J. Devlin1,2, C. Moore1,3, C. Mummery1, J. Phillips1, M. Gorno-Tempini1, U. Noppenney1, K.

Friston1, R. Frackowiak1, and C. Price1

1Wellcome Department of Cognitive Neurology2Functional Magnetic Resonance Imaging of the Brain

3Institute of Psychiatry

Is there an anatomical basis category specificity?

Cognitive theories of category specificity

Semantic theories

Evolutionary account: Individual categories recruit distinct neural regions. (Caramazza & Shelton, 1998)

Perceptional/functional account: Perceptual and functional attributes localised in different regions. (Warrington & Shallice, 1984)

Cognitive structure account: Undifferentiated semantic system at a neural level (Tyler et al., in press)

Cognitive theories and neuroimaging

Anatomical hypotheses can be best evaluated by functional neuroimaging, however, the findings to date are very variable Living things > man-made items -- only three areas identified in more

than one study: 1) bilateral anterior temporal poles(refs), 2) right inferior parietal lobe (refs), and 3) left lingual gyrus (refs)

Man-made items > living things -- three areas seen in more than one study: 1) left posterior middle temporal cortex (refs), 2) left pre-motor cortex (refs), and 3) left lingual gyrus (refs)

All other areas have been seen in only a single study

We hypothesized that this variability was due to small, context-specific effects and investigated this using a multi-factorial analysis

Design

Data from 7 PET studies of category specificity

Study Categories Stimuli Task

1. Mummery et al (1996) A, T Spoken words Category fluency

2. Mummery et al (1998) A, T Written words Semantic & syllable

decisions

3. Moore & Price (1999a) A, F, T, V Pictures Naming

4. Moore & Price (1999b) A, F, T, V Written words Matching

and pictures

5. Moore & Price (1999b) A, F, T, V Pictures Naming

6. Gorno-Tempini (2000) Fa, A, T Pictures Naming

7. Phillips et al (submitted) F, T Written words Semantic & screen

and pictures size decisions

Abbreviations: A=animals, F=fruit, Fa=famous faces, T=tools, V=vehicles.

Analysis

Single multi-factorial analysis with three factors:

1) Category (man-made vs. living things)

2) Task (semantic vs. non-semantic)3) Stimulus type (words vs. pictures)

to evaluate the effects of category in different contexts

Data collected from 60 (53M, 7F) native English speakers on a single PET scanner.

Results

No main effects of Category and no significant interactions with Stimulus type

Highly significant Category x Task interactions

Tools > Living things for semantic tasks only

Peak in left posterior middle temporal gyrus at (-62 -58 0), SPM{Z}=5.30, p=0.005 (corrected)

z = 0

L L R

Contrasts1. Location decisions2. Color decisions3. Syllable decisions4. Action decisions5. Real Size decisions6. Screen size decisions7. Category fluency8. Naming color pictures9. Naming b/w pictures10. Naming pictures11. Reading words12. Word-Picture matching

Effect size (%rCBF)

L. post. middle temporal gyrus(-62 -58 0)

Manmade - Living Contrasts

2.5

2.0

1.5

1.0

0.5

0.0

-0.5

-1.0

-1.5 1 2 3 4 5 6 7 8 9 10 11 12

Semantic tasksNonsemantic tasks

Consistent tool advantage

Word-picture matching Picture naming

L. post. middle temporal gyrus(-62 -58 0)

KeyA AnimalsF FruitV VehiclesT ToolsB BaselineCN C. N. objectsSN S. N. objects

A F V T B A F V T CN SN

Relativeeffect sizes

Tool-specific? No.

Tools > living things and vehicles BUTSimple nonobjects also activated this region

T F T F Faces A T Body Parts

Relativeeffect sizes

Semantic decisions Picture naming

Action Perceptual

{ {

KeyA AnimalsF FruitT Tools

Combine this slide into last one?

In addition, this region is activated by:1) Action decisions (for both tools and fruit) and2) Naming pictures of body parts

Living things > manmade items(Tasks requiring identification only)

z = -8 y = 8 z = -24

R. anterio-medial TP extending along medial surface. SPM{Z}=5.2, p=0.01 (corrected)

L. insula extending into anterio-medial TP. SPM{Z}=4.1, p=0.086 (corrected)

R RL RLL

Consistent living things advantage

Effect size

(%rCBF)

Nam

ing

pict

ures

Ack

now

ledg

ing

Rea

ding

Ack

now

ledg

ing

Act

ion

deci

sion

sR

eal S

ize

deci

sion

sS

cree

n si

ze d

ecis

ions

Loc

atio

n de

cisi

ons

Col

or d

ecis

ions

Syl

labl

e de

cisi

ons

Nam

ing

colo

r pi

ctur

esN

amin

g b/

w p

ictu

res

W-P

mat

chin

gN

amin

g pi

ctur

esC

ateg

ory

flue

ncy

R. Ant. Temp. Pole(24 8 -24)

L. Ant. Temp. Pole(-30 6 -18)

6

5

4

3

2

1

0

-1

-2

-3

-4

6

5

4

3

2

1

0

-1

-2

-3

-4

Living - Manmade Contrasts

Requires identificationOther tasks

Left TP(-30 6 -18)

Right TP(24 8 -24)

Relativeeffect sizes

A F V T B A F V T Faces A T BP

W-P matching Picture naming Picture naming

Category-specific? No. Separate into 2 slides??

Animals, fruit, and faces all activate these regions

*

Left TP(-30 6 -18)

Right TP(24 8 -24)

Relativeeffect sizes

A F V T B A F V T Faces A T BP

W-P matching Picture naming Picture naming

Category-specific? No. Separate into 2 slides??

Animals, fruit, and faces all activate these regions

*

Summary

Robust evidence of anatomical specialisation for semantic categories: Tools activated left posterior middle temporal region Living things activated anterio-medial temporal poles

Small, context-specific effects: Tool activation only in semantic tasks Living thing activation for tasks requiring

identification

Q: How do these findings relate to cognitive theories of category specificity?

Conclusions

Results illustrate clear specialisation within the semantic system by demonstrating an anatomical double dissociation between living things and manmade items

Inconsistent with accounts based on an undifferentiated semantic system

Conclusions

Results were not specific to individual categories “Tool area” responded to action decisions for fruit,

naming body parts, and pictures of non-objects Temporal poles were activated by animals, fruit, and

famous faces

Inconsistent with evolutionary account of neural regions specialised for individual categories

Conclusions

Results were most consistent with distinct regions for different types of information

Posterior middle temporal region responds to actions associated with graspable objects

Anterior temporal poles may respond to greater semantic integrating information RLL

Replace thiswith an axialslice through antTPs (z=-20)

References