lang and space

7/28/2019 Lang and Space

http://slidepdf.com/reader/full/lang-and-space 1/7

55OpinionOpinion TRENDS in Cognitive Sciences Vol.5 No.2 February 2001

http://tics.trends.com 1364-6613/01/$ – see front matter © 2001 Elsevier Science Ltd. All rights reserved. PII: S1364-6613(00)01598-9

Opinion

Is la ngua ge relat ed to our knowledg e of space? Thepurpose of langua ge would seem to be different from

the purpose for w hich we represent spa ce. Lang uage

mirr ors the contours of our thought 1 an d provides a

means t o communicate. Lang uage lets us encode

ma ssive am ounts of informa tion and generat e

complex ideas tha t w ould oth erw ise be impossible.

Spat ial representat ions mirror t he contours of our

external environment and provide a mea ns to reach,

search and na vigate2. Given t hese differences, one

might expect la ngua ge and spa ce to be segregated

in our cognitive system s.

In w ha t follows, I suggest tha t despite differences

in the neural a nd menta l organizat ion of langua geand space, these two domains intera ct. I review some

wa ys in w hich the spatia l context of communicat ion

can influen ce the production a nd comprehension of

langua ge. Then I review w ay s in which lang uage as a

symbolic system is likely to engage spat ial

representa tions wh en one thinks of objects a nd

events in the environment.

Neuroanatomy of language and space

At a first glance, neuroana tomical observations are

consistent with the idea that language and space are

segregated. Langua ge and space are mediat ed

primarily by different cerebra l hemispheres3.

P rofound impairments in langua ge are associated

with left hemisphere dama ge, and profound

impairments of spatia l representat ions are

associat ed with right hemisphere da ma ge.

The importa nce of lang ua ge an d space in our

menta l lives is reflected in the am ount of brain tha t is

dedicat ed to these cognitive domains. B oth langua ge

and spa ce are mediat ed by widely distributed neural

networks4. Cortically, these netw orks include the

posterior temporal-parieta l region, an d dorsolatera l

and medial prefronta l regions. Sub-cortically, t hey

include parts of the basal ga nglia and thalam us.

Distributed langua ge networks in the left hemisphere

media te components of la ngua ge, such as phonology,

lexical-semant ics and synta x. Distributed spatial

networks in t he right hemisphere mediat e components

of space, such as reference frames a nchored to the

retina, hea d or trunk, and spa tial locat ions indexed to

movements of different body part s. Thus, the net works

that mediate la nguage and space are similarly

organized, but la rgely in different hemispheres.

Despite these broad d ifferences in the neuroan at omy

of lan gua ge and spa ce, their segrega tion is unlikely to

be absolute. A lan gua ge netw ork completely

encapsulated from sensat ions w ould imply a radically

different neural organiza tion in the left a nd right

hemispheres. Prima ry sensory an d motor cortices

connect to higher order netw orks in a reciprocal

cascade4. Un imodal cortices process elementar y

sensat ions, wh ich then combine with informa tion from

other sensory m odalities to form m ore complex

representa tions. For example, neurons in th e posterior

parietal a nd dorsolatera l prefrontal cortex of macaque

monkeys a re especially responsive to combinat ions ofvisual a nd t actile stimuli t ied to movements of specific

body part s2. Evidence from bra in dam aged pat ients

suggest tha t crossmodal an d sensory–motor

informat ion similarly converge in huma ns5,6 giving rise

to th e phenomenological experience of a unified spat ial

environment in wh ich we perceive and a ct.

Why w ould the cascade of sensory informa tion

tha t modulates act ivity in the r ight temporal-

parieta l cortex not d o so in t he left? The syna ptic

connections between prima ry sensory cortices and

posterior temporal-parieta l cortex a re similar in

both hemispheres. It seems unlikely tha t sensory

informat ion, which accumulat es into complexspatia l representa tions in the right, would

completely dissipate in t he left . Alterna tively, the

sensory informa tion in the left hemisphere might

also modulate temporal-parieta l activity, but

differently from in the right , a possibility to which

we w ill return lat er. Differences in such modulat ion

ar e probably m ediat ed by hemispheric differences in

the dendritic patt erns and n euronal physiology7–9 .

Language as a means of communication

La ngua ge as a complex system of commun ication

includes verbal production a nd comprehension, a s

well as gestures, emotional prosody and t he

conventions of conversa tion. I w ill touch on thr ee

settings in w hich linguistic commun ication interacts

with space. Firs t , in American Sign La nguage

(ASL), informat ion is commun icated spa tia lly.

Second, some words r efer explicitly t o spatia l

informat ion. Third, a n issue discussed in grea ter

deta il, the direction of space in w hich some speakers

orient ma y influence their lan gua ge.

American Sign Language (ASL )

ASL conveys informa tion spatia lly using a system of

gestur es. Both h emispheres seem to be involved in

Language and space:some interactionsAnjan Chatterjee

Is language linked to mental representations of space? There are several

reasons to think that language and space might be separated in our cognitive

systems,but they nevertheless interact in important ways.These

interactions are evident in language viewed as a means of communication

and in language considered a form of representation.In communication,

spatial factors may be explicit in language itself,such as the spatial-gestural

system of American Sign Language.Even the act of conversing with others is

a spatial behavior because we orient to the locations of other participants.

Language and spatial representations probably converge at an abstract level

of concepts and simple spatial schemas.

Anjan Chatterjee

Dept of Neurology and

the Center for Cognitive

Neuroscience,

The University of

Pennsylvania,

3West Gates, 3400 Spruce

Street, Philadelphia,

PA 19104, USA.

e-mail: anjan@

mail.med.upenn.edu



http://tics.trends.com

56 OpinionOpinion

processing sentences in ASL, in contrast to En glish.

In functional neuroimaging studies, when subjects

read w ritten E nglish sentences, left B roca’s and

Wernicke’s ar eas a re activa ted. B y contra st, viewing

films of signers producing ASL sent ences activa tes

right posterior regions in add ition to the left

hemisphere langua ge ar eas10.

Ea ch hemisphere is probably mediat ing different

kinds of informa tion in ASL. In ASL, space is used to

commun icate both topographic informa tion about

the environment a nd gra mma tical rela t ionships.

Da ma ge to the right hemisphere of ASL speakers

can produce deficits in expressing spat ial

topogra phies, wh ereas da ma ge to the left

hemisphere can produce deficits in expressing

gramma tical rela t ionships11,12. I w ill not discuss

these intera ctions of lan gua ge and spa ce in

ASL further (see Refs 12,13). I simply highlight the

point th at space can play different roles in ASL

commun ication. Different neural substra tes

seem to mediate t he topographic and gra mma tical

uses of space.

Deixes

In conversation, speakers often anchor their

utt eran ces to their spatia l environment . This

an choring is r eferred to as ‘deixis’, or pointing w ith

words14,15. Deictic expressions can iden tify objects in

space, as in th e demonstra tive determiners ‘t h i s ’ or

‘that ’ . These deictic expressions a nd t heir

comprehension are ba sed on knowledge shared by

the part icipant s of the conversa tion and t he context

in w hich th e utt eran ces occur. Locative prepositions

such a s ‘above ’, and ‘behind ’ , which convey explicit

spatia l informa tion, also serve as deixes. The spatia lrelat ionship of objects m ay be anchored to th e

speaker, such as in ‘th e li ght i s above my h ead ’

or to another object in th e environment , as in ‘the

li ght i s above the table ’.

Read ing sentences with locat ive prepositions

activa tes part s of both pa rieta l cortices16. Because the

parietal cortex mediat es spatial representations,

comprehending these sentences appear s to involve

spat ial processing. The left hemisphere ma y be critical

in processing locat ive prepositions: left hemisphere

da ma ge is more likely tha n right h emisphere to produce

deficits in comprehending t hese prepositions17,18.

Spatial orientation

In conversat ion, we orient towa rds others in our

spatia l environment. Coslett a nd co-workers report

tha t the direction in which some apha sic patients

orient influences their use of langu a ge. They initia lly

observed that a pa tient w ith ischemic infarcts of the

left temporal-parieta l and left a nterior cingulate

regions w as poorer a t und erstan ding spoken

langua ge and producing words when orienting to his

right th an w hen orienting to his left 19. He was a lso

slower a t na ming pictures, and rea d single words

more poorly when st imuli were locat ed to his right

tha n to his left . A contra lesiona l at tentiona l deficit

wa s unlikely to account for his behavior because

spatia l orienta tion influenced his lang uag e even

wh en there were no externa l stimuli to be

apprehended. When generat ing nouns and

narra t ing a fa iry ta le, he was less fluent and his

story wa s less deta iled when he oriented to the right

tha n w hen he oriented to the left .

Coslett found similar spa tia l effects on langua ge

in a group of 30 individuals w ith single hemisphere

ischemic infa rcts20. Language w as a ssessed using

naming to confronta t ion, ora l rea ding and mat ching

aud itory words to pictures. Five individuals’

performa nces on some of these langua ge ta sks were

influenced by t heir spat ial orienta tion. All five had

da ma ge to the parieta l cortex. In a follow-up study

with 52 pat ients , Coslet t and Lie21 showed similar

effects of spatia l orienta tion on na ming, read ing,

synonym judgement an d sentence comprehension.

All nine people whose langua ge wa s influenced by

the direction of space into which th ey oriented had

left parieta l cortex dama ge.

The spatial registration hypothesis

Why sh ould the direction of space in w hich an

individual orients influence lan gua ge? Coslett

proposes t he ‘spat ial r egistra tion’ hypothesis20.

He a rgues tha t registering objects and events in

space is of funda menta l evolutionary im porta nce.

This registra tion determines an orga nism’s ability

to acquire sustenance and a void danger in the

environment. The locat ions of a ll stimuli a re

registered aut omatically, even w hen this

informa t ion is irrelevant to the ta sk a t hand. An

example of such automa tic registrat ion is the Simoneffect 22. I f a red target requires a r ight ha nd

response, then su bjects respond more q uickly to a

red ta rget on the r ight t han on the left , even though

ta rget location is irrelevan t to response.

Coslet t cla ims th at spat ia l regis tra t ion effects

extend beyond sensory an d motor processing to

cognit ive operat ions . Parieta l dam age impairs

contra lesional spat ia l regis tra t ion and

consequent ly impairs th e activity of even

non-spat ial operat ions like lexical retrieva l and

semantic search. The neural act ivity media t ing

langua ge is probably modulated by head a nd eye

posit ion, s imilar to the wa y in w hich tact i le

processing is influenced by head an d eye position 23.

Crossmodal (ta ct i le–visual) integrat ion in t he

posterior pariet a l cortex24 may be accompanied by

cross-ma teria l (spat ial–linguist ic) integr at ion in

the posterior left parieta l cortex.

Language as a form of representation

Langua ge as a system of symbols needs to be able to

refer to spat ial informa tion, such as the geometr y of

spat ial relat ionships, spat ial perspectives, the

separa tion of figure from ground, and t he dyna mics of

force25. However, the forma ts of linguistic and spat ial

TRENDS in Cognitive Sciences Vol.5 No.2 February 2001Opinion




57OpinionOpinion

representa tions seem to be different. La ngua ge is

usua lly algebra ic an d can convey propositional

informa tion. Individual w ords relat e arbitra rily to the

objects and event s in the w orld. For exam ple, nothing

about the word ‘dog ’ refers necessar ily to a specific

class of objects. These objects could just a s ea sily be

referred to by another word, and certainly a re in

different la nguages26. Similarly, t he structure of a

sentence need not bear a n ecessar y relat ionship to the

structu re of objects and event s in the world.

Spat ia l representat ions , in contra s t to langua ge,

are usua lly ana log a nd convey geometricinforma tion. They comprise mult iple levels, from

early un imodal sensa tions to complex multimodal

representat ions. Spa tia l representa tions often

approximate t he topogra phy of physical space. For

example, spat ial neglect is a disorder in which

pat ients with focal brain da mage are una wa re of

objects and events in spa ce contr ala tera l to their

lesion 27. These pat ients a lso frequently n eglect

contr alesional par ts of imagined visua l scenes28.

Given t hese differences in t he format s of linguistic

and spatial representations, how might language a nd

space interact? At issue is whether linguistic

descriptions of spatia l relationships are st ructured by

perceptions29–31. In t he next section, I w ill touch on the

notion tha t langua ge intera cts with space at simple

spatia l primitives, or schemas (see Box 1), a long t he

theoretica l lines developed by Ta lmy 32 a nd

J ackendoff33 (also see Miller and J ohnson-La ird34).

Then I discuss in grea ter deta il empirical evidence

suggesting tha t t he linguistic representa tion of events

relates to spatial schemas.

The interface of language and spatial representations

Different kinds of informa tion represented in t he

brain can be chara cterized by gradients a long

several para meters. Informat ion can be perceptua l

or conceptua l, geometric or algebra ic, sensorial or

amodal, a nd concrete or abstra ct . Spat ia l

representat ions tend to be perceptual, geometric

and sensoria l , whereas la nguage tends t o be

conceptual, algebra ic an d am odal. However, both

language a nd spat ia l representa t ions can be

concrete or a bstra ct. At the concrete end of

lang uag e, sounds a nd vocalizations specific to

individual la ngua ges form w ords and sentences. At

the a bstra ct end, concepts encode mean ing in a w ay

tha t is not restricted to the idiosyncra sies of anypart icular lan gua ge (see Ref. 33 for thoughts on th is

conceptual st ructure). At the concrete end of spatia l

representat ions, perceptions are derived from a ctual

spatia l scenes. At th e abstra ct end, simple spat ial

schemas a re extracted from but do not directly

reflect perceptua l informa tion. Langua ge and space

ar e likely to converge at the a bstra ct levels of

conceptual structures and spat ial schemas.

What are spatia l schemas? Ba sed on his ana lysis

of locat ive prepositions Ta lmy 32 proposes that spatial

schemas a re ‘boiled down’ featur es of a spa tia l scene.

For example, ‘across ’ refers to a schema t ha t describes

a specific pat h of movement. This pat h is

approximat ely perpendicular to the principle axis of

the r eference object, as in across a river or a cross a

plank. When a movement proceeds par a llel to the

principle a xis of the r eference object, th en ‘along ’ is

more appropriate. B oth ‘across ’ and ‘along ’ ar e

abstra cted from th e actual scene. In these schemas

only selective spa tia l aspects are deemed relevant.

Other a spects of the scene, such as w hether t he

referent object is in fa ct a river or a pla nk ar e not

relevant and are not incorporated into the schema.

Thus, t he schemas ar e simple geometric forms such

as points, lines and planes.


One approach to investigating the meaning embodied in words

is to examine how words might decompose into constituent

primitivesa. ‘Primitives’ refer to elemental properties that cannot

be further simplified. J ackendoff suggests that the conceptual

structure of verbs decomposes into primitives such as

‘movement’, ‘path’ and ‘location’. He suggests that these

primitives must somehow correspond with their linguistic

counterpartsb.

Spatial primitives, or ‘schemas’, may play a critical role in the

acquisition of concepts. Infants first learn perceptual–motor

principles about objects and events in the world. These

principles, presumably encoded as primitives, serve as the

basis for more elaborate conceptual structures. For

example, Mandlerc suggests that infants first acquire

knowledge of different kinds of motion in the world. Biological

motion is self-propelled and non-biological motion is induced

externally. Awareness of this distinction serves as the basis

for knowledge of animacy and inanimacy, a fundamental

semantic distinction.

Even if spatial primitives form the basis by which concepts

are acquired, it is not clear that these primitives remain relevant

after the concept has been acquired. They could very well be

vestigial and be discarded. Alternatively, spatial primitives might

underlie different domains of cognition. For example,

Christmand reported that pictures with a left-to-right

directionality are judged more aesthetically pleasing than

pictures with a right-to-left directionality. Spatial primitives

might also be concatenated to form more elaborate mental

models with spatial propertiese.

References

a Levin, B. and P inker, S. (1991) Introduction to Special Issue on lexical and

conceptua l semant ics.Cognition 41, 1–7

b J ackendoff, R. (1996) The ar chitecture of t he linguistic-spatia l interfa ce. In

Lan guage and Space (Bloom, P . et al ., eds), pp.1–30, MIT P ress

c Mandler, J .M. (1996) Preverbal representation an d language. In Language

and Space (Bloom, P. et al ., eds), pp. 365–384, MIT Pr ess

d Christma n, S. an d Pinger, K (1997) Lateral biases in pictorial preferences:

pictorial dimensions and neura l mechanisms. Laterality 2, 155–175

e J ohnson-Laird, P .N. (1996) Space to think. In Lan guage and Space

(Bloom, P. et al ., eds), pp. 437–462, MIT P ress

Box 1.Spatial primitives,conceptual development,and mental models




58 OpinionOpinion

Talmy identifies several importa nt fea tures of these

schemas32. Spatia l schemas a re discrete, rather tha n

being continuous. A movement cann ot be 30%‘along’

an d 70%‘a cross’, for example. Consequent ly schema s

lose some of the precision of perception. Schema s ar e

also topological ra ther t han imagistic. One might

remember a ferry moving across a river, but most

features of this image a re not incorporat ed in the schema

of ‘across’. Rat her, schema s encode spat ial fea tur es in

simple qualitative ways ra ther tha n w ith the metrics and

richness of specific ima ges. Schema s captur e only some

of the infinite possible spat ial configura tions. This design

seems a precondition of communication, in wh ich a w ide

variety of spat ial situa tions need to be described rapidly.

Thus spat ial schema s shar e properties associat ed

with both langua ge and perceptual representations.

They a re discrete and r eferential like most elements in

langua ge, and m ight be concatenat ed to form more

complex structures. And they ar e an alog (albeit simple),

like the perceptions from which they a re extra cted.

Spatial schemas in an aphasic subject

Working within a completely different t ra dition,

Cha tterjee and co-workers reported evidence tha t

conceiving events an d actions are relat ed to spat ial

schemas. This invest igat ion began w ith a man with

agr am ma tism (see B ox 2) wh ose production and

comprehension of sentences w as influenced

systematically by spat ia l factors35,36. He had

flaw less comprehension of single words an d a

superior vocabula ry. H owever, this spontan eous

speech wa s synta ctically disorgan ized. He rar ely

produced complete sentences and his ut tera nces had

few inflections, a uxiliary verbs, a nd closed class

words. H e described his problems: ‘Well , uh,

essent ia ll y lan guage abandon pr eposit ion. I

telegraph… I , I… consciously, uh, continu ity…I , I,

uh, th is subtle of pr eposit i onal ph rases this simpl y

cannot do. Un der stress, un der stress ra pi d I ju st

fl ustered … but conti nu e to do basicall y .’

The influence of space on this pa tient’s langu age

emerged w hen he wa s assessed for his a bili ty to

express or understand w ho does wha t t o whom in

sentences (thema tic role assignment ). In d escribing

pictures, he w as more likely to describe the figure on

the left a s the a gent regardless of whether this

figure wa s the doer or the recipient of the a ction.

He a lso used a s imilar spat ia l s tra tegy on an


Agrammatism is an aphasic syndrome with impairments at the

level of sentencesa. People with agrammatism do not speak

fluently, and putting words together in sequence requires effort.

Their spontaneous speech is often ‘telegraphic’. They

communicate with simple phrases, such as ‘dog eat ’ rather than

‘the dog is eating .’ They generally comprehend simple

statements, but often have difficulty comprehending

grammatically complex sentences. People with agrammatism

omit function words (like prepositions, articles and

conjunctions), more often than nounsb.

The specific deficits in agrammatic patients vary, reflecting

the selective vulnerability of linguistic processes involved in

constructing and comprehending sentencesc. However, most

people with agrammatism fall into two broad categories. Some

have difficulties with the relationship of words to each other

(syntactic deficits). Others have difficulties processing

grammatical morphemes (morphological deficits)b. Syntactic

and morphologic deficits often co-exist but may dissociate.

Saffran Schwartz and Marind drew attention to a group of

agrammatic patients with syntactic deficits who could notprocess thematic roles in sentences. These patients can

usually use general knowledge of the world to match sentences

to pictures. For example, a boy may kick a stone, but a stone

cannot kick a boy. However, reversible sentences, such as

‘The boy kisses the girl ‘ and ‘The girl kisses the boy ‘, describe

events that are both possible. Patients with thematic role

assignment deficits are especially prone to making errors with

reversible sentences.

Investigators at the turn of the last century, such as Arnold

Pick, assumed that producing sentences involves transforming a

pre-linguistic message into language in discrete stagese. Recent

models of sentence production, such as the influential one

advocated by Garrettf , also postulate cascading levels of

representation, each with different operations. Garrett

proposes a ‘message’ level prior to ‘functional’ and ‘positional’

levels in sentence production. The message level contains pre-

linguistic information. The functional level selects abstract

lexical items and establishes the argument structure of who is

doing what to whom. The positional level inserts the appropriate

grammatical morphemes. Neurolinguists have focused on the

functional and positional levels. These levels might be

considered the ‘language proper’ aspects of sentence

production. Thematic role assignment deficits occur at the

functional levelg,h. Little about the message level is known,

although some patients may have deficits at this pre-linguistic

level following left brain damagei.

References

a Chatterjee, A. and Maher, L. (2000) Gra mmar and a grammatism.

In Aphasia an d L anguage: Theory t o Practice (Gonza lez Rothi, L. et al ., eds),

pp.133–156, Gu ilford P ress

b Goodglass, H. (1993)Un derstandi ng Aphasia , Academic Pr ess

c Ber ndt , R. (1987) Sy mptom co-occurence and dissociat ion in theinterpretation of agrammatism. In Th e Cogni ti ve Neuropsychology of

Language (Coltheart, M. et al ., eds), pp.221–233, Erlbaum

d Sa f fr an , E .M .et al . (1980) The word order problem in a gra mma tism: I .

Comprehension. Brain L ang. 10, 249–262; II. Pr oduction. Brain L ang.

10, 263–280

e Pick, A. (1931)Aphasia , Thoma s

f Ga rrett, M.F. (1982) Production of speech: observations from norma l and

pathological la nguage use. In Normalit y and Pathology in Cognit ive

Functions (Ellis, A.W., ed.), pp.19–76, Acad emic P ress

g Car ama zza, A. and Miceli, G. (1991) Selective impairment of themat ic role

assignment in sentence processing. Brain L ang. 41, 402–436

h M ah er , L . et a l . (1995) Agra mma tic sent ence production: th e use of a

temporal-spatial strategy. Brain L ang. 49, 105–124

i Chatterjee, A. et a l . (1999) Conceptua l and lingu istic knowledge of thema tic

roles in aphasia . Neurology 52, A458

Box 2.Agrammatism and thematic role assignment




59OpinionOpinion

ana gram ta sk. When presented w ith w ords on

individual card s, he ordered the words int o active

sentences that were gram mat ically a ppropria te.

However, with passive sentences, he inva riably

produced sema ntically impossible sentences,

such a s ‘Th e gir ls are cli mbed by th e stai rs.’ Hepicked ‘g i r l s ’ as the a gents and placed that card on

the left of the sentence, impervious t o the sentence’s

synta ctic construction.

This individua l’s spat ial bia s in describing

pictures extended t o comprehension, as eviden ced by

his ma tching of sentences to pictures35. With a ctive

sentences he was far m ore accurate if the agent wa s

on the left and the a ction moved left-to-right t ha n th e

agent w as on the right and t he action moved

right-to-left . By contra st, w ith passive sentences

his performa nce wa s reversed.

Cha tt erjee and co-workers speculat ed tha t th eir

subject’s spat ial bia ses might r eflect a primitivestructure of menta l representa tions of events.

Hughlings J ackson in the nineteenth century

viewed the nervous system as being orga nized

hierar chically, w ith high er processes inhibiting

lower ones37. J ackson thought t ha t ‘dissolution’ of

higher functions released m ore primitive behaviors.

Accordingly , th e dissolution of our subject’s

l inguis t ic abili t ies by brain dama ge might ha ve

released a primitive pre-linguistic representa tion,

ma king the underlying spat ial schema explicit .

Spatial schemas in normal people

If events a re encoded wit h spatia l schema s, then

subtle spatia l biases might a lso influence norma l

subjects’ conception of a ctions an d them a tic roles.

Cha tt erjee and co-workers found tha t norma l

right-handed subjects tend to locate a gents to the

left of patients, a nd t o conceive of a ctions a s

proceeding from left t o right 38,39. These bia ses

emerged in several experiments: w hen subjects

drew events in response to sentences; when th ey

drew eith er the a gent or the recipient of the action

in response to sentences or phra ses; and w hen

they d rew tra jectories of actions conveyed by

verb phra ses.

As En glish is read from left to right, could th ese

spatia l biases be produced by ha bitua l exposure to

Eng lish? Perha ps, but the results of one experiment

ar e not explained ea sily by the surfa ce structur e of

wr itten E nglish. This experiment ca pitalized on the

fact th at different verbs convey opposite spatia l

trajectories39. The ver b ‘push ’ conveys an action

moving aw ay from the agent , whereas th e verb

‘pul l ’ conveys an a ction moving towa rds th e agent.

Normal subjects ma tched sentences they hea rd to

pictures fa ster w hen pictures depicted the agent on

the left an d wit h the a ction proceeding from

left-to-righ t (see Fig. 1). The influ ence of the

direction of action is not a ccounted for by th e

surface structure of English sentences. If these

subjects simply ma pped the subject–verb–object

sentence structur e onto the agent –action–patient

depiction in pictures, t hey w ould not ha ve processed

a ctions from left-to-righ t more qu ickly. When t he

direction of act ion proceeds from left-to-righ t, th e

subject–verb–object sentence sequen ce ma ps

onto ag ent–a ction–pat ient depictions with ‘push ’verbs, but to pat ient–action–agent depictions

wi th ‘pul l ’ verb s.

Speculations about the neural bases for spatial schemas

Why should a ctions correspond t o a schema w ith a

left-to-right tra jectory? We encount er event s m oving

in every direction, so the perceptual experience of

events in the environment would not produce a

schema w ith a specific direction. Perha ps these

directional bia ses follow from properties of left

hemisphere processing.

The left hemisphere seems critical to media ting

actions in general. Da ma ge to the left hemisphere isassociated wit h apra xias, or deficits of mean ingful


Agent on left

Action left-to-right

Action right-to-left

Agent on right

TRENDS in Cognitive Sciences

Fig. 1. Examples of visual

stimuli used in the

sentence–picture

matching task. One of the

following pictures

appeared on a computer

monitor after the subjectsheard sentences such as

‘The circle pushes the

square ’ or ‘The circle

pulls the square ’.

(Modified from Ref. 39.)

• What neurophysiological and neuroanatomical

properties predispose ensembles of neurons

to encode either linguistic or spatial

information?

• What parts of the brain mediate interactions of

language and space, and what are the

consequences of damage to these areas?

• Can the effects of spatial orientation on

language be used to rehabilitate aphasic

patients?

• Is there a limited set of spatial schemas, and

how are these extracted from imagistic

representations?

• Are there differences in the neural mediation of

verbs and locative prepositions and their

spatial schemas?

• How do cultural and biological variables

contribute to spatial schemas?

• Are spatial schemas used in cognitive domains

other than language?

Outstanding questions




60 OpinionOpinion

actions40. Da mage t o the left pa rieta l cortex can a lso

impair the a bili ty to prepare for and switch to

different kinds of actions, a form of motor

at tent ion41. The left hem isphere a lso directs

at tent ion with a left-to-right vector42. Thes e

featur es of the left hemisphere, the encoding of

act ions an d the deployment of spat ia l a t tent ion with

a left -to-righ t vector, migh t predispose the left

hemisphere to mediat e a left-to-right schema for

act ions (see a lso Box 3).

One possible explan at ion is tha t the left an d right

hemispheres tend to encode different kinds of

spat ia l representat ions43; the left mediates

schematic a nd t he r ight imagis t ic representa t ions.

La ngua ge that r elies on schema tic representa tions,

as in relat ional concepts encoded in verbs and

locat ive prepositions, might rely on an inta ct left

hemisphere. Conversely, langua ge tha t relies on

imagis t ic representa t ions, as in the spat ia l

topogra phy expressed in ASL, might rely more on an

inta ct right hemisphere (see Fig. 2).

Conclusion

Despite reasons to think that mental

representa t ions of langua ge and space are

likely t o be segrega ted, th ese cognitive doma ins

ma ke conta ct a t critical junctures. These points of

conta ct are evident in both t he commun ication and

the representa tion of lang uag e. The intera ctions

reviewed here suggest tha t langua ge and

space are not modular cognitive systems in th e

strong sense of being informa tionally

encapsulated from each other. Rather, a t certa in

points t he informat ion from one domain bleeds

into the other. Ca reful considera tion of these

points of conta ct is likely to reveal insight s

into how th oughts relate to actions and events

in the environment .


[EVENT] "run"

Scene Perceptual, imagisticrepresentation

Spatialschema

Conceptual structure Verbalrepresentation

TRENDS in Cognitive Sciences

Fig. 2. A general sketch of the relationship of spatial scenes in the

environment and their mental representations. The perceptual or

imagistic representation is geometric and reflects sensory features

specific to the actual scene. The spatial schema abstracts a simplified

form that retains an analog structure. The conceptual structure is

language-like in that it is algebraic and can convey propositional

information. The verbal representation encodes the actual word

representing the spatial scene.

Acknowledgements

I would like to thank

Lisa Santer and H.Branch

Coslett for their critical

reading of this paper.

Considerable evidence suggests that the brain

processes nouns and verbs differentlya–c. Lesion

studies suggest that noun retrieval deficits are

associated with left temporal lesions and verb

retrieval deficits are associated with left premotor

lesions. This neural differentiation is not

surprising, given that nouns and verbs have

different semantic and syntactic properties. Nouns

prototypically refer to objects in the world, while

verbs prototypically refer to actions in the world.

Verbs play a syntactic role in setting up the

argument structure of sentences, a role not played

by nouns. Lesion and fMRI studies suggest that the

meaning of nouns is linked to sensory features.

According to this view, access to the meaning of

nouns automatically, and perhaps necessarily,

activates brain structures that are also used to

perceive the object referred to by the noun. (For a

critique of this view, see Ref. d.)

Verb retrieval deficits following brain damagehave not been studied in the same detail as noun

retrieval deficits. Verb retrieval deficits are

associated with agrammatic production, because

verbs play a critical role in setting up the structure of

the sentence (but see also Ref. e). The syntactic

properties of verbs may be processed in different

brain regions from their semantic properties.

Furthermore, verbs that belong to different semantic

categories, such as those describing actions

(e.g. run ) versus those describing mental states

(e.g. love ), might also have different neural

underpinnings.

References

a Da masio, A.R. and Tran el, D. (1993) Nouns and verbs ar e

retrieved with differently distributed neural systems.

Proc.N atl. Acad. Sci. U. S. A. 90, 4957–4960

b Bernd t , R. S . et a l . (1997) Verb retr ieval in apha sia:

1.Cha racterizing single word impairments. Brain L ang.

56, 68–106

c Gr ossman, M. (1998) Not all words are created equal:

category-specific deficits in central nervous system disease.

Neurology 50, 324–325

d Ca ram azza , A. and Shelton, J . (1998) Domain-specific

knowledge systems in the brain: the animat e–inanimat edistinction. J . Cogn. Neurosci. 10, 1–34

e Bernd t , R . et a l . (1997) Verb retrieva l and sen tence

processing: dissociation of a n esta blished symptom

association. Cortex 33, 93–114

Box 3.Nouns and verbs




61OpinionOpinion

References

1 Chatterjee, A. et al . (1997) Tha lam ic though t

disorder: on being ‘a bit add led’. Cortex 33, 419–440

2 Colby, C.L. (1998) Action oriented spat ial

reference frames in cortex. Neuron 20, 15–24

3 Heilman, K.M. et a l . (1993)Clinical

Neuropsychology (Heilman, K.M. and Valenstein,

E., eds), Oxford U niversity P ress

4 Mesulam, M-M. (1998) From sensa tion to

cognition. Bra in 121, 1013–10525 L ad avas , E . et a l . (1998) Neuropsychological

evidence of an integra ted visuotactile

representation of peripersonal spa ce in humans.

J. Cogn. Neurosci. 10, 581–589

6 Vaishnavi, S. et a l . (in press) Binding personal

and peripersonal space: evidence from ta ctile

extinction. J. Cogn. Neurosci.

7 Ja cob s, B . et a l . (1993) A quantit at ive dendritic

ana lysis of Wernicke’s Area in huma ns: II.

Gender, hemispheric, and environmenta l factors.

J. Comp. Neurol. 327, 97–111

8 Anderson, B . an d Rutledge, V. (1996) Age and

hemispheric effects on dendritic st ructure. Brain

119, 1983–1990

9 Hutsler, J .J. a nd Ga zzaniga, M.S. (1996)

Acetylcholinesterase sta ining in huma n a uditory

and la nguage cortices: regional va riation of

structural features. Cereb. Cortex 6, 260–270

10 Bavelier , D. et al . (1998) Hem ispher ic

specialization for English a nd ASL: left invaria nce

– right variability. NeuroReport 9, 1537–1542

11 Emmorey, K. et a l . (1995) Different ial processing

of topographic and referential functions of space.

In L anguage, Gestur e and Sp ace (Emmorey, K.

and Reilly, J .S., eds), pp.43–62, Erlbaum

12 Poizner, H. et a l . (1987)What th e H and Reveals

About the Brai n , MIT Press

13 Emm orey, K. an d Reilly, J . (1995)Language,

Gestu re and S pace, Erlbaum

14 Levelt, W. (1989) Speaking: From Intention to

Ar t icu lat ion,MIT Press

15 Ga rnha m, A. (1989) A unified theory of the m eaning

of some spatial r elational terms. Cognition 31, 45–60

16 Carpenter, P. et al . (1999) Time-course of fMR I-

activation in langauge and spatial networks during

sentence comprehension. NeuroImage 10,216–224

17 Schwartz , M.F. et a l . (1980) The w ord or der

problem in agra mma tism: I. Comprehension.

Brain L ang. 10, 249–262

18 Frederici, A. (1982) Synta ctic and sema ntic

processing in a phasic deficits: the ava ilability ofprepositions. Brain Lang. 15, 249–258

19 Coslett, H.B. et al . (1993) Multi-modal hemis pat ial

deficits a fter left hemisphere str oke. Brain

116, 527–554

20 Coslett, H. (1999) Spat ial influences on motor and

langua ge function.Neuropsychologia 37, 695–706

21 Coslett, H. a nd Lie, E. (2000) Spatia l influences

on language performance. J . Cogn. Neurosci.

(Suppl.)p. 76

22 Umilta , C. a nd Niccoletti, R. (1990) Spat ial-response

compat ability. In Stim ul us–Response Compati bil ity

(Proctor, R . a nd R eeve, T., eds), pp.331–50, Els evier

23 Gross, C.G . and G ra ziano, M.S.A. (1995) Multiple

representations of space in the brain. Th e

Neuroscientist 1, 43–50

24 Vaishnavi, S. et a l . (1999) Crossmoda l an d

sensorimotor integra tion in ta ctile aw areness.

Neurology 53, 1596–1598

25 Talmy , L. (1988) Force dynamics in langua ge and

cognition. Cognit . Sci. 12, 49–100

26 Bock, K.et a l . (1992) From conceptua l roles t o

structura l relations: bridging the synta ctic cleft.

Psychol. Rev. 99, 150–171

27 Heilman, K.M. et al. (1993) Neglect a nd rela ted

disorders. In Cli ni cal Neuropsychology

(Heilman, K.M. and Valenstein, E., eds),

pp.279–336, Oxford University Press

28 Bisiach, E. (1993) Menta l representa tion in

unilateral neglect and relat ed disorders: the

twent ieth Bar tlett Memorial lecture.Q. J. Exp.

Psychol.46A, 435–461

29 Lan dau, B . and J ackendoff, R. (1993) ‘What ’ and

‘where’ in spatia l langua ge and spat ial cognition.

Behav. Brain Sci. 16, 217–265

30 Kemmerer, D. (1999) ‘Near’ and ‘far ’ in language

and perception.Cognition 73, 35–63

31 Ca rlson-Radva nsky, L.A. and Irw in, D.E. (1993)

Fra mes of reference in vision and la nguage:

where is a bove?Cognition 46, 223–244

32 Talmy , L. (1996) Fictive motion in langua ge and

‘ception’. In Lan guage and Space (Bloom, P .et al ., eds), pp. 211–276, MIT P ress

33 J ackendoff, R. (1996) The a rchitecture of the

linguistic–spacial int erface. In Lan guage and Space

(Bloom, P . et al ., eds), pp.1–30, MIT P ress

34 Miller, G. a nd J ohnson-Laird, P . (1976)Language

and Perception , Harvard U niversity Press

35 Chat terjee, A. et a l . (1995) Asyntactic thematic

role assignment: th e use of a t emporal-spatia l

strategy. Brain L ang. 49, 125–139

37 J ackson, J . (1932)Selected Wri ti ngs of John

H ughli ngs Jackson , Hodder an d Stoughton

38 Chat terjee, A. et a l . (1995) Spatia l char acteristics

of themat ic role representa tion. Neuropsychologia

33, 643–648

39 Chatt erjee, A. et al. (1999) Verbs, events a nd spat ial

representations. Neuropsychologia , 37,395–402

40 Heilman, K. and Gonzalez-Rothi, L. (1993)

Apraxia. In Cli ni cal Neuropsychology (Heilman, K.

an d Va lenstein, E ., eds), pp.141–163, Oxford

University Press

41 Rushworth, M. et al . (1997) The left pa riet al cort ex

and a ttention to action. Neuropsychologia

35, 1261–1273

42 Kinsbourne, M. (1987) Mechan isms of unilat eral

neglect. In Neurophysiological an d

Neur opsychological Aspects of Spati al N eglect

(Jea nnerod, M., ed.), pp.69–86, Elsevier

43 Kosslyn, S. et al . (1989) Evidence for tw o types of

spatia l representat ions: hemispheric specializat ion

for categorical and coordinat e relations.

J Exp. Psychol. Hum . Percept. Perform .15, 723–735


lang and space

Documents