The science of cycology: Failures to understand how everyday objects work

I think I know less than I thought.
—Participant’s comment, after completion of the
bicycle drawing task
Recent research has suggested that people often overes-
timate their ability to explain how things function. Rozen-
blit and Keil (2002) found that people overrated their un-
derstanding of complicated phenomena. This illusion of
explanatory depth was not merely due to general overcon-
fidence; it was specific to the understanding of causally
complex systems, such as artifacts (crossbows, sewing
machines, microchips) and natural phenomena (tides, rain-
bows), relative to other knowledge domains, such as facts
(names of capital cities), procedures (baking cakes), or nar-
ratives (movie plots). Rozenblit and Keil (2002; see also
Mills & Keil, 2004) investigated why people overestimated
their knowledge and which factors changed their ratings
when they were confronted with evidence of the inadequacy
of their understanding. Rozenblit and Keil found that peo-
ple reduced their estimation of their own knowledge after
having to provide functional explanations. Similarly, stud-
ies of naive physics have demonstrated that adults’ under-
standing of mechanics (e.g., trajectories of falling objects
and displacement of liquids; Kaiser, Proffitt, Whelan, &
Hecht, 1992) and optics (e.g., mirrors; Bertamini, Spooner,
& Hecht, 2003) is often sketchy, inaccurate, and inconsis-
tently applied. Together, these results demonstrate weak-
nesses in people’s understanding of everyday objects and
events, and they appear to provide evidence against expla-
nation-based (or theory-based) theories of categorization
(Murphy, 2002; Murphy & Medin, 1985).
Explanation-based theories assume that conceptual
knowledge is represented like scientific theories in that
concepts are causally related to each other and that expla-
nations underlie what can be directly observed (Ahn &
Luhmann, 2005). Murphy and Medin (1985) and, from de-
velopmental research, Carey (1985) and Keil (1989), have
proposed that our concepts are our ways of explaining the
world to ourselves and others. Explanation-based theories
contrast with similarity-based theories of categorization
(prototype and exemplar theories), which assume that
there is some simpler means of categorizing stimuli on the
basis of similarity clusters in an internal representational
space (Murphy, 2002). Explanation-based theories instead
suggest that features of concepts are related to each other
(often causally) within structured internal representations.
These more complex representations incorporate our gen-
eral knowledge of the world and our intuitive theories about
which aspects of a stimulus are most important when we
are judging similarity (e.g., causal vs. superficial features).
Concepts are thus not merely lists of independent features
but rather include an “explanation” of the appearance,
function, and other aspects of exemplars of that concept. It
is important to note that most proponents of explanation-
based theories emphasize that these “theories” are informal
explanations of the world and generally are not fully coher-
ent, organized, scientific accounts (Murphy, 2002).
Under certain circumstances, background knowledge
has been found to influence concept acquisition and cat-
egorization performance (Lin & Murphy, 1997; Murphy,
2002; Pazzani, 1991;Wisniewski, 1995). Nevertheless, this
information may not be routinely available to us (Smith &
1667 Copyright 2006 Psychonomic Society, Inc.
This research was supported by Grant RES-000-27-0162 from the
Economic and Social Research Council. Thanks to Tom Mortensen and
JulianTodd for the original inspiration for these studies, to the Liverpool
Visual Perception Lab Group for helpful discussions, and to the vol-
unteer participants from University Open Days, the Merseyside Cycle
Campaign, the Liverpool Wheelers, the Cambridge University Caving
Club, and the Red Rose Cave and Pothole Club. Correspondence should
be addressed to R. Lawson, School of Psychology, University of Liver-
pool, Eleanor Rathbone Building, Bedford Street South, Liverpool L69
7ZA, England (e-mail: rlawson@liverpool.ac.uk).
The science of cycology:
Failures to understand
how everyday objects work
REBECCA LAWSON
University of Liverpool, Liverpool, England
When their understanding of the basics of bicycle design was assessed objectively, people were
found to make frequent and serious mistakes, such as believing that the chain went around the front
wheel as well as the back wheel. Errors were reduced but not eliminated for bicycle experts, for men
more than women, and for people who were shown a real bicycle as they were tested. The results
demonstrate that most people’s conceptual understanding of this familiar, everyday object is sketchy
and shallow, even for information that is frequently encountered and easily perceived. This evidence
of a minimal and even inaccurate causal understanding is inconsistent with that of strong versions of
explanation-based (or theory-based) theories of categorization.
Memory & Cognition
2006, 34 (8), 1667-1675

1668 lawson
Sloman, 1994). In particular, if our causal understanding of
familiar and relatively simple objects is scanty or inaccu-
rate—as Rozenblit and Keil’s (2002) results suggest—then
this knowledge may not be a central component of our con-
cepts. Instead, a well specified similarity-based model of
categorization may provide a better account of our normal
conceptual processes than explanation-based theories do.
Keil and Wilson (Keil, 2003b; Wilson & Keil, 2000),
however, have argued that the finding that people have
an illusion of explanatory depth (Rozenblit & Keil, 2002)
is not necessarily inconsistent with explanation-based
theories, even though such theories assume that causal
and functional knowledge underpin our concepts. Keil
(2003b) has suggested that, even if our understanding of
how objects function is fragmentary and shallow, it may be
sufficient to track the causal structure of the world. Fur-
thermore, this minimal understanding would benefit us
by not overburdening our limited information-processing
and information-storage capacities.
Rozenblit and Keil (2002) focused on subjective mea-
sures of people’s knowledge of objects and phenomena.
Their primary measure was the change in a person’s self-
ratings of the quality of his or her written functional ex-
planations. In order to extend their findings, in the present
study I focused on measuring people’s objective under-
standing of everyday objects by asking them to draw object
parts and to answer forced-choice questions about object
function. This study assessed people’s understanding of
the functioning of a familiar and relatively simple object:
the bicycle.The bicycle was selected for two reasons. First,
Rozenblit and Keil predicted that mechanical objects with
visible parts would produce a greater overestimation of
understanding than would electrical artifacts (like com-
puters) with hidden parts. They suggested that people may
mistake being able to see the mechanism of an object (as is
the case for bicycles, zippers, and can openers) with being
able to understand how the object itself works. Second,
bicycles are familiar, everyday objects. In the present stud-
ies, of those participants who claimed to never cycle or to
only rarely cycle (n 5 193), 97% had learned to cycle and
45% owned a bicycle. Additionally, most British people
frequently see bicycles parked and being ridden.
The present study investigated the level of detail and the
accuracy of the information that people have about how
bicycles work. If people’s causal explanations are indeed
central to their conceptual knowledge, then they should
know how important parts of a bicycle are causally re-
lated. Note that this information is distinct from (and may
be independent of) the motor knowledge necessary to ac-
tually ride a bicycle. The explanation-based approach to
conceptual knowledge assumes that people have informal
theories about all their concepts (e.g., snakes, volcanoes,
nests), not just those artifacts that they use in their everyday
lives. Motor knowledge may include only minimal causal
information. For example, to ride a bicycle we need to
know that pushing the top pedal forward and down drives
us forward; we do not have to know how the pedal move-
ment is transformed into the motion of the bicycle (i.e., the
roles of the chain and the back wheel). Also, perceptual
information acquired online probably underpins people’s
ability to ride a bicycle: People who rarely cycle may rely
on seeing and feeling the bicycle to support their motor
performance. Well designed man-made artifacts have af-
fordances and physical constraints that help to specify how
they are to be used (Norman, 2002). Although we may be
able to deduce how a bicycle works by looking at it, this
information may not be stored as part of our conceptual
knowledge of bicycles.
Experiment 1
NonexpertsWhoWere Not Told
What the StudyWas Testing
Method
Participants. Eighty-one unpaid volunteers were tested in three
groups. They were prospective psychology students, their parents
who had come to the University of Liverpool for an Open Day, and
psychology undergraduates. Most participants either rarely cycled
(31%) or never cycled (49%).
Materials and Procedure. The participants were given a single
sheet of A4 (210 mm 3 297 mm) paper, folded in half. The bicycle
drawing task (see Figure 1) and the forced-choice task (see Figure 2)
were on the top and bottom of the first side, respectively, and a ques-
tionnaire was on the other side. First, the participants were asked
to rate their functional knowledge of bicycles on a scale from 1 (“I
know little or nothing about how bicycles work”) to 7 (“I have a
thorough knowledge of how bicycles work”). Next, they were asked
to draw the pedals, chain, and extra frame onto a sketch of a bicycle
(see Figure 1).Then, they were asked to select which of four alterna-
tive pictures was correct for the bicycle frame, pedals, and chain (see
Figure 2). This checked that drawing errors were not merely due to
problems sketching or to scoring drawing errors. Finally, the partici-
pants were asked to rerate their functional knowledge of bicycles
and fill out a questionnaire about their age, gender, and experience
with bicycles. The study took 5–10 min to complete. The partici-
pants were given the bicycle questionnaire without being told what
was being investigated.
Three types of errors, each of which would have severely impaired
the functioning of a bicycle (see Figure 3 for examples), were re-
corded in the drawings:
Frame errors: drawing the frame joining the front and back
wheels, making steering impossible.
Pedal errors: failing to draw the pedals between the wheels and
inside the chain so the pedals could not drive the chain (if the pedals
were separated from the chain, see Figure 3B, or were omitted) or
rendered the chain unnecessary (if the pedals were drawn attached to
the center of the front or back wheel, see Figures 3A and 3D). These
errors often also meant that cyclists would find it difficult to reach
the pedals with their feet.
Chain errors: failing to draw the chain around the pedals and the
back wheel. In most errors the chain was depicted looped around
both the front wheel and the back wheel. This design would require
the chain to stretch when the cyclist was steering.
In the forced-choice task, the correct responses were first for the
frame, second for the pedals, and fourth for the chain (see Figure 2).
All other responses were errors.
Results
Over 40% of nonexperts made at least one error in both
the drawing and the forced-choice tasks (see Table 1). In
this and the subsequent experiments, error scores were al-
most as high on the forced-choice task as on the freehand
drawing task, indicating that production problems were
not the primary cause of people’s deficiencies (Kaiser

The Science of cycology 1669
et al., 1992; Nobes, Martin, & Panagiotaki, 2005). Some
posttest comments were “I think context matters, I know
a bike when I see it but it’s different to recall [it] when sat
in a room”; “Thinking about it in more detail, I realised I
had no idea about its structure”; “I can’t believe I found it
so difficult to remember what a bike frame looks like!”;
“I thought I knew more about the workings than I actually
did”; “It’s hard to picture a bicycle when you have to”; “I
did not think long enough about the front wheel turning”;
and “I never knew how little I knew about things until I
had to draw them.”
Results from naive physics have shown that when
people are given more lifelike alternatives, their ability
to discriminate unnatural from realistic situations can im-
Figure 1. Response sheet for bicycle drawing task.
Second, as best you can, please fill in on the above schematic bicycle drawing the main bits of
the frame of the bicycle that you think are missing, the pedals and the bicycle chain. Use the
symbols given on the right of the drawing to show the frame, pedals and chain.
Frame
Pedals
Chain
First, please can you rate your knowledge of how bicycles work on a scale from 1 to 7 where 1
means “I know little or nothing about how bicycles work” and 7 means “I have a thorough
knowledge of how bicycles work”. You do not need to be an expert to give yourself a “7” rating—an
expert would be rated as 7++. Also, you should rate how much you know, not how much you think
you know compared to other people. Rating: _______
Figure 2. Questions for forced-choice task. Correct responses: frame 5 first bicycle;
pedals 5 second bicycle; chain 5 fourth bicycle.
5. CIRCLE which one of these four bicycles best shows the usual position of the chain:
3. CIRCLE which one of these four bicycles best shows the usual position of the frame:
4. CIRCLE which one of these four bicycles best shows the usual position of the pedals:

1670 lawson
prove (Bertamini et al., 2003; Kaiser et al., 1992), but this
is not always observed (Proffitt, Kaiser, & Whelan, 1990;
Rohrer, 2003).The perceptual realism manipulation in the
naive physics studies is similar to that which was used
in the drawing compared with forced-choice tasks here.
Participants continued to make many errors when they
were shown pictures of possible bicycles in the forced-
choice task. Thus, participants could not remember what
a bicycle looked like, even when they were prompted with
perceptually realistic pictures.
Experiment 2
Nonexperts Informed That Their
Functional KnowledgeWas Being Tested
In Experiment 2, 94 participants—from the same popu-
lation as Experiment 1—were tested in two groups. Most
participants either rarely cycled (27%) or never cycled
(54%). Experiment 2 replicated Experiment 1, except
that in Experiment 2 participants were informed that their
functional knowledge of bicycles was going to be inves-
tigated. Participants might have made drawing errors in
Experiment 1 because they were concentrating on produc-
ing well drawn pictures rather than focusing on drawing
functionally viable bicycles. To address this possibility,
participants in Experiment 2 were given a heavy and rather
patronizing hint about what the task was going to be about
before they began. I told them, “I am not interested in your
drawing ability. It does not matter if your lines are wobbly
or whatever. What I am interested in is what you know
about how bicycles work. I want to make you think about
what the pedals of the bike do . . . and what the chain of the
bike does . . . and why the frame of the bike is a particular
shape . . . and how you steer a bike. So I want you to think
about how a bike works before you draw in the parts.”
Surprisingly, this hint did not reduce errors in Experi-
ment 2 (see Table 1). Indeed, in comparison with Experi-
ment 1, more participants in Experiment 2 made at least
Table 1
Percentage Errors in the Bicycle Drawing and Forced-Choice Tasks
in Experiments 1, 2, and 3 for Each Error Type (Frame, Pedals, and Chain)
and Then If at Least 1 Error Occurred inAny of the Three Error Types
Experiment 1 Experiment 2 Experiment 3
Error Type Drawing Forced Choice Drawing Forced Choice Drawing Forced Choice
Frame 12 26 23 46 3 7
Pedals 7 4 11 5 4 1
Chain 36 28 50 43 6 9
Any 44 40 60 55 9 13
Note—In Experiment 1, participants (N 5 81) were nonexperts given no hints about the task. In Experi-
ment 2, participants (N 5 94) were nonexperts told to focus on function. In Experiment 3, participants
(N 5 68) were cycling experts given no hints about the task.
Figure 3. Four drawings from the bicycle drawing task (see Figure 1). Draw-
ings show pedal and chain errors in (A) and (B), frame and chain errors in (C),
and the rare triple combination of frame, pedal, and chain errors in (D). The
four participants stated that they (A) cycled at least once a month, (B) never
cycled, (C) cycled most days, and (D) rarely cycled.
(D) Male
(B) Female
(C) Female
(A) Male

one drawing error [χ2(1, n 5 175) 5 3.99, p , .05] and
at least one forced-choice error [χ2(1, n 5 175) 5 4.36,
p , .05]. These increased error rates can largely be ex-
plained by the higher proportion of females tested in Ex-
periment 2, as discussed below.
Experiment 3
Cycling ExpertsWhoWere Not Told
What the StudyWas Testing
The results of the first two experiments revealed that
nonexperts made many fundamental mistakes when they
tried to show how bicycles worked, even when they were
told what I was investigating. I was concerned that this poor
performance might simply be due to the difficulty of the
task. Experiment 3 replicated the results of Experiment 1,
but it tested 68 cycling experts to determine whether fre-
quent use of bicycles would lead to a reduction in error
rates. Of these experts, 31 were tested together at the Mer-
seyside Cycle Campaign’s Annual General Meeting and
37 were tested individually or in groups of up to 10. Most
cycled frequently, and 88% cycled at least weekly.
Experts made far fewer errors than nonexperts (see
Table 1), though they were not perfect. Relative to those
tested in Experiment 1, many fewer experts made draw-
ing errors [χ2(1, n 5 149) 5 23.17, p , .001] or forced-
choice errors [χ2(1, n 5 149) 5 12.79, p , .001]. Thus,
when participants had sufficient expertise with a bicycle,
performance was generally good; so, the task was not too
difficult or confusing per se. Note, though, that the as-
sumption behind explanation-based theories is that causal
information is central to all our concepts and not just for
a handful of objects for which we have large amounts of
experience. The accurate performance of bicycle experts
in Experiment 3 demonstrated that the task was not hard
per se. Therefore, the poor performance of nonexperts
in Experiments 1 and 2 suggests that most people do not
have a basic understanding of how bicycles work.1
Self-Ratings of Understanding
of Bicycle Function
Not surprisingly, the experts in Experiment 3 rated their
bicycle knowledge higher than the nonexperts in Experi-
ments 1 and 2 [F(2,224) 5 14.02, p , .001; see Table 2].
Posttest ratings were lower than pretest ratings in Experi-
ments 1 and 2 (though this difference was not significant)
and were higher in Experiment 3 [F(1,59) 5 9.142, p ,
.005]. Thus, participants tended to alter their ratings de-
pending on whether they did poorly on the bicycle tasks
(in Experiments 1 and 2) or were accurate (in Experi-
ment 3), consistent with Rozenblit and Keil’s (2002) find-
ings. This suggests that participants had some insight into
their own performance. However, the small size of these
differences—both across and within studies—suggests
that these self-ratings only crudely measured changes in
participants’ metaknowledge. This is consistent with re-
search showing that people are generally inaccurate at as-
sessing their own knowledge and abilities, particularly in
domains for which it is difficult to assess personal perfor-
mance, such as in the provision of explanations (Dunning,
Johnson, Ehrlinger, & Kruger, 2003; Mabe & West, 1982).
Supporting this hypothesis, participants’pretest and post-
test self-ratings of their knowledge of bicycle function did
not correlate highly to their actual knowledge, as assessed
by the total errors that they made in the task (see Table 2),
though all correlations were significant ( p , .05).
Sex,Age, and Expertise Effects
An unanticipated finding in the present studies was
a striking difference in performance between men and
women (see Table 3). For the nonexperts in Experiments
1 and 2, men made only about one third the errors that
women made. This male advantage was significant in
Experiment 1 for both drawing errors [χ2(1, n 5 80) 5
15.20, p , .001] and forced-choice errors [χ2(1, n 5
80) 5 20.12, p , .001], and likewise in Experiment 2 for
both drawing errors [χ2(1, n 5 91) 5 18.39, p , .001]
and forced-choice errors [χ2(1, n 5 91) 5 17.55, p ,
.001].The male advantage was also evident for the experts
tested in Experiment 3, for both drawing errors [χ2(1, n 5
64) 5 13.86, p , .001] and forced-choice errors [χ2(1,
n 5 64) 5 9.98, p , .01].
One reason for this male advantage may be that most
men have had more experience overall with bicycles,
even those who rarely or never cycle. The results of the
questionnaire provided some evidence for this. In the first
two experiments, men reported slightly higher levels of
cycling, and in all three experiments men stated that they
repaired bicycles more than women did. However, no dif-
ference was found between the performance of males in
Table 2
Self-Ratings of Functional Knowledge Provided Before andAfter Completing the
Bicycle Tasks in Experiments 1, 2 and 3, and Correlations Between These Pretest
and Posttest Ratings and Total Errors on the Frame, Pedals, or Chain
Across Both the Drawing and Forced-Choice Tasks*
Experiment 1 Experiment 2 Experiment 3
Rating Type Rating Correlation Rating Correlation Rating Correlation
Pretest ]4.6 ].43 ]4.2 ].36 15.4 ].35
Posttest ]4.5 ].48 ]4.0 ].32 15.6 ].29
Change ]0.1 – ]0.2 – 10.2 –
Note—Negative correlations indicate that people who rated their functional knowledge as
good generally made fewer errors in the bicycle test. Self-ratings ranged from 1 (little or no
knowledge) to 7 (thorough knowledge); see Figure 1. *Maximum of 6 errors.

1672 lawson
Experiments 1 and 2 and that of females in Experiment 3,
either for drawing errors [χ2(1, n 5 79) 5 0.18, p . .6] or
for forced-choice errors [χ2(1, n 5 79) 5 1.17, p . .2].
This provides evidence that the male advantage was not
solely due to men having greater experience with bicycles,
since these nonexpert males reported much lower levels of
cycling than did the expert women and similar levels of ex-
perience at repairing bicycles. An alternative possibility is
that men may generally have a better causal and functional
understanding of objects due to innate or environmental
differences between the sexes (see, e.g., Crowley, Callanan,
Tenenbaum, & Allen, 2001; Johnson, Alexander, Spencer,
Leibham, & Neitzel, 2004; Robert & Harel, 1996).
The age of the participants had little influence on per-
formance.Across Experiments 1 and 2, for young females
(n 5 75; mean age 5 18; age range 17–28), 71% made
drawing errors and 63% made forced-choice errors. Simi-
larly high error rates occurred for older females (n 5 34;
mean age 5 47; age range 40–54), with 71% making
drawing errors and 76% making forced-choice errors. Far
fewer errors were made by young males (n 5 28; mean
age 5 18; age range 17–23), with 21% making drawing
errors and 14% making forced-choice errors. Similarly
low error rates occurred for older males (n 5 33; mean
age 5 50; age range 42–63), with 24% making drawing
errors and 18% making forced-choice errors.
The questionnaire results revealed that 30 of the 175
nonexperts in Experiments 1 and 2 cycled at least monthly,
whereas 7 of the 68 experts tested in Experiment 3 cycled
less than weekly. The results of these participants were,
nevertheless, similar to those of the remainder of the par-
ticipants in each experiment, so they do not alter the con-
clusions drawn above.
GENERAL DISCUSSION
Many of the nonexperts tested in the first two experi-
ments revealed major inaccuracies and omissions in their
understanding of how bicycles worked, despite almost
all of them having learned to ride a bicycle and nearly
half of them owning a bicycle; and despite bicycles being
common, everyday objects. When objective assessments
were made, serious errors of functional understanding
of bicycles were found to be common. Performance im-
proved—but was still not perfect—for bicycle experts (in
Experiment 3) and for nonexperts shown a real bicycle to
copy (see note 1).
It would be unsurprising if nonexperts had failed to ex-
plain the intricacies of how gears work or why the angle
of the front forks of a bicycle is critical. Indeed, even
physicists disagree about seemingly simple issues, such
as why bicycles are stable (Jones, 1970; Kirshner, 1980)
and how they steer (Fajans, 2000). What is striking about
the present results is that so many people have virtually no
knowledge of how bicycles function. For example, many
people (but especially women) cannot specify the correct
position of the bicycle pedals and chain. They do not ap-
pear to understand a simple set of causal relations (turning
the pedals turns the chain that spins the back wheel that
drives the bicycle forward); therefore, they do not know
how turning the pedals moves a bicycle. These findings
now need to be extended to a range of everyday objects to
establish the generalizability of the present results.
The failures in understanding demonstrated in the pres-
ent study reveal that most people’s functional understand-
ing of bicycles is sketchy and shallow (see Keil, 2003b;
Wilson & Keil, 2000). The present study concentrated
on measuring functional knowledge objectively, so it
complements Rozenblit and Keil’s (2002) studies (which
focused on measuring changes to subjective self-ratings).
Together, these results suggest that people have a vague,
fragmentary, and often inaccurate understanding of how
many everyday objects function.The weakness in people’s
knowledge is particularly striking for mechanical objects
with visible parts, such as bicycles (Rozenblit & Keil,
2002).
Keil (2003b) has argued that a minimal, superficial un-
derstanding of function may actually benefit us by letting
us efficiently interpret the world and make accurate causal
predictions without overburdening our limited processing
Table 3
Percentage Errors in the Bicycle Drawing and Forced-Choice Tasks in Experiments 1, 2, and 3
for Each Error Type (Frame, Pedals, and Chain) and Then If at Least 1 Error
Occurred inAny of the Three Error Types, for Females and Males
Experiment 1* Experiment 2** Experiment 3***
Error Type Drawing Forced Choice Drawing Forced Choice Drawing Forced Choice
For Females
Frame 16 42 30 58 11 11
Pedals 09 05 13 06 17 11
Chain 56 44 63 52 17 28
Any 65 63 75 70 28 28
For Males
Frame 08 8 08 21 00 02
Pedals 05 03 04 04 00 00
Chain 14 11 21 21 00 00
Any 22 14 25 21 00 02
Note—1, 3, and 4 participants in Experiments 1, 2, and 3, respectively, did not specify their gender, so
their results are not included here. *Females, n 5 43; males, n 5 37. **Females, n 5 67; males, n 5
24. ***Females, n 5 18; males, n 5 46.

and storage capabilities. People may overestimate their
understanding of how objects work because they do not
realize how reliant they are on extracting perceptual infor-
mation directly from the world rather than from memory
(see note 1).
Similarly, Norman (2002) has argued that everyone
forms theories about how everyday objects function to
explain what they have observed. Norman, however,
suggests that these explanations may be based on weak,
fragmentary, or even inaccurate evidence. One example
involves the room thermostat. People often try to quickly
heat up a room by turning the thermostat all the way up.
Norman suggests that this is because they have an incor-
rect explanation for how the thermostat works. He notes,
“In the case of the thermostat, the design gives absolutely
no hint as to the correct answer. In the absence of external
information, people are free to let their imaginations run
free” (p. 39).The present findings are more surprising, for
they suggest that we may not acquire an understanding of
how an object such as a bicycle works even if the neces-
sary information is readily available in our everyday en-
vironment (see note 1). We may only rarely try to provide
explanations or to test the consistency of our fragmented
understanding of the world (Leiser, 2001).
If we usually store only relatively superficial and frag-
mentary information about even the most familiar objects
in our environment (Keil, 2003b; Norman, 2002); if we do
not understand the basics of how common objects work
(Rozenblit & Keil, 2002); if we rarely try to provide expla-
nations of object function; and if we do not test the consis-
tency of our knowledge (Leiser, 2001); then the role of ex-
planations and theories in our understanding of everyday
concepts may need to be reexamined. Explanation-based
(or theory-based) theories of categorization emphasize
that relatively high level, sophisticated causal theories un-
derpin our concepts and that such theories are important
in everyday behavior (Ahn & Luhmann, 2005; Murphy,
2002; Murphy & Medin, 1985). The proponents of strong
versions of explanation-based theories have gone further.
For example, Brewer, Chinn, and Samarapungavan (1998)
have suggested that children and adults both develop theo-
ries about concepts—theories that are similar to scientific
explanations. Chaigneau, Barsalou, and Sloman (2004)
have stated that “as common artefacts are encountered re-
peatedly, especially during development, people explore
the underlying causal mechanisms to understand their
functions. As a result, causal models become established
in memory to represent these understandings” (p. 620).
Yet these claims are not supported by the present results; it
is now important to establish whether the present findings
extend to other objects.
In contrast, others who support explanation-based
theories of categorization have stressed the limitations
of people’s explanations (Keil, 2003b; Rozenblit & Keil,
2002). Keil (2003a) has suggested that often only mun-
dane, general knowledge—not sophisticated, domain-
specific information—may be available to us (e.g., that
objects given the same name have similar properties; that
all animals have roughly the same internal parts; that off-
spring grow up to look like their parents). Murphy (2002)
has warned that “if explanations are often applied after
the fact and are rather shallow, then how helpful could
they be in using a concept? . . . It is possible that the limi-
tations on our knowledge will make it less than useful in
many cases. Among those cases would be situations in
which knowledge is incomplete or partially wrong, unfor-
tunately not unusual circumstances for humans” (p. 145).
The present findings support these cautious views. It
seems clear that people do sometimes use prior knowl-
edge when they categorize (Kaplan & Murphy, 2000; Lin
& Murphy, 1997; Palmeri & Blalock, 2000; Wisniewski,
1995). These cases may be exceptions, however. General
background information might be useful for difficult
categorization tasks that have relatively sparse category
information (e.g., short verbal descriptions or simple de-
pictions of stimuli). Under most circumstances, though,
our categorization processes may rely primarily on infor-
mation that is simple and largely perceptual. Even induc-
tion and category learning may not usually make use of
functional or causal information (Murphy, 2002). These
suggestions are, however, necessarily speculative: Expla-
nation-based theories are relatively recent and have not
yet been extensively tested. More empirical research is
needed to clearly establish the role played by functional
and causal information in our conceptual knowledge
across a range of tasks.
The present results are unexpected, even on accounts
of categorization that emphasize the perceptual basis of
conceptual knowledge. The bicycle frame, chain, and
pedals are large, readily perceived parts. Correctly locat-
ing them is easy if a bicycle is visible (see note 1). Even
if only visual information about bicycles was stored, the
errors observed in Experiments 1 and 2 should not have
occurred. Bicycles are visually distinctive, with no close
neighbors in shape space (toy scooters have no frame
and much smaller wheels; motorbikes and mopeds have
bulkier frames and an engine). Since it is unnecessary to
store detailed information about bicycle shape in order
to distinguish bicycles from other similar objects, our
stored visual representations of bicycles may themselves
be sketchy. People have been found to have surprisingly
poor perceptual memory for everyday objects such as
coins (Nickerson & Adams, 1979) and road signs (Mar-
tin & Jones, 1998). Marmie and Healy (2004; see also
Rinck, 1999) found that recall of the features of coins
and their spatial position was much improved after just
15 sec of intentional learning and that intentional study
produced large benefits, even when recall was delayed by
one week. Their results suggest that reports of strikingly
poor memory for common objects (e.g., Martin & Jones,
1998; Nickerson & Adams, 1979) are due to people’s
failure to encode information because it is not useful for
the everyday use of the objects (e.g., knowing the direc-
tion that a head faces on a coin). This suggests that little
detailed visual information may be stored following in-
cidental learning of even highly familiar objects. Note,
however, that most of the information tested in these stud-
ies had no functional significance. In contrast, a bicycle

1674 lawson
with pedals attached directly to the front wheel or with a
chain around both wheels would no longer function as a
normal bicycle. Nevertheless, the present results indicate
that even this functionally important and perceptually sa-
lient information is often not readily available to people.
A personal anecdote illustrates this last point. I regu-
larly scuba dive, and on one weekend trip two friends and
I decided to do a night dive. We needed our weight belts,
which were on a boat in the harbor, so I offered to fetch
them. I swam the few meters to the boat in my dry suit,
clipped my own weight belt around my waist, held one
weight belt in each hand, slipped overboard, and headed
back to the ladder. To my surprise, I immediately sank to
the bottom of the harbor. I flailed up and managed to gasp
some air but could not move forward before I was dragged
down again. Realizing that I was well on my way to drown-
ing, I dumped the two loose weight belts and managed to
reach the ladder. In retrospect, the mistake was shocking.
I have spent years carrying lead weight belts around:Their
heaviness is both perceptually highly salient and central
to their function. I had tried to carry three people’s weight
belts, even though I knew that my own weight belt was
adjusted to make me only just buoyant in a dry suit. Why,
then, did I fail to anticipate what would happen when I
jumped into the water carrying enough lead to sink me,
no matter how hard I swam? My conceptual knowledge of
weight belts let me understand what had happened after
the event, but it was not until the physical constraints of
the real world impinged that I accessed this information.
Such post hoc explanations created after perceptual expe-
riences allow us to learn from our mistakes but otherwise
may have little influence on our everyday actions.
The present findings indicate people’s striking inability
to remember the location of three causally critical parts
of a common object: the bicycle. Many people do not ap-
pear to store accurate concept-specific information about
causal function (e.g., that the pedals of a bicycle drive a
chain around that drives the back wheel forward). Most
people’s conceptual knowledge of this everyday object
seems to bear little resemblance to a scientific theory. The
present results are not consistent with those of strong ver-
sions of explanation-based theories, which propose that
high-level, domain-specific causal and functional infor-
mation is central to our conceptual knowledge. Instead,
people may rely largely on relatively simple, concept-gen-
eral information (e.g., that objects given the same name
usually have the same shape but often differ in color and
size, or that all animals have roughly the same parts inside
them; Keil, 2003a).
REFERENCES
Ahn, W.-K., & Luhmann, C. C. (2005). Demystifying theory-based cat-
egorization. In L. Gershkoff-Stowe & D. H. Rakison (Eds.), Building
object categories in developmental time: Carnegie Mellon Symposia
on cognition (pp. 277-300). Mahwah, NJ: Erlbaum.
Bertamini, M., Spooner, A., & Hecht, H. (2003). Naive optics: Pre-
dicting and perceiving reflections in mirrors. Journal of Experimental
Psychology: Human Perception & Performance, 29, 982-1002.
Brewer, W. F., Chinn, C. A., & Samarapungavan, A. (1998). Expla-
nation in scientists and children. Minds & Machines, 8, 119-136.
Carey, S. (1985). Conceptual change in childhood. Cambridge, MA:
MIT Press.
Chaigneau, S. E., Barsalou, L. W., & Sloman, S. A. (2004). Assess-
ing the causal structure of function. Journal of Experimental Psychol-
ogy: General, 133, 601-625.
Crowley, K., Callanan, M. A., Tenenbaum, H. R., & Allen, E.
(2001). Parents explain more often to boys than to girls during shared
scientific thinking. Psychological Science, 12, 258-261.
Dunning, D., Johnson, K., Ehrlinger, J., & Kruger, J. (2003). Why
people fail to recognize their own incompetence. Current Directions
in Psychological Science, 12, 83-87.
Fajans, J. (2000). Steering in bicycles and motorcycles. American Jour-
nal of Physics, 68, 654-659.
Johnson, K. E., Alexander, J. M., Spencer, S., Leibham, M. E., &
Neitzel, C. (2004). Factors associated with the early emergence of
intense interests within conceptual domains. Cognitive Development,
19, 325-343.
Jones, D. E. H. (1970). The stability of the bicycle. Physics Today, 23,
34-40.
Kaiser, M. K., Proffitt, D. R., Whelan, S. M., & Hecht, H. (1992).
Influence of animation on dynamical judgments. Journal of Experi-
mental Psychology: Human Perception & Performance, 18, 669-689.
Kaplan, A. S., & Murphy, G. L. (2000). Category learning with mini-
mal prior knowledge. Journal of Experimental Psychology: Learning,
Memory, & Cognition, 26, 829-846.
Keil, F. C. (1989). Concepts, kinds, and cognitive development. Cam-
bridge, MA: MIT Press.
Keil, F. C. (2003a). Categorisation, causation, and the limits of under-
standing. Language & Cognitive Processes, 18, 663-692.
Keil, F. C. (2003b). Folkscience: Coarse interpretations of a complex
reality. Trends in Cognitive Sciences, 7, 368-373.
Kirshner, D. (1980). Some nonexplanations of bicycle stability. Ameri-
can Journal of Physics, 48, 36-38.
Leiser, D. (2001). Scattered naive theories: Why the human mind is iso-
morphic to the Internet web. New Ideas in Psychology, 19, 175-202.
Lin, E. L., & Murphy, G. L. (1997). Effects of background knowledge
on object categorization and part detection. Journal of Experimental
Psychology: Human Perception & Performance, 23, 1153-1169.
Mabe, P. A., III, & West, S. G. (1982). Validity of self-evaluation of
ability: A review and meta-analysis. Journal of Applied Psychology,
67, 280-296.
Marmie, W. R., & Healy, A. F. (2004). Memory for common objects:
Brief intentional study is sufficient to overcome poor recall of US coin
features. Applied Cognitive Psychology, 18, 445-453.
Martin, M., & Jones, G. V. (1998). Generalizing everyday memory:
Signs and handedness. Memory & Cognition, 26, 193-200.
Mills, C. M., & Keil, F. C. (2004). Knowing the limits of one’s under-
standing: The development of an awareness of an illusion of explana-
tory depth. Journal of Experimental Child Psychology, 87, 1-32.
Murphy, G. L. (2002). The big book of concepts. Cambridge, MA: MIT
Press.
Murphy, G. L., & Medin, D. L. (1985). The role of theories in concep-
tual coherence. Psychological Review, 92, 289-316.
Nickerson, R. S., & Adams, M. J. (1979). Long-term memory for a
common object. Cognitive Psychology, 11, 287-307.
Nobes, G., Martin,A. E., & Panagiotaki, G. (2005).The development
of scientific knowledge of the Earth. British Journal of Developmen-
tal Psychology, 23, 47-64.
Norman, D. A. (2002). The design of everyday things. NewYork: Basic
Books.
O’Regan, J. K. (1992). Solving the “real” mysteries of visual percep-
tion: The world as an outside memory. Canadian Journal of Psychol-
ogy, 46, 461-488.
Palmeri, T. J., & Blalock, C. (2000). The role of background knowl-
edge in speeded perceptual categorization. Cognition, 77, B45-B57.
Pazzani, M. J. (1991). Influence of prior knowledge on concept acquisi-
tion: Experimental and computational results. Journal of Experimen-
tal Psychology: Learning, Memory, & Cognition, 17, 416-432.

Proffitt, D. R., Kaiser, M. K., & Whelan, S. M. (1990). Understand-
ing wheel dynamics. Cognitive Psychology, 22, 342-373.
Rinck, M. (1999). Memory for everyday objects: Where are the digits on
numerical keypads? Applied Cognitive Psychology, 13, 329-350.
Robert, M., & Harel, F. (1996). The gender difference in orienting
liquid surfaces and plumb-lines: Its robustness, its correlates, and the
associated knowledge of simple physics. Canadian Journal of Experi-
mental Psychology, 50, 280-314.
Rohrer, D. (2003). The natural appearance of unnatural incline speed.
Memory & Cognition, 31, 816-826.
Rozenblit, L., & Keil, F. [C.] (2002). The misunderstood limits of
folk science: An illusion of explanatory depth. Cognitive Science, 26,
521-562.
Smith, E. E., & Sloman, S. A. (1994). Similarity- versus rule-based
categorization. Memory & Cognition, 22, 377-386.
Wilson, R. A., & Keil, F. C. (2000). The shadows and shallows of ex-
planation. In F. C. Keil & R.A. Wilson (Eds.), Explanation and cogni-
tion (pp. 87-114). Cambridge, MA: MIT Press.
Wisniewski, E. J. (1995). Prior knowledge and functionally relevant
features in concept learning. Journal of Experimental Psychology:
Learning, Memory, & Cognition, 21, 449-468.
NOTE
1. A final experiment replicated the results of Experiment 1 and
tested 58 nonexperts from the same population as Experiments 1 and
2, but people could copy from a real bicycle as they completed the
questionnaire. O’Regan (1992) suggested that we may be using the
world as an “outside memory” to save us from having to store huge
amounts of information. Since much of the information that we need
in everyday life can be found simply by moving our eyes, we do not
need to store it and then retrieve it from memory (Norman, 2002). If
so, people shown a bicycle should succeed at the task. This hypothesis
was supported. Relative to those tested in Experiment 1, many fewer
nonexperts made drawing errors [χ2(1, N 5 139) 5 12.92, p , .001;
mean errors 5 16%] or forced-choice errors [χ2(1, N 5 139) 5 12.61,
p , .001; mean errors 5 12%]. Indeed, these nonexperts who could
copy a real bicycle made no more errors than did the experts tested
in Experiment 3, either in their drawing [χ2(1, N 5 126) 5 1.34, p .
.2] or in the forced-choice task [χ2(1, N 5 126) 5 0.04, p . .8]. The
errors in Experiments 1 and 2 were thus not due to participants failing
to understand the task or to their misinterpreting the bicycle sketch,
nor were most errors due to their inability to draw or to their lack of
motivation. The accuracy of nonexperts who were shown a real bicycle
demonstrates that the information needed to succeed at the task is com-
monly available in people’s everyday lives.
(Manuscript received June 21, 2005;
revision accepted for publication August 24, 2005.)

The science of cycology: Failures to understand how everyday objects work

More Related Content

Viewers also liked (6)

Similar to The science of cycology: Failures to understand how everyday objects work (20)

More from Trading Game Pty Ltd (20)

Recently uploaded (20)

The science of cycology: Failures to understand how everyday objects work