See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/255578325

Mental Spaces: Exactly When Do We Need Them?

Article in Cognitive Linguistics · January 2003

DOI: 10.1515/cogl.2003.004

CITATIONS READS

25 3,196

1 author:

Peter Harder
University of Copenhagen
70 PUBLICATIONS 1,189 CITATIONS

SEE PROFILE

All content following this page was uploaded by Peter Harder on 05 March 2015.

The user has requested enhancement of the downloaded file.

 Squib
Mental spaces: Exactly when do we
need them?

PETER HARDER

Since the first publication of Fauconnier’s Mental Spaces in 1985 there

has been a growing field of applications of the central concept that is
indicated in the title. A recent elaboration of the mental spaces approach
is the exploration of processes of conceptual integration, in which the
simultaneous invocation of two mental spaces brings about a third.
Blending spaces gives fascinating results, because previously disparate
properties can be brought to co-exist in the same mental space with pro-
perties that were found in neither of the original spaces. An example is
land yacht (cf., e.g., Fauconnier 1998: 271), where the “sailing space” and
the “driving space” are integrated to produce a mental space where the
driver of a conspicuous car is cast in the role of yacht owner.
This example simultaneously illustrates how the concept can be brought
to bear on grammatical phenomena such as compounding, since the two
spaces are blended as part of the process of forming the new compound
noun. More generally, Eve Sweetser (1999) has demonstrated how mental-
space blending can be brought to bear on the issue of compositionality.
Because of mental spaces, the usual objections against compositionality in
cases such as fake gun (that it cannot simultaneously be a gun and a fake)
can be handled: the object is a gun in one mental space but a fake in
another. The theory is attractive, since unless it is a gun in one space, it
cannot be called a fake in another; a pencil is not a gun, but that does
not qualify it to be called a fake gun. In a similar vein, Fauconnier and
Turner (cf., eg., 2002: 370) have argued that blending is a central process of
grammar. They illustrate the claim with the “caused motion” construction
(cf. Goldberg 1995), demonstrating that the insertion of a verb like sneeze
(which does not in itself designate caused motion) can be understood as
working by virtue of a blending process. Thus, when we interpret He
sneezed the napkin off the table, we integrate a space containing an
unintegrated sequence of events (including the event of sneezing) with a
space containing the integrated caused-motion schema associated with the
construction as exemplified in prototypical cases like Jack threw the ball

Cognitive Linguistics 14–1 (2003), 91–96 0936–5907/03/0014–0091

© Walter de Gruyter
92 P. Harder

into the basket. From this point of view it seems fruitful to see the blending
of mental spaces as being involved quite generally in the construction and
processing of complex expressions.
However, there is one caveat that needs to be addressed in that context.
As emphasized in relation to the blending approach by Gibbs (2000),
an attractive posthoc interpretation is not a sufficiently robust foundation
for a theory of actual mental processing (whether productive or receptive).
Gibbs suggests a number of ways in which one can make blending psycho-
logically respectable, and I would like to stress that I think one should
take a basically generous attitude to intuition-based hypotheses and avoid
all throwbacks to the “nothing buttery” mode of argument (cf. Tomasello
1998: vii). Such a fundamental generosity, however, does not exempt one
from a form of adherence to a basic principle of conservatism that ulti-
mately goes back to Ockham, i.e., that complexities in theoretical assump-
tions must be motivated by complexities in the data. If there is a theory that
can account for the same data with less heavy theoretical artillery, that
theory should be preferred until further notice.
In that respect there is a potential problem with the trajectory by which
blended mental spaces have moved into the central processes of grammar.
The star examples (cf. also Bache 2002), such as the story of the monk who
“meets himself” at a particular point on the mountain and the philosopher
having a discussion with Kant, are clear cases in which we have two distinct
spaces that are integrated—with interesting results that would be impos-
sible in either of the “parent spaces”. But by generalizing from such com-
plex cases to more mundane examples, we are reversing the directionality
recommended by the conservative approach: we are using the equipment
that was devised for complex cases to account also for the simpler ones,
instead of looking for simpler accounts until additional complexities in the
data force us to upgrade our theory.
An argument for allowing this would be to say that if we know from
elsewhere that human speakers have this capability, why not assume that it
comes for free in the simpler cases as well (Sweetser, personal communica-
tion 1998)? Parsimony in itself should not be understood as inherently
superior when there is reason to suppose that the facts support a more com-
plex explanation than one might theoretically devise; the rule-list fallacy
(cf. Langacker 1987: 29) is an example of a case where it is implausible to
use parsimony as an argument in the face of evidence that speakers also
store regular “chunks” as items. Coulson and Oakley (2000: 194) defend
the path from complex to ordinary situations in blending theory by invo-
king an analogy with gestalt psychology where the special and complex
phenomenon of optical illusions shows something general about the way
we see things. However, there must be something about the simple cases
 When do we need mental spaces? 93

that warrant their being treated on a par with the complex cases in order
for the generalization to be plausible.
In the case of mental spaces, I would like to argue that there are good
reasons to maintain the conservative view, and invoke mental-space inte-
gration only in the more complex cases. These reasons have to do with
what exactly it takes to operate with mental spaces, i.e., to understand enti-
ties and events as belonging not in one objective world, but simultaneously
in different yet overlapping mental worlds. An empirical basis for under-
standing this can be found in the experiments that have explored children’s
ability to distinguish between different pictures of reality, cf., e.g., Gopnik
(1993). One of the experiments involves an object that appears to be one
thing but is actually another, such as a green cat covered by a red filter that
makes it look black. After extensive pretraining to ensure they understand
the questions, children are asked what the object looks like and what it
really is. Three-year-olds give the same answer to both questions: the cat
looks black and really is black (or green in both cases; the choice of reality
or appearance varies with the particular object). It is only after the age of
four that children are reliably capable of holding two different pictures in
their minds at the same time.
In this context the point is that the ability to blend mental spaces must
presuppose the ability to entertain two pictures of the same thing simulta-
neously; otherwise the issue of integrating them would not arise. At age
three, blending therefore does not seem to be an option. Children, on the
other hand, know most of their syntax by that time (cf. Bates et al. 1992:
84). Therefore, I think that space building is not essentially involved in the
capacity to build composite meanings out of simpler content elements.
Since keeping apart two mental spaces is a late achievement, it is natural
to assume that it is also demanding for those who are capable of doing it,
rather than something that comes for free once you know how to do it. The
key to understanding the role of mental spaces, therefore, is to see space
building as something sophisticated that occasionally interferes with the
ordinary default assumption we know as naive realism: that there is only
one correct picture of the world, which corresponds to the way I see it. We
are born as naive realists and even when (after the age of four) we become
capable of rising above this state, we basically only do so when we have a
specific reason for it.
If we look at language from this point of view, the question is: when do
we have to invoke different, simultaneous pictures, in order to understand
what is going on? Among the clear-cut examples are epistemically complex
combinations (including conditionals, cf. also Fauconnier and Sweetser
1996). Counterfactuals clearly cannot be handled if we can only see the
world in one way. The fake gun example also survives, because there is no
94 P. Harder

way to compress the concept into a single-perspective view of the world if

we take the strict compositional path (although the phrase may of course
be learnt “lexically” as denoting a toy of a particular kind).
Space building, on the other hand, should not be invoked as a necessary
part of the picture in understanding spatial expressions like in France
or (pace Fauconnier 1998: 256) temporal expressions like in 1968. Mental
worlds, like the real world, can be assumed to have spatial and temporal
dimensions inside them. Compiling birth rates for the past fifty years, for
instance, there is no need to set up a new mental space for each figure we
take down, such as the birth rate that obtained in 1968: the whole statistical
table can be accommodated in one mental model and within one space. The
situation is similar if we compare the figures for different countries. Only
if we start playing around with assumptions that draw on alternative
assumptions about the same year or country do we have to keep track of
more than one mental space simultaneously.
For the same reason, compositional understanding of NPs does not gen-
erally require mental-space building. The core of compositionality is not in
the mental models or spaces that we construct as a result of interpretation
—rather it is in the recipe for interpretation that we follow (cf. Harder
1996: 214) in compiling the meaning, involving a particular compositional
path in Langacker’s terms. The order of modifiers typically specifies the
compilation path we have to follow, so that in Japanese fake paintings we
combine fake with painting before we make the product Japanese—while
in fake Japanese paintings we first make the paintings Japanese and then
turn them into fakes. The difference in the process means that we get dif-
ferent results in the two cases, and that the properties of the paintings
are different in spite of the fact that the words (in isolation) and the gram-
matical structures (in isolation) are the same. There may be a problem if we
view compositionality from the point of view of truth conditions; but the
results are predictable from the parts plus their manner of composition, so
there is no compositionality problem from a linguistic point of view. The
complexity lies in constructing an interpretation that will accommodate the
specifications, some of which may not readily merge, in the right way.
The main factor that prompts space building, according to this view,
is thus potential contradiction, as in the fake gun case. When you interpret
a less sophisticated case like black gun, you can add the property black
without interfering with the “gunhood” of the gun, so you can stay within
the same space—which, as a default case, I suggest you do. The situation
is similar with land animals as compared with land yacht. This, of course,
does not prevent individual listeners from building more spaces than
they (strictly speaking) need; but the process should not be understood
 When do we need mental spaces? 95

as built into the way (content-)syntactic combination works in cueing

understanding.
As a consequence of this view, we need to ask what the relationship
is between blending and mental spaces. The foregoing account does not
rule out a blending process based simply on semantic specifications, rather
than on distinct mental spaces. Fauconnier and Turner’s account of He
sneezed the napkin off the table would not be essentially altered in we
removed the space boundaries.
In Langacker’s terms, compositional relations involve “correspond-
ences”; you cannot work out the meaning of, for example, combinations
of nouns and verbs without figuring out how the semantic specifications
associated with the elaboration site of the verb can be brought to cohere
with the semantic specifications associated with the noun (Langacker 1991:
37). Blending as associated with semantic composition also frequently
involves creativity that goes beyond that of putting one lego block on top
of another, as in the case of elephant ribbon, understood to mean the ribbon
with which the zoo manager intends to wrap a present for the elephant
keeper (Langacker 1987: 282n). Here, too, I would claim that mental space
differentiation is not necessary, as long as all the required operations can
be performed in one coherent space.

Received 15 March 1999 University of Copenhagen

Revision received 10 May 2002

References
Bache, Carl
2002 Grammar, blending and conceptual disintegration. Unpublished manuscript,
University of Odense
Bates, Elizabeth, Donna Thal, and Jeri S. Janowsky
1992 Early language development and its neural correlates. In Segalowitz,
S. J. and I. Rapin (eds.), Handbook of Neuropsychology, vol. 7: Child
Neuropsychology, Amsterdam: Elsevier, 69–110.
Coulson, Seana and Todd Oakley
2000 Blending basics, Cognitive Linguistics 11–3/ 4, 175–196.
Fauconnier, Gilles
1985 Mental Spaces. Aspects of Meaning Construction in Natural Language.
Cambridge, MA: MIT Press.
1994 Mental Spaces. New York: Cambridge University Press. [Originally published
in 1985 by MIT Press.]
1998 Mental spaces, language modalities, and conceptual integration. In
Tomasello, Michael (ed.), The New Psychology of Language. Mahwah,
NJ: Lawrence Erlbaum, 251–279.
Fauconnier, Gilles and Eve Sweetser
1996 Spaces, Worlds and Grammar. Chicago: University of Chicago Press.
Fauconnier, Gilles and Mark Turner
2002 The Way We Think. New York: Basic Books.
 96 P. Harder

Gibbs, Raymond
2000 Making good psychology out of blending theory. Cognitive Linguistics
11–3/ 4, 347–358.
Goldberg, Adele
1995 Constructions: A Construction Grammar Approach to Argument Structure.
Chicago: Chicago University Press.
Gopnik, Alison
1993 How we know our minds: The illusion of first-person knowledge of intention-
ality. The Behavioral and Brain Sciences 16, 1–14.
Harder, Peter
1996 Functional Semantics. A Theory of Meaning, Structure and Tense in English.
(Trends in Linguistics. Studies and Monographs 87.) Berlin: Mouton de
Gruyter.
Langacker, Ronald W.
1987 Foundations of Cognitive Grammar, vol. 1: Theoretical Prerequisites.
Stanford: Stanford University Press.
1991 Foundations of Cognitive Grammar, vol. 2: Descriptive Applications. Stanford:
Stanford University Press.
Sweetser, Eve
1999 Compositionality and blending: Semantic composition in a cognitively realis-
tic framework. In Janssen, Theo and Gisela Redeker, Cognitive Linguistics:
Foundations, Scope and Methodology, Berlin: Mouton de Gruyter, 129–162.
Tomasello, Michael (ed.)
1998 The New Psychology of Language: Cognitive and Functional Approaches to
Language Structure. Mahwah, NJ: Lawrence Erlbaum.

View publication stats