." HP 2007, p. 395

While our terminology changes slightly -- reflecting the fact that we don't really know what's being computed -- we have been very clear that this is not a model of semantic memory/conceptual knowledge organization, but rather a claim about the area that is particularly important for interfacing such information (located elsewhere) with acoustic/phonological representations of speech.

Despite this, we have come across examples of researchers generalizing our claims to a theory of semantics. Here's one from the semantic dementia readings:

"Although the data arising from semantic dementia clearly implicate the temporal poles, bilaterally, in semantic representation, these areas are often overlooked or even disputed in other research on semantic memory [HP 2007 cited here]." Pobric et al., 2007, p. 20137

And here is an older one from Friedemann Pulvermuller's lab:

"In contrast to other authors who suggest that semantics is represented in meaning-specific brain regions that process all words alike (Hickok and Poeppel, 2000; Lichtheim, 1885; Price et al., 2001; Scott and Johnsrude, 2003; Wernicke, 1874), we proposed that semantic representations are distributed in a systematic way throughout the entire brain." Hauk, et al., 2004, p. 305.

I won't speak for Price et al., or Scott and Johnsrude, but I will defend the dead guys, Wernicke and Lichtheim. Not only did Hauk et al. mis-attribute such a claim about semantic representation to us, they completely missed the boat even on these classical authors. Here's a quote from Wernicke, 1874/1977, p.117:

"The concept of the word 'bell', for example, is formed by the associated memory images of visual, tactual and auditory perception. These memory images represent the essential characteristic features of the object, bell."

And a quote from Wernicke's later work, 1885-1886/1977, p. 177:

"... the memory images of a bell ... are deposited in the cortex and located according to the sensory organs."

For details on Wernicke's rather sophisticated distributed theory of conceptual representation of the brain, including a new translation into English of a paper by Wernicke on the topic, see Gage & Hickok, 2005. Seriously, check it out -- you will be surprised to learn that Wernicke even postulated "Hebbian learning" mechanisms decades before Hebb.

So Wernicke clearly did not believe in a focal meaning specific brain area. Now to defend Lichtheim -- here's a quote from his 1885 paper, On Aphasia, in the journal, Brain, p. 477:

"Though in the diagram B is represented as a sort of centre for the elaboration of concepts, this has been done for simplicity's sake; with most writers, I do not consider the function to be localised in one spot of the brain, but rather to result from the combined action of the whole sensorial sphere."

The point here is not so much to criticize anyone for misquoting -- it is easy to misquote, and I'm sure we are all guilty -- rather the point is (i) to make it perfectly clear what David and I are actually claiming w.r.t. semantics because I think some people are missing the distinction between a semantic memory system and an interface between this system and speech, and more importantly, (ii) to (re-)emphasize that when you are talking about "semantics", be very clear regarding what part of semantics you're talking about.


Gage, N., & Hickok, G. (2005). Multiregional Cell Assemblies, Temporal Binding, and the Representation of Conceptual Knowledge in Cortex: A Modern Theory by a "Classical" Neurologist, Carl Wernicke. Cortex, 41, 823-832.
Hauk, O., Johnsrude, I., & Pulvermuller, F. (1994). Somatotopic representation of action words in human motor and premotor cortex. Neuron, 41, 301-307.
Hickok, G., & Poeppel, D. (2000). Towards a functional neuroanatomy of speech perception. Trends in Cognitive Sciences, 4, 131-138.
Hickok, G., & Poeppel, D. (2004). Dorsal and ventral streams: A framework for understanding aspects of the functional anatomy of language. Cognition, 92, 67-99.
Hickok, G., & Poeppel, D. (2007). The cortical organization of speech processing. Nat Rev Neurosci, 8(5), 393-402.
Lichtheim, L. (1885). On aphasia. Brain, 7, 433-484.
Pobric, G., Jefferies, E., & Lambon Ralph, M.A. (2007). Anterior temporal lobes mediate semantic representation: Mimicking semantic dementia by using rTMS in normal participants. PNAS, 104:20137-41.
Wernicke, C. (1874/1977). Der aphasische symptomencomplex: Eine psychologische studie auf anatomischer basis. In G. H. Eggert (Ed.), Wernicke's works on aphasia: A sourcebook and review (pp. 91-145). The Hague: Mouton.
Wernicke, C. (1885-1886/1977). Einige neuere Arbeiten ueber Aphasie. In G. H. Eggert (Ed.), Wernicke’s works on aphasia: A sourcebook and review. The Hague: Mouton.

Tuesday, January 15, 2008

Semantics and Brain - more on the ATL as a hub

Here's a figure from the Patterson et al. reading which depicts the two views of the neural organization of semantic memory as conceptualized by the authors:

The top panel is the "distributed-only" view in which conceptual bits are widely distributed, are stored in sensory and motor areas relevant to the concept, and are bound together via connections between the various areas. The point is that there is no single area coordinating the linkages between these conceptual bits. The bottom panel is their alternative, "distributed plus hub" view, which also holds that conceptual bits are widely distributed in sensory-motor areas, BUT that the linkages area coordinated by a single "hub" living in the ATL.

To put my cards on the table, I kind of like the hub idea with respect to the ATL (I don't like it w.r.t. the planum temporale though -- future post). Like I said before, I'm not convinced yet that the evidence is all that strong, but I like the idea nonetheless. Here's something to think about though:

Suppose the evidence does pan out that the ATL is critically involved in the semantic deficits found in semantic dementia. Can we conclude that the architecture in the bottom panel of the above figure is correct? Not necessarily, as pointed out in our class by Mary Louise Kean. Just because a single region is implicated in some function, doesn't mean that computationally that region as a whole performs a single computation function. For example, it could be that the ATL contains parallel circuits (convergence zones, say) each performing a similar integrative function but across their own idiosyncratic domains. The parallel circuits in the basal ganglia are a model for this kind of architecture.

Monday, January 14, 2008

Semantics and Brain - 2nd meeting

Here's a few random highlights from our second meeting. I'll try to post more throughout the week.

Hodges & Patterson, 2007. Semantic dementia: a unique clincopathological syndrome. Lancet, 6: 1004-14. This is an excellent overview of semantic dementia. Start here if you are interested in learning about SD.

Patterson, et al., 2007. Where do you know what you know? The representation of semantic knowledge in the human brain. Nature Reviews Neuroscience, 8: 976-987. An clear, informative paper that makes the best argument I've seen for the anterior temporal lobe being critically involved in the representation of conceptual semantic knowledge. They argue that the anterior temporal lobe is an amodal "hub" that binds together modality specific bits of semantic information. They refer to this as the "distributed-plus-hub" view. This is a reasonable hypothesis, but I'm not yet convinced the data supports this view, particularly the anatomical claim.

The crux of the claim boils down to this, stated succinctly by Patterson et al.,

"The cognitive and neuroanatomical abnormalities in SD, provide trenchant [nice adjective!] evidence for the distributed-plus-hub view... The strength of this evidence hinges, however, on claims about the relatively focal nature of the pathology in SD." p. 980.

This is what I'm not so sure about, whether we can really tie the semantic deficits in SD to ATL dysfunction. I'm not opposed to the idea, in fact, it makes a lot of intuitive sense, but with so many people so unquestioningly enthusiastic about the association between SD and ATL dysfunction, it is worthwhile taking a very critical look at the data.

Here's a couple of reasons to remain cautious:

1. Herpes simplex virus encephalitis (HSVE). Patients present primarily with episodic memory impairment, not the characteristic semantic memory impairment found in SD, and when semantic memory impairment occurs in HSVE it is relatively mild. Yet, the ATL bilaterally is implicated in both SD and HSVE. In fact, according to at least one study (Noppeney et al. 2007, Brain, 130:1138-47), ATL dysfunction appears more prominent (more of the ATL involved) in HSVE than SD generally, except for a small more posterior region that was more affected in SD than HSVE. This seems problematic.

2. Pathology in SD extends beyond the ATL according to most studies. This is a bit of a complicated issue because it likely depends on how much the disease has progressed at the time of the anatomical or metabolic scan, but it is certainly NOT the case that SD involves the ATL and only the ATL. We will be looking into this issue in more depth for next week.

One thing I am sure about, though, is that the kind of "semantic" processing that David and I talk about in our papers, and that we associated with posterior temporal systems, and the kind of semantic processing that the SD folks are talking about are very different. Our claims are modality specific: a mapping between sound and meaning. Claims from the SD literature are modality independent: the deficit is thought to affect semantic knowledge at some central level. So our position, that posterior temporal regions are involved in mapping sound to meaning, is not at all incompatible with the possibility that the ATL is functioning as some sort of semantic hub.

Friday, January 11, 2008

Web based picture rating study

We (Stephen Wilson and I) are collecting concept familiarity ratings for a set of pictures using in an fMRI study of naming. We would very much appreciate it if you could run through and rate the set for us. And tell your friends, students, advisors, etc.! We promise to share data when the web study is complete...

Thank you!

-Greg and Stephen

Thursday, January 10, 2008

What's wrong with adjectives in science journals?

I just finished writing a review response -- always a good exercise to generate blog topics. The reviews were helpful in most respects, micro-managerial in others. One micro-edict was to remove terms like "strong" and "solid" from the discussion, as in "strong left dominance." These terms were deemed "unnecessarily strong." Yes, that's right, "strong" is too strong. This tendency in scientific writing -- to avoid enthusiasm shall we call it -- is not unique to this reviewer. One journal Ursula Bellugi and I submitted a paper to actually edited out the word "remarkable" (one of Ursie's favorites) because the very word itself was against journal policy. No joke.

What up? If a finding is indeed remarkable, why can't we say so? I know, I know, it's not up to the author to decide what is remarkable or strong; that is something the reader can decide for themselves. We need to write objectively, right? Well that's junk. Almost everything we write in a scientific article is subjective in one way or another. We frame the experiment in terms of the theoretical issues we deem important, and we interpret our findings according to our own theoretical, intuitive, and yes emotional biases. We can't help it, we're human, and it's ok. Other researchers are free to examine our findings and impose their own biased interpretations. That's how science works. If we wanted to be perfectly objective, we'd have to publish just the methods and results sections, or maybe just methods and raw data (analysis is subject to subjectivity too).

Disallowing words like remarkable doesn't eliminate subjectivity, it induces the characteristic robotic monotony of scientific prose. At least now, journals are allowing authors to refer to themselves in the first person. I hate writing (or reading) things like "this author has claimed..." Bluck! What an exciting step forward for us pocket-protected nerds that we get to write the word "I" in our publications. Woohoo! Think how much more fun it would be to write or read an article that not only is informative scientifically, but maybe even mildly entertaining, or at the very least include the use of actual adjectives that hinted at the human behind the keystrokes. That's what I like about blogs: I can use as many adjectives as I want, as well as the word "bluck"...

So did I cave and edit out "strong" and "solid"? Nope. Instead I made a (rather nerdly) argument that "strong" was scientifically justified. Funny thing is, I actually used restraint in using the phrase "strong left hemisphere dominance." What I really wanted to say was "freakin' strong and totally remarkable left hemisphere dominance."

Stay tuned to see what actually comes out in print. That is, if I didn't just kill my chances of getting a favorable re-review. (No really, I thought the review was quite helpful generally...)

Wednesday, January 9, 2008

Semantics and Brain course - reading set #2: semantic dementia

Having determined that semantics is complicated, our next topic is semantic dementia, a syndrome in which "semantic knowledge" is thought to be disrupted. There are a TON of papers on semantic dementia. For this week, we're going to start, somewhat arbitrarily, with the papers listed below. If anyone has any suggestions regarding critical papers that we can't afford to miss, please post a comment.

Patterson K, Nestor PJ, Rogers TT.
Where do you know what you know? The representation of semantic knowledge in the
human brain.
Nat Rev Neurosci. 2007 Dec;8(12):976-87. Review.
PMID: 18026167 [PubMed - indexed for MEDLINE]

Hodges JR, Patterson K.
Semantic dementia: a unique clinicopathological syndrome.
Lancet Neurol. 2007 Nov;6(11):1004-14. Review.
PMID: 17945154 [PubMed - indexed for MEDLINE]

Pobric G, Jefferies E, Ralph MA.
Anterior temporal lobes mediate semantic representation: mimicking semantic
dementia by using rTMS in normal participants.
Proc Natl Acad Sci U S A. 2007 Dec 11;104(50):20137-41. Epub 2007 Dec 3.
PMID: 18056637 [PubMed - in process]

Noppeney U, Patterson K, Tyler LK, Moss H, Stamatakis EA, Bright P, Mummery C,
Price CJ.
Temporal lobe lesions and semantic impairment: a comparison of herpes simplex
virus encephalitis and semantic dementia.
Brain. 2007 Apr;130(Pt 4):1138-47. Epub 2007 Jan 24.
PMID: 17251241 [PubMed - indexed for MEDLINE]

Desgranges B, Matuszewski V, Piolino P, Chételat G, Mézenge F, Landeau B, de
la Sayette V, Belliard S, Eustache F.
Anatomical and functional alterations in semantic dementia: a voxel-based MRI and
PET study.
Neurobiol Aging. 2007 Dec;28(12):1904-13. Epub 2006 Sep 15.
PMID: 16979268 [PubMed - indexed for MEDLINE]

Tuesday, January 8, 2008

Semantics and Brain course - 1st meeting summary

Our Winter quarter started this week, and the first meeting of our grad seminar on Semantics and the Brain was yesterday. In addition to a number of fabulous grad students, we have the participation of several faculty members, including Kent Johnson (whose got a very cool "world clock" on his homepage), and new UCI Cog Sci faculty, Lisa Pearl and Jon Sprouse.

Here are the high points of our discussion:

1. Semantics is complicated. Ok, not the deepest of insights, but it's worth acknowledging. The neuroscience community (Talking Brains included!) is notorious for oversimplifying "semantics." It is not hard to find references to terms such as "semantic knowledge" as if it were a single unitary thing. David and I talk about "mapping between sound and meaning" as if this were a simple computational task. In fact "mapping between sound and meaning" encompasses pretty much the totality of the field of linguistics. A good reminder of how little we know!

2. Semantics is complicated. Here's some of the ways:

(i) The meaning of words (lexical concepts) is hard (impossible) to define in terms of necessary and sufficient conditions. This is true even for concrete objects. What's a "chair"? Can't be defined by shape (think, beanbag), can't be defined a something you can sit on (think, swing or horse).

(ii) The meaning behind verbs includes a lot of structure (i.e., argument structure). Sleep requires that there be a sleeper, kick requires that there be a kicker and a kickee, and put requires that there be a putter, a thing that is put, and a place that the thing is put. The "arguments" that a verb takes can themselves vary in complexity. For example, you can kick a football, but you can't think a football, you have to think something more complex, I though ABOUT footballs, etc.

(iii) Word meanings are likely decomposable. To get an intuitive sense of this, consider the difference in meaning between walk, strut, run, swagger, and stagger. All contain a bit of meaning roughly equal to ambulate, but differ in the manner of ambulation. So we might characterize these words as ambulate+manner. (The semantics literature is full of arguments for decomposition of meaning.)

(iv) There may be multiple dimensions of meaning. For example, "The BEAR chased the lion" (there are multiple possible chasers) means something different than "The bear chased the LION" (there are multiple possible chasees), even though the propositional structure is the same.

We could go on to discuss quantifier scope, fuzzy boundaries and family resemblance, part-whole relations, reference transfer ("Chomsky is on the shelf next to Plato" means "the book by Chomsky..."), etc.

The point is that when us neuroscience types think we've identified the neural substrate of "semantics" we are barely scratching the surface, AND we're probably wrong about what we've scratched. Consider the claim that the meaning of action words is encoded in motor cortex (e.g., Hauk et al. 2004). Just because motor cortex for the hand area lights up when people read the word throw, this does not mean that this chunk of tissue is coding anywhere near the depth of the meaning of throw.

3. Semantics is complicated. When thinking about the neural basis of "semantics" we have to be careful about which aspect(s) of semantics we are talking about.

Hauk O, Johnsrude I, Pulvermüller F.
Somatotopic representation of action words in human motor and premotor cortex.
Neuron. 2004 Jan 22;41(2):301-7.