Professional Documents
Culture Documents
Affordances and Artifacts
Affordances and Artifacts
https://doi.org/10.1007/s11229-019-02297-4
Erica Cosentino1
Abstract
What are the affordances of artifacts? One view is that the affordances of artifacts,
just as the affordances of natural objects, pertain to possible ways in which they can
be manipulated (e.g., a computer keyboard affords grasping). Another view main-
tains that, given that artifacts are sociocultural objects, their affordances pertain pri-
marily to their culturally-derived function (e.g., a computer keyboard affords typ-
ing). Whereas some have tried to provide a unifying notion of affordance to capture
both aspects, here I argue that they should be kept separate. In this paper, I introduce
a distinction between standard affordances, which concern the function of artifacts,
and ad-hoc affordances, which refer to how artifacts are manipulated. I then argue
for the neuropsychological plausibility of such a distinction, linking it to the disso-
ciation between function knowledge and manipulation knowledge. Finally, I defend
the equal status of these forms of knowledge and, hence, of standard and ad-hoc
affordances, and I show that this has some implications for the debate on the role of
motor processes in the conceptual knowledge of artifacts.
1 Introduction
The notion of affordance is one of the most controversial topics in Gibson’s ecologi-
cal approach to perception (Gibson 1979). He introduces it as follows:
“The affordances of the environment are what it offers the animal, what it pro-
vides or furnishes, either for good or ill. The verb to afford is found in the
dictionary, but the noun affordance is not. I have made it up. I mean by it
something that refers both [to] the environment and the animal in a way that
no existing term does. It implies the complementarity of the animal and the
environment” (p. 127).
* Erica Cosentino
erica.cosentino@rub.de
1
Ruhr University of Bochum, Bochum, Germany
13
Vol.:(0123456789)
Synthese
This straightforward definition has generated very much debate concerning how
affordances should be characterized and what is their ontological status. Affordances
have been defined as either dispositional properties (Reed 1996; Turvey 1992) or
relations (Chemero 2003), and much discussion has been devoted to analyzing
which aspects of the environment and the animal are relevant in order to specify
affordances (Heft 1989; Warren 1984; Shaw et al. 1982). Here I do not delve into
the ontological debate on affordances. In particular, I do emphasize that affordances
arise from the combination of some features of the object/environment and some
features of the animal, and as such, they are neither primarily related to the object/
environment nor the animal. However, I do not intend to imply that they have an
ontological status as relations. I am also not going to address the issue of whether
affordances are or are not independent of perception. Since Norman (1999) intro-
duced the distinction between “real affordances” (which are closely related to physi-
cal constraints) and “perceived affordances” (which may or may not be real), the
relation between affordances and perception has been a key topic of investigation in
several fields (for example, in Human–Computer-Interaction research). Again, this
is an issue that mainly concerns the ontology of affordances, and even though it is a
very relevant one, it falls beyond the scope of this paper. Also, I am going to discuss
differences and similarities between my definition of the notion of “affordance” and
some definitions of the notion of “function,” and I claim that the two notions should
not be conflated (please see the discussion in Sect. 2 and footnote 4). However, I
am not addressing the issue of the ontological difference between affordances and
functions. The present paper does not aim at contributing to the ontological debate
on affordances. It only aims at making some conceptual distinctions, which might
be useful to provide a new epistemological framework to discuss specific problems.
As a case study, in this paper, I show how the distinction between different notions
of affordances might be useful to shed light on the debate on the role of motor pro-
cesses in the conceptual knowledge of artifacts.
In this paper, I will focus on the affordances of objects and, specifically, a sub-
set of objects, which includes things such as hammers, hats, and bottles. They
have some common features that distinguish them from other objects. First, they
are artifacts, that is, human-made objects that are different from natural objects
such as stones. Second, they are manipulable, hence different from other non-
manipulable artifacts, e.g., aircraft and skyscrapers. The analysis of the notion of
affordance with respect to artifacts generates a well-known, yet unresolved, issue
concerning what exactly the affordances of artifacts are. In the literature, at least
two different answers have been provided.
Gibson insists on the continuity between affordances of natural and artificial
objects. He writes:
“It is a mistake to separate the natural from the artificial as if there were two
environments; artifacts have to be manufactured from natural substances.”
(Gibson 1979, p. 130).
Although he discusses a variety of artifacts, such as scissors and knives, his
treatment of the affordances of artifacts does not differ from the treatment of
13
Synthese
1
Heft (1989) defines affordances as dispositional properties of the environment relative to the size of
some relevant body feature. His definition is based on the results of the classical empirical study by War-
ren (1984) on stair climbing affordances, which shows that participants’ judgements about whether or not
they could climb the stairs without using their arms or legs are based on a constant ratio between riser
height and leg length (this ratio, 0.88, is called by Warren “body-scaled” information).
13
Synthese
2
I do not mean to imply here that declarative knowledge is only of verbal nature. My claim is rather that
humans are often capable of making verbal statements on the basis of this knowledge. This is fully con-
sistent with the idea of non-verbal declarative knowledge; e.g., non-human animals may possess declara-
tive knowledge about their environment and their conspecific and express it non-verbally.
3
Declarative and procedural knowledge need not be of different nature or involve different represen-
tational formats. See Sect. 2.1 for further discussion of similar issues. See also Pezzullo (2011) for an
account of procedural and declarative knowledge in which they are defined, respectively, as on-line sen-
sorimotor anticipation and off-line simulations of potential actions.
13
Synthese
no right and wrong way of using the term affordance and both Gibson’s and Heft’s
notions capture relevant aspects related to how we interact with artifacts. Thus,
in order to accommodate both aspects without obfuscating essential differences
between them, I suggest making a distinction in the domain of affordances.
In the next section, first I introduce this distinction and explain how it relates to
other similar notions, then I argue for its neuropsychological plausibility drawing
upon evidence of a neuropsychological dissociation between function knowledge and
manipulation knowledge. The data I am going to present has also been discussed by
Garcea and Mahon (2012), who use it to support some specific claims about the role of
motor processes in the conceptual knowledge of artifacts. In Sect. 3, I analyze Garcea
and Mahon’s argument. They assume that function knowledge is more relevant than
manipulation knowledge as part of an artifact concept. I challenge this assumption
and, consequently, reject their argument. Finally, in Sect. 4, I show how my theoretical
distinction in the domain of affordances can be fruitfully applied in empirical research
to address the issue of the role of motor processes in artifact conceptual knowledge.
Building upon the previous discussion, here I introduce a distinction in the domain
of affordances between standard affordances and ad-hoc affordances.
Standard affordances are affordances of a class of objects and correspond to the
function4 of that class of objects, that is, what they are generally for. The function
of an object is culturally-derived with respect to certain goal-directed actions. For
example, mailboxes afford sending correspondence, and phones afford long-dis-
tance communication.5 Standard affordances may be stored in semantic long-term
4
Here and elsewhere in the paper, I discuss the notion of “affordance” drawing some analogies to the
notion of “function.” However, as I should make clear with the following discussion, the analogy is
restricted only to standard affordances and does not extend to ad-hoc affordances. This has two main
implications. First, the relation between standard affordances and functions is not clear-cut. My analysis
is consistent with the interpretation that standard affordances are, indeed, functions, that is, they are onto-
logically the same. Second, given that I emphasize the distinction between standard and ad-hoc affor-
dances, and ad-hoc affordances are not functions, I claim that the notion of “affordance” should not be
conflated with that of “function.”
5
In this respect, the notion of standard affordance is similar to that of canonical affordance introduced
by Costall (2012). However, I prefer the current term for two reasons. First, the term “canonical affor-
dance” has been used with different meanings in the philosophical and psychological literature. For
example, Borghi and Riggio (2009) use this term to refer to something radically different from what is
meant by Costall (2012): in their account, canonical affordances are a subset of temporary affordances
and are related to the typical orientation with which we interact with objects—for example, the typical
orientation with which we read a book. Second, the adjective “canonical” has become popular in the
neuroscientific literature to refer to a class of neurons in the F5 area of the premotor cortex of macaque
monkey, which discharge during motor act execution and during simple visual presentation of objects
(Rizzolatti and Fadiga 1998; Rizzolatti and Umiltà 2013). Thus, the term “canonical affordances” might
be misleading, because it might suggest a special relation to canonical neurons. Exploring the relation
between canonical neurons and different types of affordances is beyond the scope of this paper.
13
Synthese
memory as part of the mental concept under which that class of objects is catego-
rized. They may also be stored as part of the information generally associated to a
specific class of objects, thus including also typical uses of artifacts for purposes that
are not, strictly speaking, their intended functions (e.g., using chairs to reach things
standing on them; see below for a discussion of this point). They are relatively fixed
given a certain sociocultural environment and, as such, apply across several differ-
ent situations.6 An open question is whether or not they involve simulation of motor
processes that are also active during physical manipulation of objects. One possibil-
ity is that standard affordances are represented in a completely amodal way, that is,
decoupled from sensory-motor processes, such that their detection does not require
any simulation of processes also involved in actions and perceptions. An alternative
hypothesis is that knowing what an object is for involves knowing how to interact
with it, that is, standard affordances are represented via a concurrent implicit simu-
lation of the corresponding actual actions.7
Ad hoc affordances are instead affordances that a particular artifact has for an
individual agent in a specific situation. They are not necessarily related to the func-
tion of a certain object and are concerned, instead, with how we motorically interact
with that object, that is, how we can use it and manipulate it. For example, for a
particular agent in a given situation, a phone can take on the ad-hoc affordance of
throwing it at an intruder, and a mailbox can take on the ad-hoc affordance of sit-
ting on it. Ad hoc affordances are context-dependent because they arise in a specific
situation and are not culturally-derived (although a certain sociocultural context can
affect which ad-hoc affordances are selected by the agent8). Being tied to a certain
agent and her specific aims in a particular situation, ad-hoc affordances are not pre-
stored in semantic long-term memory and need to be processed “on the fly.” The
processing of ad-hoc affordances seems to require simulation of motor interactions
with objects in order to establish whether a particular object can afford a certain
action to a particular agent given a specific goal. This simulation is, in most cases,
implicit, but it can become explicit if the situation requires it.
It might be objected that ad-hoc affordances might not require at all the reac-
tivation of motor processes active during actual manipulation of objects, that is,
they might not require any motor simulation. For example, one could argue that in
order to decide that a pen can be used to scratch one’s back, it might be sufficient
to retrieve enough factual information about pens to infer that a pen can be used
to this aim (e.g., pens are generally a certain length and rigid enough, and so on).
6
This phrasing is consistent with the possibility that in a certain given situation, standard affordances
might not be appropriately detected. For example, a mailbox does not afford sending correspondence if it
is pulled away and lying on the ground. Still, we would maintain that mailboxes do generally afford send-
ing correspondence. The concept of a mailbox seems to include what mailboxes are generally for, that is,
what type of function is usually supported by them.
7
The paper is neutral with respect to this issue. In fact, a crucial claim of the paper is that it is not nec-
essary to put this issue at stake to address the question of the role of motor processes in the conceptual
knowledge of artifacts. See Sect. 3 for a detailed argument in support of this claim.
8
For example, at an elegant cocktail party, it might be socially inappropriate for a particular agent to
stand on the buffet table. This particular ad-hoc affordance is, then, very unlikely to be selected by the
agent in that context.
13
Synthese
This factual information can be stored in an amodal format, then retrieving it and
using it in the inferential process would not require any reactivation of motor pro-
cesses related to the actual manipulation of pens. Although this is a possibility, I
would argue that this is an empirical question and should be treated as such, that
is, by making explicit the different empirical predictions held by these two oppo-
site accounts, the simulation view, and the amodal view. Consider, for example, an
implausible scenario which would be novel for most people, such as using a fun-
nel glued to the wall to hang up the coat. The inferential steps connecting the fun-
nel to the coat-hanging affordance are quite many and with no previous experience
of using a funnel to this aim, going through these inferential steps would certainly
require a certain amount of processing time. In particular, the processing time would
be plausibly longer than the processing time required to establish that a funnel can
be used to pour water into a container, given that this information is likely to be
already stored in our long-term memory and readily accessible. Thus, comparing
the processing times for these two different conditions, one might be able to test
whether extrapolating an ad-hoc affordance for an object takes longer than retrieving
its standard affordance, which would support the amodal view. The results of some
relevant behavioral studies, in which processing time is operationalized as reading
time, are at odds with the predictions of the amodal account, showing that no addi-
tional reading time is required for the ad-hoc affordance condition compared to the
standard affordance condition (see Glenberg 1999: he does not use this jargon, but
the comparison of interest involves clearly what I call ad-hoc and standard affor-
dances). These results suggest that ad-hoc affordances are not derived from a chain
of inferential reasoning and, every other condition being equal, motor simulation
seems to be the best predictor of this behavioral data. In Sect. 4, I will briefly men-
tion the results of a recent ERP study that confirm the behavioral data and speak to
this issue.
It should be noted that neither standard nor ad-hoc affordances fully correspond
to the affordances of the environment introduced by Gibson. As noted above (see
Sect. 1), Gibson defines affordances as related to species-specific or transcultural
features of objects (size, texture, what they are made of, distance, and so on). Thus
Gibson’s affordances of the environment do not account for the socioculturally
specified information conveyed by artifacts. With respect to my distinction between
standard and ad-hoc affordances, Gibson’s definition of the “affordances of the envi-
ronment” does not account for standard affordances, in the extent to which standard
affordances are concerned with culturally-derived uses of artifacts. The notion of
the “affordances of the environment” introduced by Gibson may be partially used
to understand what type of affordances are the “ad-hoc affordances”, in the extent
to which they are concerned with transcultural features of artifacts, that is, features
that allow an agent to use artifacts in ways that are independent of any recognized
cultural practice. However, Gibson’s notion of the “affordances of the environ-
ment” cannot fully capture the notion of “ad-hoc affordances,” either. Unlike Gib-
son’s affordances of the environment, ad-hoc affordances are detected also in rela-
tion to the aims of the agent and, as such, they pertain to more complex actions,
which require a combination of simpler actions. For instance, a small ball of paper
affords capturing someone’s attention if thrown to that person’s direction. Whereas
13
Synthese
this ad-hoc affordance for the ball of paper requires the combination of several Gib-
sonian affordances (specified with respect to the material, shape, and size of the
object), it goes beyond Gibsonian affordances as it pertains to a complex action
guided by the agent’s intention of capturing someone’s attention.
For similar reasons, ad-hoc affordances are different from microaffordances as
introduced by Ellis and Tucker (2000). Microaffordances are even more precise than
Gibsonian affordances as they are specified with respect to some components of
actions, which are independent of the goals of the agent and related, for instance,
to the size or orientation of an object (see Ellis and Tucker 2000; Tucker and Ellis
1998, 2001). Microaffordances do not capture complex actions, while ad-hoc affor-
dances do, given that they are sensitive to the goals of the agent in a certain situation.
That is, ad-hoc affordances can specify new and unusual functions for an artifact.
These novel action possibilities, though, are not limitless, they are constrained by
limited motor patterns associated with how an object can be actually manipulated.
The distinction between standard and ad-hoc affordance seems similar to the dis-
tinction between proper and system function, as introduced by Preston (1998; see
also Heersmink 2016; Houkes and Vermaas 2010). The notion of proper function
traces back to Millikan (1984) and concerns “…something specific that [artifacts as
well as biological traits] are supposed to do, even though they may never perform
this function, or may be temporarily co-opted for some other use” (Preston 1998, p.
215). For example, the proper function of chairs is to support seated humans, and
this is why they are made and purchased. Sometimes chairs can also be used to stand
on them to get things off high shelves, but this is not the reason why they are there in
the first place. Preston’s pluralist theory of function acknowledges that artifacts (and
biological traits) also have system functions, which “depend on the current capaci-
ties and dispositions of things […]. System functions are also typically established
for individual things or uses of things because functionality, in this case, depends
not on the history of the thing itself, but on the containing system” (ib., p. 239).
Unlike proper functions, system functions do not concern what artifacts are designed
for, but rather the additional performances that they may end up being used for, like
using chairs to stand on them and change light bulbs. According to Preston, “the
main difference between system function and proper function is … that the latter is
normative, whereas the former is not” (ib., p. 224). The difference in normative sta-
tus also accounts for another important difference between the two notions. The nor-
mativity that defines proper function allows us to draw a clear line between function
and accident (Wright 1974), that is, between performances which fulfill a function
(e.g., the heart circulating blood) and performances which are a mere by-product of
the ‘real’ function (e.g., the heart making a characteristic noise). The notion of “sys-
tem function,” on the other hand, has no strong commitment to a distinction between
‘real’ and accidental functions, thus drawing the function-accident line much more
liberally (Cummins 1975).
The distinction between proper and system function resembles the distinction
between standard and ad-hoc affordance. The notion of “standard affordances”
could be understood similarly to that of “proper functions” of artifacts (what they
are for), whereas the notion of “ad-hoc affordances” looks similar to that of “sys-
tem functions” (what they may end up being used for). Even though there are
13
Synthese
indeed some similarities, there are also some crucial differences between mine
and Preston’s distinction. In order to see that, it may be useful to mention a fur-
ther distinction made by Preston between two types of system functions, namely,
standardized ongoing exaptations and idiosyncratic ongoing exaptations. Stand-
ardized ongoing exaptations are repeated uses of artifacts by a community of
users for purposes that are not their proper functions. For example, screwdrivers
are regularly used to open paint cans, chairs to reach things, and so on. These
additional functions are the result of widespread cultural practice; indeed, arti-
facts may end up being used at least as often for their system-functional use than
for their proper-functional use.
On the other hand, idiosyncratic ongoing exaptations are concerned with indi-
viduals’ creative uses of artifacts on specific occasions. For example, a funnel glued
to the wall by its large end can be used as a coat rack. There is no corresponding and
recognized cultural practice in this type of use.
Given my definition of standard and ad-hoc affordances (see above), the notion
of “standard affordances” does not only include that of artifacts’ “proper functions,”
but also that of “standardized ongoing exaptations.” Standard affordances concern
all types of culturally-derived uses of artifacts, which may be stored in seman-
tic long-term memory as part of the information related to artifacts. Ad hoc affor-
dances, on the other hand, are not pre-stored, but determined “on the fly,” as the
result of individuals’ creative uses. The notion of “ad-hoc affordances” covers Pres-
ton’s description of idiosyncratic ongoing exaptations, but also a third type of sys-
tem function, introduced by Heersmink (2016), namely, idiosyncratic exaptations.
Idiosyncratic exaptations are not ongoing, that is, they are uses of artifacts that are
not even well-established (or fairly well-established) for individuals (e.g., Preston’s
example of using a shoelace to tie up a tomato plant). For example, someone may
use a screwdriver as a weapon to defend oneself from an intruder. Such uses are
improvised one-offs, which may occur only once or twice in a lifetime.
Thus, despite the prima facie similarity, the distinction between proper and sys-
tem function does not correspond to the distinction between standard and ad-hoc
affordance, even though there is some overlap. More precisely, proper and system
functions are distinguished mainly according to their normative status, as we saw
above. Normativity, though, does not clear-cut the distinction between standard and
ad-hoc affordance, whereas cultural practice and the existence of pre-stored seman-
tic information do.
So far, I motivated the current distinction mainly by theoretical arguments.
However, I would now like to suggest that this distinction is also consistent with
the neuropsychological literature. More precisely, function knowledge and manipu-
lation knowledge are doubly dissociable. Given that my distinction between stand-
ard and ad-hoc affordances is primarily based on the distinction between function
and manipulation, I suggest that the neuropsychological dissociation between them
provides a further reason to distinguish between these two types of affordances and
reveals that such a distinction, independently motivated by theoretical reasons, is
also plausible from the neuropsychological point of view. Additionally, this corre-
spondence provides a significant advantage, that is, connecting the philosophical
debate on affordances to the experimental research. Before delving into the empirical
13
Synthese
13
Synthese
the dissociation between manipulation knowledge and function knowledge has also
been reported in healthy adults (Garcea and Mahon 2012).
The dissociation between function and manipulation strengthens the idea of mak-
ing a distinction in the domain of affordances between those related to the artifact
function and those related to the artifact manipulation. This dissociation, however,
has been argued to carry some specific implications which concern, in particular,
the debate on the role of motor processes in the conceptual knowledge of artifacts.
In the next section, I review an argument put forward by Garcea and Mahon (2012)
who, on the basis of the dissociation between function knowledge and manipula-
tion knowledge, claim that conceptual knowledge of artifacts does not necessarily
involve the reactivation of motor processes. I will show that the argument has some
problems because one of the assumptions on which is based can be challenged, that
is that function, and manipulation knowledge (hence, standard and ad-hoc affor-
dances) do not have an equal status in the conceptual knowledge of artifacts.
Given the dissociation between function and manipulation, one could argue that the
reactivation of motor processes, which is plausibly needed when participants make
manipulation judgments, is not required to retrieve and process function knowledge.
Garcea and Mahon (2012) endorse precisely this claim. What is more, on the basis
of that, they reject the claim that the conceptual knowledge of artifacts involves the
reactivation of motor processes active when objects are actually manipulated (see
Barsalou 1999; Gallese and Lakoff 2005). The motivation for such a conclusion is
that, in their view, function knowledge is more relevant than manipulation knowl-
edge in artifacts conceptual knowledge, given that function knowledge (but not
manipulation knowledge) is part of the core of an artifact concept. This means that
entertaining information about an artifact’s function is necessary for a subject to be
said to possess at least a minimal concept of that artifact.9 Such a minimal or core
concept, on the other hand, need not involve any simulation of motor processes, thus
leading to the conclusion that motor processes are not constitutive of artifact con-
ceptual knowledge, at least in the minimal sense.
9
Even though Garcea and Mahon phrase their claim in terms of the necessity of function knowledge
for the possession of a minimal concept of an artifact, their discussion of apraxic patients (see below in
the text) seems to involve that function knowledge is even sufficient for a minimal concept possession.
However, they are careful not to phrase their hypothesis this way, so the following discussion will only
assume the requirement of necessity. Whether or not the notion of a minimal or core concept is an ade-
quate one, will remain on the background of the current discussion. For the sake of the argument, I will
assume that this notion is adequate, however it should be noted that it is not unproblematic. Pulvermüller
(2013) has pointed out to a number of criticisms concerning this notion in his discussion of the idea, pre-
sented by Mahon and Caramazza (2008) and Bedny and Caramazza (2011), that the functional contribu-
tion of sensorimotor systems to conceptual or semantic processing would consist in enriching (“coloring”
and “dressing”, Mahon and Caramazza 2008, p. 68f) the conceptual representation.
13
Synthese
As evidence supporting their claim, the authors report on the case of apraxic
patients, who are impaired in manipulating objects (i.e., grasping and using them)
and even in pantomiming object use, even though they do not have any basic sen-
sory or motor dysfunction (e.g., Liepmann 1977; Rothi et al. 1991; for reviews see
Cubelli et al. 2000; Leiguarda and Marsden 2000; Mahon and Caramazza 2005).
Despite their deficit, apraxic patients seem to be still capable of naming objects
when visually presented. As far as naming a visually presented object is diagnos-
tic of concept possession, apraxic patients do have conceptual knowledge of arti-
facts despite not being able to manipulate objects and, thus, not being able arguably
to reactivate the related motor processes offline, when they would be required to
retrieve conceptual knowledge related to the visually presented object.10
The structure of the argument put forward by Garcea and Mahon (2012) is an
illustration of a more general strategy to adjudicate between different accounts of the
role of motor processes in conceptual knowledge of artifacts, and it is then worth-
while to analyze it more closely. Here I will argue that the argument does not go
through. The reason is that the conclusion endorsed by the authors is based on a
premise that can be challenged. In particular, I will argue that the premise by which
function knowledge is more relevant than manipulation knowledge in an artifact
conceptual knowledge should be rejected.
In their paper, Garcea and Mahon introduce two opposite views on the role of
motor processes in the conceptual knowledge of artifacts: the Embodied Cognition
Hypothesis (ECH) and their hypothesis that in other papers they call the Grounding
by Interaction Hypothesis11 (GIH; Mahon and Caramazza 2009, 2011). The main
disagreement between these hypotheses, as characterized by the authors, can be cap-
tured in the following claims.
Grounding by Interaction Hypothesis:
GIH1: “… part of the core of an artifact concept is knowledge of function
(‘what for’)…” (Garcea and Mahon 2012, p. 3)
GIH2: “… while there is significant interaction and exchange of information
between the systems that represent manipulation, function, and visual knowl-
edge, they are nonetheless functionally dissociable systems and motor knowl-
edge is not constitutive of function knowledge” (ib., p. 3).
Embodied Cognition Hypothesis:
ECH1: “… conceptual processing of tools necessarily involves the retrieval or
simulation of motor information …” (ib., p. 2).
10
It should be noted, though, that apraxic patients have been found to be able to associate an appropri-
ate hand posture to novel objects, but not to familiar objects (Buxbaum et al. 2003). Thus, manipulation
knowledge might be compromised in apraxic patients only for familiar objects, but not for novel objects.
11
The Grounding by Interaction Hypothesis holds that there is a level of conceptual content which is
“abstract” or “symbolic”, that is, not constituted by sensory and motor information. However, sensori-
motor systems functionally contribute to conceptual processing by “coloring” or “dressing” the concept
(Mahon and Caramazza 2008, p. 68f; see also note 10 above).
13
Synthese
Notably, the arguments are reconstructed by Garcea and Mahon in such a way that
both hypotheses assume the claim (a), that is, that knowledge about an artifact’s
function is part of the core of that artifact’s concept. The claim (a) is then presented
as unproblematic. Where the two hypotheses diverge is in the claims (b) and (b′),
respectively rejecting or endorsing the idea that reactivation of motor processes
active during actual manipulation of objects is required to retrieve or compute func-
tion knowledge (and, hence, is constitutive of it). From that, different conclusions
follow as to whether or not motor-relevant information about how to manipulate
an artifact is also part of that artifact concept ((c) and (c′)). The crux of the argu-
ment seems to be whether (b) or (b′) is correct, that is, whether or not manipulation
knowledge is constitutive of function knowledge. Here, I argue that this is not the
crux of the argument and that whether (b) or (b′) is correct is irrelevant to adjudi-
cate between the Grounding by Interaction Hypothesis and the Embodied Cognition
Hypothesis.
One could start from a different assumption and claim that motor-relevant infor-
mation associated to object use and manipulation can account as well as function
knowledge for the specificity of artifacts, that is, manipulation knowledge is part of
the core of an artifact concept. If we assume this starting point, then it becomes pos-
sible to make sense of some findings, which are not compatible with the view that
13
Synthese
function knowledge has a privileged status as part of the core of artifact concepts. I
will briefly mention some of these findings and explain how they challenge the view
that function knowledge is more relevant than manipulation knowledge in artifact
conceptual knowledge (for more extensive reviews see, e.g., Kellenbach et al. 2003;
Boronat et al. 2005).
Given the assumption of a privileged relationship between function knowledge
and the conceptual knowledge of artifact, one would expect that a deficit in func-
tion knowledge accompanies a deficit in representing artifacts. However, it has
been reported that functional feature knowledge is spared in patients with selective
impairments in representing artifacts compared to animals (Lambon-Ralph et al.
1998). Thus, the assumption that artifacts are characteristically represented mainly
according to their function is at odds with these results. A second relevant finding
concerns a peculiar pattern of spared and compromised knowledge in some patients
who have deficits in representing smaller artifacts but not large artifacts with clear
functions (e.g., buildings, vehicles) (Warrington and McCarthy 1983, 1987). This
pattern cannot be explained by a deficit in representing function knowledge because
the large artifacts used in the task have a clear function. Some other relevant dimen-
sion must be involved in artifact conceptual knowledge to explain the dissociation
between small and large artifacts. Finally, deficits in body-part knowledge tend to
co-occur with deficits in artifact knowledge (Sacchett and Humphreys 1992; War-
rington and McCarthy 1987). Again, this is difficult to explain assuming the priority
of function knowledge in artifact conceptual knowledge, because a deficit in rep-
resenting function knowledge would not explain why body-part knowledge is also
compromised given that body-parts do not have a clear function.
These results challenge the idea that artifacts are mainly represented according to
their function (of course, this is still consistent with the idea that function knowledge
is relevant to artifact concepts). An alternative view, which would allow us to make
sense of the findings reported above, is that another form of knowledge can account
for the specificity of artifacts, that is, manipulation knowledge. According to this
account, it is possible to explain the above findings as related to deficits in manipu-
lation knowledge. It is worth reminding at this stage that the current discussion has
been focusing on manipulable artifacts (see Introduction). What is at stake here is
the strong claim (exemplified by Garcea and Mahon’s argument) that even manipu-
lable artifacts are characteristically represented according to their function, while
motor-relevant information associated to their actual use and manipulation is not
part of the core of those artifact concepts. This strong claim is not only at odds with
the evidence so far presented but also with additional neuropsychological and neuro-
imaging results showing differences between manipulable and non-manipulable arti-
facts. Whereas these differences cannot be accounted for if one assumes that all arti-
facts, including manipulable ones, are represented mainly in terms of their function,
they are consistent with the idea that motor-relevant information related to object
use and manipulation is crucially involved in representing manipulable artifacts.
Buxbaum and Saffran (2002) showed that apraxic patients had relative deficits
in representing artifacts, and these deficits were associated with an impairment
in manipulation knowledge, but not function knowledge. This suggests the pos-
sibility that manipulation knowledge subserved by sensory-motor cortices plays a
13
Synthese
13
Synthese
section, I suggest that a valid strategy would consist in directly contrasting the rela-
tive contribution of these forms of knowledge and I illustrate an example of such
a strategy which, unlike the studies reviewed above, contrasts specifically the two
notions introduced here, that is, standard affordances and ad-hoc affordances.
The discussion in the previous paragraph has shown that the dissociation between
function knowledge and manipulation knowledge does not license any conclusion
as to the role of motor processes in artifact conceptual knowledge. I suggested that
manipulation knowledge might be part of the core of an artifact concept even though
it is not constitutive of function knowledge. Given that this dissociation provides a
criterion for the neuropsychological plausibility of the distinction between standard
and ad-hoc affordances (so I have argued), the conclusion drawn from the previous
discussion can be rephrased as follows. Ad hoc affordances related to how objects
are manipulated are as relevant as standard affordances in the conceptual knowl-
edge of artifacts because they also can be argued to be part of the core of an artifact
concept.
Assuming that my argument has so far gone through, an opponent might object
that there is an obvious problem with my proposal of considering manipulation
knowledge as part of the core of an artifact concept. Now that I have rephrased the
proposal in terms of standard and ad-hoc affordance, it seems clear that the pro-
posal cannot be accepted because ad-hoc affordances cannot possibly be part of the
core of an artifact concept. An example will illustrate the skepticism of the oppo-
nent. Consider a situation in which a person is sitting on a bench and wishes not to
meet an acquaintance who is approaching. In this special circumstance, a newspaper
bought a few minutes before can take on an ad-hoc affordance of hiding behind it.
In what sense should this ad-hoc affordance be considered as part of the core of my
concept of a newspaper with the same rights of its standard affordance of reading?
This seems highly counterintuitive because I might have never experienced such an
ad-hoc affordance for the newspaper before, and still it seems that on my account I
would be forced to include it in my concept of a newspaper.
Here is my reply. If in the example above we replace the newspaper with a match-
box, the scenario just described becomes completely senseless. Whereas it is clear
to everyone that a newspaper can afford hiding behind it, no one would doubt that
a matchbox does not possibly afford such an action. We all know that, even though
most of us are unlikely to have experienced a similar scenario before. How do we
know that? A possibility, supported by experimental studies (see, for example, Glen-
berg and Robertson 2000), is that we mentally simulate motor interactions with those
objects reactivating offline motor processes that are active during physical interactions
with them. This implicit simulation of motor-relevant information guides our current
interactions with objects even in situations never experienced before, and—what is
more—guides our conceptual processing of the situation (allowing us to establish that a
newspaper affords the hiding-behind action while a matchbox does not) when it is just
13
Synthese
linguistically described as in our current example. This is not to say that the concept of
a newspaper includes some pre-stored information concerning the possibility of using
it to hide from someone. What I mean here in claiming that motor-relevant informa-
tion is part of the core of the concept of a newspaper is that we can flexibly draw upon
this information, reactivating it via implicit motor simulation of actual interactions with
newspapers, to establish on the fly that a newspaper affords the hiding action.
Consistent with this hypothesis, in an event-related potential (ERP) study, we
showed that the ad-hoc affordances induced by the linguistic context while reading
short stories were integrated immediately during semantic comprehension (Cosen-
tino et al. 2017). The experimental task required subjects to read short stories in
which conflict was triggered between standard affordances and the contextually-
induced ad-hoc affordances. For example, in a situation in which a funnel has been
glued to the wall, and the agent wants to hang up her coat, an ad-hoc affordance of
“hanging” is induced for the funnel which conflicts with the standard affordance of
“pouring” (i.e., pouring a liquid into a container). If information about the func-
tion of the object, stored in semantic long-term memory, were more prominent than
motor-related information derived by interactions with the object (and reactivated
when processing the current situation described linguistically), we should expect
that the standard affordance of funnel is activated by default independently of the
situation at hand and only subsequently replaced by an ad-hoc affordance contex-
tually-derived in that specific situation. The timing of this default activation and
subsequent replacement during semantic comprehension is, of course, too quick to
be captured in a behavioral experiment. However, using the electroencephalography
(EEG) method is possible to track these processes millisecond by millisecond. Our
study focused on the N400, which is an ERP component thought to reflect lexical
retrieval and/or semantic integration processes (see Kutas and Federmeier 2011,
for a review). Measuring the N400 at the onset of the action verb (e.g., hang/pour)
which followed the artifact noun (e.g., funnel) it was possible to establish that given
a situation that triggers an ad-hoc affordance for an artifact, the ad-hoc affordance is
immediately integrated as part of the contextually specified meaning of the noun and
the standard affordance of that artifact is not at all activated.
Taken together, behavioral and ERP results suggest that ad-hoc affordances are
as prominent as standard affordances in the conceptual knowledge of artifacts. This
is consistent with the view that motor-relevant information related to how artifacts
are manipulated might constitute part of the core of an artifact concept regardless
of whether or not it also constitutes function knowledge. Thus, the contrast between
standard and ad-hoc affordances provides a new promising way to experimentally
address the issue of what constitutes an artifact concept and what is the role of motor
processes in artifact conceptual knowledge.
5 Conclusions
13
Synthese
Acknowledgements This research has been presented at the 24th Annual Meeting of the European Soci-
ety for Philosophy and Psychology (ESPP) in St. Andrews, at the CoSt2016 in Düsseldorf, and in the
research colloquium of the Mercator Research Group “Structure of Memory.” I would like to thank the
participants in these events for their helpful comments. Also, I thank the colleagues who contributed to
this research with their valuable insights, in particular, Anna M. Borghi, Francesco Ferretti, and Markus
Werning.
References
Barsalou, L. W. (1999). Perceptual symbol systems. Behavioral and Brain Sciences, 22, 577–660.
Bedny, M., & Caramazza, A. (2011). Perception, action, and word meanings in the human brain: The
case from action verbs. Annals of the New York Academy of Sciences, 1224, 81–95.
Borghi, A. M., & Riggio, L. (2009). Sentence comprehension and simulation of objects temporary,
canonical and stable affordances. Brain Research, 1253, 117–128.
Boronat, C. B., Buxbaum, L. J., Coslett, H. B., Tang, K., Saffran, E. M., Kimberg, D. Y., et al. (2005).
Distinctions between manipulation and function knowledge of objects: Evidence from functional
magnetic resonance imaging. Cognitive Brain Research, 23, 361–373.
Buxbaum, L. J., & Saffran, E. M. (2002). Knowledge of object manipulation and object function: Disso-
ciations in apraxic and non-apraxic subjects. Brain and Language, 82, 179–199.
Buxbaum, L. J., Schwartz, M. F., & Carew, T. (1997). The role of semantic memory in object use. Cogni-
tive Neuropsychology, 14, 219–254.
Buxbaum, L. J., Sirigu, A., Schwartz, M. F., & Klatzky, R. (2003). Cognitive representations of hand
posture in ideomotor apraxia. Neuropsychologia, 41, 1091–1113.
Buxbaum, L. J., Veramonti, T., & Schwartz, M. F. (2000). Function and manipulation tool knowledge in
apraxia: Knowing “what for” but not “how”. Neurocase, 6, 83–97.
Capitani, E., Laiacona, M., Mahon, B., & Caramazza, A. (2003). What are the facts of semantic cate-
gory-specific deficits? A critical review of the clinical evidence. Cognitive Neuropsychology, 20,
213–261.
Chemero, A. (2003). An outline of a theory of affordances. Ecological Psychology, 15(2), 181–195.
13
Synthese
Cosentino, E., Baggio, G., Kontinen, J., & Werning, M. (2017). The time-course of sentence meaning
composition. N400 effects of the interaction between context-induced and lexically stored affor-
dances. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2017.00813.
Costall, A. (2012). Canonical affordances in context. Avant: Trends in Interdisciplinary Studies, 3(2),
85–93.
Cubelli, R., Marchetti, C., Boscolo, G., & Della, Salla S. (2000). Cognition in action: Testing a model of
limb apraxia. Brain and Cognition, 44, 144–165.
Cummins, R. (1975). Functional analysis. The Journal of Philosophy, 72(20), 741–764.
Devlin, J., Russell, R., Davis, M., Price, C., Moss, H., Fadili, M., et al. (2002). Is there an anatomical
basis for category-specificity? Semantic memory studies in PET and fMRI. Neuropsychologia, 40,
54–75.
Ellis, R., & Tucker, M. (2000). Micro-affordance: The potentiation of components of action by seen
objects. British Journal of Psychology, 91, 451–471.
Gallese, V., & Lakoff, G. (2005). The brain’s concepts: The role of the sensory-motor system in concep-
tual knowledge. Cognitive Neuropsychology, 22, 455–479.
Garcea, F. E., & Mahon, B. Z. (2012). What is in a tool concept? Dissociating manipulation knowledge
from function knowledge. Memory and Cognition, 40(8), 1303–1313.
Gerlach, C., Law, I., Gade, A., & Paulson, O. B. (1999). Perceptual differentiation and category effects in
normal object recognition: A PET study. Brain, 122, 2159–2170.
Gibson, J. J. (1979). The ecological approach to visual perception. Boston: Houghton-Mifflin.
Glenberg, A. (1999). Why mental models must be embodied. Advances in Psychology, 128, 77–90.
Glenberg, A. M., & Robertson, D. A. (2000). Symbol grounding and meaning: A comparison of high
dimensional and embodied theories of meaning. Journal of Memory and Language, 43, 379–401.
Heersmink, R. (2016). The metaphysics of cognitive artefacts. Philosophical Explorations, 19(1), 78–93.
Heft, H. (1989). Affordances and the body: An intentional analysis of Gibson’s ecological approach to
visual perception. Journal for the Theory of Social Behavior, 19, 1–30.
Houkes, W. (2006). Knowledge of artefact functions. Studies in History and Philosophy of Science Part
A, 37(1), 102–113.
Houkes, W., & Vermaas, P. E. (2010). Technical functions: On the use and design of artefacts (Vol. 1).
Berlin: Springer.
Kellenbach, M., Brett, M., & Patterson, K. (2003). Actions speak louder than functions: The importance
of manipulability and action in tool representation. Journal of Cognitive Neuroscience, 15, 30–46.
Kutas, M., & Federmeier, K. D. (2011). Thirty years and counting: Finding meaning in the N400 compo-
nent of the event-related brain potential (ERP). Annual Review of Psychology, 62, 621–647.
Lambon-Ralph, M. A., Howard, D., Nightingale, G., & Ellis, A. W. (1998). Are living and non-living
category-specific deficits causally linked to impaired perceptual or associative knowledge? Evidence
from a category-specific double dissociation. Neurocase, 4, 311–338.
Leiguarda, R. C., & Marsden, C. D. (2000). Limb apraxias: Higher-order disorders of sensorimotor inte-
gration. Brain, 123, 860–879.
Liepmann, H. (1977). The syndrome of apraxia (motor asymboly) based on a case of unilateral apraxia.
(A translation from Monatschrift für Psychiatrie und Neurologie). In D. A. Rottenberg & F. H.
Hockberg (Eds.), Neurological classics in modern translation (Vol. 8, pp. 15–44). New York: Mac-
millan Publishing Co.
Magnie, M. N., Teixeira Ferreira, C., Giusiano, B., & Poncet, M. (1999). Category specificity in object
agnosia: Preservation of sensorimotor experiences related to objects. Neuropsychologia, 37, 67–74.
Mahon, B. Z., & Caramazza, A. (2005). The orchestration of the sensory-motor systems: Clues from neu-
ropsychology. Cognitive Neuropsychology, 22, 480–494.
Mahon, B. Z., & Caramazza, A. (2008). A critical look at the embodied cognition hypothesis and a new
proposal for grounding conceptual content. Journal of Physiology – Paris, 102(1–3), 59–70.
Mahon, B. Z., & Caramazza, A. (2009). Concepts and categories: A cognitive neuropsychological per-
spective. Annual Reviews in Psychology, 60, 27–51.
Mahon, B. Z., & Caramazza, A. (2011). What drives the organization of object knowledge in the brain?
Trends In Cognitive Science, 15, 97–103.
Moore, C. J., & Price, C. J. (1999). A functional neuroimaging study of the variables that generate cate-
gory-specific object processing differences. Brain, 122, 943–962.
Millikan, R. G. (1984). Language, thought, and other biological categories. Cambridge, MA: MIT Press.
Negri, G. A., Lunardelli, A., Reverberi, C., Gigli, G. L., & Rumiati, R. I. (2007). Degraded semantic
knowledge and accurate object use. Cerebral Cortex, 43, 376–388.
13
Synthese
13