Embodiment and Language Comprehension ReV PDF

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

TICS-1309; No.

of Pages 6

Opinion

Embodiment and language


comprehension: reframing the
discussion
Rolf A. Zwaan*
Department of Psychology, Erasmus University Rotterdam, 3000 DR Rotterdam, The Netherlands

The debate on whether language comprehension least some of them need to be grounded in perception and
involves the manipulation of abstract symbols or is action. Not all processing requires representations that are
grounded in perception and action has reached an im- directly grounded in perception and action; sometimes,
passe, with authors from different theoretical persua- cognitive processes can be offloaded on external memory
sions unable to agree on the diagnostic value of in the environment. Rather than contrasting models that
empirical findings. To escape this impasse, I propose a claim that language comprehension is not embodied, we
pluralist view of cognition that encompasses abstract and should examine the conditions in which the comprehension
grounded symbols. The contributions of these symbol system uses symbolic and grounded mental representa-
types to language comprehension vary as a function of tions or perhaps no representations at all.
the degree to which language use is embedded in the Language comprehension involves the construction of a
environment. I distinguish five levels of embeddedness: coherent mental representation of the state of affairs
demonstration, instruction, projection, displacement, denoted by the linguistic input [5–9]. Language compre-
and abstraction. Only through a closer analysis of context hension is situated in a specific context and occurs with a
will we make significant progress toward understanding specific purpose in mind: to perform an action [10]. Some-
language comprehension and cognition in general. times the action is immediate and sometimes it is more
distal: to commit information efficiently to long-term mem-
Introduction ory so that it can be retrieved and used for later action or
According to the traditional, symbolic view of human thought. Sometimes, we comprehend language to discon-
thought processes, cognition involves the manipulation of nect from the immediate environment altogether and es-
abstract, arbitrary, and amodal symbols. According to the cape into a fictional world (even this can have a long-term
more recent embodied view, cognition is grounded in the impact on behavior [11]). How do the symbolic and the
brain’s systems of action, perception, and emotion. Over the grounded view conceptualize language comprehension?
past 15 years, the grounded view has risen to prominence.
This rise, particularly in the study of language comprehen- Two perspectives
sion, is documented in several recent reviews [1–4]. The Symbolic cognition assumes that cognitive processing
debate between the traditional and the grounded view seems involves a small set of rules that operate on a vast set of
to have reached an impasse, with authors from different arbitrary, amodal, and abstract symbols [12–14]. Ground-
theoretical persuasions unable to agree on the diagnostic ed cognition assumes that cognitive processing involves
value of virtually all the relevant empirical findings. This the re-instantiation and recombination of experiences via
opinion is an attempt to: (i) articulate what I perceive to be symbols that are grounded in the brain’s systems of per-
the latent majority view among language researchers; and ception, action, and emotion [1,15]. These symbols are
thereby (ii) reframe the discussion about embodiment. I nonarbitrary, multimodal, and concrete.
argue that we can escape the impasse by paying closer Are these views mutually incompatible? Symbolic and
attention to how language use is embedded in the environ- grounded theories both accord a central role to mental
ment and I suggest how this might be accomplished. representations in cognition [16]. The discussion is about
We can summarize the latent majority view of language the nature of these representations (and about the pro-
processing as follows. We need mental representations. At cesses operating on them, although this issue has never
been at the forefront of the discussion). Symbolic and
Corresponding author: Zwaan, R.A. (zwaan@fsw.eur.nl, rolfzwaan@gmail.com).
Keywords: language; comprehension; embodiment; embodied cognition; mental grounded representation have a history of theoretical
representation. cohabitation in dual-coding theory [17] and grounded
*
Twitter: @RolfZwaan. approaches to language comprehension [18,19] as well
1364-6613/$ – see front matter as in ‘hybrid’ theories of cognition [20,21], with the dis-
ß 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tics.2014.02.008
cussion mostly centered on the question of which does
which share of the work.
As I detail below, at this juncture the most fruitful
option is to adopt a pluralist perspective [21–23]. We need
Trends in Cognitive Sciences xx (2014) 1–6 1
TICS-1309; No. of Pages 6

Opinion Trends in Cognitive Sciences xxx xxxx, Vol. xxx, No. x

both perspectives to solve two major problems that face object or event in the world that they are denoting. How-
cognitive theorizing: the grounding problem and the scal- ever, words are nonarbitrarily related to their referents in
ing problem. terms of situational co-occurrence. Concrete words are
often encountered in the presence of the entity or event
Two problems and their proposed solutions they denote, which is a basis for word learning [28]. Simi-
Cognition cannot exist as a free-floating set of symbols and larly, recent evidence suggests that some abstract words
rules; the symbols have to be connected to perception and acquire their grounding in emotional contexts [29]. Be-
action. This is the grounding problem that poses a threat to cause words systematically co-occur with specific agents,
symbolic cognition [24]. Grounded cognition has arisen as an objects, and events, links are forged between words and
attempt to address the grounding problem, but the need to their referents in memory.
address the grounding problem is now also felt by research- Referents do not occur in isolation and neither do words.
ers of other theoretical persuasions (e.g., [21,25]). The cur- Referents occur in spatiotemporal settings that include
rent discussion centers more on how and to what extent the other agents, objects, and events that co-occur with some
grounding problem needs to be addressed. One possible view systematicity with these referents. Words are spoken and
is that cognition exclusively involves grounded symbols. Not written in the context of other words and this also occurs
all proponents of grounded cognition subscribe to this view, with some systematicity. The systematicity in the word co-
because they accord a role to word-like representations in occurrences derives from the patterns of entity co-
cognition [1,21,26]. Another view is that symbolic represen- occurrences in the world [19]. This explains why particular
tations do all the initial work in terms of meaning activation ‘grounding effects’ such as geographical distances can be
and conceptual combination and that these symbols may obtained by sophisticated analyses of word co-occurrences
then activate grounded representations; activation ‘cas- (or rather, the similarity of contexts in which certain words
cades down’ from abstract symbols to grounded ones in occur) [30]. It should be noted that this type of distribu-
primary sensorimotor areas of the brain [25]. A much more tional (bag-of-words) semantics does not inform about
extreme way to deal with the grounding problem is to do issues that are central to language comprehension (e.g.,
away with symbols altogether [27]. However, both a view predication, compositionality, anaphoric resolution) and
that does away with mental representations and a view that therefore needs to be augmented with structural informa-
assumes that comprehension exclusively involves detailed tion [31] to avoid creating its own scaling problem.
mental simulations involving grounded representations en- The key point about word-like representations is that
counter the scaling problem. they are consistent with the idea that language compre-
The scaling problem occurs because an account of most hension does not necessarily invoke a sensorimotor simu-
processes that we associate with cognition (language un- lation of the word’s referent. In other words, they allow for
derstanding, reasoning, problem solving, decision making) abstraction. Accepting that we need both abstract and
that does not rely on mental representations seems remote, grounded symbols leads to a new set of questions. When
if not impossible [22,26]. What we might call the secondary do we need which and how do they interact? To answer this
scaling problem is that it also seems unlikely that a system we need to consider the context of language use.
comprising solely grounded symbols can account for ab-
stract thought [21,25,20]. Even for understanding a simple Context
phrase like ‘a thief having to pay for stolen goods’, a ‘tower A major problem with most of the work on embodiment and
of abstraction’ has to be erected [20]. This tower rests on language (and with psycholinguistics in general) is that the
concepts such as ownership of property, theft, social com- focus has almost exclusively been on individual words and
pulsion, and payment via fines or imprisonment, the latter sentences [32]. Psychological experiments on language
concepts resting on notions of money and freedom. The typically use decontextualized words and sentences as
secondary scaling problem is that this tower of abstraction stimuli [33,34]. This is problematic because the resulting
cannot be surmounted by systems that are equipped with patterns of brain and behavioral responses that are
grounded concepts only. Several researchers have pro- obtained in experiments cannot simply be extrapolated
posed hybrid symbolic-grounded positions to address both to discourse comprehension. During discourse comprehen-
the grounding and the scaling problem. These proposals sion, people generate extensive mental representations [9],
differ in whether the symbolic representations are them- elements of which may form the context of an utterance
selves grounded in perception of an action [1,19], whether and can override word-level patterns of association [35].
they are at least in part amodal [21], or whether they are Context has been a sleeping giant in the discussion on
exclusively amodal, with activation cascading down to embodiment. It is likely to be a major factor in mediating
grounded representations [25]. the relative impact of symbolic and grounded representa-
We need abstraction to account for any form of mean- tions. Although action verbs may trigger pre-motor activa-
ingful language comprehension, but do the symbols also tion when presented in isolation [36] in literal expressions,
need to be amodal and arbitrary? Word-like representa- this activation is diminished or eliminated in figurative
tions, for example, can be considered multimodal, involv- expressions [37,38]. Conversely, reading narrative texts
ing visual, auditory, tactile, and various motor might lead to elaborate grounded representations that
representations; we can speak, sign, write, type, hear, facilitate the formation of further grounded representa-
and read words and some of us can even feel them [19]. tions [39,40]. Thus, firm claims about the relative impor-
Are words arbitrary? They certainly are in the sense tance of symbolic versus grounded representations cannot
that they are not analogous to the representation of the be made without taking context into account [4].
2
TICS-1309; No. of Pages 6

Opinion Trends in Cognitive Sciences xxx xxxx, Vol. xxx, No. x

Often the context will not be sufficiently constraining for Five levels of environmental embeddedness
the comprehension system to activate all the way down to a I distinguish different forms of language comprehension in
detailed grounded representation. Otherwise the system terms of how deeply they are embedded in the environ-
might commit to an overly specified representation that ment. The level of embeddedness can be characterized in
might be contradicted by subsequent input. Suppose a terms of the overlap between the communicative situation
sentence states that ‘There were 50 red roses in the vase.’ and the referential situation (the situation that the dis-
There is no utility to a detailed visual representation of all course is about). Situations are characterized by a spatio-
50 roses, even if it were possible to create one. A systematic temporal framework that involves agents, objects, and
activation (one that is roughly similar across different events [9]. Each sentence in discourse conveys a situation.
comprehenders) of grounded representations might just When embeddedness is maximal, the referential situation
involve the visual (and olfactory) experience of a vase with overlaps completely with the communicative situation.
red roses, no matter how many it contains. The less the referential situation overlaps with the com-
However, the fact that the story mentions the number of municative, the more comprehenders will have to rely on
roses may hold significance for the continuation of the their long-term memory to activate representations of the
discourse; perhaps the recipient has just turned 50 years relevant agents, objects, and events.
of age. This information can be gleaned from the sentence Figure 1 depicts five levels of embeddedness; a more
just based on word associations among ‘50’, ‘roses’, ‘vase’, intuitive account follows here. The five levels rarely occur
and ‘red’. A detailed visual representation of exactly 50 in their pure form; most forms of discourse are mixtures
roses would add nothing to the evolving representation. To and comprehenders are highly skilled at shifting between
develop this idea further, consider the following sentence: these levels. Nevertheless, there is some conceptual clarity
‘Several people entered the house.’ What detailed repre- to be gained by distinguishing among them, particularly
sentations might be relevant for understanding this sen- because it allows us to analyze more closely the relative
tence? How many people were there? How many men contributions of the symbolic and grounded symbols as well
versus women? Were there any children? What was every- as sensorimotor processes.
one wearing? Granted, there is no law against idiosyncratic (i) Demonstrations are the most completely embedded
activations of experiential representations, but detailed level of language use. Agents, objects, and actions are
mental simulations of the number of people, their gender all present in the communicative situation. An
balance, and their respective appearances are not licensed
by the text. So although representations may be grounded,
they do not need to be detailed. Communicave situaon Referenal representaon Long-term memory

Once we consider language in the form of connected SxTx SxTx


discourse, the division of labor between symbolic and
(1) AXOXEX AXOXEX
grounded representations becomes clearer. Let us assume
that the story continues with ‘Among them was a young
woman with long blonde hair in a red coat.’ Now the SxTx SxTx
activation of grounded symbols becomes more extensive (2) AXOXEX AXOyEX Oy
(and similar across individuals). If the story continues with
‘She held a white iPhone in a white case by her left ear’, the SxTx SxTy
activation of grounded representations becomes even more
extensive. At some point the activation of grounded sym- (3) AXOXEX AxOxOyEX AxOyEX
bols may be so extensive that they would dominate the
comprehension process. If the narrative shifted to a new SxTx SxTy
situation, the reliance on symbolic representations may
become greater again. In other words, questions regarding (4) AxOxEXCX AxOxEXCX
the interaction between symbolic and grounded represen-
SxTx S0T0
tations cannot be answered by analyzing decontextualized
snippets of language. (5) CX CX
Suppose further that the story about the blonde woman
is told in a conversation between two friends, a joint action TRENDS in Cognitive Sciences
that is situated in the world [10]. One of the conversation
partners can pick up his own iPhone and hold it by his ear Figure 1. Five levels of situational embeddedness for language use. The columns
depict: (1) the communicative situation (where the communication occurs); (2) the
to demonstrate woman’s action. How does this enactment referential representation (what the communication is about); and (3) the
of the story impact the comprehension process? Is a mental comprehender’s long-term memory (LTM). The thickness of the arrows indicates
the extent to which the referential representation draws on the communicative
representation of the action still necessary, given that it
situation and long-term memory for referential information. S and T denote a
unfolds before our eyes? We need to consider both the spatiotemporal framework. A, O, and E refer to agents, objects, and events,
linguistic context and the communicative situation dis- respectively. C refers to abstract concepts (e.g., democracy). At Level 1, all relevant
referential information is derived from the communicative situation. At Level 2,
course context if we want to make informed predictions one referential element is derived from LTM. At Level 3, the referential situation is
about the role of sensorimotor processes, grounded repre- in the same location as the communicative, but is situated at a different time,
sentations, and symbolic representations in language com- which involves different situational elements (e.g., different objects). At Level 4,
there is a different spatiotemporal framework and all representational information
prehension. Specifically, we need to know how a particular is derived from LTM and may involve abstract concepts. At Level 5, there is no
act of language use is embedded in the environment. spatiotemporal framework as an abstraction is represented.

3
TICS-1309; No. of Pages 6

Opinion Trends in Cognitive Sciences xxx xxxx, Vol. xxx, No. x

example is a cooking show in which the chef gives a perception of actions and object, leading to slight modifica-
concurrent account of his actions. The bell pepper he tions of object and action representations.
is talking about is the bell pepper he is cutting at this Returning to the culinary domain, suppose that Jamie
moment. If deictic terms like ‘here’, ‘now’, and ‘you’ Oliver uses a device to shave something from the zest of a
are used, they refer to the situation that the chef is in. lemon. He calls the device a ‘zester’. If we did not already
(ii) Instructions describe a desired or required state of the know what a zester was, we have now formed a connection
world that (slightly) differs from the present one. The between an auditory event ‘zester’ coupled with a visual
chef might tell his sous-chef to fetch him a parsnip. event involving a tool, an action (removing the zest), and a
For this to work, the sous-chef needs to have a result (lemon zest, which is subsequently used in a mari-
grounded representation in long-term memory of nade). We now know what a zester looks like, how it should
what a parsnip looks like to recognize it among the be used, and what it is for. We may not have had an up-
assembled vegetables in the refrigerator. close look at the zester – it was partly covered by Jamie’s
(iii) Projections map a past or future state of the hand and he was moving quickly – but we may have
environment on the current one. Therefore, they retain glimpsed enough to recognize a zester when we see one
the spatial component of the communicative situation in a store on a later occasion.
but have a different temporal one, located either in the The new visual experience of a zester modifies our repre-
past or in the future. A tour guide might explain what sentation of this nifty tool. Moreover, the act of recognition
the Coliseum looked like in its heyday when it was activates the word representation (we might even ‘hear’
filled with spectators, bears, gladiators, shields, and Jamie say the word), which might prompt us to exclaim
swords, but this requires access to even more memory ‘ah, a zester!’, producing a new articulatory and auditory
representations because agents, props, and actions event that will strengthen the link between word represen-
(not to mention important parts of the Coliseum itself) tation and visual object representation even further. Each
are missing. Similarly, a contractor may explain to his time we see someone using a zester, grounded and symbolic
clients what the house will look like after a proposed representations will resonate, which may lead us to make
renovation. Some objects and spatial relations may predictions. For example, when Jamie picks up his zester,
remain unchanged (e.g., windows) whereas others we anticipate seeing a lemon, lime, or orange in his hand and
(e.g., walls) will be broken down. we can predict what he is going to do with it and how and
(iv) Displacements describe an environment that is what the result of that action will be.
unrelated to the current environment, as is the case In instructions, grounded representations interact with
in narratives (e.g., jokes, novels, news reports, perceptuomotor processes. An object is referred to that is
historical accounts). Here, deictic terms do not refer not within view. Recognizing it requires a grounded repre-
to the reader’s environment but to some distal one. sentation. Once the object is retrieved and brought into
Grounded representations are very common at this view so it can be acted on, grounded representations may
level but share the stage with symbolic representa- recede into the background. For example, infants use
tions, which often function as placeholders until more motor actions to make a linguistically denoted object visu-
grounded information accumulates over the course of ally more salient by moving it closer to the eyes, an action
the discourse. that leads to better learning [41]. The so-called visual-
(v) Abstractions do not refer to any environment in world paradigm often employs instructions in which peo-
particular – they do not have a spatiotemporal ple’s eye (and hand) movements are simultaneously affect-
framework – but describe a constellation of mostly ed by the linguistic input and the visible and actionable
abstract concepts (e.g., scientific articles, philosophi- context. We often spatially index aspects of the visual
cal treatises, legal documents). The primary vehicle world without encoding the objects, because they can be
for comprehension will be abstract symbols. However, readily accessed in the environment via an eye movement
when the author assumes it is beneficial to compre- to the indexed location [42,43].
hension, he or she may invoke grounded representa- In projections, there is a mixture of sensorimotor and
tions by using metaphors and similes. grounded representations. The sensorimotor processes
(e.g., eye movements) are used to identify specific locations
A further assumption is that symbolic and grounded in the environment and grounded representations are then
symbol systems are simultaneously active – along with used to project on those locations (e.g., a long-deceased
perceptuomotor processes (e.g., eye movements) – each op- gladiator or an envisioned curtain). Projections are likely
erating in accordance with their own constraints as well as to involve mental imagery, which can be viewed as a
constraining each other. Which system dominates the com- conscious use of grounded symbols [1,4].
prehension process depends on the level of embeddedness. In displacements, both grounded and symbolic repre-
In demonstrations, perceptuomotor processes dominate sentations are active. One possibility is that symbolic and
and the mental representations (grounded and symbolic) grounded networks are interactive and constrain each
that resonate may have only a small role to play. Interest- other to enhance processing fluency [44]; another is that
ingly however, this passive resonance process that occurs activity cascades down from symbolic to grounded repre-
during demonstrations (and instructions and projections) sentations [25]. In reading comprehension, sensorimotor
is a major mechanism for word learning [28]. During processes (eye fixations and saccades) are used to acquire
sensorimotor interaction, grounded representations are the linguistic input, but in listening comprehension they
formed or, when already formed, may resonate with the can be used for indexing purposes [45].
4
TICS-1309; No. of Pages 6

Opinion Trends in Cognitive Sciences xxx xxxx, Vol. xxx, No. x

Finally, in abstraction, symbolic representations pre- Concluding remarks


dominate. Relatively few grounded representations might Reframing the discussion about language and embodiment
be activated or they might be activated only weakly due to leads us to ask fundamentally different research questions
being indirectly connected to grounded representations. about language comprehension and embodiment than
Interestingly, however, recent research suggests that ab- have been asked over the past 15 years (Box 1). Rather
stract words achieve their grounding via emotional states than trying to find evidence for or against a particular
[29]. theoretical framework and argue over the diagnostic value
Comprehenders are highly adept at switching between of a particular experimental finding regarding the explan-
these levels. For example, abstractions can instantaneous- atory superiority of a particular framework, the proposed
ly turn into demonstrations through the use of gestures reframing prompts us to ask when and how the different
and enactments and the use of diagrams and doodles. types of process and representation interact. A more de-
Abstractions can also invoke grounded representations tailed analysis of the levels of embeddedness is likely to
by making use of metaphors. Importantly, however, yield novel predictions about the relative contributions of
demonstrations and metaphors are used in the service of sensorimotor processes and grounded and symbolic repre-
abstraction, meaning that successful comprehension sentations in language comprehension. This move will help
requires abstracting away from them, although the dem- us explain that most human of all cognitive feats: under-
onstration or metaphor itself can be stored as a helpful standing language.
mnemonic device and be retrieved into working memory on
a later occasion to help ‘re-simulate’ the abstraction. Acknowledgments
The degree to which language comprehension is ground- The author thanks Vic Ferreira, Art Markman, and others for a helpful
ed depends on the level of environmental embeddedness. discussion on Facebook regarding matters discussed in this opinion, and
Art Glenberg and an anonymous reviewer for helpful comments on a
Another important factor to consider is the depth of previous version of the manuscript.
comprehension.
References
Comprehension is fault tolerant 1 Barsalou, L.W. (2010) Grounded cognition: past, present, and future.
Comprehension is not all or none. Whether people build Top. Cogn. Sci. 2, 716–724
coherent mental representations using grounded and sym- 2 Glenberg, A.M. et al. (2013) From the revolution to embodiment: 25
years of cognitive psychology. Perspect. Psychol. Sci. 8, 573–585
bolic representations during demonstrations depends on
3 Kiefer, M. and Pulvermüller, F. (2012) Conceptual representations in
their processing goals, their linguistic skills and knowl- mind and brain: theoretical developments, current evidence and future
edge, and their (sensorimotor) experience. If we watch a directions. Cortex 48, 805–825
cooking show for relaxation, we might not construct very 4 Meteyard, L. et al. (2012) Coming of age: a review of embodiment and
elaborate mental representations and allow sensorimotor the neuroscience of semantics. Cortex 48, 788–804
5 Bower, G.H. and Morrow, G.G. (1990) Mental models in narrative
processes to do the bulk of the work, simply taking in the
comprehension. Science 247, 44–48
information and following along. However, if we watch the 6 Bransford, J.D. et al. (1972) Sentence memory: a constructive vs
show to learn how to cook a certain dish, we are likely to interpretive approach. Cogn. Psychol. 3, 193–209
construct a more elaborate representation that will be 7 Johnson-Laird, P.N. (1983) Mental Models: Towards a Cognitive
stored in long-term memory, which (possibly with the help Science of Language, Inference and Consciousness, Cambridge
University Press
of a recipe) will allow us to prepare the food on a future 8 Van Dijk, T.A. and Kintsch, W. (1983) Strategies of Discourse
occasion. Comprehension, Academic Press
If someone has extensive experience with a domain, 9 Zwaan, R.A. and Radvansky, G.A. (1998) Situation models in language
more grounded representations will be activated compared comprehension and memory. Psychol. Bull. 123, 162–185
with a novice [46]. An expert might form a full-blown first- 10 Clark, H.H. (1996) Using Language, Cambridge University Press
11 Mar, R.A. et al. (2006) Bookworms versus nerds: exposure to fiction
person mental simulation of the described actions (involv- versus non-fiction, divergent associations with social ability, and the
ing many grounded representations), a domain novice simulation of fictional social worlds. J. Res. Pers. 40, 694–712
might form a shallow representation based on word-like 12 Fodor, J.A. (1983) Modularity of Mind: An Essay on Faculty Psychology,
representations, and someone with limited language pro- MIT Press
13 Pylyshyn, Z.W. (1984) Computation and Cognition: Toward a
ficiency might form a rudimentary interpretation based on
Foundation for Cognitive Science, MIT Press
the closed-class words and bound morphemes. Even the 14 Turing, A.M. (1950) Computing machinery and intelligence. Mind 59,
third group will have achieved some understanding. Com- 433–460
prehension is fault tolerant [47]. 15 Glenberg, A.M. and Gallese, V. (2011) Action-based language: a theory
of language acquisition, comprehension, and production. Cortex 48,
905–922
Box 1. Outstanding questions 16 Markman, A.B. and Dietrich, E. (2000) Extending the classical view of
representation. Trends Cogn. Sci. 4, 470–475
 How does the level of environmental embeddedness impact the 17 Paivio, A. (2007) Mind and its Evolution: A Dual Coding Theoretical
role of grounded representations in language comprehension? Approach, Erlbaum
 Do grounded representations become more detailed over time 18 Barsalou, L.W. et al. (2008) Language and simulation in conceptual
during discourse comprehension? processing. In Symbols, Embodiment, and Meaning (De Vega, M. et al.,
 How are comprehenders able to switch back and forth between eds), pp. 245–284, Oxford University Press
levels of embeddedness? 19 Zwaan, R.A. and Madden, C.J. (2005) Embodied sentence
 What are the cognitive costs and benefits of these switches? comprehension. In Grounding Cognition: The Role of Perception and
 Do comprehenders generate more detailed representations than Action in Memory, Language, and Thinking (Pecher, D. and Zwaan,
is licensed by the discourse? R.A., eds), pp. 224–245, Cambridge University Press

5
TICS-1309; No. of Pages 6

Opinion Trends in Cognitive Sciences xxx xxxx, Vol. xxx, No. x

20 Arbib, M.A. et al. (2014) Language is handy but is it embodied? 36 Hauk, O. et al. (2004) Somatotopic representation of action words in
Neuropsychologia 55, 57–70 human motor and premotor cortex. Neuron 41, 301–307
21 Dove, G.O. (2009) Beyond perceptual symbols: a call for 37 Raposo, A. et al. (2009) Modulation of motor and premotor cortices by
representational pluralism. Cognition 110, 412–431 actions, action words and action sentences. Neuropsychologia 47, 388–
22 Dale, R. (2010) Critique of radical embodied cognitive science. J. Mind 396
Behav. 31, 127–140 38 Schuil, K.D.I. et al. (2013) Sentential context modulates the
23 Dale, R. et al. (2009) Explanatory pluralism in cognitive science. Cogn. involvement of the motor cortex in action language processing: an
Sci. 33, 739–742 fMRI study. Front. Hum. Neurosci. 7, 1–13
24 Harnad, S. (1990) The symbol grounding problem. Phys. D 42, 335–346 39 Chow, H.M. et al. (2014) Embodied comprehension of stories:
25 Mahon, B.Z. and Caramazza, A. (2008) A critical look at the embodied interactions between language regions and modality-specific neural
cognition hypothesis and a new proposal for grounding conceptual systems. J. Cogn. Neurosci. 26, 279–295
content. J. Physiol. Paris 102, 59–70 40 Kurby, C.A. and Zacks, J.M. (2013) The activation of modality-specific
26 Markman, A. and Dietrich, E. (2000) Extending the classical view of representations in naturalistic discourse processing. Brain Lang. 216,
representation. Trends Cogn. Sci. 4, 470–475 338–349
27 Chemero, A. (2009) Radical Embodied Cognitive Science, MIT Press 41 Yu, C. and Smith, L.B. (2012) Embodied attention and word learning
28 Smith, L.B. and Yu, C. (2008) Infants rapidly learn word–referent by toddlers. Cognition 125, 244–262
mappings via cross-situational statistics. Cognition 106, 333–338 42 Spivey, M. and Richardson, D.C. (2008) Language embedded in the
29 Kousta, S-T. et al. (2011) The representation of abstract words: why environment. In The Cambridge Handbook of Situated Cognition
emotion matters. J. Exp. Psychol. Gen. 140, 14–34 (Robbins, P. and Aydede, M., eds), pp. 383–400, Cambridge
30 Louwerse, M.M. (2011) Symbol interdependency in symbolic and University Press
embodied cognition. Top. Cogn. Sci. 3, 273–302 43 Tanenhaus, M.K. et al. (1995) Integration of visual and linguistic
31 Kintsch, W. and Mangalath, P. (2011) The construction of meaning. information in spoken language comprehension. Science 268, 1632–
Top. Cogn. Sci. 3, 346–370 1634
32 Fischer, M.H. and Zwaan, R.A. (2008) Embodied language – a review of 44 Zwaan, R.A. (2008) Experiential traces and mental simulations in
the role of the motor system in language comprehension. Q. J. Exp. language comprehension. In Symbols, Embodiment, and Meaning
Psychol. 61, 825–850 (DeVega, M. et al., eds), pp. 165–180, Oxford University Press
33 Graesser, A.C. et al. (1997) Discourse comprehension. Annu. Rev. 45 Spivey, M. and Geng, J. (2001) Oculomotor mechanisms activated by
Psychol. 48, 163–189 imagery and memory: eye movements to absent objects. Psychol. Res.
34 Spivey, M.J. et al. (2002) Eye movements and spoken language 65, 235–241
comprehension: effects of visual context on syntactic ambiguity 46 Beilock, S.L. et al. (2010) Sports experience changes the neural
resolution. Cogn. Psychol. 45, 447–481 processing of action language. Proc. Natl. Acad. Sci. U.S.A. 105,
35 Nieuwland, M.S. and Van Berkum, J.J.A. (2006) When peanuts fall in 13269–13273
love: N400 evidence for the power of discourse. J. Cogn. Neurosci. 18, 47 Taylor, L.J. and Zwaan, R.A. (2009) Action in cognition: the case of
1098–1111 language. Lang. Cogn. 1, 45–58

You might also like